Steroidal contraceptives: effect on carbohydrate metabolism in women without diabetes mellitus

  • Review
  • Intervention

Authors


Abstract

Background

Many hormonal contraceptives have been associated with changes in carbohydrate metabolism. Alterations may include decreased glucose tolerance and increased insulin resistance, which are risk factors for Type 2 diabetes mellitus and cardiovascular disease. These issues have been raised primarily with contraceptives containing estrogen.

Objectives

To evaluate the effect of hormonal contraceptives on carbohydrate metabolism in healthy women and those at risk for diabetes due to overweight.

Search methods

In April 2014, we searched the computerized databases MEDLINE, POPLINE, CENTRAL, and LILACS for studies of hormonal contraceptives and carbohydrate metabolism. We also searched for clinical trials in ClinicalTrials.gov and ICTRP. The initial search also included EMBASE.

Selection criteria

All randomized controlled trials were considered if they examined carbohydrate metabolism in women without diabetes who used hormonal contraceptives for contraception. Comparisons could be a placebo, a non-hormonal contraceptive, or another hormonal contraceptive that differed in drug, dosage, or regimen. Interventions included at least three cycles. Outcomes included glucose and insulin measures.

Data collection and analysis

We assessed all titles and abstracts identified during the literature searches. The data were extracted and entered into RevMan. We wrote to researchers for missing data. For continuous variables, the mean difference (MD) was computed with 95% confidence interval (CI) using a fixed-effect model. For dichotomous outcomes, the Peto odds ratio with 95% CI was calculated.

Main results

We found 31 trials that met the inclusion criteria. No new trials were eligible in 2014. Twenty-one trials compared combined oral contraceptives (COCs); others examined different COC regimens, progestin-only pills, injectables, a vaginal ring, and implants. None included a placebo. Of 34 comparisons, eight had any notable difference between the study groups in an outcome.

Twelve trials studied desogestrel-containing COCs, and the few differences from levonorgestrel COCs were inconsistent. A meta-analysis of two studies showed the desogestrel group had a higher mean fasting glucose (MD 0.20; 95% CI 0.00 to 0.41). Where data could not be combined, single studies showed lower mean fasting glucose (MD -0.40; 95% CI -0.72 to -0.08) and higher means for two-hour glucose response (MD 1.08; 95% CI 0.45 to 1.71) and insulin area under the curve (AUC) (MD 20.30; 95% CI 4.24 to 36.36).

Three trials examined the etonogestrel vaginal ring and one examined an etonogestrel implant. One trial showed the ring group had lower mean AUC insulin than the levonorgestrel-COC group (MD -204.51; 95% CI -389.64 to -19.38).

Of eight trials of norethisterone preparations, five compared COCs and three compared injectables. In a COC trial, a norethisterone group had smaller mean change in glucose two-hour response than a levonorgestrel-COC group (MD -0.30; 95% CI -0.54 to -0.06). In an injectable study, a group using depot medroxyprogesterone acetate had higher means than the group using norethisterone enanthate for fasting glucose (MD 10.05; 95% CI 3.16 to 16.94), glucose two-hour response (MD 17.00; 95% CI 5.67 to 28.33), and fasting insulin (MD 3.40; 95% CI 2.07 to 4.73).

Among five recent trials, two examined newer COCs with different estrogen types. One showed the group with nomegestrel acetate plus 17β-estradiol had lower means than the levonorgestrel group for incremental AUC glucose (MD -1.43; 95% CI -2.55 to -0.31) and glycosylated hemoglobin (HbA1c) (MD -0.10; 95% CI -0.18 to -0.02). Two trials compared extended versus conventional (cyclic) regimens. With a dienogest COC, an extended-use group had greater mean change in AUC glucose (MD 82.00; 95% CI 10.72 to 153.28). In a small trial using two levonorgestrel COCs, the lower-dose group showed smaller mean change in fasting glucose (MD -3.00; 95% CI -5.89 to -0.11), but the obese and normal weight women did not differ significantly.

Authors' conclusions

Current evidence suggests no major differences in carbohydrate metabolism between different hormonal contraceptives in women without diabetes. We cannot make strong statements due to having few studies that compared the same types of contraceptives. Many trials had small numbers of participants and some had large losses. Many of the earlier studies had limited reporting of methods.

We still know very little about women at risk for metabolic problems due to being overweight. More than half of the trials had weight restrictions as inclusion criteria. Only one small trial stratified the groups by body mass index (obese versus normal).

Résumé scientifique

Contraceptifs stéroïdiens : effet sur le métabolisme des glucides chez les femmes sans diabète

Contexte

De nombreux contraceptifs hormonaux ont été associés à des modifications du métabolisme des glucides. Les altérations peuvent inclure une baisse de la tolérance au glucose et une hausse de la résistance à l'insuline, qui sont des facteurs de risque pour le diabète de type 2 et les maladies cardiovasculaires. Ces questions ont été soulevées principalement avec les contraceptifs contenant de l'œstrogène.

Objectifs

Évaluer l'effet des contraceptifs hormonaux sur le métabolisme des glucides chez les femmes en bonne santé et celles à risque pour le diabète en raison du surpoids.

Stratégie de recherche documentaire

En avril 2014, nous avons effectué des recherches dans les bases de données informatisées MEDLINE, POPLINE, CENTRAL et LILACS pour trouver des études des contraceptifs hormonaux et du métabolisme des glucides. Nous avons également recherché des essais cliniques dans ClinicalTrials.gov et ICTRP. La recherche initiale incluait également EMBASE.

Critères de sélection

Tous les essais contrôlés randomisés ont été considérés s'ils examinaient le métabolisme des glucides chez les femmes sans diabète qui utilisaient des contraceptifs hormonaux pour la contraception. Les comparaisons pouvaient concerner un placebo, un contraceptif non-hormonal ou un autre contraceptif hormonal qui différait par le médicament, la dose ou le schéma posologique. Les interventions incluaient au moins trois cycles. Les critères de jugement comprenaient des mesures du glucose et de l'insuline.

Recueil et analyse des données

Nous avons évalué tous les titres et résumés identifiés au cours des recherches de littérature. Les données ont été extraites et entrées dans RevMan. Nous avons contacté des chercheurs pour obtenir les données manquantes. Pour les variables continues, la différence moyenne (DM) était calculée avec un intervalle de confiance (IC) à 95 % en utilisant un modèle à effets fixes. Pour les résultats dichotomiques, le rapport des cotes de Peto était calculé avec un IC à 95 %.

Résultats principaux

Nous avons trouvé 31 essais qui répondaient aux critères d'inclusion. Aucun nouvel essai n'était éligible en 2014. Vingt-et-un essais comparaient des contraceptifs oraux combinés (COC) ; d'autres examinaient différents schémas posologiques de COC, des pilules de progestérone uniquement, des contraceptifs injectables, un anneau vaginal et des implants. Aucun ne comprenait de placebo. Sur les 34 comparaisons, huit ont révélé une quelconque différence notable entre les groupes d'étude dans un résultat.

Douze essais avaient étudié les COC contenant du désogestrel, et les quelques différences par rapport aux COC de lévonorgestrel étaient contradictoires. Une méta-analyse de deux études a révélé que le groupe prenant du désogestrel avait une glycémie à jeun moyenne plus élevée (DM 0,20 ; IC à 95 % 0,00 à 0,41). Lorsqu'il n'était pas possible de combiner les données, des études individuelles ont révélé une glycémie à jeun moyenne plus basse (DM -0,40 ; IC à 95 % -0,72 à -0,08) et des moyennes plus élevées pour la réponse au glucose à deux heures (DM 1,08 ; IC à 95 % 0,45 à 1,71) et l'aire sous la courbe de l'insuline (AUC) (DM 20,30 ; IC à 95 % 4,24 à 36,36).

Trois essais avaient examiné l'anneau vaginal d'étonogestrel et un avait examiné un implant d'étonogestrel. Un essai a révélé que le groupe utilisant l'anneau avait une aire sous la courbe (AUC) moyenne pour l'insuline plus basse que le groupe prenant des COC de lévonorgestrel (DM -204,51 ; IC à 95 % -389,64 à -19,38).

Sur les huit essais de préparations de noréthistérone, cinq avaient comparé les COC et trois avaient comparé les contraceptifs injectables. Dans un essai des COC, un groupe prenant de la noréthistérone présentait des modifications moyennes de la réponse au glucose à deux heures plus faibles qu'un groupe prenant du lévonorgestrel (DM -0,30 ; IC à 95 % -0,54 à -0.06). Dans une étude des contraceptifs injectables, un groupe utilisant de l'acétate de médroxyprogestérone-retard avait des moyennes plus élevées que le groupe utilisant de l'énanthate de noréthistérone pour la glycémie à jeun (DM 10,05 ; IC à 95 % 3,16 à 16,94), la réponse au glucose à deux heures (DM 17,00 ; IC à 95 % 5,67 à 28,33), et l'insuline à jeun (DM 3,40 ; IC à 95 % 2,07 à 4,73).

Parmi cinq essais récents, deux avaient examiné des COC plus récents avec différents types d'œstrogène. Un essai a révélé que le groupe prenant de l'acétate de nomégestrel plus 17β-estradiol avait des moyennes plus basses que le groupe prenant du lévonorgestrel pour une aire sous la courbe (AUC) incrémentale du glucose (DM -1,43 ; IC à 95 % -2,55 à -0,31) et l'hémoglobine glycosylée (HbA1c) (DM -0,10 ; IC à 95 % -0,18 à -0,02). Deux essais comparaient des schémas posologiques prolongés à des schémas posologiques conventionnels (cycliques). Avec un COC de diénogest, un groupe ayant une utilisation prolongée présentait des modifications moyennes plus importantes de l'aire sous la courbe (AUC) du glucose (DM 82,00 ; IC à 95 % 10,72 à 153,28). Dans un essai de petite taille utilisant deux COC de lévonorgestrel, le groupe prenant la dose plus basse a révélé des modifications moyennes plus faibles de la glycémie à jeun (DM -3,00 ; IC à 95 % -5,89 à -0,11), mais les femmes obèses et de poids normal ne différaient pas significativement.

Conclusions des auteurs

Les preuves actuelles ne suggèrent pas de différences majeures dans le métabolisme des glucides entre les différents contraceptifs hormonaux chez les femmes sans diabète. Nous ne pouvons émettre des recommandations fiables étant donné que les études qui comparaient les mêmes types de contraceptifs étaient rares. Un grand nombre d'essais avaient un petit nombre de participants et certains ont eu de nombreux arrêts prématurés. La notification des méthodes d'un grand nombre d'études antérieures était limitée.

Nous en savons toujours très peu sur les femmes à risque pour les problèmes métaboliques en raison du surpoids. Plus de la moitié des essais incluaient les restrictions de poids dans les critères d'inclusion. Seul un essai de petite taille avait stratifié les groupes par indice de masse corporelle (obèse par rapport à normal).

Plain language summary

Hormone contraceptives and how the body uses carbohydrates in women without diabetes

Hormone contraceptives may change how the body handles carbohydrates (starches and sugars). Changes may include lower ability to use sugar from food and more problems with the body's insulin. Insulin is a hormone that helps the body use sugar. Problems with blood sugar can increase risk for diabetes and heart disease. These issues have been raised mainly with birth control methods that contain the hormone estrogen.

In April 2014, we looked for randomized trials of how the body handles carbohydrates when using birth control methods with hormones. Outcomes were blood glucose or insulin levels. Birth control methods could contain estrogen and progestin or just progestin. The type could be pills, shots (injections), implants (matchstick-size rods put under the skin), the vaginal ring, or an intrauterine device (IUD). The studies had to compare two types of birth control or one type versus a placebo or 'dummy' method.

We included 31 trials. None had a placebo. Of 34 pairs of birth control methods compared, eight showed some difference by study groups. Twelve trials studied pills with desogestrel. The few differences were not consistent. Three trials looked at the etonogestrel ring. One showed the ring group had lower insulin than the pill group.

Eight trials looked at the progestin norethisterone. A group using norethisterone pills had less glucose change than those taking other pills. In another study, a group using the injectable ‘depo’ (depot medroxyprogesterone acetate) had higher glucose and insulin than the group using another injectable.

Of five new trials, two used different estrogen types. In one study, a group taking a pill with ethinyl valerate had lower glucose than a group taking a standard pill. Two other trials compared taking pills for several cycles without stopping (extended use) versus usual use. In one using a dienogest pill, the extended-use group had more glucose change. A small trial used two levonorgestrel pills, and looked at obese and normal weight women. The outcomes did not differ much between those groups.

In women without diabetes, hormone contraceptives have little effect on the body's carbohydrate use. Few studies compared the same types of birth control. Therefore, we cannot make strong statements. Many trials had small numbers of women, and many women dropped out. Older trials often did not report all the study methods. Many trials did not include overweight women.

Résumé simplifié

Contraceptifs hormonaux et mode d'utilisation par l'organisme des glucides chez les femmes sans diabète

Les contraceptifs hormonaux peuvent modifier la manière dont l'organisme se sert des glucides (amidons et sucres). Les modifications peuvent comprendre une capacité affaiblie à utiliser le sucre contenu dans les aliments et davantage de problèmes avec l'insuline de l'organisme. L'insuline est une hormone qui aide l'organisme à utiliser le sucre. Les problèmes liés à la glycémie peuvent augmenter le risque de diabète et de cardiopathie. Ces questions ont été soulevées principalement avec les méthodes de contraception qui contiennent l'hormone œstrogène.

En avril 2014, nous avons cherché des essais randomisés ayant étudié la manière dont l'organisme se sert des glucides dans le cas de l'utilisation de méthodes de contraception hormonale. Les critères de jugement étaient les niveaux de glucose dans le sang ou les niveaux d'insuline. Les méthodes de contraception pouvaient contenir de l'œstrogène et de la progestérone ou juste de la progestérone. Le type pouvait être des pilules, des piqûres (injections), des implants (tiges de la taille d'une allumette placées sous la peau), l'anneau vaginal, ou un dispositif intra-utérin (DIU). Les études devaient comparer deux types de contraception ou un type à un placebo ou à une méthode 'factice'.

Nous avons inclus 31 essais. Aucun ne comprenait de placebo. Sur les 34 paires de méthodes de contraception comparées, huit ont révélé quelques différences par groupes d'étude. Douze essais avaient étudié les pilules de désogestrel. Les quelques différences n'étaient pas cohérentes. Trois essais avaient examiné l'anneau d'étonogestrel. Un essai avait montré que le groupe utilisant l'anneau avait un niveau d'insuline inférieur à celui du groupe prenant la pilule.

Huit essais avaient examiné la noréthistérone (progestatif). Un groupe utilisant des pilules de noréthistérone présentait moins de modifications des niveaux de glucose que ceux prenant d'autres pilules. Dans une autre étude, un groupe utilisant le contraceptif injectable «Â Depo » (acétate de médroxyprogestérone-retard) avait des niveaux de glucose et d'insuline plus élevés que le groupe utilisant un autre contraceptif injectable.

Sur les cinq nouveaux essais, deux utilisaient des types d'œstrogène différents. Dans une étude, un groupe prenant une pilule avec de l'éthinyle valérate avait des niveaux de glucose plus bas qu'un groupe prenant une pilule standard. Deux autres essais comparaient la prise de pilules pendant plusieurs cycles sans interruption (utilisation prolongée) par rapport à une utilisation habituelle. Dans un essai utilisant une pilule de diénogest, le groupe ayant une utilisation prolongée présentait plus de modifications des niveaux de glucose. Un essai de petite taille utilisait deux pilules de lévonorgestrel, et s'intéressait aux femmes obèses et de poids normal. Les résultats n'étaient pas très différents entre ces groupes.

Chez les femmes sans diabète, les contraceptifs hormonaux n'ont que peu d'effet sur l'utilisation des glucides par l'organisme. Peu d'études comparaient les mêmes types de contraception. Par conséquent, nous ne pouvons émettre des recommandations fiables. Plusieurs essais avaient un petit nombre de femmes, et beaucoup de femmes avaient abandonné. Des essais plus anciens ne précisaient pas toujours toutes les méthodes de l'étude. Un grand nombre d'essais n'avaient pas inclus de femmes en surpoids.

Notes de traduction

Traduit par: French Cochrane Centre 1st June, 2012
Traduction financée par: Ministère du Travail, de l'Emploi et de la Santé Français

Laički sažetak

Steroidni kontraceptivi: učinak na metabolizam ugljikohidrata u žena koje nemaju diabetes mellitus (šećernu bolest)

Hormonski kontraceptivi mogu imati učinak na metabolizam ugljikohidrata (šećer i škrob). To znači smanjenu sposobnost iskorištavanja šećera iz hrane i lošiji učinak inzulina. Inzulin je hormon koji omogućuje tijelu korištenje šećera. Povišene koncentracije šećera (glukoze) u krvotoku povećavaju rizik nastanka dijabetesa i bolesti srca. Ova problematika se do sada uglavnom povezivala s učinkom kontraceptiva koji sadrže hormon estrogen.

U travnju 2014. godine pretražena su randomizirana klinička istraživanja o učinku hormonskih lijekova za sprječavanje nastanka neželjene trudnoće na metabolizam ugljikohidrata. Analizirani su rezultati koncentracije šećera (glukoze) ili inzulina u krvi. Metode sprječavanja nastanka neželjene trudnoće mogu sadržavati estrogene i progestine ili samo progestin. Mogu biti u obliku tableta, injekcija, subdermalnih implanatata (štapići veličine šibice koji se umetnu pod kožu), vaginalni prsteni ili maternični ulošci (tzv. „spirala“). Istraživanja su usporedila dva različita oblika kontracepcije ili jedan oblik u odnosu na placebo ili „lažni“ pripravak.

Analizirani su rezultati iz 31 kliničkog istraživanja. Ni u jednom se nije koristio placebo. Od ukupno 34 parova načina kontracepcije, za 8 je nađena razlika između ispitivanih skupina. U 12 kliničkih istraživanja su korištene tablete s dezogestrelom. Nađeno je nekoliko razlika, ali nisu bile dosljedne (sustavne). Tri klinička istraživanja su pratila učinak etonogestrelskog prstena. U jednom istraživanju je nađeno da je u skupini žena koje su koristile vaginalni prsten inzulin u krvi bio niži nego u skupini žena koje su uzimale tablete.

U 8 kliničkih istraživanja je ispitivan učinak progestin noretisterona. U skupini koja je koristila tablete s noretisteron je uočena manja promjena glukoze u krvi nego u skupini koja je koristila drugu vrstu tableta. U istraživanju u kojem je ispitivan učinak potkožnog depo-preparata (medroksiprogesteron acetat) koncentracije glukoze i inzulina su bile veće nego u skupini koja je koristila neku drugu vrstu preparata koji se primjenjuje injekcijom.

Od 5 novijih kliničkih istraživanja, u dva su korišteni drugi oblici estrogena. U studiji u kojoj su ispitivane tablete s etinil-valeratom je koncentracija glukoze bila niža nego u onih žena koje su koristile standardne tablete. U dvije studije je proučavan učinak produženog uzimanja tableta (tj. bez prekida) u odnosu na uobičajeni režim. Produženo uzimanje dienogest tableta je imalo više učinaka na glukozu. Tablete s levonogestrelom su u pretilih ispitanica i onih s normalnom tjelesnom težinom imale podudaran učinak.

U žena bez dijabetesa hormonska kontracepcija ima mali učinak na korištenje ugljikohidrata. Mali broj kliničkih istraživanja je usporedio isti oblik prevencije neželjene trudnoće. Zaključno nije moguće ustanoviti nedvojbeni učinak ovih preparata na metabolizam ugljikohidrata. U više istraživanja je broj ispitanica bio malen, i mnogo je ispitanica odustalo tijekom istraživanja. U istraživanjima starijeg datuma nisu navedene istraživačke metode, a također mnoga istraživanja nisu uključila pretile žene.

Bilješke prijevoda

Hrvatski Cochrane ogranak.
Prevela: Vesna Kušec

Background

Many hormonal contraceptives have been associated with changes in carbohydrate metabolism (Dorflinger 2002; Kahn 2003). Alterations may include decreased glucose tolerance and increased insulin resistance, which are risk factors for Type 2 diabetes mellitus and cardiovascular disease (Reaven 2005). These issues have been raised primarily with contraceptives containing estrogen.

To reduce side effects of steroidal contraceptives, changes have been made in the estrogen dose and newer progestins have been developed (Van der Mooren 1999). New formulations of combined oral contraceptives (COCs) are often examined for their relationship with carbohydrate metabolism (Gaspard 2003; Sitruk-Ware 2011). In addition, estrogens other than ethinyl estradiol have been developed for contraception (Agren 2011; Sitruk-Ware 2011), and these may have different metabolic effects. Progestin-only contraceptives do not carry the same risk of vascular complications as contraceptives that contain estrogen, so they may be recommended for women at risk for diabetes or vascular disease (Kahn 2003). However, progestin-only contraceptives may influence carbohydrate metabolism; results from studies with varying designs have been inconsistent (Dorflinger 2002; Kahn 2003; Kivela 2001).

Little is known about the effect of hormonal contraceptives among women at risk for metabolic problems as opposed to healthy populations (Dorflinger 2002). Glucose changes have been noted in users of depo-medroxyprogesterone acetate (DMPA) with greater body weight (Kahn 2003). Due to the risk of cardiovascular complications, the World Health Organization classifies combined contraceptives as category 2 for women with diabetes who do not have vascular disease and 3/4 if vascular disease is evident (WHO 2009). For category 2, the advantages of using the method generally outweigh the theoretical or proven risks. In category 3, use of the method is not usually recommended, unless more appropriate methods are not feasible. Category 4 implies the method should not be used under the circumstances. Progestin-only contraceptives are generally classed as category 2 for women with diabetes (WHO 2009). The exception is DMPA when vascular disease has been identified; then DMPA is category 3.

The prevalence of overweight (including obesity) has increased worldwide, along with the effects on insulin resistance (Reaven 2005). Chronic diseases have not been emphasized in development efforts, although they are a major problem in less developed countries (Strong 2005). Low- and middle-income countries account for 80% of deaths due to chronic disease. Overweight is associated with impaired glucose tolerance and increased insulin resistance (Reaven 2005), which may indicate increased risk for diabetes (CDC 2005; Kahn 2003). Overweight and obesity are generally determined with the body mass index (BMI), which is based on weight and height [BMI = weight (kg) / height (m)2] (CDC 2009). BMI generally reflects the amount of fat, whereas body weight mainly addresses overall body size. Frequently used BMI categories are 25 to 29.9 (kg/m2) for overweight and 30 or higher for obesity, although these cutoffs may not be optimal for all ethnic groups (Lopez 1992; Huxley 2005). A Cochrane review examined the effectiveness of hormonal contraceptives in among women who are overweight or obese versus women in a lower weight or BMI group (Lopez 2013a). The researchers only found three studies using BMI rather than body weight. Only one of the three studies showed a higher pregnancy risk for overweight women. The risk of oral contraceptive failure among overweight or obese women may depend on whether the assessment is based on perfect use or typical use (Trussell 2009).

We examined whether steroidal contraceptives affect carbohydrate metabolism among women without previously known diabetes mellitus. We also searched for data on women who were overweight and therefore at risk for diabetes. At the time of the initial review, no systematic review on this topic included only randomized controlled trials.

Objectives

To evaluate the effect of hormonal contraceptives on carbohydrate metabolism in healthy women and those at risk for diabetes due to overweight. The hypotheses are the following:

  • combined hormonal contraceptives do not cause clinically important changes in carbohydrate metabolism;

  • progestin-only contraceptives cause changes in carbohydrate metabolism among women at risk for diabetes due to being overweight.

Methods

Criteria for considering studies for this review

Types of studies

All randomized controlled trials (RCTs) were considered if they examined carbohydrate metabolism in women who used hormonal contraceptives for contraception. We excluded trials that focused on women with known diabetes, as another review covers that topic (Visser 2013).

We did not include trials if the reports had insufficient data for analysis. In 2012, we excluded such reports that had been in earlier versions of this review.

Types of participants

Women of reproductive age who participated in the identified trials.

Types of interventions

Interventions included comparisons of a hormonal contraceptive with a placebo, a non-hormonal contraceptive, or with another hormonal contraceptive that differed in drug, dosage, or regimen. Trial drug interventions must have included at least three consecutive cycles to be eligible. The contraceptives must have been provided for contraception.

We excluded interventions of hormone replacement therapy for postmenopausal women as well as hormonal contraceptives used to treat a specific health condition such as dysmenorrhea or polycystic ovary syndrome.

Types of outcome measures

Outcomes included glucose and insulin levels, which were reported as fasting value or response to a glucose or insulin tolerance test. In the case of a tolerance test, the measure was usually the area under the curve (AUC) during the three-hour test. If available, data were included on the number of women with glucose outside the normal range. Other outcome measures included glycosylated hemoglobin (HbA1c), which reflects longer-term glucose control. In addition, C-peptide levels reflect the amount of insulin produced by the body.

Search methods for identification of studies

Electronic searches

In April 2014, we searched for studies of hormonal contraceptives and carbohydrate metabolism in PubMed, POPLINE, Cochrane Central Register of Controlled Trials (CENTRAL), and LILACS. In addition, we searched for recent clinical trials through ClinicalTrials.gov and the International Clinical Trials Registry Platform (ICTRP). The search strategies are given in Appendix 1. The previous search strategies can be found in Appendix 2.

Searching other resources

For the initial review, we examined references lists of relevant articles. We also wrote to known investigators for information about other published or unpublished trials not found in our search.

Data collection and analysis

Selection of studies

We assessed for inclusion all titles and abstracts identified during the literature searches with no language limitations. Studies were generally included if they met the criteria noted above. However, two studies were excluded because of scientific fraud (Briggs 1980; Briggs 1982) as noted in Rossiter 1992.

Data extraction and management

One author extracted the data and entered the information into RevMan. A second author conducted the secondary data extraction and verified correct data entry. Any discrepancies were resolved by discussion. We wrote to researchers for missing data such as means and standard deviations for results shown in figures only. For the initial review, we also asked about study design issues missing from the reports. However, for the updates, the studies were recent and could be expected to follow CONSORT (Moher 2001; Schulz 2010).

Assessment of risk of bias in included studies

Studies were examined for methodological quality, according to recommended principles (Higgins 2011). Factors considered included study design, randomization method, allocation concealment, blinding, and losses to follow-up and early discontinuation.

Measures of treatment effect

We generally entered the data in the units in which they were reported and the units are specified in the tables. The exception was fasting glucose for Ludicke 2002, which was converted from g/L to μg/dl in order to combine the data with those from Rechberger 2004. Similar outcomes were often measured with similar units across studies, e.g., fasting serum glucose or fasting serum insulin, so the reader can examine the values across studies. Researchers may have measured the outcome based on serum, plasma, or whole blood, which can affect the absolute values. However, the primary interest for this review was the difference between groups within a study, so the actual units were less important than size of the mean difference.

In some cases, the insulin results from different studies were not in the same range, even when units were converted into the same system of measurement, so the data were not combined in a meta-analysis (Bowes 1989 and Gillespy 1991; Ludicke 2002 and Rechberger 2004). The researchers may have used different assays. Also, studies were not combined if different measures were used, such as plasma versus blood values. For one report (WHO 1998), standard deviations were calculated from the 95% CI.

Glucose and insulin measures were generally presented as fasting values or the area under the curve (AUC) during an oral glucose tolerance test (OGTT). If the AUC data were not available, the two-hour response values were used. If change data were available, those results were used rather than values at a point in time. While several studies reported change, few included the variation data needed for analysis.

Assessment of heterogeneity

We tested for heterogeneity where relevant; none was evident. Fixed-effect and random-effects will give the same result if no heterogeneity exists, which is also the case if a comparison includes a single study. There is no consensus regarding the use of either model.

Data synthesis

For continuous variables, such as the mean area under the curve for glucose, the mean difference was computed with 95% confidence interval (CI) using a fixed-effect model. RevMan uses the inverse variance approach (Higgins 2011). A fixed-effect model does not require the assumption of normal distribution for the effects. Fixed and random effects give the same result if no heterogeneity exists, as when a comparison includes only one study.

For dichotomous outcomes, the Peto odds ratio (OR) with 95% CI was calculated. An example is the proportion of women who had blood glucose categorized as outside the normal range. The Peto OR is useful when treatment effects are small and when events are not very common (Higgins 2011). This approach performs well under many circumstances, except when the study arms are severely unbalanced, which rarely occurs in RCTs (Deeks 2001).

Results

Description of studies

Results of the search

In 2014, the search produced 36 unduplicated references. No trial met the eligibility criteria. Searches of ClinicalTrials.gov and ICTRP yielded 39 unduplicated listings, but none were relevant.

The 2011 electronic search produced 58 references; 5 were potentially eligible and full text was obtained for each. In addition, two previously ongoing trials had reports available. Results for one were posted on ClinicalTrials.gov and we located the relevant publications (Agren 2011), a primary paper, and a secondary report. For the other, we found an 'in press' article with the carbohydrate metabolism results; the report has since been published (Beasley 2012).

Included studies

A total of 31 trials met the inclusion criteria and provided sufficient data for analysis in this review. Most (N=21) compared different types of combined oral contraceptives (COCs). The COCs varied in the types and doses of progestin and estrogen. Two trials compared different regimens of the same COC (Machado 2010; Wiegratz 2010). The remaining eight trials compared progestin-only pills (Ball 1991), injectables (Benagiano 1997; Fahmy 1991; WHO 1998), a vaginal ring versus a COC (Cagnacci 2009b; Duijkers 2004; Elkind-Hirsch 2007), and implants (Biswas 2001).

Most of the studies measured glucose and insulin by fasting levels or by area under the curve (AUC) during an oral glucose tolerance test (OGTT). Others included C-peptide or glycosylated hemoglobin (HbA1c). The percent glycosylation reflects the amount of glucose available in the prior two to six weeks (Kivela 2001). C-peptide levels during OGTT reflect the amount of insulin being produced by the body, and may be more accurate than insulin levels in the blood (Skouby 2005). Some recent studies included calculated estimates of insulin resistance and insulin sensitivity.

We found little data to assess the effects of steroidal contraceptives on carbohydrate metabolism in women who were overweight. One small trial stratified by weight (Beasley 2012). More than half of the trials (18/31) had weight restrictions as eligibility criteria, such as body mass index (BMI) less than 25 or 30 kg/m2 or body weight within 20% of 'ideal'. Some of the included trials reported on weight gain in participants. However, weight changes have been examined for users of combination contraceptives (Gallo 2011) and for users of progestin-only contraceptives (Lopez 2013b).

Duration of the trials ranged from 3 cycles to 24 months. Of the 31 trials, 15 had treatment duration of 6 or 7 cycles or 6 months and 9 trials reported durations of 12 months or 13 cycles. The remainder were 3 to 5 cycles (N=3), 9 cycles (N=2), or 24 cycles or months (N=2).

Sample sizes were generally small. WHO recommendations for metabolic studies include 40 subjects per group (Michal 1989), which only eight trials had (Agren 2011; Biswas 2001; Bowes 1989; Duijkers 2004; Reisman 1999; WHO 1985; Van der Mooren 1999; WHO 1998). Nine trials had fewer than 25 women in a comparison group.

Excluded studies

For the 2012 update, we excluded 16 trials from previous versions that did not have sufficient data for analysis. Two other potentially eligible trials were also excluded due to insufficient data (Rad 2011; Winkler 2010). Appendix 3 has descriptive results from those trials. In addition, a previously included crossover trial was excluded (Song 1992). Outcome data for this small trial (N=12) were not available for each treatment segment prior to crossover.

Risk of bias in included studies

Reporting was limited for many trials, but particularly for those preceding CONSORT (Moher 2001; Schulz 2010).

Allocation

We obtained information on randomization for 17 trials. Eleven studies initially reported some information on randomization, while six provided further information via correspondence (Duijkers 2004; Fahmy 1991; Gaspard 2003; Luyckx 1986; Prasad 1989; WHO 1998). Methods included a computer-generated list or randomization code (Beasley 2012; Bowes 1989; Cagnacci 2009a; Cagnacci 2009b; Duijkers 2004; Elkind-Hirsch 2007; Junge 2011; Klipping 2005; Machado 2010; Prasad 1989; WHO 1998; Wiegratz 2010), randomly permuted block sizes (Agren 2011; Gaspard 2003; Luyckx 1986), allocation at 1:1 and stratified by site (Reisman 1999), and paper slips placed into sealed envelopes and mixed (Fahmy 1991).

Adequate methods for allocation concealment include a centralized telephone system and the use of sequentially-numbered, opaque, sealed envelopes (Schulz 1995; Schulz 2002a). Pharmacy distribution of pills is another good method. Seven trials reported on concealment and five provided further information via correspondence (Duijkers 2004; Fahmy 1991; Gaspard 2003; Junge 2011; Luyckx 1986; WHO 1998). Ten trials were considered low risk due to the use of an interactive voice response system (Agren 2011), central randomization service (Gaspard 2003; Luyckx 1986), sequentially-numbered, sealed envelopes (Beasley 2012; Fahmy 1991; Machado 2010; Reisman 1999), numbered sealed packages for study drug (Wiegratz 2010), or a concealed list (Cagnacci 2009a; Cagnacci 2009b). Two reported no allocation concealment (Duijkers 2004; WHO 1998). Most trials (N=19) had unclear methods or provided no information on allocation concealment; five of those trials were dated after 2001.

Blinding

Of 31 trials, 22 provided information on whether the study had any blinding. Of the 22 trials, 19 initially mentioned any blinding or not and 3 provided information via correspondence. A total of 14 were open-label, 3 were double-blind, 1 was single-blind, 2 noted the analysis was blinded, 1 said the lab assessments were blinded, and 1 said the physician and lab analysts were blinded. Nine trials had no information on blinding; two were published after 2001.

Incomplete outcome data

Losses greater than 20% threaten trial validity (Strauss 2005). Eight trials had losses, exclusions, or discontinuations totaling more than 20%: Ball 1991 (35%); Beasley 2012 (36%); Benagiano 1997 (27%); Biswas 2001 (31%); Elkind-Hirsch 2007 (35%); Prasad 1989 (35%); Reisman 1999 (43%); and WHO 1985 (44%). Furthermore, three had differential losses between the comparison groups (Ball 1991; Beasley 2012; Biswas 2001). One study did not provide enough information to determine losses (Gillespy 1991). The losses included data from women who did not return for follow-up, had protocol deviations, or discontinued early. Exclusion of participants after randomization can bias the results (Schulz 2002b). Trials excluded data due to measurement error (Benagiano 1997), women who were not consistent users of the assigned OC (Beasley 2012), and those who did not take any study medication (Agren 2011; Junge 2011; Machado 2010; Wiegratz 2010).

Effects of interventions

Of the 34 comparisons analyzed, 8 showed any notable differences between the study groups. Results are organized by the type of comparison, which is focused on the type and dose of progestin. While estrogen may also have an effect, the type was generally ethinyl estradiol (EE) although two recent trials examined 17β-estradiol (E2). No trial included a placebo. The first group has comparisons with desogestrel, including a subgroup for etonogestrel, a metabolite of desogestrel. The next has COC and injectable preparations containing norethisterone. 'Other comparisons' are mostly from recent trials, including those examining newer COC preparations, extended versus conventional regimens, and women with normal BMI versus obese women.

Of the 31 trials, 7 mentioned whether any individuals had abnormal values for glucose or insulin (dichotomous data). Two studies reported that all individual values were within normal ranges or that no impaired tolerance was evident (Bowes 1989; Gaspard 2003). Five trials reported that some women had values outside the normal range (Ball 1991; Beasley 2012; Elkind-Hirsch 2007; WHO 1985; WHO 1998).

Due to the types of progestin compared or the doses of progestin or estrogen, meta-analysis was possible for only five comparisons, most of which included older trials. Data were combined as follows:

Desogestrel

Combined oral contraceptives

Three trials compared desogestrel 150 µg + EE (20 or 30 µg) to gestodene (60 or 75 µg) + EE (15, 20, or 30 µg) (Ludicke 2002; Rechberger 2004; Van der Mooren 1999). Regardless of the preparations compared, the study arms were not significantly different for the glucose or insulin measures (Analysis 1.1 to Analysis 4.2), including one meta-analysis for fasting glucose (Analysis 2.1).

Another three trials compared desogestrel 150 µg + EE (20 or 30 µg) versus desogestrel 150 µg + EE 30 µg (Basdevant 1993; Cagnacci 2009b) or versus desogestrel 125 µg + EE 50 µg (Luyckx 1986). The study arms were not significantly different for the glucose and insulin measures (Analysis 5.1 to Analysis 6.2). Cagnacci 2009a compared desogestrel 150 µg + EE 20 µg versus chlormadinone acetate 2 mg + EE 30 µg, and also showed that the study arms were not significantly different for the carbohydrate metabolism measures (Analysis 7.1 to Analysis 7.3).

Desogestrel-containing COCs were compared with levonorgestrel-containing COCs in four trials. When desogestrel 150 µg + EE 30 µg was compared with levonorgestrel 150 µg + EE 30 µg, glucose results were not consistent across measures and trials. In a meta-analysis of fasting glucose for Liukko 1987 and Prasad 1989, the desogestrel-COC group had a higher mean at cycle 6 (mean difference (MD) 0.20; 95% CI 0.00 to 0.41) (Analysis 8.1), but not at cycle 12 (Analysis 8.2). Where the trial data could not be combined, the desogestrel-COC group had a lower mean for fasting glucose at cycle 24 (Liukko 1987) (MD -0.40; 95% CI -0.72 to -0.08) (Analysis 8.3) and a higher mean for two-hour glucose response at cycle 12 (Prasad 1989) (MD 1.08; 95% CI 0.45 to 1.71) (Analysis 8.5) that was not apparent at cycle 6 (Analysis 8.4). When desogestrel 150 µg + EE 30 µg was compared with levonorgestrel 50-75-125 μg + EE 30-40-30 μg (Luyckx 1986; Prasad 1989), the groups were not significantly different in the glucose or insulin measures (Analysis 9.1 to Analysis 9.6). Luyckx 1986 examined desogestrel 125 µg + EE 50 µg versus the levonorgestrel 50-75-125 μg + EE 30-40-30 μg. The mean AUC for glucose was not significantly different for the groups after six months (Analysis 10.1), but the mean AUC insulin was higher for the desogestrel-COC group (MD 20.30; 95% CI 4.24 to 36.36) (Analysis 10.2). Knopp 2001 compared desogestrel 50-100-150 μg + EE 35-30-30 μg to levonorgestrel 50-75-125 μg + EE 30-40-30 μg, and again the groups were not significantly different for changes in the glucose or insulin measures (Analysis 11.1; Analysis 11.2).

Gaspard 2003 and Klipping 2005 compared drospirenone 3 mg + EE 30 µg versus a COC containing desogestrel 150 µg + EE 30 µg and EE 20 µg, respectively. The study arms within each trial were not significantly different for changes in the measures of carbohydrate metabolism (Analysis 12.1 to Analysis 13.8).

Etonogestrel vaginal ring and implant

Etonogestrel is the active metabolite of desogestrel. Three trials compared a vaginal ring (etonogestrel 120 µg + EE 15 µg) to a COC (Cagnacci 2009b; Duijkers 2004; Elkind-Hirsch 2007). In Elkind-Hirsch 2007 at cycle five, insulin sensitivity was not significantly different for the ring group compared to the group assigned to levonorgestrel 100 µg + EE 20 µg (Analysis 14.1; Analysis 14.2). For Cagnacci 2009b, changes in the carbohydrate metabolism measures were not significantly different for the ring group versus a group assigned to desogestrel 150 µg plus either EE 30 µg or EE 20 µg (Analysis 15.1 to Analysis 15.3). The only difference noted was in Duijkers 2004; the ring group had lower a mean at cycle six for insulin AUC (MD -204.51; 95% CI -389.64 to -19.38) (Analysis 16.2) compared to the group with levonorgestrel 150 µg + EE 30 µg. For other outcomes, the groups were not significantly different (Analysis 16.1; Analysis 16.3).

In Biswas 2001, an implant with etonogestrel (68 mg) was compared to a levonorgestrel (216 mg) implant. The groups were not significantly different in the carbohydrate metabolism measures (Analysis 17.1 to Analysis 17.10).

Norethisterone

Combined oral contraceptives

Several trials examined preparations containing norethisterone, sometimes referred to as norethindrone. We used the terminology provided in the reports. Two studies compared a COC with norethisterone 1000 µg + EE 35 µg versus levonorgestrel 150 µg + EE 30 µg (Loke 1992; WHO 1985). The study arms were not significantly different for most of the glucose measures (Analysis 18.1 to Analysis 18.3). However, in WHO 1985 , the mean change in glucose two-hour response was lower for the norethisterone-COC group than the levonorgestrel group (mean difference -0.30; 95% CI -0.54 to -0.06) (Analysis 18.4). WHO 1985 included data from seven centers but had a 44% loss; Loke 1992 was a small trial. WHO 1985 reported that 8% showed impaired glucose tolerance at 3 and 12 months, but the report did not provide group-level data for that measure. A third study compared two progestin-only pills in a small trial (Ball 1991). At six months, the glucose measures were not significantly different for the norethisterone 350 µg and levonorgestrel 30 µg groups (Analysis 19.1 to Analysis 19.4).

Three trials examined triphasic COCs containing norethindrone. The study arms were not significantly different in the glucose or insulin outcomes. Reisman 1999 examined norethindrone 500-750-1000 μg + EE 35 μg versus levonorgestrel 100 μg + EE 20 μg for four cycles (Analysis 20.1). Two trials compared norethindrone 500-750-1000 μg + EE 35 μg versus levonorgestrel 50-75-125 μg + EE 30-40-30 μg (Bowes 1989; Gillespy 1991). The study groups were not significantly different in the glucose or insulin outcomes (Analysis 21.1 to Analysis 21.10), including in a meta-analyses for fasting glucose and for glucose AUC at cycle six. Bowes 1989 also examined norethindrone 500-1000-500 μg + EE 35 μg versus levonorgestrel 50-75-125 μg + EE 30-40-30 μg (Analysis 22.1 to Analysis 22.4) and norethindrone 500-1000-500 μg + EE 35 μg versus norethindrone 500-750-1000 μg + EE 35 μg (Analysis 23.1 to Analysis 23.4). Again, the comparison groups were not significantly different in the carbohydrate metabolism outcomes.

Injectable contraceptives

Three trials examined injectables containing norethisterone. Fahmy 1991 compared progestin-only methods: depot medroxyprogesterone acetate 150 mg (DMPA) versus norethisterone enanthate 200 mg. After 12 months, the DMPA group had higher means compared to the norethisterone injectable for fasting serum glucose (mean difference 10.05; 95% CI 3.16 to 16.94) (Analysis 24.1), glucose two-hour response (mean difference 17.00; 95% CI 5.67 to 28.33) (Analysis 24.2), and fasting serum insulin (mean difference 3.40; 95% CI 2.07 to 4.73) (Analysis 24.3) but not for insulin two-hour responses (Analysis 24.4). Two studies examined combined injectables. Benagiano 1997 studied norethisterone enanthate 50 mg + EV 5 mg versus a COC of norethisterone 500-750-1000 µg + EE 35 µg. WHO 1998 compared the same norethisterone enanthate injectable versus medroxyprogesterone acetate 25 mg + estradiol cypionate 5 mg. Both Benagiano 1997 and WHO 1998 showed the groups were not significantly different in carbohydrate metabolism after six and nine cycles, respectively (Analysis 25.1 to Analysis 26.3). Benagiano 1997 was a small study while WHO 1998 included data from four centers.

Other comparisons

Other multiphasic COCs

Two older trials compared multiphasic COCs. Ball 1990 examined gestodene 50-70-100 μg + EE 30-40-30 μg versus levonorgestrel 50-75-125 μg + EE 30-40-30 μg). The groups were not significantly different in the carbohydrate metabolism measures at cycle six (Analysis 27.1; Analysis 27.2). Bloch 1979 studied levonorgestrel 50-125 μg + EE 50 μg versus levonorgestrel 150 μg + EE 30 μg. The groups were not significantly different for 'random' blood glucose, which is not taken after fasting or a glucose tolerance test (Analysis 28.1).

COCs containing other estrogens

Two trials studied newer COCs containing 17β-estradiol (E2) or estradiol valerate (E2V); 1 mg E2V corresponds to 0.76 mg E2. Agren 2011 examined nomegestrel acetate 2 mg (NOMAC) + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg. Compared to the levonorgestrel-COC group at cycle six, the NOMAC group had a lower mean for incremental AUC glucose (MD -1.43; 95% CI -2.47 to -0.39) (Analysis 29.2) and for HbA1c (MD -0.10; 95% CI -0.17 to -0.03) (Analysis 29.5). However, the study arms were not significantly different for AUC glucose, AUC insulin, and incremental AUC insulin (respectively, Analysis 29.1; Analysis 29.3; Analysis 29.4). Junge 2011 compared multiphasic preparations: dienogest 0-2-3-0 mg + estradiol valerate (E2V) 3-2-2-1 mg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg. At cycle seven, the study groups were not significantly different for AUC glucose (Analysis 30.1) or AUC insulin (Analysis 30.2).

Extended versus cyclic regimens

Two trials compared different regimens of the same study drug. With dienogest 2 mg + EE 30 μg, Wiegratz 2010 compared an extended-cycle regimen (84+7 days) versus conventional treatment (21+7 days). After 12 cycles, the extended-use group had a greater mean change in AUC glucose compared to the conventional-use group (MD 82.00; 95% CI 10.72 to 153.28) (Analysis 31.5). However, the study groups were not significantly different for change in fasting HbA1c, glucose, insulin, AUC insulin, AUC C-peptide, and calculated insulin resistance and insulin sensitivity (Analysis 31.1 to Analysis 31.4; Analysis 31.6 to Analysis 31.9). Machado 2010 compared continuous use (168 days) versus cyclic use (6 cycles) of drospirenone 3 mg + EE 30 µg. The study groups were not significantly different for fasting glucose or insulin at the last assessment (Analysis 32.1; Analysis 32.2).

Obese versus normal weight women

Beasley 2012 was the only trial to examine changes in carbohydrate metabolism for obese women versus normal women. The COCs used were levonorgestrel 100 µg + EE 20 µg and levonorgestrel 150 µg + EE 30 µg. The COC groups were stratified by BMI (normal, BMI 19 to 24.9 kg/m2; obese, BMI 30 to 39.9 kg/m2). The BMI groups were not significantly different for changes in fasting glucose or insulin or in calculated insulin resistance (natural log used for analysis) (Analysis 33.1 to Analysis 33.3). When the levonorgestrel COC groups were compared, the group assigned to levonorgestrel 100 µg + EE 20 µg had a smaller mean change in fasting glucose compared to the group with levonorgestrel 150 µg + EE 30 µg (MD -3.00; 95% CI -5.89 to -0.11) (Analysis 34.1). The COC groups were not significantly different for changes in the other measures (Analysis 34.2; Analysis 34.3). The report noted impaired fasting glucose at follow-up in 4% of the normal women and 8% of the obese women but reportedly none developed overt diabetes.

Discussion

Due to differences in the types of progestin compared or in the doses of progestin or estrogen, only five comparisons involved meta-analysis. For many of the trials, carbohydrate metabolism was a secondary outcome.

Summary of main results

Our first hypothesis appears to be supported by the results: combined hormonal contraceptives do not cause clinically important changes in carbohydrate metabolism in women without diabetes. In the trials that analyzed change, the amount of change was generally small. When newer formulations were compared with other COCs, including 'standard' pills like levonorgestrel 150 µg plus EE 30 µg, few differences were noted in the measures of carbohydrate metabolism. However, no trial was placebo-controlled, which is not unusual for contraceptive studies due to ethical concerns.

Of 34 comparisons, 31 examined compared different hormonal contraceptives, two examined different regimens, and one stratified by body mass index groups within the contraceptive groups. Eight of the comparisons showed any notable differences. These results were from trials that examined COCs containing desogestrel, norethisterone, levonorgestrel, nomegestrel acetate plus 17β-estradiol; an extended regimen of a dienogest-containing COC; the etonogestrel vaginal ring; and the injectable depot medroxyprogesterone acetate (DMPA).

Ten trials compared desogestrel-containing COCs versus levonorgestrel-containing COCs. A few differences were noted in three older trials, but the results were not consistent across trials, outcomes, or time points. A meta-analysis of two studies showed higher mean fasting glucose with the desogestrel preparation, while analyses within the individual studies showed a lower mean fasting glucose level and a higher two-hour glucose response. The third study showed higher mean AUC insulin with the desogestrel-COC; AUC is the area under the curve during a glucose tolerance test. These seemingly contradictory results were discussed in Henzl 2000. However, the doses of desogestrel and estrogen varied across studies and the comparison COCs varied in composition, too.

Drospirenone is one of the 'newer' progestins, and is derived from spironolactone (Sitruk-Ware 2011). Two studies of drospirenone COCs versus desogestrel COCs indicated that the effects on carbohydrate metabolism were limited and not significantly different from those of the comparison formulations.

Three trials examined the etonogestrel-releasing vaginal ring versus a COC containing levonorgestrel or desogestrel. Etonogestrel is the active metabolite of desogestrel. The one notable difference showed the ring group had a lower mean AUC for insulin. In addition, one trial studied an etonogestrel implant with no notable results.

Norethisterone is part of the estrane group of progestins, which are derived from testosterone (Sitruk-Ware 2011). The gonane group was developed later, and includes levonorgestrel and its derivatives, e.g., desogestrel and gestodene among others (Henzl 2000). These groups of progestins differ metabolically (Wallach 2000). Five trials examined COC preparations containing norethisterone, also referred to as norethindrone. One difference was noted in which the norethisterone-COC had a lower mean change in glucose two-hour response compared to a levonorgestrel-COC group, and that large trial had a high loss. Henzl 2000 had suggested the effects of norethindrone with EE on carbohydrate metabolism were modest. In addition, three older trials examined injectables containing norethisterone. In one study, the group assigned to depot medroxyprogesterone acetate (DMPA) had higher means for fasting glucose, glucose two-hour response, and fasting insulin compared to the group using norethisterone enanthate. Medroxyprogesterone acetate is a part of the early group of progestins called pregnanes (Henzl 2000).

The five newest trials had other comparisons. Two examined newer COCs containing 17β-estradiol (E2) or estradiol valerate (E2V). In one, the group with nomegestrel acetate plus E2 had a lower mean than a levonorgestrel-COC group for incremental AUC glucose and for glycosylated hemoglobin (HbA1c). Another two trials compared extended versus conventional (cyclic) regimens. With a dienogest-containing COC, the extended-use group had a greater mean change in AUC glucose. One trial using two levonorgestrel COCs stratified by body mass index (normal or obese). While the BMI groups did not differ, the group assigned to the lower dose preparation showed a smaller mean change in fasting glucose. The study was small and had high losses.

Overall completeness and applicability of evidence

The 34 comparisons provided adequate Information to examine the effects of various hormonal contraceptives on carbohydrate metabolism. However, only 7 of 31 trials mentioned whether any individuals had abnormal values for glucose or insulin (dichotomous data). While no trial was placebo-controlled, many studies compared a newer contraceptive to an older or more commonly used one. Those trials examined whether the newer formulation had a greater effect on carbohydrate metabolism than the comparison contraceptive.

Data were insufficient to test our second hypothesis that progestin-only contraceptives cause changes in carbohydrate metabolism among women at risk for diabetes due to being overweight. More than half of the trials had weight restrictions as inclusion criteria. Only one small trial with COCs stratified by body mass index (obese versus normal) but the study had high losses. Therefore, we still know very little about women at risk for metabolic problems due to being overweight or obese.

Quality of the evidence

Reporting was limited for many trials, but particularly for those preceding CONSORT (Schulz 2010; Moher 2001). Of 14 trials without information on randomization, only 2 were published after 2001. Ten trials were considered low risk for allocation concealment and two had no concealment. Of the 19 with unclear methods or no information for allocation concealment, only 5 were dated after 2001. Regarding any blinding, 14 were open-label and 9 had no information, only of which two were published after 2001.

Carbohydrate metabolism was a secondary objective for many studies. Sample sizes were usually determined for the primary outcomes and may have been insufficient for the outcomes of interest in this review. Many trials had small sample sizes; nine had fewer than 25 women in a group, and only eight studies had more than 40 women per group.

Eight trials had losses, exclusions, or discontinuations totaling more than 20%, and three had differential losses between the comparison groups. The losses included data from women who did not return for follow-up, had protocol deviations, or discontinued early. Six trials excluded participants from the primary analysis after randomization: an older trial excluded due to measurement error, a recent study excluded women who were not consistent users of the assigned OC, and four recent trials excluded those who did not take any study medication.

Potential biases in the review process

After we drafted the review, we explored whether pharmaceutical company involvement was related to showing any statistically significant difference. We added the pharmaceutical involvement to the Risk of bias tables in Characteristics of included studies. We summarized the information in Table 1. Some statistically significant result was shown in 5 of 13 trials funded by pharmaceutical companies. Another four trials had report authors from pharmaceutical companies, but the reports did not have funding information. One of the four had a statistically significant difference. While the proportion of differences appears slightly larger for those with pharmaceutical funding, the number of studies is small. In addition, the statistically significant differences did not consistently favor the experimental intervention.

Table 1. Summary of trial sponsorship and study results
  1. 1Details are in Characteristics of included studies, Risk of bias tables.

    2Any statistically significant result; details can be found in Effects of interventions.

Pharmaceutical company involvement1Number of studies with
significant difference2
Number of
studies
Proportion
Funding5130.38
Authorship; funding not mentioned140.25
Study product provided120.5
None likely; other support identified030
No information290.22
Total9310.29

Authors' conclusions

Implications for practice

Current evidence suggests no major differences in carbohydrate metabolism between different hormonal contraceptives in women without diabetes. Most effects seemed to be limited, yet not consistent across the types of contraceptives studied. We cannot make strong statements due to having few studies that compared the same types of contraceptives. However, we still know very little about women at risk for metabolic problems due to being overweight or obese. Most of the women in these trials were healthy and of normal weight.

Implications for research

Some progestins appeared to have more favorable results regarding carbohydrate metabolism, but none were clearly superior. More studies of the same contraceptives would help in making recommendations for clinical practice. Many trials had small numbers of participants and large losses to follow-up. More recent trials had improved study design and more complete reporting. Only one small study compared normal and obese women. The field needs trials that include women at risk for metabolic problems due to being overweight or obese.

Acknowledgements

From FHI 360:

  • Carol Manion developed and implemented several of the database searches.

  • For the 2012 update, Mario Chen and Sarah Mullins did the secondary data abstraction.

  • In 2014, Thomas Grey helped review the search results.

Data and analyses

Download statistical data

Comparison 1. Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting blood glucose (mmol/L) at cycle 61118Mean Difference (IV, Fixed, 95% CI)0.0 [-0.15, 0.15]
2 Fasting blood insulin (mlU/L) at cycle 61118Mean Difference (IV, Fixed, 95% CI)-0.40 [-2.34, 1.54]
3 Fasting C-peptide (pmol/L) at cycle 61118Mean Difference (IV, Fixed, 95% CI)-0.04 [-0.12, 0.04]
4 Glucose AUC (min x mmol/L) at cycle 61116Mean Difference (IV, Fixed, 95% CI)42.00 [-26.35, 110.35]
5 Insulin AUC (min x mlU/L) at cycle 61116Mean Difference (IV, Fixed, 95% CI)-327.00 [-1807.37, 1153.37]
6 C-peptide AUC (min x nmol/L) at cycle 61116Mean Difference (IV, Fixed, 95% CI)-1.0 [-38.14, 36.14]
Analysis 1.1.

Comparison 1 Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg, Outcome 1 Fasting blood glucose (mmol/L) at cycle 6.

Analysis 1.2.

Comparison 1 Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg, Outcome 2 Fasting blood insulin (mlU/L) at cycle 6.

Analysis 1.3.

Comparison 1 Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg, Outcome 3 Fasting C-peptide (pmol/L) at cycle 6.

Analysis 1.4.

Comparison 1 Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg, Outcome 4 Glucose AUC (min x mmol/L) at cycle 6.

Analysis 1.5.

Comparison 1 Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg, Outcome 5 Insulin AUC (min x mlU/L) at cycle 6.

Analysis 1.6.

Comparison 1 Desogestrel 150 µg + EE 20 µg versus gestodene 60 µg + EE 15 µg, Outcome 6 C-peptide AUC (min x nmol/L) at cycle 6.

Comparison 2. Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dL) after 6 cycles257Mean Difference (IV, Fixed, 95% CI)0.34 [-3.99, 4.67]
2 Fasting plasma insulin (µU/ml) after 6 cycles131Mean Difference (IV, Fixed, 95% CI)-0.80 [-3.16, 1.56]
3 Fasting plasma insulin (mIU/ml) after 6 cycles126Mean Difference (IV, Fixed, 95% CI)4.10 [-0.57, 8.77]
4 Fasting C-peptide (pmol/L) after 6 cycles131Mean Difference (IV, Fixed, 95% CI)13.0 [-45.80, 71.80]
5 Glucose AUC (h x g/L) after 6 cycles131Mean Difference (IV, Fixed, 95% CI)-0.21 [-0.85, 0.43]
6 Insulin AUC (h x IU/ml) after 6 cycles131Mean Difference (IV, Fixed, 95% CI)14.0 [-76.09, 104.09]
7 C-peptide AUC (h x pmol/L) after 6 cycles131Mean Difference (IV, Fixed, 95% CI)-89.0 [-587.05, 409.05]
8 Fasting glucose (g/L) after 13 cycles131Mean Difference (IV, Fixed, 95% CI)-0.02 [-0.07, 0.03]
9 Fasting insulin (IU/ml) after 13 cycles131Mean Difference (IV, Fixed, 95% CI)-1.0 [-3.74, 1.74]
10 Fasting C-peptide (pmol/L) after 13 cycles131Mean Difference (IV, Fixed, 95% CI)4.0 [-42.84, 50.84]
11 Glucose AUC (h x g/L) after 13 cycles131Mean Difference (IV, Fixed, 95% CI)0.29 [-0.93, 1.51]
12 Insulin AUC (h x IU/ml) after 13 cycles131Mean Difference (IV, Fixed, 95% CI)0.0 [-80.75, 80.75]
13 C-peptide AUC (h x pmol/L) after 13 cycles131Mean Difference (IV, Fixed, 95% CI)494.00 [-694.17, 1682.17]
Analysis 2.1.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 1 Fasting plasma glucose (mg/dL) after 6 cycles.

Analysis 2.2.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 2 Fasting plasma insulin (µU/ml) after 6 cycles.

Analysis 2.3.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 3 Fasting plasma insulin (mIU/ml) after 6 cycles.

Analysis 2.4.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 4 Fasting C-peptide (pmol/L) after 6 cycles.

Analysis 2.5.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 5 Glucose AUC (h x g/L) after 6 cycles.

Analysis 2.6.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 6 Insulin AUC (h x IU/ml) after 6 cycles.

Analysis 2.7.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 7 C-peptide AUC (h x pmol/L) after 6 cycles.

Analysis 2.8.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 8 Fasting glucose (g/L) after 13 cycles.

Analysis 2.9.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 9 Fasting insulin (IU/ml) after 13 cycles.

Analysis 2.10.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 10 Fasting C-peptide (pmol/L) after 13 cycles.

Analysis 2.11.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 11 Glucose AUC (h x g/L) after 13 cycles.

Analysis 2.12.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 12 Insulin AUC (h x IU/ml) after 13 cycles.

Analysis 2.13.

Comparison 2 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 20 µg, Outcome 13 C-peptide AUC (h x pmol/L) after 13 cycles.

Comparison 3. Desogestrel 150 µg + EE 20 µg versus gestodene 75 µg + EE 20 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dL) after cycle 6128Mean Difference (IV, Fixed, 95% CI)-1.0 [-6.24, 4.24]
2 Fasting plasma insulin (mIU/mL) after cycle 6128Mean Difference (IV, Fixed, 95% CI)4.20 [-3.18, 11.58]
Analysis 3.1.

Comparison 3 Desogestrel 150 µg + EE 20 µg versus gestodene 75 µg + EE 20 µg, Outcome 1 Fasting plasma glucose (mg/dL) after cycle 6.

Analysis 3.2.

Comparison 3 Desogestrel 150 µg + EE 20 µg versus gestodene 75 µg + EE 20 µg, Outcome 2 Fasting plasma insulin (mIU/mL) after cycle 6.

Comparison 4. Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dL) after cycle 6125Mean Difference (IV, Fixed, 95% CI)-1.70 [-7.93, 4.53]
2 Fasting plasma insulin (mIU/mL) after cycle 6125Mean Difference (IV, Fixed, 95% CI)3.5 [-1.88, 8.88]
Analysis 4.1.

Comparison 4 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 30 µg, Outcome 1 Fasting plasma glucose (mg/dL) after cycle 6.

Analysis 4.2.

Comparison 4 Desogestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 30 µg, Outcome 2 Fasting plasma insulin (mIU/mL) after cycle 6.

Comparison 5. Desogestrel 150 µg + EE 20 µg versus desogestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting blood glucose (mmol/L) at cycle 6149Mean Difference (IV, Fixed, 95% CI)0.0 [-0.22, 0.22]
2 Fasting blood insulin (µU/ml) at cycle 6149Mean Difference (IV, Fixed, 95% CI)0.30 [-2.41, 3.01]
3 Change in insulin AUC (µU/mL) by 6 months120Mean Difference (IV, Fixed, 95% CI)-0.69 [-2.38, 1.00]
Analysis 5.1.

Comparison 5 Desogestrel 150 µg + EE 20 µg versus desogestrel 150 µg + EE 30 µg, Outcome 1 Fasting blood glucose (mmol/L) at cycle 6.

Analysis 5.2.

Comparison 5 Desogestrel 150 µg + EE 20 µg versus desogestrel 150 µg + EE 30 µg, Outcome 2 Fasting blood insulin (µU/ml) at cycle 6.

Analysis 5.3.

Comparison 5 Desogestrel 150 µg + EE 20 µg versus desogestrel 150 µg + EE 30 µg, Outcome 3 Change in insulin AUC (µU/mL) by 6 months.

Comparison 6. Desogestrel 150 µg + EE 30 µg versus desogestrel 125 µg + EE 50 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Glucose AUC (mg/dl) at 6 months125Mean Difference (IV, Fixed, 95% CI)-14.90 [-38.08, 8.28]
2 Insulin AUC at 6 months125Mean Difference (IV, Fixed, 95% CI)-16.30 [-38.06, 5.46]
Analysis 6.1.

Comparison 6 Desogestrel 150 µg + EE 30 µg versus desogestrel 125 µg + EE 50 µg, Outcome 1 Glucose AUC (mg/dl) at 6 months.

Analysis 6.2.

Comparison 6 Desogestrel 150 µg + EE 30 µg versus desogestrel 125 µg + EE 50 µg, Outcome 2 Insulin AUC at 6 months.

Comparison 7. Desogestrel 150 µg + EE 20 µg versus chlormadinone acetate 2 mg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Glucose AUC (mmol/L) at 6 months124Mean Difference (IV, Fixed, 95% CI)0.10 [-1.49, 1.69]
2 Insulin AUC (pmol/L) at 6 months124Mean Difference (IV, Fixed, 95% CI)7.70 [-21.47, 36.87]
3 C-peptide AUC (pmol/L) at 6 months124Mean Difference (IV, Fixed, 95% CI)201.60 [-615.19, 1018.39]
Analysis 7.1.

Comparison 7 Desogestrel 150 µg + EE 20 µg versus chlormadinone acetate 2 mg + EE 30 µg, Outcome 1 Glucose AUC (mmol/L) at 6 months.

Analysis 7.2.

Comparison 7 Desogestrel 150 µg + EE 20 µg versus chlormadinone acetate 2 mg + EE 30 µg, Outcome 2 Insulin AUC (pmol/L) at 6 months.

Analysis 7.3.

Comparison 7 Desogestrel 150 µg + EE 20 µg versus chlormadinone acetate 2 mg + EE 30 µg, Outcome 3 C-peptide AUC (pmol/L) at 6 months.

Comparison 8. Desogestrel 150 µg + EE 30 µg versus levonorgestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting blood glucose (mmol/l) at cycle 6272Mean Difference (IV, Fixed, 95% CI)0.20 [0.00, 0.41]
2 Fasting blood glucose (mmol/l) at cycle 12262Mean Difference (IV, Fixed, 95% CI)0.15 [-0.08, 0.38]
3 Fasting blood glucose (mmol/l) at cycle 24117Mean Difference (IV, Fixed, 95% CI)-0.40 [-0.72, -0.08]
4 Glucose 2-h response (mmol/l) at cycle 6152Mean Difference (IV, Fixed, 95% CI)-0.01 [-0.97, 0.95]
5 Glucose 2-h response (mmol/l) at cycle 12144Mean Difference (IV, Fixed, 95% CI)1.08 [0.45, 1.71]
Analysis 8.1.

Comparison 8 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 Fasting blood glucose (mmol/l) at cycle 6.

Analysis 8.2.

Comparison 8 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 2 Fasting blood glucose (mmol/l) at cycle 12.

Analysis 8.3.

Comparison 8 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 3 Fasting blood glucose (mmol/l) at cycle 24.

Analysis 8.4.

Comparison 8 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 4 Glucose 2-h response (mmol/l) at cycle 6.

Analysis 8.5.

Comparison 8 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 5 Glucose 2-h response (mmol/l) at cycle 12.

Comparison 9. Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Glucose AUC (mg/dl) at 6 months127Mean Difference (IV, Fixed, 95% CI)-9.90 [-28.33, 8.53]
2 Insulin AUC at 6 months127Mean Difference (IV, Fixed, 95% CI)4.0 [-12.61, 20.61]
3 Fasting blood glucose (mmol/l) at cycle 6152Mean Difference (IV, Fixed, 95% CI)0.21 [-0.04, 0.46]
4 Fasting blood glucose (mmol/l) at cycle 12142Mean Difference (IV, Fixed, 95% CI)0.12 [-0.16, 0.40]
5 Glucose 2-h response (mmol/l) at cycle 6152Mean Difference (IV, Fixed, 95% CI)-0.20 [-1.11, 0.71]
6 Glucose 2-h response (mmol/l) at cycle 12142Mean Difference (IV, Fixed, 95% CI)0.37 [-0.41, 1.15]
Analysis 9.1.

Comparison 9 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 Glucose AUC (mg/dl) at 6 months.

Analysis 9.2.

Comparison 9 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 Insulin AUC at 6 months.

Analysis 9.3.

Comparison 9 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 3 Fasting blood glucose (mmol/l) at cycle 6.

Analysis 9.4.

Comparison 9 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 4 Fasting blood glucose (mmol/l) at cycle 12.

Analysis 9.5.

Comparison 9 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 5 Glucose 2-h response (mmol/l) at cycle 6.

Analysis 9.6.

Comparison 9 Desogestrel 150 µg + EE 30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 6 Glucose 2-h response (mmol/l) at cycle 12.

Comparison 10. Desogestrel 125 µg + EE 50 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Glucose AUC (mg/dl) at 6 months124Mean Difference (IV, Fixed, 95% CI)5.0 [-14.94, 24.94]
2 Insulin AUC at 6 months124Mean Difference (IV, Fixed, 95% CI)20.30 [4.24, 36.36]
Analysis 10.1.

Comparison 10 Desogestrel 125 µg + EE 50 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 Glucose AUC (mg/dl) at 6 months.

Analysis 10.2.

Comparison 10 Desogestrel 125 µg + EE 50 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 Insulin AUC at 6 months.

Comparison 11. Desogestrel 50-100-150 µg + EE 35-30-30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in glucose (mg/dL) from fasting to 1 h after 400 kcal drink at cycle 6157Mean Difference (IV, Fixed, 95% CI)1.90 [-8.07, 11.87]
2 Change in insulin (µU/ml) from fasting to 1 h after 400 kcal drink at cycle 6157Mean Difference (IV, Fixed, 95% CI)12.60 [-1.27, 26.47]
Analysis 11.1.

Comparison 11 Desogestrel 50-100-150 µg + EE 35-30-30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 Change in glucose (mg/dL) from fasting to 1 h after 400 kcal drink at cycle 6.

Analysis 11.2.

Comparison 11 Desogestrel 50-100-150 µg + EE 35-30-30 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 Change in insulin (µU/ml) from fasting to 1 h after 400 kcal drink at cycle 6.

Comparison 12. Drospirenone 3 mg + EE 20 µg versus desogestrel 150 µg + EE 20 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in fasting blood insulin (mU/L) at cycle 7151Mean Difference (IV, Fixed, 95% CI)1.1 [-0.77, 2.97]
2 Change in glucose AUC (h x mg/L) at cycle 7153Mean Difference (IV, Fixed, 95% CI)0.23 [-1.12, 1.58]
3 Change in fasting C-peptide (ng/mL) at cycle 7153Mean Difference (IV, Fixed, 95% CI)0.09 [-0.17, 0.35]
4 Change in C-peptide 2-h response (ng/mL) at cycle 7153Mean Difference (IV, Fixed, 95% CI)0.76 [-0.27, 1.79]
Analysis 12.1.

Comparison 12 Drospirenone 3 mg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 1 Change in fasting blood insulin (mU/L) at cycle 7.

Analysis 12.2.

Comparison 12 Drospirenone 3 mg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 2 Change in glucose AUC (h x mg/L) at cycle 7.

Analysis 12.3.

Comparison 12 Drospirenone 3 mg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 3 Change in fasting C-peptide (ng/mL) at cycle 7.

Analysis 12.4.

Comparison 12 Drospirenone 3 mg + EE 20 µg versus desogestrel 150 µg + EE 20 µg, Outcome 4 Change in C-peptide 2-h response (ng/mL) at cycle 7.

Comparison 13. Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in fasting plasma glucose (mg/dL) by cycle 13150Mean Difference (IV, Fixed, 95% CI)0.56 [-5.29, 6.41]
2 Change in fasting plasma insulin (µU/mL) by cycle 13150Mean Difference (IV, Fixed, 95% CI)-1.86 [-4.72, 1.00]
3 Change in fasting C-peptide (µmol/L) by cycle 13150Mean Difference (IV, Fixed, 95% CI)14.4 [-83.00, 111.80]
4 Change in fasting free fatty acids (µEq/L) by cycle 13150Mean Difference (IV, Fixed, 95% CI)-42.17 [-194.32, 109.98]
5 Change in glucose AUC (h x mg/dL) by cycle 13150Mean Difference (IV, Fixed, 95% CI)6.5 [-18.07, 31.07]
6 Change in insulin AUC (h x µU/mL) by cycle 13150Mean Difference (IV, Fixed, 95% CI)6.50 [-29.37, 42.37]
7 Change in C-peptide AUC (h x µmol/L) by cycle 13150Mean Difference (IV, Fixed, 95% CI)0.29 [-0.60, 1.18]
8 Change in free fatty acids AUC (h x µEq/L) by cycle 13150Mean Difference (IV, Fixed, 95% CI)-91.3 [-212.17, 29.57]
Analysis 13.1.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 1 Change in fasting plasma glucose (mg/dL) by cycle 13.

Analysis 13.2.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 2 Change in fasting plasma insulin (µU/mL) by cycle 13.

Analysis 13.3.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 3 Change in fasting C-peptide (µmol/L) by cycle 13.

Analysis 13.4.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 4 Change in fasting free fatty acids (µEq/L) by cycle 13.

Analysis 13.5.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 5 Change in glucose AUC (h x mg/dL) by cycle 13.

Analysis 13.6.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 6 Change in insulin AUC (h x µU/mL) by cycle 13.

Analysis 13.7.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 7 Change in C-peptide AUC (h x µmol/L) by cycle 13.

Analysis 13.8.

Comparison 13 Drospirenone 3 mg + EE 30 µg versus desogestrel 150 µg + EE 30 µg, Outcome 8 Change in free fatty acids AUC (h x µEq/L) by cycle 13.

Comparison 14. Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 100 µg + EE 20 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Impaired glucose tolerance at cycle 5142Odds Ratio (M-H, Fixed, 95% CI)0.16 [0.01, 3.64]
2 Insulin sensitivity at cycle 5142Mean Difference (IV, Fixed, 95% CI)1.9 [-5.92, 9.72]
Analysis 14.1.

Comparison 14 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 100 µg + EE 20 µg, Outcome 1 Impaired glucose tolerance at cycle 5.

Analysis 14.2.

Comparison 14 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 100 µg + EE 20 µg, Outcome 2 Insulin sensitivity at cycle 5.

Comparison 15. Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus desogestrel 150 µg + (EE 20 µg or EE 30 µg)
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in glucose AUC (mg/dL) by 6 months130Mean Difference (IV, Fixed, 95% CI)-2.37 [-6.72, 1.98]
2 Change in insulin AUC (µU/mL) by 6 months130Mean Difference (IV, Fixed, 95% CI)-2.58 [-6.63, 1.47]
3 Change in C-peptide AUC (ng/mL) by 6 months130Mean Difference (IV, Fixed, 95% CI)0.04 [-1.76, 1.84]
Analysis 15.1.

Comparison 15 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus desogestrel 150 µg + (EE 20 µg or EE 30 µg), Outcome 1 Change in glucose AUC (mg/dL) by 6 months.

Analysis 15.2.

Comparison 15 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus desogestrel 150 µg + (EE 20 µg or EE 30 µg), Outcome 2 Change in insulin AUC (µU/mL) by 6 months.

Analysis 15.3.

Comparison 15 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus desogestrel 150 µg + (EE 20 µg or EE 30 µg), Outcome 3 Change in C-peptide AUC (ng/mL) by 6 months.

Comparison 16. Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Glucose AUC (hr x mmol/L) at cycle 6169Mean Difference (IV, Fixed, 95% CI)-0.06 [-1.41, 1.29]
2 Insulin AUC (hr x pmol/L) at cycle 6165Mean Difference (IV, Fixed, 95% CI)-204.51 [-389.64, -19.38]
3 Fasting glycosylated hemoglobin (%) at cycle 6168Mean Difference (IV, Fixed, 95% CI)-0.16 [-0.33, 0.01]
Analysis 16.1.

Comparison 16 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 Glucose AUC (hr x mmol/L) at cycle 6.

Analysis 16.2.

Comparison 16 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 150 µg + EE 30 µg, Outcome 2 Insulin AUC (hr x pmol/L) at cycle 6.

Analysis 16.3.

Comparison 16 Etonogestrel 120 µg + EE 15 µg (vaginal ring) versus levonorgestrel 150 µg + EE 30 µg, Outcome 3 Fasting glycosylated hemoglobin (%) at cycle 6.

Comparison 17. Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant)
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mmol/L) at 12 months175Mean Difference (IV, Fixed, 95% CI)0.01 [-0.24, 0.26]
2 Fasting plasma glucose at 24 months168Mean Difference (IV, Fixed, 95% CI)-0.02 [-0.24, 0.20]
3 Fasting plasma insulin (pmol/L) at 12 months175Mean Difference (IV, Fixed, 95% CI)-6.45 [-37.68, 24.78]
4 Fasting plasma insulin (pmol/L) at 24 months168Mean Difference (IV, Fixed, 95% CI)12.09 [-4.17, 28.35]
5 Fasting glycosylated hemoglobin (%) at 12 months175Mean Difference (IV, Fixed, 95% CI)-0.03 [-0.27, 0.21]
6 Fasting glycosylated hemoglobin (%) at 24 months168Mean Difference (IV, Fixed, 95% CI)0.13 [-0.11, 0.37]
7 Glucose incremental AUC (min x mmol/L) at 12 months175Mean Difference (IV, Fixed, 95% CI)0.0 [-79.44, 79.44]
8 Glucose incremental AUC (min x mmol/L) at 24 months168Mean Difference (IV, Fixed, 95% CI)-59.00 [-143.87, 25.87]
9 Insulin incremental AUC (min x nmol/L) at 12 months175Mean Difference (IV, Fixed, 95% CI)3.20 [-11.69, 18.09]
10 Insulin incremental AUC (min x nmol/L) at 24 months168Mean Difference (IV, Fixed, 95% CI)-9.40 [-28.91, 10.11]
Analysis 17.1.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 1 Fasting plasma glucose (mmol/L) at 12 months.

Analysis 17.2.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 2 Fasting plasma glucose at 24 months.

Analysis 17.3.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 3 Fasting plasma insulin (pmol/L) at 12 months.

Analysis 17.4.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 4 Fasting plasma insulin (pmol/L) at 24 months.

Analysis 17.5.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 5 Fasting glycosylated hemoglobin (%) at 12 months.

Analysis 17.6.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 6 Fasting glycosylated hemoglobin (%) at 24 months.

Analysis 17.7.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 7 Glucose incremental AUC (min x mmol/L) at 12 months.

Analysis 17.8.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 8 Glucose incremental AUC (min x mmol/L) at 24 months.

Analysis 17.9.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 9 Insulin incremental AUC (min x nmol/L) at 12 months.

Analysis 17.10.

Comparison 17 Etonogestrel 68 mg (implant) versus levonorgestrel 216 mg (implant), Outcome 10 Insulin incremental AUC (min x nmol/L) at 24 months.

Comparison 18. Norethisterone 1000 µg + EE 35 µg versus levonorgestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting serum glucose (mg/dL) at 12 months158Mean Difference (IV, Fixed, 95% CI)-1.30 [-6.01, 3.41]
2 Change in fasting blood glucose (mg/dL) at 12 months1470Mean Difference (IV, Fixed, 95% CI)0.0 [-0.12, 0.12]
3 Glucose 2-h response (mg/dL) at 12 months158Mean Difference (IV, Fixed, 95% CI)6.5 [-4.75, 17.75]
4 Change in glucose 2-h response (mg/dL) at 12 months1470Mean Difference (IV, Fixed, 95% CI)-0.30 [-0.54, -0.06]
Analysis 18.1.

Comparison 18 Norethisterone 1000 µg + EE 35 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 Fasting serum glucose (mg/dL) at 12 months.

Analysis 18.2.

Comparison 18 Norethisterone 1000 µg + EE 35 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 2 Change in fasting blood glucose (mg/dL) at 12 months.

Analysis 18.3.

Comparison 18 Norethisterone 1000 µg + EE 35 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 3 Glucose 2-h response (mg/dL) at 12 months.

Analysis 18.4.

Comparison 18 Norethisterone 1000 µg + EE 35 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 4 Change in glucose 2-h response (mg/dL) at 12 months.

Comparison 19. Norethisterone 350 µg versus levonorgestrel 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mmol/L) at 6 months133Mean Difference (IV, Fixed, 95% CI)-0.62 [-1.26, 0.02]
2 Fasting glycosylated hemoglobin (%) at 6 months133Mean Difference (IV, Fixed, 95% CI)0.09 [-0.66, 0.84]
3 Abnormal fasting plasma glucose at 6 months133Peto Odds Ratio (Peto, Fixed, 95% CI)0.07 [0.00, 1.30]
4 Abnormal fasting glycosylated hemoglobin at 6 months133Peto Odds Ratio (Peto, Fixed, 95% CI)0.97 [0.14, 6.61]
Analysis 19.1.

Comparison 19 Norethisterone 350 µg versus levonorgestrel 30 µg, Outcome 1 Fasting plasma glucose (mmol/L) at 6 months.

Analysis 19.2.

Comparison 19 Norethisterone 350 µg versus levonorgestrel 30 µg, Outcome 2 Fasting glycosylated hemoglobin (%) at 6 months.

Analysis 19.3.

Comparison 19 Norethisterone 350 µg versus levonorgestrel 30 µg, Outcome 3 Abnormal fasting plasma glucose at 6 months.

Analysis 19.4.

Comparison 19 Norethisterone 350 µg versus levonorgestrel 30 µg, Outcome 4 Abnormal fasting glycosylated hemoglobin at 6 months.

Comparison 20. Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 100 µg + EE 20 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in glucose (mmol/L) from baseline to cycle 41235Mean Difference (IV, Fixed, 95% CI)-0.04 [-0.28, 0.20]
Analysis 20.1.

Comparison 20 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 100 µg + EE 20 µg, Outcome 1 Change in glucose (mmol/L) from baseline to cycle 4.

Comparison 21. Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dl) at cycle 62144Mean Difference (IV, Fixed, 95% CI)-0.88 [-2.83, 1.06]
2 Fasting plasma glucose (mg/dl) at cycle 12157Mean Difference (IV, Fixed, 95% CI)-1.0 [-5.38, 3.38]
3 Fasting plasma insulin (µg/dl) at cycle 6189Mean Difference (IV, Fixed, 95% CI)-0.79 [-4.31, 2.73]
4 Fasting plasma insulin (µU/ml) at cycle 6148Mean Difference (IV, Fixed, 95% CI)-2.0 [-4.77, 0.77]
5 Fasting plasma insulin (µU/ml) at cycle 12148Mean Difference (IV, Fixed, 95% CI)1.0 [-1.77, 3.77]
6 Glucose AUC (h x mg/dl) at cycle 62144Mean Difference (IV, Fixed, 95% CI)-10.83 [-29.62, 7.96]
7 Glucose AUC (h x mg/dl) at cycle 12157Mean Difference (IV, Fixed, 95% CI)-14.0 [-44.99, 16.99]
8 Insulin AUC (h x µg/dl) at cycle 6189Mean Difference (IV, Fixed, 95% CI)-12.0 [-58.88, 34.88]
9 Insulin AUC (h x µU/mL) at cycle 6148Mean Difference (IV, Fixed, 95% CI)3.00 [-33.46, 39.46]
10 Insulin AUC (h x µU/mL) at cycle 12148Mean Difference (IV, Fixed, 95% CI)21.0 [-15.46, 57.46]
Analysis 21.1.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 Fasting plasma glucose (mg/dl) at cycle 6.

Analysis 21.2.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 Fasting plasma glucose (mg/dl) at cycle 12.

Analysis 21.3.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 3 Fasting plasma insulin (µg/dl) at cycle 6.

Analysis 21.4.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 4 Fasting plasma insulin (µU/ml) at cycle 6.

Analysis 21.5.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 5 Fasting plasma insulin (µU/ml) at cycle 12.

Analysis 21.6.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 6 Glucose AUC (h x mg/dl) at cycle 6.

Analysis 21.7.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 7 Glucose AUC (h x mg/dl) at cycle 12.

Analysis 21.8.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 8 Insulin AUC (h x µg/dl) at cycle 6.

Analysis 21.9.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 9 Insulin AUC (h x µU/mL) at cycle 6.

Analysis 21.10.

Comparison 21 Norethindrone 500-750-1000 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 10 Insulin AUC (h x µU/mL) at cycle 12.

Comparison 22. Norethindrone 500-1000-500 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dl) at cycle 6188Mean Difference (IV, Fixed, 95% CI)-0.60 [-3.32, 2.12]
2 Glucose AUC (h x µg/dl) at cycle 6188Mean Difference (IV, Fixed, 95% CI)-2.0 [-27.51, 23.51]
3 Fasting plasma insulin (µg/dl) at cycle 6191Mean Difference (IV, Fixed, 95% CI)-0.34 [-3.66, 2.98]
4 Insulin AUC (h x µg/dl) at cycle 6191Mean Difference (IV, Fixed, 95% CI)-10.0 [-55.87, 35.87]
Analysis 22.1.

Comparison 22 Norethindrone 500-1000-500 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 Fasting plasma glucose (mg/dl) at cycle 6.

Analysis 22.2.

Comparison 22 Norethindrone 500-1000-500 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 Glucose AUC (h x µg/dl) at cycle 6.

Analysis 22.3.

Comparison 22 Norethindrone 500-1000-500 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 3 Fasting plasma insulin (µg/dl) at cycle 6.

Analysis 22.4.

Comparison 22 Norethindrone 500-1000-500 µg + EE 35 µg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 4 Insulin AUC (h x µg/dl) at cycle 6.

Comparison 23. Norethindrone 500-1000-500 µg + EE 35 µg versus norethindrone 500-750-1000 µg + EE 35 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dl) at cycle 6185Mean Difference (IV, Fixed, 95% CI)1.14 [-1.74, 4.02]
2 Fasting plasma insulin (µg/dl) at cycle 6190Mean Difference (IV, Fixed, 95% CI)0.45 [-2.63, 3.53]
3 Glucose AUC (h x µg/dl) at cycle 6185Mean Difference (IV, Fixed, 95% CI)8.0 [-17.31, 33.31]
4 Insulin AUC (h x µg/dl) at cycle 6190Mean Difference (IV, Fixed, 95% CI)2.0 [-35.64, 39.64]
Analysis 23.1.

Comparison 23 Norethindrone 500-1000-500 µg + EE 35 µg versus norethindrone 500-750-1000 µg + EE 35 µg, Outcome 1 Fasting plasma glucose (mg/dl) at cycle 6.

Analysis 23.2.

Comparison 23 Norethindrone 500-1000-500 µg + EE 35 µg versus norethindrone 500-750-1000 µg + EE 35 µg, Outcome 2 Fasting plasma insulin (µg/dl) at cycle 6.

Analysis 23.3.

Comparison 23 Norethindrone 500-1000-500 µg + EE 35 µg versus norethindrone 500-750-1000 µg + EE 35 µg, Outcome 3 Glucose AUC (h x µg/dl) at cycle 6.

Analysis 23.4.

Comparison 23 Norethindrone 500-1000-500 µg + EE 35 µg versus norethindrone 500-750-1000 µg + EE 35 µg, Outcome 4 Insulin AUC (h x µg/dl) at cycle 6.

Comparison 24. Depot medroxyprogesterone acetate 150 mg versus norethisterone enanthate 200 mg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting serum glucose (mg/dl) after 12 months140Mean Difference (IV, Fixed, 95% CI)10.05 [3.16, 16.94]
2 Glucose 2-hr response (mg/dl) after 12 months140Mean Difference (IV, Fixed, 95% CI)17.0 [5.67, 28.33]
3 Fasting serum insulin (nU/ml) after 12 months140Mean Difference (IV, Fixed, 95% CI)3.40 [2.07, 4.73]
4 Insulin 2-hr response (nU/ml) after 12 months140Mean Difference (IV, Fixed, 95% CI)20.0 [-47.41, 87.41]
Analysis 24.1.

Comparison 24 Depot medroxyprogesterone acetate 150 mg versus norethisterone enanthate 200 mg, Outcome 1 Fasting serum glucose (mg/dl) after 12 months.

Analysis 24.2.

Comparison 24 Depot medroxyprogesterone acetate 150 mg versus norethisterone enanthate 200 mg, Outcome 2 Glucose 2-hr response (mg/dl) after 12 months.

Analysis 24.3.

Comparison 24 Depot medroxyprogesterone acetate 150 mg versus norethisterone enanthate 200 mg, Outcome 3 Fasting serum insulin (nU/ml) after 12 months.

Analysis 24.4.

Comparison 24 Depot medroxyprogesterone acetate 150 mg versus norethisterone enanthate 200 mg, Outcome 4 Insulin 2-hr response (nU/ml) after 12 months.

Comparison 25. Norethisterone enanthate 50 mg + EV 5 mg (injected) versus norethisterone 500-750-1000 µg + EE 35 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dl) at cycle 6148Mean Difference (IV, Fixed, 95% CI)0.60 [-2.33, 3.53]
2 Fasting plasma insulin (µU/ml) at cycle 6148Mean Difference (IV, Fixed, 95% CI)0.28 [-0.84, 1.40]
3 Plasma glucose rate of disappearance (mg/kg/min) at cycle 6148Mean Difference (IV, Fixed, 95% CI)0.11 [-0.19, 0.41]
Analysis 25.1.

Comparison 25 Norethisterone enanthate 50 mg + EV 5 mg (injected) versus norethisterone 500-750-1000 µg + EE 35 µg, Outcome 1 Fasting plasma glucose (mg/dl) at cycle 6.

Analysis 25.2.

Comparison 25 Norethisterone enanthate 50 mg + EV 5 mg (injected) versus norethisterone 500-750-1000 µg + EE 35 µg, Outcome 2 Fasting plasma insulin (µU/ml) at cycle 6.

Analysis 25.3.

Comparison 25 Norethisterone enanthate 50 mg + EV 5 mg (injected) versus norethisterone 500-750-1000 µg + EE 35 µg, Outcome 3 Plasma glucose rate of disappearance (mg/kg/min) at cycle 6.

Comparison 26. Norethisterone enanthate 50 mg + EV 5 mg versus medroxyprogesterone acetate 25 mg + estradiol cypionate 5 mg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting serum glucose (mmol/L) at cycle 91298Mean Difference (IV, Fixed, 95% CI)0.07 [-0.06, 0.20]
2 Serum glucose 2-h response (mmol/L) at cycle 91298Mean Difference (IV, Fixed, 95% CI)0.14 [-0.06, 0.34]
3 Abnormal fasting serum glucose at cycle 91298Odds Ratio (M-H, Fixed, 95% CI)0.93 [0.06, 15.08]
Analysis 26.1.

Comparison 26 Norethisterone enanthate 50 mg + EV 5 mg versus medroxyprogesterone acetate 25 mg + estradiol cypionate 5 mg, Outcome 1 Fasting serum glucose (mmol/L) at cycle 9.

Analysis 26.2.

Comparison 26 Norethisterone enanthate 50 mg + EV 5 mg versus medroxyprogesterone acetate 25 mg + estradiol cypionate 5 mg, Outcome 2 Serum glucose 2-h response (mmol/L) at cycle 9.

Analysis 26.3.

Comparison 26 Norethisterone enanthate 50 mg + EV 5 mg versus medroxyprogesterone acetate 25 mg + estradiol cypionate 5 mg, Outcome 3 Abnormal fasting serum glucose at cycle 9.

Comparison 27. Gestodene 50-70-100 µg + EE 30-40-30 µg levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mmol/L) at cycle 6138Mean Difference (IV, Fixed, 95% CI)0.10 [-0.18, 0.38]
2 Fasting glycosylated hemoglobin (%) at cycle 6138Mean Difference (IV, Fixed, 95% CI)-0.30 [-0.85, 0.25]
Analysis 27.1.

Comparison 27 Gestodene 50-70-100 µg + EE 30-40-30 µg levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 Fasting plasma glucose (mmol/L) at cycle 6.

Analysis 27.2.

Comparison 27 Gestodene 50-70-100 µg + EE 30-40-30 µg levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 Fasting glycosylated hemoglobin (%) at cycle 6.

Comparison 28. Levonorgestrel 50-125 µg + EE 50 µg versus levonorgestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Blood glucose ('random') (mg/dl) at 12 months154Mean Difference (IV, Fixed, 95% CI)-0.90 [-8.61, 6.81]
Analysis 28.1.

Comparison 28 Levonorgestrel 50-125 µg + EE 50 µg versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 Blood glucose ('random') (mg/dl) at 12 months.

Comparison 29. Nomegestrel acetate 2 mg + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 AUC glucose (h x mmol/L) at cycle 61102Mean Difference (IV, Fixed, 95% CI)-0.60 [-1.80, 0.60]
2 Incremental AUC glucose (h x mmol/L) at cycle 61102Mean Difference (IV, Fixed, 95% CI)-1.43 [-2.55, -0.31]
3 AUC insulin (h x pmol/L) at cycle 6193Mean Difference (IV, Fixed, 95% CI)-63.0 [-173.86, 47.86]
4 Incremental AUC insulin (h x pmol/L) at cycle 6193Mean Difference (IV, Fixed, 95% CI)-69.0 [-165.75, 27.75]
5 Serum HbA1c (%) at cycle 61104Mean Difference (IV, Fixed, 95% CI)-0.10 [-0.18, -0.02]
Analysis 29.1.

Comparison 29 Nomegestrel acetate 2 mg + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 1 AUC glucose (h x mmol/L) at cycle 6.

Analysis 29.2.

Comparison 29 Nomegestrel acetate 2 mg + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 2 Incremental AUC glucose (h x mmol/L) at cycle 6.

Analysis 29.3.

Comparison 29 Nomegestrel acetate 2 mg + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 3 AUC insulin (h x pmol/L) at cycle 6.

Analysis 29.4.

Comparison 29 Nomegestrel acetate 2 mg + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 4 Incremental AUC insulin (h x pmol/L) at cycle 6.

Analysis 29.5.

Comparison 29 Nomegestrel acetate 2 mg + E2 1.5 mg versus levonorgestrel 150 µg + EE 30 µg, Outcome 5 Serum HbA1c (%) at cycle 6.

Comparison 30. Dienogest 0-2-3-0 mg + E2V 3-2-2-1 mg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 AUC insulin (mU x h/L) at cycle 7158Mean Difference (IV, Fixed, 95% CI)-14.40 [-30.26, 1.46]
2 AUC glucose (mU x h/L) at cycle 7158Mean Difference (IV, Fixed, 95% CI)-0.60 [-1.53, 0.33]
Analysis 30.1.

Comparison 30 Dienogest 0-2-3-0 mg + E2V 3-2-2-1 mg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 1 AUC insulin (mU x h/L) at cycle 7.

Analysis 30.2.

Comparison 30 Dienogest 0-2-3-0 mg + E2V 3-2-2-1 mg versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg, Outcome 2 AUC glucose (mU x h/L) at cycle 7.

Comparison 31. Extended versus conventional regimen of dienogest 2 mg + EE 30 μg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in fasting HbA1c (%) after 12 months156Mean Difference (IV, Fixed, 95% CI)-0.1 [-0.23, 0.03]
2 Change in fasting glucose (mmol/l) after 12 months156Mean Difference (IV, Fixed, 95% CI)0.09 [-0.12, 0.30]
3 Change in fasting insulin (mU/l) after 12 months156Mean Difference (IV, Fixed, 95% CI)-0.8 [-2.63, 1.03]
4 Change in fasting C-peptide ng/ml) after 12 months156Mean Difference (IV, Fixed, 95% CI)-0.10 [-0.42, 0.22]
5 Change in AUC glucose (mmol/l x min) after 12 months156Mean Difference (IV, Fixed, 95% CI)82.0 [10.72, 153.28]
6 Change in AUC insulin (mU/l x min) after 12 months156Mean Difference (IV, Fixed, 95% CI)310.0 [-754.51, 1374.51]
7 Change in AUC C-peptide (ng/ml x min) after 12 months156Mean Difference (IV, Fixed, 95% CI)10.00 [-81.50, 101.50]
8 Change in calculated insulin resistance (HOMA-IR) after 12 months156Mean Difference (IV, Fixed, 95% CI)0.06 [-0.36, 0.48]
9 Change in calculated insulin sensitivity index after 12 months156Mean Difference (IV, Fixed, 95% CI)-1.48 [-3.13, 0.17]
Analysis 31.1.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 1 Change in fasting HbA1c (%) after 12 months.

Analysis 31.2.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 2 Change in fasting glucose (mmol/l) after 12 months.

Analysis 31.3.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 3 Change in fasting insulin (mU/l) after 12 months.

Analysis 31.4.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 4 Change in fasting C-peptide ng/ml) after 12 months.

Analysis 31.5.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 5 Change in AUC glucose (mmol/l x min) after 12 months.

Analysis 31.6.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 6 Change in AUC insulin (mU/l x min) after 12 months.

Analysis 31.7.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 7 Change in AUC C-peptide (ng/ml x min) after 12 months.

Analysis 31.8.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 8 Change in calculated insulin resistance (HOMA-IR) after 12 months.

Analysis 31.9.

Comparison 31 Extended versus conventional regimen of dienogest 2 mg + EE 30 μg, Outcome 9 Change in calculated insulin sensitivity index after 12 months.

Comparison 32. Continuous versus cyclic use of drospirenone 3 mg + EE 30 μg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Fasting plasma glucose (mg/dL) after 6 cycles177Mean Difference (IV, Fixed, 95% CI)-1.20 [-3.95, 1.55]
2 Plasma insulin (mU/mL) after 6 cycles177Mean Difference (IV, Fixed, 95% CI)0.5 [-1.46, 2.46]
Analysis 32.1.

Comparison 32 Continuous versus cyclic use of drospirenone 3 mg + EE 30 μg, Outcome 1 Fasting plasma glucose (mg/dL) after 6 cycles.

Analysis 32.2.

Comparison 32 Continuous versus cyclic use of drospirenone 3 mg + EE 30 μg, Outcome 2 Plasma insulin (mU/mL) after 6 cycles.

Comparison 33. Obese (BMI 30 to 39.9) versus normal (BMI 19 to 24.9): levonorgestrel + EE
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in fasting serum glucose (mg/dL) by cycle 31109Mean Difference (IV, Fixed, 95% CI)0.0 [-3.30, 3.30]
2 Change in fasting serum insulin (μU/mL) by cycle 31109Mean Difference (IV, Fixed, 95% CI)3.1 [-0.97, 7.17]
3 Change in log calculated insulin resistance by cycle 31109Mean Difference (IV, Fixed, 95% CI)0.09 [-0.13, 0.31]
Analysis 33.1.

Comparison 33 Obese (BMI 30 to 39.9) versus normal (BMI 19 to 24.9): levonorgestrel + EE, Outcome 1 Change in fasting serum glucose (mg/dL) by cycle 3.

Analysis 33.2.

Comparison 33 Obese (BMI 30 to 39.9) versus normal (BMI 19 to 24.9): levonorgestrel + EE, Outcome 2 Change in fasting serum insulin (μU/mL) by cycle 3.

Analysis 33.3.

Comparison 33 Obese (BMI 30 to 39.9) versus normal (BMI 19 to 24.9): levonorgestrel + EE, Outcome 3 Change in log calculated insulin resistance by cycle 3.

Comparison 34. Levonorgestrel 100 μg + EE 20 μg versus levonorgestrel 150 μg + EE 30 μg
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Change in fasting serum glucose (mg/dL) by cycle 31109Mean Difference (IV, Fixed, 95% CI)-3.00 [-5.89, -0.11]
2 Change in fasting serum insulin (μU/mL) by cycle 31109Mean Difference (IV, Fixed, 95% CI)-1.7 [-4.96, 1.56]
3 Change in log calculated insulin resistance by cycle 31109Mean Difference (IV, Fixed, 95% CI)-0.1 [-0.31, 0.11]
Analysis 34.1.

Comparison 34 Levonorgestrel 100 μg + EE 20 μg versus levonorgestrel 150 μg + EE 30 μg, Outcome 1 Change in fasting serum glucose (mg/dL) by cycle 3.

Analysis 34.2.

Comparison 34 Levonorgestrel 100 μg + EE 20 μg versus levonorgestrel 150 μg + EE 30 μg, Outcome 2 Change in fasting serum insulin (μU/mL) by cycle 3.

Analysis 34.3.

Comparison 34 Levonorgestrel 100 μg + EE 20 μg versus levonorgestrel 150 μg + EE 30 μg, Outcome 3 Change in log calculated insulin resistance by cycle 3.

Appendices

Appendix 1. Search 2014

PubMed (01 Sep 2011 to 14 Apr 2014)

(contraceptive agents, female OR (steroid* AND contracept*) OR orthoevra OR "ortho evra" OR "norelgestromin" OR (contraceptive devices, female and ring) OR NuvaRing OR cyclofem OR lunelle OR mesigyna OR cycloprovera OR (medroxyprogesterone 17-acetate AND (contracept* OR inject* OR depo OR depot)) OR depot medroxyprogesterone OR depo medroxyprogesterone OR depotmedroxyprogesterone OR depomedroxyprogesterone OR dmpa OR "net en" OR norethisterone-enantate OR norplant OR uniplant OR jadelle OR implanon OR ((levonorgestrel OR etonogestrel) AND implant) OR (levonorgestrel AND intrauterine devices) OR mirena OR ((progestational hormones OR progestin) AND contracept* AND (oral OR pill* OR tablet*))) AND ((metabolism AND carbohydrate) OR blood glucose OR insulin/blood)
AND (Clinical Trial[ptyp] AND ( "2011/09/01"[PDat] : "3000/12/31"[PDat] ) )

CENTRAL (2012 to 10 Apr 2014)

Title, Abstract, or Key Words: contracept* AND (carbohydrate OR glucose OR insulin)
NOT Record Title: (polycystic OR PCOS OR exercise OR performance OR postmenopaus* OR perimenopaus*)
Search Limits: Publication Date from 2012 to 2014 in Trials

POPLINE (2012 to 10 Apr 2014)

All fields: (carbohydrate AND metabol*) OR glucose OR insulin
Keyword: Contraceptive Agents Female
OR Keyword: Contraceptive Methods

LILACS (14 Apr 2014)

contraceptive agents, female or agentes anticonceptivos femeninos or anticoncepcionais femeninos or ((steroid or steroidal or steroids or esteroides) and (contraceptive or contraceptives or contraception or anticonceptivos or anticoncepcionais)) [Words] and ((metabolism or metabolize or metabolismo) and (carbohydrate or carbohydrates or carbohidratos or carboidratos)) or (blood glucose or glucose blood level or glucosa de la sangre or glicemia) or (insulin or blood insulin level or insulin/blood or insulina) [Words]

ClinicalTrials.gov (01 Jun 2011 to 04 Apr 2014)

Search terms: (contraceptive OR contraception) AND (glucose OR insulin OR carbohydrate)
Study type: Interventional studies
Conditions: NOT diabetes
Gender: Studies with female participants

ICTRP (2011 to 08 Apr 2014)

We ran 3 basic searches with no other limitations.
Search terms:
contracept* AND carbohydrate
contracept* AND insulin
contracept* AND glucose

Appendix 2. Previous searches

2012

PubMed (01 Mar 2009 to 13 Feb 2012)

(contraceptive agents, female OR (steroid* AND contracept*) OR orthoevra OR "ortho evra" OR "norelgestromin" OR (contraceptive devices, female and ring) OR NuvaRing OR cyclofem OR lunelle OR mesigyna OR cycloprovera OR (medroxyprogesterone 17-acetate AND (contracept* OR inject* OR depo OR depot)) OR depot medroxyprogesterone OR depo medroxyprogesterone OR depotmedroxyprogesterone OR depomedroxyprogesterone OR dmpa OR "net en" OR norethisterone-enantate OR norplant OR uniplant OR jadelle OR implanon OR ((levonorgestrel OR etonogestrel) AND implant) OR (levonorgestrel AND intrauterine devices) OR mirena OR ((progestational hormones OR progestin) AND contracept* AND (oral OR pill* OR tablet*))) AND ((metabolism AND carbohydrate) OR blood glucose OR insulin/blood)
Limits Activated: Clinical Trial, Randomized Controlled Trial

CENTRAL (2009 to 13 Feb 2012)

contracept* AND (carbohydrate OR glucose OR insulin) in Title, Abstract, or Key Words NOT (polycystic OR PCOS OR exercise OR performance OR postmenopaus* OR perimenopaus*) in Title

POPLINE (5 years through 13 Feb 2012)

(Contraceptive Agents Female/depo provera/dmpa/medroxyprogesterone/(steroid* & contracept*) /orthoevra/ortho evra /norelgestromin/(contraceptive devices, female and ring)/ NuvaRing /cyclofem /lunelle/ mesigyna/ cycloprovera/ (medroxyprogesterone 17-acetate & (contracept* /inject*/depo/depot))/ depot medroxyprogesterone/ depo medroxyprogesterone/ depot medroxyprogesterone/depo medroxyprogesterone/dmpa/ net en/ norethisterone-enantate/norplant/uniplant/jadelle/implanon/((levonorgestrel/ etonogestrel) & implant)/(levonorgestrel & intrauterine devices)/mirena /((progestational hormones/progestin) & contracept* & (oral/pill*/tablet*))) & ((metaboli* & carbohydrate?)/glucose metabolism effects/insulin)

LILACS (13 Feb 2012)

contraceptive agents, female or agentes anticonceptivos femeninos or anticoncepcionais femeninos or ((steroid or steroidal or steroids or esteroides) and (contraceptive or contraceptives or contraception or anticonceptivos or anticoncepcionais)) [Words] and ((metabolism or metabolize or metabolismo) and (carbohydrate or carbohydrates or carbohidratos or carboidratos)) or (blood glucose or glucose blood level or glucosa de la sangre or glicemia) or (insulin or blood insulin level or insulin/blood or insulina) [Words]

ClinicalTrials.gov (01 Jan 2009 to 16 Nov 2011)

Search terms: (contraceptive OR contraception) AND (glucose OR insulin OR carbohydrate)
Gender: Studies with female participants

ICTRP (16 Nov 2011)

Search terms:
contracept* AND carbohydrate
contracept* AND insulin
contracept* AND glucose

2009

MEDLINE using PubMed (20 Mar 2006 to 01 Apr 2009)

(contraceptive agents, female OR (steroid* AND contracept*) OR orthoevra OR "ortho evra" OR "norelgestromin" OR (contraceptive devices, female and ring) OR NuvaRing OR cyclofem OR lunelle OR mesigyna OR cycloprovera OR (medroxyprogesterone 17-acetate AND (contracept* OR inject* OR depo OR depot)) OR depot medroxyprogesterone OR depo medroxyprogesterone OR depotmedroxyprogesterone OR depomedroxyprogesterone OR dmpa OR "net en" OR norethisterone-enantate OR norplant OR uniplant OR jadelle OR implanon OR ((levonorgestrel OR etonogestrel) AND implant) OR (levonorgestrel AND intrauterine devices) OR mirena OR ((progestational hormones OR progestin) AND contracept* AND (oral OR pill* OR tablet*))) AND ((metabolism AND carbohydrate) OR blood glucose OR insulin/blood)

POPLINE (2006 to 26 Mar 2009)

(Contraceptive Agents Female/depo provera/dmpa/medroxyprogesterone/(steroid* & contracept*) /orthoevra/ortho evra /norelgestromin/(contraceptive devices, female and ring)/ NuvaRing /cyclofem /lunelle/ mesigyna/ cycloprovera/ (medroxyprogesterone 17-acetate & (contracept* /inject*/depo/depot))/ depot medroxyprogesterone/ depo medroxyprogesterone/ depot medroxyprogesterone/depo medroxyprogesterone/dmpa/ net en/ norethisterone-enantate/norplant/uniplant/jadelle/implanon/((levonorgestrel/ etonogestrel) & implant)/(levonorgestrel & intrauterine devices)/mirena /((progestational hormones/progestin) & contracept* & (oral/pill*/tablet*))) & ((metaboli* & carbohydrate?)/glucose metabolism effects/insulin)

LILACS (26 Mar 2009)

contraceptive agents, female or agentes anticonceptivos femeninos or anticoncepcionais femeninos or ((steroid or steroidal or steroids or esteroides) and (contraceptive or contraceptives or contraception or anticonceptivos or anticoncepcionais)) [Words] and ((metabolism or metabolize or metabolismo) and (carbohydrate or carbohydrates or carbohidratos or carboidratos)) or (blood glucose or glucose blood level or glucosa de la sangre or glicemia) or (insulin or blood insulin level or insulin/blood or insulina) [Words]

EMBASE (2005 to 31 Mar 2009)

contraceptive agent? or steroid?(w)contracept?
and
(metabolism(w)carbohydrate? or glucose blood level or insulin)

CENTRAL (2005 to 07 Apr 2009)

contracept* AND (carbohydrate OR glucose OR insulin) in Title, Abstract, or Key Words NOT (polycystic OR exercise OR performance OR postmenopaus* OR perimenopaus*) in Title, Abstract, or Key Words

ClinicalTrials.gov (20 Apr 2009)

Interventions: contraceptive OR contraception
Outcome measures: glucose OR insulin OR carbohydrate

ICTRP (20 Apr 2009)

Title or Condition: contraceptive OR contraception OR COC
Title or Condition: carbohydrate OR insulin OR glucose

2006

The original search (2006) used strategies similar to those for 2009 but did not include ClinicalTrials.gov and ICTRP.

Appendix 3. Reports with insufficient data for analysis (2012)

Eighteen studies had sufficient data for analysis. Previously, we had included 16 with the results as reported by the investigators. In 2012, we excluded them along with two new studies, because we could not adequately compare the study groups. The descriptive results and data limitations are given below. For some trials, medians were reported rather than means, which could be appropriate for skewed data. Several had carbohydrate metabolism as a secondary outcome and the results were mentioned in the text. We wrote to investigators for additional information, but many studies were too old to obtain further data.

Desogestrel COCs

Kuhl 1985 studied desogestrel 150 µg + EE 30 µg versus a triphasic levonorgestrel preparation. The researchers reported significant increases in fasting glucose within groups at cycle three but normal levels during the OGTT (Kuhl 1985). Between-group comparisons were not available.

After comparing desogestrel 150 µg + EE 20 µg versus gestodene 75 µg + EE 30 µg, Petersen 1991 reported no changes in the glucose and insulin measures at 6 or 12 months. Rakoczi 1985 studied a desogestrel preparation similar to those in Luyckx 1986, and the triphasic levonorgestrel was the same. The two groups were reportedly not significantly different after three months.

Additional trials studied OCs containing desogestrel versus other preparations:

  • Winkler 2009 studied a COC containing desogestrel 150 µg + EE 30 µg versus chlormadinone acetate 2 mg + EE 30 µg. The groups were reportedly not significantly different for the outcomes measured.

  • Jandrain 1990 reported within-group changes for the same desogestrel COC as Winkler 2009 and for cyproterone acetate 2 mg + EE 35 µg. At 12 months, the desogestrel group had decreases in plasma insulin (fasting and AUC) but increases in C-peptide (fasting and AUC) and in glucose AUC. The cyproterone acetate group reportedly had increases in fasting values and AUC for glucose and C-peptide.

  • Miccoli 1989 studied desogestrel 150 µg + EE 20 µg versus the same cyproterone acetate COC as Jandrain 1990 versus a triphasic gestodene preparation. The researchers reported a few changes for the cyproterone acetate group after six months: higher means for glucose after the OGTT and for plasma insulin (fasting and during the OGTT).

  • In a trial of two progestin-only pills (desogestrel 75 µg versus levonorgestrel 30 µg), Kivela 2001 reported no significant changes in the glucose or insulin measures. The small changes in glycosylated hemoglobin were reportedly not significantly different for the two groups.

Levonorgestrel

Several trials examined levonorgestrel COCs versus other contraceptives. Both Endrikat 2002 and Skouby 2005 studied levonorgestrel 100 µg + EE 20 µg versus levonorgestrel 150 µg + EE 30 µg for 13 cycles, and provided data as medians and ranges. Endrikat 2002 reported a significant decrease in insulin AUC for the EE 20 µg group versus an increase for the EE 30 µg group. Skouby 2005 reported no significant treatment differences in the carbohydrate measures. Pakarinen 1999 examined a levonorgestrel 30 µg OC versus a levonorgestrel intrauterine system versus a copper-releasing intrauterine device. The researchers reported that fasting glucose and insulin did not change significantly in any group in three months.

Two trials compared gestodene and levonorgestrel preparations. Lepot 1987 compared the same triphasic preparations as Ball 1990 (levonorgestrel 50-75-125 µg + EE 30-40-30 µg versus gestodene 50-70-100 µg + EE 30-40-30 µg). Reportedly, glucose AUC increased for gestodene, and insulin AUC increased for levonorgestrel by six cycles. Rabe 1987 examined levonorgestrel 150 µg + EE 30 µg versus gestodene 75 µg + EE 30 µg. The researchers reported that the glucose and insulin measures did not change significantly within groups by six cycles.

Oelkers 1995 compared three drospirenone-containing COCs versus a levonorgestrel COC: 1) drospirenone 3 mg + EE 30 µg; 2) drospirenone 3 mg + EE 20 µg; 3) drospirenone 3 mg + EE 15 µg; 4) levonorgestrel 150 µg + EE 30 µg. The researchers reported that glucose AUC increased significantly in all groups after six cycles, but the drospirenone-COC groups reportedly were not significantly different from the levonorgestrel group.

Rad 2011 compared a continuous regimen of levonorgestrel 90 μg + EE 20 μg versus a cyclic regimen (21 days on, 7 days off) of levonorgestrel 100 μg + EE 20 μg. Fasting glucose, insulin, and HbA1c were assessed after 7 and 13 pill packs, and change from baseline was computed. The study groups were reportedly not significantly different for the carbohydrate metabolism measures after cycles 7 and 13. For HbA1c, the continuous group had a smaller mean change than the cyclic group after cycle 7 but the groups were not significantly different after cycle 13.

Other comparisons

Van der Vange 1987 compared seven COC preparations for six cycles. The formulations were levonorgestrel 150 µg + EE 30 µg, norethisterone 1 mg + EE 35 µg, desogestrel 150 µg + EE 30 µg, levonorgestrel 50-75-125 µg + EE 30-40-30 µg, cyproterone acetate 2 mg + EE 35 µg, gestodene 50-70-100 µg + EE 30-40-30 µg, and desogestrel 25-125 µg + EE 40-30 µg. Significant changes were reported as follows: lower fasting glucose for the monophasic desogestrel and higher glucose during OGTT for the triphasic levonorgestrel and for the norethisterone preparations (at one point in time each). Differences included lower glycosylated hemoglobin for the monophasic levonorgestrel compared to the other preparations and for the cyproterone acetate preparation versus the monophasic norethisterone.

Mostafavi 1999 examined a COC containing norgestrel 150 µg + EE 30 µg versus medroxyprogesterone acetate (150 mg intramuscularly) versus a levonorgestrel implant (6 rods at 36 mg each). Reportedly, fasting blood glucose increases were not significantly different in the three groups.

Petersen 1999 compared norgestimate 180-215-250 µg + EE 35 µg versus gestodene 50-70-100 µg + EE 30-40-30 µg. Reportedly, the groups were not significantly different in the fasting insulin increases by six months. In the gestodene group, C-peptide and insulin AUC reportedly increased by six months but the groups were not significantly different.

Winkler 2010 compared the effects of chlormadinone acetate (CMA) 2 mg + EE 20 µg for 24 days each cycle versus desogestrel 150 µg + EE 20 µg or levonorgestrel 150 µg + EE 30 µg for 21 days each cycle. The researchers reported the median percentage changes in the carbohydrate measures, but did not provide means and standard deviations nor any P values. Reported results included the following: AUC glucose increased about half as much in the CMA group than in the other two groups; AUC insulin decreased slightly in both the CMA and levonorgestrel group and hardly changed for the desogestrel group; and AUC C-peptide decreased slightly in the CMA group while the levonorgestrel and desogestrel groups had increases.

What's new

DateEventDescription
16 April 2014New citation required but conclusions have not changedNo new trials met the inclusion criteria.
15 April 2014New search has been performed

Searches updated.

In 2012 version, Background had 2 paragraphs inadvertently inserted from the Plain Language Summary. We corrected that and apologize for the error.

History

Protocol first published: Issue 3, 2006
Review first published: Issue 1, 2007

DateEventDescription
13 February 2012New search has been performedSearches updated.
13 February 2012New citation required but conclusions have not changedFive new trials added (Agren 2011; Beasley 2012; Junge 2011; Machado 2010; Wiegratz 2010).
5 January 2012Amended

For the 2012 update, trials without sufficient data for analysis were classed as 'Excluded'. The descriptive results from those reports have been included in Appendix 3, along with new trials that did not have sufficient data either.

A previously included crossover trial (N=12) was excluded (Song 1992). Outcome data were not available for each treatment segment prior to crossover.

7 May 2009New citation required but conclusions have not changedAdded 4 new trials (Cagnacci 2009a; Cagnacci 2009b; Elkind-Hirsch 2007; Winkler 2009).
24 April 2009New search has been performedSearches were updated; added searches for ongoing trials.
15 April 2008AmendedConverted to new review format.
15 January 2007New citation required and conclusions have changedSubstantive amendment

Contributions of authors

LM Lopez developed the idea and some database searches, reviewed the search results, conducted the primary data abstraction, and drafted the review and updates through 2014. DA Grimes did the secondary data abstraction through 2009, helped interpret the results, and helped update the literature in 2012. KF Schulz provided statistical oversight. All authors reviewed and edited the manuscript.

Declarations of interest

Dr. Grimes has consulted with the pharmaceutical companies Bayer Healthcare Pharmaceuticals and Merck & Co, Inc.

Sources of support

Internal sources

  • No sources of support supplied

External sources

  • National Institute of Child Health and Human Development, USA.

    Support for conducting the review and updates at FHI 360

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Agren 2011

Methods

Randomized open-label trial at 5 centers in Finland; randomization 1:1 using blocks with randomly permuted block sizes. Women stratified by age group (18 to 35 years and 36 to 50 years)

Sample size based on detecting differences in several metabolic measures, especially hemostatic indices. With effect size (i.e., differences between the treatment groups divided by the standard deviation) of 0.6, sample estimated at 42 evaluable participants per group for 80% power and 5% significance. Assuming 20% early discontinuation, 60 women were needed per group.

Participants

121 women, 18 to 50 years.

Inclusion Criteria:

  • Sexually active

  • Need contraception and willing to use an OC for 6 cycles

  • Body mass index 17 to 29 kg/m2

  • Good physical and mental health.

Exclusion criteria:

  • Use of hormonal treatment in past 2 months, including sex hormones (other than contraceptives), insulin, thyroid and corticosteroid hormones (except for local dermatological use)

  • Contraindications for contraceptive steroids

  • History (in past year) of alcohol or drug abuse

  • Abnormal cervical smear at screening or documentation of abnormal smear in past 6 months

  • Clinically relevant abnormal laboratory result at screening

  • Use of injectable contraception in past 6 months for injection with 3-month duration, in past 4 months for injection with a 2-month duration, in past 2 months for injection with 1-month duration

  • Before spontaneous menstruation following a delivery or abortion

  • Breastfeeding in past 2 months

  • Use in past 2 months: phenytoin, barbiturates, primidone, carbamazepine, oxcarbazepine, topiramate, felbamate, rifampicin, nelfinavir, ritonavir, griseofulvin, ketoconazole, lipid-lowering drugs,anticoagulants and herbal remedies containing Hypericum perforatum (St John's Wort);

  • Use of pharmacological agents which affect the hemostatic system during the pretreatment blood sampling

  • Use of investigational drugs or participation in another clinical trial in past 2 months.

Interventions

1) Nomegestrol acetate (NOMAC) 2.5 mg + 17β-estradiol (E2) 1.5 mg (N=60)

2) Levonorgestrel 150 µg + EE 30 µg (N=61)

6 treatment cycles

Outcomes

Primary (selected): glucose and insulin AUC over 3 hours (AUC3) (oral glucose tolerance test); incremental AUC3 for glucose and insulin (AUC3 - 3* fasting concentration); serum HbA1c

Publication provides medians and interquartile ranges. ClinicalTrials.gov (NCT00511355) had the means and standard deviations posted.

Notes

Analysis excluded from LNG group 3 women randomized who were not treated.

Losses (with exclusions and discontinuations): overall 16/121 = 13%; NOMAC 7/60 = 12%, LNG 9/61 = 15%

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskInteractive voice response system
Pharmaceutical company involvementHigh riskFunded by Schering Plough

Ball 1990

Methods"Randomly allocated" study conducted in Oxford, UK.
Participants43 healthy women, 16 to 30 years, attending family planning clinic and requesting oral contraceptives (OC).
Exclusion criteria: hormonal contraceptive use in past 6 months, contraindication to OCs, body weight 25% above or below ideal for height, alcoholism, diabetes, diastolic BP > 90 mm Hg, smoked > 10 cigarettes/day, family history of hyperlipidemia or use of drugs that affect lipid metabolism.
Interventions

1) Gestodene 50-70-100 µg + ethinyl estradiol (EE) 30-40-30 µg (N=22)

2) Levonorgestrel 50-75-125 µg + EE 30-40-30 µg (N=21);

6 treatment cycles

OutcomesFasting plasma glucose, glycosylated hemoglobin
NotesNo information on method for randomization or blinding.
Losses: 5/43 = 12%
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskStudy drug provided by Schering Chemicals

Ball 1991

MethodsRandomized trial; "single-blind" (unspecified); stratified according to prior OC use
Participants51 women, 17 to 41 years, requesting OC. New OC users had not used an OC or hormone therapy for 3 months; switchers were changing from low-dose combined OC.
Exclusion criteria: hypertension (diastolic BP > 100 mm Hg, systolic BP > 140 mm Hg), smoking > 20 cigarettes per day, or diabetes.
InterventionsProgestin-only pills: norethisterone 350 µg (N=23) versus levonorgestrel 30 µg (N=23); 6 treatment cycles.
OutcomesPlasma glucose and glycosylated hemoglobin
NotesNo information on method for randomization. No mention of ethical committee approval or informed consent.
Five women did not return for follow-up, and the researchers excluded them from the analysis. Nine withdrew after 3 months; 4 in each group reportedly did not return for fasting blood sample. Analysis included 33 women; loss 18/51 =35%.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskNo information

Basdevant 1993

Methods"Allocated at random"; conducted in France.
Participants58 healthy women. Inclusion criteria: regularly menstruating, at least 3 months postpartum or postabortion, not lactating, and no steroid treatment in past 3 months.
Exclusion criteria: symptoms or history of venous or arterial disease, diabetes (WHO criteria), hyperlipidemia (cholesterol > 6.4 mmol/l and triglycerides > 2.5 mmol/l), hypertension (BP > 140/90), body mass index (BMI) >27, eating disorders, smoked > 10 cigarettes/day, gynecological tumors, known or suspected cancer, treatment with antibiotics, barbiturates or other drugs that interfere with hepatic metabolism.
InterventionsDesogestrel 150 µg + EE 30 µg (N=25) versus desogestrel 150 µg + EE 20 µg (N=33); 6 treatment cycles
OutcomesFasting blood glucose and insulin
NotesNo information on method for randomization or blinding.
Losses: 4 (2 from each group) before baseline and 5 discontinued early (9/58 =16%); their data were excluded from the analysis.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskNo information

Beasley 2012

Methods

Randomized double-blind trial conducted at one New York City site from 2006 to 2008. Sequence for 1:1 allocation determined with random-number table constrained by randomly permutated blocks stratified by BMI class: either normal weight (BMI 19.0 to 24.9) or obese (BMI 30.0 to 39.9). An investigator not involved with participant contact generated the allocation schedule.

Blinding: product labeling obliterated by opaque sticker on pill pack to blind participants to OC formulation. Laboratory personnel blinded to treatment assignment and whether blood specimens belonged to a normal-weight or obese woman.

Sample size chosen for the primary objective of ovarian follicular suppression during OC use. Power analysis was retrospective; using the observed sample size and variance, the analysis reportedly had 80% power to detect mean HOMA-IR change (homeostatic model assessment) of 1.4 between BMI groups and 1.3 between pill groups, mean insulin change of 4.7 μIU/mL between BMI groups and 4.5 μIU/mL between pill groups, and mean glucose change of 4.4 mg/dL between BMI groups and 4.1 mg/dL between pill groups.

Participants

169 women, 18 to 35 years old, either normal weight (BMI 19.0 to 24.9) or obese (BMI 30.0 to 39.9). Inclusion criteria: regular, spontaneous menstrual cycles; agreed to use a study OC and to undergo eight biweekly study visits.

Exclusion criteria: medical contraindications to the use of combination OCs; recently pregnant or breastfeeding; pre-existing renal disease, diabetes or thyroid, pituitary, adrenal or ovarian disorders; smoking >10 cigarettes per day.

Additional exclusion criteria from ClinicalTrials.gov posting: oophorectomy or polycystic ovary syndrome, taken oral contraceptives to regulate menses recently, weight reduction surgery, used Depo-Provera in past 12 months, desiring pregnancy in next 4 months.

Interventions

1) Levonorgestrel 100 µg plus EE 20 µg (53 normal; 39 obese)

2) Levonorgestrel 150 µg plus EE 30 µg (44 normal; 33 obese)

duration 3 months (4 cycles)

OutcomesSecondary: changes in carbohydrate metabolism (fasting glucose and insulin; homeostatic model assessment to estimate insulin resistance (HOMA-IR) = (fasting insulin μU/mL x fasting glucose mg/dL)/405)
Notes

Excluded from analysis women who were not consistent users of the assigned OC; non-compliant included 9 normal weight and 15 obese women.

Losses: overall 60/169 (36%); normal weight 26/97 (27%), obese 34/72 (47%)

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskSequentially numbered opaque envelopes containing
four OC packs
Pharmaceutical company involvementLow riskStudy drug provided by Duramed Pharmaceuticals. Funded by National Institutes of Health (USA) and other sources.

Benagiano 1997

Methods"Randomized" study conducted in Italy.
Participants60 healthy women. Inclusion criteria: 18 to 35 years; at least 12 weeks post-pregnancy; no diabetes, liver disease or thromboembolic disorder; within 20% ideal body weight; regularly menstruating.
InterventionsNorethisterone enanthate 50 mg + estradiol valerate 5 mg (intramuscular every 30 days) (N=30) versus norethisterone 500-750-1000 µg + EE 35 µg (OC) (N=30); 6 treatment cycles.
OutcomesFasting glucose and insulin; glucose rate of disappearance (euglycemic glucose clamp test)
NotesNo information on method for randomization or on blinding.
Losses: 5 in injectable group and 7 in OC group; 6 women excluded from analysis due to variation coefficient > 5 for euglycemic glucose clamp test (indicating test error) (16/60 = 27%).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskFunded by Schering AG

Biswas 2001

MethodsOpen-label "randomized" study in Singapore. Sample size was based on WHO recommendations for metabolic studies (40 per group)
Participants80 women. Inclusion criteria: 18 to 40 years, sexually active and of childbearing potential, normal menstrual cycles with mean length 24 to 35 days and intra-individual variation + 3 days, good physical and mental health, and no contraindications to contraceptive use.
InterventionsEtonogestrel implant (N=40) versus levonorgestrel implant (N=40) for 24 treatment cycles
OutcomesFasting glucose, insulin, and glycosylated hemoglobin; incremental AUC for glucose and insulin during OGTT, 2-hour response for glucose and insulin.
NotesNo information on method for randomization. No mention of ethical committee approval or informed consent.
Losses: at 12 months, samples were 39 (etonogestrel) and 36 (levonorgestrel) (94%) for 6% loss; at 24 months, samples were 24 and 31, respectively; (25/80 = 31%).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskAuthorship, NV Organon

Bloch 1979

Methods"Randomly allocated"; study conducted in Witwatersrand, South Africa.
ParticipantsWomen "matched for age and parity" (unspecified why or how), proven fertility, attending family planning clinics.
Exclusion criteria: amenorrhea, liver or gastrointestinal disease, breast cancer, gynecological lesions, OC in past 3 months or injection in past 6 months.
InterventionsLevonorgestrel 150 µg + EE 30 µg (N=28) versus biphasic levonorgestrel 50-125 µg + EE 50 µg (N=26); study duration of 1 year.
OutcomesBlood glucose ("random")
NotesNo information on method of randomization or blinding. No mention of ethical committee approval or informed consent.
Researchers did not specify how many women were allocated. Report indicates that 54 women completed the trial.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskNo information

Bowes 1989

MethodsOpen-label, randomized trial conducted in the US. Assignment with computer-generated randomization code.
Participants157 healthy women, 18 to 35 years, within 20% of ideal body weight; menstrual cycles normal (25 to 32 days) and regular (+/- 2 days) ; no contraindications to OC use, no injectable or OC in past 3 months.
Exclusion criteria: Nursing mothers, pregnancy in past 3 months, smoking in past year, diabetes, primary relative with diabetes, birth of infant > 9 pounds, clinically significant disease or taking chronic medication; abnormal lab values: fasting serum cholesterol > 240 mg/dl, serum triglyceride > 145 mg/ml, serum glucose >120 mg/ml; abnormal Pap smear (class III, IV, V) or positive culture for gonorrhea.
InterventionsLevonorgestrel 50-75-125 µg + EE 30-40-30 µg (N=51) versus norethindrone 500-750-1000 µg + EE 35 µg (N=50) versus norethindrone 500-1000-500 µg + EE 35 µg (N=56); 6 treatment cycles.
OutcomesAUC for glucose and insulin, plasma glucose and insulin concentration during OGTT.
NotesLosses: 27/157 = 17%.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskNo information

Cagnacci 2009a

MethodsRandomized trial in Italy; enrollment Mar to Jun 2007; randomization following a "computer-generated list." Sample size estimate based on detecting significant change in glucose test results.
Participants28 healthy women, 18 to 35 years old. Inclusion criteria: BMI < 25 kg/m2; normal menses, no OC or medicine that could affect glucose or lipid metabolism in past 12 months.
InterventionsDesogestrel 150 µg plus EE 20 µg (N=14) versus chlormadinone acetate 2 mg plus EE 30 µg (N=14); duration 6 months
OutcomesAfter 'frequently sampled intravenous glucose tolerance test': AUC for glucose, insulin, and C-peptide
NotesNo mention of blinding except for the data analysis.
Losses: 14% overall; 2 from each group (2 discontinued for personal reasons, 1 was lost, and 1 withdrew)
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear risk"The list was concealed" and an independent physician did the allocation
Pharmaceutical company involvementLow riskUnrestricted university funds

Cagnacci 2009b

Methods

Randomized trial in Italy; enrollment Jan 2005 to Jan 2007; randomization following a "computer-generated list."

Sample size estimate based on detecting significant change in glucose test results.

Participants36 healthy women, 18 to 35 years old. Inclusion criteria: BMI 20 to 25 kg/m2; normal menses, no OC or medicine that could affect glucose or lipid metabolism in past 6 months.
InterventionsDesogestrel 150 µg plus EE 30 µg (high EE) (N=12) versus desogestrel 150 µg plus EE 20 µg (low EE) (N=12) versus vaginal ring (etonogestrel 120 µg plus EE 15 µg) (N=12); duration 6 months
OutcomesAfter 'frequently sampled intravenous glucose tolerance test': AUC for glucose, insulin, and C-peptide. Data for the 2 COC groups were not shown separately unless they differed or were relevant.
NotesNo mention of blinding except for the data analysis.
Losses: 17% overall; by group 8% high EE, 25% low EE, 17% ring
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear risk"The list was concealed" and an independent physician did the allocation
Pharmaceutical company involvementLow riskUnrestricted university funds

Duijkers 2004

MethodsOpen-label, randomized trial in the UK. According to the researcher, a computer-generated list was used for randomization. Block size of 4 for either ring or OC.
Participants85 women, 18 to 40 years, at risk of pregnancy and requesting contraception. Inclusion criteria: menstrual cycle 24 to 35 days (+/- 3 days), BMI >=18 and <= 29 kg/m2.
Exclusion criteria: abnormal carbohydrate metabolism, adrenal or thyroid function; contraindications to contraceptive use; use of injectable in past 6 months or other hormonal contraceptive in past 2 months; genital prolapse, vaginitis or bleeding cervical erosion, Pap smear class III-V, severe or chronic constipation, dyspareunia or other coital problems, use of drug that interferes with sex steroid metabolism, drug or alcohol abuse.
InterventionsVaginal ring releasing etonogestrel 120 µg + EE 15 µg daily (N=44) versus OC with levonorgestrel 150 µg + EE 30 µg (N=41); 6 treatment cycles.
OutcomesAUC during OGTT for glucose and insulin, incremental AUC for glucose and insulin, fasting glycosylated hemoglobin.
NotesCorresponding researcher (Dieben) provided design information and data (means, SD, and Ns for carbohydrate measures).
Eight women withdrew prior to study medication (6 due to abnormal lab values); intent to treat population included 40 (OC) and 37 (ring).
Early discontinuations: 2 in OC group and 6 in ring group. Protocol violations: 3. Per protocol population was 74; loss: 11/85 = 13%.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)High riskAccording to researcher, none used.
Pharmaceutical company involvementHigh riskFunded by NV Organon

Elkind-Hirsch 2007

MethodsRandomized trial conducted in Louisiana (USA); randomization with a "computer-generated program." Modified intent-to-treat analysis used (at least one treatment dose). Post hoc power analysis.
Participants65 healthy women, 18 to 40 years old, seeking contraception or contraceptive steroids for cycle control. All were nonsmokers or had not smoked for 3 months. Exclusion criteria: contraindication to hormonal contraceptive, used drugs that interfered with carbohydrate metabolism, used injectable contraceptive in past 6 months or hormonal IUD or OC in past 2 months; had condition relevant to ring use, such as cervicitis or vaginitis; non-normal Papanicolaou smear; prolapse of cervix, cystocele, or rectocele.
InterventionsVaginal ring (releasing etonogestrel 120 µg plus EE 15 µg daily) (N=34) versus OC containing levonorgestrel 100 µg plus 20 µg (N=31); duration 5 cycles
OutcomesInsulin sensitivity after OGTT
NotesNo mention of blinding.
Losses: none to follow-up reported. Overall: 35% ring and 35% COC; includes excluded women who never used study product and those who discontinued early.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskFunded by Organon USA

Fahmy 1991

MethodsAccording to the researcher, randomization was conducted with paper slips (20 DMPA and 20 NET-EN) and the lab assessments were blinded. Study was conducted in Egypt.
Participants40 multiparous women, attending family planning clinic. Inclusion criteria: willingness and fitness for injectables, having at least 2 living children, not lactating, and no hormonal contraception in past 6 months.
Exclusion criteria: liver disease, bilharziasis (schistosomiasis), diabetes mellitus.
InterventionsInjectables: DMPA 150 mg (N=20) every 90 (+/- 5) days for 12 months versus norethisterone enanthate 200 mg (N=20) every 60 (+/- 5) days for 6 months, then every 84 (+/-) 5 days for another 6 months. Study duration was 12 months.
OutcomesFasting serum insulin; during OGTT, serum glucose and 2-hour serum insulin.
NotesResearcher provided information on study design and additional data.
Analysis included all participants (N=40).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskAccording to the researcher, serially-numbered, sealed and mixed envelopes were used to conceal allocation until after assignment. Did not specify if the envelopes were opaque.
Pharmaceutical company involvementUnclear riskNo information

Gaspard 2003

MethodsOpen-label randomized trial at one center in Belgium. According to correspondence with the researcher, randomization done with permuted block technique with block size of four.
Participants60 women. Inclusion criteria: wanted contraception for at least 13 cycles, new OC users or switchers with at least 2 OC-free cycles before study. Exclusion criteria: contraindications to OC use, no parenteral depot contraceptive in past 6 months, co-existing diseases such as diabetes or endocrinopathies, use of medications that interfere with lipid or carbohydrate metabolism, diagnostically unclassified genital bleeding, history of migraine with menstruation.
InterventionsDrospirenone 3 mg + EE 30 µg (N=30) versus desogestrel 150 µg + EE 30 µg (N=30); 13 treatment cycles.
OutcomesArea under the curve for glucose and insulin during OGTT.
NotesLosses included 6 women who did not start study medication (3 each group) and 4 who discontinued early (2 each group) (10/60 = 17%). Per protocol analysis included the remaining 50.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskAccording to the researcher, central telephone system used; participant identifiers were recorded before allocation was revealed.
Pharmaceutical company involvementHigh riskAuthorship, Schering AG

Gillespy 1991

Methods"Randomly assigned"; conducted in the US.
ParticipantsWomen who responded to newspaper advertisements. Inclusion criteria: 18 to 35 years, at least 90 days postpartum, within 20% of average weight for height and age, regular menses, no change in diet or exercise pattern for study duration, no contraindications to OCs, limited alcohol and cigarette use, no hormone use in 30 days, and no other medications within 7 days of study.
InterventionsLevonorgestrel 50-75-125 µg + EE 30-40-30 µg (N=27) versus norethindrone 500-750-1000 µg + EE 35 µg (N=30); study duration was 12 months.
OutcomesDuring OGTT, plasma glucose and insulin.
Notes

No information on method of randomization or blinding. Corresponding researcher was unable to provide design information as records were no longer available.

Complete insulin data for 48 women (incomplete 9/57 = 16%)

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskFunded by Wyeth-Ayerst

Junge 2011

Methods

Randomized, open-label, single-center study conducted in Germany between March 2005 and March 2006. Randomization list generated by Central Randomization Service of sponsor.

Sample size based on having 20 per group; 30 per group recruited, assuming 30% drop-out rate.

Participants60 healthy women aged 18 to 50 years requiring contraception. Inclusion criteria: cigarette consumption <=10 per day if aged 18 to 30 years; nonsmokers aged >30 years.
Exclusion criteria: pregnancy; lactation; < 3 regular menstrual cycles following childbirth, abortion or breastfeeding; amenorrhea; and other typical exclusion criteria for COC studies such as known liver diseases; known alcohol, drug or medicine abuse; body mass index < 17.5 kg/m2 or > 30 kg/m2; prohibited concomitant medication; and use of sex hormones prior to treatment.
Interventions

1) Quadriphasic dienogest (DNG) + estradiol valerate (E2V) (N=30): E2V 3 mg on days 1 and 2, DNG 2 mg + E2V 2 mg on days 3 to 7, DNG 3 mg + E2V 2 mg on days 8 to 24, E2V 1 mg on days 25 to 26, and placebo on days 27 to 28 (N=30);

2) Triphasic levonorgestrel (LNG) + EE (N=28): LNG 50 µg + EE 30 µg on days 1 to 6, LNG 75 µg + EE 40 µg on days 7 to 11, LNG 125 µg + EE 30 µg on days 12 to 21, and placebo on days 22 to 28 (N=28);

treatment duration 7 cycles

OutcomesInsulin and glucose AUC over 120 minutes (oral glucose tolerance test)
Notes

Analysis excluded women who did not take any study medication.

Losses (with exclusions, discontinuations): overall 10/60 = 17%; 5/30 per group = 17%.

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskCorrespondence with researcher (sponsor representative) indicated no concealment of allocation: "assigned in ascending order according to the volunteers’ admission" to study
Pharmaceutical company involvementHigh riskFunded by Bayer HealthCare

Klipping 2005

MethodsOpen-label randomized trial at one site in The Netherlands. Use of computer-generated randomization code. Recruited 30 women per group in order to have 20 per group for per protocol analysis (assumed 30% drop outs).
Participants60 healthy women, 18 to 35 years, including smokers up to age 30 years, requesting OC. Exclusion criteria: pregnancy or lactation, contraindications to steroidal contraceptives, BMI > 30 kg/m2; uncontrolled thyroid disorders, clinically significant findings that might worsen with hormonal therapy, depression in past year, vascular disease (or its risks), diabetes or impaired glucose tolerance, sickle cell anemia, disturbance of lipid metabolism; use of medication that affects metabolism or pharmokinetics of OCs, such hydantoins, barbiturates, rifampicin, or St John's Wort; use of OC in past 2 cycles, sex hormones in past 3 cycles, injectable or implant in past 6 months; uncontrolled hypertension; and malignant or premalignant tumors.
InterventionsDrospirenone 3 mg + EE 20 µg (N=30) versus desogestrel 150 µg + EE 20 µg (N=30); 7 treatment cycles.
OutcomesChange in fasting insulin, AUC for glucose and C-peptide during OGTT.
NotesLosses: 7 discontinued early and were excluded from the per protocol analysis (7/60 = 12%).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskFunded by Schering AG

Knopp 2001

Methods"Randomly assigned"; conducted in the US.
ParticipantsHealthy women. Inclusion criteria: 21 to 35 years, serum LDL-cholesterol < 160 mg/dl and triglycerides < 250 mg/dl, regular menstrual cycle of 25 to 35 days.
Exclusion criteria: illnesses affecting lipoprotein metabolism (hypothyroidism, liver or renal disease, diabetes); use of OCs in past 90 days or injectable in past 6 months; drugs that affect lipid or OC metabolism; gastrointestinal, gynecologic, hematologic, neurologic, respiratory, endocrine, or cardiovascular disease; diastolic BP > 90 or systolic BP > 150, smoking in past 12 months, alcohol or drug abuse, pregnancy or breastfeeding in past year.
InterventionsDesogestrel 50-100-150 µg + EE 35-30-30 µg (N=33) versus levonorgestrel 50-75-125 µg + EE 30-40-30 µg (N=34); 9 treatment cycles
OutcomesGlucose and insulin: change from fasting to one hour after ingesting 400 kcal drink (glucose, protein, fat). Data on carbohydrate measures were provided for cycle 6. Means for fasting values were presented without any measure of variance.
NotesNo information on method of randomization or blinding.
Loss: one woman became pregnant in baseline cycle, and took no study medication; she excluded from the analysis. By cycle 6, losses were 11/68 (16%).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskFunded by Organon, National Institutes of Health (USA), and a foundation

Liukko 1987

Methods"Randomly assigned"; study conducted in Finland.
ParticipantsTwenty healthy women, motivated for OC. Inclusion criteria: regular menstrual period of approximately 28 days, no abnormal carbohydrate metabolism, body weight <= 115% ideal, and no pregnancy or hormonal therapy in past 3 months.
InterventionsLevonorgestrel 150 µg + EE 30 µg (N=10) versus desogestrel 150 µg + EE 30 µg (N=10); study duration was two years.
OutcomesFasting blood glucose
NotesNo information on method of randomization or blinding. No mention of ethical committee approval or informed consent.
Losses: 3 withdrew due to moving (3/20 = 15%).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskAuthorship, Organon

Loke 1992

MethodsRandomized, double-blind trial in Singapore.
Participants58 healthy women were recruited from outpatient and fertility control clinics. Inclusion criteria: 18 to 34 years without diabetes, liver disease, or thromboembolic disorders, at least 70 days postpartum.
InterventionsNorethisterone 1 mg + EE 35 µg (N=29) versus levonorgestrel 150 µg + EE 30 µg (N=29); 12-month duration of study.
OutcomesSerum glucose: fasting and 2 hours after glucose loading
NotesNo information on method of randomization. No mention of ethical committee approval or informed consent.
Losses: none reported.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementLow riskFunded by World Health Organization

Ludicke 2002

MethodsOpen-label, randomized trial in one center
Participants36 lean, healthy women, 19 to 29 years. No other inclusion or exclusion criteria were reported.
InterventionsDesogestrel 150 µg + EE 20 µg (N=18) versus gestodene 75 µg + EE 20 µg (N=18); 13 treatment cycles
OutcomesFasting values as well as AUC during OGTT for glucose, insulin, and C-peptide
NotesNo information on method of randomization. Researchers were from Switzerland and Belgium; study site not specified. Numbers randomized to each group were not clearly stated, but appear to have been 18 each. Sample sizes were specified for the analysis.
Losses: 3 women discontinued in pretreatment cycles; 2 women discontinued during study (5/36 = 14%)
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskNo information

Luyckx 1986

MethodsAccording to correspondence with a researcher (Gaspard): randomization used permuted block technique with block size of four; study was not blinded.
Participants40 healthy women with no family history of diabetes, within 15% ideal body weight, and no OC use in past 8 weeks.
InterventionsLevonorgestrel 50-75-125 µg + EE 30-40-30 µg (N=13) versus desogestrel 125 µg + EE 50 µg (N=13) versus desogestrel 150 µg + EE 30 µg (N=14); 6 treatment months.
OutcomesDuring OGTT, AUC for blood glucose and insulin (units not specified).
NotesResearchers were from Belgium; study site not specified.
Losses: 2 discontinued early and were excluded from the analysis (2/40 = 5%).
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskAccording to a researcher, central telephone system used; participant identifiers were recorded before allocation was revealed.
Pharmaceutical company involvementHigh riskFunded in part by Schering AG

Machado 2010

Methods

Randomized, open-label trial conducted at 3 Brazilian health care centers. Randomization sequence generated by publicly available Random Allocation software.

Randomization and analysis done by independent specialist unrelated to study.

Sample size based on primary objective of evaluating absence of bleeding. Estimates were 80% amenorrhea in continuous group and 1.6% in cyclic group. Assuming 35 to 40% drop-out rate, sample size was estimated at 24 participants (12 per group) with 5% significance and 80% power. Sample of 78 chosen to study secondary outcomes (without statistical consideration).

Participants

78 women, sexually active, requesting contraceptive method and not using hormonal contraception or IUD. Inclusion criteria: age 18 to 35 years old, BMI 19 to 30, at least 8 years of schooling, understand oral and written instructions and willingness to comply with the study requirements.

Exclusion criteria: pregnant (suspected or confirmed); abnormal cervicovaginal colposcopy; history of genital or breast neoplasia, cardiovascular disease, arterial hypertension, venous or arterial thromboembolism, diabetes mellitus, acute or chronic hepatopathies or alcoholism; use of hormonal contraceptives in past 2 months; concomitant use of barbiturates, hydantoins, carbamazepine, phenylbutazone, meprobamate, rifampicin or other drugs that may interact with OCs.

Interventions

Drospirenone 3 mg + EE 30 µg:

1) continuous use for 168 days (N=39) versus

2) cyclic use for 6 cycles (N=39); 28-day cycles with 7-day hormone-free interval between cycles

Outcomes

Primary: bleeding patterns

Secondary: metabolic parameters including glucose (fasting) and insulin (timing not specified)

Notes

Analysis included all women randomized except one in cyclic group who did not take medication; analysis of 77 women (39 continuous and 38 cyclic).

Losses: overall 20/78 (26%); by group, 10/39 (26%). Losses include losses to follow and discontinuations.

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskSequentially numbered opaque envelopes used
Pharmaceutical company involvementHigh riskFunded by Libbs Farmaceutica

Prasad 1989

MethodsAccording to the researcher: Randomization was accomplished with computer-generated random numbers. Physician and lab analyst were blinded; nurse who allocated the pills and the subjects knew the treatment group. Drug companies would not repackage the pills for blinding.
Participants105 Chinese women of proven fertility, 18 to 34 years, seeking OC. Inclusion criteria: regular menstrual cycles (28 + 4 days), no hormonal contraception in past 6 months.
InterventionsLevonorgestrel 50-75-125 µg + EE 30-40-30 µg (N=39) versus desogestrel 150 µg + EE 30 µg (N=32) versus levonorgestrel 150 µg + EE 30 µg (N=34); study duration of 1 year.
OutcomesBlood glucose: fasting and during OGTT (30, 60, 90, 120 minutes).
NotesResearchers were from Singapore; study site not specified.
Losses: 37 discontinued early (37/105 = 35%)
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskFunding and study drug provided by Schering AG and Organon Pte

Rechberger 2004

MethodsRandomly allocated
Participants53 women, 18 to 45 years. Inclusion criteria: no pathological findings from gynecological assessment, no contraindications to OCs.
InterventionsDesogestrel 150 µg + EE 20 µg (N=13) versus gestodene 75 µg + EE 20 µg (N=15) versus desogestrel 150 µg + EE 30 µg (N=11) versus gestodene 75 µg + EE 30 µg (N=14); 6 treatment cycles.
OutcomesFasting plasma glucose and insulin
NotesNo information on method of randomization or blinding. Researchers were from Poland; study site not specified.
Losses: none reported
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementUnclear riskNo information

Reisman 1999

MethodsOpen-label, randomized study at 11 sites in the US. Treatment allocation 1:1, stratified by site. A priori sample size determination.
Participants387 women. Inclusion criteria: 18 to 35 years for smokers (< 15 cigarettes/day), no upper age for nonsmokers, regular menstrual cycles (25 to 31 days) for past 3 months, no OC in past 3 months.
Exclusion criteria: contraindications for OC; use of IUD, injectable, or implant in past 6 months; use of concomitant medication without investigator consent.
InterventionsMonophasic levonorgestrel 100 µg + EE 20 µg (N=192) versus triphasic norethindrone 500-750-1000 µg + EE 35 µg (N=195); study duration 4 cycles.
OutcomesChange from baseline in blood glucose
NotesReport notes that assignment was from 1 to 4 cycles.
Losses: 65 did not take study medication, 102 discontinued early (167/387=43%); change data reported (cycle 4 minus baseline) for 235 women.
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low risk"sequentially numbered, sealed envelopes." Did not specify if envelopes were opaque.
Pharmaceutical company involvementHigh riskFunded by Wyeth-Ayerst

Van der Mooren 1999

MethodsOpen-label, randomized, multicenter trial; intent-to-treat analysis include all those assigned to treatment and who took at least one pill.
Participants124 women. Inclusion criteria: 18 to 40 years, regular menstrual cycles (25 to 31 days), BMI <=28 kg/m2, smoked <= 10 cigarettes/day.
Exclusion criteria: thrombophlebitis; thromboembolic disorders; clotting disorders; hyperlipidemia; abnormal HDL-C, LDL-C, thyroid hormone, or thyroid-stimulating hormone levels.
InterventionsGestodene 60 µg + EE 15 µg (N=62) versus desogestrel 150 µg + EE 20 µg (N=62); 6 treatment cycles.
OutcomesFasting glucose, insulin, and C-peptide; AUC for glucose, insulin, and C-peptide during OGTT.
NotesNo information on method of randomization or allocation concealment. Study sites not specified.
Losses: 4 discontinued early (4/124 = 3%)
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementHigh riskAuthorship, Wyeth-Ayerst

WHO 1985

MethodsDouble-blind randomized trial conducted at 7 WHO Collaborating Centers (Hungary, Yugoslavia, Egypt, India (2), Singapore, Cuba)
Participants

847 women recruited from family planning clinics; women who chose IUD (N=195) are not included in this review; another 43 excluded after admission due to abnormal lab values.

Inclusion criteria: 18 to 39 years, parous, apparently healthy, not breastfeeding, no hormonal contraceptive use in past 3 months, no injectable or implant use, no contraindications to OCs.

Interventions

Norethisterone 1000 µg + EE 35 µg (n=429) versus levonorgestrel 150 µg + EE 30 µg (n=418);

12-month duration of study.

OutcomesGlucose: fasting and 2 hours after glucose load
Notes

No information on method of randomization.
Losses: 43 women excluded after admission due to abnormal lab values (no information on which group), and 429 discontinued early (472/1085 = 44%)

In original review, we labeled this trial as 'Sadik 1985'; Sadik is the first person listed on Task Force.

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Unclear riskNo information
Pharmaceutical company involvementLow riskInitiated by WHO Special Program

WHO 1998

MethodsRandomized trial in 4 centers: Hangzhou, Havana, Jakarta, and Shanghai. According to correspondence with researcher, randomization was stratified by center; computer-generated random number sequence; permuted blocks; no blinding.
Participants

357 women requiring contraception to avoid pregnancy.

Inclusion criteria: 18 to 35 years, at least two regular menstrual cycles since last contraceptive method, not lactating in past 6 months, at least 6 months postpartum or postabortion, and one normal cycle since delivery or abortion.
Exclusion criteria: diabetes; Pap smear grade III, IV, or V; history of thromboembolism; vaginal bleeding of unknown etiology; hypertension hyperlipidemia; recent severe liver disease; malignancy; abnormal nipple discharge; low hemoglobin; nulliparous or pregnant; used barbiturates, anticonvulsants, rifamicin, systemic corticosteroids, drugs affecting cardiovascular or hepatic systems, or any drug long-term; used OCs in past 3 months or injectable in 6 months; BMI > 25 kg/m2, smoker.

InterventionsInjectables: norethisterone enanthate 50 mg + estradiol valerate 5 mg (N=184) versus medroxyprogesterone acetate 25 mg + estradiol cypionate 5 mg (N=173); 9 treatment cycles.
OutcomesSerum glucose - fasting and during OGTT (30-minute intervals) for all 4 sites combined; change in insulin AUC was presented by site (1 site did not have insulin data) but samples sizes were not available by center.
NotesResearcher provided additional information on design via correspondence.
Losses: 59 did not finish the treatment period (59/357 = 17%); total of 73 did not finish full follow-up period (73/357 = 20%)
Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)High riskAccording to researcher, none used.
Pharmaceutical company involvementLow riskUnder WHO Special Program

Wiegratz 2010

  1. a

    Interventions were oral contraceptives (OC) unless otherwise specified.
    AUC = area under the curve; incremental AUC is adjusted for fasting value
    BMI = body mass index (kg/m2)
    BP = blood pressure
    Broca index = weight (kg) as % of height (cm) - 100
    DMPA = depot medroxyprogesterone acetate
    EE = ethinyl estradiol
    HbA1c = glycosylated hemoglobin
    IUD = Intrauterine device
    IUS = Intrauterine system
    OGTT = oral glucose tolerance test
    WHO = World Health Organization

Methods

Randomized open-label trial conducted in Germany; randomization 1:1 in blocks of 4 generated via SAS.

Part of larger study on efficacy and safety (51 clinical sites) that randomized 1323 women.

Sample size for sub-study set at 60 (30 each arm) without statistical consideration. Reportedly, sample size was commonly accepted by regulatory authorities for this type of trial.

Participants

60 healthy women, 18 to 40 years of age, seeking contraception. Inclusion criteria: no hormonal medication in past 4 weeks, no drugs known to influence the effects of COCs, no drugs known to influence insulin action in past 4 weeks.

Exclusion criteria: body mass index > 30 kg/m2, thalassemia minor, post-traumatic syndrome, Hashimoto's thyroiditis, laboratory values outside inclusion range and withdrawal of consent.

Interventions

Dienogest (DNG) 2 mg + EE 30 μg:

1) conventional treatment (13 cycles of 21 days treatment +7 days hormone-free) (N=29)

2) extended-cycle regimen (4 extended cycles of 84 days treatment +7 days hormone-free) (N=30);

12 treatment months

Outcomes

a) Serum HbA1c, glucose, insulin, C-peptide (fasting and AUC);

b) calculated insulin resistance (homeostasis model assessment (HOMA-IR) = fasting insulin mU/l x fasting glucose (mmol/l22.5);

c) insulin sensitivity index (ISI) = 10,000/square root of [fasting glucose (mg/dl) x fasting insulin (mU/l)] x [mean OGTT glucose concentration (mg/dl) x mean OGTT insulin concentration (mU/l)]

Notes

Excluded one participant (conventional-cycle group) who became pregnant in control cycle and did not receive any study medication.

'Full analysis' included only those who took at least one dose of study drug and had at least one observation.

Losses: overall (exclusions and discontinuations) 4/60 = 7%; conventional 1/30 = 3%, extended 3/30 = 10%

Risk of bias
BiasAuthors' judgementSupport for judgement
Allocation concealment (selection bias)Low riskInvestigators provided with sealed volunteer packages for each random number (study medication and allocation to treatment).
Pharmaceutical company involvementHigh riskFunded by Jenapharm GmbH

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
Amatayakul 1988Intervention included insulin-induced hypoglycemia; glucose measured afterward.
Aznar 1976Participants had personal or family history of diabetes. This review excludes studies of women with known diabetes.
Bergink 1984Trial did not include any of the measures for carbohydrate metabolism.
Briggs 1980Briggs is suspected of scientific fraud (Rossiter 1992). This is an earlier conference report of what was later published as Briggs 1982.
Briggs 1982Briggs is suspected of scientific fraud (Rossiter 1992).
Burkman 1992Study was not RCT.
Endrikat 2001Carbohydrate metabolism was studied in a subgroup of 27 volunteers (out of the original 69 who had been randomized).
Endrikat 2002Insufficient data for analysis; report contained medians and ranges. Researcher was not able to provide additional information, due to the time since publication.
Jandrain 1990Insufficient data for analysis; report contained means but no variance measure. Unable to obtain further information from researcher.
Kivela 2001Insufficient data for analysis; report contained medians and ranges. Corresponding researcher was not able to locate any more information on design or data.
Kuhl 1985Insufficient data for analysis. Researcher was not able to provide additional information, due to the time since publication.
Lepot 1987Insufficient data for analysis; report contained means but no variance measure. Researcher provided design information; study was too old to request data.
Liukko 1979No mention of random or allocation method.
Miccoli 1989Insufficient data for analysis; results were presented only in figures. Unable to obtain further information from researcher.
Mostafavi 1999Insufficient data for analysis; no sample sizes for comparison groups. Unable to obtain further information from researcher.
Oelkers 1995Insufficient data for analysis; report contained means but no variance measure. Researcher forwarded correspondence to study sponsor (co-author), who provided information on design.
Pakarinen 1999Insufficient data for analysis; results were presented only in figures. Researcher provided design information.
Petersen 1991Insufficient data for analysis; report contained medians and ranges. Other results were presented only in figures. Unable to obtain further information from researcher.
Petersen 1999Insufficient data for analysis; report contained medians and ranges rather than means. Unable to obtain further information from researcher.
Rabe 1987Insufficient data for analysis; results were presented only in figures. Requested design information but not data due to age of study.
Rad 2011Insufficient data for analysis. Standard errors reported; no sample sizes provided for carbohydrate results. Excluded participants from analysis if fasting status was unknown. Unable to obtain additional information from researcher.
Rakoczi 1985Insufficient data for analysis; report contained means but no variance measure. Requested design information but not data due to age of study.
Saleh 1977Report does not mention random assignment.
Skouby 2005Insufficient data for analysis; report contained medians and ranges rather than means.
Song 1992This was a small crossover trial (N=12). Outcome data were not available for each treatment segment prior to crossover.
Spellacy 1972Study was not RCT.
Spellacy 1990Study was not randomized, according to correspondence with the researcher.
Straznicky 1998Randomized to diet not oral contraceptive.
Van den Ende 1997Intervention was only two months in duration.
Van der Vange 1987Insufficient data for analysis; report contained means but no variance measure.
Winkler 2009Insufficient data for analysis; results were presented only in figures. Unable to obtain further information from researchers.
Winkler 2010Insufficient data for analysis; results were presented as medians and ranges. Unable to obtain further information from researchers.

Characteristics of studies awaiting assessment [ordered by study ID]

UW 2011

MethodsRandomized open-label trial; crossover assignment
Participants36 women, 18 to 50 years old
InterventionsCOC (norgestimate 250 µg plus EE 35 µg); patch (norelgestromin 6 mg plus EE 75 µg)
1: visits 2 to 6: patch; visits 6 to 11: COC; visits 11 to 15: extended use of patch
2: visits 2 to 6: patch; visits 6 to 11: extended use of patch; visits 11 to 15: COC
3: visits 2 to 6: COC; visits 6 to 11: patch; visits 11 to 15: extended use of patch
4: visits 2 to 6: COC; visits 6 to 11: extended use of patch; visits 11 to 15: patch
5: visits 2 to 6: extended use of patch; visits 6 to 11: patch; visits 11 to 15: COC
6: visits 2 to 6: extended use of patch; visits 6 to 11: COC; visits 11 to 15: patch
OutcomesPrimary (selected): glucose and insulin; measured at baseline and months 2, 4 and 6 
Notes

Start date: Feb 2007; estimated completion Dec 2011.

ClinicalTrials.gov listing has not been updated since Feb 2011.

Robert H Knopp MD was the PI, but is now deceased (according to University of Washington web site; May 2010).