Since the introduction of combined estrogen and progestin oral contraceptives (COCs) in the early 1960s, the dose of estrogen has been reduced progressively. COCs containing 50 μg of estrogen or more comprised over 99% of OC retail prescriptions in the United States (USA) in 1968 (Gerstman 1991). Twenty years later, less than 2% of these prescriptions were for 50-μg estrogen COCs. The reduction in the estrogen dose has been in response to two main discoveries. First, concern about estrogen-related adverse effects has driven the search for lower-dose estrogen COCs. COC use has been linked in epidemiological studies to breast cancer (CGHFBC 1996), and estrogen has been associated with a number of adverse events, including cerebrovascular complications, thromboembolic incidents, and myocardial infarction (Anonymous 2000). Low-estrogen COCs have been formulated in an attempt to reduce the risk of these rare events. Second, the discovery that estrogen and progestin act synergistically to inhibit ovulation revealed that contraceptive efficacy could be maintained with lower doses of each component. COCs with less than 50 μg estrogen contain ethinyl estradiol (EE) as the estrogenic component in combination with a progestin (Nelson 2007). Twenty-μg EE COCs ( Table 1) first became available in the 1970s and, by 1998, accounted for about 8% of COC prescriptions in the USA (Wallach 2000). COC pills with 15 μg of estrogen contain the lowest estrogen dose available ( Table 2) and have been approved for use in some countries in Europe, South America and elsewhere (IPPF 2013).
Reducing the estrogen dose to improve safety also could decrease contraceptive effectiveness and cycle control. Contraceptive effectiveness depends both on individual susceptibility and compliance. Determining the optimal estrogen dose required is complicated by high biological variation. The inter-individual variation in the blood levels of exogenous hormones has been estimated to vary tenfold (Guillebaud 1989) and intra-individual differences also occur. Consequently, the lowest estrogen dose needed to prevent pregnancy while also maintaining acceptable cycle control and minimal adverse side effects could vary substantially both among women and within an individual woman at different times. Also, the contraceptive effectiveness of low-estrogen COCs could be influenced more by missed pills or drug interactions (Elstein 1994).
Contraceptives with lower doses of estrogen also may compromise cycle control. Evaluating these concerns is difficult due to the lack of uniformity in the analyses of bleeding patterns. Recognizing the ambiguities in the interpretation of trial results and the comparison of contraceptive products, the World Health Organization (WHO) issued recommendations to standardize the collection and analysis of bleeding pattern data (Belsey 1986). Recently, Mishell 2007a reviewed methods used for collecting and analyzing bleeding data for trial reports of combined hormonal contraceptives. Mishell 2007b then developed recommendations for designing trials and for data collection and analysis of bleeding data. However, when this review was initially conducted, Belsey 1986 was considered the standard. 'Bleeding' was defined as a bloody vaginal discharge that requires sanitary protection whereas 'spotting' does not require protection. A 'bleeding or spotting episode' consists of one or more days with bleeding or spotting that is bounded by days without bleeding or spotting. The WHO advised that the woman, rather than the cycle, be used as the unit of analysis; this avoids disproportionate weight in the analysis from women who contribute more cycles and also prevents artificially precise confidence intervals (Belsey 1986). Outcomes should be measured using reference periods of at least 90 days, an amount of time that was sufficiently short to allow the identification of changes over time, while also long enough for bleeding patterns to be characterized properly. The reference period was modified to 84 days in a recent trial of COCs to correspond to the typical length of three pill cycles (Miller 2001). The WHO identified five bleeding outcomes to be included: the proportion of women with prolonged, frequent, infrequent, or irregular bleeding or spotting episodes and the proportion with amenorrhea during the reference period (Belsey 1986). They noted that the timing of bleeding is an additional issue particular to COC research and that terms for assessing bleeding associated with the pill-free interval (e.g., intermenstrual or breakthrough bleeding) should be defined and evaluated. However, the reporting of trial results often does not conform to the WHO recommendations. Instead many trials report only cyclical data for outcomes related to the control of the menstrual cycle, often without specifying precise definitions for the terms used.
Low-estrogen COCs have been attributed with more breakthrough bleeding and spotting (Nelson 2007). Although bleeding irregularities do not threaten health, sub-optimal cycle control impairs the acceptance of and adherence to the contraceptive. A large, six-month USA study found that 46% of pill users discontinued the method due to side effects or physician recommendations, and about 12% of these women identified bleeding irregularities as the primary reason for discontinuation (Rosenberg 1998). The progestin type and dosing regimen are also thought to affect cycle control. Reviews of COCs suggest that levonorgestrel results in better cycle control than norethindrone and that gestodene performs better than desogestrel and levonorgestrel (Rosenberg 1992; Maitra 2004). Given the possible relationship between progestin type and bleeding patterns, a meta-analysis of low-estrogen contraceptives should combine only trials that compare identical drugs, dosages, and regimens.
A secondary concern related to reducing the estrogenic component involves possible decreases in the non-contraceptive benefits of COCs. Combined oral contraception confers health benefits, such as the prevention of ovarian cancer and endometrial cancer and the reduction of acne (Nelson 2007), and the effect of reducing the estrogen level on these preventive benefits is unclear. Ness 2000 found that the reduction in the risk of ovarian cancer was similar for COCs containing less than 50 μg estrogen compared to those with 50 μg or more. Because 20-μg estrogen COCs were not analyzed separately from the other low-estrogen pills, the question remains as to whether this low dose is sufficient to maintain the protective effect. Contemporary low-estrogen COC pills may grant no protection against functional ovarian cysts, as was observed with COCs containing higher doses of estrogen (Holt 1992). As to acne reduction, a COC containing 20 μg estrogen and 100 μg levonorgestrel had better results than a placebo in Thiboutot 2001. However, whether the low-dose estrogen contraceptive is as effective in reducing acne as a higher-dose estrogen COC is unknown. Furthermore, low-estrogen contraceptives might vary in their ability to improve acne.
Combined contraceptives should have high contraceptive effectiveness, while maintaining cycle control and causing minimal adverse effects. The present review evaluates COCs containing 20 μg EE or less with those containing a higher dose of EE regarding these key outcomes. Rare adverse events, though, were not a focus because randomized controlled trials generally do not have sufficient power to study infrequent events, even when combined.
To test the hypothesis that COCs containing ≤ 20 μg of EE perform similarly as those containing > 20 μg in terms of contraceptive effectiveness, bleeding patterns, discontinuation, and side effects.
Criteria for considering studies for this review
Types of studies
Published or unpublished randomized controlled trials that compare a COC containing 20 μg of EE or less with a COC containing more than 20 μg EE were eligible. Trials were restricted to those reported in English (Higgins 2011).
Types of participants
Women of reproductive age without medical contraindications to COCs, irrespective of previous COC history were eligible.
Types of interventions
Eligible interventions included any COC containing ≤ 20 μg of EE that was compared with a COC containing > 20 μg EE. Trial interventions had to be designed to be administered for a minimum of three consecutive cycles to be eligible for inclusion. Studies were excluded if the interventions were used primarily as treatment for non-contraceptive conditions (e.g., acne, anovulation, dysmenorrhea, menorrhagia, pelvic pain, or endometriosis).
Types of outcome measures
To be included, trials had to report on contraceptive effectiveness, bleeding patterns, side effects, or trial discontinuation due to bleeding-related reasons or other side effects. Those that measured only biochemical changes were not eligible. The outcomes were measured as follows:
1) Contraceptive effectiveness
- Cumulative life-table or Kaplan-Meier pregnancy rate
- Pregnancy Pearl index
- Proportion of women pregnant
2) Discontinuation (overall, due to bleeding-related reasons, and due to other side effects)
- Cumulative life-table or Kaplan-Meier discontinuation rate
- Proportion of women who did not complete the trial
- Discontinuation Pearl index
3) Bleeding patterns
- Proportion of women with amenorrhea
- Proportion of women with prolonged (i.e., longer than 14 days) bleeding or spotting episodes
- Proportion of women with frequent (i.e., more than 5) bleeding or spotting episodes
- Proportion of women with infrequent (i.e., less than 3) bleeding or spotting episodes
- Proportion of women with irregular bleeding (i.e., 3 to 5 bleeding or spotting episodes and less than 3 bleeding or spotting-free intervals of at least 14 days)
Proportion of women with amenorrhea or breakthrough (also known as intermenstrual) bleeding or spotting for cycle three or cycle six, or if data for these cycles were not reported, for the last cycle of follow up
4) Side effects
Proportion of women experiencing any side effect reported
Search methods for identification of studies
We searched the computerized databases the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, and POPLINE for eligible trials using a list of brand names of COCs containing 20 μg of EE or less (IPPF 2013). We also searched for trials via ClinicalTrials.gov and ICTRP. The strategies are shown in Appendix 1. Previous strategies, which also included EMBASE, can be found in Appendix 2.
Searching other resources
We assessed the references of eligible trials. For the initial review, we wrote to COC manufacturers to request information about any other published or unpublished trials not discovered in our search.
Data collection and analysis
One author evaluated the titles and abstracts located in the literature searches to determine whether they met the inclusion criteria. Two authors independently extracted data from the studies identified for inclusion. We wrote to the researchers when clarifications or additional data were needed. Data were entered and analyzed with RevMan. For trials that included more than two intervention groups, regardless of whether the authors identified a control group, we treated the ≤ 20 μg EE contraceptive as the referent and compared it to the other study groups.
We calculated Peto odds ratios (ORs) with 95% confidence intervals (CIs) for all dichotomous outcomes and mean differences using fixed-effect models with 95% CIs for all continuous outcomes. Contraceptive effectiveness and early discontinuation also were presented as Pearl indices or survival analysis estimates when these measures were available. Pearl indices are calculated by the number of events divided by the total person-time at risk of the event (Trussell 1998). Because contraceptive failure rates typically decrease with duration of use, the Pearl index rate tends to decline as the amount of person-time contributed by each woman increases. Thus, rates from two studies will not be comparable if the studies differed in duration. Although the Pearl method is a sub-optimal measurement, it was included in the review because its use remains prevalent. Survival analyses, which include life-table and Kaplan-Meier methods, are preferred measures. Because RevMan 4.2, used for the initial review, was not designed to include Pearl indices or measures of survival analysis, these estimates were presented in Table 3 and Table 4. Most of the studies that included contraceptive effectiveness data simply reported the number of women who became pregnant, and we used these proportions of women to calculate ORs in RevMan. However, these measures potentially are biased in favor of OCs that result in higher discontinuation rates since women who discontinue the study treatment or study participation or who are lost to follow up were unlikely to have been followed to determine whether they became pregnant.
We included all bleeding-related outcomes reported in the eligible trials (e.g., WHO bleeding terms, breakthrough bleeding, spotting, and amenorrhea). Following WHO recommendations (Belsey 1986), we used women, rather than cycles, as the unit of analysis when possible. For studies that did not use the recommended 90- or 84-day reference period, we extracted bleeding-related data from the third and the sixth cycle, or when neither of these was available, from the last cycle of follow up. Although the choice of these cycles was arbitrary, the standardization in the data extraction facilitated comparisons across trials. Definitions of bleeding-related outcomes were described in Characteristics of included studies.
Interval estimation considers both the magnitude of the estimate and the width of the confidence intervals when assessing potential relationships (Rothman 1998). While all available data are presented in the tables, we highlighted the findings that suggest possible differences in COCs based on interval estimation as well as P values.
The review was limited to the analytic method (e.g., intent-to-treat, per-protocol, or a modification of either type) used in the trial report. We combined study results for meta-analysis only when identical drugs, dosages, and regimens were compared. Homogeneity was assessed by examining the results of both a fixed-effect model and a random-effects model. Because the chi-squared test for heterogeneity is a low-power test, the alpha-level was set at 0.10. We conducted sensitivity analyses by examining the effect of deleting each study in turn to test the robustness of any result that appeared to be based on heterogeneous combinations. We critically appraised validity of trials by assessing the potential for bias resulting from the study design, blinding, randomization method, group allocation concealment, and loss to follow up.
Description of studies
Results of the search
The 2013 search resulted in 201 unduplicated references: 162 from the main databases, 10 from other sources such as reference lists, and 29 ongoing or recent clinical trials. We did not find any RCTs that met our eligibility criteria for this review. One new study was excluded because it reported on pooled analysis from two uncontrolled trials (Marr 2012).
The search strategy yielded 21 primary articles that were eligible for the review. Most trials recruited healthy, reproductive-age women without contraindications to COC use. Trials ranged in duration from 3 to 24 treatment cycles with most designed for 6 or 12 treatment cycles. The location was not described for 7 trials; the remaining 14 trials were conducted in North America, South America, Europe or Asia. The trials ranged from a single site (four trials) to 110 sites except for three trials with unspecified number of sites. Five trials included more than two eligible intervention arms. The trials compared 20 different COC pairs.
Risk of bias in included studies
The reporting of randomized clinical trials should include details on the participants, blinding, randomization method, group allocation concealment method, participant flow, and statistical methods (CONSORT 2009). One trial blinded only the study observer (Appel 1987). Three trials blinded the participants and either the clinic staff (Bounds 1979; WHO 1982) or the investigator (Endrikat 2001). Two articles described double-blinding but did not specify who was blinded (Akerlund 1993; Endrikat 1997). The remaining trials either were open (12 trials) or did not mention blinding (3 trials). Fourteen reports did not provide any details of the method of randomization. Reisman 1999 reported the use of sealed envelopes for allocation concealment; the remaining trials did not provide any details regarding attempts to conceal the allocation process. None of the articles described the person responsible for implementing the randomization process or the use of a centralized location for randomization.
The number of participants who were recruited, randomized, received treatment, lost to follow up, discontinued early, and excluded from the analysis should be reported in the details of the participant flow (CONSORT 2009). Only WHO 1982 reported the number of recruited women who were screened for trial participation. All trials reported the total number of women randomized, although these figures may have been the total numbers analyzed rather than randomized. The number of women reportedly randomized in the trials ranged from 20 to 2894. Four trials did not specify the number of women randomized to each study group (Bounds 1979; Appel 1987; Teichmann 1995; Endrikat 2001). Eleven trials reported excluding randomized women from the analysis, usually for protocol violations. Kluft 2006 used intent-to-treat analysis. Hampton 2001 reported using an intent-to-treat analysis for contraceptive effectiveness but did not specify the analytic method used for the remaining outcomes. Endrikat 2001 reported both modified intent-to-treat and per-protocol analyses without specifying which method was used for the results presented. Seven trials described using a per-protocol or modified per-protocol analysis based on the exclusion of women or treatment cycles from the analysis for failure to comply with the protocol (Akerlund 1993; Brill 1996; Winkler 1996; Chavez 1999; Kaunitz 2009; Reisman 1999; Skouby 2005). However, the decision to exclude the woman or the cycle was unclear in Chavez 1999 because different sections of the article reported conflicting procedures. Inauen 1991 had complete follow up and no early discontinuation. The remaining trials did not specify the analytic method used. Three trials did not report the proportion of women completing the study (Bounds 1979; Brill 1996; Winkler 1996), and the proportion ranged from 46% to 94% in the remaining trials.
Effects of interventions
No significant differences were found in contraceptive effectiveness for the 13 COC pairs for which this outcome was reported ( Table 3).
The COC containing EE 20 μg and norethindrone acetate 1 mg had higher life-table rates of early discontinuation (overall, due to medical reasons, and due to amenorrhea) than its five comparison COCs ( Table 4), but P values were the only measure of variability provided for the estimates (WHO 1982). The remaining six comparisons with data for overall discontinuation found no significant differences between groups. Differences in bleeding-related discontinuations were apparent for three comparisons. Women in the EE 20 μg and desogestrel 150 μg group had an OR of discontinuation due to irregular bleeding 2.59 (95% CI 1.35 to 5.00) times that of the EE 30 μg and desogestrel 150 μg group (Akerlund 1993). Women assigned to use the COC with EE 20 μg and desogestrel 150 μg were 2.4 times as likely to discontinue due to metrorrhagia 2.35 (95% CI 1.16 to 4.77) than those in the EE 30 μg and gestodene 75 μg group (Kirkman 1994; Bruni 2000). In addition, women on EE 20 μg and norethindrone acetate 1 mg were more likely to discontinue due to bleeding than women on EE 30 μg and levonorgestrel 150 μg (Bounds 1979).
EE 20 mg and desogestrel COCs versus higher-estrogen COCs
Women in the EE 20 μg and desogestrel 150 μg group were more likely than those in the EE 30 μg and desogestrel 150 μg group to report irregular bleeding (OR 1.56; 95% CI 1.10 to 2.20) and a longer duration of irregular bleeding during the third cycle (mean difference 0.70 days; 95% CI 0.30 to 1.10) (Akerlund 1993). No significant differences in amenorrhea or the duration of irregular bleeding remained by the sixth cycle. The lower-dose estrogen group was also more likely to report the occurrence of irregular bleeding (OR 1.69; 95% CI 1.07 to 2.69) or prolonged withdrawal bleeding (OR 1.98; 95% CI 1.03 to 3.78) at least once throughout the trial than the higher-dose estrogen group. The EE 20 μg and desogestrel 150 μg group versus the EE 30 μg and gestodene 75 μg had an OR of irregular bleeding of 2.51 (95% CI 1.77 to 3.56) during the third cycle and 1.72 (95% CI 1.15 to 2.55) during the sixth cycle (Kirkman 1994). Women assigned to use EE 20 μg and desogestrel 150 μg were also more likely to report metrorrhagia (i.e., bleeding from the uterus that is not associated with menstruation) at least once during the trial than women using EE 30 μg and gestodene 75 μg (OR 1.67; 95% CI 1.05 to 2.66) (Brill 1996) or women using EE 30-40-30 μg and gestodene 50-70-100 μg (OR 2.28; 95% CI 1.39 to 3.73) (Bruni 2000).
Inauen 1991 compared a COC containing EE 20 μg and desogestrel to a second COC and also reported bleeding outcomes. However, the trial had insufficient power to detect major differences between the groups.
EE 20μg and gestodene COCs versus higher-estrogen COCs
Five trials compared the same gestodene dose (75 μg) but differing EE doses (20 μg versus 30 μg) with a standard (Brill 1996; Winkler 1996; Endrikat 1997; Taneepanichskul 2002) or an extended cycle length (Endrikat 2001). Limitations to assessing bleeding pattern changes included small sample sizes, insufficient data reported, and varying definitions.
EE 20 μg and levonorgestrel COCs versus higher-estrogen COCs
EE 20 μg and drospirenone COC versus higher-estrogen COC
In Kaunitz 2009, the low dose group (EE 20 µg plus drospirenone) had more unscheduled bleeding days than the group with EE 25 μg and norgestimate 180-215-250 µg (mean difference 1.00; 95% CI 0.44 to 1.56).
EE 20 μg and norethindrone acetate COCs versus higher-estrogen COCs
The low-dose COC containing EE 20 μg and norethindrone acetate 1 mg fared worse than the COC EE 30 μg and levonorgestrel 150 μg for three bleeding outcomes during the first to third cycles: 1) frequent bleeding (OR 2.92; 95% CI 2.08 to 4.09); 2) infrequent bleeding (OR 2.84; 95% CI 1.80 to 4.46); and 3) irregular bleeding (OR 4.01; 95% CI 2.12 to 7.61) (WHO 1982). The low-dose COC also had worse bleeding outcomes during the first to third cycles when compared to a COC with the same norethindrone acetate dose (1 mg) but more estrogen (EE 50 μg): 1) frequent bleeding (OR 4.59; 95% CI 3.24 to 6.51); 2) infrequent bleeding (OR 3.08; 95% CI 1.95 to 4.86); 3) irregular bleeding (OR 5.33; 95% CI 2.74 to 10.34); and 4) prolonged bleeding (OR 3.11; 95% CI 1.83 to 5.27).
Similarly, EE 20 μg and norethindrone acetate 1 mg resulted in a longer duration of bleeding or spotting days during the third cycle when compared to three higher-dose COCs containing the same progestin type: 1) EE 30 μg and norethindrone acetate 1.5 μg (mean difference 1.10 days; 95% CI 0.37 to 1.83); 2) EE 50 μg and norethindrone acetate 1 mg (mean difference 1.20 days; 95% CI 0.43 to 1.97); and 3) EE 20-30-50 μg and norethindrone acetate 1-1.5-1 mg (mean difference 1.60 days; 95% CI 0.94 to 2.26) (Appel 1987). Women in the low-dose COC group (EE 20 μg and norethindrone acetate 1 mg) also were more likely to report frequent bleeding (OR 1.97; 95% CI 1.42 to 2.73), infrequent bleeding (OR 1.95; 95% CI 1.27 to 3.00), and irregular bleeding (OR 2.38; 95% CI 1.31 to 4.31) during the first to third cycles than those in the EE 50 μg and norethindrone acetate 1 mg group (WHO 1982). (The groups did not differ significantly in the frequency of prolonged bleeding during the first three cycles.) Likewise, women assigned to this low-dose COC reported more infrequent bleeding (OR 1.88; 95% CI 1.22 to 2.90) and irregular bleeding (OR 1.92; 95% CI 1.08 to 3.43) than those using EE 35 μg and norethindrone acetate 400 μg, but no difference in frequent or prolonged bleeding (WHO 1982).
The sole trial to compare EE 20 μg and norethindrone acetate 1 mg to mestranol 50 μg and norethindrone 1 mg found the low-dose COC group had a higher risk of frequent bleeding (OR 2.82; 95% CI 2.00 to 3.97), infrequent bleeding (OR 2.49; 95% CI 1.58 to 3.91), irregular bleeding (OR 4.85; 95% CI 2.49 to 9.43), and prolonged bleeding (OR 2.67; 95% CI 1.58 to 4.52) during the first to third cycles (WHO 1982). Finally, the COC with EE 20 μg and norethindrone acetate 1 mg resulted in more breakthrough bleeding during the third cycle (OR 2.79; 95% CI 2.09 to 3.74) and during the sixth cycle (OR 2.40; 95% CI 1.78 to 3.24) compared to the COC containing EE 25 μg and norgestimate 180-215-250 μg (Hampton 2001). When breakthrough spotting was included with the outcome breakthrough bleeding, the ORs were similar to those for breakthrough bleeding alone. The bleeding data were re-analyzed with new criteria for a secondary article; the pattern was the same. However, the percentages with unscheduled bleeding at cycle six were 33.5% for EE 20 μg and norethindrone acetate 1 mg versus 21% for EE 25 μg and norgestimate 180-215-250 μg, while the earlier estimates of breakthrough bleeding or spotting were 22.2% and 10.3%, respectively.
Side effects were measured as the proportion of women experiencing the event during the study. Three of the six COC pairs with side effects reported found differences. The ORs of headache (1.71; 95% CI 0.94 to 3.11), dizziness (7.65; 95% CI 1.54 to 38.08), mood changes (1.93; 95% CI 1.05 to 3.56), and weight gain (2.46; 95% CI 1.04 to 5.84) were higher for the COC EE 20 μg and desogestrel 150 μg group than the COC EE 30 μg and desogestrel 150 μg group (Akerlund 1993). However, women in the COC EE 20 μg and desogestrel 150 μg group were less likely to report breast pain (OR 0.70; 95% CI 0.47 to 1.05) and dizziness (OR 0.40; 95% CI 0.17 to 0.93), than those in the EE 30-40-30 μg and gestodene 50-70-100 μg group (Bruni 2000). Also, compared to women in the 35 μg and norethindrone 500-750-1000 μg, group women in the 20 μg and levonorgestrel 100 μg were less likely to report breast pain (OR 0.45; 95% CI 0.22 to 0.93) or vomiting (OR 0.33; 95% CI 0.11 to 0.96) (Chavez 1999).
Few comparisons could be combined in meta-analysis because most studies differed in the COC pairs that were compared. Only two of the comparisons that were eligible for meta-analyses appeared to combine heterogeneous estimates with either fixed-effect or random-effects models. The risk of headache differed in the two studies that compared COCs with gestodene 75 μg but different EE doses (Brill 1996; Endrikat 1997), but neither trial found significant ORs for this outcome. The increased rate of discontinuation due to adverse events in the EE 20 μg and levonorgestrel 100 μg group versus the EE 35 μg and norethindrone 500-750-1000 μg group was the result of Reisman 1999; Chavez 1999 did not show a difference in discontinuation for this reason between the two study groups.
Summary of main results
The included trials provide insufficient evidence to determine whether the contraceptive effectiveness of COCs containing 20 μg of EE differs from those with higher estrogen doses. Most studies were underpowered to study pregnancy as a primary outcome; larger sample sizes are required because COCs are highly effective and few pregnancies occurred. Furthermore, most pregnancy estimates could not be combined in meta-analysis because the study COCs contained different progestins. Research using ultrasonography to measure the growth of follicle-like structures suggests that low-estrogen pills could have a reduced margin of safety from missed pills or drug interactions (Teichmann 1995; Spona 1996). Increased risk of pregnancy from this possible lack of 'forgiveness' for missed pills among the low-dose estrogen users might not have been detected in the present review. Also, many studies excluded women from the analysis for protocol violations, including failure to adhere to the daily pill intake, which could have biased the results.
Early discontinuation from the trial can be considered a proxy measure of method acceptability. Overall discontinuation rates did not vary substantially for most COCs compared. Four comparisons showed higher rates of discontinuation due to adverse events for EE 20 μg pills than their higher-estrogen comparison COCs, and one trial found more medical-related discontinuations for the low-dose estrogen COC than its five high-dose estrogen counterparts. However, discontinuation due to adverse events or medical reasons provides limited information because this outcome can combine disparate reasons for quitting the trial early. For example, amenorrhea is different than excessive or irregular bleeding, and studies should separate these. On the other hand, the included trials might have been underpowered to detect differences between specific reasons for discontinuation. Few trials reported data for discontinuation due to specific adverse events.
Several COCs containing 20 μg EE resulted in higher rates of outcomes related to lack of bleeding (amenorrhea and infrequent bleeding) as well as changes in bleeding (irregular, prolonged, frequent, and breakthrough bleeding) than their higher-estrogen comparison pills. Comparing bleeding data from studies is complicated by lack of uniformity in the outcomes used. Also, determining the clinical importance of changes in bleeding patterns is difficult since women in different cultures may view menstrual patterns and assess the acceptability of any changes differently. For example, amenorrhea or infrequent bleeding may be more acceptable in certain cultural settings than others. Finally, the progestin type and dose as well as the ratio of the progestin and estrogen doses could also affect bleeding patterns. Research suggests that certain progestin types could result in better cycle control than others (Rosenberg 1992). While bleeding irregularities do not pose a health risk, they reduce method acceptability and adherence by users.
Overall completeness and applicability of evidence
The randomized controlled trial design, generally, is not suited for evaluating the risk of rare adverse events (e.g., thromboembolic events or myocardial infarction). Studies assessing various hemostatic outcomes as markers for these rare events indicate that the effects on these intermediate variables might be less with 20 μg pills (Basdevant 1993; Norris 1996; Winkler 1996). However, usual tests of hemostasis do not predict clinical events, so these intermediate outcomes have no clinical utility.
Quality of the evidence
Trials often failed to report a measure of variability for outcome data (e.g., reporting dichotomous outcomes in percentages, rather than absolute numbers, not including standard deviation or confidence interval for continuous outcomes). The lack of variability estimates severely constrains the interpretation of a point estimate and, consequently, prevented their inclusion in this review. A second limitation was that none of the trials described using adequate steps to conceal the allocation process during randomization. Failure to conceal the allocation sequence can lead to biased results (Schulz 2002). Furthermore, most trials were unblinded or did not report blinding of group assignment, and the knowledge of the group assignment could have introduced bias. Losses to follow up were high in many trials, which can bias the results (Strauss 2005). Finally, all of the trials appear to have been funded by pharmaceutical companies except for WHO 1982 and two that did not specify an industry relationship (Basdevant 1993; Teichmann 1995). Pharmaceutical sponsorship represents a potential conflict of interest and can introduce bias in terms of the study design, analysis or reporting of unfavorable results (Lexchin 2003).
Implications for practice
While COCs containing 20 μg EE may be theoretically safer, this review did not focus on the rare events required to assess this hypothesis. Data from existing randomized controlled trials are inadequate to detect possible differences in contraceptive effectiveness. Twenty μg EE COCs resulted in higher rates of bleeding pattern disruptions. However, most trials compared oral contraceptives containing different progestin types, and changes in bleeding patterns could be related to progestin type as well as estrogen dose.
Implications for research
Large randomized controlled trials comparing regimens with the same progestin type are needed to determine whether the contraceptive effectiveness of 20-μg EE COCs is similar to that of their higher-estrogen counterparts. Likewise, studies of bleeding patterns should compare pills containing the same progestin type but different estrogen doses. Trials should use standardized methods for collecting and analyzing bleeding data (Mishell 2007b). Trials should also employ better research methods, e.g., adequate allocation concealment (Schulz 2002), and follow CONSORT guidelines for reporting the results (CONSORT 2009). Higher follow-up rates are essential for meaningful interpretation of the results.
Carol Manion of FHI 360 helped develop and execute the literature searches.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. Search strategies 2013
OC BRAND LIST (used within MEDLINE and CENTRAL searches)
Adoless OR Alesse OR Aliane OR Allestra OR Anulette 20 OR April OR Arianna OR Beyaz OR Careza OR Ciclidon OR Ciclomex OR Ciclotab OR Cycleane-20 OR Dal OR Desmin OR Desorelle OR Desoren OR Diminut OR Diva OR Dschess OR Dzhess OR Eloine OR Elyfem OR Estinette OR Eve OR Fedra OR Femadiol-Mepha OR Femexin OR Femiane OR Femilon OR Femina OR Feminet OR Feminol-20 OR Femodette OR Gedarel OR Ginelea OR Ginesse OR Gynera OR Gynostat-20 OR Gynovin OR Gyselle OR Harmonet OR Jasminelle OR Jasminellecontinu OR Leois OR Lerogin OR Levlite OR Lindynette OR Liofora OR Loestrin OR Loette OR Logest OR Lovelle OR Lovette OR Lowette OR Marvelon OR Meliane OR Meloden OR Melodene OR Melodia OR Mercilon OR Microgen OR Microgynon OR Microlevlen OR Microlite OR Midalet OR Millinette OR Minesse OR Minestril-20 OR Minestrin OR Minian OR Minima OR Minifem OR Minigeste OR Minisiston OR Miranova OR Mirelle OR Myralon OR Neolette OR Norvetal OR Novinet OR Novynette OR Ortho Tri-cyclen Lo OR Primera OR Secret OR Securgin OR Siblima OR Suavuret OR Sunya OR Tamisa OR Vivelle OR Yasmin OR Yasminelle OR Yaz
MEDLINE via PubMed (01 Mar 2010 to 10 Jul 2013)
contraceptives, oral[MeSH] AND ("low dose" OR "low-dose"[title/abstract word] OR "ultra low dose" OR "ultra-low-dose"[title/abstract word] OR [OC BRAND LIST])
AND (Clinical Trial[ptyp])
CENTRAL (2010 to 10 Jul 2013)
Title, abstract, keywords: oral AND contracept*
AND Title, abstract, keywords: low dose OR low-dose OR ultra low dose OR ultra-low-dose OR [OC BRAND LIST]
POPLINE (2010 to 11 Jul 2013)
oral contraceptives, low-dose OR (contraceptive agents, female)
Filter by keywords: clinical trials, oral contraceptives combined
ClinicalTrials.gov (01 Sep 2010 to 08 Jul 2013)
Search terms: 20 μg OR 15 μg
Study type: Interventional studies
Condition: NOT (HIV OR acne OR PMDD OR post-menopausal OR postmenopausal OR polycystic OR PCOS OR dysmenorrhea OR cancer OR anorexia)
Intervention: oral AND (contraceptive OR contraception)
Gender: Studies with female participants
ICTRP (01 Sep 2010 to 08 Jul 2013)
Condition: contraceptive OR contraception
Intervention: 20 μg OR 15 μg
Appendix 2. Previous search strategies
OC BRAND LIST (used within database searches below)
Alesse OR Allestra OR Anulette OR April OR Arianna OR Careza OR Ciclidon OR Ciclomex OR Ciclotab OR Cycleane-20 OR Dal OR Desmin OR Desorelle OR Desoren OR Diminut OR Estrostep Eve OR Fedra OR Femexin OR Femiane OR Femilon OR Femina OR Feminet OR Feminol-20 OR Femodette OR Ginelea OR Ginesse OR Gynera OR Gynostat-20 OR Gynovin OR Harmonet OR Lamuna OR Leois OR Lerogin OR Levlite OR Lindynette OR Loestrin OR Leotte OR Logest OR Lovelle OR Lovette OR Lowette OR Lovina OR Marvelon OR Meliane OR Meloden OR Melodene OR Melodia OR Mercilon OR Microdosis OR Microgen OR Microgynon OR Microlevlen OR Microlite OR Midalet OR Minesse OR Minestril-20 OR Minestrin OR Minian OR Minima OR Minifem OR Minigeste OR Miranova OR Mircette OR Mirelle OR Myralon OR Neolette OR Norvetal OR Novinet OR Novynette OR Primera OR Secret OR Securgin OR Segurin OR Siblima OR Suavuret OR Tamisa OR YAZ
MEDLINE via PubMed (to 21 Sep 2010)
((randomized controlled trials [pt] OR controlled clinical trial [pt] OR randomized controlled trials [mh] OR random allocation [mh] OR double-blind method [mh] OR single-blind method [mh] OR clinical trial [pt] OR clinical trials [mh] OR ("clinical trial" [tw]) OR ((singl* [tw] OR doubl* [tw] OR trebl* [tw] OR tripl* [tw]) AND (mask* [tw] OR blind* [tw])) OR ("latin square" [tw]) OR placebos [mh] OR placebo* [tw] OR random* [tw] OR research design [mh:noexp] OR comparative study [mh] OR evaluation studies [mh] OR follow-up studies [mh] OR prospective studies [mh] OR cross-over studies [mh] OR control* [tw] OR prospectiv* [tw] OR volunteer* [tw]) NOT (animal [mh] NOT human [mh])) AND (eng [la] AND contraceptives, oral[MeSH] AND ("low dose" OR "low-dose"[title/abstract word] OR "ultra low dose" OR "ultra-low-dose"[title/abstract word] OR [OC BRAND LIST])
CENTRAL (to 21 Sep 2010)
(oral AND (contracept*) in Title, Abstract or Keywords AND (low dose OR low-dose OR ultra low dose OR ultra-low-dose OR [OC BRAND LIST]) in Title, Abstract or Keywords
EMBASE (to 03 Nov 2010)
((low dose oral contraceptive OR (oral contraceptive agent AND low(W)dose)) OR (oral contraceptive agent AND [OC BRAND LIST])) AND (clinical trial OR controlled study OR randomized controlled trial OR (controlled(W)clinical(W)trial) OR (random(W)allocation) OR multicenter study OR (comparative(W)study) OR (evidence(W)based(W)medicine) OR (research(W)design) OR (double(W)blind(W)procedure) OR (single(W)blind(W)procedure) OR random) AND human
POPLINE (to 01 Nov 2010)
(oral contraceptives, low-dose) / (contraceptive agents, female & [OC BRAND LIST])
ClinicalTrials.gov (to 28 Sep 2010)
Search terms: 20 μg OR 15 μg
Condition: NOT (HIV OR acne OR PMDD OR post-menopausal OR postmenopausal OR polycystic OR PCOS OR dysmenorrhea OR cancer OR anorexia)
Intervention: oral AND (contraceptive OR contraception)
Study type: Interventional studies
Gender: Studies with female participants
ICTRP (to 28 Sep 2010)
Condition: contraceptive OR contraception
Intervention: 20 μg OR 15 μg
Last assessed as up-to-date: 15 July 2013.
Protocol first published: Issue 1, 2003
Review first published: Issue 2, 2005
Contributions of authors
M Gallo extracted data for the initial review and drafted the original review. K Nanda developed the idea and extracted data for the initial review. K Nanda, D Grimes and K Schulz revised and approved the initial review. L Lopez conducted the updates from 2008 to 2013, extracted new data, and revised the review accordingly. D Grimes did the second data extraction for the 2008 and 2010 updates.
Declarations of interest
DA Grimes has consulted with the pharmaceutical companies Bayer Healthcare Pharmaceuticals and Merck & Co, Inc.
Sources of support
- No sources of support supplied
- National Institute of Child Health and Human Development, USA.Support for conducting the review at FHI 360
- US Agency for International Development, USA.Support for conducting the review at FHI 360
Medical Subject Headings (MeSH)
Contraceptives, Oral, Combined [*administration & dosage; adverse effects]; Contraceptives, Oral, Hormonal [*administration & dosage; adverse effects]; Desogestrel [administration & dosage; adverse effects]; Estrogens [*administration & dosage; adverse effects]; Ethinyl Estradiol [*administration & dosage; adverse effects]; Menstruation Disturbances [chemically induced]; Randomized Controlled Trials as Topic
MeSH check words
* Indicates the major publication for the study