Dr HR Verhoeve, Department of Obstetrics and Gynaecology, Onze Lieve Vrouwe Gasthuis, Oosterpark 9, 1090 AC Amsterdam, the Netherlands. Email firstname.lastname@example.org
Background Guidelines recommend diagnostic laparoscopy in subfertile women with known co-morbidities in their medical history. Aggregated evidence underpinning these recommendations is, however, currently lacking.
Objective The objective of this study was to perform a systematic review and meta-analysis of the available evidence on the association between items reported during medical history taking and tuboperitoneal pathology.
Search strategy MEDLINE (from 1966 to May 2007), EMBASE (from 1960 to January 2007) and bibliographies of retrieved primary articles.
Selection criteria All relevant studies that compared medical history with the presence or absence of tubal pathology.
Data collection and analysis Studies comparing medical history with the presence or absence of tubal pathology were included. A diagnosis of tubal pathology had to be made by hysterosalpingography, laparoscopy or a combination of both. In the absence of invasive tubal testing, tuboperitoneal pathology was considered to be absent in case of intrauterine pregnancy. Homogeneity between studies was assessed, and the association between medical history and tubal pathology was expressed as a common odds ratio with a 95% CI. No language restriction was applied.
Main results We included 32 studies. In cohort studies, strong associations were found for a history of complicated appendicitis (OR 7.2, 95% CI 2.2–22.8), pelvic surgery (OR 3.6, 95% CI 1.4–9.0) and pelvic inflammatory disease (PID) (OR 3.2, 95% CI 1.6–6.6), and in case–control studies, for a history of complicated appendicitis (OR 3.3, 95% CI 1.8–6.3), PID (OR 5.5, 95% CI 2.7–11.0), ectopic pregnancy (OR 16.0, 95% CI 12.5–20.4), endometriosis (OR 5.9, 95% CI 3.2–10.8) and sexually transmitted disease (OR 11.9, 95% CI 4.3–33.3).
Author’s conclusions Subfertile women reporting a history of PID, complicated appendicitis, pelvic surgery, ectopic pregnancy and endometriosis are at increased risk of having tuboperitoneal pathology. In these women, diagnostic laparoscopy should be offered early in the fertility work-up.
Approximately one in ten couples who are trying to conceive have not succeeded after 1 year and suffer from subfertility.1,2 It has been estimated that in about 10–30% of these couples, subfertility is due to tuboperitoneal factors.3 Therefore, tubal patency testing is part of the standard fertility work-up. Diagnostic laparoscopy with tubal testing is considered the reference standard for diagnosing tubal disease but requires general anaesthesia and is invasive and expensive, limiting its routine use.4 Selection of women at highest risk for tubal pathology could be an option to restrict the use of laparoscopy and medical history taking might be a tool to do so.
Guidelines5,6 recommend hysterosalpingography (HSG) to screen for tubal pathology in women who are not known to have co-morbidities (such as pelvic inflammatory disease [PID], previous ectopic pregnancy and endometriosis). Laparoscopy with dye (DLS) is recommended in women who are thought to have co-morbities in their medical history. However, aggregated evidence to underpin this recommendation is currently lacking.
The aims of the present study were to systematically identify possible risk indicators in the medical history for tubal pathology and to quantify their associations by means of meta-analysis.
Materials and methods
Identification of studies
We performed this systematic review and meta-analysis according to standard methodology.7–9 We performed a computerised search in MEDLINE (from 1966 to May 2007) and EMBASE (from 1960 to January 2007) to identify articles reporting on risk indicators for tubal pathology. Language restrictions were not applied. Reference lists of included studies were scrutinised for articles not identified by the electronic searches. For pragmatic reasons, we only included the specific risk indicators: sexually transmitted diseases (STDs), PID and appendicitis as search terms. Keywords used were anamn*, medical history taking (MeSH), medical records (MeSH), history (MeSH), medical record* (text word), infertility evaluation (text word), anamneses (text word), anamnesis (text word), sexually transmitted disease (text word), pelvic inflammatory disease (text word), appendicitis (text word), subfertility (text word), infertility (MeSH), infertility (text word), adnexal diseases (MeSH), genital diseases, female (MeSH), infertility, female (MeSH), fallopian tube diseases (MeSH), fallopian tube patency test (MeSH), fallopian tubes (MeSH), tubal pathology (text word), tubal occlusion (text word), tubal blockage (text word), tubal disease (text word), tubal subfertility (text word), hysterosalpingography (MeSH), laparoscopy (MeSH), hysterosalpingogram (text word) and HSG (text word).
Selection of studies and data extraction
Two independent reviewers (F.Y.L. and S.F.P.J.C.) screened the electronic searches for eligible articles by reading the title and abstract. Cohort studies, case–control and cross-sectional studies were eligible for inclusion. In case–control studies, cases had to be women with tubal pathology, proven with HSG or DLS. Controls were women without tubal pathology proven on either HSG or DLS or assumed because women were pregnant or had delivered a liveborn child at the same period of time. Cohort studies were defined as studies reporting on medical history taking in a subfertile population, in which presence of tubal pathology was subsequently diagnosed by testing with HSG, DLS or a combination of both. All identified articles were retrieved in full and assessed by two reviewers for eligibility. Any disagreements were resolved by consensus. In case of persistent disagreement, the judgement of a third reviewer (B.W.J.M.) was decisive.
All studies were scored by a combination of three independent reviewers (F.Y.L., H.R.V. and S.F.P.J.C.) on methodological quality by a data extraction form based on the QUADAS tool.10 We scored each study on the following characteristics: (1) study design (cohort study, case–control or cross-sectional study), (2) sampling (consecutive women or other), (3) data collection (prospective or retrospective), (4) blinding (present or absent), (5) verification bias (partial or differential verification bias), (6) reported definition of tubal pathology and (7) missing results or withdrawals.10
A two-by-two table had to be available from the articles, either directly or retractable from the data supplied, in which items from medical history were classified against the presence or absence of tubal pathology. We studied the following risk indicators: pelvic surgery, appendicectomy, normal appendicectomy, appendicitis, complicated appendicitis, PID, STD, chlamydia, intrauterine contraceptive device (IUCD), induced abortion, miscarriage, history of endometriosis, abnormal vaginal discharge, dysmenorrhoea, chronic abdominal pain, ectopic pregnancy, number of sexual partners in the past and dyspareunia.
Tubal pathology was defined as any of the following: unilateral or bilateral tubal occlusion (proximal or distal), tubo-ovarian or pelvic adhesions.
From the data in each study, two-by-two tables were constructed in which items from the medical history were related to the presence or absence of tubal pathology. Odds ratios and their 95% CI were calculated from these tables. For each risk indicator, homogeneity was tested by means of the Breslow–Day test. P values of <0.05 were considered to indicate statistically significant heterogeneity. In case homogeneity could not be rejected, a common odds ratio with a 95% CI was calculated for each exposure by means of the Mantel–Haenszel method. If homogeneity was rejected, a random effect model was used to calculate the common odds ratio.11 Statistical analysis was performed using SAS software (SAS Institute, Inc., Cary, NC, USA). RevMan 4.2 software (Cochrane Collaboration) was used to construct forest plots and visualise the data. An increase in the odds for risk of tubal pathology was displayed graphically to the right of the centreline; a decrease in the odds was displayed to the left of the centreline.
Number of retrieved studies
The search in MEDLINE and EMBASE detected 1675 studies, of which 132 articles were selected for further reading based on the titles and abstracts. Of these 132 studies, 100 were excluded after reading the papers in full. No additional papers were identified in cross-references of the selected articles that fulfilled the inclusion criteria, leaving 32 articles available for analysis (Figure 1). Papers in languages other than English were translated. However, we were not able to appraise one study in Bulgarian.12 The English abstract of this study described a cohort of 114 women with primary tubal infertility and the prevalence of previous appendicectomy, but it was unclear whether or not a control group was used. For this reason, the study was excluded from the analysis.
Characteristics of the included studies
Of the 32 studies, 18 studies used a case–control design13–30 and 14 studies were cohort studies.31–44
The exact definition of tubal pathology was not described in nine studies.14,15,18,22,24,27–29,31 In the other 23 articles, the definition was mentioned according to predefined criteria. In 18 studies, both HSG and DLS were used as reference tests.13–15,17,18,20,23–25,27,32,33,35–39,43 In five studies, a HSG was used to identify tubal pathology.16,22,30,40,44 Eight studies reported laparoscopy as the reference standard.21,28,29,31,34,41,42,44 One study did not report any reference test.19 As tubal pathology was clearly defined as ‘tubal occlusion and/or peritoneal adnexal adhesions’, we assume that DLS was performed as a reference test in this study.
In 15 studies, a reference test was not performed if a woman was not at risk for tubal pathology, leading to partial verification bias.13–25,30,39 This was mainly performed in case–control studies where pregnant women or women who delivered a liveborn child were used as control group. Ten studies showed differential verification bias as they used a HSG if women did not report a positive history of assumed risk factors, whereas DLS was used in case of a positive medical history 14,15,17,18,20,23–25,36,39 In the other 22 studies, the same reference test was performed in all women to identify tubal pathology.
The reference test was blinded in five studies.15,23,26,41,43 Six studies reported that the interpretation of the tubal patency test was not blinded,14,21,27,28,30,40 whereas in 21 studies, blinding was not reported. The overall methodological quality of included articles is shown in Figure 2. Characteristics of the included studies are outlined in Tables 1 and 2.
Table 1. Characteristics of included cohort studies
Prevalence of tubal pathology (%)
NL, Netherlands; NP, number of patients; OCP, oral contraception pill; CAT, Chlamydia antibody titer; IVF, in vitro fertilisation.
Women with proven tubal patency and primigravid women
Surgery-related risk indicators
Three studies reported on a history of complicated appendicitis. The pooled results of two case–control studies (OR 3.3, 95% CI 1.8–6.3; Figure 3) and one cohort study (OR 7.2, 95% CI 2.2–23.8; Figure 4) demonstrated a strong association with tubal pathology. Eight studies (four cohort and four case–control) examined the association between a history of uncomplicated or complicated appendicectomy and tubal pathology. Overall results indicated a mildly increased risk of tubal disease with a common odds ratio of 2.0 (95% CI 1.5–2.6) in case–control studies, but no increased risk in the cohort studies (OR 0.9, 95% CI 0.7–1.3). Seven studies (five cohort and two case–control) assessed a history of pelvic surgery. The data of the case–control studies showed significant heterogeneity. A history of pelvic surgery increased the likelihood of tubal pathology, with an odds ratio of 3.6 (95% CI 1.4–9.0; Figure 5) in the pooled cohort studies but not in the case–control studies (OR 1.5, 95% CI 0.2–11.6; Figure 6). A positive history of induced abortion (three case–control studies) showed a statistically significant association with tubal pathology, with an odds ratio of 1.7 (95% CI 1.3–2.1; Figure 7).
Infection-related risk indicators
The relation between previous PID and the probability of tubal pathology was examined extensively, with 9 cohort and 12 case–control studies reporting. For previous PID, a strong association with tubal disease was found in the pooled cohort studies (OR 3.2, 95% CI 1.6–6.6; Figure 8) and case–control studies (OR 5.5, 95% CI 2.7–11.0; Figure 9). Five studies (three cohort and two case–control) reported a history of STDs in general but did not specify the underlying pathogens. Analysis of these data showed a nonsignificant association in the cohort studies (OR 0.6, 95% CI 0.4–1.3) but a significant association in the case–control studies (OR 11, 95% CI 4.3–33.3). Self-reported previous chlamydia infection (one cohort and one case–control study) was not found to be statistically significantly associated with tubal pathology (Table 3).
Table 3. Risk indicators for tubal pathology in cohort and case–control studies
Risk indicator (study design)
Homogeneity was rejected; therefore, data were pooled with a random effects model.
A history of endometriosis (two case–control studies) strongly increased the risk for tubal pathology (OR 5.9, 95% CI 3.2–10.8; Figure 10). The previous use of an IUCD was examined in 16 studies (5 cohort and 11 case–control). Homogeneity was rejected in the case–control studies, but pooling of the data using a random effects model still showed a mild association with tubal pathology, which was statistically significant (OR 2.0, 95% CI 1.6–2.6). The same applied to cohort studies (OR 1.9, 95% CI 1.3–2.8). A history of abnormal vaginal discharge (three cohort studies) was not associated with tubal pathology (OR 1.3, 95% CI 0.9–2.0). Women recalling a history of chronic or severe pelvic pain (two cohort studies) showed an increased risk for tubal pathology (OR 1.9, 95% CI 1.0–3.6). A total of seven studies (two cohort and five case–control) examined the effect of a previous ectopic pregnancy. With an odds ratio of 16.0 (95% CI 12.5–20.4; Figure 11) for case–control studies, ectopic pregnancy showed a strong association with tubal disease, which was not found in the cohort studies (OR 3.7, 95% CI 0.3–47.6; Figure 12).
We did find studies reporting an increased risk of tubal pathology when a woman recalled a history of more than one sexual partner in the past.18,24 However, these numbers were too small to include in the statistical analysis. We were not able to extract two-by-two tables for uncomplicated appendicectomy and dysmenorrhoea. For one study reporting on the association between induced abortion and tubal pathology, it was not possible to extract data as this study combined miscarriage and induced abortion.35 Only one study reported deep dyspareunia as a risk indicator, but an association with tubal pathology was not found.41
The results of all analyses specified by risk indicator and design, including whether data were pooled with fixed or random effects models, are summarised in Table 3.
This meta-analysis summarises the available evidence on the value of medical history as risk indicator for tubal pathology. We found a history of PID and complicated appendicitis to be strong risk indicators for tubal pathology in both cohort and case–control studies. Strong associations between tuboperitoneal pathology and a history of endometriosis, ectopic pregnancy and STD were found in case–control studies, whereas a strong association with tuboperitoneal pathology and a history of pelvic surgery was found in cohort studies. For the indicators appendicectomy in general, chronic pelvic pain, previous use of an IUCD and a history of induced abortion, the association was less strong, while for a history of previous chlamydial infection and vaginal discharge, only nonsignificant associations were found.
This study is the first systematic comparison of the relation between certain aspects of the medical history and the probability of tubal pathology in a subfertile woman. Although previous narrative reviews have already suggested these factors to be strongly related to tubal disease,45 our study provides the evidence-based support to guidelines that are already recommending diagnostic laparoscopy to women with known co-morbidities, such as PID, previous ectopic pregnancy and endometriosis.5,6
In this meta-analysis, we used odds ratios to quantify clinical factors that indicate a risk factor for tuboperitoneal pathology. Although likelihood ratios are considered an important tool in the interpretation of diagnostic tests in clinical practice, there is no consensus on whether they should be used in meta-analysis. Strong arguments against the pooling of likelihood ratios have been raised, which are in our opinion valid.46 Moreover, we believe that odds ratios are useful in clinical practice because the odds ratio is a good estimate of risk ratio when the disease/event is rare.47
Although the present study was performed according to established methodology, it has limitations. First, we were only able to perform univariable analyses, as the construction of two-by-two tables per risk indicator does not allow for correcting their mutual dependence. Multivariable analyses of larger studies or analysis of individual patient data (IPD) may well find less strong associations than we found.48 For instance, a recent study that described a multivariable model to estimate the probability of tubal pathology in subfertile couples showed that female age and duration of subfertility were also strongly correlated with tubal disease.49
A second limitation of our study is the fact that we defined tubal pathology as any abnormality disturbing the integrity of one or both tubes, even though it is known that only bilateral tubal occlusion is strongly correlated with infertility.50 We made this decision for pragmatic reasons. Due to a large variability in the definition used by the original studies, numerous subanalyses according to the definition of tubal pathology used would otherwise have to be performed.
A third limitation is that we included five studies in which only HSG was used to verify tubal pathology (one cohort40 and four case–control16,22,26,30). Inclusion of these studies could have led to an overestimation of the summary estimates. However, in a subanalysis excluding these studies, the summary estimates of the value of historical risk factors did not change, for example the odds ratio for ectopic pregnancy excluding these five studies was 16.4 (95% CI 9.6–27.8) for case–control studies compared with an odds ratio of 16, whereas the odds ratio remained 3.7 for the cohort studies (data not shown).
A fourth limitation concerns the design of the studies. In this meta-analysis, 18 case–control studies were included. The controls in these articles were mainly selected on the basis of pregnancy or delivery of a liveborn child, and these women were classified as having no tubal pathology. One should, however, be careful with such an assumption as conception is no guarantee for absence of tubal pathology. Only bilateral tubal occlusion can almost certainly be ruled out, but it is still possible that women who conceived spontaneously have unilateral tubal abnormalities. It has been shown that studies comparing a diseased population with healthy controls in general overestimate the value of the index parameter under study.51,52 Also, pregnant controls might be less likely to remember or reveal specific items from their history compared with subfertile women who are under evaluation in a fertility clinic. This potential recall bias is a common problem with case–control studies and will probably lead to an overestimation of associations. These factors explain the higher odds values (with the exception of complicated appendicitis and pelvic surgery) in the case–control studies. Most evidently, this is seen for the risk indicators ectopic pregnancy and STD, which showed a discrepancy with the cohort studies. Contrary to the studies that addressed STD, the number of women in the studies who addressed ectopic pregnancy was large, which makes an association with tuboperitoneal pathology more likely. For endometriosis, a strong association with tuboperitoneal pathology was found, but only two studies with less than 1000 women were included, and no cohort studies were found. Therefore, these results should be interpreted with caution. Similarly, risk indicators with odds ratios below 2.0, albeit statistically significant, that is the previous use of an IUCD and chronic pelvic pain, should be interpreted with caution.
Meta-regression and sensitivity analysis could give insight into how certain study characteristics are related to summary estimates of studied risk factors. However, the availability of data that could be extracted from the original papers for performing a meta-regression analysis was limited, which makes such an analysis complicated and can potentially lead to biased results. Although we used a large number of studies, only few studies were available for specific risk items. Given such low numbers of studies, the potential for robust conclusions from meta-regression analysis is very limited.53 It would even be better to use IPD54 as IPD meta-analyses are generally considered to be the gold standard to explore subgroups. However, getting IPD from all the observational studies we used is practically impossible.
A fifth limitation is that we were not able to extract from the articles which type of pelvic surgery was performed. A history of pelvic surgery for severe endometriosis will have a different effect on adhesion formation than coagulation for mild endometriosis or puncture of an ovarian cyst and as a consequence will influence the odds ratio for risk of tuboperitoneal pathology.
In conclusion, this review shows that women with a positive history of PID, complicated appendicitis, pelvic surgery, ectopic pregnancy and endometriosis are at increased risk for tubal pathology. Women with these risk factors in their medical history should be offered a diagnostic laparoscopy with possible surgical procedures by competent surgeons early in the fertility work-up. To strengthen the evidence outlined in this meta-analysis, large cohort studies addressing all possible risk indicators for tubal pathology and their mutual (in)dependence are welcomed.
Disclosure of interest
None of the authors report any financial interests.
Contribution of authorship
F.Y.L., H.R.V., S.F.P.J.C. were responsible for the data collection and scoring, analysis and writing of the paper; B.W.J.M. was responsible for the analysis and writing of the paper and H.J.v.D. and F.v.d.V. were responsible for the writing of the paper.
Details of ethics approval
Ethics approval was not obtained since this is a systematic review of the literature.
This systematic review was performed without funding.