Article Text
Abstract
Objective: To summarise evidence on the attributable risk of infertility after chlamydial infection in women.
Methods: Twelve databases were searched, limited to peer-reviewed literature published from January 1970 to September 2007. Conference abstracts and reference lists from reviews published since 2000 and from key articles were hand-searched. Studies were selected for review if they met the following criteria: (1) the study population comprised women of child-bearing age (defined as 15–45 years) and incorporated a comparison group of women documented as “chlamydia negative”; (2) the study outcomes included either infertility or successful pregnancy; and (3) the study design was one of the following: cohort, randomised controlled trial, “before and after” study, screening trial and systematic review. Studies were excluded if they described genital infections that either did not include Chlamydia trachomatis or described genital chlamydial co-infection, in which no data were available for C trachomatis infection alone.
Results: 3349 studies were identified by the search. One study satisfied the inclusion criteria, a longitudinal investigation measuring pregnancy rates in adolescent women with and without current chlamydial infection at baseline. That study reported no significant difference in subsequent pregnancy rates; however, it had serious methodological limitations, which restricted its conclusions.
Conclusions: This systematic review demonstrates the absence of valid evidence on the attributable risk of post-infective tubal factor infertility after genital chlamydial infection. The findings contribute empirical data to the growing debate surrounding previous assumptions about the natural history of chlamydial infection in women.
Statistics from Altmetric.com
Concerns about the public health impact of genital chlamydial infection have generated considerable policy interest in the UK; a National Chlamydia Screening Programme is currently being implemented throughout England and national guidelines on chlamydial infection have been widely adopted in Scotland.1–3 Proactive targeted screening for chlamydial infection has been justified by four attributes of chlamydial infection: its high general population prevalence (recently estimated at 3% in UK residents aged under 25 years);4 its substantial transmission potential;5 the recent development of simple, non-invasive tests;2 5 and its potential for acute and chronic morbidity.5 6
In this last respect, however, the evidence base appears weak; although numerous case–control studies have reported an association between serological evidence of previous chlamydial infection and tubal infertility, this evidence is of limited value for two reasons. First, case–control studies have to rely on serological methods for the ascertainment of previous genital Chlamydia trachomatis infection; serological methods (before the advent of peptide-based, species-specific assays) are universally acknowledged to exhibit poor validity and reliability for this purpose.7 Second, case–control studies generate ratio measures rather than absolute measures of effect. Although ratio measures can demonstrate an association between chamydial infection and infertility, they cannot directly quantify the excess risk of infertility attributable to chlamydial infection, which is the crucial information required to counsel patients and, at a population level, to estimate the proportion of infertility cases that might be averted by chlamydia screening programmes. In addition, control participants are often selected from populations (commonly pregnant women), which generally underestimate infertility risk in the unexposed group and thus generate exaggerated estimates of the effect of chlamydial infection on infertility.
Quantifying excess (or attributable) risk is fundamentally important at many levels; individual women diagnosed with apparently uncomplicated chlamydial infection frequently request prognostic information on their subsequent risk of infertility and policy makers need robust cost-effectiveness evidence to underpin prevention strategies. One of the authors (LAW) recently reviewed patient information materials used by genitourinary medicine clinics across the United Kingdom; almost invariably, these contained no quantitative information on the risk of infertility, despite the importance of this issue to women diagnosed with chlamydial infection.8 We therefore conducted a systematic review of the literature to summarise existing evidence on the attributable risk of infertility after chlamydial infection.
METHODS
The review question
The review sought to answer the following question: “What is the attributable risk of infertility following one or more episodes of genital C trachomatis infection in women of reproductive age?” Attributable risk is defined as the proportion of women infected with C trachomatis who subsequently develop infertility attributable to the chlamydial infection.
Search terms, databases and search strategy
A search strategy was developed in collaboration with information scientists at the Centre for Reviews and Dissemination, University of York, UK (see table published online only).
A comprehensive list of databases was searched, with appropriate adaptation of the key words for each database search (table 1). The search was limited to published peer-reviewed papers from January 1970 to September 2007; the earlier time boundary was selected because it was judged to reflect the earliest timepoint from which diagnostic methods for chlamydial infection became available.
Although this was primarily a systematic review of peer-reviewed literature, conference abstracts were also searched, using the “Inside Conferences” and “Biosis Previews” electronic databases to September 2004. Reference lists from reviews published since 2000 and from key articles were hand-searched to identify any further relevant papers. We did not personally contact researchers to identify ongoing research studies or unpublished reports; non-English language papers were not included in the review, as our resources would not have supported their translation.
Inclusion and exclusion criteria
Studies were selected for review if they met the following criteria: (1) the study population comprised women of child-bearing age (defined as 15–45 years) and, in addition to women with genital chlamydial infection, also incorporated a comparison group of women documented as uninfected with C trachomatis; (2) the study outcomes included either infertility (defined as the failure of a couple to achieve pregnancy despite 12 months of regular unprotected sexual intercourse and/or referral to a specialist infertility service) or successful pregnancy; and (3) the study design was one of the following: cohort, randomised controlled trial, “before and after” study, screening trial (in which information on outcomes is given) and systematic review.
Conversely, studies were excluded if: (1) the study population comprised women under 15 years or older than 45 years; (2) they described genital infections that either did not include C trachomatis or described genital chlamydial co-infection, in which no data were available for C trachomatis infection alone; (3) they focused on pelvic inflammatory disease (PID) only and data on pregnancy and/or infertility outcomes were unavailable.
The first reviewer scanned the abstracts and titles using the criteria described above and categorised them into two libraries: “papers for further analysis” and “papers not relevant”. The former consisted of abstracts of original articles, reviews, and titles, for which no abstract was available, but which incorporated key words relevant to the study. To assess reliability, a second reviewer evaluated a 10% random sample of both libraries. After the resolution of any differences, both reviewers agreed on those abstracts or titles for which full papers were retrieved and the papers to be included in the review.
Study quality assessment
Papers selected for review were assessed for methodological quality by the two reviewers using the criteria described by Levine et al.9
RESULTS
Twelve databases were searched for relevant articles, generating a library of 3349 abstracts and titles for review. Fifty papers were identified as potentially relevant to the research question; these included nine primary research papers, 10 review articles published since 2000 and 31 papers retrieved from the “title only” group. Of these 50 selected papers, 49 failed to satisfy the inclusion criteria; 33 were review, narrative or news articles and not primary research studies, eight articles did not address the outcome of interest and eight did not employ an appropriate study design. One additional paper was identified from a review paper but this was also rejected on the basis of study design10 (fig 1).
The one study that satisfied the inclusion criteria was a longitudinal investigation, measuring pregnancy rates in adolescent women with and without evidence of current chlamydial infection at baseline; no statistically significant difference in pregnancy rates at the end of the follow-up period was found (table 2).11 The study’s methodological limitations, however, diminish the precision of its statistical estimates and the overall validity of its conclusions.
First, the study was conducted in Indianapolis, USA, between 1985 and 1990, at a time when the gold standard laboratory diagnostic test (tissue culture) was relatively insensitive. This is likely to have resulted in the incorrect misclassification of a proportion of chlamydia-infected women into the “uninfected” group at the study outset.
Second, the sample size afforded limited statistical power, compounded by substantial attrition of the study population, with only 104 (21%) of the original cohort of 496 women available for participation in the follow-up telephone survey. There was uncertainty about the mean follow-up time and whether this differed between the three study groups.
Third, the measurement and analysis of confounding variables (including age, socioeconomic circumstances, contraceptive use, frequency of intercourse and number of sexual partners) were inadequate.
Crucially, more than half of the overall survey sample reported current contraceptive use (oral contraception, condoms or both); however, pregnancy outcomes were neither stratified by contraceptive use nor analysed by multivariable methods to investigate the confounding effects of contraceptive use on pregnancy outcomes.
In summary, Katz et al11 found no association between a history of treated chlamydial infection and infertility in their study population, but the caveats described above seriously limit these conclusions. As no other eligible studies could be located in our systematic review, it was not possible to obtain valid evidence on the attributable risk of infertility after genital chlamydial infection in women.
DISCUSSION
Our systematic review demonstrates the absence of valid evidence on the attributable risk of infertility after genital chlamydial infection and an overall dearth of research on the natural history of chlamydial infection in women.
Key messages
This systematic review demonstrates an absence of evidence on the attributable risk of infertility after genital chlamydial infection and an overall lack of research on the natural history of chlamydial infection in women.
Chlamydia screening programmes are based on assumptions about the natural history of infection, in particular the likelihood of developing infertility. This evidence is derived almost exclusively from case–control studies, which do not allow an accurate estimation of post-chlamydial infertility risk.
This study, in combination with other recent publications, indicates that previous assumptions on progression rates to infertility may no longer be valid.
The proportion of PID and infertility preventable by chlamydia screening remains unknown.
Although an extensive search for evidence was systematically conducted, our study does have the following limitations: researchers in the field were not personally contacted about any unpublished research and non-English language papers were not included in the review. The authors, however, have sufficient awareness of current health services research on chlamydia testing policy to be satisfied that this boundary (selected for pragmatic reasons) was unlikely to have introduced bias.
To our knowledge, this is the first published systematic review of the evidence base on the attributable risk of infertility after genital chlamydial infection. Although our focus was to examine the role of genital chlamydial infection, we fully recognise the polymicrobial nature of PID and thus subsequent infertility. Accordingly, we defined our exclusion criterion to select only studies in which data were analysable on genital chlamydial infection, even in those with concurrent infections.
Genital chlamydial infection has been judged to fulfil the required criteria for the establishment of a screening programme.12 This judgement was, however, based on assumptions about the natural history of infection that are now being increasingly questioned.13–20 A brief summary of the principal arguments follows.
Data from two randomised controlled trials of screening versus no screening/normal care demonstrated a greater than 50% decrease in the incidence of PID.16 17 The studies have, however, since been criticised because of the considerable potential for both selection and measurement bias. Scholes et al16 in the mid-1990s analysed outcomes in only a small proportion of the women randomly selected and ascertainment of the main outcome measure (PID) was made by case note review (as opposed to laparoscopic diagnosis or the use of systematic, criterion-based clinical examination). In addition, the study did not evaluate an opportunistic approach to screening, which is the current strategy being advocated in the United Kingdom and elsewhere. In the study by Ostergaard et al17 in the late 1990s there was a high level of loss to follow-up of both the screening and the control group. This, combined with the possible underreporting of PID, creates considerable potential for bias and limits the conclusions of that screening trial.
Furthermore, in a one-year follow-up study performed in the late 1990s, Morre et al18 reported a 45% clearance rate of infection per year in women who had received no antibiotic treatment and none of the women developed clinical PID; however, unanswered questions regarding the persistence of chlamydial infection remain.
Van Valkengoed et al19 and Low et al20 have recently added further valuable contributions to the debate. Van Valkengoed et al19 generated modelled estimates of the risk of tubal factor infertility at 0.02%, substantially lower than previous estimates that ranged between 1.5 and 16%. The authors concluded that current assumptions overestimate the probability of complications and accordingly the potential health gain and cost savings associated with chlamydia screening. In 2006, Low et al20 published a keynote study using a retrospective population-based cohort of women in Uppsala, Sweden. That study did not fulfil the inclusion criteria of our review because it could not be categorised as a conventional cohort study,21 which necessitates the follow-up of a defined group of participants, classified as “exposed” or “unexposed” to chlamydial infection, and comparison of outcomes according to exposure status. Rather, the longitudinal study used record linkage methods, on data extracted from separate laboratory, hospital and population registers, to estimate infertility risk in women stratified by their previous chlamydia test status. Using this methodology, it was only possible to infer an exposure history in approximately half of the women in the study population and other causes of infertility could not be excluded. Accordingly, no measure of attributable risk could be determined. The cumulative incidence of infertility by the age of 35 years was estimated at 5.6% (4.7% to 6.7%) in women who ever tested positive for chlamydia, 6.7% (5.7% to 7.9%) in those with negative tests and 3.1% (2.8% to 3.3%) in those who were never screened. The authors concluded that the estimated risk of infertility attributable to chlamydial infection was likely to be considerably lower than previous published estimates.13 20 Although we excluded the study from our systematic review, we regard the powerful and well-conducted study, containing 709 000 woman-years of follow-up, as a major contribution to the evidence base on the reproductive sequelae of chlamydial infection.
There is thus a growing body of published data that suggests that the reduction in reproductive morbidities resulting from chlamydia screening programmes may be overestimated and that further research is required. Our systematic review advances the current debate about the potential health benefits gained from chlamydia screening by demonstrating empirically the lack of evidence on attributable risk.
This review has highlighted the absence of high quality evidence to answer questions commonly posed by the increasing numbers of women daily who receive a diagnosis of lower genital tract chlamydial infection within proactive opportunistic screening programmes. Current evidence is unable to provide women with any reliable estimates of the likelihood of serious reproductive health complications. At a wider population level, because the absolute risk of infertility after chlamydial infection is unknown, policy makers find themselves similarly ill-equipped to quantify the population impact of chlamydial infection on reproductive morbidity, and therefore face uncertainty about the proportion of PID and infertility that is preventable by chlamydia screening.
There are a number of possible approaches to address the major gap in the evidence base. The optimal epidemiological method, a cohort study design, for estimating the absolute risk of infertility attributable to chlamydial infection would now be considered unethical; it would require an extensive follow-up period in untreated women, careful measurement and control of potential confounding variables (including age, contraceptive usage, socioeconomic circumstances and sexual behaviour) and an accurate measurement of infertility/successful pregnancy outcomes. This necessitates an extremely large sample size and, accordingly, considerable resources. It would also be extremely difficult to gain ethical approval for a cohort study designed to monitor adverse outcomes among women with untreated chlamydial infection within a policy context that actively promotes testing and treatment.22 There is already good evidence that treating people with genital chlamydial infection is beneficial in relation to short-term morbidity;2 5 moreover, numerous case–control studies have demonstrated that women with infertility have a significantly greater chance of having been infected with chlamydia than those who are fertile.5 6 What is still not known, however, is the absolute magnitude of the increased risk of infertility that is attributable to C trachomatis infection.
If undertaking a cohort study is not feasible, the remaining options for the estimation of attributable risk would involve statistical modelling of the risk of progression to PID after chlamydial infection. With this approach, assumptions about the polymicrobial aetiology of PID and other dynamic transmission-related factors that may influence the outcome of chlamydial infection could be factored into the analysis. Alternatively, ecological analyses comparing nations or regions might offer useful insights.23 Such an analysis might involve comparing reproductive data from countries with differing approaches to screening policy. The resulting data will ensure that the future direction of chlamydia screening programmes in the United Kingdom and elsewhere is based on the best available evidence.
Acknowledgments
The authors gratefully acknowledge the expertise of information scientists at the Centre for Reviews and Dissemination, University of York, UK. PW’s research career award is funded by the Scottish Executive Health Department Chief Scientist Office.
REFERENCES
Footnotes
Contributions: AS, GH, MR, PW and DJG designed the study, obtained funding and provided comments on the review protocol. LAW performed the systematic review with AS in a supervisory role and as second reviewer. LAW, AS and DJG drafted the manuscript and all authors contributed to manuscript revisions. LAW and AS prepared the final paper.
Funding: The systematic review was funded by the West of Scotland Primary Care Sexual Health Research and Development Fund, project ID 03SH05. AS is partly funded by the Chief Scientist Office of the Scottish Executive Department of Health. PW is funded by a primary care research career award from the Scottish Executive Chief Scientist Office. The researchers were wholly independent from the funders of the study.
Competing interests: None.