Risk of reproductive complications following chlamydial testing: A population-based retrospective cohort study in Denmark

,

Consensus estimates of the risk of complications following chlamydia are necessary for clinical consultations and chlamydia control policy decisions.Since the 1970s, Denmark has had a comprehensive electronic population register that can be linked to electronic health-care records and a well established chlamydia control programme with voluntary reporting of cases since 1984.We used a national cohort to improve estimates of the contemporary association between diagnosed chlamydia and pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility in women in Denmark, including never-tested women and after repeat diagnosed infections.

Research in context
Evidence before this study We searched PubMed from Jan 1, 2008, to April 1, 2016, for studies that report the prospective risk of reproductive complications in low-risk women following chlamydia infection compared with a chlamydia-negative control group.We used the search term "Chlamydia trachomatis" combined with "pelvic infl ammatory disease", "salpingitis", "endometritis", "infertility", or "ectopic pregnancy" and identifi ed relevant studies restricted to the English language.We added these results to relevant studies identifi ed by a 2010 systematic review and a study from the reference list of an included publication.The prospective risk of pelvic infl ammatory disease following untreated genital chlamydia infection, treated infection, and a negative chlamydia test is presented in a single randomised controlled trial of 2529 women at low risk (9•5%, 1•6%, and 1•3%, respectively).However, the sample size is too small to detect a diff erence between groups in this secondary analysis.No prospective studies of the risk of ectopic pregnancy or infertility following untreated chlamydia infection were identifi ed.Five retrospective cohort studies compare the lifetime risk of reproductive complications by chlamydia exposure in the general population.These studies show a 30-70% increased risk of pelvic infl ammatory disease in women who test positive for chlamydia compared with women who test negative.The three studies of ectopic pregnancy have not reached consensus and the single study of infertility uses data from before the widespread introduction of highly sensitive diagnostic tests.There is limited information on the role of repeat infection in reproductive complications.

Added value of this study
This cohort is considerably larger than the examples in the published literature and it is representative of the population of a country.It uses contemporary chlamydia test methods and the never-tested cohort has a straightforward defi nition: women who do not have a chlamydia test during the cohort period.The size of the cohort has allowed us to report the contribution of a repeat diagnosed chlamydia infection to all three reproductive complications (pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility) for the fi rst time in a single population.

Implications of all available evidence
Women with at least one diagnosed chlamydia infection have a 30% increased risk of pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility throughout their reproductive lifetime.The duration of the elevated risk of pelvic infl ammatory disease is longer than the estimated 1 year duration of untreated chlamydia infection.Therefore, in the case of pelvic infl ammatory disease, chlamydia is likely to act as a proxy for other risk factors of adverse reproductive health (eg, sexual behaviour, repeat chlamydia infection) or a diagnostic bias.The absolute impact of diagnosed chlamydia infections on reproductive complications might be small.We observed a small diff erence in the lifetime cumulative incidence of pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility between women who tested positive compared with women who tested negative.Women tested for chlamydia, who are negative, have an increased risk of complications compared with women who are never-tested.This fi nding might indicate that widespread testing in Denmark is targeting the at-risk population.The risk of pelvic infl ammatory disease following chlamydia infection is increased by a further 20% following a repeat diagnosed infection but we found no evidence that repeat infections further increase the risk of ectopic pregnancy or tubal factor infertility.Control interventions need to prevent fi rst infections to reduce the lifetime risk of pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility, and repeat infections to further reduce the risk of pelvic infl ammatory disease.

Dataset and data selection
We used a subset of the national Danish Chlamydia Study (appendix).The master dataset contains all residents of Denmark (including Greenland) who had a positive chlamydia test recorded by a public health microbiology laboratory from Jan 1, 1992 For this analysis, the master dataset was limited to women resident in Denmark (excluding Greenland) who entered the dataset after Jan 1, 1995, when aged 15-44 years, and who were in a fi ve-person set (one chlamydia-positive and four chlamydia-negative or never tested) who met the study chlamydia exposure defi nitions.This start date was chosen because chlamydia notifi cation became compulsory in Denmark in 1994 and in 1994-95 the DNPR changed from International Classifi cation of Disease (ICD)-9 to ICD-10 coding and added outpatient and emergency department presentations.Individuals with a positive chlamydia test entered the cohort on the date of their fi rst positive test during the study period.Their four matched individuals were assigned the same entry date.
During the study period, chlamydia diagnostic methods varied between laboratories and the change from nonnucleic acid methods to nucleic acid amplifi cation tests (NAATs) occurred predominantly in 1999-2000.Records with an invalid or missing date, site, or result were removed along with tests not done on genital, rectal, or urinary samples.Where multiple eligible tests were recorded on the same date, positive results were retained in preference to negative results.Chlamydia tests were then restricted to one per 30-day period.The exposure in this study is chlamydia status and women were defi ned as chlamydia-positive if their fi rst test in the study cohort was positive, negative if they had at least one negative test, and no positive tests during the follow-up period or never-tested if they did not have a recorded test during follow-up.

Statistical analysis
We calculated the crude cumulative incidence of each complication by the end of follow-up (defi ned as the number of women diagnosed with the complication by the end of follow-up divided by the number of women at baseline).We constructed Kaplan-Meier survival plots of time to each complication by chlamydia exposure status and used Cox proportional-hazards regression to explore the association between chlamydia exposure status and each complication separately, adjusted for age at cohort entry, and year of entry.Negative women formed the baseline group and time since entry was the time axis.
We then divided the follow-up period of positive women into the interval following their fi rst positive test and the interval following their second positive test and estimated the association between this repeat diagnosed

Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.The corresponding author had full access to all the data in the study and had fi nal responsibility for the decision to submit for publication.
Kaplan-Meier survival curves are presented in fi gure 2. The adjusted hazard of each outcome was 31-50% higher in women who tested positive compared with women who tested negative (pelvic infl ammatory disease, AHR

Discussion
In this cohort of women from a contemporary setting with widespread chlamydia control who were tested for chlamydia, a positive chlamydia test increased the risk of pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility by at least 30%.This risk remained throughout 17 years of follow-up, which is considerably longer than the estimated 1 year duration of chlamydia infection. 17Therefore, the (treated) index infection might, in part, be a marker for other risk factors for pelvic infl ammatory disease.The impact of the index positive chlamydia test on reproductive health was small in absolute terms because the diff erence between the lifetime risks (cumulative incidence) of each outcome between women who tested positive and negative were small (0•1-0•6%).Finally, repeat diagnoses were a risk factor for pelvic infl ammatory disease but not ectopic pregnancy or tubal factor infertility.These fi ndings suggest that the optimal strategy for improving women's long-term reproductive health is to provide interventions that prevent both fi rst and repeat infections.
This nationally representative cohort of over 500 000 women is an order of magnitude larger than previously published cohorts (Uppsala, 43 715; 5 Sør-Trøndelag, 24 947; 13 North Jutland, 13 693; 16 Manitoba, 72 883 14 ).The size of the cohort together with the high volume of repeat testing allowed us to explore the risk of complications in never-tested women and following repeat diagnosed infections.There was complete ascertainment of chlamydia tests and hospital presentations for pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility in Denmark throughout the study.
4][15] We used chlamydia test records to assign lifetime chlamydia status but undiagnosed infections are likely to have occurred. 5Misclassifi cation of outcome status is likely to be most marked for pelvic infl ammatory disease in which the diagnostic criteria Age (years) *Adjusted for chlamydia exposure, age at entry to cohort, and year of entry to cohort.

Table 2: Adjusted hazard ratios (HRs) of pelvic infl ammatory disease, ectopic pregnancy, and tubal factor infertility following a repeat chlamydia diagnosis
are non-specifi c, there is no gold-standard non-invasive diagnostic test, 18 and a diagnostic bias towards pelvic infl ammatory disease in women with a past history of chlamydia. 2This proposed diagnostic bias would increase the association between chlamydia and pelvic infl am matory disease if it aff ected hospital episodes of health care and could contribute to the observed increased risk of pelvic infl ammatory disease after a repeat diagnosed infection.Primary care data were not available in this study.This incomplete ascertainment of complications is likely to have had the greatest impact on the analysis of pelvic infl ammatory disease where there has been a steady increase in management in the outpatient or community setting. 19,20If there is no association between the cause of pelvic infl ammatory disease and clinical presentation then the absence of cases managed solely in primary care will reduce the cumulative incidence but not change the relative risk across exposure groups.However, pelvic infl ammatory disease due to chlamydia might have a milder clinical presentation than other causes. 21If pelvic infl ammatory disease in chlamydia-negative women is more likely to be recorded in the DNPR this would underestimate the association between chlamydia and pelvic infl ammatory disease.This is unlikely to be relevant in this study because Neisseria gonorrhoea, a major cause of severe pelvic infl ammatory disease, is rare in Denmark (80% of the 490 patients with gonorrhoea in Denmark in 2011 were men). 22here is uncertainty in the validity of our ICD defi nitions to accurately represent a relevant diagnosis in the clinical record.This uncertainty has been well documented for pelvic infl ammatory disease [23][24][25] but might be less pronounced for ectopic pregnancy and tubal factor infertility because they have more specifi c diagnostic criteria.To improve validity, we restricted the cohort to women who entered after 1995 when ICD-10 coding was introduced, and emergency department and outpatient data were added to the DNPR.This period also coincided with the shift from voluntary to compulsory notifi cation of chlamydia in 1994.
The structure of the study cohort was a trade-off between the completeness of exposure and outcome measures and the duration of follow-up.By starting on Jan 1, 1995, it was possible for the youngest women (at study entry) to reach the mean age of fi rst childbirth in Denmark by the end of follow-up, although they did not reach the end of their reproductive lifetime.Exposure and outcome ascertainment was also comprehensive.The validity of a chlamydia test result will have changed following the introduction of NAATs in 1999-2000.NAATs have a high sensitivity compared with enzyme immunoassays and direct immuno fl uorescence tests.However, the pathogenicity of detected infections might vary with test type and be lowest for NAATs.In support of this, infection resolution rates have been shown to be higher in people diagnosed by the use of NAATs compared with enzyme immun oassay or culture. 26herefore, the association between diagnosed infections and reproductive complications might alter over time as the predominant diagnostic test changes.
To control for this, we adjusted the analysis for year of entry to the cohort.We were unable to control for many potential confounders, including sexual behaviour, reproductive intent, and sexual network, due to the absence of suitable data.We chose to include all women in our analysis of ectopic pregnancy rather than restricting the cohort to women who had a recorded pregnancy.This might explain the lower AHR of ectopic pregnancy following diagnosed chlamydia observed in women aged 30-44 years at cohort entry compared with women aged 25-29 years.Specifi cally, the rate of conception is likely to be lower in the older age group, and therefore if the true risk of ectopic pregnancy as the outcome of pregnancy is the same across age groups, the AHR would be lower in older women if the incidence of pregnancy was lower.
We used the Cox proportional hazards model in this study.4][15][16] The assumption of proportional hazards was violated in our analysis but this was not unexpected given the structure of our dataset: events at baseline are likely to become poorer predictors of events that occur after increasing time intervals.We explored several alternative methods for analysing the data; 27 however, we felt it was appropriate to report the fi ndings of this Cox regression to enable direct comparison of the risk of complications in Denmark to other settings.
Our fi ndings of an increased risk of pelvic infl ammatory disease in chlamydia-positive women compared with chlamydia-negative women (AHR 1•50, 95% CI  13 and Manitoba, 1•55 [1•43-1•70]) 14 despite diff erences in cohort structure, defi nition of chlamydia exposure, and pelvic infl ammatory disease ascertainment.This increased risk is present throughout the reproductive lifetime and therefore the index chlamydia test is likely, at least in part, to be a surrogate for other risk factors for pelvic infl ammatory disease, most likely undiagnosed sexually transmitted infections, including repeat chlamydia infections, which might be mediated through sexual or health-care seeking behaviour.
There is confl icting evidence in the literature about the relation between a chlamydia diagnosis and ectopic pregnancy.We found that a positive test increased the risk of ectopic pregnancy (as the outcome of any pregnancy) by 31% (AHR 1•31, 95% CI 1•25-1•38) compared with women with a negative test.This fi nding is consistent with the other study that included all women (Uppsala, AHR 1•26, 0•94-1•67). 5Two further studies limited their analysis to the fi rst pregnancy in women with a pregnancy during follow-up and found a 45% decreased risk (Northern Jutland, AHR 0•55, 95% CI 0•31-3•59) 16 and 7% increased risk (Sør-Trøndelag, 1•07, 1•01-1•12). 15hlamydia is thought to be a risk factor for tubal factor infertility rather than infertility more widely, and this is the only retrospective cohort study that has restricted the defi nition of infertility to tubal factor infertility.As tubal factor infertility is a subset of the overall infertility diagnoses it is expected that the cumulative incidence of tubal factor infertility would be lower than the cumulative incidence of infertility and we report a ten times lower cumulative incidence of tubal factor infertility compared with that of infertility in Uppsala.It is diffi cult to explain the similar increased HR for tubal factor infertility and infertility in chlamydia-positive compared with chlamydia-negative women in the two cohorts (Denmark, AHR 1•37 [95% CI 1•24-1•52]; Uppsala, 1•31 [1•09-1•57]). 5n this cohort, women with a diagnosed (and presumably treated) chlamydia infection had a low absolute risk of complications (cumulative incidence in positive women: pelvic infl ammatory disease, 3•1%; ectopic pregnancy, 2•2%; tubal factor infertility, 0•6%) that was only slightly higher than that observed in women with a negative test (0•1-0•6% higher in positive than negative women).This information might provide reassurance to women who have a chlamydia test.However, this diff erence in risk is equivalent to 9•1-19•3% of the recorded complications in positive women.Therefore, chlamydia control interventions that prevent fi rst infections and retain women in the lifetime-negative cohort could have an important eff ect on reproductive health.
In Denmark, widespread chlamydia testing appears to have been eff ectively targeted to women at risk of reproductive complications because the observed risks in never-tested women were extremely low.Nonetheless, some never-tested women did experience complications and might have missed out on appropriate health care.We observed a much lower risk of all complications in this group (compared with negative women) than the Uppsala study (eg, pelvic infl ammatory disease: AHR 0•32 [95% CI 0•30-0•34] in Denmark and 0•72 [0•63-0•82] in Uppsala.) 5 This fi nding might refl ect diff erences in the defi nition of the never-tested cohorts (lifetime never-tested in Demark compared with the interval before the fi rst test in Uppsala) or an increased uptake of testing in Denmark due to the more recent time period.
This study adds substantially to the evidence that a repeat diagnosis of chlamydia further increases the risk of pelvic infl ammatory disease 14,28 and we observed a very similar increased risk following a second diagnosed infection (11-31%) to the cohort in Manitoba (13-23%). 14nlike in Sør-Trøndelag we did not fi nd that a repeat diagnosed infection was a risk factor for ectopic pregnancy but this could refl ect diff erences in study defi nitions (including cohort and ectopic pregnancy) rather than biological diff erences. 15Our observation that a repeat diagnosed infection was only a risk factor for pelvic infl ammatory disease could be because each episode of chlamydia carries a discrete risk of pelvic infl ammatory disease, whereas damage to the oviducts, suffi cient to cause ectopic pregnancy and tubal factor infertility, might occur before the median duration of a fi rst infection.
A single diagnosed chlamydia infection increases the risk of all reproductive complications in women and a repeat diagnosis further increases the risk of pelvic infl ammatory disease.To improve women's long-term reproductive health and wellbeing, chlamydia control programmes must be designed to prevent both fi rst and repeat infections.To inform the design of widespread chlamydia control interventions, more information is required on the relative contribution of chlamydia control activity (ie, interventions that reduce the incidence of untreated infections), change in the incidence of gonorrhoea, and change in the pathogenicity of diagnosed chlamydia infections after the introduction of NAATs 29 on the risk of reproductive complications.

*
Adjusted for chlamydia exposure, age at entry to cohort, and year of entry to cohort.

Contributors
BD, KMET, MF, HWa, TB, and HWe contributed to study design, data interpretation, and writing.BD and KMET did the data analysis.MF, TB, HWe, SR, and the Danish Chlamydia Study Group designed the Danish Chlamydia Study and collected the dataset.MF cleaned the Danish Chlamydia Study dataset.MTM provided statistical advice and contributed to data interpretation and writing.Members of the Danish Chlamydia Study Group additionally contributed to writing.Declaration of interestsBD and HWa declare personal fees from the European Centre for Disease Prevention and Control (ECDC).KMET declares personal fees from Aquarius Population Health.MF, SR, and MTM declare no competing interests.TB declares personal fees from GlaxoSmithKline, Bristol Myers Squibb, Gilead, Symphogen, and Abbvie, and grants from Pfi zer.HWe declares grants from Roche Molecular, grants and personal fees from Hologic, and personal fees from Novo Nordic.KMET was funded by a National Institute for Health Research (NIHR) Post-Doctoral Fellowship 2009-02-055 (January, 2009, to December, 2014).This Article presents independent research funded by the National Institute for Health Research (NIHR).The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.(Note only KMET received direct funding during the preparation of this paper.)The Danish Chlamydia Study Group Berit S Andersen (Afdeling for Folkeundersøgelser, Region Midtjylland, c/o Randers Hospital), Jette Bangsborg (Department of Clinical Microbiology, Herlev Hospital, University of Copenhagen), Claus Bohn Christiansen (Department of Clinical Microbiology, Rigshospitalet University of Copenhagen), Ram B Dessau (Department of Clinical Microbiology, Slagelse Hospital), Esad Dzajic (Department of Clinical Microbiology, Esbjerg Hospital), Steen Hoff mann (Neisseria and Streptococcus Reference Laboratory, Department of Microbiology and Infection Control, Division of Diagnostics and Infection Control, Statens Serum Institut), Poul Kjaeldgaard (Department of Clinical Microbiology, Sønderborg Hospital), Jørgen Skov Jensen (Department of Microbiology and Infection Control, Division of Diagnostics and Infection Control, Statens Serum Institut), Tove E Jensen (Department of Clinical