Chlamydia trachomatis (chlamydia) is the most common sexually transmitted infection (STI) diagnosed in the UK.1 It can lead to serious reproductive complications including pelvic inflammatory disease (PID), ectopic pregnancy, and infertility.2 Chlamydia screening has been widely introduced as a strategy to reduce the incidence of these complications. However, it is difficult to obtain robust estimates of the cost-effectiveness of different screening approaches. Models are very sensitive to assumptions about the natural history of infection3,4 including the risk of progression to PID,4,5 and there is no consensus estimate of the risk of PID after a positive chlamydia test result.2,6
Two large retrospective population-based cohort studies have shown that women diagnosed as having chlamydia have a higher risk of PID compared with those with only negative test results,7,8 but there was variation in the observed proportion of women who developed PID after an infection: 5.6% (95% confidence interval [CI], 4.7–6.7) by the age of 35 years in Uppsala, Sweden,8 and 1.1% (95% CI, 0.8–1.4) after 15 years of follow-up in Sør-Trøndelag, Norway.7
There are several factors that may contribute to this uncertainty in PID risk, including the role of repeat infection9,10 and the contribution of other STIs.2 We will use an existing cohort of sex workers who underwent repeat testing for multiple STIs to estimate the risk of PID after a case of chlamydia.
MATERIALS AND METHODS
This study uses a prospective cohort of sex workers, the Praed Street Project (PSP) cohort, recruited between 1985 and 1993, and observed until 1993 (full details of the cohort have been published previously11,12). The cohort was recruited through the department of Genito-Urinary Medicine (GUM) at St Mary’s Hospital, London, and referrals from the walk-in sexual health clinic, a drop-in center run by PSP and outreach services to sex work venues and local streets. Eligible women defined themselves as “prostitutes” (sex workers) and had worked in the 3 months before their clinic visit.
During the study, women were able to visit the clinic at their discretion. Self-reported history of chlamydia and Neisseria gonorrhea (gonorrhea) was collected at the first clinic visit. Information about age, sexual behavior, and symptoms was also collected, and tests for chlamydia (direct immunofluorescence [DIF; Syva, Microtrak] and enzyme immunoassay [EIA; Chlamydiazyme, Abbott]), gonorrhea (gram stain with culture confirmation), Trichomoniasis vaginalis (trichomoniasis; direct microscopy), bacterial vaginosis (BV; clinical diagnosis based on modified Amsel criteria), and candidiasis (direct microscopy) were conducted, depending on clinical presentation and participant’s wishes, at this and subsequent clinic visits.
We applied additional inclusion criteria for this analysis: (1) attended clinic more than once during the study period, (2) had a complete set of key variables (clinic date, age, ≥1 chlamydia test result), and (3) not diagnosed with PID at the first visit. Duration of follow-up was calculated from the first visit to either first PID diagnosis or final visit before the end of the cohort in 1993.
A case of chlamydia was defined as a positive DIF test result or a positive EIA test result for visits where a DIF test was not performed. Cases diagnosed at the first clinic visit are labeled “prevalent,” and “incident” cases were diagnosed during subsequent visits. Prevalent and incident cases are collectively referred to as “cases during the cohort.” Follow-up time was split based on the timing of chlamydia diagnoses. A “recent” diagnosis was defined as any case that occurred within the most recent 6 months of follow-up, and a “previous” diagnosis was defined as any case that occurred before entry to the cohort, during follow-up, but more than 6 months ago, or the second case during follow-up. The same definitions were applied to gonorrhea. Exposure to trichomoniasis, BV, and Candida was defined as missing until the first test, negative from the date of the first negative test, and positive from the date of the first positive test. Pelvic inflammatory disease was diagnosed using the following clinical criteria: the presence of cervical motion, uterine or adnexal tenderness together with more than 5 leucocytes on endocervical gram stain, with or without pyrexia, and with or without lower abdominal pain. Referral for gynecologic opinion and further investigations (laparoscopy) occurred where symptoms were severe or to rule out ectopic pregnancy.
For chlamydia and gonorrhea, we calculated the proportion of women with a self-reported history, prevalence at first visit, number of incident infections, and the proportion of women diagnosed with a case during the cohort. We compared these measures between women who did and did not develop PID using Wilcoxon rank sum test for difference between proportions. We estimated the crude incidence rate of PID overall, by age and by infection status, and used Cox proportional hazards regression to explore the association between a recent case of chlamydia and PID. A multivariate model was constructed using variables that were significant (P < 0.05) in the univariate analysis. All statistical analyses were performed using STATA/SE 11.1, and 2-sided tests at the 95% level of significance were used to interpret the results of statistical tests.
The study received approval from Parkside Health Authority Research Ethics Committee (1089), Kensington, Chelsea and Westminster Local Research Ethics Committee (3279), and St Mary’s Hospital Local Research Ethics Committee (reference no. 97/Bi/158E). These approvals covered recruitment, data collection, storage of data and samples, and analysis of HIV, STI, and other health risks. Written informed consent was obtained from each participant.
Three hundred seven (51.3%) of the 598 sex workers recruited to the PSP between 1985 and 1993 were eligible for this study (Fig. 1). The cohort made 1973 visits to clinic (mean [SD], 6.4 [6.6]; median, 4; range, 2–44) during 401.2 person-years (py) of observation. The median duration of follow-up was 9.9 months (range, 1 week to 6 years and 8 months; mean, 15.7 months), and the mean (SD) age of women at their first visit was 26.2 (6.4) years (range, 16–52 years). There were missing data for some variables, including 30.3% missing a self-reported history of chlamydia and 18.9% missing a self-reported history of gonorrhea (but this will include those women who had never had previous tests). Tests were conducted at most clinic visits, but they were not carried out (or results were not available) at 30.1% (n = 593) of visits for chlamydia and at 0.3% (n = 6) for gonorrhea. Most chlamydia tests were performed using a DIF test, 11.5% (159/1380) were from an isolated test, of which 17 (10.7%) had positive results and changed the chlamydia status of 13 women.
At the first visit, 38.8% (83/214) of women reported a history of chlamydia and 46.2% (115/249) for gonorrhea. The prevalence of chlamydia was 8.5% (21/246) overall and 7.7% (5/65) in women who reported a history of chlamydia. The same figures for gonorrhea are 2.6% (8/307) and 3.5% (4/115). During follow-up, the incidence of chlamydia was 3.8% (43/1134) and that of gonorrhea was 1.9% (31/1660). Fifty women (16.3%) were diagnosed as having chlamydia during the study (prevalent or incident case), and 29 (9.4%) were diagnosed as having gonorrhea. The rate of chlamydia diagnosis (from the date of the first test) was 15.2 cases per 100 py (95% CI, 11.8–19.6), and the rate of gonorrhea was 8.0 per 100 py (95% CI, 5.6–11.3).
Forty-eight (15.6%) women were diagnosed as having at least 1 case of PID during the study. At the time of their first case, 8 women had a recent exposure to chlamydia, 4 to gonorrhea, and 1 to both infections. There was no significant difference between women who were diagnosed as having PID and those who were not in the mean number of clinic visits (6.5 compared with 6.4 [P = 0.622]) and age at the first visit (24.5 years compared with 26.5 years [P = 0.053]; see Table, Supplemental Digital Content 1, http://links.lww.com/OLQ/A52). However, women diagnosed as having PID had a higher incidence of infection (chlamydia: 6.3% and 3.3%, P = 0.048; gonorrhea: 3.4% and 1.6%, P = 0.046; trichomoniasis: 3.0% and 1.0%, P = 0.031; and BV: 26.2% and 15.5%, P < 0.001).
The crude incidence rate of PID in the cohort was 12.0 per 100 py (95% CI, 9.0–15.9) (Table 1). The rate of PID was approximately twice as high in women with a recent or previous case of chlamydia or gonorrhea compared with women without the exposure. The number of cases of PID was too small to compare the rate after the first known infection (chlamydia n = 3; gonorrhea n = 0) and the rate after a recurrent infection (chlamydia n = 2; gonorrhea n = 3).
The crude hazard ratio (HR) of PID was significantly higher in women with a recent case of chlamydia, a previous case of chlamydia, or a previous case of gonorrhea and was inversely associated with age (Table 1). The relationship between other infections (trichomoniasis, BV, Candida) and PID was investigated but were nonsignificant in the univariate analysis. The multivariate model was adjusted for age, previous chlamydia, recent gonorrhea, and previous gonorrhea. In the adjusted model, the association with recent or past chlamydia was no longer significant; only a history of gonorrhea (adjusted HR [aHR], 2.3; 95% CI, 1.1–4.6) and age (aHR, 0.9; 95% CI, 0.9–1.0) remained significant predictors of PID.
In this cohort, women who had been diagnosed as having chlamydia at any time had a higher risk of PID than did women who had never been diagnosed as having chlamydia. When controlling for age and recent or past gonorrhea, this association was no longer significant, although the size of the effect only declined slightly, suggesting that this was caused by lack of power once these multiple comparisons were included. In multivariate analysis, only past gonorrhea and younger age were significantly associated with an increased risk of PID.
Strengths and Limitations
This prospective cohort is well characterized and was collected using consistent methods and clinical care. Women had a high median number of follow-up visits, and symptomatic and asymptomatic women were offered repeat testing for a range of genital infections, which allowed us to report the risk of PID after a chlamydia diagnosis controlled for gonorrhea. There are limitations to the methods used to recruit the cohort and to collect the data. The clinic-based cohort, although supplemented by recruitment through fieldwork, is likely to include some selection bias through missing some sections of the population, including some at high risk for STI but who do not wish to engage with services and others at low risk who do not disclose their sex work or attend services. However, an early analysis of HIV risk showed little difference between clinic and community samples.13 Measurement bias may have occurred because some elements of data collection in this study evolved over time and because of the introduction of new technologies for diagnosis.11,12 The additional inclusion criteria used in this analysis could introduce further selection bias, and that the cohort was conducted in a functional service setting may have contributed to the high proportion of missing values for some variables.
A major limitation is the potential underascertainment of chlamydia cases. Women’s previous chlamydia status was classified using self-reported history of chlamydia before enrollment, and at the time of this study, chlamydia testing was not routine in GUM clinics. Therefore, cases of chlamydia before the study may have gone undiagnosed. During the study, information about chlamydia status was only available from tests performed by the study clinic, so infected women diagnosed elsewhere or who went untested (e.g., asymptomatic women who declined testing) would have been misclassified as uninfected. Furthermore, cases may have been undiagnosed because of the low sensitivity of the diagnostic tests (compared with culture: DIF 74.5% [95% CI, 67.6–80.5] and EIA 61.9% [95% CI, 55.6–68.7]).14 If Chlamydia-positive women were misclassified in these ways, our reported risk of PID will underestimate the true risk of PID after previous or current infection.
Pelvic inflammatory disease cases may also have been incompletely ascertained or misdiagnosed. As with chlamydia, no information was collected about cases treated outside the study setting, and the diagnostic method used, clinical criteria, has a low sensitivity and specificity, but there is no gold-standard noninvasive diagnostic test.15 Alternatively, it is possible that PID was overdiagnosed in this cohort because knowledge of a woman’s occupation as a sex worker or knowledge of a previous exposure to an STI could (appropriately) bias the differential diagnosis toward PID.2,16 The risk of PID reported in this study may overestimate the risk of PID after chlamydia diagnosed using nucleic acid amplification test (NAATs). Nucleic acid amplification tests detect the presence of chlamydia DNA rather than elementary bodies (DIF) or specific chlamydia antigens (EIA), and it has been postulated that infections detected using DIF or EIA may have a higher organism load than those detected by NAAT and therefore may be more likely to develop into PID.2
Comparison With Other Studies
We observed a lower prevalence of chlamydia than that in 2 contemporary London-based cohorts that also used DIF for diagnosis17 (8.5% in this cohort, 10.7% in women attending general practitioner practices pre-1987,18 and 29% in women attending a GUM clinic in 1990/199119). The lower prevalence in this study may reflect earlier uptake of safe sexual practices in this sex worker community. This is supported by published analysis of this cohort that demonstrated consistent use of condoms with clients and an increase in condom use over time.12 We found that the proportion of women who developed PID after a case of chlamydia was similar to that reported in a contemporary cohort of female sex workers in Nairobi (26% [13/50] in this study; 30% [44/146] in Kimani et al.10), which had a similar mean duration of follow-up (15.7 months compared with 17.6 months).
More recent studies in general population settings report a lower prevalence of chlamydia and lower rates of progression to PID than were seen in this study. Chlamydia prevalence in students participating in the Prevention of Pelvic Infection study was 5.7% at baseline, and during 12 months of follow-up, Chlamydia-positive screened participants had a crude rate of PID equivalent to 1.6 cases per 100 py.20 Although the rate of hospital diagnosed PID in women with a positive chlamydia test result in the large population-based cohort in Norway was 0.2 per 100 py (95% CI, 0.1–0.2),7 which is much lower than we observed (current exposure: 27.4 per 100 py [95% CI, 13.1–57.5] and previous exposure: 16.0 per 100 py [95% CI, 10.7–23.9]).
There are several factors that could explain these differences. First, we studied a high-risk population who may have been exposed to multiple agents that can cause PID. Second, modern NAATs used in the Prevention of Pelvic Infection study and the later years of the Norway cohort have a much higher sensitivity than the diagnostic tests used in this study (NAAT sensitivity compared with culture, 80%–100%21), which means that a higher proportion of infections should have been detected and these infections may be less likely to develop into PID, as discussed previously.2 Finally, the Norway cohort looked at hospitalized episodes of PID, whereas we studied cases diagnosed in a GUM clinic, which could be expected to include milder and more severe presentations, increasing the observed rate. However, the rate of PID in women with a recent chlamydia diagnosis in this study was still much higher than the combined rate of inpatient and outpatient PID after a diagnosis of chlamydia or gonorrhea reported in the US general population between 1988 and 1991 (5.6 cases per 100 py).22 This supports the suggestion that part of the elevated risk of PID in this study could derive from the high-risk nature of the cohort.
Although we report a high rate of PID, the adjusted hazard of PID after a previous chlamydia infection (aHR, 1.8 [95% CI, 1.0–3.5]) is comparable with the adjusted hazard in women with at least 1 positive test result in the Norway cohort (1.7 [95% CI, 1.2–2.4])7 and in a similar population-based cohort in Sweden (1.3 [95% CI, 1.0–1.6]).8 Interpreting this comparison is challenging because there are large differences in the sample size (307 in this cohort compared with 25,0007 and 44,0008), ascertainment of PID (GUM clinic vs. hospital episodes7,8), and controlling for gonorrhea in this study.
Meanings of the Study
This study suggests that a case of chlamydia or gonorrhea in the previous 6 months approximately doubles the rate of PID diagnosis and that the risk of PID is elevated in the 6 months after a diagnosis of chlamydia and remains elevated after any previous case of chlamydia or gonorrhea. However, these findings should be interpreted with caution given the small number of PID cases that followed infection in this study. The elevated risk of PID in the 6 months after a diagnosed case of chlamydia may be caused by continuing exposure (e.g., from an untreated partner, inadequate compliance with treatment, or treatment failure) or the fact the index infection may be a marker for an elevated risk of other infections that cause PID.
If chlamydia infections went undiagnosed, the observed elevated risk of PID in women with a history of chlamydia would support the evidence for an increase in the risk of PID after repeat infection.9,10,16 However, as we observed, a high uptake of testing this relationship may be more likely to represent a diagnostic bias. It is plausible that age is a proxy for number of previous infections, so in the absence of protective immunity, we would expect to observe an increasing risk of PID with age. Therefore, our findings could be interpreted as support for the presence of protective immunity after previous chlamydia infections, or age may be a proxy for health care–seeking behaviors including prompt treatment of STIs or a reduction in risk behavior, which would both reduce the risk of PID.
The prevalence of chlamydia in this cohort was lower than that seen in contemporary studies in health care settings.17–19,22 Also, compared with more recent population-based cohorts, we observed a higher rate of progression to PID with a similar adjusted hazard of PID after chlamydia.7,8 One interpretation is that not all cases of chlamydia are at equal risk for progressing to PID. Factors including the diagnostic method, the presence of other STIs, and individual biological factors may all attenuate this risk. However, in this study, previous lifetime exposure to gonorrhea was an important factor in increasing the risk of PID after a recent chlamydia infection.
We were unable to explore the impact of immunity to chlamydia on the risk of PID because there were too few cases to estimate the risk after a first infection compared with a recurrent infection. A larger study is needed and should also explore the association between recent gonorrhea and PID. To provide useful information for policy makers, more research into the natural history of chlamydia is needed from populations where NAATs are the main testing method, where it is possible to have a high ascertainment of PID cases (from both community and hospital settings) and where other etiologies of PID can be considered to allow for calculation of the attributable fraction of PID due to chlamydia.
1. Health Protection Agency. Genital Chlamydia trachomatis diagnoses in young adults in England, 2011. HIV/STIs. HPR. 2012: 6.
2. Haggerty CL, Gottlieb SL, Taylor BD, et al.. Risk of sequelae after Chlamydia trachomatis genital infection in women. J Infect Dis 2010; 201 (suppl 2): S134–S155.
3. Althaus CL, Heijne JC, Roellin A, et al.. Transmission dynamics of Chlamydia trachomatis affect the impact of screening programmes. Epidemics 2010; 2: 123–131.
4. Hu D, Hook EW 3rd, Goldie SJ. The impact of natural history parameters on the cost-effectiveness of Chlamydia trachomatis screening strategies. Sex Transm Dis 2006; 33: 428–436.
5. Adams EJ, Turner KM, Edmunds WJ. The cost effectiveness of opportunistic chlamydia screening in England. Sex Transm Infect 2007; 83: 267–274; discussion 274-265.
6. Risser WL, Risser JM. The incidence of pelvic inflammatory disease in untreated women infected with Chlamydia trachomatis: A structured review. Int J STD AIDS 2007; 18: 727–731.
7. Bakken IJ, Ghaderi S. Incidence of pelvic inflammatory disease in a large cohort of women tested for Chlamydia trachomatis: A historical follow-up study. BMC Infect Dis 2009; 9: 130.
8. Low N, Egger M, Sterne JAC, et al.. Incidence of severe reproductive tract complications associated with diagnosed genital chlamydial infection: The Uppsala Women’s Cohort Study. Sex Transm Infect 2006; 82: 212–218.
9. Hillis SD, Owens LM, Marchbanks PA, et al.. Recurrent chlamydial infections increase the risks of hospitalization for ectopic pregnancy and pelvic inflammatory disease. Am J Obstet Gynecol 1997; 176: 103–107.
10. Kimani J, Maclean IW, Bwayo JJ, et al.. Risk factors for Chlamydia trachomatis pelvic inflammatory disease among sex workers in Nairobi, Kenya. J Infect Dis 1996; 173: 1437–1444.
11. Ward H, Day S, Green A, et al.. Declining prevalence of STI in the London sex industry, 1985 to 2002. Sex Transm Infect 2004; 80: 374–376.
12. Ward H, Day S, Weber J. Risky business: health and safety in the sex industry over a 9 year period. Sex Transm Infect 1999; 75: 340–343.
13. Ward H, Day S, Mezzone J, et al.. Prostitution and risk of HIV: Female prostitutes in London. BMJ 1993; 307: 356–358.
14. Newhall WJ, Johnson RE, DeLisle S, et al.. Head-to-head evaluation of five chlamydia tests relative to a quality-assured culture standard. J Clin Microbiol 1999; 37: 681–685.
15. Simms I, Warburton F, Westrom L. Diagnosis of pelvic inflammatory disease: Time for a rethink. Sex Transm Infect 2003; 79: 491–494.
16. Gottlieb SL, Martin DH, Xu F, et al.. Summary: The natural history and immunobiology of Chlamydia trachomatis genital infection and implications for Chlamydia control. J Infect Dis 2010; 201 (suppl 2): S190–S204.
17. Simms I, Catchpole M, Brugha R, et al.. Epidemiology of genital Chlamydia trachomatis in England and Wales. Genitourin Med 1997; 73: 122–126.
18. Longhurst HJ, Flower N, Thomas BJ, et al.. A simple method for the detection of Chlamydia trachomatis infections in general practice. J R Coll Gen Pract 1987; 37: 255–256.
19. Hay PE, Thomas BJ, Horner PJ, et al.. Chlamydia trachomatis in women: The more you look, the more you find. Genitourin Med 1994; 70: 97–100.
20. Oakeshott P, Kerry S, Aghaizu A, et al.. Randomised controlled trial of screening for Chlamydia trachomatis to prevent pelvic inflammatory disease: The POPI (Prevention of Pelvic Infection) trial. BMJ 2010; 340: c1640.
22. Rothman KJ, Lanza L, Lal A, et al.. Incidence of pelvic inflammatory disease among women treated for gonorrhea or chlamydia. Pharmacoepidemiol Drug Saf 1996; 5: 409–414.