Chlamydia trachomatis infection and persistence of human papillomavirus
Version of Record online: 8 MAR 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 116, Issue 1, pages 110–115, 10 August 2005
How to Cite
Silins, I., Ryd, W., Strand, A., Wadell, G., Törnberg, S., Hansson, B. G., Wang, X., Arnheim, L., Dahl, V., Bremell, D., Persson, K., Dillner, J. and Rylander, E. (2005), Chlamydia trachomatis infection and persistence of human papillomavirus. Int. J. Cancer, 116: 110–115. doi: 10.1002/ijc.20970
- Issue online: 12 MAY 2005
- Version of Record online: 8 MAR 2005
- Manuscript Accepted: 23 NOV 2004
- Manuscript Received: 20 AUG 2004
- Swedish Cancer Society
- Europe against Cancer
- European Union Biomed 5 Consortium on HPVCCS
- HPV infection;
- HPV persistence;
- Chlamydia trachomatis;
Human papillomavirus (HPV) persistence is the major cause of cervical cancer, but most HPV infections will not persist and risk factors for HPV persistence are not well known. Chlamydia (C.) trachomatis infection seems to also be associated with cervical cancer. We investigated whether C. trachomatis infection is a risk factor for HPV persistence. In a cohort of 12,527 women participating in a population-based HPV screening trial in Sweden, 6,418 women completed testing for HPV DNA by general primer PCR and typing by reverse dot blot hybridization. On average 19 months later, 303 women that had been HPV-positive and had normal cytology at enrollment completed a new HPV test. Environmental exposures were assessed by an 87-item questionnaire. Previous sexually transmitted infections were also investigated by serology. At follow-up, 44% of the women were positive for the same type of HPV DNA as at enrollment. Persistence correlated with length of follow-up (p < 0.01) and condom use seemed to protect against HPV persistence (p < 0.05). The most significant risk factor for persistent presence of HPV DNA was self-reported history of previous C. trachomatis infection (relative risk in multivariate model = 2.09; 95% confidence interval = 1.05–4.18). We conclude that persistence of oncogenic HPV infections is more likely among women with a previous C. trachomatis infection. © 2005 Wiley-Liss, Inc.
Persistent oncogenic HPV infection is recognized as a necessary factor in development of high grade cervical intraepithelial neoplasia and invasive cervical cancer.1 Genital infections with oncogenic HPV types are very common among sexually active women, but only a minority of infected women will have a persistent HPV infection.2, 3, 4 Type-specific persistence of HPV (defined as repeated detectability of the same type of HPV DNA in serial samples) is a much stronger risk factor for high-grade CIN5 and cervical cancer6 than HPV infection per se.
Chlamydia (C.) trachomatis has been found repeatedly to associate with cervical neoplasia and invasive cancer in cross-sectional case-control studies,7, 8 although the association has commonly been thought to be the result of confounding by HPV. During recent years, an association with C. trachomatis has also been found in several biobank-based longitudinal studies with invasive cervical cancer as endpoint.9, 10, 11, 12 The vexed question of whether the association could be due to confounding by HPV infection is, however, only possible to address in cohort studies comprising only HPV-infected women.
A possible explanation for the association of C. trachomatis and cervical cancer might be that the C. trachomatis-induced inflammation results in an impaired ability to clear HPV infections.
To investigate these issues, we carried out a prospective, population-based cohort study restricted to HPV-infected women with HPV persistence as the endpoint.
Subjects and Methods
A population-based randomized multicenter study was initiated in Sweden in May, 1997. The study has the main aim to evaluate the effect of primary HPV screening at the age of 32–38 years for protection against CIN II–III or invasive cervical cancer at the next round of invitational cervical screening. A total of 12,527 women (mean age = 35.1, SD = 2.19) attending the population-based organized invitational screening in 5 Swedish cities (Gothenburg, Malmö, Stockholm, Umeå, Uppsala) were enrolled after informed consent. A total of 6,257 women were randomized to immediate HPV testing using computer-generated random numbers (Fig. 1). The randomization code was not released until samples had arrived at the laboratory. Only the virus laboratory, not any other personnel in the study nor the participating women were aware of the randomization status. A total of 168 of 6,257 (2.7%) samples were not adequate for HPV testing (tested negative in PCR for human β-globin). A total of 433 of 6,089 (7.1%) adequate cervical samples were HPV DNA-positive. A total of 341 of the HPV-positive women had a concomitant normal Pap smear and were, together with 398 randomly selected women with normal cytology, invited for a new Pap smear and HPV DNA test at least 12 months later (on average 19 months later). An 87-item questionnaire covering contraceptive history, reproductive history, sexual history, history of sexually transmitted diseases, other diseases and medications was administered and a serum sample was taken for C. trachomatis and Herpes simplex-2 (HSV-2) antibody testing. A significant number of women consented to participate in the main study (HPV DNA testing) but did not want to provide serum samples or did not respond to the questionnaire or to specific items in the questionnaire. The exact numbers of women with information available are detailed for each variable in the tables. To avoid differential losses or other differences in follow-up or possible recall bias in the self-reported questionnaires the women in the study and all personnel involved in administering questionnaires or scheduling visits were blinded to the HPV status of the woman. A total 270 of 341 (79.2%) of invited HPV-positive women and 337 of 398 (84.7%) of women invited at random attended the follow-up visit. The 337 baseline HPV samples from the attending women who had been invited at random were retrieved and HPV tested in parallel with the follow-up samples (Fig. 1). It should be emphasized that the invitation of random women and of HPV-positive women are based on a random-number generated subsample of the overall cohort. We carried out successful baseline HPV tests on a population-based sample of 6,418 women (Fig. 1).
Persistence was defined as detection of the same type of HPV DNA in both the base-line and follow-up samples.
For the subpopulation of women that were positive for HPV 16 DNA at the first visit and corresponding controls (206 participants), we also tested the human leukocyte antigen (HLA) allele DQB1*06.
For analysis of correlates of Chlamydial and herpes infections, we also included the interviewed women who had been HPV-negative at baseline into the analysis (n = 298) (Fig. 1).
C. trachomatis testing
IgG antibodies to C. trachomatis were detected by microimmunofluorescence (MIF).13, 14, 15 In brief, slides with antigens of C. trachomatis and C. pneumoniae were used. The pre-assigned cut-off of titers of ≥1/16 in the absence of cross reactivity with C. pneumoniae was considered a positive result. Personnel who carried out the serological tests were blinded to the case/control status of subjects. The assay was evaluated for specificity using the panel of monogamous women in the study, of whom 2 of 42 were positive. Specificity was also indicated by the strong correlation with sexual history of the participants (Table I).
|Correlate||Clearance group1||Persistence group2||Crude OR||Adjusted OR1||95% CI|
|Age (per year)||0.62||0.95||0.83–1.07|
|Pills used, ever (yes vs. no)||149/15||107/15||0.72||1.02||0.37–2.79|
|Copper IUD, ever (yes vs. no)||55/110||35/87||0.80||0.74||0.39–1.41|
|HPV 16 seropositivity (yes vs. no)||69/80||54/50||1.24||1.09||0.61–1.96|
|C. trachomatis self-reported history (ever vs. never)||36/125||39/79||1.70||2.09||1.05–4.18|
|C. trachomatis MIF (pos. vs. neg.)||44/104||42/60||1.64||1.61||0.86–3.01|
|HSV-2 seropositivity (yes vs. no)||51/98||24/77||0.67||0.63||0.32–1.22|
|Smoking history (yes vs. no)||101/65||73/49||0.97||1.00||0.55–1.85|
|Condom use (yes vs. no)||126/39||80/42||0.59||0.48||0.25–0.92|
|Length of follow-up|
|p for trend||0.008|
|Lifetime sexual partners|
HPV DNA testing
Cervical brush samples were analyzed using a general primer GP5+/6+ mediated PCR followed by enzyme immunoassay (EIA) with digoxigenin-labeled human papillomavirus (HPV) type-specific oligonucleotide probes for detection of 14 high-risk (Types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68) HPVs,16, 17 followed by reverse dot blot hybridization (RDBH) for confirmation and HPV typing in a single assay.18
The presence of antibodies specific for HSV-2 was determined using an established enzyme-linked immunosorbent assay (EIA).19 The optical density values transformed into ELISA units using the PLL (parallel line) method.20 Each set of experiments included a validation panel with HSV-1-positive serum samples from children (that were always negative). An initial comparison with EIAs based on purified glycoprotein G from HSV-2 (a kind gift of Professor T. Bergström) found identical specificity, but slightly better sensitivity of our EIA, as determined using validation panels of serum samples from individuals with isolation-verified HSV-2 infection.
HLA DQB1*06 testing
Blood samples were separated into serum, lymphocytes and granulocytes. The DNA was extracted from the granulocytes by phenol-chloroform extraction and dissolved in sterile H2O and stored at −20°C. Olerup SSP PCR kit (Olerup SSP AB, Saltsjöbaden, Sweden) was used to detect the *0602 allele, according to the manufacturer's instructions.
Relative risks (RR), odds ratios (OR) and 95% confidence intervals (CI) were calculated by logistic regression using Statistica® software. All analyzed variables were initially included in the multivariate logistic regression model. Determinants not significant statistically were then removed from the model, except variables with biological rationale for inclusion in the multivariate model.
The absolute risk of persistent positivity for the same HPV type was 42.6% (129 of 303 baseline HPV-positive women). For 15 women (5.0%) there was a change of HPV type that was classified as a virus clearance. The remaining 52.5% of the women were HPV-negative at the second visit (Fig. 1).
Correlates of HPV DNA persistence were evaluated using the women who had cleared their HPV as the reference group. Self-reported history of ever having had a C. trachomatis infection was the only factor associated with increased risk for HPV persistence (RR = 2.09, 95% CI = 1.05–4.18) (Table I). A strong tendency for increased risk was observed also for C. trachomatis seropositivity (Table I). There was a decreased risk of HPV persistence with increasing length of follow-up (RR = 0.33, 95% CI = 0.16–0.69) (Table I) and regular condom use was associated with a significant protective effect against HPV persistence (Table I). When the multivariate analysis was stratified to include only women who did not use condoms, the increased risk of HPV persistence in case of history of C. trachomatis infection seemed to become stronger (RR = 2.65, 95% CI = 1.16–6.07) (data not shown).
To further assist in the consideration of whether the HPV persistence risk associated with C. trachomatis infection could be due to confounding, we also determined the correlates of self-reported Chlamydia infection and C. trachomatis seropositivity. The most important risk determinant for both of these measures of Chlamydia exposure was an increasing number of life-time sexual partners (Tables II, III). Late coitarche and regular condom use were associated with protection against Chlamydia seropositivity (Table III). Chlamydia serology and self-reported history of Chlamydial infection also correlated with each other (OR = 2.10, 95% CI = 1.28–3.44).
|Negative C. trachomatis history||Positive C. trachomatis history||Adjusted OR1||CI 95%|
|Age (per year)||0.92||0.84–1.02|
|Pills used, ever (yes vs. no)||177/25||70/5||3.10||1.04–9.28|
|Age at coitarche (>16 vs. −16)||90/110||32/43||1.15||0.69–1.91|
|Lifetime sexual partners|
|Smoking history (yes vs. no)||120/83||48/27||1.02||0.60–1.72|
|Condom use (yes vs. no)||146/57||54/21||0.91||0.53–1.58|
|C. trachomatis MIF (pos vs. neg)||50/129||35/27||2.10||1.28–3.44|
|HSV-2 seropositivity (yes vs. no)||53/128||23/40||1.24||0.74–2.08|
|Negative C. trachomatis MIF||Positive C. trachomatis MIF||Adjusted OR1||95% CI|
|Age (per year)||1.10||1.00–1.20|
|Pills used, ever (yes vs. no)||141/19||80/6||1.37||0.62–3.05|
|Age at coitarche (>16 vs. −16)||84/76||28/58||0.53||0.33–0.84|
|Lifetime sexual partners|
|Smoking history (yes vs. no)||95/67||57/29||1.12||0.69–1.81|
|Condom use (yes vs. no)||119/42||55/31||0.51||0.31–0.83|
|C. trachomatis self-reported history (ever vs. never)||27/129||35/50||2.10||1.28–3.44|
|HSV-2 seropositivity (yes vs. no)||45/118||32/54||1.13||0.70–1.82|
Presence of the HLA allele DQB1*06 (reported to associate with HPV persistence21 and CIN risk in an HPV-16-specific manner21, 22) did not correlate with HPV 16 DNA persistence (OR = 0.60, 95% CI = 0.20–1.80). Because HLA typing was only carried out for the HPV 16 positive women we were not able to analyze this variable in the multivariate models.
The major hypothesis evaluated in our present study was whether impaired ability to clear an HPV infection is associated with C. trachomatis infection. Indeed, self-reported history of C. trachomatis infection was the only positively associated determinant of HPV DNA persistence in our prospectively followed, population-based cohort of HPV DNA positive women. Our finding suggests a plausible mechanism whereby C. trachomatis infection could be associated with increased risk for cervical cancer, as has been reported in several epidemiological studies.7, 10, 11, 12, 23, 24 It has been reported that C. trachomatis infection may induce large local amounts of pro-inflammatory chemokines, including IL-1, IL-8 and granulocyte-macrophage colony-stimulating factor25 producing a chronic inflammatory background that may conceivably contribute to HPV DNA persistence in the cervix.
Although Chlamydial seropositivity7, 10, 11, 12, 23, 24 and oncogenic HPV seropositivity26 associate with cervical cancer risk, the joint risk of being positive for both seropositivities is less than would be expected for independent effects.24 Our present study has used HPV DNA testing, rather than HPV serology, as a measure of HPV infection. Antibodies to HPV may be absent despite infection and seropositivity is more common among women with persistent infection.27 If the association between Chlamydia and cancer is mediated by an increased risk of HPV persistence, an association of Chlamydia with cancer risk would not be expected to be seen among women who already have HPV persistence.
Confounding should be considered as an explanation for all epidemiological studies. The analysis of C. trachomatis covariates found the life-time number of sexual partners as the only strong covariate, raising the possibility that some unknown aspect of sexual history (e.g., repeated past exposures to HPV infections or other STI) could account for the association. None of the questionnaire items related to sexual history (life-time number of sexual partners, number of current sexual partners, age at coitarche, duration, length or age at start of oral contraceptive use or number of births, pregnancies or age at first pregnancy) had any association with HPV persistence (Table I and data not shown). Other conceivable confounders, such as smoking or condom use, were included in the multivariate model and did not affect the estimate of association between Chlamydia and HPV persistence. Further, the association was at least as strong in analysis restricted to include only women who did not use condoms. Finally, none of the other specific STI investigated (history of genital herpes, history of genital warts, history of vaginal discharge, seropositivity for HPV16 or seropositivity for HSV-2) had any association with HPV persistence, despite also being strongly associated with sexual history in a similar manner as C. trachomatis was (Table IV).
|No. of life-time sexual partners||C. trachomatis serology (MIF)||C. trachomatis history||HSV-2 serology||HSV-2 history||Genital warts history||Age of coitarche >16||Age at first pregnancy >19|
|1||4.8% (2/42)||7.0% (3/43)||9.5% (4/42)||0% (0/43)||0% (0/43)||86.0% (37/43)||81.6% (31/38)|
|2–3||14.6% (8/55)||6.7% (4/60)||10.9% (6/55)||5.0% (3/60)||0% (0/60)||62.5% (40/64)||94.5% (52/55)|
|4–5||20.3% (15/74)||15.7% (13/83)||14.7% (11/75)||4.8% (4/83)||15.7% (13/83)||54.2% (45/83)||91.4% (64/70)|
|6–10||34.3% (57/166)||23.9% (44/184)||29.5% (49/166)||10.9% (20/184)||16.8% (31/184)||40.5% (75/185)||80.1% (125/156)|
|>10–20||41.1% (37/90)||32.6% (32/98)||36.2% (34/94)||8.2% (8/98)||16.3% (16/98)||33.3% (34/102)||87.7% (71/81)|
|>20||46.5% (20/43)||49.0% (25/51)||45.4% (20/44)||25.5% (13/51)||31.4% (16/51)||34.0% (18/53)||77.3% (34/44)|
|p for trend||<0.0001||<0.0001||<0.0001||<0.0002||<0.0001||<0.0001||<0.05|
We used 2 different measures for previous C. trachomatis exposure: serology using microimmunofluorescence (MIF) and self-reported history. Both measures have their weaknesses. Microimmunofluorescence, although recognized as the “gold standard” for C. trachomatis serology has modest sensitivity.28, 29 Self-reported history is also not a completely accurate measure of exposure as infections are commonly asymptomatic and are often under-reported.30, 31 Women who have been diagnosed with a Chlamydia infection are mostly aware of it, however, because they have been treated and also participated in partner tracing.
Although both measures reflect the same exposure, they overlapped only partially. Agreement beyond chance (κ-value) for these 2 variables was only 0.20, in line with the sensitivity problems discussed above. The fact that both measures of C. trachomatis infection showed a similar strength of their association with HPV DNA persistence suggests that the association is not a coincidence. The fact that both measures of the exposure have substantial misclassification suggests that actual risks for HPV persistence may have been substantially underestimated.
The protective effect against HPV persistence for condom users could be explained in several ways. Chance finding and residual confounding due to the association with C. trachomatis exposure (assessed by measures of low sensitivity) are plausible explanations. Protection against repeated type-specific re-infection with HPV, which would by design have been classified as “persistence,” is also a possibility.
In summary, we found an association between Chlamydia infection and HPV persistence in the cervix that could not be explained by investigated confounding factors. The association could be relevant for planning and evaluation of practices for control of these infections and merits further investigations.
W.R., A.S., G.W., B.G.H., J.D. and E.R. constitute the steering group of the HPV screening trial and have participated in its design and implementation. E.R. also developed and coordinated the questionnaire administration. S.T. has provided epidemiological advice and enrolled most of the women. V.D., D.B. and K.P. advised on and carried out Chlamydial analyses, X.W. advised on and carried out H.S.V. analyses, L.A. advised on and carried out HLA analyses. I.S. administrated the database and undertook the statistical analysis. J.D. coordinated study design and writing of the paper and is guarantor. All authors contributed to the final version. All authors declare that the answer to the questions on competing interest form are all No and therefore have nothing to declare. All researchers are independent of funding. Ethical approval was provided by the Research Ethics Committee of the Karolinska Institute, decision number 98:469. All women gave informed consent.
- 28Nongonococcal urethritis and related infections. Washington, DC: American Society for Microbiology, 1977., , , , .