A population-based prospective study of Chlamydia trachomatis infection and cervical carcinoma



Persistent human papillomavirus (HPV) infection is an established cause of cervical cancer, but the role of other sexually transmitted agents, most notably Chlamydia trachomatis, has not been well defined. The women participating in the population-based cervical cancer screening program in Västerbotten county of Northern Sweden were followed up for up to 26 years to identify 118 women who developed cervical cancer after having had a normal Pap smear (on average 5.6 years later; range 0.5 months–26 years). As controls, we selected another 118 women who were matched by birth cohort, time-point of sampling of the baseline normal smear and who had a normal smear at the time when the corresponding case was diagnosed with cancer. The Pap smears and cervical cancer biopsies were analyzed by PCR for C. trachomatis DNA and for HPV DNA. At baseline, C. trachomatis DNA was present in 8% of cases but not among any one of the controls. The relative risk for cervical cancer associated with past C. trachomatis infection, adjusted for concomitant HPV DNA positivity, was 17.1 (95% CI 2.6–∞).The presence of C. trachomatis and of HPV were not interrelated. Whereas C. trachomatis was primarily found in specimens taken many years before cancer diagnosis, HPV DNA was associated with a short lag time before cancer diagnosis. Whereas most women who were HPV DNA-positive in the prediagnostic smear were also positive for the same virus in the cervical cancer biopsy, none of the women were positive for C. trachomatis in both the prediagnostic smear and in the subsequent cervical cancer. In conclusion, a prior cervical C. trachomatis infection was associated with an increased risk for development of invasive cervical cancer. © 2002 Wiley-Liss, Inc.

Cervical cancer is one of the leading causes of death among women especially in developing countries.1 The etiologic role of human papillomavirus (HPV) in cervical cancer is established from both experimental and epidemiologic evidence,2 including population-based prospective studies with invasive cervical cancer as an endpoint.3 Several other environmental and behavioral factors such as smoking, age at first intercourse, number of lifetime sexual partners and other sexually transmitted infections (STIs) are also associated with cervical cancer.4

Chlamydia trachomatis infection is the most common bacterial STI. Although most infections are asymptomatic, they may cause pelvic inflammatory disease (PID), adverse pregnancy outcome and tubal factor infertility in women.5, 6C. trachomatis is also a common cause of urethritis, cervicitis, epididymitis, proctitis and reactive arthritis. Exposure to C. trachomatis has been consistently associated with cervical cancer and its precursors.7, 8, 9, 10, 11, 12 However, this association has commonly been attributed to confounding by infection with HPV. Poor epidemiologic study designs and/or inaccurate diagnostic assays used have made it particularly difficult to disentangle the effects of these STIs. Therefore, we performed a population-based prospective study on the risk of invasive cervical cancer among healthy women. The presence of C. trachomatis and HPV DNA in prediagnostic cervical smears was analyzed by PCR. We report that women who are positive for C. trachomatis are at a substantially increased risk for subsequent development of invasive cervical cancer.


Study base and study design

Women participating in the population-based invitational cervical cancer screening program in Västerbotten county in northern Sweden (total population was 260,472 in 1995 of whom 130,651 were women), who had no history of operative treatment of the cervix, who had taken at least 1 cytologically normal smear and at least 1 additional Pap smear constituted the study base. The program started in 1969 with invitations issued at 4-year intervals. The attendance rate has been more than 80%. From the start of the program to the early 1980s, a wooden Ayers spatula and a cotton tip applicator were used for cervical sampling. In the mid 1980s the latter was replaced by a cytologic brush (Cytobrush®, Medscand Medical AB, Malmö, Sweden). Since 1969 all patients with cervical cancer have been treated at the University Hospital, Umeå. All cytologic diagnoses on Pap smears taken within the organized program and on spontaneously taken smears have been recorded at the Cytology Laboratory, University Hospital, Umeå. Similarly, all cervical cancer patients from the Västerbotten population have been treated at this hospital. The histologic specimens and all data on histopathologic diagnosis, clinical stage and survival have been archived in the same hospital.

Linkage between the cytology registry and the Swedish Cancer Registry from 1969–1995 identified 133 women in the study base who had been diagnosed with invasive cervical cancer after the date of a normal smear. Four women were excluded because of incorrect entry in the cervical cancer registry and 11 women because histopathologic review found noninvasive cervical neoplasia in their histologic samples, leaving 118 cases with invasive cervical cancer (25 adenocarcinoma and 93 squamous cell carcinoma). The mean time between the sampling of the baseline smear and cancer diagnosis was 5.6 years (range 0.5 months–26.2 years). Eligible controls were women in the study base who did not develop cervical cancer before the time-point of diagnosis of the corresponding case. They were matched 1:1 individually to the corresponding case according to age (same calendar year of birth) and time-point of sampling of the baseline smear. The date of smear taking differed 1 month on average. The average age when the prediagnostic Pap smear was taken was 44.2 years (range 19.1–74.1) among the cases and 44.1 years (range 19.5–74.4) among the controls. The controls were also required to have been followed up for at least the same length of time as the corresponding case women, i.e., it was mandated that a normal smear should have been taken after the diagnosis of cancer in the matched-case woman. If there were several normal smears stored after the date of cancer diagnosis of the corresponding case, the smear sampled closest after the diagnosis date of the matched-case woman was chosen. The time-point of sampling differed on average 9.6 months. In some cases, the time difference was more than 1 year. The average age at cancer diagnosis was 49.8 years (range 23.7–78.9), and the age among controls at their corresponding normal smear was 50.5 years (range 24.0–79.1).

All smears were reevaluated before laboratory analysis as previously described in detail.3 Women with negative cytology with abnormal findings on re-review were not excluded. The exclusion of these women from our study would have made the results difficult to interpret, since it would have redefined the study population using additional tests that are related to both the exposure and the outcome. Thus, the study base consisted of all women who were at risk for cervical cancer during follow-up. All histologic slides of the cases were reexamined and only blocks confirmed to include cancer lesions were used for PCR analysis.

Ethical approval of our study was obtained from the Institutional Review Boards (IRB) of the Umeå University and the Karolinska Institute. As approved by the IRB, informed consent was not obtained but a press conference of the study was held on May 23, 1995 (resulting in coverage by major regional newspapers).

DNA extraction

DNA was extracted from archival smears and biopsies as previously described.14, 15 From 12 cases the histopathologic specimens could not be retrieved for PCR and 2 cases had inadequate material for PCR analysis. The DNA was dissolved in 400 μl of Tris buffer with 10 mM ethylenedinitrilotetraacetic acid. Four 5 μm thick sections were cut from each paraffin block. Knives were changed and an empty paraffin block sectioned in between each biopsy to prevent cross-contamination. All samples were tested for DNA integrity by PCR using human ribosomal gene S14 primers14, 15 that gave 150 bp amplimeres. In the S14 PCR, 1 of the cervical smears and 2 of the biopsies were negative. Samples positive for S14 PCR but negative for HPV were alcohol precipitated and the PCR was repeated.

PCR analysis for Chlamydia trachomatis

PCR was performed using the COBAS AMPLICOR™ CT Test (Roche Molecular Diagnostics, Branchburg, NJ) with primers (CP24 and CP27) targeting the conserved cryptic plasmid of approximately 7,500 bp common to all serovars of C. trachomatis, giving a PCR product of 207 bp. This test has been reported to be highly sensitive and reproducible.16 The sensitivity was found to be 5–10 elementary bodies per sample. An internal control was added to the processed specimens to test for inhibitors that might interfere with amplification. Five microliters of extracted DNA were incubated at room temperature for 5 min together with 22.5 μl of lysis buffer. A further 22.5 μl of specimen dilution buffer was added prior to PCR amplification. Detection of positive results was performed according to the manufacturer's instructions including negative and positive controls. The samples were analyzed blinded, with random mixing of cases and controls.

HPV PCR analysis

The HPV analysis methods and data have been reported previously (Wallin NEJM).3 Briefly, PCR was performed with the GP5+/GP6+ primers, which amplify 150 bp products, a similar size as that generated by the S14 PCR. The sensitivity was determined from dilutions of DNA from the cervical cancer cell line SiHa, in each PCR run, and was consistently 1.0 fg or less. The PCR analyses were performed blindly, with cases and controls analyzed in the same run. Blanks without DNA were included after every 24 samples. For HPV typing, direct automated sequencing of the GP5+/GP6+ PCR products was performed using an ABI PRISM® 310 Genetic Analyzer (Applied Biosystems, Foster City, CA), and the sequence matched to the Genbank database using the BLAST program (http://www.ncbi.nlm.nih.gov:80/BLAST/).

Statistical methods

Relative risks with 95% confidence intervals (CI) were estimated as odds ratios (OR) by exact conditional logistic regression using LogXact 4 software (Cytel, Cambridge, MA). The conditional likelihood function was maximized whenever possible. Otherwise the less reliable median unbiased estimates are reported.17 To increase the statistical efficiency, matched pairs were pooled, resulting in 74 strata with a fixed matching ratio of 1. Differences between HPV or C. trachomatis-positive and C. trachomatis-negative women for the continuous variables were tested by Wilcoxon's rank sum test. All p-values are 2-sided.


At baseline, C. trachomatis DNA was present in 10 (8%) of the 118 cases who subsequently developed invasive cervical cancer, but in none of the controls (Table I). Of the 10 C. trachomatis-positive cases, 9 developed squamous cell carcinoma and 1 carcinoma that could not be typed histologically due to missing block. All prediagnostic Pap smears taken less than 3 years before cancer diagnosis were negative for C. trachomatis DNA. The estimated relative risk of cervical cancer associated with past C. trachomatis infection was high (RR 19.0) and changed little after adjusting for concomitant HPV DNA (OR 17.1; 95% CI 2.6–∞) (Table I). The point estimates of the relative risk associated with C. trachomatis increased with increasing lag time (Table I).

Table I. Relative Risks (95% CI) for Development of Invasive Cervical Carcinoma Associated with C. Trachomatis in Prediagnostic Smears
Lag time4 (years)CasesControlsRR3 (95% CI)RR3 (95% CI)
  • 1

    Positive for C. trachomatis

  • 2

    Adjusted for concomitant HPV DNA presence.

  • 3

    The median unbiased estimate of the risk needs to be interpreted with caution, when no controls were exposed to C. trachomatis. The exact confidence interval is the more informative and reliable statistic.

  • 4

    Time between the prediagnostic Pap smear and the subsequent cancer diagnosis in the case.

0–3480 (0%)480 (0%)  
3–6354 (11%)350 (0%)7.5 (0.8–∞)5.8 (0.6–∞)
>6356 (17%)350 (0%)10.5 (1.4–∞)10.5 (1.4–∞)
Total11810 (8%)1180 (0%)19.0 (2.8–∞)17.1 (2.6–∞)

Three cases had C. trachomatis DNA in the cervical cancer biopsies. All smears taken from the controls at the time of the cases' diagnoses were negative for C. trachomatis DNA. In only 1 case was the prediagnostic smear positive for both C. trachomatis DNA and HPV DNA. None of the cases were positive for C. trachomatis in both the prediagnostic smear and in the cervical cancer biopsy.

Results of the HPV DNA testing have been reported previously.3 In brief, 35 (30%) of the 118 cases had HPV DNA at baseline compared to 3 (3%) of 118 of the controls (OR 16.4; CI 4.4–75.1). At cancer diagnosis, HPV DNA was amplifiable in 80 (77%) of the 104 cervical cancer biopsies and in 4 (4%) of the 104 Pap smears of healthy women. Of the 104 cases with biopsy specimen type-specific persistence of high-risk HPV types was seen in 26% of the cases compared to none of the controls.3

The median and mean ages of the C. trachomatis-positive and HPV-positive women were similar (41.1 and 39.6 vs. 39.0 and 41.0, respectively) (Table II). However, C. trachomatis DNA positivity was associated with long lag time between prediagnostic smear and cancer diagnosis in the case. The average lag time until diagnosis of cancer among C. trachomatis-positive cases was 9.2 years compared to 5.2 years among C. trachomatis-negative cases (p = 0.01). By contrast, HPV DNA positivity was associated with short lag time. The average lag time among HPV-positive cases was 4.1 years compared to 6.2 years (p = 0.05) among HPV-negative cases (Table II).

Table II. The Presence of C. Trachomatis and HPV DNA in Prediagnostic Smears by Age (A) and Lag Time (B)
A. Age at prediagnostic smear among cases
  • 1

    Wilcoxon rank sum test.

C. trachomatis      
HPV DNA      
B. Lag time between prediagnostic smear and diagnostic biopsy among cases
C. trachomatis      
HPV DNA      


This population-based prospective study found that C. trachomatis infection is associated with increased risk for development of cervical carcinoma. In cervical cancer, the presence of HPV DNA appears to be virtually required for cancer development. However, because only a small minority of HPV-exposed women actually develop cancer, it is conceivable that other risk factors are involved at some stage in the carcinogenic process. Longitudinal studies are necessary to study effects of possible risk factors during the process of cervical carcinogenesis because information of temporal associations cannot be obtained from case-control studies. In addition to the longitudinal design, our study has several other advantages. First, invasive cervical cancer was used as the endpoint. This is important since intraepithelial neoplasias often do not progress to more severe lesions and, in fact, a majority may regress spontaneously. Second, the study was nested in a cohort where all cervical smears taken had been stored. This study design minimizes the risk for selection biases, since samples from all cases and controls in the study base were available for analysis, the equivalent to a 100% attendance rate in a case-control study. Third, the study was population-based, making the results generalisable. Finally, sensitive and specific molecular diagnostic assays were used to minimize the risk for misclassification.

The risk associated with C. trachomatis differed from that of HPV in 2 respects. First, it was associated with long lag time before the diagnosis rather than short lag time, in spite of the fact that HPV-positive and C. trachomatis-positive women were of similar age. Similar results were obtained in a recent prospective seroepidemiologic study on C. trachomatis, HPV and invasive cervical cancer.12 By contrast, HPV was found primarily in specimens taken with short lag time until cancer.18 Secondly, none of the cases were positive for C. trachomatis both in the prediagnostic smear and in the cancer biopsy, while most women with HPV DNA in their baseline smear were persistently positive for the same virus type in the subsequent cervical cancer.3 Thus, persistence of HPV appears to be a prerequisite for the development of cervical cancer.19 By contrast, if C. trachomatis has a role in the cervical carcinogenesis, it is not dependent on additional risk factors in the presence of another extremely strong risk factor, in this instance HPV. Although the relative risks were adjusted for concurrent HPV infection, it is possible that C. trachomatis may be a risk marker for acquisition of HPV infection in the future. However, the C. trachomatis-associated RR was strong and changed little after HPV adjustment, indicating that cosegregation of the risk to acquire C. trachomatis and HPV infection is not strong and thus would be unlikely to explain the entire C. trachomatis-associated relative risk. Several seroepidemiologic studies have also found that C. trachomatis is a risk determinant for the development of cervical neoplasia.7, 8, 9, 10

The fact that our study was based on archival smears rather than fresh specimens might conceivably have resulted in some false-negative results, e.g., if the DNA is degraded on storage. However, although the proportion of prediagnostic smears positive for HPV DNA decreased with increasing lag time (time until cancer was diagnosed), HPV positivity was not dependent on storage time (calendar time between smear taking and testing of the archival specimen), suggesting that there was no major degradation over time.3 Since the sensitivity for HPV detection on archival specimens was at least 77%3 and adjustment for HPV DNA had only a small effect on the risk associated with C. trachomatis, possible underestimation of the HPV DNA presence is not likely to have affected our results significantly. Underestimation of C. trachomatis DNA prevalence would, if present, have affected both the cases and the controls equally and would, if anything, have underestimated the relative risks.

The concept of a relationship between C. trachomatis and cervical cancer is not new. C. trachomatis causes severe inflammation of the cervix associated with metaplastic atypia of the transformation zone of the cervix.20, 21 The prevalence of C. trachomatis is higher among women with abnormal cytologic findings compared to women with normal cytology.22, 23 A number of seroepidemiologic studies have also found an association between C. trachomatis and cervical cancer.9, 10, 11, 24 Several mechanisms could explain how C. trachomatis might be involved in the development of cervical cancer. C. trachomatis can inhibit apoptosis by blocking the release of mitochondrial cytochrome C and caspase 3, which may then allow infected cells to escape CD8+ killer T-cell attack.25 It has also been reported that the function of the natural killer cells producing interferon gamma (IFN-γ) and the development of cell-mediated Th1 response are impaired by C. trachomatis infection.26, 27, 28C. trachomatis can also suppress IFN-γ-induced MHC class II expression by selective disruption of the IFN-γ signaling pathways, thereby evading the host immune response.29 It would therefore be most interesting to perform cohort studies investigating whether the outcome of an HPV infection, in terms of clearance or persistence of HPV, is dependent on prior exposure to C. trachomatis.

Our study confirms the epidemiologic association between C. trachomatis and cervical cancer and suggests that the association is not only due to confounding resulting from covariation of the risks to acquire HPV and C. trachomatis. If the association is of etiologic significance, it could have major implications for preventive strategies. Future studies would be needed in 2 areas: better definition of C. trachomatis-associated relative risks in larger studies encompassing substantial numbers of exposed control women and studies investigating possible carcinogenic mechanisms, e.g., investigations on possible influence of C. trachomatis on the HPV clearance rate.


We are grateful to Profs. G. Wadell, T. Hakulinen and M. Kenward for stimulating discussions.