Joint effect of HPV16 with Chlamydia trachomatis and smoking on risk of cervical cancer: antagonism or misclassification (Nordic countries).

Abstract

To estimate the joint effects of infections with human papillomavirus type 16 (HPV16) and Chlamydia trachomatis and smoking on the risk of cervical cancer. To study whether the joint effects can be accounted for by misclassification in the HPV serology. A nested case-control study with incidence density sampling was conducted in three cohorts of 530,000 women, who donated serum samples to three Nordic serum banks in 1973-1994. The main outcome measure is the odds ratio (OR) of incidence rates of invasive cervical squamous cell carcinoma (SCC) among those seropositive for HPV16 and/or C. trachomatis and/or with increased levels of cotinine in serum compared to those negative for all the three exposures. Two hundred eight women with SCC and 624 matched controls were identified during a mean follow-up of 5 years through linkage to the national cancer registries. Exposure to past infections and smoking was defined by presence of specific IgG antibodies to HPV16 and C. trachomatis and increased levels of serum cotinine. Observed ORs were compared to OR = 20 for HPV16 and accounting the differences for by misclassification bias. OR = 20 was elected as a gold standard on the basis of other studies with PCR-based analyses and a follow-up design. Each of the three exposures was associated with an increased risk of SCC (OR = 5.4 for HPV16, 3.4 for C. trachomatis and 1.8 for cotinine). The interaction was antagonistic (observed OR = 2.5 among those positive for all three exposures as compared to OR = 33 expected on the basis of multiplicative single effects (p = 0.047)). The antagonism could not totally be accounted for by any credible combination of sensitivity and specificity of HPV16 serology. HPV16, C. trachomatis, and smoking are likely to be risk factors of SCC with strong antagonistic joint effect. Non-differential misclassification in serology for HPV16 could be ruled out (but only some types of differential) as an alternative explanation for the observed antagonism.