Can we predict reproductive complications after Chlamydia trachomatis infection? A cohort study integrating demographic, behavioural, host genetic and infection-related factors

OBJECTIVES: (chlamydia) infection has a variable course; most infections resolve but some lead to reproductive complications. Behavioural, immunological and host genetic factors likely influence risk. We aimed to identify predictors of complications in persons with and without chlamydia and develop prediction models for risk stratification. METHODS: We analysed data from a long-term cohort of Dutch individuals registered as female in municipal registries and previously screened for chlamydia, including demographic, behavioural, chlamydia-infection and serology variables, and 24 single nucleotide polymorphisms (SNPs). We constructed SNP sets capturing shared biological pathways. Using multivariable Cox regression, we developed and internally validated composite and outcome-specific models for reproductive complications (pelvic inflammatory disease (PID), ectopic pregnancy and tubal factor infertility) stratified by chlamydia status. Model performance was assessed. Point-based risk scores were derived from final models, estimating 15-year risks. Thresholds assessed the proportion classified as high risk and the proportion of outcomes captured. RESULTS: Among the 5704 participants (35 523 and 66 511 person-years in chlamydia-positive and chlamydia-negative groups), 7.1% and 5.4% experienced reproductive complications, respectively. Across outcomes, predictors included behavioural, sociodemographic and immune-related genetic factors, with most identified SNPs known to be involved in immune pathways. Model discrimination was modest (area under the curve ~0.65). Using a ≥15% predicted risk threshold, 54% of chlamydia-positive individuals were classified as high risk of reproductive complications, capturing 75% of composite outcomes (positive predictive value (PPV) 9.7%). For PID, 27% were classified as high risk, capturing 39% of cases (PPV 6.9%), and for ectopic pregnancy, 1.7% were classified as high risk, capturing 10% of cases (PPV 11.8%). CONCLUSIONS: Prediction models integrating behavioural, sociodemographic and host genetic factors showed modest performance, with better classification for chlamydia-positive individuals and PID than for ectopic pregnancy. Individual-level risk stratification is unlikely to be clinically useful, supporting population-level or group-level targeting of prevention and care.