Background There is considerable discussion about which sexual behaviour variables are most strongly associated with chlamydia. We investigated this in a study conducted within a chlamydia screening trial.

Methods A consecutive sample of patients aged 16–29 attending 134 GP clinics in 54 postcodes was recruited. Patients completed a questionnaire and chlamydia test. Using random effects logit regression models we estimated (1) the significance of a variable’s association with chlamydia (likelihood ratio test for model fit), and; (2) the strength of association with chlamydia (odds ratio[OR]). Number of partners in the last 12 months and partnership duration (years) were fitted as continuous variables. Each model included age, gender and a sexual behaviour variable. A multivariate model including all sexual behaviour variables was also run. All analyses accounted for intra-cluster correlation within postcode.

Results 1257 men and 3025 women participated (66–71% response rate). Chlamydia positivity was 4.6%(95% CI: 3.9–5.4); similar between men (5.2%; 95% CI: 3.9–6.4) and women (4.4%; 95% CI: 3.5–5.2). The likelihood ratio test found number of partners to be most significantly associated with chlamydia, followed by partnership duration, ≥ 1 concurrent partnerships (yes vs no), condom use (inconsistent vs consistent) and frequency of sex (daily/weekly/monthly vs less). The association was strongest for ≥ 1 concurrent partnerships (OR = 2.4; 95% CI: 1.7–3.4) followed by condom use (OR = 2.0; 95% CI: 1.3–2.9), partnership duration (OR = 0.5; 95% CI: 0.4–0.6) and number of partners (OR = 1.2; 95% CI: 1.1–1.3). Frequency of sex was not associated with chlamydia. When all variables were included in the model, condom use (OR = 2.1; 95% CI: 1.4–3.1) had the strongest association with chlamydia followed by partnership duration (OR = 0.5; 95% CI: 0.4–0.7), concurrent partnership (OR = 1.5; 95% CI: 1.0–2.3) and number of partners (OR = 1.1; 95% CI: 1.0–1.2), with the latter two highly correlated (p < 0.01).

Conclusion Sexual behaviour is difficult to capture accurately in questionnaires, but these results suggest that number of partners, partnership duration, concurrent partnerships and condom use are important. It is difficult to separate the effect of concurrency from number of partners.