Background Although the importance of age in the prevalence of Chlamydia is well recognised, its importance in the relationship between Chlamydia and pelvic inflammatory disease (PID) has received little attention in epidemiology.

Methods We generate and compare several sets of estimates of the population attributable fraction (PAF) of PID due to chlamydia by age-group using a number of data sources. Estimates are obtained using data from case-control studies and Chlamydia population prevalence in England. A second set of estimates is obtained from data on the incidence of PID, the incidence of Chlamydia, and the risk that a Chlamydia infection causes PID. We estimate the incidence of all-cause PID by age in England from routine data sources, and evidence on the proportion of PID episodes that are diagnosed. We synthesise these data with data from the control arm of the POPI trial. We estimate Chlamydia incidence by age in a multi-parameter evidence synthesis of studies of Chlamydia incidence, prevalence, and duration of infection. Finally we estimate the risk of PID following Chlamydia from a statistical synthesis of randomised controlled trials using a multistate model. A third estimate of the PAF is obtained for the POPI trial population.

Results Prospective estimates of the PAF fall from 50% (15%, 100%) in women aged 16–19 to 20% (6%, 49%) in women aged 35–44, and estimates from retrospective data drop from 34%(17%,55%) to 6% (2%, 14%). Changes with age groups are most likely due to changes in the aetiology of PID, but part of the effect would be explained if the risk of PID due to CT, or proportion of PIDs that are diagnosed, increases with age.

Conclusions The PAF of PID due to Chlamydia reduces dramatically with age. More attention needs to be given to age when designing and reporting results from epidemiological studies.