Understanding the burden of Chlamydia trachomatis in populations is based primarily on studies of prevalence of current infection measured using sensitive and specific nucleic acid amplification tests on genital or urinary specimens. Such studies have produced widely varying results, but most have emphasised the high prevalence in young sexually active women, with fewer studies in men.1 Studies have frequently been based on clinic populations, which suffer from problems of selection bias (towards those at higher risk) and limited demographic and behavioural data. The few studies based on random and representative population samples have generally demonstrated lower prevalence and increased risk linked to increasing numbers of sexual partners.2 Studies of chlamydia incidence in population samples are few and far between. Data from national surveillance programmes consistently show rising rates of diagnosed infection in most developed countries over the last decade. But this cannot measure true incidence of infection because they are biased by the introduction and wider spread use of new sensitive tests leading to increased diagnosis of asymptomatic infection. All these studies measure current infection. They cannot account for past exposure or measure cumulative risk of infection over time that screening programmes aim to minimise.

In the last issue, Lyytikainen et al reported on Chlamydia trachomatis antibody seroprevalence in a stratified probability sample of 8000 sera from a large Finnish population serum bank.3 Participants were women under 29 years, having at least two pregnancies. The paper illustrated an approach that could potentially be used more widely. The paper assessed prevalence of past exposure to genital C trachomatis and explored changes over time by comparing age-specific prevalence at different time points.

Lyytikainen et al are able to demonstrate apparent clustering of infection in large cities but were surprised to find a declining prevalence of C trachomatis antibody between 1983–1989 and 1997–2003. They postulated that this may be attributable to more intensive sexually transmitted infection (STI) testing and possible earlier diagnosis resulting in poorer humoral antibody response. Other explanations are also worth exploring. They included possible bias in the sample selected (two pregnancies required for inclusion), uncertainty about the methods of stratification in sample selection and how this was dealt with in the analysis and an unexplained smaller sample size in the later cohort. Changing fertility patterns over the 20 years of the study may have altered the characteristics of those having children under the age of 30. The possible role of sexual behaviour change needs consideration. STI rates dropped markedly in most European countries from around 1983 in response to behavioural change in the face of the HIV epidemic, and started to rise again only in the mid-1990s when risk behaviour began to increase again.4 These changes may have been sufficient to reduce cumulative risk of exposure over time to chlamydia while being consistent with apparent rising current incidence of new diagnoses.

The authors rightly highlighted the difficulties of interpretation but their approach is novel. Could the method be more widely used in exploring chlamydia transmission dynamics and specifically the impact of chlamydia prevention and screening programmes in populations? We note three requirements for C trachomatis seroprevalence studies to be more widely applicable.

First, we need confidence in the sensitivity and specificity of available antibody tests. There is no single sensitive and specific C trachomatis antibody test that has been consistently used to investigate chlamydial serology.5 In addition, current assays have not generally been rigorously evaluated for sensitivity and specificity against large samples of well-characterised sera from patients ever exposed to C trachomatis compared to those never exposed.

The majority of commercially available assays have limitations, which can include variable sensitivity and specificity and cross-reactivity with Chlamydophila pneumoniae—a common respiratory pathogen. C pneumoniae antibody prevalence increases from a few per cent in pre-school children to 20–30% in teenagers and 40–60% in adults.6 C pneumoniae and C trachomatis are from the family Chlamydiaceae, and the genomes contain many similar genes (orthologues), including ompA, the major outer membrane protein (MOMP) gene. There is considerable amino acid identity (average 62%) between proteins encoded by orthologues from C trachomatis and C pneumonia, and thus the potential that these proteins may contain cross-reactive epitopes.7 However, cross-reactive C pneumoniae anti-MOMP antibodies with C trachomatis MOMP appear to be uncommon.8 9 At least three companies have developed C trachomatis antibody specific enzyme immunoassays (EIA) tests using synthetic peptides derived from C trachomatis specific regions of MOMP. Nevertheless, caution needs to be exercised given the DNA homology and the evidence that some C trachomatis MOMP peptides are recognised by individuals previously exposed to C pneumoniae.7 8 A study by Land et al, which evaluated three assays,10 suggested that cross-reactivity may occur with at least two.

The actual sensitivities might be lower than those quoted by manufacturers when used to assess previous exposure in population-based studies. Assays have been characterised using sera from chlamydia detection-positive individuals from clinical settings who may be more likely to develop antibodies than asymptomatic individuals in the community. Although little is known about persistence of antibody titres following acute infection, there is evidence that some initially chlamydia antibody-positive people may become antibody negative.11 How often and when this happens needs further investigation. Assay specificity can also be affected by the selection of negative controls. Specificity will be underestimated by assays that use controls who are nucleic acid detection-negative but might have been previously exposed and subsequently resolved the infection.12 Specificity might be overestimated by assays developed using sera from pre-school children with low exposure to C pneumoniae as controls.

Second, it is important to have studies that include both men and women if population impact is to be assessed. Pregnant women are a selected group of the population and may have lower rates of partner change than the population as a whole, although the completeness of coverage and availability of blood samples is a clear advantage. Similar serum banks from men are generally not available.

Third, availability of more extensive demographic and behavioural data linked to C trachomatis serology status would greatly enhance interpretation and extend use of the approach. The method could usefully be applied to the large-scale repeated population health surveys available in some countries. The US NHANES study13 has usefully explored the changing epidemiology of HSV2 in this way. In the UK, the Health Survey for England might be an appropriate vehicle but does not currently explore infectious disease epidemiology.14

C trachomatis antibody has been overlooked as a potential research tool for understanding chlamydia transmission, perhaps because the test had limited application in clinical practice. If test characteristics can be better quantified and improved, antibody tests have the potential to be a highly informative tool in better understanding chlamydia transmission and pathogenesis. Used in this way they may improve our ability to assess the public health impact of current efforts to reduce the population burden of genital C trachomatis.