Why have numbers of reported chlamydia cases been going up for at least a decade in many developed countries? In this issue, Rekart and Brunham (see page 87) and Miller (see page 82) debate whether or not the observed trend means that public health measures to control chlamydia are failing.1 2 Their opinions are “no” and “we don’t know”.

Rekart and Brunham argue that “arrested immunity” is the main explanation for the increasing trend.1 The starting point for this debate is their hypothesis that widespread early treatment has impaired the development of immune responses that would protect against reinfection, resulting in a paradoxical increase in population susceptibility to chlamydia.3 At a population level, they argue that this phenomenon would fit observed trends in reported cases from British Columbia, Canada and some European countries, where declines in the late 1980s and early 1990s have reversed and increased continuously since. Miller argues, however, that trends in routine surveillance data for chlamydia cannot provide the answer.2 He describes many factors influencing the numbers of reported cases, including changes in the numbers of people being tested and re-tested, diagnostic test performance and the risk profile of people being tested. Miller contends that a combination of factors could lead to the observed surveillance trend, even if the underlying prevalence of chlamydia infection in the population did not change.2

There are two fundamental flaws in the arrested immunity hypothesis. First, the main assumption is that widespread early detection and treatment successfully reduced population chlamydia rates in British Columbia. The mathematical model predicting the decline and rebound in chlamydia reproduced the surveillance data best when test coverage was 80% per year.3 Actual numbers of chlamydia cases reported each year in British Columbia are not compatible with this. Expected test coverage would be 10–20% per year if it is assumed that all infections and tests were in 15–34 year olds and population prevalence was 3–5%. Leaving 80–90% of infected individuals untreated is unlikely to affect either population level transmission or immunity. Second, the hypothesis relies on case reports being a true representation of chlamydia incidence. Rekart and Brunham discount alternative explanations because they do not fully account for the trend individually or because exceptions can be found.1 As pointed out by Miller, however, external factors rarely occur in isolation and a combination of these could plausibly reproduce the observed trend without the need for an additional immunological explanation. It has been previously suggested that declining chlamydia rates in the late 1980s and early 1990s were caused by a reduction in risky sexual behaviour in response to the threat of HIV/AIDS and safer sex campaigns.4 This would explain why the decline also occurred in countries without any organised chlamydia control activities.

See Editorial, page 79 and linked articles, page 82 and page 87

So, what do chlamydia surveillance trends mean? The answer is not much if they are interpreted without contextual information about changes in concurrent interventions, change in test volume, population coverage of testing, reporting coverage, diagnostic tests used, case definition and sexual behaviour of those being tested. These factors affect the interpretation of all surveillance data. They are, however, particularly influential when the data are dominated, not by clinical case reports, but by asymptomatic infections, which has become the case with most chlamydia detected by nucleic acid amplification tests. chlamydia case report data alone do not reflect either trends in prevalence or the impact of control measures. They should be used together with the many additional sources of epidemiological, behavioural and health services data to interpret current trends and guide future intervention measures and research.

The arrested immunity hypothesis is already being suggested as a likely explanation for paradoxical study findings without a critical appraisal of its theoretical basis, plausibility or appropriateness.5 6 In this issue, Johnson and Horner (see page 79) comment on the Finnish study, providing several behavioural and methodological explanations for an ecological association between decreasing chlamydia seroprevalence and increasing case reports.7 Atik et al concluded that antibiotic treatment of trachoma might be damaging, based on a sub-group analysis of a study in Vietnam where Chlamydia trachomatis reinfection rates increased in two communities treated with antibiotics but not in one community receiving surgery only.6 Mabey and others point out important methodological and immunological limitations of this interpretation.8 Host responses to treated and untreated C trachomatis infections are clearly important and the need for further immunological research is not in doubt. Intrinsic weaknesses in the arrested immunity hypothesis, however, mean that it is premature and counterproductive to prioritise vaccine development1 over other public health interventions to control chlamydia.