Prevalence-incidence bias in longitudinal studies of rheumatic heart disease in Fiji.

backgroundRheumatic heart disease (RHD), a consequence of a dysregulated immune response to Streptococcus pyogenes infection, remains the most commonly acquired cardiovascular condition in under 25-year-olds. Predominantly occurring in low- and middle-income countries (LMICs), RHD has an estimated prevalence of 40.5 million cases globally, although RHD complication rates in endemic settings remain uncertain. Accordingly, we undertook a prospective cohort study of RHD mortality and morbidity in Fiji with the aim of comparing complication rates to those estimated in our previous retrospective studies based on routine data and record-linkage from this setting.

methodsWe prospectively ascertained RHD patients in the Central Division of Fiji from: (i) prevalent cases reviewed in echocardiography clinic prior to the study (May 2014-September 2015), and (ii) incident cases diagnosed in hospital (May 2014-October 2016). The primary endpoint comprised the earliest of RHD-attributable death, new onset heart failure or new onset stroke (assessed to October 2016). Differences between groups were assessed using a log-rank test (LRT). Additionally, to further investigate the issue of prevalence-incidence bias, rates of these events in the prospective cohort were also compared to those among patients with similar characteristics from our previously reported retrospective studies.

resultsDuring a median of 19.3 months follow-up, the primary endpoint occurred in six out of a total of 163 patients recruited to the study, with an incidence rate of 1.8 (95% CI, 0.6–5.5) and 5.3 (95% CI, 1.7–16.5) per 100 person-years in the prevalent and incident groups respectively (LRT, p = 0.13). Among 617 comparable individuals from our previous retrospective studies, the primary endpoint occurred in 24 patients with an incidence of 1.4 (95% CI, 0.7–2.5) and 6.4 (95% CI, 3.8–10.8) per 100 person-years in prevalent and incident groups respectively (LRT, p = 0.002).

conclusionsComplications of RHD were less frequent in the prevalent compared to incident groups, reaching statistical significance in the larger retrospective cohort. Reflecting the impact of systematic differences between the groups, these findings highlight the vulnerability of longitudinal studies of RHD to prevalence-incidence bias, underscoring the need for further robust and representative studies of RHD outcomes in LMICs to inform efforts to limit the impact of RHD globally.