This increased risk peaked in the first 6 months after individuals started ART and then gradually declined. Immune reconstitution inflammatory syndrome (IRIS) is a possible explanation for this observed initial increase in risk. When ART first became available in this cohort, individuals starting ART would have included those with advanced HIV infection and low

CD4 cell count, who were therefore at increased risk of IRIS. Those commencing ART were also seen more frequently Kinase Inhibitor Library mouse in clinical follow-up, especially during the first 6 months, and hence were more likely to have HIV-related illnesses diagnosed in this early period compared with the later periods. Strengths of our study include the long follow-up period, the general population source, the high levels of follow-up (93% in seroconverters), and the availability

of an estimated date of HIV seroconversion. Taken together, these features of the study enabled us to estimate find more rates of WHO stage-defining diseases before and after ART introduction. Most previous studies in developing countries have been limited to cohorts of prevalent HIV cases with no known HIV seroconversion dates. There are also several limitations to our data. Firstly, although the date on which an episode of morbidity commenced was documented, there was no documentation of when it ended. The time ‘not at risk’ of future episodes STK38 while experiencing an episode may

have been under- or overestimated, and may have influenced our incidence rates. However, the same criteria were used in all follow-up periods, and while on or off ART, so this is unlikely to have biased our measures of effect. Secondly, diseases requiring invasive diagnostic procedures and histology such as lymphoma and cytomegalovirus infections were not documented in this cohort, so our overall rate of any WHO stage-defining disease may be an underestimate, as was also observed in an earlier study in Cote d’Ivoire [10]. The use of cotrimoxazole may be an alternative explanation for the reduction in morbid events following the introduction of ART, or may explain the residual trend with calendar time after adjusting for the use of ART. Though cotrimoxazole prophylaxis was prescribed for all HIV-infected participants, we did not adjust for its effect on morbidity in this analysis. The first edition of the National Policy guidelines for cotrimoxazole prophylaxis was issued in 2005 [18], but we did not have a separate code in our database for cotrimoxazole prophylaxis until 2008. The slightly higher response rates for male than female subjects may have resulted in a slight underestimation of our incidence rate, as female subjects had a slightly higher rate of acquiring any WHO stage disease than male subjects (adjusted HR 1.35; 95% CI 0.97–1.9).