Accurate information on morbidity after HAART initiation in low-resource settings is still needed to improve treatment and monitoring guidelines. The present study presents the incidence of new ADIs in such a setting and the main predictors of ADI occurrence: CD4 cell count and viral load.
The ADI incidence rates observed in the Senegalese cohort were consistent with those seen in other low-resource settings [10, 39] but higher than those seen in industrialized settings [5, 20]. The most frequent ADIs that occurred during HAART were similar to those seen in other low-resource settings, tuberculosis being among the leading ones [6, 7, 10]; its estimated incidence in the Senegalese cohort was comparable to that reported in other similar studies [39, 40]. Besides, the rate at which the incidence rate of ADIs decreased was different between ADIs, which is also consistent with previous observations [2, 5].
A decrease in the ADI incidence rate over the first years after HAART initiation has been observed in other low-resources settings [1, 4–7, 10, 19, 41]. Interestingly, similar rates of decrease in the ADI incidence were found among patients with different CD4 levels while these rates were different among patients with different virological responses. This result underlines the importance of the initial virological response. Also, patients who already experienced an ADI before starting HAART had an increased risk of new ADIs during the first four years of treatment, indicating that antiretroviral therapy should be initiated well before the occurrence of a first ADI.
Despite the decrease in the ADI incidence rate during the first four years on HAART, an overall increase was observed afterwards, which was mainly due to the high incidence rate of ADIs in patients with less than 200 CD4 cells/mm3 and a VL persistently above 1,000 cp/mL. Moreover, after four years on HAART, that decrease persisted only in patients with more than 350 CD4 cells/mm3. These results indicate that the association between the CD4 cell count and the occurrence of ADIs changes with time (non-proportional effect) and that the CD4 cell count presents an increased predictive value after 4 years on HAART. It should be noted that, during the second period, the current CD4 value was a stronger predictor of ADI than the individual variation in the CD4 slope.
The effect of the virological response on the occurrence of new ADIs could be separated into a direct effect independent of the CD4 response and an indirect effect mediated by the effect of the virological response on the CD4 response. This may explain the two-fold decrease in the rate ratio associated with the current viral load between the univariate (marginal) and the multivariate (conditional) model.
The role of the viral load in monitoring AIDS patients after HAART initiation in low-resource settings is still not sufficiently clear or recognized [22–24]. One finding of the present study is that, along with the CD4 cell count, the viral load was an important predictor of the occurrence of new ADIs, especially during the first years of ART. We have also found that, after 42 months, ADIs could occur in patients who are immunologically stable but have a detectable viral load. These results suggest that the viral load may complement the CD4 cell count in identifying patients with high risks of disease progression. The treatment strategy in such patients should not be limited to a change in antiretroviral drugs but extended to an intensive and comprehensive clinical care that enhances adherence to the treatment.
The present study has several strengths and limitations. Despite its limited size, the study cohort boasts a long follow-up period and few lost to follow-ups. Regular visits with clinical examination allowed frequent information on new ADI occurrence. However, it must be noted that, in our setting, the diagnostic procedures were more limited than in industrialized settings. It is therefore likely that the incidence of some ADIs, such as pneumocystosis, that require advanced diagnostic procedures, could have been underestimated and some misclassification bias introduced. However, those diagnostic procedures did not change over time, protecting from a differential misclassification with time.
Another limitation could have been the lack of information on cause-specific mortality. However, the outcome of this study was restricted to the occurrence of ADIs. The dependence between mortality and morbidity could have biased the estimations of the rate ratios but this was taken into account by weighting the models with the inverse probability of surviving.