We have demonstrated that short term adherence measured by pharmacy refill was the strongest predictor of VF on second line ART. Short term pharmacy refill adherence was also associated with virologic suppression with an “adherence dose response” relationship. Our study is one of few which have evaluated adherence on second line ART and is novel in that adherence was measured over the short term. We found that the “interval gap” method, which is a method not previously used for calculating adherence from ART pharmacy refills, outperformed the usual methods that average adherence over an interval. We observed a trend towards an association between longer duration of second line ART and risk for VF, but this was not statistically significant. We found an association between lower CD4 count at the time of starting second line ART and VF on second line ART, which adds support to data showing that second line outcomes are improved with early detection of failure of first line ART and prompt initiation of second line ART before immunological deterioration [
18].
A previous report of an earlier cohort from the same clinic as the present study, [
9] reported an association between virologic suppression on second line ART and adherence measured by pharmacy refill over the long term since second line ART initiation. Pharmacy refill as an adherence measure over shorter time periods is more pragmatic and implementable [
19]. In our study short term “interval gap” refill performed similarly to long term “overall” refill on ROC analysis, and out-performed other methods of determining short term adherence using pharmacy refills. We explored the ability of differing durations of pharmacy refill from three to twelve months to predict VF: “interval crude” and “interval average” performed better with longer durations, but the best performing method, “interval gap”, performed similarly over all of the interval durations assessed. Grossberg et al. demonstrated that a 90 day period of pharmacy refill was associated with VL change, [
19] but refill periods shorter than 60 days may overcall imperfect adherence leading to unnecessary clinical interventions [
17]. We found that 80% adherence by pharmacy refill over 4 months appeared to be a threshold for predicting virologic suppression (Figure
2). However, there were small numbers of patients in the lower adherence strata, which limited our ability to determine a threshold. Others have reported an increased risk of VF with adherence <80% in observational studies of patients on boosted PI regimens [
8],[
20]. A threshold of 80% adherence measured by pharmacy refill in the previous 4 months could be used to identify patients needing enhanced adherence support and rationalise use of VL testing in resource-limited settings. Most patients on second line ART experiencing VF were able to achieve virologic suppression with intensified adherence support in a study at a clinic in South Africa [
21]. VF on second line ART is likely a result of poor adherence rather than resistance as several studies have found a low proportion of major PI mutations in patients with VF on second line ART [
10]-[
12]. Unfortunately, as a result of high cost, the routine use of GART in patients with VF on second line will not be widely available in most low-middle income countries. Van Zyl et al. [
12] suggested an algorithm to select patients in VF on second line for GART using lopinavir plasma and hair therapeutic drug monitoring. However, these pharmacokinetic measures are costly (although less costly than GART) and have extremely limited availability in resource-limited settings. By contrast, short term adherence measured by pharmacy refill can be easily implemented, especially in clinics with electronic dispensing, without incurring large additional costs.
Our study has several limitations. First, the rate of VF on second line in the McCord ART clinic was lower than reported in a recent systematic review of second line treatment outcomes in resource limited settings, possibly due to a high physician/nurse to patient ratio and reliable antiretroviral drug supply [
2]. Therefore the findings may not be generalisable to public sector ART clinics in other settings. Second, pharmacy staff shortages in clinics with large patient numbers may be unable to apply the “interval gap” method to calculate adherence. However, the “interval crude” method, which is easy to calculate, could be instituted in pharmacies with staff shortages or those that only keep manual pharmacy refill records. Third, we excluded patients with zero adherence and suppressed VL on the grounds that they must have been collecting ART at another clinic. However, we had no way of determining this and it is possible that other patients may also have collected ART at other clinics, which would weaken associations between adherence and the virologic outcomes we assessed. Fourth, we lacked power for some of the associations we assessed, notably the duration of second line ART and risk for VF, and the determination of an adherence threshold for virologic suppression.