Sequential Antiretroviral Adherence

Dr Shuter et al reported the results of a study examining the stability of the adherence to antiretroviral therapy (ART) measured by Medication Event Monitoring System (MEMS) in 2 separate time points. ¹ It is very interesting to discern the extent that short-term adherence measured by MEMS technology predicts the adherence to antiretroviral (ARV) medications over the long run. Adherence to ART is the mainstay of HIV treatment since excellent adherence is associated with favorable viral suppression. ² Clinical trials and longitudinal studies usually provide the capacity to assess the adherence to ART over a short period of time (typically 24-week measurements); however, it is the long-run adherence that identifies the adequacy of treatment to achieve the desirable viral suppression and HIV clinical outcomes.  For this purpose, the authors compared the adherence measured by MEMS cap monitoring among a cohort of HIV-infected patients that participated in a study in 2004 to 2005 (study I) and the same cohort recruited again between 2008 and 2009 (study II) to repeat measuring ART adherence after 4 years. The authors showed that the mean adherence rates in the first and second periods were not significantly different and are, in fact, strongly correlated. They conclude that the ART adherence rates measured by MEMS caps are closely correlated at 2 time points several years apart. The findings of this study are remarkable; however, there are a few issues that need to be considered before drawing any inference.

First, seemingly patients with more favorable adherence to ARV medications and/or less-advanced HIV disease are more likely to be included in the study cohort. In study I, 84 patients were primarily enrolled and 64 completed the study. ³ The characteristics of participant and nonparticipants were assessed. Participants had a higher CD4 count compared with those who did not participate. In study II (2008-2009), of  64 participants who completed study I, authors included in the final analysis only 48 participants who managed to complete study II, whereas 16 (25%) were excluded (4 had died, 8 had been lost to follow-up [hence not recruited for study II], and 4 did not complete study II). The mean adherence rate in study I of the 16 participants who did not complete study II was not significantly different from the 48 participants who were included in the analysis. Nevertheless, it is not unlikely to contemplate that these individuals’ adherence and behaviors have changed over time from their initial adherence profile in study I. This potential change in adherence in turn can explicate, at least in part, their outcomes being mortality, loss to follow-up, or dropout. In addition, the proportion of patients meeting the criteria to be diagnosed with AIDS was 80% in the first period, while this proportion decreased to 75% in the second cohort. This is despite the fact that provided the “time elapsed since HIV diagnosis” was relatively high in the beginning of study I, it is not unreasonable to expect more, rather than less, AIDS cases after 4 more years. Exclusion of individuals with more advanced HIV disease and/or with less favorable adherence to ART might have had the sample in the second period mostly comprising “survivor” cases. Notably, there is a trend of decrease in adherence rate from study I to study II (74.2% vs 68.9%; P = .09), however, the proportion of virally suppressed patients (viral load <75 copies/mL) increased significantly from study I to study II (60.4% vs 79.2%; P = .05), which supports the abovementioned observation.

Second, 12 of 48 cases included in study II had switched to new drug regimens. The adherence rate of periods I and II were found to be significantly correlated among those who remained on their initial regimen (N = 36), whereas there was no significant correlation among those who switched to a new drug regimen (N = 12). New regimens are usually prescribed to those who do not achieve favorable outcomes with their initial regimen (in fact, 8 of 12 had detectable viral loads preceding switching to new regimens). These regimens confer less daily dosing frequency and hence confer the potential of increased adherence.

In summary, exclusion of some cases from the sample might have introduced biases, overestimated the adherence rate in the second study, and restrained its generalizability. Also, more robust statistical approaches besides correlation and quartile stratification may be employed to ascertain the agreement between the absolute adherence rates of 2 measurements.