Evidence from clinical trials and observational studies in a variety of therapeutic applications supports the ability of FDC products to improve adherence to therapy, compared to separate pills.7
To our knowledge, this is the first study to empirically quantify the spillover effect of FDCs in improving adherence to another regimen component. Our results confirm the hypothesis that use of an FDC as the backbone of an ART regimen improves adherence to the third agent, compared to a regimen that utilizes separate pills as a regimen backbone. This improvement in adherence is in addition to the increase associated solely with the backbone component of the regimen.
While some evidence suggests that newer antiretrovirals and ART combinations may be more “forgiving” of poor adherence than older agents and un-boosted regimens, the standard for adherence in HIV remains high.4
In our population of managed care enrollees, patients receiving an FDC backbone were nearly 50% more likely to achieve an adherence rate of 95% or better to the third agent in the regimen. Even at the lower 90% threshold, FDC backbone use was associated with a nearly 40% increase in the likelihood of third-agent adherence. The magnitude of the FDC spillover adherence effect observed in our sample, while smaller than the approximately 300% direct effect estimated by Legorreta et al,8
is nonetheless statistically and clinically meaningful.
Results for adherence thresholds of 80% and 85% and for continuous adherence were also positive, but did not reach statistical significance. The inability to detect a spillover effect in these other model specifications may reflect the adherence behavior of our sample. Mean regimen adherence among the patients in this study was 88%, relatively high compared to other reports, though not unprecedented.21
Examination of adherence effects from FDCs in patient populations with lower baseline levels of adherence is needed to fully understand the spillover phenomenon.
Across the spectrum of available antiretrovirals, adherence is positively correlated with virologic suppression.4
FDC backbones, then, may play an important role in improving adherence and likelihood of suppression for the remaining regimen component(s), regardless of drug class. Additionally, adherence plays an important role in the development of drug resistance, although the specific relationships may differ across drug classes.22
As such, the spillover adherence effect of FDCs may be especially clinically meaningful for PI-based regimens. Studies examining the relationship between adherence and resistance suggest that the development of primary and secondary PI mutations peaks at levels of adherence just below those required to maintain virologic suppression.22
For example, among patients with relatively high overall adherence, even small, absolute improvements in PI adherence of only a few percentage points may be sufficient to substantially reduce mutation development. The use of an FDC backbone may be one tool for achieving such an improvement in PI adherence.
The use of administrative records to examine the spillover effect of FDCs offers several advantages over other data sources. First, these data reflect the experiences of patients in actual clinical practice, potentially avoiding the adherence bias associated with clinical trials.24
Second, these observational data allow for a relatively large sample size, thereby increasing the ability to detect differences between study groups. Additionally, while not a nationally representative sample, the IHCIS data include individuals from all regions of the US. To the extent that heterogeneity in adherence is correlated with regional heterogeneity in HIV risk, demographic characteristics, and other factors that may affect adherence, the generalizability of our results is improved relative to smaller studies of more homogeneous cohorts.
There are differences between the IHCIS sample, and the US HIV population as a whole. In 2006, approximately 73% of individuals living with a diagnosis of HIV or AIDS in the US in 2006 were males, compared to nearly 83% in our study.25
During the same time period, roughly 38% of persons living with HIV/AIDS were aged 18–34 years, compared to less than 15% in our sample. The older age and increased proportion of males observed in our sample may reflect an increased access to health insurance for this group compared to other individuals diagnosed with HIV/AIDS. It is possible that spillover effects measured in different populations may differ from those seen in the IHCIS sample, especially if factors associated with insurance status, such as education and income (which are unobserved in the IHCIS sample), also influence adherence behavior. While socioeconomic status does appear to be associated with adherence in many chronic diseases, such relationships have not been documented conclusively for HIV.26
As with other observational studies, this work is subject to limitations which should be considered when evaluating the results. The retrospective nature of the data source raises the possibility of selection bias because the assignment of patients to study groups is not random, as in a clinical trial. For example, if individuals who received Epzicom were more likely to be adherent to a third regimen component than those who received a backbone of separate NRTIs, the impact of FDCs on spillover adherence would be overestimated. Conversely, if physicians who were particularly concerned with patient adherence tended to prescribe FDCs to that end, then the spillover effect observed here may be underestimated. It should be noted that the study inclusion requirement of at least one refill in the 60 days following the index prescription – necessary to capture the intent to treat with a specific agent – may bias the sample toward more adherent patients. In such a case, the ability of the FDC to improve adherence even among those already adherent is notable. Still, further efforts are needed to understand adherence among very poor adherers.
As noted in , subjects in the two study groups were similar in terms of age and gender, but other differences did exist. We controlled for all the factors listed in in our statistical analysis. We did not have access to subjects’ complete HIV treatment histories and could not definitively identify those who were completely naïve to treatment. While there is some evidence that adherence differs between treatment-experienced and treatment-naïve patients,27
the likelihood that the marginal adherence effect estimated here would differ between these two groups of patients was less likely and, indeed, was what we observed. There may exist other factors for which we were unable to control. If those factors were associated with adherence differentially by study group, the study results may be biased.
Epzicom dosing is labeled as once per day, while most other NRTI combinations used during the study period were dosed twice per day, per product labeling. Thus, some of the spillover attributed to the FDC group may be due to reduced dosing frequency. However, some of the combinations in the NRTI Combo group included one or two agents that could be dosed once per day. Interestingly, patients in the NRTI Combo were more likely to receive an NNRTI as the third regimen component. Compared to PIs, NNRTIs have been associated with improved adherence.28
The approach we use to construct the follow-up period over which adherence was assessed is subject to the inherent limitation that some patients may discontinue study therapy before expiration of the last observed refill. Since discontinuation in that case cannot be observed in automated pharmacy claims, our method may lead to follow-up durations that are slightly overestimated for some patients. Similarly, no gold standard for measurement of adherence exists. Adherence levels measured from pharmacy records have been shown to be higher than those measured by electronic monitoring, although the two measures are highly correlated.29
Importantly, any overestimation of adherence would be attributable to both study groups, minimizing its effect.
This study examines only the use of an FDC containing two antiretrovirals and intended to be combined with at least one additional agent. Due to sample size constraints, we were unable to assess whether spillover effects extend to all components of regimens that consist of an FDC plus two or more antiretrovirals taken as separate pills. FDCs containing three antiretrovirals are also available and can be used alone or as part of a larger regimen in combination with other antiretrovirals. It is unclear, a priori, whether such effects would be greater or lesser than those seen with two-agent FDCs in a typical, three-agent regimen. We have shown that adherence to a third regimen component is increased with the use of an FDC backbone and that this adherence level is high. As a result, in populations where medication-taking behavior is relatively high, triple-agent FDCs may not offer much of an adherence advantage over dual-agent FDCs. Additional research on the direct and spillover adherence effects of dual-agent and triple-agent FDCs as components of regimens consisting of more than three antiretrovirals is warranted.
Successful HIV therapy requires individuals to maintain nearly perfect adherence, and clinical practice guidelines have long recommended that physicians consider adherence when selecting regimen components. This study highlights an additional adherence advantage associated with the use of FDCs that has not previously been identified: a spillover effect on a non-fixed regimen component. While additional efforts are needed to assess the extent of adherence spillover associated with other FDCs, our work supports the use of fixed-dose NRTI backbones as a means of encouraging adherence to the entire ART regimen.