Abstract Body

Data created during the care continuum are challenging to assemble, and disparate sources may account for varied results in observational studies. To assess the limitation of one source, we contrasted adherence, duration, and regimen composition between prescription (RX) and pharmacy dispense (PD) data generated during care of PLWH.

Antiretroviral (ARV) RX and PD data were obtained for 1270 treatment-experienced PLWH from the TRIO network, consisting of 11 HIV treatment centers servicing 39 US states. Follow up was ≥12 months (m) post index, defined as the first ARV regimen switch between 2014 to 2017 with final data collection Jun 2019. Adherence was based upon proportion of days (d) with all drugs. Regimen discontinuation was dated at exhaustion of all regimen components and/or upon addition of a new ARV drug. Time to discontinuation was assessed by Kaplan-Meier with log-rank. Univariate analyses were via chi-square, exact, or T-test.

Discontinuation rates (46% RX v 43% PD, p=0.060) and median time to discontinuation (29 m RX v. 29 m PD, p=0.448) were not significantly different by data source, though time to discontinuation/censoring differed by >90 d (+/-) for 29% (374) of PLWH, with 20% (258) discontinuing therapy >90 d before the end of the RX-based regimen [FIGURE]. ≥80% adherence was calculated for 90% (1143) PLWH based on RX v 92% (1166) PD (p=0.129) and ≥95% adherence for 86% (1087) RX v 87% (1110) PD (p=0.202). Of PLWH with <80% adherence by PD, 28% (29/104) were classified with ≥80% adherence by RX. Conversely, 41% (52/127) patients classified with <80% adherence by RX had ≥80% adherence by PD. Changes in multi-tablet regimen (MTR) due to early discontinuation of a component (>90 d before discontinuation of remaining regimen drugs) were indicated in 16% (75/478) PLWH by RX and 13% (63/478) by PD (p=0.311). Of PLWH with a change in MTR by PD, 14% (9/63) were not reflected by RX as having any early drug discontinuation and an additional 37% (23/63) were indicated as having a change that differed in time >90 d from observed by PD. In total, 37% (473/1270) of study patients had one or more of these differences in duration, adherence, and/or MTR composition.

These data suggest a lack of concordance between what is prescribed and dispensed for over a third of PLWH. As dispensing data are more likely to reflect actual treatment, observational studies should include this information whenever possible.