Background:  To date, interpretations of data suggest that the highest levels (>95%) of ART adherence are key to viral suppression. However, differences by regimen in the amount of adherence needed to maintain suppression are often ignored.

Methods:  We assessed the association between adherence to different regimens and viral suppression in 1634 British Columbians in the HOMER cohort with ≥2 consecutive viral loads <500 copies/mL. Time 0 was the first of at least 2 consecutive viral loads <500 copies/mL. Event date for breakthrough was the second of 2 viral loads ≥1000 copies/mL and non-events are censored at the time of last viral load <1000 ccopies/mL until October 30, 2004. ART adherence over every 2 dispensing periods was calculated as: number days of medication was supplied by number of days between refill dates. It is a time-updated measure dichotomized as ≥95% vs <95%. Primary outcome was time to viral breakthrough and models were constructed separately for protease inhibitor (PI), non-nucleoside reverse transcriptase inhibitor (NNRTI), and boosted-PI initial regimens. Cox proportional hazard regression analyses included baseline variables:  sex, age, injection drug use, CD4 count, AIDS diagnosis, and physician experience.

Results:  The participants included 1376 (84%) males and 383 (23%) subjects had a history of injection drug use. The median age was 39 years (IQR 33 to 46), with a median CD4 count of 200 cells/mm³ (IQR 80 to 350) and median follow-up of 29 months. A total of 752 (46%) initiated a PI-, 631 (39%) an NNRTI-, and 251 (15%) a boosted PI-based regimen. Overall, 606 (37%) subjects had viral breakthrough. In adjusted analyses, <95% adherence was most strongly associated with breakthrough with PI (HR 1.78; 1.41 to 2.24), followed by NNRTI (HR 1.47; 1.01 to 2.14). In contrast, there was no association between <95% adherence and breakthrough for boosted PI (HR 1.05; 0.46 to 2.42).

Conclusions:  The lack of an association between adherence and maintained suppression at a threshold of 95% adherence suggests that boosted PI are more forgiving of non-adherence than either unboosted PI or NNRTI. These results may be explained in part by higher trough concentrations with respect to the IC₉₀ of the virus, higher genetic barrier to resistance of PI vs NNRTI, and prolonged half-lives due to the ritonavir boosting. For individuals at high risk of suboptimal adherence, using boosted PI-based regimens may achieve higher effectiveness rates. Adherence studies should address these differences between regimens in the evaluation of their results.