The level of adherence to antiretroviral therapy is an important predictor of success in the treatment of HIV infection, with earlier studies suggesting that 95% adherence was necessary to optimize response.

Whether this level of adherence is required for treatment success with current therapies, including potent ritonavir-boosted protease inhibitor-based regimens, has been questioned. In addition, the degree of adherence needed to achieve plasma HIV RNA (viral load) suppression might differ from that needed to maintain viral load suppression.

In the current study, presented at the 11th European AIDS Clinical Society (EACS) meeting this week in Madrid, Spain (October 24-27, 2007), researchers investigated the relationship between the level of adherence required first to achieve and then to maintain viral load suppression.

M99-056 and M02-418 were similarly designed prospective, randomized, parallel arm, 96-week trials evaluating the safety and efficacy of lopinavir/ritonavir (Kaletra) dosed once or twice in antiretroviral-naive HIV positive subjects.

The studies differed with respect to sample size and NRTI backbone: patients in study M99-056 (n=38) used d4T (stavudine; Zerit) plus 3TC (lamivudine; Epivir) plus twice daily lopinavir/ritonavir, while those in study M02-418 (n=190) used tenofovir (Viread) plus emtricitabine (Emtriva) plus once-daily lopinavir/ritonavir.

Plasma HIV RNA was measured every 4 weeks from baseline to

Week 24, every 8 weeks from Week 24 to Week 48, and every 12 weeks from Week 48 to Week 96. For the analysis, viral load suppression was defined as the first of 2 consecutive viral load measurements < 50 copies/mL after initiating treatment. Loss of viral load suppression, or viral rebound, was defined as at least 1 viral load measurement ≥ 50 copies/mL after initial viral suppression.

MEMS monitors (electronic caps for medicine containers that record how often the bottle was opened) were used to compile dosing histories for lopinavir/ritonavir. Subjects underwent a 5-14 day placebo lead-in period during which instruction on use of the MEMS monitors and feedback about adherence were provided.

“Taking adherence,” defined as the percentage of prescribed doses taken, was used to summarize adherence. In evaluating the association between the time to viral load suppression and adherence to lopinavir/ritonavir, “taking adherence” was estimated from the time the subject initiated treatment until the initial detection of viral load suppression, or the subject’s last visit in cases where the subject never achieved full viral suppression.

In relating adherence and time to viral rebound, “taking adherence” was estimated from the subject’s dosing history during the 30-day period before a viral rebound was detected, or the subject’s last visit in cases where viral load rebound did not occur.

The pattern of the relationships between “taking adherence” and either time to viral load suppression or time to viral rebound was assessed using Cox proportional hazards models with smoothing splines used to transform the explanatory variable.

As output from this model, the relationships between the log hazard ratio (a relative measure of the rate of viral suppression/rebound events at a certain time) and “taking adherence” were plotted. The threshold of “taking adherence” above which there was no further improvement in the log hazard ratio of virological suppression/viral rebound was visually identified. The significance of this threshold was further tested using a Log-rank test.

Results

Of 228 subjects enrolled, 214 (94%) were evaluable and 189 (83%) achieved viral load suppression at some point during the 96-week studies. Demographic and baseline clinical characteristics were similar for subjects who received the once-daily and twice-daily regimens.

The model suggested a plateau of log hazard ratio near a “taking adherence” of 85%, indicating the percentage of prescribed doses that need to be taken over time to achieve viral load suppression.

Subjects were then stratified post-hoc into 2 distinct groups using their estimated “taking adherence” over time (i.e., < 85% vs ≥ 85%). After stratification, subjects with “taking adherence” ≥ 85% were more likely to achieve viral load suppression than those with “taking adherence” < 85% (Log rank test P = 0.001).

Of the 189 subjects who achieved viral load suppression, 65 (34%) subsequently experienced viral rebound with at least 1 viral load ≥ 50 copies/mL.

“Taking adherence” was measured during the 30-day period prior to viral rebound (or for 30 days prior to the last visit if rebound did not occur). For subjects with missing adherence data around this time, 2 analyses were performed: patients who experienced a “non-monitored” period of adherence during this time were excluded from the analysis (n = 15); and patients who experienced a non-monitored period during this time were included, but assumed to have “taking adherence” of zero.

The model suggested a plateau in the log hazard ratio near a “taking adherence” level of 75%, indicating the level of adherence necessary to maintain viral load suppression in those subjects who previously achieved viral load suppression. The shape of the log-hazard function was similar regardless of the assumptions made with respect to “taking adherence” during non-monitored periods.

Subjects with “taking adherence” < 75%  were more likely to experience at least one viral load measurement ≥ 50 copies/mL than those with “taking adherence” ≥ 75% (Log rank test P=0.015).

Discussion

Previous adherence research has mainly focused on the percentage of prescribed doses taken to provide an explanation for untoward clinical events that arise from patient non-adherence.

Unfortunately, the percentage of prescribed doses taken is a one-dimensional expression of dosing history data that excludes the important dimension of time: when doses were missed and the length of sequential omissions of doses (“drug holidays”).

As an example, the clinical impact of missing 8 single doses in the course of a 2-month period is probably considerably less than that of a drug holiday of 8 consecutive days of missed doses. Yet each of these 2 patterns of dose omission produces the same percentage of prescribed doses taken. Given the documented diversity of temporal patterns of dosing errors, there is a clear need to understand how these various types of error impact viral load evolution and emergent drug resistance.