How many doses can be missed[ix][ix]?

The most conservative answer to this question is `as few as possible', since any falls in drug levels will offer an opportunity for drug-resistant virus to replicate.

The extent to which drugs levels which become inadequately low as a consequence of one missed or incorrectly taken dose, will vary between individuals; in part because different people process drugs at different speeds; but also because there are differences between antiretroviral drugs themselves in terms of their metabolic characteristics. For example, indinavir levels fall much more sharply after eight hours than ritonavir levels, so the timing and frequency of indinavir dosing may be more critical than ritonavir dosing. Moreover, resistance to some antiretrovirals emerges more easily than for others. Efavirenz (Sustiva), nevirapine (Viramune) and 3TC (lamivudine, Epivir) are ineffective once HIV acquires just one mutation in its genes, whereas several others require a more complex pattern of mutations to emerge. These three drugs are amongst the slowest to leave the body once doses are stopped, but because HIV can become resistant to them so easily, it's possible that even very limited problems with adherence could allow them to fail.

The best evidence we have on this subject comes from large trials which follow people taking treatment for long periods, and measure their adherence using validated methods. The landmark study which is quoted most often was published by Paterson and colleagues in 2000 (Paterson 2000). This study involved people taking protease inhibitor-containing antiretroviral therapy, and measured their adherence by MEMScap, an electronic device fitted to pill bottles which records when the cap is removed. The study found that adherence levels above 95% were associated with superior virological and immunological response. 99 people were followed for a median of six months on treatment. Viral load failure was detected in 22% of patients with adherence above 95%; in 61% of patients with adherence levels between 80% and 94.9%; and in 80% of those with levels below 80%. Patients with greater than 95% adherence had fewer days in hospital than those with lower levels. To put this in perspective, 95% adherence is the equivalent of missing no more than one dose a week of a regimen dosed three times daily, or three doses a month for a twice-daily dose.

A more recently published study provided a longitudinal analysis of the effects of adherence levels on a range of treatment combinations, both protease inihibitor- and NNRTI-containing. This study is important because it demonstrated clearly that the consistency of adherence over time is integral to the success of HIV treatment. Eleven hundred people (a mixture of drug-naïve and drug-experienced individuals) began antiretroviral therapy within two large randomised trials. Their adherence was monitored via a self-complete questionnaire administered at regular study visits. Viral load was below 50 copies/ml in 72% of participants who reported 100% adherence at all four follow-up visits over twelve months, compared with 66%, 41%, 35%, and 13% of participants who reported 100% adherence at three, two, one, or zero follow-up visits, respectively (Mannheimer 2002).

A study where antiretroviral therapy was directly administered to prison inmates provided further evidence that 'perfect' adherence levels of 100% produce even better responses than levels below this (Fischl 2000).

A study of 116 people found that 85% of those with over 95% adherence had undetectable viral load compared with only 45% of those with less than 95% adherence after 48 weeks of treatment (Tuldra 2000).

Another study which confirmed the importance of 95% adherence found that although over 80% adherence may be associated with an initial virological response, less than 95% adherence was associated with virological failure at three months (Flandre 2002).

In addition to the number of doses taken, a recent study has also shown that taking doses on time every day results in lower viral loads, at least in patients taking their first antiretroviral drug regimen (Liu 2006). The study also questioned the concept of a 95% 'threshold' value for optimal adherence, finding that adherence below this level is often associated with good responses to treatment. However, this is no cause for complacency: the study showed that patients with higher levels of adherence had lower viral loads after a year's HIV treatment.

Contradictory findings - the need to consider plasma drug levels and the dynamics of drug resistance

Though these data indicate a clear relationship between high adherence and effective HAART, there may be some circumstances where even high adherence results in drug resistance.

Dr John Walsh and colleagues at the Chelsea and Westminster Hospital, London, reported their findings from a small study involving patients whose PI-containing HAART was failing. Blood plasma drug levels, genotypic and phenotypic resistance to protease inhibitors and adherence behaviour were examined in a group of prospectively recruited patients failing a combination containing either nelfinavir (16) or indinavir (18). Failure was defined as two consecutive viral load measurements above 1,000 copies. A control group of patients taking the same PIs but with three consecutive viral load measurements below 50 copies was also recruited.

Adherence was measured by pill counts, questionnaire and MEMScap. Plasma drug levels were measured two, four and six hours after observed dosing.

There was a strong association between genotypic resistance mutations associated with PI treatment and a high level of adherence (99%) in those experiencing viral load rebound, suggesting that continued high adherence was in fact maintaining drug resistance. Individuals with poor adherence (average 75%) were significantly more likely to have wild type virus or naturally occurring genotypic changes (polymorphisms) than PI-associated mutations.

Plasma drug level testing showed a significant association between viral genotypic resistance and a lower nelfinavir trough level. These patients may have poor drug absorption or rapid drug elimination; coupled with good adherence this would be expected to lead to frequently sub-optimal drug levels that would select for drug resistant virus.

On the other hand, in those with poorer adherence, adherence may be so low as to fail to drive the selection of resistance mutations.

Another study, conducted in San Francisco, used unannounced pill counts at patients' homes to monitor adherence in people with viral load that remained above 100 copies/ml after three months of stable HAART, and continued to remain above 100 copies/ml for a further six months. Drug resistance mutations were monitored six months apart, and those with the highest levels of adherence (but ongoing detectable viral load) had the highest risk of developing new resistance mutations (p=0.02) (Bangsberg 2002). These findings suggest that where adherence is good, adequate drug levels may permit the selection of drug resistant mutations, whereas if a drug is absent from the blood for up to 50% of the time, there is less risk of drug resistant mutants gaining an advantage, since they are less fit in the absence of the selective pressure exerted by the drug(s).

Another study, in 195 patients in Baltimore, found that the risk of resistance declined when adherence fell below 69%, presumably because drug levels were too low to select for drug resistant virus (Sethi 2003). The risk of clinically significant drug resistance was greatest in patients with adherence of between 70% and 89%, and in those who missed more than one clinic visit.

References are listed at the end of Interventions to improve adherence in Anti-HIV therapy: Adherence.