Several factors determine class-specific adherence–resistance relationships. First, antiretroviral regimen potency is important, as individuals with very low levels of viral replication are unlikely to develop resistance. Second, in the setting of viremia, circulating viral populations are determined by the interplay of the fold-change in resistance and fold-change in fitness caused by drug resistance mutations. Third, the genetic barrier to antiretroviral resistance determines the rate of development of resistance mutations at levels of drug exposure that favor resistant over wild-type virus. During multidrug therapy, differential drug exposure increases the likelihood of developing resistance. Long half-life drugs, in the presence of short half-life drugs, may be particularly susceptible to the development of resistance at low-adherence levels due to periods of differential drug exposure during intermittent dosing. Finally, antiretroviral medications with higher potency and higher genetic barrier to resistance decrease the incidence of resistance for companion antiretroviral medications.
The complexities of adherence–resistance relationships are related to characteristics of the virus, the medications, and to their interactions. Despite this complexity, adherence–resistance relationships have been consistent using diverse methods of adherence assessment (electronic prescription bottle caps, pill-count, self-report, or pharmacy refill data), study methodology (cross-sectional or prospective), and type of resistance testing (genotypic or phenotypic).
It is also important to understand the type of study when evaluating adherence–resistance relationships. Incident resistance describes new resistance mutations accumulating over time in individuals initiating antiretroviral therapy. Prevalent or cross-sectional resistance describes resistance present in individuals at the time they fail antiretroviral therapy. Both perspectives are useful in settings with limited availability of resistance testing, such as in many resource-poor settings, and in resource-rich settings in which loss to follow-up, transfers of care, and cyclical engagement in healthcare are common [48
The World Health Organization supports a public health approach for the treatment of HIV infection [49
], which necessitates that salvage therapy for a population be chosen in a way that provides effective treatment for most of the individuals [50
]. This requires knowledge not only of typical adherence levels and adherence patterns, but also an understanding of what types of resistance are predicted in individuals failing a particular therapy. Although some of this knowledge can be gained through experience, an understanding of the mechanisms behind adherence–resistance relationships may make it possible to predict expected resistance patterns for new medications and new classes of medications in the future. This understanding may also facilitate clinical trial design, including designs used to evaluate antiretroviral regimen sequencing and the use of specific combinations of medications, such as designing regimens with symmetrical half-lives. Following are brief examples of the application of this information for these purposes.
Predicting adherence–resistance relationships for other antiretroviral agents
Unknown adherence–resistance relationships can be hypothesized based on knowledge of drug potency, the fitness of resistant virus, and the genetic barrier to antiretroviral resistance ().
- (1)Nucleoside/nucleotide analogues other than the deoxycytidine analogues are mostly of moderate potency; there is impaired fitness with resistance and a moderate genetic barrier to resistance. This pattern is most similar to that of nonboosted protease inhibitors, so resistance would be expected at moderate-to-high levels of adherence.
- (2) Enfuvirtide is of high potency, resistant virus has impaired fitness, and the genetic barrier to resistance is low. This pattern is most similar to the deoxycytidine analogue NRTIs, so resistance would be expected to occur at moderate levels of adherence.
- (3) Raltegravir is of high potency, resistant virus has impaired fitness , and there is a low genetic barrier to resistance. This pattern is most similar to the deoxycytidine analogue NRTIs, so resistance would be expected to occur at moderate levels of adherence.
- (4) Etravirine is of high potency, fitness of mutated virus is not impaired, and the genetic barrier to resistance is moderate. This pattern predicts a similar adherence–resistance relationship to other NNRTIs except that the increased barrier to resistance would make the incidence of resistance lower at all adherence levels.
- (5) Maraviroc is of high potency but the fitness of mutated virus is hard to predict. The resistance pathways are complex and include overgrowth of low-frequency populations that use the CXCR4 coreceptor . In addition, some mutations allow the mutant virus to preferentially use drug-bound CCR5 . For these reasons, it is not possible to predict the adherence–resistance relationship for this class of drugs.
In the absence of preexisting resistance, poor adherence is the major risk factor for virological failure and the development of resistance. shows the expected risks for resistance with typical initial drug combinations. Overall, resistance is most common for NNRTIs and deoxycytidine analogue NRTIs, followed by nonboosted protease inhibitors and nondeoxycytidine analogue NRTIs, and is least common for boosted protease inhibitors. also presents potential associations between differential drug exposure, due to asymmetric medication half-lives or differential adherence, and the development of class-specific resistance.
Likelihood of resistance at initial virological failure and likelihood of differential drug exposure and differential adherence based on regimen composition.
Integrase inhibitors are currently being studied as initial therapy for HIV-1 infection [54
]. The adherence–resistance relationship for integrase inhibitors is expected to be similar to deoxycytidine NRTIs. Based on the relatively short serum half-life of raltegravir, the potential for differential drug exposure based on pharmacokinetics should be similar to protease inhibitors. However, the situation may be more complex as recent evidence suggests that raltegravir is essentially an irreversible inhibitor of HIV-1 DNA integration [55
]. Differential adherence is unlikely, as raltegravir appears well tolerated. These characteristics suggest that resistance will be common in individuals failing integrase inhibitors and this has been seen in heavily pretreated patients [56
]. Limited data suggest that dual-class resistance at first failure may also be relatively common [54
There are several important gaps in our current knowledge. Adherence–resistance relationships in the setting of transmitted or preexisting mutations may differ. Also, most studies have assessed class-specific relationships in the setting of antiretroviral regimens composed of a nucleoside backbone and one other component. How alternative combinations as initial or salvage therapy will interact is unclear. Recent studies have only begun to explore patterns of nonadherence, such as treatment gaps and differential adherence, which may be important in creating differential drug exposure leading to resistance. Adherence–resistance relationships for newer antiretroviral agents are not well characterized; future research should help to delineate these relationships. Finally, to date studies reporting adherence–resistance relationships have used traditional resistance assays with sensitivities down to 10–20% of the circulating viral population. Failure with ‘susceptible’ virus as defined by standard assays may hide a more complex mixture of circulating and/or archived resistant viruses that could impact the effectiveness of future treatment regimens [57
]. More sensitive resistance assays are now available and will help to further delineate class-specific adherence–resistance relationships.