Abstract

Public health debates about providing HIV antiretroviral therapy to impoverished populations have centred on the relationship between adherence and risk of drug resistance. Recent data indicate that each antiretroviral therapeutic class has a unique adherence–resistance relationship. Resistance to single protease inhibitor therapy occurs most frequently at moderate to high levels of adherence, resistance to non-nucleoside reverse transcriptase inhibitor therapy occurs at low to moderate levels of adherence, and resistance to ritonavir-boosted protease inhibitor therapy is most likely to occur at middle ranges of adherence. These dynamic relationships should be considered in balancing the individual and public health benefits of therapy.

‘Non-adherence leads to drug-resistant HIV.’ Or, at least, this is what we have been trained to believe. Failure to take the prescribed doses of antiretroviral drugs leads to ongoing viral replication in the presence of drug and the selection of drug-resistant HIV. This view forms the basis of domestic15 and international6 public health debates regarding the potential benefits and dangers of providing antiretroviral therapy to populations at risk for non-adherence. Some have argued that the risk of spreading resistant virus justifies withholding therapy, both domestically and internationally, until adherence support mechanisms are in place.1,7,8 These debates hinge on an accurate understanding of the relationship between adherence and resistance to HIV antiretroviral therapy. However, only recently have empirical studies directly addressed this issue. As we will discuss here, these studies indicate that the relationship between adherence and HIV drug resistance is more complicated than assumed initially. For some regimens, resistance may be more likely in those who take more rather than less of their medications. For others, the opposite may be true.

The association between antiretroviral adherence and viral suppression,912 progression to AIDS13 and progression to death14,15 is well established. Initial studies of adherence and resistance were limited by a small sample size, the use of monotherapy,16 or the use of incompletely characterized measures of adherence.17 Given the limitations in these early studies, the early proposition that non-adherence leads to resistance was influenced heavily by the prior epidemic of multidrug-resistant tuberculosis in New York City where resistance was almost exclusively seen in individuals at risk for non-adherence due to addiction, mental illness and unstable housing.5,18

The widely accepted premise that non-adherence breeds resistance is now being challenged, at least as it pertains to anti-HIV therapy. Recent data from cohorts of individuals with well-characterized measures of adherence suggest that resistance to both protease and nucleoside reverse transcriptase inhibitors occurs primarily in highly adherent patients.9 In separate studies, Walsh et al.19 and Howard et al.20 demonstrated linear and direct associations between adherence and the number of drug resistance mutations. Gallego et al.21 found protease inhibitor resistance was limited to those individuals reporting more than 90% adherence.

The above studies were largely cross-sectional and were limited by the possibility that resistance occurred during unmeasured lapses in adherence. Subsequent studies with more extensive, concurrently obtained longitudinal adherence and resistance measures have confirmed the early findings. Longitudinal studies by our group and by Miller et al. found that increasing adherence independently predicts the rate of accumulation of drug resistance mutations among patients with persistent detectable viraemia.22,23 When all patients were included in our study, including those with undetectable viraemia, we estimated that one-quarter of all drug resistance mutations occurred in patients with 92–100% adherence. A subsequent mathematical model based on these data estimated that population-level resistance occurs most frequently at 81% adherence and declines only modestly with perfect adherence.24 This estimate is remarkably similar to independent estimates recently provided in abstract form by Harrigan et al.25 and King et al.26 Both of these longitudinal studies involved well-characterized individuals treated with standard treatment regimens, and both concluded that drug resistance is most common in patients with 80–90% adherence. Since the average level of adherence in most studies with objective measures of adherence is 70%,27 clearly it is not the ‘non-adherent’ who are at highest risk of drug resistance.

Most of the preceding data were derived from cohorts of individuals treated with either sequential monotherapy or earlier, less potent three-drug regimens (e.g. two nucleoside analogues and a single protease inhibitor). Given the inherent lack of anti-HIV potency when antiretroviral drugs are used in this manner, it is not surprising that high-level adherence did not routinely suppress all viral replication nor that high-level adherence did not prevent the emergence of resistant variants. Standard of care now focuses on the use of more effective first-line regimens, including two nucleoside analogues and either a non-nucleoside reverse transcriptase inhibitor (e.g. efavirenz) or a ritonavir-boosted protease inhibitor (e.g. ritonavir/lopinavir). We estimate that any regimen that durably suppresses viraemia to undetectable levels (<50 copies/mL) in 95% of perfectly adherent patients will reduce the population burden of resistance by 45% compared to historical regimens.24 Given the very high virological suppression rates observed with either efavirenz- or ritonavir/lopinavir-based regimens,28,29 it is reasonable to predict that the balance will shift towards more complete viral suppression and less drug resistance in highly-adherent individuals.

Emerging data indicating that antiretroviral classes may have different adherence–resistance relationships complicate the choice between efavirenz- and ritonavir/lopinavir-based regimens. A direct adherence–resistance relationship appears strongest for non-ritonavir-boosted protease inhibitors,1921,23,26 and it has also been observed for most nucleoside analogues.9,30 No such relationship has been observed for the non-nucleoside reverse transcriptase inhibitors (NNRTIs). In contrast to studies of protease inhibitors, Parienti et al. found that NNRTI resistance is associated with interruptions of therapy,31 and in a mostly NNRTI-treated cohort, Sethi et al. found resistance occurring at lower levels of adherence than that observed in patients who develop resistance to protease inhibitor therapy.32 A different adherence–resistance relationship for NNRTIs is also consistent with reports of resistance to these medications occurring after single-dose or after short-course therapy given during perinatal HIV prevention trials.33,34 Since single-dose therapy is analogous to the lowest level of adherence possible, it appears that the NNRTIs have an adherence–resistance relationship wherein almost any exposure in the absence of full viral suppression is sufficient to cause resistance.

Why would the relationship between adherence and resistance differ substantially between drug classes? Although the answer to this question remains unknown, substantial in vivo data and theoretical considerations suggest that NNRTIs have several characteristics that might result in an unfavourable adherence/resistance relationship. First, NNRTIs are very potent and therefore exert a strong selective pressure. Second, NNRTIs act at a site distant from the active site of their target enzyme. Therefore, mutations that confer drug resistance do not dramatically impact enzymic efficiency and—by extension—viral replicative capacity (or viral ‘fitness’). Third, NNRTIs have long half-lives and remain in plasma for extended periods after several missed doses; this allows the virus an opportunity to replicate in the presence of suboptimal drug exposure (note that this pharmacological property might be beneficial in some instances since a few missed doses separated in time will be unlikely to allow the virus to replicate). Fourth, resistance to one NNRTI almost universally confers cross-resistance to all other NNRTIs. Finally, NNRTI resistance persists indefinitely after the drug is discontinued.35 Thus, many factors appear to favour the virus over the drug in terms of NNRTI resistance. NNRTIs should be considered a relatively fragile drug class across the entire range of adherence too low to suppress viral replication.

Many of these factors regarding adherence and resistance to NNRTIs do not apply to ritonavir-boosted protease inhibitors. For example, resistance to protease inhibitors requires multiple mutations, each of which significantly reduces enzymic efficiency and viral fitness. Therefore, high-level drug resistance requires both viral replication and sufficient drug exposure to create a selective advantage for less-fit, resistant virus and should emerge over a relatively narrow range of drug exposure. Also, since the half-life of a protease inhibitor is relatively short, protease inhibitor concentrations are likely to remain in a suboptimal therapeutic range for only a brief time during periods of non-adherence.25 Given these factors, it is perhaps not surprising that protease inhibitor resistance is rarely observed during early virological failure of ritonavir-boosted protease inhibitor-based regimens.26,28 Even if resistance does emerge during protease inhibitor therapy, there are theoretical arguments and some empirical data that indicate that these variants will be less fit and therefore less infectious than NNRTI-resistant variants.36,37

Based simply on these resistance arguments, it would appear that the widespread use of NNRTIs will have greater public health costs than the widespread use of ritonavir-boosted protease inhibitors. Although these issues argue for the use of dual protease inhibitor-based regimens over NNRTI-based regimens, it remains difficult to ignore the fact that NNRTIs are far less expensive than protease inhibitors (especially generic formulations available in the developing world), and that NNRTIs are generally easy to store and easy to administer. In our opinion, regimen choice on both an individual- and a population-based level should be driven by many factors, including fiscal constraints, clinical effectiveness, tolerability and the risk of drug resistance.

In summary, the relationship between adherence and resistance to HIV antiretroviral therapy is more complex than ‘non-adherence increases the risk of drug resistance.’ For non-boosted protease-based regimens, most drug resistance occurs in patients who take most of their medications, and there is likely to be a bell-shaped relationship between adherence and resistance accumulation (with the peak of the curve near 70–80% adherence). For ritonavir-boosted protease inhibitor regimens, limited resistance occurs at any level of adherence. For NNRTIs, resistance mutations are uncommon in highly adherent patients but will likely be very common in patients with any level of adherence insufficient for full viral suppression (Figure 1).

How should these data be used to formulate treatment strategies? Most physicians’ first priority is to prevent illness in individual patients rather than to bring down the population burden of resistant HIV. This therapeutic framework may change, however, with the increasing need to maximize treatment benefit on a community basis in resource-constrained areas. Since governments and other large funding agencies will be purchasing drugs in such settings, the public health implications of any treatment strategy will likely become more relevant. Rational predictions of how deployment of therapy will impact population levels of drug resistance will require a sound understanding of the complex relationships between adherence and development of drug resistance. Since this relationship appears to be drug-specific, we propose that defining this relationship be a critical part of the drug development process. Recently, resistance to HIV has been the price for patients adhering to therapy so that they can live longer. Hopefully, a better understanding of adherence and resistance in combination with more complete viral suppression will better align the individual and public health benefits of antiretroviral therapy.

Acknowledgements

Funding for this study has been from NIMH, The Joint Fogarty/GIVI-UCSF Center for AIDS Research Program and The Doris Duke Charitable Foundation.

*

Corresponding author. Present address: Box 1372, San Francisco General Hospital, UCSF, 1001 Potrero Avenue, Building 100, Room 301, San Francisco, CA 94110, USA. Tel: +1-415-206-3462; Fax: +1-415-206-4360; E-mail: db@epi-center.ucsf.edu

Figure 1. Schematic figure outlining the relationship between medication adherence and the risk of developing PI or NNRTI drug resistance. NNRTI-treated individuals rarely develop resistance at high levels of adherence due to the virological effectiveness of these regimens. NNRTI resistance develops rapidly at moderate to low levels of resistance due to the low ‘fitness’ costs associated with single mutations. Single PI-treated individuals may develop resistance at high levels of adherence because residual viral replication is often seen in such patients. PI resistance is uncommon at low levels of adherence because of the significant fitness costs associated with these mutations. Resistance to a ritonavir-boosted PI is only possible in a narrow range of adherence where there is sufficient drug around to select for mutations that reduce ‘fitness’ while still allowing residual viral replication. Data in this figure are conceptual and based on trends observed in a number of recent studies (see text). PI, protease inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor.

Figure 1. Schematic figure outlining the relationship between medication adherence and the risk of developing PI or NNRTI drug resistance. NNRTI-treated individuals rarely develop resistance at high levels of adherence due to the virological effectiveness of these regimens. NNRTI resistance develops rapidly at moderate to low levels of resistance due to the low ‘fitness’ costs associated with single mutations. Single PI-treated individuals may develop resistance at high levels of adherence because residual viral replication is often seen in such patients. PI resistance is uncommon at low levels of adherence because of the significant fitness costs associated with these mutations. Resistance to a ritonavir-boosted PI is only possible in a narrow range of adherence where there is sufficient drug around to select for mutations that reduce ‘fitness’ while still allowing residual viral replication. Data in this figure are conceptual and based on trends observed in a number of recent studies (see text). PI, protease inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor.

References

1.
Sontag, D. & Richardon, L. (
1997
). Doctors withhold HIV pill regimen from some. New York Times, 2 March 1997, p. A1.
2.
Altice, F. L. & Friedland, G. H. (
1998
). The era of adherence to HIV therapy.
Annals of Internal Medicine
 
129
,
503
–5.
3.
Friedland, G. H. & Williams, A. (
1999
). Attaining higher goals in HIV treatment: the central importance of adherence.
AIDS
 
13
, Suppl. 1,
S61
–72.
4.
Wainberg, M. A. & Friedland, G. (
1998
). Public health implications of antiretroviral therapy and HIV drug resistance.
Journal of the American Medical Association
 
279
,
1977
–83.
5.
Bangsberg, D., Tulsky, J. P., Hecht, F. M. et al. (
1997
). Protease inhibitors in the homeless.
Journal of the American Medical Association
 
278
,
63
–5.
6.
Harries, A. D., Nyangulu, D. S., Hargreaves, N. J. et al. (
2001
). Preventing antiretroviral anarchy in sub-Saharan Africa.
Lancet
 
358
,
410
–4.
7.
Popp, D. & Fisher, J. D. (
2002
). First, do no harm: a call for emphasizing adherence and HIV prevention interventions in active antiretroviral therapy programs in the developing world.
AIDS
 
16
,
676
–8.
8.
Jackson, P. (
2003
). Can Africa handle AIDS drugs? [Online]. http://news.bbc.co.uk/1/hi/health/3067345.stm (9 March 2004, date last accessed).
9.
Bangsberg, D. R., Hecht, F. M., Charlebois, E. D. et al. (
2000
). Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population.
AIDS
 
14
,
357
–66.
10.
Paterson, D. L., Swindells, S., Mohr, J. et al. (
2000
). Adherence to protease inhibitor therapy and outcomes in patients with HIV infection.
Annals of Internal Medicine
 
133
,
21
–30.
11.
Arnsten, J. H., Demas, P. A., Farzadegan, H. et al. (
2001
). Antiretroviral therapy adherence and viral suppression in HIV-infected drug users: comparison of self-report and electronic monitoring.
Clinical Infectious Diseases
 
33
,
1417
–23.
12.
McNabb, J., Ross, J. W., Abriola, K. et al. (
2001
). Adherence to highly active antiretroviral therapy predicts virologic outcome at an inner-city human immunodeficiency virus clinic.
Clinical Infectious Diseases
 
33
,
700
–5.
13.
Bangsberg, D. R., Perry, S., Charlebois, E. D. et al. (
2001
). Non-adherence to highly active antiretroviral therapy predicts progression to AIDS.
AIDS
 
15
,
1181
–3.
14.
Hogg, R. S., Heath, K., Bangsberg, D. et al. (
2002
). Intermittent use of triple-combination therapy is predictive of mortality at baseline and after 1 year of follow-up.
AIDS
 
16
,
1051
–8.
15.
Garcia de Olalla, P., Knobel, H., Carmona, A. et al. (
2002
). Impact of adherence and highly active antiretroviral therapy on survival in HIV-infected patients.
Journal of Acquired Immune Deficiency Syndromes
 
30
,
105
–10.
16.
Vanhove, G. F., Schapiro, J. M., Winters, M. A. et al. (
1996
). Patient compliance and drug failure in protease inhibitor monotherapy.
Journal of the American Medical Association
 
276
,
1955
–6.
17.
Montaner, J. S., Reiss, P., Cooper, D. et al. (
1998
). A randomized, double-blind trial comparing combinations of nevirapine, didanosine, and zidovudine for HIV-infected patients: the INCAS Trial. Italy, The Netherlands, Canada and Australia Study.
Journal of the American Medical Association
 
279
,
930
–7.
18.
Belkin, L. (
1991
). TB threat: not taking the medicine. Partly cured patients are the deadliest carriers. New York Times, 18 November 1991, p. B1.
19.
Walsh, J. C., Pozniak, A. L., Nelson, M. R. et al. (
2002
). Virologic rebound on HAART in the context of low treatment adherence is associated with a low prevalence of antiretroviral drug resistance.
Journal of Acquired Immune Deficiency Syndromes
 
30
,
278
–87.
20.
Howard, A., Arnsten, J., Gardner, L. et al. (
2002
). Lack of multi-drug resistance in nonresponders to antiretroviral therapy with poor adherence. In First International AIDS Society Conference on HIV Pathogenesis and Treatment, Buenos Aires, 2002. Abstract 604. International AIDS Society, Stockholm, Sweden.
21.
Gallego, O., de Mendoza, C., Perez-Elias, M. J. et al. (
2001
). Drug resistance in patients experiencing early virological failure under a triple combination including indinavir.
AIDS
 
15
,
1701
–6.
22.
Miller, L., McCutchan, J., Keiser, P. et al. (
2003
). The impact of medication adherence.
Antiviral Therapy
 
8
,
S167
.
23.
Bangsberg, D. R., Charlebois, E. D., Grant, R. M. et al. (
2003
). High levels of adherence do not prevent accumulation of HIV drug resistance mutations.
AIDS
 
17
,
1925
–32.
24.
Bangsberg, D. R., Kagay, C., Porco, T. et al. (
2004
). Modeling the relationship between adherence and the accumulation of protease inhibitor drug resistance mutations.
Journal of Infectious Diseases
 , in press.
25.
Harrigan, R., Dong, W., Alexander, C. et al. (
2003
). The association between drug resistance and adherence determined by two independent methods in a large cohort of drug naive individuals starting triple therapy. In Second International Conference on HIV Treatment and Pathogenesis, Paris, 2003. Abstract LB12. IAS International AIDS Society, Stockholm, Sweden.
26.
King, M., Brun, S., Tschampes, J. et al. (
2003
). Exploring the effects of adherence on resistance: use of local linear regression to reveal relationships between adherence and resistance in antiretroviral-naive patients treated with lopinavir/ritonavir or nelfinavir.
Antiviral Therapy
 
8
,
S118
.
27.
Bangsberg, D. & Deeks, S. (
2003
). Is average adherence to HIV antiretroviral therapy enough?
Journal of General Internal Medicine
 
17
,
812
–3.
28.
Walmsley, S., Bernstein, B., King, M. et al. (
2002
). Lopinavir-ritonavir versus nelfinavir for the initial treatment of HIV infection.
New England Journal of Medicine
 
346
,
2039
–46.
29.
Staszewski, S., Morales-Ramirez, J., Tashima, K. T. et al. (
1999
). Efavirenz plus zidovudine and lamivudine, efavirenz plus indinavir, and indinavir plus zidovudine and lamivudine in the treatment of HIV-1 infection in adults. Study 006 Team.
New England Journal of Medicine
 
341
,
1865
–73.
30.
Kuritzkes, D., Ickovics, J., Bassett, R. et al. (
2001
). HIV-1 drug resistance and medication adherence in patients receiving NRTIs. In Fifth International Workshop on HIV Resistance & Treatment Strategies, Scottsdale, 2001. Abstract 84.
31.
Parienti, J., Veronique, M., Descampes, D. et al. (
2004
). Predictors of virologic failure and resistance in HIV-1 infected patients treated with nevirapine or efavirenz-based antiretroviral therapy.
Clinical Infectious Diseases
 , in press.
32.
Sethi, A. K., Celentano, D. D., Gange, S. J. et al. (
2003
). Association between adherence to antiretroviral therapy and human immunodeficiency virus drug resistance.
Clinical Infectious Diseases
 
37
,
1112
–8.
33.
Jackson, J. B., Becker-Pergola, G., Guay, L. A. et al. (
2000
). Identification of the K103N resistance mutation in Ugandan women receiving nevirapine to prevent HIV-1 vertical transmission.
AIDS
 
14
,
F111
–5.
34.
Clarke, J. R., Braganza, R., Mirza, A. et al. (
1999
). Rapid development of genotypic resistance to lamivudine when combined with zidovudine in pregnancy.
Journal of Medical Virology
 
59
,
364
–8.
35.
Mellors, J., Palmer, S., Nissley, D. et al. (
2003
). Low frequency non-nucleoside reverse transcriptase inhibitor (NNRTI)-resistant variants contribute to failure of efavirenz-containing regimens in NNRTI-experienced patients with negative standard genotypes for NNRTI mutations.
Antiviral Therapy
 
8
, S166.
36.
Grant, R. M., Hecht, F. M., Warmerdam, M. et al. (
2002
). Time trends in primary HIV-1 drug resistance among recently infected persons.
Journal of the American Medical Association
 
288
,
181
–8.
37.
Little, S. J., Holte, S., Routy, J. P. et al. (
2002
). Antiretroviral-drug resistance among patients recently infected with HIV.
New England Journal of Medicine
 
347
,
385
–94.

Author notes

1The Epidemiology and Prevention Interventions Center, San Francisco General Hospital, UCSF; 2The San Francisco General Hospital AIDS Program, UCSF; 3The Department of Epidemiology and Biostatistics, San Francisco General Hospital, UCSF, 1001 Potrero Avenue, San Francisco, CA 94110, USA