Antiviral Therapy 9:695–702

Current patterns in the epidemiology of primary HIV drug resistance in North America and Europe Deenan Pillay Department of Virology, Windeyer Institute, Royal Free and University College Medical School, University College London, London, UK Correspondence: Tel: +44 20 7679 9490; Fax: +44 20 7580 5896; E-mail: [email protected]

Despite numerous studies in recent years, it is still difficult to draw general conclusions about the extent to which drugresistant HIV-1 is transmitted. In addition to the highly stratified nature of primary resistance itself, true epidemiological surveillance has been rare and studies to date have suffered from wide variability in their designs, definitions and datasets. In the absence of consensus standards, this has resulted in a large number of isolated 'snapshots' with little scope for data-pooling and comparison. This brief review examines some of the major confounding factors that restrict the utility of individual studies and prevent the combination of studies to increase statistical power. Despite these limitations, data from North America and Europe lead to the tentative conclusion that

transmission rates in these areas have generally fallen or remained stable in the past 2–3 years. However, data for the UK seem to indicate an ongoing rise in the transmission of drug resistance mutations, currently present in up to 20% of new infections. Transmission of resistant HIV represents a clinically important phenomenon, although the scale and relevance are being obscured by methodological variations and non-clinical definitions of resistance. Those of us with an interest in the epidemiology of drug resistance, whether in primary transmission or on-treatment, must learn to speak the same language if we are to establish meaningful correlations between survey datasets and the HIV-infected population as a whole.

Introduction As antiretroviral therapy moves into its 17th year of existence – and durable multi-drug treatment regimens into their eighth – the scale and distribution of drugresistant HIV in a growing and increasingly treatment-experienced patient population is a matter of serious clinical concern. The potential for transmission of resistant virus from this population has significant clinical implications, but studies into the epidemiology of both primary and treatment-emergent resistance have long been complicated by assay limitations and differences in methodology, in assumptions about what constitutes resistance and in the populations under study. This paper reviews data on the incidence of primary resistance transmission in Europe and North America – the sources of most available information on the epidemiology of resistance – and discusses some of the key influences and confounding factors that make inter-study comparisons and trend analyses difficult.

Factors influencing the reported prevalence of primary drug resistance Primary resistance is a highly stratified phenomenon and local incidence rates are exquisitely sensitive to the acute ©2004 International Medical Press 1359-6535/02/$17.00

effects of changes in treatment practice that influence both on-therapy resistance and HIV transmission frequency. Even within a particular demographic stratum, the ongoing rate can vary tremendously over a short period of time. The lack of a unified and standard methodology of study design, sampling and reporting format that takes account of the numerous influences on observed rates makes it exceedingly difficult to compare survey results internationally, when the distinction between real differences in the demography of a local epidemic and unrepresentative bias in the dataset is unclear. As a result, generalizations regarding the incidence and trends in transmitted drug resistance should be tempered with an understanding that any conclusions are tentative and subject to the variability of available data. Nevertheless, the clinical importance of transmission makes extrapolation within the limits imposed by this variability a worthwhile exercise despite the difficulties involved. The most significant influences and confounders include study design, study location and sample population, and definitions and assay methodology.

Study design Most surveys of primary transmission performed to date have been opportunistic or ‘snapshot’ captures of 695

D Pillay

data from an available population that do not represent true surveillance in the epidemiological sense. Study catchment areas are often limited and localization introduces a degree of sample stratification that may render it unrepresentative. Sample sizes have usually been small with no confidence intervals – the result of limiting sampling to relatively uncommon cases of acute or very recent infection. Longitudinal data, which may help provide a context for the findings of a single round of data capture, are often unavailable.

all mutations minimized. The surprising persistence of some transmitted mutations allows for larger sample sizes through incorporating patients with a longer history of infection. However, primary resistance in these individuals may not be representative of new infections. The informative value of a dataset based on mixed recent and chronic infections will be diluted both by the different time points for infection and the potential for different rates of loss for different mutations over the range of time-since-infection represented in the sample set.

Study location and sample population No epidemic is static, and this is especially true for HIV. At least within the developed world, HIV infection within a particular country comprises multiple co-existing epidemics, each with their own dynamics of spread and antiretroviral (ARV) susceptibility (often associated with differential access to healthcare). Demographic or social variables that correlate with availability of and access to treatment, adherence to medication, transmission mode and propensity toward risk-taking transmission behaviours [1,2], will all influence the rate of transmitted drug resistance. Other influences on both the rate and pattern of transmission will include differences in ARV prescribing between distinct populations and local medical standards of care for HIV infection (such as the availability of sensitive viral load assays or resistance testing). Evidence exists for differing transmission efficiencies between different resistance mutations, apparently linked to considerations of HIV replicative fitness [3,4]. While this implies that overall local patterns of transmitted resistance will not always mirror the patterns of local on-therapy resistance, it could also, in principle, create subtly stratified sub-populations with different incidence rates for transmission based on differential ARV prescription. Additionally, viral fitness may be related to transmission through viral load, rather then the virus being less transmissible per se. Furthermore, for any given population the rate of transmission will be influenced by whether infection was acquired within that population or outside it. An example of the influence of human travel and interversus intra-population infection can be found in the significantly higher incidence of primary resistance in Europe for HIV subtype B versus non-B subtypes predominantly acquired elsewhere, compared with similar incidences for B and non-B in Israel, where the epidemic has historically shown a greater subtype diversity [5]. The implications of a study will also vary according to whether individuals with recent or chronic infection are sampled. Clearly, the advantage of sampling recent/acute infections is that the timing of transmission can be estimated and the potential for loss of some or 696

Definitions and assay methodology Genotyping and phenotyping provide different information and are not directly interchangeable. Furthermore, the rules which define ‘resistance’ can significantly affect the outcome of a survey [6–8]. One of the most significant problems in all studies of resistance to date has been lack of standardization with regards to the criteria used to define it. For phenotypic surveys, the problem of different assay cutoffs, not necessarily related to ontherapy response, makes data pooling and cross-study comparison difficult. However, most studies have been based on genotype not phenotype, and the definitions of ‘resistance’ used are seldom the same as those used clinically, as, for example, when making new drug choices after therapeutic failure. As a result, interpreting the results of epidemiological surveys in terms of their clinical significance is extremely difficult. The majority of surveys score resistance as the presence of one or more resistance-associated mutations from a given set, most commonly as defined by the IAS-USA Resistance Panel [9]. While such an approach has the advantage of simplicity and portability, without a second layer of interpretation to relate observed mutations to in vivo response – or even in vitro drug susceptibility – the presence of mutations tells us only that a non-wild-type genome derived from selective drug pressure has been transmitted. It does not necessarily tell us whether the isolate is itself physically drug-resistant at a clinically relevant level, and hence the reported prevalence may be overestimated. The scale of the problem can also be overestimated when, as is commonly the case, the presence of mutations or reduced susceptibility to one or more drugs in a given class is reported as ‘class resistance’. Clearly, this is an oversimplification that presents a risk of overstating the extent of resistance observed. Only for the current non-nucleoside reverse transcriptase inhibitors (NNRTIs) could the presence of mutations to any given drug be reasonably interpreted as ‘class resistance’ in a semantically appropriate sense. Similarly, for the nucleoside reverse transcriptase inhibitor (NRTI) analogue drugs only the presence of the multinucleoside resistance mutation Q151M or a ©2004 International Medical Press

Epidemiology of primary HIV drug resistance

codon 69 insertion would define true class resistance without an additional layer of interpretation on a drugby-drug basis. Even when based on strictly objective parameters, such as interpretation according to a clinically validated genotype rule or phenotypic cutoff, definitions of resistance or sensitivity may change as more information becomes available or treatment practices change. Lists of resistance mutations, such as that generated by IAS-USA, reflect the genetic changes observed following drug selection pressure. For instance, the V118I within the reverse transcriptase gene is associated with nucleoside analogue resistance, but in reality is invariably present with other well-recognized and important mutations. By contrast, the presence of this mutation alone in a drug-naive individual is more probably a reflection of a polymorphic change not depicting prior drug selection. Indeed, individuals infected with such viruses do not appear to respond sub-optimally to nucleoside analogues [10].

Primary resistance in North America Primary transmission of NRTI-resistant virus peaked in the mid-1990s and then fell sharply after 1997, following the introduction of protease inhibitors (PIs) and triple-drug highly active antiretroviral therapy (HAART) as standards of care [11]. Following this decline, transmission of resistance to all drug classes across North America as a whole increased again to a new peak in 1999–2000 [11–13], after which levels appear to have stabilized or declined slightly, with a shift towards less NRTI and PI resistance but no significant change in the extent of NNRTI resistance [14,15]. This trend is reflected in part by a New York study of 249 mainly homosexual primary HIV-infected (PHI) subjects enlisted between 1995 and 2002 [16,17]. In this study, the overall prevalence of any resistance-associated mutations increased from 13% in 1995–1998 to 20% in 1999–2000. The subsequent decline to 17% in 2001–2002 was due mainly to the decrease in transmitted NRTI resistance. However, NNRTI genotypic and phenotypic resistance increased steadily during the study and PI resistance also increased, although it remained lower at below 5%. Reported absolute transmission rates vary somewhat by study, with San Francisco-based surveys showing higher rates for all drug classes than multicentre surveys. Grant et al. [14] noted a 27% overall incidence of resistance mutations to any drug class in 90 patients presenting with recent HIV infection in San Francisco from January 2000 to June 2001. For the period July 2001 to December 2002, this incidence remained stable at 25% for 89 newly infected patients, and primary NNRTI resistance also remained stable at Antiviral Therapy 9:5

12–13% across these two periods. However, the incidence of NRTI resistance mutations halved to 10% in 2001–2002 from 21% in the previous period. Contrary to most other US studies, primary PI resistance increased in San Francisco from 8% in 2000–2001 to 14% in 2001–2002, although this increase was not statistically significant (P=0.088). Other studies show similar trends but with a slightly lower incidence and a more obvious trend towards less transmission after 2000. Little et al. [15] observed high-level phenotypic resistance (>10-fold increase in drug IC50) to any ARV drug in 12% of 123 recently infected patients in 10 cities across North America in 1999–2000, whereas for 135 patients in 2000–2002 this had fallen (though not statistically significantly) to 7%. As in the Grant study, there was a significant drop in primary NRTI resistance (>10-fold) from 5.7% in 1999–2000 to 0.7% in 2000–2001. NNRTI resistance remained at 6–7% across these two periods, but, unlike in the Grant study, high-level resistance to PIs did fall significantly in the latter period (0.7% in 2000–2002 vs 8% in 1999–2000), as did resistance to multiple drug classes (0.7% vs 6.5%, respectively). In part, the difference between rates observed over similar time periods between the Grant and Little studies could be due to their differing definitions of resistance. However, a 10-city multi-centre US survey by Bennett et al. [18] covering a range of risk groups, undertook evaluation of the incidence of drug-resistant infection by the presence of resistance-associated mutations in a similar way to the Grant study. The results were still lower than the San Francisco data, but comparable with the Little study regarding overall rates of infection with virus resistant to any drug (12–13% for 1999 and 2000–2001). The data therefore suggest that while transmission of drug resistance across the US as a whole is currently about 10% of new infections, ‘hot-spots’ such as San Francisco may have a much higher local incidence rate. A slightly lower and more stable rate of transmission has been reported across Canada over the period 1997–2001, according to data from the Canadian HIV Strain and Drug Resistance Surveillance Program [19]. Overall, primary resistance mutations were noted in 7% of newly diagnosed, treatment-naive patients over this time, with no obvious trends other than a later emergence of primary NNRTI (post-1999) and PI (post-1998) resistance than in the US. Most resistance noted in the Canadian programme was to NRTIs, even at later time points (5% in 2001, vs 2% each for NNRTI, PI and multi-class resistance). Further Canadian data from the Montreal Primary Infection Cohort [2] suggested a significant drop in primary resistance in homosexual transmission after 2000 relative to the period 1996–2000. Overall, 697

D Pillay

Transmission in Europe The largest European-wide survey of primary drug resistance to date has been the recent CATCH study [5]. This multi-centre cross-sectional retrospective survey assessed primary resistance in acutely and chronically infected patients naive to drugs across 16 countries over the period 1996–2002. Of 1400 patients assessed, 138 (10%) displayed genotypic evidence of resistance to any drug, with the majority of that resistance being to the NRTIs (7%). Interestingly, compared with the US data, NNRTI resistance was observed in only 3%, which may represent the influence of samples from earlier time points over the period (>80% of primary resistance was NRTI resistance in 1996–2000 vs