Combination antiretroviral therapy is used to minimize HIV replication and reduce clinical progression and death in HIV-1-infected individuals [1,2]. Although combination antiretroviral therapy is highly potent, its efficacy can be compromised by the development of mutations that cause resistance to one or more antiretroviral drugs in the regimen. The development of resistance can lead to increased viral replication, which in turn can lead to the development of more mutations to antiretroviral drugs in the regimen and to therapy failure .
Genotypic resistance testing is used to guide therapy selection by producing lists of the type and positioning of each mutation [4–6]. Studies exploring the relationship between resistance and response to therapy have so far been limited by a low prevalence of each mutation or inadequate data on virological response. Some mutations cause resistance to certain antiretroviral drugs but resensitize others [7–10], other mutations enhance the overall fitness of the virus in the presence of selective pressure by antiretroviral drugs [11,12], and some mutations are key mutations in the pathway for developing major resistance mutations . Consequently there are disagreements between experts on how each mutation affects the response to each antiretroviral drug, and over 20 rules-based genotypic interpretation systems (GIS) have been proposed [14–21].
Differences between GIS have been illustrated previously [22–25]. They tend to agree on the level of resistance associated with common resistance mutations; however, high levels of discordance exist for antiretroviral drugs that require complex mutational patterns to confer resistance (e.g. zidovudine, didanosine, stavudine and abacavir) [23,24,26–29]. This is also true for protease inhibitors (PI) when disagreements cannot be attributed to the presence of a single mutation or specific mutational pattern .
Ritonavir is currently used in low doses to boost the levels of other PI. Because of the potency of these combinations ritonavir-boosted protease inhibitors (PI/r) may require alternative or more sensitive interpretations to those for unboosted PI, although the current systems have also been shown to apply to PI/r . Some GIS have been updated to account for the use of ritonavir boosting, but still most discordance between GIS occurs among PI .
Individual mutations and resistance to the overall regimen, as illustrated by a genotypic sensitivity score (GSS), has been associated with HIV-RNA decline at various timepoints [6,18,22]. As limited information exists on the prognostic value of GIS for PI/r we investigated concordance between predicted PI/r resistance levels using four freely available GIS, and related the resulting interpretations to virological response after 12 weeks follow-up for patients receiving a PI/r. Current treatment guidelines identify week 12 as the optimal timepoint for assessing virological responses to a new combination regimen, so we selected this timepoint for analysis . We also assigned a GSS to the rest of the regimen and explored the relationship between this GSS and virological response.
This analysis was initially performed using data from four independent data sources: a large ongoing European cohort study (EuroSIDA) and data from three randomized controlled trials (MaxCmin1, MaxCmin2 and COLATE). Although EuroSIDA contained a more treatment-experienced population compared with the trials, a separate analysis of the two datasets provided consistent results. As a result, we merged datasets to improve power because the frequency of resistance to PI/r in each study individually was relatively low. To our knowledge, this is the largest combined database of patients receiving PI/r in which it was possible to examine the relationship between predicted sensitivity and viral load outcome.
Methods and materials
Details of EuroSIDA and the trials have already been published [31–34], and can be found at www.cphiv.dk. EuroSIDA is an ongoing, observational cohort of 13 601 HIV-1-infected patients from 30 European countries, Israel and Argentina. Six cohorts of consecutive HIV-1-infected patients with prebooked clinic appointments were followed up until May 2005 (date of censoring). At each visit, all CD4 cell counts and HIV-RNA levels taken since the last follow-up visit were collected and dates of starting and stopping each antiretroviral drug were recorded. From 13 601 individuals, 3893 (32.6%) started a PI/r, of whom 376 (9.7%) initiated a single PI/r (baseline), had genotyping in the year before baseline and had HIV-RNA values available at baseline (−3 months to +1 week) and between 4 and 24 weeks. If patients had more than one HIV-RNA measurement the value taken closest to 12 weeks was included in our analyses so that patients only had one outcome measurement.
Between June 1999 and May 2002, 799 HIV-1-infected patients were enrolled into the three 48-week, multicentre trials [MaxCmin1 (N = 324), MaxCmin2 (N = 339) and COLATE (N = 136)]. The MaxCmin studies compared ritonavir-boosted saquinavir with ritonavir-boosted indinavir (MaxCmin1) or against kaletra (lopinavir/ritonavir) (MaxCmin2). COLATE investigated the virological benefit of retaining lamivudine in a failing lamivudine-containing regimen in patients receiving combination antiretroviral therapy. Patients in COLATE who received a PI/r in addition to their lamivudine component were identified and combined with the MaxCmin patients to maximize numbers. The trials had relatively unrestrictive entry criteria, including patients who were antiretroviral naive, PI naive, or PI experienced with virological failure or PI intolerance. All trials collected baseline data on medical history, demographics and clinical and laboratory parameters. Data on clinical evaluation, safety, HIV-RNA levels and CD4 cell counts were collected at follow-up visits, which occurred after 4, 12, 24, 36 and 48 weeks. The week 4 HIV-RNA value may be too early to determine the success of a regimen and as the week 12 virological response has been shown to reflect long-term success , this was chosen as the timepoint of interest in this study. Out of 799 patients, 368 (46.1%) initiated a PI/r, had a successful baseline resistance test in the 3 weeks before baseline and had HIV-RNA values available at 12 weeks.
Overall, there were 28, 29, 55, 231 and 33 EuroSIDA patients and 0, 0, 89, 126 and 153 trial patients who received amprenavir/ritonavir, atazanavir/ritonavir, indinavir/ritonavir, lopinavir/ritonavir and saquinavir/ritonavir, respectively. Patients in EuroSIDA were similar to patients in the trials in terms of baseline characteristics (data not shown).
Plasma samples were used for sequencing if baseline HIV-RNA levels were greater than 500 copies/ml. HIV-RNA was quantified using polymerase chain reaction technology in combination with the Roche Ultrasensitive Amplicor HIV-1 Monitor Assay (Roche Molecular Systems, Alameda, California, USA) and the Chiron branched chain (bDNA) assay (Chiron Corp., Emeryville, California, USA).
Different sequencing procedures were used for each study. Virological sequencing was performed centrally (Advanced Biological Laboratories, Luxembourg) in batches using cryopreserved plasma samples in MaxCmin2 and COLATE. Sequence data for the reverse transcriptase (RT) and protease reading frames were obtained by extracting HIV-RNA from 500 μl plasma using the NucliSens isolation kit. Through nested polymerase chain reaction amplification we generated a 1.8-kb amplicon encompassing the protease and the first 1005 nucleotides of the RT gene, spanning a total of 1302 nucleotides. Purified amplicons were sequenced using the ABI Prism BigDye terminator cycle sequencing kit (Applied Biosystems, Foster City, California, USA) and the resulting nucleotide sequences were translated into amino acids.
In MaxCmin1 virological sequencing was performed at the International Clinical Virology Center (High Wycombe, UK), where they used the QIAamp kit (Qiagen, Barcelona, Spain) according to the manufacturer's instructions. Sequence data for the RT and protease genes were obtained using the Trugene HIV-1 Genotyping Kit and OpenGene DNA Sequencing System (Visible Genetics, Barcelona, Spain).
In EuroSIDA, virological sequencing was either performed at the IrsiCaixa Foundation (Badalona, Spain) or at the International Clinical Virology Center, depending on the time of the resistance test. For all four studies we identified and excluded sample mix-up or laboratory contamination by drawing phylogenetic trees.
Genotypic interpretation systems
Genotypes were provided as amino acid substitutions (full or mixed) from a reference clade B strain (i.e. HXB2) for both RT and protease. The amount of resistance was quantified by running the amino acid sequences through ANRS (version 13) , DMC (October 2004) , REGA (version 6.4)  and Stanford (version 4.2.0)  to produce a score for each antiretroviral drug in the regimen relating to whether a patient had a virus showing resistance, intermediate resistance or sensitivity to that drug. In each case a score of 0 for an antiretroviral drug demonstrates full resistance to that antiretroviral drug, 0.5 indicates intermediate resistance and 1 relates to antiretroviral sensitivity. Resistance levels ascribed by Stanford were reduced from a five-level system to a three-level system, so an overall score of 0–14 represented susceptibility, 15–59 intermediate resistance and ≥ 60 high-level resistance.
A GSS was generated for the background regimen by summing the level of resistance to each antiretroviral drug in the regimen other than the PI/r. Separate GSS were calculated for each GIS. Resistance to ritonavir was not included in this calculation because it was not administered in therapeutic doses. Neither DMC nor Stanford had GIS available for all PI/r so interpretations for the unboosted PI were used instead. These four GIS were chosen because they are rules-based, updated regularly, freely available and possible to programme. They were programmed in STATA (version 9.2; Stata Corp., College Station, Texas, USA) and cross-checked using SAS (version 8.2; SAS Institute Inc., Cary, North Carolina, USA).
We evaluated concordance between predicted PI/r resistance levels for each GIS using kappa statistics . A kappa greater than 0.8 reflects very good concordance, kappa 0.6 ≤ 0.8 good agreement, kappa 0.4 ≤ 0.6 moderate agreement, kappa 0.2 ≤ 0.4 fair agreement and kappa ≤ 0.2 poor concordance or no relationship.
We related PI/r resistance levels to HIV-RNA reductions from baseline after 12 weeks (range 4–26 weeks). The most commonly used HIV-RNA assay has a lower limit of quantification of 50 copies/ml, therefore we could not estimate exact HIV-RNA declines for patients with an HIV-RNA level less than 50 copies/ml. Factors associated with HIV-RNA change were identified through censored regression analysis to take into account the partial observation of the extent of viral load reduction caused by this censoring . If a more sensitive assay was used we modelled the full HIV-RNA reductions and did not consider the patient to have a censored HIV-RNA measurement. All variables associated with HIV-RNA decline in unadjusted analysis were entered into separate multivariable models for each GIS. Each multivariable model is adjusted for the GSS to the rest of the regimen calculated using the same GIS. Statistical analyses were performed using STATA software (version 9.2; Stata Corp.).
The 744 patients started one of the following single PI/r (as opposed to double-boosted PI/r): amprenavir/ritonavir (N = 28, 4%); atazanavir/ritonavir (N = 29, 4%); indinavir/ritonavir (N = 144, 19%); lopinavir/ritonavir (N = 357, 48%); or saquinavir/ritonavir (N = 186, 25%). The characteristics of patients at the time of starting the PI/r (baseline), as shown in Table 1, are reasonably comparable for each PI/r. Overall, patients were a median (interquartile range; IQR) age of 40 years (35, 47 years), primarily male (76%), antiretroviral experienced (85%) and PI experienced (67%). Patients were receiving a varied number and type of backbone antiretroviral drugs at baseline (data not shown). Median (IQR) baseline CD4 cell counts were 246 (133, 381) cells/μl and median (IQR) HIV-RNA levels were 4.4 (3.5, 5.1) log10 copies/ml.
Concordance between genotypic interpretation systems
At baseline 730 patients were receiving nucleoside reverse transcriptase inhibitors (NRTI), 135 patients were receiving non-nucleoside reverse transcriptase inhibitors (NNRTI) and 744 patients were receiving PI/r. Overall, 371 (51%) patients had a virus with an International AIDS Society – USA panel  resistance mutation to one of their baseline NRTI, 31 (23%) had a resistance mutation to a baseline NNRTI and 133 (18%) had a major PI resistance mutation to a PI/r they were receiving at entry. Nearly double the number of patients (N = 273, 37%) had any PI mutation (i.e. including minor mutations) to a baseline PI/r. Using REGA, 110 (15%) patients had a virus exhibiting full or intermediate resistance to the PI/r they received (all PI/r combined). ANRS calculated similar levels of resistance to REGA but both DMC and Stanford predicted nearly double the number of patients with a virus exhibiting full or intermediate resistance to their baseline PI/r. The breakdown of resistance by PI/r and GIS is shown in Table 2.
Concordance between GIS on the predicted level of resistance to the PI/r was moderate. Kappas ranged from 0.01 to 0.38 for amprenavir/ritonavir; 0.31 to 1.00 for atazanavir/ritonavir; 0.45 to 0.77 for indinavir/ritonavir; 0.38 to 0.82 for lopinavir/ritonavir and 0.36 to 0.68 for saquinavir/ritonavir (Table 3).
Figure 1a crudely shows the codons that cause discordance between GIS for each drug class. Although ANRS considers 25 codons in the protease gene to be associated with resistance to a PI/r in this study (i.e. three more mutations than any other GIS), it estimates that fewer patients have a virus with full or intermediate resistance to their PI/r compared with the other GIS. This is because mutations occurring at these codons are either infrequent or need to be considered in combination with other mutations to infer resistance using ANRS. Figure 1b shows a high level of discordance when examining interpretations for each PI/r separately. DMC predicts significantly higher levels of resistance to amprenavir/ritonavir, indinavir/ritonavir and saquinavir/ritonavir than any other GIS. This is primarily driven by the presence of the I93L protease mutation, which occurs in 275 (37%) of our patients, but is not considered to have an effect on resistance using any GIS other than DMC.
Prognostic value of genotypic interpretation systems for predicting viral load decline
Ritonavir-boosted protease inhibitor sensitivity
The median (IQR) baseline HIV-RNA level was 4.4 (3.5, 5.1) log10 copies/ml. After a median (IQR) 12 (9, 13) weeks from the start of a PI/r-containing regimen this was reduced by a mean [95% confidence interval (CI)] 2.2 (2.1, 2.3) log10 copies/ml. There were 236 (32%) patients with a follow-up HIV-RNA recorded as 50 copies/ml or less in whom the reduction was censored in the regression analysis.
The level of resistance to the PI/r (all PI/r combined) was related to HIV-RNA reductions for all GIS in unadjusted analysis (P ≤ 0.0001). Patients with a virus that is considered to be resistant to the PI/r, using the REGA GIS, experienced a mean (95% CI) 1.4 (0.9, 1.8) log10 decline, patients harbouring a virus with intermediate resistance had a mean 1.6 (1.2, 1.9) log10 decline, and patients with a sensitive virus had a 2.3 (2.2, 2.5) log10 copies/ml decline.
After adjusting for baseline HIV-RNA, the PI/r used, the time between baseline and follow-up HIV-RNA measurement and the GSS to the rest of the regimen (calculated using the same GIS that was used to calculate the level of PI/r resistance), all GIS still showed significantly greater reductions as sensitivity to the PI/r increased. Patients with a virus that was sensitive to the PI/r experienced a 0.82 greater log10 reduction between baseline and week 12 compared with patients with a virus exhibiting full resistance using REGA (Table 4). We then examined these trends after adjusting for a fixed GSS to the backbone regimen (i.e. adjusting for the REGA GSS rather than each corresponding GSS individually) while varying the system used for the PI/r. This led to a reduction in the noise from the background regimen, but the same patterns between PI/r resistance levels and virological response were observed (data not shown).
The genotypic sensitivity score to the rest of the regimen
There were 490 (66%), 483 (65%), 438 (59%) and 469 (63%) patients with a virus that was susceptible to two or more antiretroviral drugs in their background regimen (i.e. antiretroviral drugs other than the PI/r) using ANRS, DMC, REGA and Stanford, respectively.
We explored the relationship between sensitivity to the PI/r and the GSS to the rest of the regimen to assess whether patients with high degrees of viral sensitivity to their regimen also had PI/r sensitivity. The GSS to the rest of the regimen was similar for each PI/r resistance level: using REGA, the mean (95% CI) GSS to the rest of the regimen was 1.8 (1.4, 2.2) for patients with viral resistance at baseline, 1.7 (1.5, 1.9) for patients who had a virus with intermediate baseline resistance and 1.6 (1.6, 1.7) for patients who had a virus with sensitivity to the PI/r.
Differences existed between GIS in their abilities to predict HIV-RNA declines according to the GSS to the rest of the regimen. There was a trend for greater HIV-RNA reductions with increasing GSS to the rest of the regimen in both unadjusted and adjusted analysis for REGA and ANRS (Table 4). Both GIS showed greater HIV-RNA declines with increasing PI/r sensitivity and with increasing GSS (Fig. 2). Neither DMC nor Stanford showed a relationship between the GSS to the rest of the regimen and the HIV-RNA response.
Sensitivity analysis excluding protease inhibitor-naive patients
The GIS were created using information on drug-experienced patients so we performed a subgroup analysis on our 502 PI-experienced patients. PI-experienced patients display higher levels of baseline resistance: there were 130 (26%) PI-experienced patients with a major International AIDS Society – USA panel PI-resistance mutation to a PI/r they were receiving at baseline. Totals of 295 (59%), 293 (59%), 253 (50%) and 294 (59%) patients had a virus that was susceptible to two or more antiretroviral drugs in their background regimen using ANRS, DMC, REGA, and Stanford, respectively. We saw similar trends between PI/r resistance levels and virological response in this subgroup of patients to the study overall (Fig. 3).
Drug-resistant HIV-1 can adversely impact on virological and clinical outcomes in both treated and untreated individuals [39–41]. Genotyping provides lists of mutations that are present as the majority in viral RNA so they can be translated into a sensitivity score using a GIS. In our study we compared four GIS and found moderate concordance between them on their predicted levels of resistance to a PI/r. Other studies that have evaluated concordance compared more GIS or investigated a larger number of antiretroviral drugs, resulting in more discordance [23,24].
Ravela et al.  investigated concordance between ANRS, REGA, Stanford and Visible Genetics and found discordance between 33.6% of interpretations, mainly driven by the NRTI. Among the PI/r, they found similar kappas for indinavir [from 0.61 (REGA versus Stanford) to 0.80 (REGA versus NRS)] and lopinavir/ritonavir [from 0.40 (ANRS versus Stanford) to 0.67 (REGA versus ANRS)], but higher kappas for amprenavir/ritonavir [from 0.48 (ANRS versus Stanford) to 0.72 (REGA versus Stanford)] and saquinavir/ritonavir [from 0.63 (REGA versus Stanford) to 0.77 (ANRS versus Stanford)] compared with our study. They defined discordance as one GIS assigning a resistant score and another a sensitive score. If a patient had intermediate viral resistance they were considered neutral, and no discordance between either resistant or sensitive strains was noted. We would therefore expect to see larger kappa values in the study by Ravela et al.  compared with ours.
The comparison by Ravela et al.  uses GIS that have since been updated to account for the latest developments in research and some to account for the use of ritonavir boosting. Originally the GIS were created using the same data with similar expert interpreters so interpretations were comparable. As time progressed more drugs entered the market, more resistance mutations were identified, and opinions on what mutations contribute towards resistance started to diverge. In addition, only some GIS took into account the clinically derived genotypic susceptibility scores published in recent years [42–45]. Now antiretroviral drugs have been used more frequently so there is more of a consensus on resistance profiles. This should have resulted in higher kappa values in our study, but this was not the case. We found substantially more discordance for amprenavir/ritonavir, which could result from the use of different formulations (i.e. amprenavir/ritonavir, amprenavir alone or fosamprenavir). Different formulations may lead to different rates of non-compliance and to disagreements between experts on the degree of resistance caused by each mutation. There is a consensus on whether most mutations affect amprenavir/ritonavir resistance, but the weighting attributed to each mutation differs according to the GIS, resulting in high levels of discordance. Examination of a larger group of patients receiving amprenavir/ritonavir would improve these comparisons.
As virological outcomes are pivotal for testing the clinical use of GIS we examined the relationship between PI/r resistance and HIV-RNA change. Although each GIS considers different mutations to be relevant for each antiretroviral drug (Fig. 1), they all predict virological response to a similar extent. HIV-RNA was reduced by 2.2 (2.1, 2.3) log10 after a median 12 (9, 13) weeks. This is larger than other studies comparing GIS for predicting virological decline [4,5,46], but the other studies were conducted on patients receiving salvage therapy who were less likely to experience large reductions. Our HIV-RNA reductions are greater because our population contains more varied patients, including patients who were naive to antiretroviral therapy, failing a first-line regimen or failing a later regimen. In addition, we modelled HIV-RNA decline using a censored regression analysis to take into account the fact that a patient's HIV-RNA level could fall below the lower limit of detection during follow-up.
All GIS showed significantly greater HIV-RNA reductions as sensitivity to the PI/r increased. The GSS to the rest of the regimen was not associated with response for either DMC or Stanford, although in the latter case there was a trend in the expected direction. Our results are consistent with a study by Miller et al. , in which resistance to PI was shown to be a stronger predictor of virological failure at week 24 than NRTI resistance. Conversely, De Luca et al.  found that the GSS to the whole regimen (including the PI/r) did not predict response to a PI/r-containing regimen, but the independent effect of PI/r sensitivity may have been overshadowed by sensitivity to the rest of the regimen in their study because the regimen was studied as a whole rather than PI/r separately.
Although current GIS are able to identify differences in virological response according to PI/r resistance levels, patients with a virus deemed fully resistant to the PI/r still experience large HIV-RNA reductions (i.e. a reduction of > 1 log10 copies/ml). PI/r may exert antiviral effects in the presence of resistance; however, this observation is probably partly caused by susceptibility to the nucleoside backbone. If a patient has a high GSS to the other drugs in the regimen they appear to experience large HIV-RNA reductions despite the presence of PI/r resistance (Fig. 2). Patients with a virus that is fully resistant to all of the antiretroviral drugs in their regimen only experienced small HIV-RNA reductions.
In our study more than 70% of patients were susceptible to the PI/r they were receiving. In those patients the GSS to the rest of the regimen was not an important determinant of HIV-RNA decline because sensitivity to the PI/r is the major factor influencing response. Even though GIS vary on what resistance mutations they consider to be important, as reflected by the low-to-moderate kappa values, the relationship between resistance and HIV-RNA change for each GIS illustrates that the GIS do effectively discriminate viruses according to sensitivity to PI/r. This is somewhat surprising considering that some of the GIS have not yet been updated to take into account ritonavir boosting.
Our study has several limitations: we do not have a measure of adherence in our patients and as the predictive ability of the GIS is stronger in adherent patients , non-adherence could dilute the strength of association between each GIS and response. In the trials, pharmacokinetic analysis after 4 weeks of follow-up revealed high PI drug levels in the majority of patients (data not shown), reflecting good adherence overall. Although adherence rates are known to diminish with time on a regimen , the amount of bias introduced by examining the 12-week virological response as opposed to an earlier timepoint is likely to be minimal. No adherence data were available in EuroSIDA. Another limitation is that a large proportion of our study population is PI-naive and will respond to therapy differently to experienced patients. When we repeated the analysis in the subgroup of PI-experienced patients we found similar trends.
A wider limitation of genotypic resistance testing concerns the practical implications of dealing with consensus sequences rather than clones. The use of this method could result in a virus being randomly selected from the quasispecies in circulation, and therefore clinically significant resistant mutations present in minor populations may be overlooked. The presence of undetected or archived species is a limitation to our study and a major problem with building GIS in general. Treatment success beyond the first few weeks of the start of a regimen is likely to be dependent on minor populations and therefore ideally they need to be incorporated into GIS.
To conclude, differences exist between GIS; however, the level of disagreement for predicting short-term HIV-RNA change is relatively minor. Although GIS have been shown to work effectively, when patients with a virus deemed to be sensitive to their regimen experience the largest HIV-RNA declines, GIS need further refinement to improve discrimination and concordance.
1. Mocroft A, Vella S, Benfield TL, Chiesi A, Miller V, Gargalianos P, et al
. Changing patterns of mortality across Europe in patients infected with HIV-1. EuroSIDA Study Group. Lancet 1998; 352:1725–1730.
2. Palella FJJ, Delaney KM, Moorman AC, Loveless MO, Fuhrer J, Satten GA, et al
. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med 1998; 338:853–860.
3. Hammer SM, Saag MS, Schechter M, Montaner JSG, Schooley RT, Jacobsen DM, et al
. Treatment for adult HIV infection – 2006 recommendations of the International AIDS Society – USA panel. JAMA 2006; 296:827–843.
4. Durant J, Clevenbergh P, Halfon P, Delgiudice P, Porsin S, Simonet P, et al
. Drug-resistance genotyping in HIV-1 therapy: the VIRAD APT randomised controlled trial. Lancet 1999; 353:2195–2199.
5. Baxter JD, Mayers DL, Wentworth DN, Neaton JD, Hoover ML, Winters MA, et al
. A randomized study of antiretroviral management based on plasma genotypic antiretroviral resistance testing in patients failing therapy. AIDS 2000; 14:F83–F93.
6. Zolopa AR, Shafer RW, Warford A, Montoya JG, Hsu P, Katzenstein D, et al
. HIV-1 genotypic resistance patterns predict response to saquinavir–ritonavir therapy in patients in whom previous protease inhibitor therapy had failed. Ann Intern Med 1999; 131:813–821.
7. Larder BA, Kellam P, Kemp SD. Convergent combination therapy can select viable multidrug-resistant HIV-1 in vitro
. Nature 1993; 365:451–453.
8. Bazmi HZ, Hammond JL, Cavalcanti SCH, Chu CK, Schinazi RF, Mellors JW. In vitro selection of mutations in the human immunodeficiency virus type 1 reverse transcriptase that decrease susceptibility to (−)-beta-D-dioxolane-guanosine and suppress resistance to 3′-azido-3′-deoxythymidine. Antimicrob Agents Chemother 2000; 44:1783–1788.
9. Margot NA, Isaacson E, McGowan I, Cheng AK, Schooley RT, Miller MD. Genotypic and phenotypic analyses of HIV-1 in antiretroviral-experienced patients treated with tenofovir DF. AIDS 2002; 16:1227–1235.
10. Tozzi V, Zaccarelli M, Narciso P, Trotta MP, Ceccherini-Silberstein F, De Longis P, et al
. Mutations in HIV-1 reverse transcriptase potentially associated with hypersusceptibility to nonnucleoside reverse-transcriptase inhibitors: effect on response to efavirenz-based therapy in an urban observational cohort. J Infect Dis 2004; 189:1688–1695.
11. Garcia-Lerma JG, MacInnes H, Bennett D, Weinstock H, Heneine W. Transmitted HIV-1 carrying D67N or K219Q evolve rapidly to zidovudine resistance in vitro
and show a high replicative fitness in the presence of zidovudine. Antiviral Ther 2003; 8:U81–U82.
12. Cong M, Bennett DE, Heneine W, Garcia-Lerma JG. Fitness cost of drug resistance mutations is relative and is modulated by other resistance mutations: implications for persistence of transmitted resistance. Antiviral Ther 2005; 10:S169.
13. Boyer PL, Sarafianos SG, Clark PK, Arnold E, Hughes SH. Why do HIV-1 and HIV-2 use different pathways to develop AZT resistance? PLoS Path 2006; 2:e10.
14. Been-Tiktak A, Korn K, Keulen W, Schwingel E, Walter H, Schmidt B, et al
. Evaluation of an open expert-based genotype interpretation program: RetroGram. In: 41st Interscience Conference on Antimicrobial Agents and Chemotherapy. Chicago, IL. 15–19 December 2001. Abstract I-1745.
15. Agence Nationale de Recherches sur le SIDA (ANRS). ANRS genotypic resistance guidelines – version 13. Electronic citation, 2005. Available at: http://www.hivfrenchresistance.org/
.Accessed: May 2007.
16. De Luca A, Vendittelli M, Boldini F, Di Giambenedetto S, Trotta MP, Cingolani A, et al
. Construction, training and clinical validation of an interpretation system for genotypic HIV-1 drug resistance based on fuzzy rules revised by virological outcomes. Antiviral Ther 2004; 9:583–593.
17. MacArthur RD. An updated guide to genotype interpretation. AIDS Reader 2004; 14:256–266.
18. Ormaasen V, Sandvik L, Asjo B, Holberg-Petersen M, Gaarder PI, Bruun JN. An algorithm-based genotypic resistance score is associated with clinical outcome in HIV-1-infected adults on antiretroviral therapy. HIV Med 2004; 5:400–406.
22. De Luca A, Perno CF. Impact of different HIV resistance interpretation by distinct systems on clinical utility of resistance testing. Curr Opin Infect Dis 2003; 16:573–580.
23. Kijak GH, Rubio AE, Pampuro SE, Zala C, Cahn P, Galli R, et al
. Discrepant results in the interpretation of HIV-1 drug-resistance genotypic data among widely used algorithms. HIV Med 2003; 4:72–78.
24. Sturmer M, Doerr HW, Staszewski S, Preiser W. Comparison of nine resistance interpretation systems for HIV-1 genotyping. Antiviral Ther 2003; 8:239–244.
25. Braun P, Helm M, Ehret R, Schmidt B, Sturner KH, Walter H, et al
. Predictive value of different drug resistance interpretation systems in therapy management of HIV-infected patients in daily routine. Antiviral Ther 2002; 7(Suppl. 1):S77.
26. Cabrera C, Cozzi-Lepri A, Phillips AN, Loveday C, Kirk O, It-Khaled M, et al
. Baseline resistance and virological outcome in patients with virological failure who start a regimen containing abacavir: EuroSIDA study. Antiviral Ther 2004; 9:787–800.
27. De Luca A, Antinori A, Di Giambenedetto S, Cingolani A, Colafigli M, Perno CF, et al
. Interpretation systems for genotypic drug resistance of HIV-1. Scand J Infect Dis Suppl 2003; 106:29–34.
28. Snoeck J, Kantor R, Shafer RW, Van Laethem K, Deforche K, Carvalho AP, et al
. Discordances between interpretation algorithms for genotypic resistance to protease and reverse transcriptase inhibitors of human immunodeficiency virus are subtype dependent. Antimicrob Agents Chemother 2006; 50:694–701.
29. Ravela J, Betts BJ, Brun-Vezinet F, Vandamme AM, Descamps T, Van Laethem K, et al
. HIV-1 protease and reverse transcriptase mutation patterns responsible for discordances between genotypic drug resistance interpretation algorithms. J Acquir Immune Defic Syndr 2003; 33:8–14.
30. Gazzard B. British HIV Association (BHIVA) guidelines for the treatment of HIV-infected adults with antiretroviral therapy. HIV Med 2006; 7:487.
31. Mocroft A, Ledergerber B, Katlama C, Kirk O, Reiss P, D'Arminio-Monforte A, et al
. Decline in the AIDS and death rates in the EuroSIDA study: an observational study. Lancet 2003; 362:22–29.
32. Dragsted UB, Gerstoft J, Youle M, Fox Z, Losso M, Benetucci J, et al
. A randomized trial to evaluate lopinavir/ritonavir versus saquinavir/ritonavir in HIV-1-infected patients: the MaxCmin2 trial. Antiviral Ther 2005; 10:735–743.
33. Fox Z, Dragsted UB, Gerstoft J, Phillips AN, Kjaer J, Mathiesen L, et al
. A randomized trial to evaluate continuation versus discontinuation of lamivudine in individuals failing a lamivudine-containing regimen: the COLATE trial. Antiviral Ther 2006; 11:761–770.
34. Bak Dragsted U, Gerstoft J, Pedersen C, Peters B, Duran A, Obel N, et al
. Randomized trial to evaluate indinavir/ritonavir versus saquinavir/ritonavir in human immunodeficiency virus type 1-infected patients: the MaxCMin1 Trial. J Infect Dis 2003; 188:635–642.
35. Raffi F, Katlama C, Saag M, Wilkinson M, Chung J, Smiley L, et al
. Week-12 response to therapy as a predictor of week 24, 48, and 96 outcome in patients receiving the HIV fusion inhibitor enfuvirtide in the T-20 versus Optimized Regimen Only (TORO) trials. Clin Infect Dis 2006; 42:870–877.
36. Sim J, Wright CC. The kappa statistic in reliability studies: use, interpretation, and sample size requirements. Phys Ther 2005; 85:257–268.
37. Marschner IC, Betensky RA, DeGruttola V, Hammer SM, Kuritzkes DR. Clinical trials using HIV-1 RNA-based primary endpoints: statistical analysis and potential biases. J Acquir Immune Defic Syndr Hum Retrovirol 1999; 20:220–227.
38. Johnson VA, Brun-Vezinet F, Clotet B, Kuritzkes DR, Pillay D, Schapiro JM, et al
. Update of the drug resistance mutations in HIV-1: fall 2006. Topics HIV Med 2006; 14:125–130.
39. Weinstein MC, Goldie SJ, Losina E, Cohen CJ, Baxter JD, Zhang H, et al
. Use of genotypic resistance testing to guide HIV therapy: clinical impact and cost-effectiveness. Ann Intern Med 2001; 134:440–450.
40. Sax PE, Islam R, Walensky RP, Losina E, Weinstein MC, Goldie SJ, et al
. Should resistance testing be performed for treatment-naive HIV-infected patients? A cost-effectiveness analysis. Clin Infect Dis 2005; 41:1316–1323.
41. Johnson JA, Lil JF, Wei X, Craig C, Stone C, Horton JH, et al
. Baseline detection of low-frequency drug resistance-associated mutations is strongly associated with virological failure in previously antiretroviral-naive HIV-1 infected persons. In: 15th HIV Drug Resistance Workshop. Sitges, Spain. 13–17 June 2006. Abstract 69.
42. Marcelin AG, Dalban C, Peytavin G, Lamotte C, Agher R, Delaugerre C, et al
. Clinically relevant interpretation of genotype and relationship to plasma drug concentrations for resistance to saquinavir-ritonavir in human immunodeficiency virus type 1 protease inhibitor-experienced patients. Antimicrob Agents Chemother 2004; 48:4687–4692.
43. Marcelin AG, Chazallon C, Gerard L, Saidi Y, Aboulker JP, Girard PM, et al
. External validation of atazanavir/ritonavir genotypic score in HIV-1 protease inhibitor-experienced patients. J Acquir Immune Defic Syndr 2006; 42:127–128.
44. Vora S, Marcelin AG, Guinthard HF, Flandre P, Hirsch HH, Masquelier B, et al
. Clinical validation of atazanavir/ritonavir genotypic resistance score in protease inhibitor-experienced patients. AIDS 2006; 20:35–40.
45. Marcelin AG, Flandre P, Pavie J, Schmidely N, Wirden M, Lada O, et al
. Clinically relevant genotype interpretation of resistance to didanosine. Antimicrob Agents Chemother 2005; 49:1739–1744.
46. De Luca A, Cingolani A, Di Giambenedetto S, Trotta MP, Baldini F, Rizzo MG, et al
. Variable prediction of antiretroviral treatment outcome by different systems for interpreting genotypic human immunodeficiency virus type I drug resistance. J Infect Dis 2003; 187:1934–1943.
47. Miller V, Cozzi-Lepri A, Hertogs K, Gute P, Larder B, Bloor S, et al
. HIV drug susceptibility and treatment response to mega-HAART regimen in patients from the Frankfurt HIV cohort. Antiviral Ther 2000; 5:49–55.
48. Daar ES, Cohen C, Remien R, Sherer R, Smith K. Improving adherence to antiretroviral therapy. AIDS Reader 2003; 13:81–90.