©2011 International Medical Press 1359-6535 (print) 2040-2058 (online) 395

Antiviral Therapy 2011; 16:395–404 (doi: 10.3851/IMP1759)

Background: Maraviroc-containing regimens are known to achieve virological suppression in many treatment- experienced patients. This study aimed to evaluate a more rigorous methodological approach to resistance–response analysis in large clinical studies and to better establish which subpopulations of patients were most likely to ben-efit from maraviroc by refining and extending previous sub-group analyses from the MOTIVATE studies.Methods: Individual weighted optimized background ther-apy (OBT) susceptibility scores were calculated by combin-ing genotypic or phenotypic resistance testing with prior drug use information. Virological response (HIV-1 RNA<50 copies/ml at week 48) using each of these methods was compared with a commonly used method of counting active drugs. Baseline predictors of virological response, including weighted or unweighted scoring, maraviroc use, baseline

CD4+ T-cell count, HIV-1 plasma viral load and tropism, were assessed by logistic regression modelling.Results: Genotypic or phenotypic weighted methods were similarly predictive of virological response and bet-ter than counting active drugs. Weighted scoring and baseline CD4+ T-cell count were the strongest predic-tors of virological response (P<0.0001): ≈70% of mara-viroc patients with a weighted score ≥2 had a virological response, rising to ≈80% when the baseline CD4+ T-cell count was ≥50 cells/mm3.Conclusions: Approximately 80% of patients with a CD4+ T-cell count ≥50 cells/mm3 receiving maraviroc with the equivalent of at least two fully active agents achieved HIV-1 RNA<50 copies/ml at week 48 in the MOTIVATE studies. Genotypic and phenotypic weighted scores were similarly predictive of virological response.

Maraviroc is an orally available antagonist that binds to the chemokine receptor CCR5 and prevents its interaction with the HIV-1 envelope [1]. Maraviroc is approved for twice-daily administration in the combi-nation treatment of CCR5-tropic HIV-1 infection. The clinical efficacy of maraviroc was initially established in treatment-experienced patients in two Phase III clinical studies: MOTIVATE 1 (Maraviroc plus Opti-mized Therapy In Viraemic Antiretroviral Treatment-

Experienced patients 1) in the USA and Canada, and MOTIVATE 2 in Europe, Australia and the USA. In both studies, treatment-experienced patients with a CCR5 (R5) screening tropism result were randomized to receive maraviroc (once or twice daily) or placebo along with an optimized background therapy (OBT) of 3–6 approved agents. Individual and pooled analyses from these studies showed significantly greater viro-logical and immunological responses over 48 weeks

Original article

Baseline CD4+ T-cell counts and weighted background susceptibility scores strongly predict response to maraviroc regimens in treatment-experienced patients

Jonathan M Schapiro1, Charles AB Boucher 2,3, Daniel R Kuritzkes4, David A van de Vijver5, Josep M Llibre6, Marilyn Lewis7, Paul Simpson7, Christophe Delogne7, Lynn McFadyen7, Douglass Chapman8, Manos Perros7, Hernan Valdez8*, Elna van der Ryst7, Mike Westby7

1National Hemophilia Center, Tel Hashomer, Israel 2Department of Virology, Erasmus Medical Centre, Erasmus University, Rotterdam, the Netherlands 3Department of Medical Microbiology, University Medical Center Utrecht, Utrecht, the Netherlands 4Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA 5Erasmus Medical Centre, University Medical Centre Rotterdam, Rotterdam, the Netherlands 6Lluita contra la SIDA Foundation, Hospital Germans Trias i Pujol, Badalona, Spain 7Pfizer Global Research & Development, Sandwich, UK 8Pfizer Inc, New York, NY, USA

*Corresponding author e-mail: hernan.valdez@pfizer.com

Introduction

AVT-10-OA-1711_Schapiro.indd 395 6/5/11 09:09:03

©2011 International Medical Press396

JM Schapiro et al.

in patients receiving maraviroc compared with those receiving placebo [2,3].

Subgroup analyses have established that simple binary susceptibility estimates of OBT regimen activity, maraviroc use, viral load, CD4+ T-cell count and the first use of enfuvirtide or selected protease inhibitors (PIs) as fully active new agents are associated with virologi-cal outcome in the MOTIVATE studies [4]. These data are consistent with other treatment-experienced patient studies of the association between the number of active drugs and the rate of virological response [5,6].

The commonly used approach to estimating regimen activity in treatment-experienced patient studies, includ-ing the MOTIVATE subanalyses described above, has several limitations. In the binary assignation, active agents with full susceptibility are coded as 1, whereas any reduced susceptibility is coded as 0. Regimen activity scores are typically derived by counting all coadministered agents deemed active by resistance testing. This approach implicitly assumes an equal contribution by each active agent towards the total antiviral activity of the regimen. This is not always the case, as recycled but active nucleo-side reverse transcriptase inhibitors (NRTIs) contrib-ute less to the antiviral activity of a regimen than active agents from other classes [7–9]. Moreover, agents that experience a cumulative loss of susceptibility over a wide dynamic range, such as protease inhibitors (PIs), might exhibit significant partial activity. In addition, for patients with detectable baseline viraemia on a pre-study regimen, scoring agents carried forward from the failing regimen into the study regimen will confound correlations with on-study virological responses, given that such agents are unlikely to contribute to post-baseline reductions in plasma HIV-1 RNA. Finally, alterations to regimen com-position over the period of active on-study therapy are seldom controlled for in activity–response analyses that score the initial set of drugs received. Resistance–response studies are important for establishing the activity of new agents as a function of pre-existing resistance to older companion drugs. However, the result of these methodo-logical problems is that the clinical datasets used in such analyses are seldom ‘clean’ and the final response out-come will in part be determined by associations with hid-den covariates not included in the evaluation.

Current treatment guidelines for treatment- experienced patients recommend maximally suppress-ing HIV RNA levels (for example, to <50 copies/ml) and to prevent further selection of resistance mutations [10]: two goals that are intimately linked. New regi-mens should include at least two, and preferably three, fully active agents that should include new antiretro-viral drugs, including some with new mechanisms of action [10]. With the availability of four agents active against multidrug-resistant HIV (maraviroc, ralte-gravir, darunavir and etravirine) over the past few

years, questions exist regarding how many agents are needed for complete virological control in treatment- experienced patients, which patients have the highest probability of response and whether a more rigorous method of assessing regimen activity might improve patient management by more accurately establishing the predictive associations between baseline conditions and virological outcome. The availability of both genotypic and phenotypic methodologies for assessing drug activ-ity also raises the question as to whether one approach correlates better with response than the other.

We have undertaken a post-hoc analysis on pooled data from the MOTIVATE studies in which we com-pare week 48 virological outcomes for a standard ‘drug counting’ approach to estimating regimen activity ver-sus a weighted system based on both genotypic and phenotypic resistance testing.

Methods

MOTIVATE study populationMOTIVATE 1 and 2 (ClinicalTrials.gov numbers NCT00098306 and NCT00098722, respectively) enrolled treatment-experienced adults (89% male and 84% White) with exclusively CCR5-tropic (R5) HIV-1, plasma viraemia ≥5,000 RNA copies/ml at screening and either ≥6 months prior experience of antiretroviral agents from at least three classes (NRTIs, non-nucleo-side reverse transcriptase inhibitors [NNRTIs], PIs and enfuvirtide) or documented HIV-1 resistance to ≥3 of these classes. The design and 48-week outcome of these studies have been described elsewhere [3]. Raltegravir, etravirine and darunavir were not permitted in the OBT. Tipranavir was initially excluded, but was later allowed following the availability of maraviroc drug interaction data. The pooled study population was mostly (93.6% [982/1,049]) infected with subtype B HIV-1.

Analysis populationsThe analysis populations were defined to minimize confounding by non-virological failures and on-study reg-imen changes and their derivation is shown in Figure 1. Briefly, patients were included in this analysis if they had a complete resistance and tropism testing dataset avail-able at screening, achieved a virological outcome (defined by time to loss of virological response; see Statistical anal-yses) by week 48 [11], began their OBT regimen within a 7-day window of their ‘day 1’ baseline study visit (that is, day -6 to +7) and remained on an unchanged drug regi-men during the entire treatment phase. As not all patients possessed both genotypic and phenotypic drug resistance data at baseline, three analysis populations were derived from the full analysis set of 1,049 patients in MOTIVATE 1 and 2 by application of the above selection criteria: a genotype population of patients with genotypic resistance

AVT-10-OA-1711_Schapiro.indd 396 6/5/11 09:09:03

Predictors of maraviroc response

Antiviral Therapy 16.3 397

data (n=604), a phenotype population with phenotype data (n=631) and a combined genotype/phenotype popu-lation with both (n=598). For the logistic regression anal-ysis described below, one patient who lacked complete baseline covariate data was excluded from all three of the analysis populations described to yield modelling popula-tions of 603, 630 and 597 patients, respectively (Figure 1). Of note, all patients (n=145) who discontinued the study over 48 weeks for non-virological failures such as adverse events or loss to follow-up were excluded.

Scoring methodologiesTwo weighted and one unweighted OBT scoring methods were defined, based on OBT drug susceptibilities assessed in screening visit samples. All three scores excluded mar-aviroc (which as a new drug class is assumed to be fully active) and ritonavir when it was used at subtherapeutic doses to boost other PIs. Reverse transcriptase and pro-tease sequencing and phenotypic resistance testing were performed by Monogram Biosciences (South San Fran-cisco, CA, USA). All three scoring methods used gp41

Full analysis setn=1,049

g-wOBTSSversus OSS

p-wOBTSSversus OSS

g-wOBTSSversus p-wOBTSS

Non-virologicalfailures excluded

n=145

On-study week 48or virological failure

n=904

Missing correspondingOSS result

n=13

Incomplete baselineinformation

n=1

Incomplete baselineinformation

n=1

Final genotypicpopulation

n=604

g-VOn=617

p-VOn=636Missing corresponding

p-wOBTSS/g-wOBTSSvalues

g-VO: n=19p-VO: n=38

287 Exclusions for ARVadministration events:• Baseline visit >7 days before randomized therapy started (n=8)• OBT changes before date of last active therapy or ARV changes between screening and start of OBT window (n=214)• OBT drugs interrupted and restarted within OBT window (n=33)• Unavailable genotypic resistance test result for an OBT drug (n=61)

268 Exclusions for ARVadministration events:• Baseline visit >7 days before randomized therapy started (n=8)• OBT changes before date of last active therapy or ARV changes between screening and start of OBT window (n=214)• OBT drugs interrupted and restarted within OBT window (n=33)• Unavailable phenotypic resistance test result for an OBT drug (n=27)

Final combinedpopulation

n=598

Final phenotypicpopulation

n=631

Phenotypic regressionmodel population

n=630

Combined regressionmodel population

n=597

Genotypic regressionmodel population

n=603

Missing correspondingOSS result

n=5

Incomplete baselineinformation

n=1

Figure 1. Patients included in the analyses

ARV, antiretroviral; g-VO, genotypic virological outcome; g-wOBTSS, genotypic weighted optimized background therapy sensitivity score; OBT, optimized background therapy; OSS, overall susceptibility score; p-VO, phenotypic virological outcome; p-wOBTSS, phenotypic weighted optimized background therapy sensitivity score.

AVT-10-OA-1711_Schapiro.indd 397 6/5/11 09:09:07

JM Schapiro et al.

©2011 International Medical Press398

genotypic sequencing alone (BC Centre for Excellence in HIV/AIDS, Vancouver, Canada) to determine suscep-tibility to enfuvirtide. Tropism was determined at the screening and baseline visits using the original Trofile™ assay (Monogram Biosciences).

Unweighted scoring was performed using the overall susceptibility score (OSS), a simple binary drug-counting approach used as the major protocol-specified assessment of OBT activity in the MOTIVATE studies in previous subgroup analyses [4]. The OSS was defined as the sum of all OBT drugs received to which the patient’s virus was interpreted as susceptible. For antiretroviral agents other than enfuvirtide, susceptibility was defined by a combi-nation of phenotypic and genotypic data according to a proprietary methodology (‘PhenoSense GT’; Monogram Biosciences) used to generate the screening resistance test reports provided to MOTIVATE investigators. To assess the influence of recycling drugs from the failing, pre-study regimen, drugs from these regimens that were continued into the OBT were included in the OSS if there was no evidence of resistance. All drugs with full sus-ceptibility contributed 1 point, whereas drugs with any reduced susceptibility were coded as 0 [4].

Weighted optimized background therapy sensitivity scores (wOBTSS) were derived using either phenotypic (p-wOBTSS) or genotypic (g-wOBTSS) resistance test-ing in the appropriate population. Both weighted methodologies excluded OBT drugs recycled from the failing pre-study regimen, scoring them as 0 irrespec-tive of resistance test result. The p-wOBTSS method, which has previously been used to explore antiviral phenotypic cutoffs [12], used a binary (sensitive/resistant) classification of susceptibility and scored all OBT agents (other than enfuvirtide) according to phenotypic susceptibility alone (PhenoSense assay; Monogram Biosciences) using the assay-defined lower clinical cutoff for each drug. Unlike the OSS, which also used a binary scoring system, the p-wOBTSS did not assign the same weight to all active agents: active PIs, NNRTIs and enfuvirtide contributed 1.0 to the p-wOBTSS, whereas active NRTIs contributed 0.5.

The g-wOBTSS method used a trinary (sensitive/intermediate/resistant) classification of susceptibility based on the widely used Agence Nationale de Recher-ches sur le Sida (ANRS) genotypic algorithm. This trinary classification was used to ascertain whether the additional refinement of acknowledging partial drug activities added predictive value to the weighting method explored. For all agents except unboosted PIs, the 2008 revision of the algorithm was used [13]; the nevirapine rule was used to score delavirdine (n=1). For unboosted PIs (n=9), the earlier ANRS 2002 algo-rithm was used [14]. As with the p-wOBTSS, all agents to which the screening virus was ranked as ‘sensitive’ scored 1.0 except for NRTIs, which scored 0.5. In

addition, however, in the g-wOBTSS approach PIs with an ‘intermediate’ ranking (whether boosted or not) scored 0.5, whereas all other drugs ranked as ‘interme-diate’ scored 0.

Note that although the OSS is a direct estimate of the number of active drugs in an OBT regimen, both the g-wOBTSS and the p-wOBTSS are semi- quantitative estimates of overall regimen activity and do not necessar-ily correspond to the number of active drugs received.

Statistical analysesIn the published primary efficacy analysis of the MOTIVATE studies, virological response was defined as HIV-1 RNA<50 copies/ml at week 48. For these analyses of outcomes by susceptibility score, virologi-cal response at week 48 was defined by the time to loss of virological response algorithm with a threshold for response of <50 copies/ml [15].

The restrictions on patient inclusion necessary to prevent confounding resulted in the exclusion of a sig-nificant proportion (40%) of the overall MOTIVATE patient set. To address the possibility of bias, baseline characteristics (median HIV-1 RNA, CD4+ T-cell count and numbers of resistance mutations) were compared between patients included in and excluded from the final genotypic and phenotypic analysis populations by a two-sample Student’s t-test.

Logistic regression analysis was used to assess pre-dictors of virological response for OBT scoring by g-wOBTSS or p-wOBTSS versus OSS. In addition to OBT scores, each model also included the fixed covariates of baseline CD4+ T-cell count (arithmetic mean of screen-ing and baseline visits), baseline viral load (geometric mean of up to three visits at screening, randomization and baseline), baseline tropism (DM/X4 versus R5 and DM/X4 versus unreportable) and treatment (maraviroc once- or twice-daily, combined, versus placebo).

Goodness-of-fit for each model was assessed by the Hosmer and Lemeshow (H&L) test [16], which gener-ates a χ2 test statistic for the null hypothesis that the model is correct. A larger P-value for this test implies a better fit. If the H&L P-value is <0.05 this hypothesis is rejected and it can be concluded that the model does not provide a reasonable fit.

Results

Weighted versus unweighted scoringFigure 2A shows the distribution of OSS and g-wOBTSS results in the genotypic analysis population. The propor-tions of patients with intermediate scores were broadly comparable between the two methods but, compared with the OSS, use of the g-wOBTSS resulted in a redis-tribution of patients at very high and very low score strata. The number of patients estimated to be receiving

AVT-10-OA-1711_Schapiro.indd 398 6/5/11 09:09:07

Predictors of maraviroc response

Antiviral Therapy 16.3 399

at least three active drugs in their OBT (not counting randomized maraviroc) was sevenfold higher by OSS than g-wOBTSS, whereas the number of patients esti-mated to be receiving either no active OBT drugs (OSS and g-wOBTSS =0) or only a single active nucleoside (g-wOBTSS =0.5) was 2.5-fold higher by g-wOBTSS.

In the placebo arm, where virological response was entirely attributable to OBT activity, this redistribution of patients by score was associated with incremental increases in the proportion of virological responders for increasing levels of regimen activity by g-wOBTSS estimation. By contrast, the proportion of responders remained low and similar for all patients estimated to be receiving fewer than three fully active drugs by OSS (Figure 2B). Similar results were observed for the distri-bution of patients according to p-wOBTSS values rela-tive to OSS in the phenotypic analysis population (DC, data not shown).

Figure 3 shows multivariate predictors of virological success in models that contain OSS versus g-wOBTSS scoring (Figure 3A) or OSS versus p-wOBTSS scoring (Figure 3B) in the appropriate genotypic or pheno-typic analysis populations. Owing to the low number of patients with weighted scores of three or more (Fig-ure 2), the highest OBT activity stratum was set to ≥2 for both weighted and unweighted modelling. Keeping other covariates constant, the two wOBTSS models gave a better fit to the data in their respective popula-tions than did the models that incorporated the OSS, as indicated by higher H&L P-values (0.1241 and 0.0936 for g-wOBTSS and OSS, respectively; 0.1570 and 0.0090 for p-wOBTSS and OSS, respectively). Of note, the small H&L P-value for the OSS model in the phenotypic population implies a poor fit (Figure 3B).

Increasing OBT susceptibility score (weighted or unweighted), a baseline CD4+ T-cell count ≥50 cells/mm3 and treatment with maraviroc were the strongest inde-pendent predictors of virological response in each model; baseline HIV-1 RNA<100,000 copies/ml was also signifi-cant. HIV-1 tropism at baseline visit was not a significant predictor of response in these patients, who had all pre-viously returned an R5 tropism result at their screening visit (original Trofile assay; Monogram Biosciences).

Genotypic versus phenotypic weighted scoringFigure 4 shows g-wOBTSS and p-wOBTSS logistic regression models for patients with matched genotype and phenotype data, excluding those who received pla-cebo. The two methods gave similar results for all cov-ariates and baseline tropism remained non-significant even after exclusion of the placebo arms.

Similar response rates were seen for genotypic and phenotypic weighted scoring. For example, in the mar-aviroc twice daily arm, HIV-1 RNA at week 48 was <50 copies/ml in 36%, 54% and 71% of patients with

g-wOBTSS of <1, 1 to <2 and ≥2, respectively, and in 33%, 50% and 73% of patients with p-wOBTSS of <1, 1 to <2 and ≥2, respectively. In the subpopulation with baseline CD4+ T-cell counts ≥50 cells/mm3, HIV-1 RNA at week 48 was <50 copies/ml in 46%, 62% and 77% of patients with g-wOBTSS of <1, 1 to <2 and ≥2, respectively, and in 42%, 58% and 81% of patients with p-wOBTSS of <1, 1 to <2 and ≥2, respectively.

Representativeness of the analysis populationComparisons of baseline characteristics between MOTIVATE patients who were included (n=598) or excluded (n=451) from the combined regression model population are shown in Table 1. In each treatment arm, included and excluded patients showed similar baseline HIV-1 RNA levels and CD4+ T-cell counts with no evidence of a statistical difference between the two groups (P>0.05 for all comparisons). The median number of major PI mutations, minor PI mutations and thymidine analogue mutations (International Antiviral Society USA 2008) [17] was also very similar between included and excluded patients.

Discussion

Although it is known that maraviroc-containing regi-mens achieve virological suppression in a substantial proportion of treatment-experienced patients [4,18], this post hoc study sought both to evaluate a more rig-orous methodological approach to resistance– response analysis in large clinical studies and to better establish which particular subpopulations of patients were most likely to benefit from maraviroc by refining and extend-ing previous subgroup analyses from MOTIVATE [4].

These additional analyses show that, in a study population defined to minimize confounding, baseline CD4+ T-cell count and a weighted activity scoring of only those background antiretrovirals newly initiated at study baseline were the strongest independent predic-tors of response to maraviroc in MOTIVATE 1 and 2. Overall, approximately 70% of MOTIVATE patients receiving maraviroc with a weighted OBT susceptibil-ity score of ≥2 by either phenotypic or genotypic test-ing achieved <50 HIV-1 RNA copies/ml at 48 weeks. This number rose to approximately 80% for partici-pants who received maraviroc with the equivalent of at least two fully active OBT drugs by either method and who had a baseline CD4+ T-cell count ≥50 cells/mm3. However, a reduced virological response was observed in patients with very low CD4+ T-cell counts, consist-ent with results reported for other antiretroviral drugs, such as enfuvirtide, darunavir, etravirine and raltegravir in their pivotal salvage trials [6,19], and consistent with reports that baseline CD4+ T-cell count is an independent predictor of outcome in HIV-infected patients [20,21].

AVT-10-OA-1711_Schapiro.indd 399 6/5/11 09:09:07

JM Schapiro et al.

©2011 International Medical Press400

0

0

18

13 13

21n=

Score

Score

30 33

1

OSS

A

B

g-wOBTSS

OSS (n=176) versus g-wOBTSS (n=25): 86% reduction in score ≥3

OSS (n=84) versus g-wOBTSS (n=209): 149% increase in score ≤1

25

3432

28 77 83

2

2624

28

30 54 72

≥3

31 29 28

35 67 74

Pat

ient

sa , %

50

40

30

20

10

0

0–0.5

39

31

36

44n= 70 95

1–1.5

3942

40

44 96 104

2–2.5

18

2320

21 53 52

≥3

4 4 4

5 9 11P

atie

ntsa ,

%

50

40

30

20

10

g-wOBTSS

0

0–0.5

2

3137

44nb= 70 95

1–1.5

21

5854

44 96 104

2–2.5

52

7073

21 53 52

≥3

40

78

64

5 9 11

Pat

ient

s, %

100

80

60

40

20

0

0

0

20

30

21nb= 30 33

1

OSS

7

48 51

28 77 83

2

PBO+OBT MVC QD+OBT MVC BID+OBT

10

67

57

30 54 72

≥3

51

6458

35 67 74

Pat

ient

s, %

100

80

60

40

20

Figure 2. Weighted versus unweighted scoring

(A) Distribution of patients by overall susceptibility score (OSS) and genotypic weighted optimized background therapy sensitivity score (g-wOBTSS) results and treatment arm (genotypic analysis population). (B) Proportion of patients with virological response (<50 HIV-1 RNA copies/ml at week 48) by OSS and g-wOBTSS result. aPercentage of patients within the indicated treatment arm: denominators are placebo (PBO) n=114; maraviroc once daily (MVC QD) n=228; and maraviroc twice daily (MVC BID) n=262. bFor each subgroup, this is the number of patients with a post-baseline observation used to calculate the percentage. OBT, optimized background therapy.

AVT-10-OA-1711_Schapiro.indd 400 6/5/11 09:09:08

Predictors of maraviroc response

Antiviral Therapy 16.3 401

0.1 1.0 10.0Odds ratio

Genotypic model population (n=603)

Phenotypic model population (n=630)

A

B

100.0

0.1 1.0 10.0Odds ratio

100.0

7.3 (4.3–12.3)7.2 (3.7–13.8)

10.2 (6.0–17.3)7.6 (3.9–14.6)

2.9 (1.9–4.6)3.2 (1.6–6.4)

6.4 (3.4–12.1)6.2 (3.3–11.4)

4.2 (2.2–8.1)3.9 (2.1–7.5)

2.3 (1.1–4.9)2.2 (1.0–4.5)

1.7 (1.1–2.6)1.7 (1.1–2.6)

2.5 (0.4–15.2)2.2 (0.4–13.0)

1.8 (0.8–3.8)1.8 (0.8–3.8)

5.0 (2.9–8.5)5.0 (3.0–8.5)

g-wOBTSS: n=209 vs 150OBT Score 0 vs ≥2

OBT Score 0 vs 1

OSS: n=84 vs 331

g-wOBTSS: n=205 vs 244OSS: n=84 vs 188

p-wOBTSS: n=201 vs 193OBT Score 0 vs ≥2

OBT Score 0 vs 1

OSS: n=88 vs 342

p-wOBTSS: n=201 vs 236OSS: n=88 vs 200

BL CD4+ T-cell count <50 vs >200n=113 vs 266

BL CD4+ T-cell count <50 vs 101–200n=113 vs 148

BL CD4+ T-cell count <50 vs 50–100n=113 vs 76

BL viral load ≥105 vs <105

n=249 vs 354

BL tropisma DM/X4 vs NR/NP/BLQb

n=48 vs 10

BL tropisma DM/X4 vs R5n=48 vs 545

Treatment PBO vs MVC QD or BIDn=114 vs 489

3.0 (1.9–4.8)3.3 (1.7–6.4)

7.8 (4.1–14.8)6.9 (3.7–12.6)

4.9 (2.5–9.5)4.5 (2.4–8.5)

3.0 (1.4–6.2)2.5 (1.2–5.2)

2.1 (1.3–3.2)1.7 (1.2–2.6)

2.4 (0.4–15.4)2.4 (0.4–14.4)

1.8 (0.8–3.9)1.8 (0.9–3.8)

5.7 (3.3–9.9)5.1 (3.0–8.5)

BL CD4+ T-cell count <50 vs >200n=121 vs 276

BL CD4+ T-cell count <50 vs 101–200n=121 vs 156

BL CD4+ T-cell count <50 vs 50–100n=121 vs 77

BL viral load ≥105 vs <105

n=261 vs 369

BL tropisma DM/X4 vs NR/NP/BLQb

n=51 vs 10

BL tropisma DM/X4 vs R5n=51 vs 569

Treatment PBO vs MVC QD or BIDn=117 vs 513

Figure 3. Multivariate predictors of virological response

Odds ratios and 95% confidence intervals for predictors of virological response (HIV-1 RNA<50 copies/ml at week 48) in logistic regression models using (A) genotypic weighted optimized background therapy sensitivity score (g-wOBTSS; square markers) versus overall susceptibility score (OSS; circular markers) and (B) phenotypic weighted optimized background therapy sensitivity score (p-wOBTSS; diamond markers) versus OSS (circular markers). Fractional values were rounded down in both wOBTSS models. aAll patients had an R5 tropism result at screening visit. bTropism result not available: NR, non-reportable; NP, non-phenotypable; BLQ, viral load below assay limit. DM/X4, dual-mixed tropic or CXCR4-tropic. All CD4+ T-cell counts, viral load and tropism covariates refer to baseline values (mean of screening and baseline visits for CD4+ T-cell counts; mean of log10-transformed screening, randomization and baseline visits for viral load; baseline only visit for tropism). Hosmer and Lemeshow statistics were calculated for both model populations: g-wOBTSS model 12.6601 (P=0.1241) and corresponding OSS model 12.2172 (P=0.0936); p-wOBTSS model 11.8719 (P=0.1570) and corresponding OSS model 18.7509 (P=0.0090). BID, twice daily; BL, baseline; QD, once daily; PBO, placebo; vs, versus.

AVT-10-OA-1711_Schapiro.indd 401 6/5/11 09:09:09

JM Schapiro et al.

©2011 International Medical Press402

Interestingly, both phenotypic and genotypic resistance testing returned similar results, with no observable improvement in the association between activity score and outcome accruing from the inclu-sion of partial PI activity in the score over and above

a simple binary categorization. The ANRS genotypic algorithm was used here because of its public availabil-ity and well-established use in routine resistance test-ing, but the comparable performance of most modern genotype algorithms for predicting clinical response

0.1 1.0 10.0Odds ratio

100.0

5.7 (3.3–10.0)8.9 (5.0–15.8)

2.8 (1.8–4.4)2.8 (1.7–4.6)

6.2 (3.3–11.9)7.0 (3.6–13.7)

3.7 (1.9–7.3)3.8 (1.9–7.7)

2.1 (1.0–4.4)2.4 (1.1–5.1)

1.6 (1.0–2.6)1.9 (1.2–3.1)

2.5 (0.4–15.0)2.1 (0.3–13.6)

1.8 (0.8–4.0)1.7 (0.8–3.9)

g-wOBTSS: n=166 vs 123wOBTSS 0 vs ≥2

wOBTSS 0 vs 1

p-wOBTSS: n=156 vs 147

g-wOBTSS: n=166 vs 198p-wOBTSS: n=156 vs 184

BL CD4+ T-cell count <50 vs >200n=95 vs 217

BL CD4+ T-cell count <50 vs 101–200n=95 vs 115

BL CD4+ T-cell count <50 vs 50–100n=95 vs 60

BL viral load ≥105 vs <105

n=192 vs 295

BL tropisma DM/X4 vs NR/NP/BLQb

n=41 vs 9

BL Tropisma DM/X4 vs R5n=41 vs 437

Figure 4. Logistic regression models for patients with matched genotype and phenotype

Odds ratios and 95% confidence intervals for predictors of virological response (HIV-1 RNA<50 copies/ml at week 48) in patients on maraviroc once or twice daily (n=487) in logistic regression models in the combined model population using genotypic weighted optimized background therapy sensitivity score (g-wOBTSS; square markers) or phenotypic weighted optimized background therapy sensitivity score (p-wOBTSS; diamond markers). aAll patients had an R5 tropism result at screening visit. bTropism result not available: NR, non-reportable; NP, non-phenotypable; BLQ, viral load below assay limit. DM/X4, dual-mixed tropic or CXCR4-tropic. All CD4+ T-cell counts, viral load and tropism covariates refer to baseline values (mean of screening and baseline visits for CD4+ T-cell counts; mean of log10-transformed screening, randomization and baseline visits for viral load; baseline only visit for tropism). Hosmer and Lemeshow statistics were calculated for both model populations: g-wOBTSS model 10.5538 (P=0.2283); and p-wOBTSS model 11.4369 (P=0.1781). vs, versus.

Table 1. Baseline characteristics of included and excluded patients

Only HIV-1 RNA and CD4+ T-cell count variables were tested for difference between included and excluded groups. P>0.05, comparing included and excluded patients by treatment arm for baseline viral load and CD4+ T-cell counts (two-sample Student’s t-test). BID, twice daily; IQR, interquartile range (1st–3rd quartile); MVC, maraviroc; PBO, placebo; PI, protease inhibitor; QD, once daily; TAMs, thymidine analogue mutations.

Patients included in the analysis Patients excluded from the analysisCharacteristic PBO MVC QD MVC BID PBO MVC QD MVC BID

Number of patients, n 110 226 262 99 188 164Baseline HIV-1 RNA, median log10 copies/ml (IQR) 4.9 4.8 4.8 4.8 4.9 4.9 (4.5–5.2) (4.3–5.2) (4.4–5.3) (4.3–5.3) (4.5–5.4) (4.5–5.4)Baseline CD4+ T-cell count, median cells/mm3 (IQR) 174 173 181 163 168 151 (84–290) (81–288) (75–299) (84–273) (64–284) (60–270)Major PI mutations, median (IQR) 3 3 3 2 2 3 (1–4) (1–4) (1,4) (0–4) (0–3) (0–4)Minor PI mutations, median (IQR) 7 8 8 8 7 7 (5–10) (5–10) (5–10) (4–9) (4–9) (5–10)TAMs, median (IQR) 3 3 3 3 3 3 (2–4) (2–4) (1–4) (0–4) (1–4) (2–4)

AVT-10-OA-1711_Schapiro.indd 402 6/5/11 09:09:10

Predictors of maraviroc response

Antiviral Therapy 16.3 403

[22] infers that similar observations would result from the use of an alternative platform. It is also interest-ing that a baseline non-R5 tropism result following a previous R5 screening result was not significantly asso-ciated with virological outcome. However, the caveat applies here that this is established only for a relatively small number of patients who probably had low-level CXCR4-using populations that went undetected at screening, and not for patients with established non-R5 infections where maraviroc is known to confer little or no virological benefit [23]. A newer version of the Trofile assay has become available since MOTIVATE trial screening, with greater discrimination for low-level non-R5 populations, which would probably have identified and excluded at screening most or all of those patients with non-R5 baseline results in these analyses.

Compared with the weighted methods, the OSS sub-stantially overestimated the numbers receiving a fully active HAART regimen from the OBT alone (OSS≥3) and underestimated those with little or no OBT activity. This finding is consistent with both substantial OBT recycling of ‘active’ agents from the failing pre-MOTIVATE regi-men and with the high use of nucleoside analogues with lower weighted scores. The redistribution of patients with very high and very low OSS values when weighted scoring was used was associated with a much more linear associa-tion between virological response and OBT score among those not receiving maraviroc as a new agent. OSS-type scoring is common in resistance–response analyses and correlates with virological outcomes [24,25]; our data are consistent with those from studies of darunavir and ralte-gravir that used OSS-like scoring of OBT activity [5,6]. However, comparison with the weighted approaches suggests that OSS-like methods lack discrimination and might introduce other sources of confounding.

This study is subject to several limitations. The availa-ble resistance data were cross-sectional and based on the results of a single sample at screening visit. These were the data used to guide OBT selection in MOTIVATE and hence the most appropriate data for establishing a link between the drugs chosen and subsequent response in this particular dataset. However, it is known that cross-sectional testing can underestimate the total resist-ance burden in treatment-experienced patients [26] and that a longitudinal approach incorporating previous test results can improve response to a subsequent regimen [27]. It is therefore unclear whether longitudinal data would have changed the relationship between the resist-ance assessed and the response observed. In addition, the MOTIVATE studies excluded the use of raltegravir, etravirine and darunavir – which were all investiga-tional agents at the time of study recruitment – and the use of tipranavir was limited. It is unclear whether the same similarity of outcome for genotypic versus pheno-typic resistance assessment would be observed in more

modern regimens using newer agents with less clinical outcome data for both methods of assessment. Related to this latter point, it is also unclear whether, for regi-mens involving newer agents than those in MOTIVATE, the same anticipated similarity of outcome between dif-ferent genotypic algorithms would pertain.

Formal OBT scoring methods such as these, although essential to establish statistical associations with response, are of more limited value in the routine clinic to guide new regimen selection. In conclusion, however, the data from the g-wOBTSS and p-wOBTSS models suggest that using maraviroc with two fully active drugs from classes other than NRTIs – whether activity is assessed by genotype or phenotype – should result in undetectable viraemia in the majority of patients with CCR5-tropic virus. This is particularly true for those patients who present earlier in the disease/treatment process before a profound loss of CD4+ T-cells has occurred. The avail-ability of new agents such as raltegravir, darunavir and etravirine makes it increasingly possible to construct such regimens for highly treatment-experienced patients without nucleoside analogue recycling.

Acknowledgements

The authors would like to acknowledge and thank Vijay Sharma (Eliassen Group, Inc.), Alexa Parliyan (Pfizer Inc.), John Sullivan (Pfizer Global Research and Develop-ment) and Eoin Coakley (Monogram Biosciences, Inc.) for their valuable contributions to these analyses. Edi-torial assistance was provided by Nick Fitch of Health Interactions, UK with funding by ViiV Healthcare.

These data have been presented in part at the 48th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC) and the 46th Meeting of the Infectious Diseases Society of America (IDSA) joint con-ference, Washington, DC, USA, 25–28 October 2008 (poster H-1221); the 7th European HIV Drug Resistance Workshop, 25–27 March 2009, Stockholm, Sweden; and the XVIII International HIV Drug Resistance Workshop, Fort Myers, FL, USA, 9–13 June 2009 (poster 48).

The study was sponsored and funded by Pfizer Global Research and Development.

Disclosure statement

JMS has received research support, honoraria and/or con-sulting fees from Abbott, Merck, Roche, Gilead, Glaxo-SmithKline, Pfizer, Monogram Biosciences, Tibotec, Sie-mens, Ambrillia and BMS. CABB has received research support, honoraria and/or consulting fees from Abbott, Pfizer, Roche, GlaxoSmithKline, BMS, Merck and Mon-ogram. DRK is a consultant to, and has received hono-raria and/or research grants from, Monogram (Lab-Corp), Pfizer, Schering-Plough and Tobira. DAvdV has

AVT-10-OA-1711_Schapiro.indd 403 6/5/11 09:09:10

JM Schapiro et al.

©2011 International Medical Press404

received research funding from GlaxoSmithKline and lecture sponsorship from Abbott, Boehringer-Ingelheim and Bayer. JML has received research funding, consul-tancy fees or lecture sponsorships from, or has served on advisory boards for, Abbott, Boehringer-Ingelheim, Gilead Sciences, GlaxoSmithKline, Jansen-Cilag, Merck, Pfizer and Roche. ML, PS, CD, LM, DC, MP, HV, EvdR and MW are employees of Pfizer and hold stock and stock options in Pfizer.

References1. Dorr P, Westby M, Dobbs S, et al. Maraviroc (UK-427,857):

a potent, orally bioavailable and selective small-molecule inhibitor of the chemokine receptor CCR5 with broad-spectrum anti-human immunodeficiency virus type 1 activity. Antimicrob Agents Chemother 2005; 49:4721–4732.

2. Asmuth DM, Goodrich J, Cooper DA, et al. CD4+ T-cell restoration after 48 weeks in the maraviroc treatment-experienced trials MOTIVATE 1 and 2. J Acquir Immune Defic Syndr 2010; 54:394–397.

3. Gulick RM, Lalezari J, Goodrich J, et al. Maraviroc for previously treated patients with R5 HIV-1 infection. N Engl J Med 2008; 359:1429–1441.

4. Fätkenheuer G, Nelson M, Lazzarin A, et al. for the MOTIVATE 1 and MOTIVATE 2 Study Teams. Subgroup analyses of maraviroc in previously treated R5 HIV-1 infection. N Engl J Med 2008; 359:1442–1455.

5. Clotet B, Bellos N, Molina JM, et al. Efficacy and safety of darunavir-ritonavir at week 48 in treatment-experienced patients with HIV-1 infection in POWER 1 and 2: a pooled subgroup analysis of data from two randomised trials. Lancet 2007; 369:1169–1178.

6. Cooper DA, Steigbigel RT, Gatell JM, et al. Subgroup and resistance analyses of raltegravir for resistant HIV-1 infection. N Engl J Med 2008; 359:355–365.

7. Baxter JD, Schapiro JM, Boucher CA, et al. Genotypic changes in human immunodeficiency virus type 1 protease associated with reduced susceptibility and virologic response to the protease inhibitor tipranavir. J Virol 2006; 80:10794–10801.

8. De Luca A, Weidler J, Di Giambenedetto S, et al. Association of HIV-1 replication capacity with treatment outcomes in patients with virologic treatment failure. J Acquir Immune Defic Syndr 2007; 45:411–417.

9. Swanstrom R, Bosch RJ, Katzenstein D, et al. Weighted phenotypic susceptibility scores are predictive of the HIV-1 RNA response in protease inhibitor-experienced HIV-1-infected subjects. J Infect Dis 2004; 190:886–893.

10. Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the use of antiretroviral agents in HIV-1 infected adults and adolescents. Department of Health and Human Services. (Updated 10 January 2011. Accessed 11 March 2011.) Available from http://www.aidsinfo.nih.gov/ContentFiles/AdultandAdolescentGL.pdf

11. Naeger LK, Struble KA. Food and Drug Administration analysis of tipranavir clinical resistance in HIV-1-infected treatment-experienced patients. AIDS 2007; 21:179–185.

12. Coakley E, Chappey C, Flandre P, et al. Defining lower (L) and upper (U) phenotypic clinical cutoffs (CCO) for tipranavir (TPV), lopinavir (LPV), saquinavir (SQV) and amprenavir (APV) co-administered with ritonavir (r) within the RESIST dataset using the PhenoSense assay (Monogram [MGRM] Biosciences). Antivir Ther 2006; 11 Suppl 1:S81.

13. ANRS-AC11: Resistance Group. Genotype interpretation: nucleoside and nucleotide reverse transcriptase inhibitors. (Updated July 2008. Accessed 11 March 2011.) Available from www.hivfrenchresistance.org/2008/Algo-2008.pdf

14. HIV French Resistance: HIV-1 genotypic drug resistance interpretation’s algorithms. (Updated 2002. Accessed 11 March 2011.) Available from www.hivfrenchresistance.org/2002/tab2.html

15. DHHS/FDA/CDER guidance for industry. Antiretroviral drugs using plasma RNA measurements - clinical considerations for accelerated and traditional approval. (Updated October 2002. Accessed 11 March 2011.) Available from http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm070968.pdf

16. Hosmer DW, Lemeshow S. Applied logistic regression. New York: John Wiley and Sons, Inc. 1989; pp. 140–145.

17. Johnson VA, Brun-Vezinet F, Clotet B, et al. Update of the drug resistance mutations in HIV-1: Spring 2008. Top HIV Med 2008; 16:62–68.

18. Hardy WD, Gulick R, Mayer H, et al. Efficacy and safety of maraviroc in treatment-experienced patients infected with R5 HIV-1: 96-week combined analysis of the MOTIVATE 1 and 2 studies. 9th International Congress on Drug Therapy in HIV Infection. 9–13 November 2008, Glasgow, UK. Abstract O425.

19. Nelson M, Arasteh K, Clotet B, et al. Durable efficacy of enfuvirtide over 48 weeks in heavily treatment-experienced HIV-1-infected patients in the T-20 versus optimized background regimen only 1 and 2 clinical trials. J Acquir Immune Defic Syndr 2005; 40:404–412.

20. Hogg RS, Yip B, Chan KJ, et al. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA 2001; 286:2568–2577.

21. May MT, Sterne JA, Costagliola D, et al. HIV treatment response and prognosis in Europe and North America in the first decade of highly active antiretroviral therapy: a collaborative analysis. Lancet 2006; 368:451–458.

22. Liu L, May S, Richman DD, et al. Comparison of algorithms that interpret genotypic HIV-1 drug resistance to determine the prevalence of transmitted drug resistance. AIDS 2008; 22:835–839.

23. Saag M, Goodrich J, Fätkenheuer G, et al. A double-blind, placebo controlled trial of maraviroc in treatment-experienced patients infected with non-R5 HIV-1. J Infect Dis 2009; 199:1638–1647.

24. Anderson JA, Jiang H, Ding X, et al. Genotypic susceptibility scores and HIV type 1 RNA responses in treatment-experienced subjects with HIV type 1 infection. AIDS Res Hum Retroviruses 2008; 24:685–694.

25. Castor D, Vlahov D, Hoover DR, et al. The relationship between genotypic sensitivity score and treatment outcomes in late stage HIV disease after supervised HAART. J Med Virol 2009; 81:1323–1335.

26. Harrigan PR, Wynhoven B, Brumme ZL, et al. HIV-1 drug resistance: degree of underestimation by a cross-sectional versus a longitudinal testing approach. J Infect Dis 2005; 191:1325–1330.

27. Zaccarelli M, Lorenzini P, Checcherini-Silberstein F, et al. Historical resistance profile helps to predict salvage failure. Antivir Ther 2009; 14:285–291.

Accepted 25 September 2010; published online 21 March 2011

AVT-10-OA-1711_Schapiro.indd 404 6/5/11 09:09:10