1. Comparison of PI vs PI  ATV vs ATV/r BMS 089  LPV/r mono vs LPV/r + ZDV/3TCMONARK  LPV/r QD vs BIDM02-418 M05-730 A5073  LPV/r + 3TC vs LPV/r + 2 NRTIGARDEL  ATV/r vs FPV/rALERT  ATV/r vs DRV/rATADAR  FPV/r vs LPV/rKLEAN  SQV/r vs LPV/rGEMINI  ATV/r vs LPV/rCASTLE  DRV/r vs LPV/rARTEMIS

2. Molina JM. Lancet 2008;372:646-55 CASTLE  Design  Objective –Non inferiority of ATV/r vs LPV/r at W48: % HIV RNA < 50 c/mL by intention to treat, confirmed virologic response, non completer equals failure (CVR, NC = F) (lower margin of the 2-sided 95% CI for the difference = - 10%, 90% power) ATV/r 300/100 mg QD TDF/FTC fdc QD LPV/r 400/100 mg BID TDF/FTC fdc QD Randomisation* 1 : 1 Open-label > 18 years ARV-naïve or < 1 week of prior ARV exposure HIV RNA > 5,000 c/mL Any CD4 cell count *Randomisation was stratified by HIV RNA ( 100,000 c/mL) at screening and geographic region CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC N = 443 N = 440 W48W96

3. CASTLE CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC ATV/r N = 440 LPV/r N = 443 Median age, years3436 Female31% HIV RNA (log 10 c/mL), median5.014.96 HIV RNA > 100,000 c/mL51%47% CD4 cell count (/mm 3 ), median205204 CD4 < 50 per mm 3 13%11% Hepatitis B / hepatitis C coinfection5% / 9%5% / 7% Discontinuation by W489%13% For lack of efficacyN = 5N = 8 For adverse eventN = 10N = 14 Discontinuation on or after W483% For lack of efficacyN = 7N = 1 For adverse eventN = 1 Baseline characteristics and patient disposition LPV/r was administered as soft-gel capsules during the first 48 weeks Molina JM. Lancet 2008;372:646-55

4. Mean CD4/mm 3 increase at W48 (observed values): 203 (ATV/r) vs 219 (LPV/r) BaselineATV/rLPV/r RNA < 5 log 10 c/mL RNA > 5 log 10 c/mL 82% 74% 81% 72% HIV RNA < 50 c/mL at W48 (CVR, NC = F) by baseline HIV RNA Response to treatment at week 48 Post hoc analysis: lower virologic response rate associated with lower CD4 counts for LPV/r (p = 0.0085) but not for ATV/r (p = 0.51) CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC Molina JM. Lancet 2008;372:646-55 HIV RNA < 50 c/mL (ITT) 25 50 100 75 78 76 % 95% CI for the difference = - 3.8; 7.1 78 76 Primary analysis 95% CI for the difference = - 3.6; 7.4 ATV/r (N = 440) LPV/r (N = 443) CVR, NC = FTLOVR CASTLE CVR, NC = F : confirmed virologic response, non completer equals failure

5. CASTLE CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC  Virologic failure –Definition: failure to achieve confirmed HIV RNA 400 c/mL after achieving confirmed HIV RNA < 400 c/mL without re-suppression, or discontinuation due to insufficient HIV RNA response before W48 Emergence of resistance in virologic failure ATV/r N = 440 LPV/r N = 443 Confirmed virologic response-defined virologic failure25 (6%)26 (6%) Failure to achieve confirmed HIV RNA < 400 c/mL113 Rebound after confirmed HIV RNA < 400 c/mL1017 Discontinuation due to insufficient HIV RNA response46 Assessed for emergence of resistance mutations1920 Any PI-resistance mutation Polymorphic mutations Major mutations 10 8 2 * 880880 M184I/V / TDF-resistance mutation / TAM5 / 1 / 14 / 0 / 1 * 1 patient with M46I and N88S; 1 patient with L10F, V32I, K43T, M46I, A71I, G73S, L90M Molina JM. Lancet 2008;372:646-55

6. CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC  Safety at W48: ATV/r vs LPV/r –Serious adverse events occurred in similar proportions: 12% vs 10% –Grade 2 to 4 treatment-related nausea and diarrhoea were less frequent with ATV/r: 4% vs 8% and 2% vs 11%, respectively Initiation of anti-diarrhoeal medication: 9% vs 22% –Grade 2 to 4 jaundice: 4% of ATV/r patients vs none of LPV/r patients –Grade 3/4 hyperbilirubinaemia: 34% vs < 1% –Grade 3/4 elevations in triglycerides and total cholesterol were significantly less frequent with ATV/r: < 1% vs 4% and 7% vs 18%, respectively –Median change in calculated creatinine clearance (Cockroft) was similar in both groups = - 1% –Mean increases in total cholesterol, non-HDL cholesterol and triglycerides were significantly lower on ATV/r (p < 0.0001 for the 3) –Use of lipid-lowering agents through W48: 2% vs 8% CASTLE Molina JM. Lancet 2008;372:646-55

7. CASTLE CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC  Summary –ATV/r QD was non inferior to LPV/r BID, when co-administered with TDF/FTC –Similar virologic reponse of the 2 PI/r in patients with high HIV RNA at enrolment –Results suggest reduced virologic response to LPV/r in patients with baseline CD4 < 50/mm 3 mainly because of intolerance to LPV/r in this highly immunosuppressed subgroup –Development of major PI-associated resistance mutations occurred in 2 ATV/r patients and no LPV/r patients –Incidence of diarrhoea and nausea was lower with ATV/r than with LPV/r –Incidence of hyper bilirubinemia with ATV/r was high, but less than 1% of patients discontinued due to jaundice –Lipid elevations were less pronounced with ATV/r Molina JM. Lancet 2008;372:646-55

8. CASTLE CASTLE Study: ATV/r QD vs LPV/r BID, in combination with TDF/FTC  Conclusion –ATV/r QD demonstrated similar antiviral efficacy to LPV/r BID, when coadministered with TDF/FTC (1) With less gastrointestinal toxicity But with a higher rate of hyperbilirubinemia –At W96 (2), HIV RNA < 50 c/mL was obtained in 74% of ATV/r patients vs 68% of LPV/r patients (p < 0.05 in the intent-to-treat analysis) confirming non-inferiority of ATV/r to LPV/r –Safety analysis at W96 confirmed W48 results Treatment-related gastrointestinal adverse events were more frequent with LPV/r Hyperbilirubinemia and/or jaundice was the most frequent ATV/r-related adverse event Lipid elevations were significantly higher with LPV/r –These results support current recommendation of ATV/r + TDF + FTC QD as a preferred first-line regimen for the treatment of HIV-infected patients (1) Molina JM. Lancet 2008;372:646-55 ; (2) Molina JM. JAIDS 2010;53:323-32