1. Efficacy of initial combination antiretroviral therapy for HIV-1: a meta-analysis Frederick J. Lee 1, Janaki Amin 2, Andrew Carr 1 Centre for Applied Medical Research, St Vincent’s Hospital 1 Kirby Institute, University of New South Wales 2 Sydney, Australia 7 th IAS Conference on HIV Pathogenesis, Treatment & Infection July 2013, Kuala Lumpur, Malaysia

2.  Initial antiretroviral therapy (ART) −‘backbone’ (2 NRTIs) + third drug (NNRTI / rPI / INSTI)  DHHS guidelines: when to start −mostly driven by CD4 count / clinical stage −HIV viral load not a criterion since 2007  DHHS guidelines: what to start −‘Preferred’ &‘Alternative’ regimens −current ‘Preferred’ regimens: TDF-FTC/3TC + EFV / rDRV / rAZV / RAL Background Overview

3.  Guidelines based on serial assessment of individual trials  Systematic review / meta-analysis −more patients / regimens, so more power to evaluate subpopulations identify predictors of ART success  Limitations of previous ART meta-analyses −only some regimens −short follow-up durations −more recent studies: new drugs / studies longer follow-up Background Overview

4.  Primary = overall efficacy −determined by undetectable viral load −all studies over maximum follow-up period  Secondary  efficacy over time −48, 96, & 144 weeks  efficacy by 2012 DHHS guidelines −‘Preferred’ vs. ‘Alternative’ ART  efficacy by baseline viral load −≥100,000 vs. <100,000 copies/mL  predictors of efficacy and of failure Objectives Primary and secondary

5.  Inclusion criteria −treatment-naïve, HIV-1+ adults −prospective design −≥48 weeks duration −intent-to-treat (ITT) efficacy analysis  Exclusion criteria −retrospective or cross-sectional design −combinations not recommended for toxicity/poor efficacy (e.g. monotherapy) −directly observed therapy Methods Study selection

6.  Databases sourced (to Dec 31, 2012) −MEDLINE −clinical trial registries (Cochrane, NIH, ISRCT) −conference abstracts & presentations (CROI, IAS, ICAAC, Glasgow) −product labels & medical reviews (FDA, EMA) −study synopses from manufacturers  Manufacturers approached for missing data, kindly provided by: −BMS, Gilead, MSD, ViiV Healthcare Methods Data sources

7.  Database construction −study characteristics −eligibility criteria −participant & disease characteristics −ART characteristics Methods Data collection

8.  Descriptive −unit of analysis = treatment group −variables expressed as percentage mean, weighted for group size  Predictive −linear regression approach multivariable backwards, step-wise variable selection  Analyses performed using STATA (v.11) −parameters not displayed if not significant or not of particular interest Methods Statistics

9. Characteristics Groups (n=216) Participants (n=40,124) Mean, % Randomised yes 19636,53491.1 Placebo yes 6817,63043.9 Phase 2 50 4,38610.9 3 9626,32565.6 4 70 9,41323.5 Sponsorship academic 5510,68126.6 industry 12326,54766.2 both 38 2,896 7.2 Year commenced pre-1996 10 641 1.6 1997-1999 53 8,10820.2 2000-2002 43 9,33123.3 2003-2005 6113,06232.6 2006-2008 31 6,59716.4 2009-2010 17 2,346 5.8 Study characteristics

10. Characteristics Groups (n=216) Participants (n=40,124) Mean, % Eligibility restrictions viral load 18433,91584.5 ALT / AST 17331,53678.6 haemoglobin 13324,22460.4 CD4 count 12219,42848.4 genotype 6212,56331.3 AIDS (CDC C) 11 1,436 3.6 prior IDU 5 223 0.6 Duration (weeks) 48 13120,16550.3 96 6011,62929.0 144 25 8,33020.8 Efficacy analysis ITT M=F 62 7,12717.8 ITT NC=F 8413,67634.1 TLOVR 7019,32148.2 Study characteristics

11. CharacteristicMean (SD) Age37 (2) Sex men 76% (15) Race white 65% (17) black 27% (17) Risk factors MSM 52% (19) heterosexual 38% (15) IDU 10% (9) Previous AIDS12% (13) CD4 count248 (81) HIV RNA log 10 cp/mL 4.9 (0.2) ≥100,000 cp/mL 43% (11) Hepatitis C10% (9) Participant characteristics

12. CharacteristicsGroups, n Participants, n Mean, % Pills per day, mean (SD) 21640,1246.3 (3.6) Doses per day, mean (SD) 21640,1242.0 (0.7) Dosing with food, % fasting only 59 9,75424.3 fasting + food 22 5,10812.7 food only 6511,94729.8 no restriction 7013,31533.2 ART characteristics Dosing

13. Groups, n Participants, n Mean, % NRTI backbones AZT-3TC5610,83227.0 TDF-FTC4510,12325.2 d4T-3TC27 3,988 9.9 ABC-3TC26 5,51613.7 d4T-ddI20 2,349 5.9 TDF-3TC 9 1,970 4.9 Third drug classes NNRTI9419,51248.6 PI (boosted)56 9,72424.2 PI (unboosted)38 5,68614.2 NRTI12 1,771 4.4 INSTI 9 2,150 5.4 ART characteristics Key NRTI backbones & third drug classes

14. Efficacy All studies Follow-up, weeks 4896144 Groups, n 216 8525 Participants, n 40,124 19,9598,330 Follow-up, weeks (SD) 82 (38) 4896144 ART efficacy, % (SD) 60 (16) 66 (16) 60 (16) 52 (18)

15. Efficacy All studies Follow-up, weeks 4896144 Groups, n 216 8525 Participants, n 40,124 19,9598,330 Follow-up, weeks (SD) 82 (38) 4896144 ART efficacy, % (SD) 60 (16) 66 (16) 60 (16) 52 (18) ART cessation, % (SD) 25 (11) 20 (9) 28 (11) 34 (10) adverse events8.16.97.510 patient decision118.51415 virological failure3.52.44.84.4 other4.73.36.48.3

16. Efficacy: all studies from 1994 to 2010

17. Efficacy Predictors (all studies): NRTI backbone Efficacy, %, (SD) Multivariable analysis Coeff95% CIpp group TDF-FTC73 (10) Ref ABC-3TC63 (7) -7.6 -12.7, -2.6 0.003 AZT-3TC48 (15) -13.3 -18.0, -8.5 <0.001 ddI-FTC78 (-) 7.9 -11.6, 27.4 0.426 TDF-3TC69 (5) -1.2 -8.5, 6.2 0.758 ddI-3TC65 (8) -7.9 -16.6, 0.8 0.075 d4T-3TC55 (12) -11.4 -16.8, -5.9 <0.001 d4T-ddI44 (13) -18.4 -24.9, -12.0 <0.001 AZT-ddI42 (4) -35.3 -51.9, -18.7 <0.001 r 2,NRTI backbone type = 35.3%

18. Efficacy Predictors (all studies): third drug class Efficacy, %, (SD) Multivariable analysis Coeff95% CIpp group NNRTI61 (15) Ref INSTI84 (5) 11.9 4.6, 19.2 0.002 PI (boosted)67 (9) -0.9 -4.7, 3.0 0.660 NRTI51 (12) -10.7 -17.4, -4.1 0.002 PI (unboosted)42 (11) -15.0 -19.4, -10.6 <0.001 r 2, third drug class = 43.0%

19. Efficacy Predictors (all studies): study design Efficacy, % (SD) r 2 (%) Multivariable analysis Coeff95% CIpp group Study phase 2 64 (15) 3 62 (16) 4 54 (16) 3.8 Analysis ITT M=F 53 (17) Ref ITT NC=F 59 (15) -6.4 -10.1, -2.7 0.001 TLOVR 64 (15) 2.1-7.5 -11.5, -3.6 <0.001 Eligibility restriction Genotype yes 72 (10) no 55 (15) 21.7 CD4 yes 54 (18) no 66 (11) 6.2 Treatment Pills / day 19.2 Doses / day 22.3

20. Efficacy By pre-treatment viral load All studies Pre-treatment viral load <100,000≥100,000 Groups, n 21698 Participants, n 40,12413,1849,694 Follow-up, weeks (SD) 82 (38) 81 (36) ART efficacy, % (SD) 60 (16) 70 (15) 62 (15) ART cessation, % (SD) 25 (11) no data Mean difference for <100,000 vs. ≥100,000 subgroups = 8.4% (95% CI 6.0 to 10.9), p<0.001

21. Efficacy Predictors (all studies): viral load ≥100,000 Efficacy, % (SD) r 2 (%) Multivariable analysis Coeff95% CIpp group Placebo yes 67 (13) Ref no 57 (15) 7.9-6.3 -12.0, -0.5 0.034 Pills per day 24.0-1.8 -2.8, -0.8 <0.001 Third drug class NNRTI 64 (11) Ref INSTI 81 (5) 14.5 4.5, 24.4 0.005 PI (boosted) 65 (11) 6.5 0.2, 12.8 0.042 NRTI 43 (17) -20.0 -29.1, - 10.8 <0.001 PI (unboosted) 43 (11) -10.0 -19.9, -0.2 0.045

22. Efficacy By DHHS regimen Mean difference for ‘Preferred’ vs. ‘Alternative’ regimens = 10% (95% CI 7.6 to 15.4), p<0.001 DHHS recommendation ‘Preferred’‘Alternative’other Groups, n 2736153 Participants, n 5,6778,55625,891 Follow-up, weeks (SD) 99 (41) 86 (35) 77 (37) ART efficacy, % (SD) 75 (8) 65 (7) 55 (17)

23. Efficacy By DHHS regimen DHHS recommendation ‘Preferred’‘Alternative’other Groups, n 2736153 Participants, n 5,6778,55625,891 Follow-up, weeks (SD) 99 (41) 86 (35) 77 (37) ART efficacy, % (SD) 75 (8) 65 (7) 55 (17) ART cessation, % (SD) 20 (8) 25 (8) 27 (12) adverse events5.37.39.0 patient decision9.011.0 virological failure3.62.83.7 other3.44.55.0

24. Efficacy By DHHS regimen TDF-FTC/3TC plus Groups, n Participants, n Follow-up, weeks Efficacy, % (SD) EFV142,72910872 (8) rAZV 91,776 8775 (7) rDRV 1 343 9679 (-) RAL 3 829 9982 (8) Similar efficacy across DHHS-’Preferred’ regimens, although trend to superior efficacy with raltegravir

25. Cessation less likely in industry-sponsored studies Cessation less likely in industry-sponsored and phase 2 studies Cessation: all studies, 1994 to 2010 (overall cessation 25%) Adverse events (8%) Participant decision (11%)

26.  Overall mean efficacy of initial ART is low −only 60% over 82 weeks −higher with ‘Preferred’ regimens (75% over 99 weeks)  Treatment determinants of greater success −TDF-FTC (vs. ABC-3TC) −INSTI (vs. NNRTI or boosted PI) - including when pre-treatment viral load ≥100,000 cp/mL  Suboptimal efficacy −most patients will interrupt initial ART −TDF-FTC + INSTI efficacy 81% when pre- treatment viral load ≥100,000 cp/mL need for study of 3 vs. 4 drugs? Conclusions

27.  Ongoing loss in efficacy over time −participant decision > adverse events >> virological failure  Significant 8.4% difference in efficacy between higher & lower viral load groups −similar to 10% difference between ‘Preferred’ & ‘Alternative’ DHHS regimens −guidelines should recommend initiating ART before viral load reaches 100,000 cp/mL  Despite focus on co-formulation, fewer daily pills & doses not independent predictors of overall efficacy Conclusions

28.  Groups the base unit, not individuals  Only some unpublished data available, mostly from industry-sponsored studies  Efficacy by viral load not randomised  No analysis of: −individual drugs within third drug class e.g. NVP vs. EFV; rLPV vs. other PI −clinical outcomes or resistance  Multivariable method of analysis −clinically irrelevant associations may emerge, relevant associations missed −dependent upon data completeness Limitations

29.  Data: Fraser Drummond (ViiV) Silke Schweizer (BMS) Carolee Welebob (MSD) Howard Wraight (Gilead) Rebekah Puls, Kathy Petoumenos (UNSW)  Funding: NHMRC of Australia (FJL)  Potential conflicts of interest (AC) research funding - Baxter, Gilead, MSD, Pfizer consultancies - Gilead, MSD, ViiV lectures - Gilead, MSD, Serono, ViiV advisory boards - Gilead, MSD, ViiV Acknowledgements