1. Adherence and Resistance to HIV Antiretroviral Therapy David Bangsberg, MD, MPH Massachusetts General Hospital Harvard Medical School Harvard Initiative for Global Health February, 2009

2. Will “widespread, unregulated access to antiretroviral drugs in sub-Saharan Africa, lead to the rapid emergence of drug resistant viral strains, spelling doom for the individual, curtailing future treatment options, and [leading] to transmission of resistant virus?” “Preventing antiretroviral anarchy in sub-Saharan Africa” Harries et al Lancet 2001; 358:410-4.

3. Bell-shaped Adherence and Resistance Curve Increasing probability of selecting mutation Increasing Adherence Inadequate Drug Pressure To Select Resistant Virus Drug Pressure Selects Resistant Virus Complete Viral Suppression

4. Adherence and Prospective Accumulation of Drug Resistance Mutations in The REACH Cohort >1mo HAART 6 mo HAART Genotype #1 VL>50 copies Genotype #2 VL >50 copies >3 mo pill count Outcome: # IAS-USA primary or secondary drug resistant mutations at Genotype #2 not present at Genotype #1 >7 mo HAART w/o change in regimen Bangsberg et al AIDS 2003:17:1325

5. New Drug Resistance Mutations Over 6 Months in by Adherence Quintile in Viremic Patients REACH Cohort n=57 p=0.0002 #New DRM Bangsberg et al AIDS 2003:17:1325

6. Proportion VL>50 copies/ml by Adherence Quintile REACH Cohort n=148 p=<0.0001 Proportion VL>50 Bangsberg et al AIDS 2003:17:1325

7. Resistance Risk by Adherence and Regimen Class Bangsberg et al J. Antimicrob Chem; 2002 53(5):696-9.

8. Ritonavir Boosted PIs Lead to Better Viral Suppression at Moderate Adherence Levels Viral Suppression <50 copies/ml for RTV Boosted and Unboosted PI N=46 N=67 N=83n=71 Bangsberg et al Int Conference on Adherence to HIV Treatment 2007 P=0.04

9. Resistance Risk by Adherence and Regimen Class Bangsberg et al J. Antimicrob Chem; 2002 53(5):696-9.

10. Why NNRTI Might Have A Different Adherence-Resistance Relationship NNRTI potent and exert high selective pressure NNRTI act distant to the active site – little impact on fitness NNRTI resistance seen with single dose therapy

11. NNRTI Lead to Better Viral Suppression (<400 copies/ml) than Unboosted PIs at Moderate Electronic Medication Monitor Adherence n=65 p=0.01 Bangsberg CID 2006 : 43:939-41

12. Prevalence of NNRTI Resistance by Adherence Bangsberg AIDS 2006 20:223-232 N=54

13. Resistance Risk by Adherence and Regimen Class Bangsberg et al J. Antimicrob Chem; 2002 53(5):696-9.

14. Subjects selecting for viral mutations (NN = any mutation; PI =  1 major or  3 minor) Adherence % Percent of patients selecting for mutations at by adherence level Maggiolo et al HIV Clin Trials. 2007 Sep-Oct;8(5):282-92.

15. Resistance Risk by Adherence and Regimen Class Bangsberg et al J. Antimicrob Chem; 2002 53(5):696-9.

16. Patient Plasma Replicative Capacity Purify Viral RNA AAAA RT-PCR HIV PR and RT Sequences Transfection Pool of Patient-Derived Recombinant Viruses Containing Luciferase + Luciferase + PR-RT Luciferase A-MLV env

17. Luciferase Activity (Replication) of Sensitive “Wild-Type” Virus Decreases at Higher Drug Levels 100 1,000 10,000 100,000 1,000,000 10,000,000 1101001,000 Drug concentration, nM Luciferase 0 WT Control (NL4-3)

18. Replication of Sensitive vs. Resistant Virus Drug concentration, nM 100 1,000 10,000 100,000 1,000,000 10,000,000 1101001,000 Luciferase 0 WT Control (NL4-3) Resistant (pt-derived)

19. Resistant:WT ratio 0.01 0.1 1 10 100 01101001,000 Drug concentration (nM) Resistant virus favored Resistance:WT >1 Wildtype virus favored Resistance:WT <1 Sensitive HIV is More Fit than Resistant HIV at Lower Drug Concentrations and Becomes Less Fit at Higher Drug Concentration 100 1,000 10,000 100,000 1,000,000 10,000,000 1101001,000 Luciferase 0 WT Control (NL4-3) Resistant (pt-derived pol) Low RC High RC Resistant : Wildtype Replication Ratio Comparing Resistant Subject Isolates With Sensitive Reference Strain Bangsberg et al AIDS 2006 20:223-232

20. Methods Derive average resistant/WT fitness curve Convert adherence adjusted predicted in vivo concentrations to comparable in vitro concentrations Bangsberg et al. AIDS 2006; 20:223-231

21. Level of adherence above which the resistant virus is more fit than the wild-type virus is ~ 2% for efavirenz and nevirapine and ~ 85% for nelfinavir Bangsberg et al. AIDS 2006; 20:223-231

22. Impact of initial mutations on resistance Impact of initial mutations on fitness (no drug) Resistance at low adherence? NNRTIs++++↓Yes PI+↓↓No 3TC++++↓↓Yes TNF, ZDV, ddI, ABC +↓↓No T20++++↓↓Yes Integrase++++↓↓Possibly Maraviroc, R5 inhibitors ???

23. Simple Stuff to Improve Adherence Medication Errors Pill Boxes Case management

24. Pill Identification Test Parienti et al. JAMA 2001;285:412. P<.001; OR, 3.7, 95% CI = 1.9-7.2

25. Computer-Assisted Self-Interviewing (CASI) Advantages of CASI –Privacy may improve disclosure –Visual ARV recognition –Standardizes adherence assessment –Not personnel intensive –Could be administered in waiting room or at home via the web Bangsberg et al. AIDS Care 2002:14:3-15 Purposes of CASI –Determine patient’s understanding of medication regimen –Determine patient’s adherence over 3-day period http://www.edermpda.com/hivadheredemowww.edermpda.com/hivadheredemo

26. Audio CASI Adherence Measurement 114 patient- provider pairs 18% of patients misunderstood regimen Providers missed 74% of non- adherent patients Bangsberg et al. AIDS Care 2002:14:3-15

27. Pill Box Organizers are Associated with Improved HIV Antiretroviral Adherence and Viral Suppression: A Marginal Structural Model Analysis ML Petersen, Y Wang, MJ van der Laan, D Guzman, E Riley, DR Bangsberg Clin Infect Dis. 2007 Oct 1;45(7):908-15 Objective: Estimate the effect of pill box organizer (Mediset) use for a given month on adherence to ART and viral load the same month REACH Cohort: –Recruited from San Francisco homeless shelters, free meal programs and low-income single-room-occupancy hotels –Adherence assessed using unannounced pill count Study sample: –237 individuals followed for 3170 person-months –March 1998 - September 2005 –Majority non-white men –High proportion injection drug users

28. Results: Pill box organizers improve adherence and reduce viral load MSM Estimator Difference in % Adherence 95% CIDifference in Log VL 95% CIOR VL<400 95% CI G-Comp 4.5% (2.0, 7.0) -0.34 (0.08, 0.60) 1.81 (1.25, 2.62) IPTW 4.1% (0.0, 8.3) -0.37 (0.05, 0.69) 1.91 (1.27, 2.90) Double Robust 4.1% (1.1, 7.1) -0.36 (0.09, 0.63) 1.91 (1.27, 2.90) 4% better adherence 1.9 odds better viral suppression $5.00/pill box: extremely cost-effective intervention Should be standard-of-care ML Petersen, Y Wang, MJ van der Laan, D Guzman2, E Riley, DR Bangsberg Clin Infect Dis. 2007 Oct 1;45(7):908-15

29. Routine Case Management in HIV+ Homeless and Marginally Housed Kushel and Bangsberg CID 2006 Jul 15;43(2):234-42. Outcome adjusted odds ratio 95% confidence interval Sustained Primary Care4.01.2 – 14.0 No Hospitalization2.71.2 – 5.8 No ER Visit2.31.3 – 4.1 HIV Viral RNA <4002.11.1 – 3.9 Increasing CD42.51.3 – 4.6

30. Simple Stuff Summary Ask patients to construct their regimen –Pictures are better than words Give patients a pill box organizer Send patients to a good case manager

31. Meta-analysis of HIV Adherence Interventions J Simoni et al JAIDS 2006 Dec 1;43 Suppl 1:S23-35 19 studies reported between 1998-2005 -Published since 200374% Studies conducted in the -U.S. 74% -Spain11% -France11% -Switzerland 5% Study sites -Outpatient HIV primary care clinics84%

32. Intervention Characteristics J Simoni et al JAIDS 2006 Dec 1;43 Suppl 1:S23-35 n (%) Intervention provided by: Health care provider 8 (47%) Pharmacist 2 (12%) Counselor 5 (29%) Median Range Number of intervention sessions 41 – 54 Duration of each session 1 hr 45 min–2.5 hrs Intervention duration (days) 70 1 day – 1 yr Post Intervention follow-up (k=16) 56 days 14 days-1 yr

33. Intervention Components J Simoni et al JAIDS 2006 Dec 1;43 Suppl 1:S23-35 HAART information 79% (One-way didactic provision of information about HIV, HAART, and the prescribed regimen) Cognitive strategies 79% (Two-way discussion involving patient-specific information addressing cognitions such as motivations and expectations) Behavioral strategies 84% (Such as providing external rewards or two-way discussion involving patient-specific information addressing behaviors such as plans or coping or cue-dosing) External reminders32% (Such as pagers, diaries, or calendars)

34. 95% Adherence at First Follow-up J Simoni et al JAIDS 2006 Dec 1;43 Suppl 1:S23-35 Study Intervention Control OR (95% CI) (n/N) (n/N) DiIorio 8/8 6/99.29 (3.15, 27.35) Knobel 46/60 58/1102.95 (2.32, 3.76) Margolin 23/3712/32 2.74 (1.03, 7.28) Weber 21/31 12/272.42 (0.78, 7.52) Safren-life 16/30 8/262.30 (1.42, 3.74) Remien 30/86 18/952.30 (1.82, 2.90) Rathbun 6/16 4/171.94 (1.16, 3.25) Pradier 75/64 62/701.92 (1.56, 2.36) Tuldra 37/40 35/651.76 (1.05, 2.95) Murphy 14/1711/141.27 (0.69, 2.35) Andrade 14/32 12/321.25 (0.44, 3.53) Rawlings 15/5118/571.13 (0.88, 1.46) Samet 33/53 40/650.96 (0.74, 1.24) Goujard 86/101 73/85 0.94 (0.71, 1.25) Jones 40/92 40/820.79 (0.43, 1.43) Rigsby 4/15 4/12 0.75 (0.43, 1.33) Safren-pager 1/34 1/360.62 (.02, 19.33) Rotheram 15/19 12/13 0.30 (0.14, 0.67) Overall 484/786 426/847 1.50 (1.16,1.94) 0.010.101.0010.00 100.00 OR=1.5 (1.16-1.94)

35. Study Intervention Control OR (95% CI) (n/N) (n/N) Rathbun 16/16 12/17 13.48 (4.81, 37.79) Smith 7/115/13 2.90 (1.64, 5.14) Tuldra 22/28 17/26 2.03 (1.33, 3.07) Knobel 39/60 60/110 1.55 (1.24, 1.94) Pradier 79/123 65/121 1.51 (1.27, 1.81) Goujard 49/77 37/62 1.21 (0.96, 1.54) Rawlings 53/66 43/54 1.13 (0.88, 1.46) Remien 37/86 39/95 1.09 (0.89, 1.33) Samet 19/31 24/38 0.96 (0.69, 1.34) Andrade 10/29 11/29 0.86 (0.60, 1.25) Rigsby 3/15 3/12 0.84 (0.44, 1.58) Margolin 11/25 11/20 0.64 (0.43, 0.97) Weber 27/29 23/24 0.58 (0.25, 1.35) Rotheram 4/9 2/3 0.52 (0.21, 1.29) Overall 376/605 352/642 1.25 (.99, 1.59) Undetectable VL Post-Intervention J Simoni et al JAIDS 2006 Dec 1;43 Suppl 1:S23-35 OR 1.25 (.99-1.59)

36. Correlates of Successful Interventions J Simoni et al JAIDS 2006 Dec 1;43 Suppl 1:S23-35 >7 day recall period Didactic component Interactive discussion of cognition, expectations, and motivation Conducted outside of US

37. Antiretroviral therapy in Africa Warren Stevens, Steve Kaye, Tumani Corrah BMJ 2004;328:280-282 [In sub-Saharan Africa]….the potential short term gains from reducing individual morbidity and mortality may be far outweighed by the potential for the long term spread of drug resistance…. In Africa, a higher proportion of patients are likely to fall into the category of potential poor adherers unless resource intensive adherence programmes are available.

38. Adherence to HIV Therapy in the Industrialized North San Francisco Bangsberg AIDS 2000 67% Pittsburgh Paterson Annals Int Med 2000 74% Los Angeles Liu Annals Int Med 2001 63% New York City Arnsten CID 2001 57% Hartford McNabb CID 2001 53% Philadelphia Gross AIDS 2001 79%

39. Mbarara, Uganda

40. Adherence in Patients Purchasing Generic D4T/3TC/NVP in Uganda N=36 MEMSUnannounced Pill Count Self Report 93% (SD 16%) 92% (SD 16%) 94% (SD 16%) Oyugi et al JAIDS 2004 36:1100-1102

41. Meta-Analysis of Barriers to Adherence in Africa and North America Mills and Bangsberg JAMA 2006:296:679-690 Systematic review of adherence –28,689 patients in 228 studies North America Brazil, Uganda, Cote d’Ivoire, South Africa, Malawi, Bostwana, Costa Rica, Romania Resource-Rich Country Summary 54.7% (95CI: 48.0-61.3%) Resource-Poor Country Summary 77.1% (95CI:67.3%-85.6%)

42. UARTO Adherence Over 12 Months on Free ARV Therapy n=274 Bangsberg et al CROI 2008

43. A Social Model of Adherence for sub-Saharan Africa Ware and Bangsberg PLoS Medicine 2009 Improving Health

44. Resource Scarcity Resource Scarcity Improving Health A Social Model of Adherence for sub-Saharan Africa Ware and Bangsberg PLoS Medicine 2009

45. Resource Scarcity Resource Scarcity Social Capital Improving Health A Social Model of Adherence for sub-Saharan Africa Ware and Bangsberg PLoS Medicine 2009

46. Resource Scarcity Resource Scarcity Social Capital Improving Health A Social Model of Adherence for sub-Saharan Africa Ware and Bangsberg PLoS Medicine 2009

47. Resource Scarcity Resource Scarcity Social Capital Improving Health A Social Model of Adherence for sub-Saharan Africa Ware and Bangsberg PLoS Medicine 2009

48. Social Structural: Patterns of Inequality, e.g., stigma, gender inequality Social Capital Infrastructural: Few treatment sites Distance to care Cost/Availability of Transportation Cultural: Religious Beliefs Respect for Authority Importance of having children Individual: HIV knowledge Med side effects Cognitive function Mental health Alcohol Use Resource Scarcity Resource Scarcity Improving Health A Social Model of Adherence for sub-Saharan Africa Ware and Bangsberg PLoS Medicine 2009

49. D4T/3TC/Nevirapine 17 USD per month Triomune

50. Stopping drugs with different half lives 0244836 12 Time (hours) Drug concentration Zone of potential replication IC 90 IC 50 Last Dose Day 1 Day 2 MONOTHERAPY S. Taylor et al. 11th CROI Abs 131

51. NNRTI Resistance and Treatment Discontinuation Parienti et al CID 2004:38:1311-6 No. patients at Risk ≤1 drug holiday52473830194 >= 2 drug holidays1917131061

52. Frequency and Duration of Treatment Interruptions >48hrs over 24 weeks on Self-pay ART Oyugi and Bangsberg AIDS 2007

53. Correlates: Financial difficulty securing ARVs and pharmacy stockouts

54. Frequency and Duration of Treatment Interruptions >48hrs over 24 weeks on Self-pay ART Oyugi and Bangsberg AIDS 2007 Correlates: Financial difficulty securing ARVs and pharmacy stockouts 90% of all missed doses occur during an interruption

55. MEMS-Defined 48 Hour Treatment Interruptions Predict Resistance to Self-pay ART in Uganda Oyugi and Bangsberg AIDS 2007 Resistant Interruption >48 hours Yes 8/32 (63%) No 0/56 (0%) P=0.04

56. Duration of MEMS Defined Treatment Interruption and Probability of NNRTI Resistance Parienti and Bangsberg PLoS One 2008 n=72 + Controls O Cases Estimated 95% confidence interval Longer interval of treatment discontinuation in days Estimated probability of viral control

57. Africans “don’t know what Western time is,”and “do not know what you are talking about,” when asked to take drugs at specific times. Andrew Natsios USAID Administrator

58. How to Take ARVs on Time in Rural Uganda Without a Watch: John’s Adherence Story Maier, Mwebesa, Emenyonu, Pepper, Bangsberg PLOS 2006 No education Works as a farmer. Lives with his brother, sister-in-law, and three nieces in a three room mud-walled house without electricity. Owns a lantern, bed, sofa, bike, and a radio, but no watch. HIV in April 2005 and started generic D4T/3TC/NVP (Triomune) after disseminated herpes zoster and Kaposi’s sarcoma CD4 count of 151

59. Electronic medication monitor record of time of bottle openings for am and pm doses.

60. Adherence 90% of doses within 10 minutes of 7:20 90% of doses within 17 minutes of 7:20 pm Overall adherence 98.9%

61. John’s Adherence: 0-9 and 10-18 months Initial MEMS assessment (August 2005 to April 2006 (9 months)) Subsequent MEMS assessment (May 2006 to January 2007 (9 months))

62. Summary Most resistance has occurred in highly adherent patients on partially suppressive regimens Potent regimens reduce resistance at all levels of adherence NNRTI resistance: low adherence and treatment discontinuation Internationally: stable drug supply and distribution

63. Summary Most resistance has occurred in highly adherent patients on partially suppressive regimens Potent regimens reduce resistance at all levels of adherence NNRTI resistance: low adherence and treatment discontinuation Internationally: stable drug supply and distribution

64. Summary Most resistance has occurred in highly adherent patients on partially suppressive regimens Potent regimens reduce resistance at all levels of adherence NNRTI resistance: low adherence and treatment discontinuation Internationally: stable drug supply and distribution

65. Summary Most resistance has occurred in highly adherent patients on partially suppressive regimens Potent regimens reduce resistance at all levels of adherence NNRTI resistance: low adherence and treatment discontinuation Internationally: stable drug supply and distribution

66. Andrew Moss, PhDUCSF Epi/Biostat Ed Acosta Huyen Cao, MD Univ of Alabama Ca Sate Health Department Tom Coates, PhD Edwin Charlebois, MPH, PhD UCLA UCSF EPI Center Barry Bredt, PhDUCSF Center for AIDS Prevention Richard Clark, MPHUCSF Epi/Biostat Steven Deeks, MDUCSF Positive Health Program Nneka Emenyonu Robert Grant, MD, MPH UCSF Epi Center UCSF Gladstone Institute Norma Ware, PhD Gwen Hammer, PhD Rick Hecht, MD Harvard Medical School UCSF EPI Center UCSF Positive Health Program Mark Holodniy, MDPalo Alto VA Jeff Martin, MD Neil Parkin, PhD Jennifer Free Travis Porco, PhD UCSF Epidemiology Monogram Bioscience UCSF Epi Center SF Department of Public Health Irene Andia, MMedMbarara University Elise Riley, PhDUCSF EPI Center Neil Parkin, PhDVirologic Richard Harrigan, PhDUniversity of British Columbia Andrew Zolopa, MDStanford Positive Care Program Funding: NIMH, NIAAA, The Doris Duke Charitable Foundation, Bill and Melinda Gates Foundation, University-Wide AIDS Research Program, UCSF Center for AIDS Research