1. Increased risk of suicidal behaviour with use of efavirenz: Results from the START Trial Alejandro Arenas-Pinto, Birgit Grund, Shweta Sharma, Esteban Martinez, Nathan Cummins, Julie Fox, Karin Klingman, Dalibor Sedlacek, Simon Collins, Patricia Flynn, William Chasanov, Eynat Kedem, Christine Katlama, Juan Sierra-Madero, Vidar Ormaasen, Pim Brouwers, David Cooper for the INSIGHT START Study Group Abstract: THAB0202

2. START Study Design HIV-infected individuals who are ART-naïve with CD4+ count > 500 cells/mm 3 Immediate ART Group Initiate ART immediately following randomization N=2,326 Deferred ART Group Defer ART until CD4+ count drops to <350 cells/mm 3 or AIDS develops N=2,359 We use data accrued through May 27, 2015, when the START study results were unblinded Primary composite endpoint Serious AIDS or death from AIDS Serious Non-AIDS* Events and death not attributable to AIDS * CVD, ESRD, decompensated liver disease, & non-AIDS defining cancers Combination ART pre-specified before randomization

3. 3 START Results* a Serious AIDS-defining events, Serious non-AIDS events and death of any cause b DAIDS grading system – only symptomatic events c Medical Dictionary for Regulatory Activities (MedDRA®) Version 18.0, High Level Group Term (HLGT) Endpoint Immediate ARTDeferred ART Hazard Ratio P-value NRate/100PYN Primary endpoint a 420.60961.38 0.43 (0.30-0.62) <0.001 Grade 4 event b, hospitalization or death 2834.363114.78 0.91 (0.77-1.07) 0.25 Suicidal/self harming behaviour c 270.39240.34 1.15 (0.66-1.99) 0.63 *Lundgren et al. N Engl J Med 2015; 373 (9): 795-807

4. 4 Composition of “Suicidal Behaviour” a Medical Dictionary for Regulatory Activities (MedDRA), Version 18.0; HLGT under Psychiatric SOC; by Preferred Terms. Suicidal Behavior a Immediate ARTDeferred ART Suicidal ideation 106 Suicidal attempt 1713 Completed suicide 03 Self-injurious ideation 01 Intentional self-injury 01 Any of the above 2724

5. Methods Aims –Assess effects of EFV on “suicidal behaviour” by comparing the immediate versus deferred ART groups. What is the correct control group for those using EFV in the immediate ART group? –Identify factors associated with high risk of suicidal behaviour during EFV use “Suicidal and self-injurious behaviour”Outcome: “Suicidal and self-injurious behaviour” –Collected as part of “serious events” in START, irrespective of ART use –MedDRA coded blinded to treatment arm –HLGT* under the MedDRA Psychiatric System Organ Class 5 *HLGT: High Level Group Term

6. Methods: Statistical Analysis Used Cox proportional hazards models to: Compare the immediate versus deferred ART groups*: 1.Intent-to-treat among all START participants 2.Intent-to-treat within subgroups: those with or without EFV use pre-specified. 3.Among those with pre-specified EFV, we censored follow-up at ART start in the deferred group. Estimate associations of factors with suicidal behaviour 6 *Models were stratified by history of psychiatric diagnoses (because of higher rates of suicidal behaviour with history)

7. 7 Baseline Characteristics by Pre-specified ART EFV Pre-specified (N=3516) Other ART Pre-specified (N=1169) Age (years), Med [IQR] 36 [29, 43]36 [29, 45] Gender (female), N (%) 995 (28.3)262 (22.4) Geographic location, N (%) High-income regions a 1206 (34.3)949 (81.2) Low-Mid income regions b 2310 (65.7)220 (18.8) CD4 cell count (cell/mm 3 ), Med [IQR] 651 [583, 768]652 [585, 755] HIV-RNA (copies/mL), Med [IQR] 12217 [2873, 41549]14304 [3750, 47703] Pre-specified ART contains emtricitabine+tenofovir 3312 (88.5)1052 (90.0) Prior psychiatric diagnosis c, N (%) 108 (3.1)162 (13.9) Psychiatric drug treatment d, N (%) 140 (4.0)176 (15.1) Use of recreational drugs, N (%) Ever use 858 (24.4)442 (37.8) Current use ≥ 1 day a week 78 (2.2)47 (4.0) Current use < 1 day a week 212 (6.0)126 (10.8) Heavy alcohol use, N (%) 155 (4.4)45 (3.8) a US, Europe, Israel, Australia b Latin America, Africa, Asia c Including major depression, bipolar disorder and psychotic disorders d Antidepressants, antipsychotic drugs, drugs for bipolar disorder, methadone, other opiates

8. Event Rates by Pre-specified ART and Prior Psychiatric Diagnosis 8

9. 9 Overall Rate: 0.28 per 100 PY Overall Rate: 0.63 per 100 PY

10. ART Use: Percent of Participants, Percent of Follow-up Time 10 EFV Other ART EFVOther Pct. of Time Spent on ART Pre-specified ART

11. a Estimated in Cox proportional hazards models, stratified by psychiatric diagnosis. b Interaction between indicators for treatment group and pre-specified regimen. c Of these events, 6 and 0, in the immediate vs deferred arms, respectively, occurred among 108 participants with prior psychiatric diagnoses. d Of these events, 5 and 2, in the immediate vs deferred arms respectively, occurred among 162 participants with prior psychiatric diagnoses. Of the 1169 participants without EFV in the pre-specified regimen, 32 were excluded (in the immediate group, 7 never started ART, and for 25, the first ART regimen contained EFV). Follow-up in the immediate group was censored at EFV start. Suicidal/self harming events by randomisation arm N Immediate ARTDeferred ART HR a 95% CIPInt. P b EventsRateEventsRate Intention to treat (ITT) analysis, mean follow-up 3 years EFV pre-specified 3516180.34110.211.42(0.7, 3.0)0.37 0.23 Other ART pre-specified 116990.53130.720.74(0.3, 1.8)0.50 ITT analysis, year 1 only EFV pre-specified 351690.5220.113.75(0.8, 17.5)0.09 0.15 Other ART pre-specified 116971.2571.191.02(0.4, 2.9)0.96 Censoring deferred arm participants at ART initiation EFV pre-specified c 3516170.3530.084.16(1.2, 14.4)0.02 0.05 Other ART pre-specified d 113790.5980.691.04(0.4, 2.7)0.93

12. Time to suicidal/self harming behaviour Immediate arm excluding pts who never started ART or had event prior to ART initiation Deferred arm follow-up censored at ART initiation Mean follow-upPY Imm.3.0 yrs4834 Def.2.2 yrs3864 Mean follow-upPY Imm.2.8 yrs1527 Def.1.9 yrs1167

13. Predictors of suicidal/self harming behaviour Immediate arm, pts who were pre-specified EFV and started ART Multiple Regression a HR95% CIP-value Age (per 10 yrs older)0.52(0.30, 0.91)0.02 Gender & risk group Female vs male0.69(0.19, 2.53)0.58 MSM vs other male0.48(0.15, 1.57)0.23 Prior psychiatric diagnosis b 12.8(4.7, 34.9)<0.001 Recreational drug use, ever2.9(1.0, 7.9)0.04 Heavy alcohol use6.1(1.9, 19.6)0.003 EFV started (time-updated) c 1.1(0.1, 8.2)0.96 a Estimated from a Cox proportional hazards model with all listed variables b Major depression, bipolar disorder, psychotic disorder incl. schizophrenia c Indicator variable, switches to 1 at the date of EFV start (if ever)

14. Limitations Small number of events Only 3 completed suicides The comparison of the immediate versus deferred ART groups after censoring follow-up at ART initiation is not protected by randomization –Last 2 rows in table and Kaplan-Meier curves 14

15. Conclusions In the START trial, participants using EFV in the immediate ART group had an increased risk of suicidal behaviour compared to their ART-naïve controls. The impact of EFV exposure was particularly high in those with a prior psychiatric diagnosis. This supports the recommendation to screen for pre-existing depression (and other conditions) before EFV initiation. 15

16. Acknowledgments 16 START participants Members of the START study group INSIGHT International Coordinating Center staff INSIGHT Minnesota Coordinating Center staff University of Minnesota (Study Sponsor) Funding: The START study is primarily funded by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (Award Numbers UM1-AI068641 and UM1AI120197) Other NIH institutes International funding support from Europe, France, Germany, Denmark, U.K., and Australia ART provided by AbbVie, Bristol-Meyer Squibb, Gilead Sciences, GSK/ViiV Healthcare, Janssen Scientific Affairs, Merck & Co.