Background

1. 46th Interscience Conference on Antimicrobial Agents and Chemotherapy, September 27th –30th, 2006; San Francisco, CA (USA). Presentation H-1059 When is the Optimal Moment to Start HAART in HIV Infected Patients from PISCIS Cohort Study (Spain)? Jaén A, Esteve A, Montoliu A, Miro JM*, Tural C, Podzamczer D, Riera M, Segura F, Force L, Vilaró J, García I, Masabeu A, Altés J, Sued O,Clotet B, Ferrer E, Casabona J, and PISCIS study group. Catalonia and Balearic Islands (Spain)

2. Background • After the introduction of HAART,the mortality and morbidity of HIV infection decreased dramatically—HIV infection now a chronic disease. • However, HAART has some disadvantages (adverse effects, resistance). • The decision when to start HAART in asymptomatic individuals with >200 and <350 CD4 remains unclear. • In the context of observational cohort studies (= no randomization), a major limitation to resolving this issue is the need to account for leadtimeand the unseen events.

3. Aims • To evaluate progression to AIDS or death according to CD4+ count and plasma HIV-RNA levels prior to initiating HAART among patients in the PISCIS Cohort. • To estimate the optimal time at which to initiate HAART before progression to AIDS using methods accounting for leadtimeand the unseen eventsin the context of data from observational cohort studies

4. Methods • PISCIS Cohort is an ongoing multicenter observational study of HIV infected individuals (≥ 16 years old)from • Nine hospitals within Catalonia, and • One from the Balearic Islands, (Spain). • Collects clinical and epidemiological data through computer applications during routine clinical follow-up. • PISCIS Cohortrepresents 70% of the new HIV infections notified in these two regions. • From January 1998 to December 2004: • 6,922 patients were included, • 17,766 person-years of follow-up. Catalonia Balearic Islands

5. Inclusion criteria For the present study we included: • HIV-infectedantiretroviral naïve patients • Who initiated HAART from January 1998 to June 2004, • AIDS-free before antiretroviral treatment was started, • Who were on HAART for at least one month. 2,035 individuals met the inclusion criteria

6. Methods • HAART was defined as the combination of 3 antiretroviral drugs, according to Spanish and International Guidelines. • Treatment regimens: • Boosted or unboosted protease inhibitors (PI) plus 2 nucleoside reverse transcriptase inhibitors (NRTI); • 1 non-nucleoside reverse transcriptase inhibitor (NNRTI) plus 2 NRTI; • 3 NRTI; • other combinations • According to the year of initiation, HAART was categorized into: • “Early HAART” period: Jan1998 - Dec2000, • “Late HAART” period: Jan2001-Jun2004. • Analysis by intention to treat.

7. Statistical analysis:HIV disease progression or death • Descriptive analysis: • median values, interquartile ranges (IQR), and percentages. • comparisons among CD4 groups using the Kruskal-Wallis and Pearson’s chi-square tests. • HIV progression to AIDS/death: • analyzed using Cox regression models. • multivariate model was adjusted by calendar year of initiation of HAART (time-dependent variable).

8. PISCIS Cohort Study 1998-2004 Epidemiological characteristics at initiation of HAART.

9. PISCIS Cohort Study 1998-2004Clinical characteristics at initiation of HAART.

10. PISCIS Cohort Study 1998-2004Outcome

11. PISCIS Cohort Study 1998-2004 Risk of HIV progression to AIDS/death.Multivariate Cox model* * Adjusted by sex, age and calendar period of initiation of HAART.

12. PISCIS Cohort Study 1998-2004 Survival curves and hazard ratios from initiation of HAART to AIDS/death by CD4/ plasma HIV-RNA VL group HR: 1.24; 95%CI: 0.49-3.14 HR: 1.59; 95% CI: 0.76-3.30 HR: 1.93; 95% CI: 0.83-4.47 HR: 3.39; 95%CI: 1.72-6.67 HR: 5.88; 95% CI: 3.17-10.91 Number of patients at risk Patients at risk n=656 n=398 n=176 n=2,035 n=1,887 n=1,263 n=946

13. 2. Statistical analysis: Accounting for leadtime In absence of randomization, comparisons among patients who initiate HAART at 200-350 CD4 with those who chose to defer HAART at <200 CD4 do not answer the question of when to initiate HAART. • Appropriate statistical analysis should take into account the leadtimeand unseen events • Leadtime:defined as the time it took the deferred group with an early disease stage (e.g. CD4>350) to reach the latter stage (e.g. CD4 200-350)*. • Unseen events:previous events in patients who initiate HAART at late stage (fast progressors). * Cole, Rui, Anastod, Detels, Young, Chmiel, Muñoz. Statist Med 2004; 23:3351-3363.

14. Ideal randomized clinical trial (RCT) RCT: “initiate HAART” versus “defer HAART”S: observed survival time from randomization to AIDS Initiate HAART S AIDS “Initiate HAART “ at 200-350 CD4 S AIDS Initiate HAART S AIDS “Defer HAART “ to <200 CD4 S AIDS 350-200 CD4 <200 CD4 Common origin for the time-to-event analysis

15. Observational cohort study T: observed survival time from initiation of HAART to AIDS Initiate HAART T AIDS Initiate HAART at 200-350 CD4 T AIDS Initiate HAART T AIDS Initiate HAART at <200 CD4 350-200 CD4 <200 CD4 Different origin for the time-to-event analysis

16. Observational cohort study T: observed survival time from initiation of HAART to AIDSL: unobserved survival time to reach the latter stage (prior leadtime) Initiate HAART T AIDS Initiate HAART at 200-350 CD4 T AIDS Initiate HAART L T AIDS Initiate HAART at <200 CD4 350-200 CD4 <200 CD4

17. Observational cohort study T: observed survival time from initiation of HAART to AIDSL: unobserved survival time to reach the latter stage (prior leadtime) fast progressors: unseen individuals who develop AIDS before initiating HAART Initiate HAART T AIDS Initiate HAART at 200-350 CD4 T AIDS Initiate HAART L T AIDS Initiate HAART at <200 CD4 L AIDS “fast progressors” 350-200 CD4 <200 CD4

18. 2. Statistical analysis: Accounting for leadtime • Accounting for leadtime was analyzed using statistical methods described in Cole et al. 2004. Briefly : • Leadtime : • obtained from a log-normal distribution estimated with data from the pre-HAART period. • assigned to each individual in the deferred group multiple times. • added to the observed survival time to AIDS. • The number of fast progressors(unseen events) and their leadtimes were estimated. • Finally, leadtime-adjusted hazard ratios were computed through Cox regression models, using multiple imputation techniques. Cole, Rui, Anastod, Detels, Young, Chmiel, Muñoz. Statist Med 2004; 23:3351-3363.

19. Probability of the time to the CD4 transition and time to AIDS in the pre-HAART era (MACS cohort study)*. 200-350 CD4 350-500 CD4 • 11.6% fast progressors: • Median time: • to 200 CD4: 0.96 years • to AIDS: 4.5 years • 3.0% fast progressors: • Median time: • to 350 CD4: 0.88 years • to AIDS: 1.1 years * Cole, Rui, Anastod, Detels, Young, Chmiel, Muñoz. Statist Med 2004; 23:3351-3363.

20. PISCIS Cohort Study 1998-2004Survival curves for the time to AIDS: comparing initiation of HAART at 200-350 CD4 to initiation of HAART at <200 CD4

21. PISCIS Cohort Study 1998-2004Survival curves for the time to AIDS: comparing initiation of HAART at 200-350 CD4 to initiation of HAART at <200 CD4

22. PISCIS Cohort Study 1998-2004Survival curves for the time to AIDS: comparing initiation of HAART at >350 CD4 to initiation of HAART at 200-350 CD4

23. PISCIS Cohort Study 1998-2004Survival curves for the time to AIDS: comparing initiation of HAART at >350 CD4 to initiation of HAART at 200-350 CD4

24. Limitations • Intention-to-treat analysis: changes of treatment and poor adherence were not taken into account. • Using probability models of CD4 transition and fast progression (=natural history of HIV) estimated from external data (MACS cohort study). • Advisable to examine probability models according to combinations of biomarkers (e.g. CD4/plasma HIV-RNA VL) and other clinical endpoints (e.g.,death). • Alternative analysis: models allowing causal inferences from observational studies. • a large sample size is needed (pooling cohorts).

25. Conclusions • Major determinants of HIV progression or death: • having CD4 count <200, • HIV-1 RNA >=100,000 copies/ml, • HCV coinfection, • and having initiated HAART before January 2001. • Important findings were found after performing methodology accounting by leadtime: • Statistically significant risk of progression to AIDSinpatients with 200-350 CD4. • These results provide valuable information for clinical decision-making.

26. PISCIS study group Steering Committee: J. Casabona, A. Esteve, A. Jaén, M. Granell, from Center for Epidemiological Studies on HIV/AIDS of Catalonia (CEESCAT), Badalona, Spain; JM. Gatell, JM. Miró from Clínic Hospital, Barcelona, Spain; C. Villalonga and S. Riera from Son Dureta Hospital, Palma de Mallorca, Spain; D. Podzamcer and E. Ferrer from Bellvitge Hospital, Hospitalet de Llobregat, Spain; B. Clotet and C.Tural from Germans Trias i Pujol University Hospital, Badalona, Spain; F. Segura and G. Navarro from Parc Taulí Hospital, Sabadell, Spain; L. Force from Mataró Hospital, Mataró, Spain; J. Vilaró from General Vic Hospital, Vic, Spain; A. Masabeu from Palamós Hospital, Palamós, Spain; I. García from Creu Roja Hospital, Hospitalet de Llobregat, Spain; E. Dorca from Manresa Hospital, Manresa, Spain; and J. Altés from Alt Penedès Hospital, Vilafranca, Spain. Data manager and computerized support: E. Puchol, Y. Alvaro, A. Montoliu (Centre d'Estudis Epidemiològics sobre la sida de Catalunya); Modesto Sanchez (Hospital Clínic- Idibaps, Universitat de Barcelona). Clinical Staff: JL Blanco, F. Garcia-Alcaide, E. Martinez, J. Mallolas (Hospital Clínic- Idibaps, Universitat de Barcelona); G. Sirera, J. Romeu, A. Bonjoch, A. Jou. A. Ballesteros, E. Negredo, D. Fuster, J.C Martinez (Hospital Universitari Germans Trias i Pujol, Universitat Autónoma de Barcelona); M. Santin, MJ Barbera, M. Olmo, P. Robres, F. Bolao, J Carratala, C. Cabellos; C. Peña.  (Hospital de Bellvitge de Barcelona); P. Barrufet (Hospital de Mataró); M. Guadarrama (Hospital Alt Penedès de Vilafranca). Acknowledgements: The PISCIS Cohortwas funded by 3084/99 and 36354/02 projects from Fundación para la Investigación y la Prevención del Sida en España (FIPSE).