Use of observational databases to evaluate the effectiveness of antiretroviral therapy for HIV infection: comparison of cohort studies with randomized trials. EuroSIDA, the French Hospital Database on HIV and the Swiss HIV Cohort Study Groups.

摘要: OBJECTIVES: It is important to assess the extent of bias when comparing clinical efficacy antiretroviral regimens in observational databases because, with current lack planned large trials, such analyses may represent only means assessing risk serious events associated new regimens. We aimed compare results from those randomized trials. METHODS: Three treatment comparisons trials [Delta, AIDS Clinical Trials Group (ACTG) 175, Community Programs for Research on (CPCRA) 007 and ACTC 320] were mimicked cohorts: (i) zidovudine monotherapy versus combination two nucleoside analogues; (ii) combined either didanosine or zalcitabine; (iii) a dual triple regimen including protease inhibitor. Data over 10 000 patients French Hospital Database HIV, EuroSIDA study Swiss HIV cohort analysed each comparisons. Progression disease death was Cox models, adjusting baseline differences, compared RESULTS: For comparison adjusted relative estimates cohorts between 0.61 0.84, favouring combinations monotherapy, 0.57 0.63 ranged 0.81 1.01 0.77 0.92 (iii), showed similar ACTG 320 trial but one indicated higher progression therapy [relative 1.20, 95% confidence interval (CI) 1.01-1.441, direct contrast result (relative 0.50, CI 0.33-0.76). CONCLUSION: Serious biases can be present outcomes use studies. However, not inevitable careful interpretation several studies considered together likely informative, guiding design