This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Survival Analysis for Comparative Clinical Trials We get a brief overview here. STA/ BST 222 covers Survival Analysis in much more detail. The focus here is on its use in clinical trials. From N Engl J Med in 2006, the first few issues show the importance of survival analysis in the clinical trials literature: January 5: Randomized trial of two chemotherapy regimens in stage III ovarian cancer. Primary outcome: progressionfree, overall survival. Statistical summary included KaplanMeier curves and proportional hazards models, adjusted for residual disease status and secondlook surgery option. January 12: efficacy and safety of benazepril for advanced chronic renal disease. Primary outcome a composite of disease progression endpoints or death. KaplanMeier, logrank test, Cox regression models. Example of a randomized clinical trial with survival data with censoring: Henry Kaplan, Hodgkins disease trial, late 1950s. The first trial of extended radiation therapy. There were 12 patients who got conventional therapy and 17 who got extended radiation. The outcome is reported as survival time in days, and a + denotes that the person was still alive at the end of the study, i.e. censored data. Conventional therapy: 150, 180+, 482+, 496, 503, 538, 558+, 716, 806, 1140, 2055+, 2141+. Extended radiation: 202+, 370+, 383+, 418+, 425+, 447+, 542+, 583+, 593+, 625+, 655+, 802+, 857+, 1308+, 1357+, 2001+. Common summary statistics in clinical trials One common summary is 5year survival. For conventional, 2/12 still alive, 7 died before 5 ears, and 3 were censored before 5 years. Should this count as 2/12, 2/9, 5/12? KaplanMeier curves: calculated like life tables. At each DEATH, you stop and recalculate and update the cumulative fraction surviving. For the controls, we have: 1. At times before 150 days, all of 12 survive, 100%. 2. At 150 days, 1/12 died, so 11/12 survive, 0.92 fraction. 3. At 180 days, loss to followup so now 10 in cohort but we do not change survival fraction. At 482 another person is lost, now 9 left. 4. At 496 one of the 9 died so 8/9 survive. We calculate survival as (8/9) (11/12) and get 0.81....
View Full
Document
 Winter '06
 Tsodikov

Click to edit the document details