Skip to main content

Advertisement

Table 2 Results of Polyp Prevention Trial

From: A simple method for analyzing data from a randomized trial with a missing binary outcome

stratum s   adenoma difference in observed weight bias factor ε(max)s
stratum s   recurrence rates of recurrence d s w s  
sex age group no yes missing    
   control 573 374 94 (9%)    
   study 578 380 76 (7%)    
men 30–49 control 33 22 5 (8%) -.23 .07 .09
   study 58 12 3 (4%)    
  40–59 control 99 76 7 (4%) .01 .17 .05
   study 94 76 9 (5%)    
  60–69 control 122 105 25 (10%) -.04 .23 .11
   study 144 105 18 (7%)    
  70–79 control 65 76 26 (16%) -.04 .13 .20
   study 70 71 29 (17%)    
women 30–49 control 54 11 3 (4%) .03 .10 .07
   study 47 12 4 (6%)    
  40–59 control 69 24 4 (4%) .02 .11 .04
   study 69 27 4 (4%)    
  60–69 control 77 31 13(11%) .08 .12 .11
   study 68 40 5 (4%)    
  70–79 control 54 29 11(12%) .22 .07 .12
   study 28 37 4 (6%)    
  1. The overall estimate of the difference in probabilities of recurrence between study and control groups is = Σ s d s w s = -.003 with a standard error .022. We define ε(max)s = max((1 - π0s )/π1s , (1 - π1s )/π0s ), where π zs equals one minus the fraction missing in group z and stratum s. The anticipated maximum bias is ψ max Σ s ε(max)s w s = ± .10 ψ max , where ψ max is the anticipated bias if there were complete confounding of the unobserved covariate and treatment.