Clinical Trials

Clinical Trials: A Practical Guide ■❚❙❘Table 2. Summary and analysis of the vitamin A trial described in Example 2 [5].Study groupsResultsNumbers Vitamins A Placebo Odds ratio (95% CI); P-valueChildren randomized 12,094 11,588Compliant subjects 9,675Noncompliant subjects 2,419Deaths among randomized subjects 46 74Deaths among compliant subjects 12Deaths among noncompliant subjects 34AnalysisMortality rates in:Randomized subjects 0.38% 0.64%Compliant subjects 0.12%Noncompliant subjects 1.41%Mortality rates according to:Intention-to-treat analysis 0.38% (46/12,094) 0.64% (74/11,588) 0.59 (0.40, 0.87); P = 0.005Per-protocol analysis 0.12% (12/9,675) 0.64% (74/11,588) 0.19 (0.10, 0.36); P = 0.001suggesting an 81% reduction in mortality in children who actually receivedvitamin A as part of the supplementation program.The information on compliance in Table 2 shows that, of the 2,419 children in thesupplementation group who did not receive vitamin A, 34 (1.41%) died. This ratewas twice that of the no-supplementation group (0.64%), despite theoreticallyhaving similar treatments (no supplementation). The reason for this was that thesame issues responsible for noncompliance – transportation and distributiondifficulties – were also responsible for other healthcare deficiencies in the samevillages, leading to a higher rate of infant mortality. In the trial, mortality wasassociated with the bias of whether the child came from a village that could orcould not receive delivery of vitamin A. In this situation, comparing only thecompliant children (PP analysis) would overestimate the efficacy.To achieve an unbiased result, the risk of mortality among compliant children inthe supplementation group should be compared with the risk of mortality of acomparable subgroup in the no-supplementation group. This comparison wouldbe more appropriate if it can be assumed that randomization led to an equalproportion of noncompliant children in both groups. For further results and anexplanation of the methodology used to analyze the data in Table 2, see [5].261