A Randomized Controlled Trial may be blinded, (also called "masked") by "procedures that prevent study participants, caregivers, or outcome assessors from knowing which intervention was received." Wood L,et al."Empirical evidence of bias in treatment effect estimates in controlled trials with different interventions and outcomes: meta-epidemiological study". BMJ 336 (7644): 601–5 (2008).
During the conduct of this Phase IIb 2nd line NSCLC double-blinded trial, neither the patients nor the clinicians knew who was receiving what treatment. As this was the case, the conduct of the trial itself could not have been "tainted" by any miscoding, because no one knew who was getting what in the first place. All of the patients simply received their coded treatments. They did not know what the codes meant.
It was only after unblinding that this discrepancy between sample test result and treatment code was found. What this means is only that patient "A" was coded to receive treatment "X" but actually received treatment "Y". What does this mean? Does it mean that all the patients could have received placebo? Does it mean that all the patients could have received Bavi? Does it mean that only 1/3 of the patients could have received Bavi and 2/3 could have received placebo?
What was known is that 1/3 of the patients had to receive placebo, 1/3 had to receive Bavi 1mg, and 1/3 had to receive Bavi 3mg. As such, for any one patient "A" coded to receive treatment "X" but actually received treatment "Y", there had to be another patient "B" coded to receive treatment "Y" but actually received treatment "X", otherwise, the overall numbers of patients in each arm would not have corresponded to the 1/3, 1/3, 1/3 groupings.
To put it another way, we know that one third of the 121 patients received placebo, one third received Bavi 1mg, and one third received Bavi 3mg. The only correction needed is to verify which patients those were.
So the way I am seeing this, is that the identifications of the treatments actually received by the patients must be sorted out and verified. When this is accomplished, it will be verified which 1/3 received the Bavi 1 mg, which 1/3 received the Bavi 3 mg, and which 1/3 receive the placebo. The post-hoc data analysis then will be performed on the now verified and confirmed data correctly identifying which patient received what treatment.
All IMO.
