RT Journal
A1 Goodman S, Berry D, Wittes J
T1 BIas and trials stopped early for benefit
JF JAMA
JO JAMA
YR 2010
FD July 14
VO 304
IS 2
SP 156
OP 159
DO 10.1001/jama.2010.931
UL http://dx.doi.org/10.1001/jama.2010.931
AB
The authors' claim of a large bias is spurious because they have confused “bias” with the observation that results from trials addressing the same question will differ because of random variation. Consider an unbiased group of trials that go to completion. Trials with the largest effects will have similarly large interim effects, potentially leading to truncation.3 So comparing the truncated trials to the nontruncated trials is similar to comparing completed trials with large effects with those with lower effects. The difference the authors observed was both predictable and uninformative.