Second, the survey cohort is a subgroup of the full study population and the findings in that cohort may not fully repre- sent findings for the whole cohort. However, we found no evidence of sample-selection bias. The intensive-therapy group and the standard-therapy group did not differ significantly with respect to any baseline characteristics, and a sensitivity analy- sis that used data from the complete cohort showed results very similar to the results of analy- ses that used data from the survey cohort