Response to Letter by Bath et al
We read with interest the comments of Dr Bath and his associates related to our article. The focus of our report was the persistence of the effect of chlorthalidone-based therapy in significantly decreasing cardiovascular mortality up to 14.3 years after randomization in spite of the advice to all SHEP participants to receive active therapy.
However, a statistically significant effect on fatal stroke was not observed nor was there any interaction between treatment and the occurrence of stroke during SHEP on subsequent mortality.
Dr Bath and associates present an interesting way of looking at the stroke data. In our opinion, their statistical approach has limitations as follows. First, ordinal logistic regression assumes 1 category per person. As our Table 1 states, there were some participants who had both a fatal and a nonfatal stroke while others had both a TIA and a stroke. Second, ordinal logistic regression does not account for time to event; for example, the proposed method will not account precisely for a short time to a TIA event versus a long time to a stroke. Third, ordinal logistic regression makes the strong assumption that the reduction of risk for chlorthalidone versus placebo is the same in going from fatal stroke to nonfatal stroke as it is in going from nonfatal stroke to no stroke or the equivalent sequence when including TIAs. Lastly, the proposed method in the 3-tier analysis equates stroke case fatality with stroke severity. Stroke severity also includes the extent and clinical importance of the neurological deficit, and its effect on activities of daily living and quality of life.