Monday, January 16, 2006

BMJ Letter Attributes Helena and Pueblo Heart Attack Declines to Random Variation

In a rapid response published on the BMJ online, Drs. Brad Rodu (Professor of Medicine at the University of Louisville) and Philip Cole have presented data which seems to confirm that the observed changes in heart attack admissions in Helena and Pueblo could be attributed to simply representing random variation in the underlying secular trend in that variable.

These authors used data on heart attack mortality in Lewis and Clark county in Montana (home of Helena) and Pueblo county in Colorado (home of Pueblo) to gauge trends in heart attack deaths over time (1979 to 2002) in these two cities.

First, the authors found that there was no decline in heart attack deaths from 2001 to 2002 (the year of the ban) in Helena, suggesting that whatever happened in Helena, fewer people did not die from heart attacks because of the smoking ban during its first six months.

Second, the authors found very high variability in year-to-year trends in heart attack mortality in both Pueblo and Helena, with several particularly striking variations, such that the single observed change in heart attack admissions from 2001 to 2002 in Helena and 2001-02 and 2003-04 in Pueblo could easily be attributed simply to random variation.

Demonstrating how much variability there was in these data, the authors point out that in Pueblo, there was a 28% decline in heart attack mortality between 2001 and 2002, even in the absence of a smoking ban. Had the smoking ban been implemented in 2001, the reasoning used by the Pueblo study authors could have concluded that 28% decline was caused by the smoking ban.

Moreover, these data demonstrate that there was a pre-existing decline in heart attack mortality in both Helena and Pueblo going into the time at which the smoking bans were implemented. So it would be expected that heart attack mortality would continue to decline.

The Rest of the Story

The bottom line is that I think Rodu and Cole have provided a compelling argument, backed up by convincing data, that there is tremendous variability in heart attack trends in Helena and Pueblo, large enough such that one cannot and should not make too much out of the simple observation of a change from one year to the next (or one 18-month period to the next) in this particular variable.

In other words, the observed changes in heart attack admissions in Helena and Pueblo could easily be attributed to random variation, rather than to an effect of the smoking ban.

Do these data prove that the heart attack decline was not due to the smoking ban? No. But they provide a plausible alternative explanation for the observed findings, and that is enough to render invalid the definitive causal conclusions that have been drawn from these studies.

No comments: