Current issues of ACP Journal Club are published in Annals of Internal Medicine


Letter

Applying the findings of clinical trials to individual patients

ACP J Club. 1995 July-Aug;123:26. doi:10.7326/ACPJC-1995-123-1-026



To the Editor

The editorial by Guyatt and colleagues (1) contained an incongruity as glaring as a cubist figure squatting on a Monet water lily.

The statement was made that because the confidence interval of a relative risk reduction (RRR) includes zero, "the results are compatible with no treatment effect." This sounds like radiology-speakā€”a bland, noncommittal structure free of information. Blue skies are "compatible with no treatment effect." So is rain.

I consider your journal to be a leader in making words mean something. This jars.

Thomas E. Finucane, MD
Johns Hopkins University
Baltimore, Maryland

In response

In our editorial, we described a randomized trial of low-dose amiodarone in patients with heart failure in which 87 of 260 patients and 106 of 256 control patients died (RRR, 19% [95% CI, 11% to 36%]). This indicates that our best estimate (or point estimate) of the true underlying treatment effect is an RRR of 19%. That 19%, however, is just an estimate, and the true treatment effect may be greater or smaller. The further from the point estimate of 19%, the less likely it is that a particular RRR represents the true underlying effect. An RRR > 36% or a detrimental treatment effect with increases in relative risk > 1% are very unlikely. Thus, although the treatment may appreciably reduce the risk for death, if asked whether we have excluded a treatment effect of 0, we must answer "no, we have not."

Dr. Finucane considers our statement that the results are compatible with a treatment effect of 0 to be free of information. We can understand his frustration, but the limitation of information is in the trial rather than in our statement.

Because of the limited sample size and the resulting width of the CI, we must conclude that, although blue skies (an appreciable treatment effect) are most likely, there remains an outside chance of rain (a treatment effect of 0). Until we have sufficiently large methodologically rigorous trials of our therapies, or meta-analyses combining the results of several of these trials, we will have to accept the same uncertainty with our treatments as we do when we listen to the weather report.

Gordon H. Guyatt, MD
Roman Z. Jaeschke, MD
Deborah J. Cook, MD
McMaster University
Hamilton, Ontario, Canada


Reference

1. Guyatt GH, Jaeschke RZ, Cook DJ. Applying the findings of clinical trials to individual patients. ACP J Club. 1995 Mar-Apr;122:A12-3.