[Skip to Content]
[Skip to Content Landing]
Citations 0
April 26, 2000

Intra-arterial Prourokinase for Acute Ischemic Stroke

Author Affiliations

Phil B.FontanarosaMD, Deputy EditorIndividualAuthorStephen J.LurieMD, PhD, Contributing EditorIndividualAuthor

JAMA. 2000;283(16):2102-2104. doi:10.1001/jama.283.16.2101

To the Editor: I am puzzled that the PROACT II study1 investigators presented their results as odds ratios (ORs) rather than relative risks (RRs). The ORs are estimates of RRs, and their use should be restricted to situations in which true RRs cannot be calculated. They should be used in case-control studies in which multivariate analysis requires the use of logistic regression, or cohort studies in which the rare disease assumption (≤10% of patients have the outcome of interest) allows the OR to be a good estimate of RR.2-4 Given that this was a randomized trial, that logistic regression was not used, and that "good outcomes" were found in 25% and 40% of the placebo and prourokinase arms respectively, it would have been more appropriate to compute the true RR. Using the data presented in Table 3 of the article, I calculated the National Institutes of Health Stroke Scale strata specific RRs and used the Mantel-Haenszel method to adjust overall RR of a good outcome in patients treated with prourokinase vs placebo. These RRs are compared with the ORs from the article (Table 1). The analysis was performed using EPI Info (version 6.0B, Centers for Disease Control and Prevention, Atlanta, Ga).