LAUSR

The Insulin Resistance Intervention After Stroke (IRIS) trial has reported that treating insulin resistance with the peroxisome proliferator–activated receptor γ agonist pioglitazone hydrochloride reduced recurrent stroke or myocardial infarction (MI) by about onefourth compared with placebo (pioglitazone, 9.0% vs placebo, 11.8%; hazard ratio [HR], 0.76; 95% CI, 0.62-0.93) in 3876 patients with recent (<6 months) ischemic stroke or transient ischemic attack and insulin resistance but without diabetes, heart failure, or bladder cancer.1 Pioglitazone was also associated with less incident diabetes vs placebo (3.8% vs 7.7%; HR, 0.48 [95% CI, 0.330.69]) but more weight gain above 4.5 kg (52.2% vs 33.7%), more peripheral edema (35.6% vs 24.9%), and more bone fracture requiring surgery or hospitalization (5.1% vs 3.2%). The weight gain reflects an increase in adipose tissue and fluid due to renal sodium retention. As the latter may predispose to heart failure, patients with a history of heart failure were excluded from the IRIS trial. Patients with a history of bladder cancer were also excluded because pioglitazone may increase the risk of bladder cancer.2-4 The results of the IRIS trial are consistent with the effects of pioglitazone in other patients with insulin resistance, prediabetes, and type 1 or type 2 diabetes,5,6 and are therefore likely to be valid and generalizable. Treating 100 patients with ischemic stroke or transient ischemic attack and insulin resistance with pioglitazone for nearly 5 years can be expected to prevent a recurrent stroke or MI in about 3 patients (number need to treat for 5 years to prevent [or cause] 1 event [NNT] = 36) and a new diagnosis of diabetes in 4 patients (NNT = 26), but also to cause weight gain of more than 4.5 kg in 18 patients (NNT = 5), edema in 10 patients (NNT = 9), and bone fracture in 2 patients (NNT = 53). These numbers suggest that the overall net benefits of pioglitazone may be offset by the risks. However, they also raise the question of whether patients likely to benefit from pioglitazone therapy can be identified and distinguished from those unlikely to benefit, so that treatment can be targeted to the former group. An analysis of 13 prespecified subgroups defined by single variables did not reveal significant heterogeneity of the reduction in relative risk of stroke or MI with pioglitazone vs placebo in any subgroup.1 Such consistency of the relative treatment effect across subgroups is the rule, rather than the exception.7 What did vary within some subgroups was the absolute rates of stroke and MI, as is also generally the rule. For example, among participants allocated to receive placebo, the absolute rates of stroke and MI at 5 years were higher for those 65 years of age or older (13.8%) than for those younger than 65 years of age (10.2%), for those with coronary heart disease (21.3%) than for those without (10.6%), and for those with hypertension (13.4%) than for those without (7.7%).1 Hence, the consistent one-fourth reduction in the relative risk of stroke or MI with pioglitazone across subgroups translated to greater absolute reductions in stroke or MI within the subgroups with a higher absolute risk (eg, the elderly and those with coronary heart disease or hypertension) than within the subgroups with lower risk.1 Subgroup analyses of single variables do not, however, account for multiple characteristics among patients that simultaneously affect their prognosis and response to treatment. A potentially more informative strategy of identifying patients who may benefit from treatment is to use multivariable risk analysis to stratify individual patients according to their baseline risk of major outcome events.8,9 In this issue of JAMA Neurology, Kernan et al10 report the results of a secondary, post hoc analysis of the IRIS trial in which Cox proportional hazards regression multivariate analysis of the whole trial population was undertaken to determine the following: (1) the independent effect of all potential prognostic factors that were recorded at baseline before treatment, (2) the absolute risk of stroke or MI for all patients based on their prognostic factor profile, and (3) the effect of intention to treat with pioglitazone vs placebo on Kaplan-Meier 5-year rates of major outcome events according to each participant’s baseline estimated risk of stroke or MI, stratified into quantiles (eg, halves, tertiles, quartiles, and quintiles). The baseline factors most strongly associated with stroke and MI were increasing age; a history of prior stroke, coronary artery disease, or hypertension; current smoking; and aphasia.10 Each factor, including a 10-year increase in age, independently increased the relative risk of stroke or MI by about two-thirds. For the half of participants with an estimated risk of stroke or MI above the median, the 5-year risk of stroke or MI was 19.6% in the placebo group vs 14.7% in the pioglitazone group (HR, 0.75 [95% CI, 0.60-0.95]; absolute risk reduction [ARR], 4.9% [95% CI, 1.2%-8.6%]).10 For the other half of participants with an estimated risk of stroke or MI below the median, the 5-year risk of stroke or MI was 7.9% in the placebo group vs 6.0% in the pioglitazone group (HR, 0.77 [95% CI, 0.531.11]; ARR, 1.9% [95% CI, –0.6% to 4.4%]).10 The absolute benefit of pioglitazone was therefore greater for participants with Related article Opinion