A new look at an old therapy.

DESPITE MAJOR ADVANCES IN THE CARE OF PAtients with acute coronary syndrome over the past 3 decades, rates of early morbidity and mortality associated with this condition remain unacceptably high. In the 1980s, the concept of administering glucose, insulin, and potassium (GIK) to patients with suspected myocardial ischemia garnered substantial interest in both the research and practice communities. The therapy is remarkably simple and available globally at a low cost—an especially attractive feature given the increasing incidence of cardiovascular disease in low-income countries. The results of several clinical trials of moderate size seemed to suggest a bright future for GIK therapy; however, a subsequent large pragmatic trial found no benefit, and discussion of the treatment subsided. Selker and colleagues, however, were not comfortable with the conventional wisdom that GIK should be retired from the therapeutic armamentarium. Stimulated both by the prodding of Apstein and by evidence from an animal model that the major effect of GIK was seen during the very early phase of coronary occlusion, the investigators were concerned that the therapy might have benefit but only if applied early, preferably within the first hour from onset of symptoms. They hypothesized that GIK therapy, if given early and prior to reperfusion, would reduce infarct size and lower mortality by preventing sudden arrhythmias. As reported in this issue of JAMA, Selker et al proposed an innovative study, the Immediate Myocardial Metabolic Enhancement During Initial Assessment and Treatment in Emergency care (IMMEDIATE) trial—funded by the National Heart, Lung, and Blood Institute (NHLBI)—and created a network to study GIK in the setting of emergency medical services. The IMMEDIATE trial was designed as a large pragmatic effectiveness trial with the primary goal of examining the effect of GIK on the rate of total mortality at 30 days and at 1 year. However, a combination of ambitious investigators and the particulars of the study mechanism combined caused the trial to fall far short of its enrollment goals. With support from the study’s data monitoring committee and from the NHLBI, the investigators reconfigured the trial to address a primary end point of progression to infarction. The more important end points (survival, cardiac arrest, and progression to heart failure) were designated as secondary end points, and in an interesting and unusual approach, these outcomes were examined with an overt statement that the 5 major secondary and 8 other secondary end points would be evaluated with an level of .05 for each hypothesis. By conventional criteria for multiple hypothesis testing, none of these tests would attain statistical significance. The cardiovascular community has been chastened by appealing biological hypotheses that were supported by secondary hypothesis testing that turned out to be random noise. The IMMEDIATE trial showed no effect of GIK on progression to infarction, but all major clinical end points trended in favor of treatment with GIK, as did infarct size in the small number of participants included in the mechanistic substudy. These results raise 4 questions that should be the subject of keen interest among the medical and public health communities. First, how should the trial findings be interpreted for practical purposes? The GIK therapy failed with regard to its primary end point, but there are tantalizing signs of benefit in more important secondary end points representing morbid and mortal disease outcomes. Given the massive amount of data from previous trials that on balance showed neither benefit nor significant risk, the clinical community is presented with a conundrum. Were previous trials flawed by late treatment with GIK that was given past the point when it can show benefit? Or was the IMMEDIATE trial a purely negative trial that exhausted its allotted statistical power with a negative primary question, rendering secondary trends of interest but of limited use in the quest to definitively decipher signal from random noise? There are no clear answers to these questions. There should be no rush to treatment, but more data are needed to answer the question definitively. Second, does the trial add new information that points toward the future use of this therapy? In an ideal scenario, this trial would be followed immediately with a definitive study, one that enrolled patients only within the first hour of symptom onset, with total mortality as the primary outcome. Such a trial probably could be done inexpensively, measuring the primary end point with a postcard or by harvesting electronic health records for vital status and symptoms of heart failure, depending on the environment. A rigorous meta-analysis of all previous data, specifically