Obstructive Sleep Apnea and Postoperative Major Adverse Cardiac or Cerebrovascular Events: Was the Composite End Point Appropriate and Valid?

Abstract

To the Editor We read with interest Nagappa et al’s1 recent systematic review and meta-analysis on the associations between obstructive sleep apnea (OSA) and postoperative complications in cardiac surgical patients. Their main finding was that after cardiac surgery, the estimated odds ratio for major adverse cardiac or cerebrovascular events (MACCEs) was 2.4 (95% confidence interval, 1.38–4.2) in OSA versus non-OSA patients. This topic is interesting and potentially important, given the high prevalence of OSA in the cardiac surgical population. However, we worry about the validity of their result because the authors used inappropriate statistical methods. The primary end point of this meta-analysis was MACCE, which, as defined by the authors, was a composite end point of all-cause death, myocardial infarction, nonfatal cardiac arrest, revascularization, pulmonary embolism, deep venous thrombosis, atrial fibrillation, stroke, and congestive heart failure. The use of this composite end point was inappropriate. First, the component end points were of widely differing importance to patients. Similar severity of components is the most important of the criteria for a good composite end point. If this was not achievable, the authors could have introduced severity weights to attempt to address the heterogeneity among the severities of the components.2 Although weights are usually subjective and inaccurate, they would have made the analysis and interpretation of results much stronger. Second, the frequencies of component end points were not comparable. For example, death accounted for only 8.4% of MACCEs, while atrial fibrillation accounted for 72.0% of total events. The association of OSA with the component with the largest frequency will tend to drive the overall estimate. To address this issue, the average relative effect odds ratio method could be considered.3 This method has the desirable property of not allowing the components with higher incidence to drive the results. It also allows introduction of severity weights. However, raw data from original studies are needed to use this method. Third, the strength of associations between OSA and component end points was not expected to be similar due to their different pathophysiology. Using the data obtained from Nagappa et al’s1 meta-analysis, we found that OSA was significantly associated with increased risk of all-cause death, but not myocardial infarction, congestive heart failure, or stroke (Figure). Nagappa et al1 could have reported on the individual components to assess how different the associations with outcomes were. They could also have conducted a statistical test for homogeneity of odds ratios. The combination of the visual display of the odds ratios and the test would have most likely concluded that it was inappropriate to combine the components. On the other hand, if all components were similarly severe, it would have made much less difference that there was heterogeneity.Figure.: Summary estimates for the associations of obstructive sleep apnea with the composite end point MACCE and some important components, calculated using data obtained from Nagappa et al’s1 meta-analysis. CI indicates confidence interval; MACCE, major adverse cardiac or cerebrovascular event; OR, odds ratio.Composite end points are frequently used in studies of anesthesia and other clinical specialties to better capture the disease of interest and/or to increase statistical power. However, concerns about their use have emerged.4 Composite end points should be defined strictly and analyzed with advanced statistical methods.2 Their use in meta-analyses warrants particular caution because the raw data needed in many sophisticated statistical methods are often unavailable. Bing-Cheng Zhao, MDFang-Ling Zhang, MDKe-Xuan Liu, MD, PhDDepartment of AnesthesiologySouthern Medical University Nanfang HospitalGuangzhou, Guangdong, China[email protected]