Hematocrit Level and Postsurgical Outcome: Title and subTitle BreakPowers of Observation

In this issue of JAMA, Wu and colleagues¹ report the results of a cohort study examining the relationship between preoperative hematocrit levels and adverse events among older veterans (most of whom were men) undergoing noncardiac surgery. Incremental departures less than or greater than normal hematocrit levels for mostly men were associated with increased risk of early death and postoperative cardiac events. The authors recommend future studies to evaluate interventions aimed at reducing the risk of postsurgical events attributable to either anemia or polycythemia.

This work highlights the strengths of using high-quality, large observational databases—such as the Veterans Affairs (VA) National Surgical Quality Improvement Program (NSQIP)—to explore the relationship between patient factors and adverse outcomes. A large sample size allowed for adjustment of many potentially confounding variables that are common to surgical practice. Moreover, the sample size augments natural strengths of a cohort study, including evaluation of rare exposures (abnormally high preoperative hematocrit levels) and multiple outcomes.²

Although the authors appropriately state that they cannot establish a cause-effect relationship, they provide several classic lines of evidence consistent with a causal relationship between preoperative hematocrit and postoperative adverse events. They describe an appropriate temporal relationship, a strong association of predictor and outcome, an apparent “dose-response” relationship, and biological plausibility related to cardiac physiology.² Including data from many different hospitals and surgeons allows for broad generalizability to a large, unique, and important subset of the nation's population. While nearly all patients included in the study were men, and the cohort included only those aged 65 years or older, these epidemiologic features speak to a real relationship between abnormal hematocrit levels and outcome.

Assuming the relationship of hematocrit and outcome is real and generalizes to other cohorts, the central issue to be determined is whether modifying hematocrit improves outcome. The theory linking preoperative anemia and postoperative events is that the stress of an operation combined with the limited compensatory ability of the heart in older individuals with anemia may lead to cardiac ischemia and death. Interventions correcting anemia aimed at preventing cardiac stress might be expected to save lives, but expectation and reality are often at odds. Several randomized trials have compared blood transfusion for moderate compared with more severe degrees of anemia in cardiac surgery^{3 - 4} and critically ill patients.^{5 - 8} Although varying in the timing of transfusion, none of these studies identified important differences in clinical outcomes attributable to transfusion for hematocrit levels in the moderate anemia range specified as problematic in the VA study. In an observational study, the possibility exists that identified associations may be markers of adverse outcomes rather than predictors of outcome.

Since no intervention is without risk, clinicians should avoid using these findings reported by Wu et al to justify interventions—use of transfusion, erythropoietic agents, iron supplementation—outside the research setting. In other clinical arenas involving patients with anemia, such as those with renal failure and cancer, clinicians may have prematurely embraced the aggressive use of erythropoietic agents to boost red blood cell production in the absence of sufficient evidence, only to learn later that despite their best intentions they may have been causing more harm than benefit.⁹ The community of clinicians caring for patients preoperatively should learn from these experiences and resist the urge to “not just stand there—but do something.”

Despite these caveats, the power of a rigorously defined and validated observational cohort such as the NSQIP should not be underestimated. For instance, it is unlikely that the link between hematocrit level and outcome could have been as well described using other research designs. In this way, the VA health care system has made and continues to make, important contributions to clinical epidemiology, health services research, and a nationwide effort to improve the quality of care that cannot be overstated. Physician leaders like Kenneth Kizer—who led the transformation of the VA in the last decade—and Shukri Khuri—a cardiac surgeon who built the NSQIP—together with many others have made the VA health care system a model of integrated care and data-driven quality improvement. The effort to transform the VA serves US veterans well, and VA efforts like NSQIP will have far-reaching implications for the health of all Americans.