Beutler and Waalen1 undertake the Herculean task of tackling the nearly 40-year-old World Health Organization (WHO) definition of anemia. The authors mention several reasons why these old parameters cannot be taken as valid, and we fully agree with all of them. Furthermore, they describe 2 newer large epidemiologic studies, the NHANES-III (the third US National Health and Nutrition Examination Survey) and Scripps-Kaiser databases, which demonstrate very similar hemoglobin concentrations among the different segments of the US population. Both of these studies should be used as a platform to establish the lower limit of normal blood hemoglobin concentration. However, we disagree with the authors that drawing the line of normality should be done by choosing an arbitrary 95%, 97.5%, or any other percentage value out of the Gaussian distribution. We should not be afraid to raise the bar to a level where possibly 22% or more of supposedly healthy men or women in the US would be anemic. The question that needs to be asked is where we start to define what is healthy and what is not. An approach that might help us out of this eternal dilemma and lead us to define the lower limit of normal hemoglobin was presented by Zakai et al2 in their prospective epidemiologic study, where they take into account the effect of hemoglobin concentration as an independent mortality risk predictor. We are well aware that using 133 g/L and 145 g/L (13.3 g/dL and 14.5 g/dL) for women and men 65 years or older, respectively, could create a tremendous public-health burden. However, we should face it, just as we have come to understand in the last 25 years the value of establishing a healthy low-density lipoprotein (LDL) cholesterol level, rather than the previous normal levels by our population standards. Even now, the appropriate LDL cholesterol level is being readjusted for some high-risk populations to ever-lower limits. The other lesson, therefore, that the LDL cholesterol example should give us when we search for a new lower normal limit of hemoglobin concentration is to stratify the population by meaningful underlying comorbidities, rather than simply aiming for a universal value according to age, sex, and race.
Establishing normal limits for blood hemoglobin concentration
We agree with Boehringer and Darden that establishing normal limits is fraught with difficulty and that cutoff values are, of necessity, always somewhat arbitrary. If physicians choose to employ the 95% cutoff they are merely allocating a probability to the patient being a part of the normal distribution of hemoglobin values. As Boehringer and Darden state, it is entirely possible that it is better for hemoglobin levels to be higher. In the game of basketball, being of average height is a disadvantage even though it is clearly normal; perhaps average hemoglobin levels are not optimal in the game of life, either.
We also agree that it is important to try to arrive at some type of functional definition of normal. In the case of their example of LDL cholesterol and heart disease, the data required to establish functional cutoffs have taken decades to acquire and continue to be modified with additional data. In the case of anemia, we are only beginning to accumulate the necessary data. Although the study of Zakai et al,1 which shows increased mortality among subjects 65 and older with hemoglobin levels even in the “normal range,” is intriguing, in other studies2,3 confounding comorbid conditions accounted for most or all of the difference between anemic and nonanemic groups. Furthermore, even if 22% of the population were at risk for early demise because their hemoglobin levels were found to be low, it would remain to be shown that therapy directed at the hemoglobin level would have any effect on survival. In addition, given that high hemoglobin levels also have deleterious effects on function, it is likely that the normal distribution of hemoglobin represents a balance struck by evolutionary forces. Future studies, some of which we hope to be able to perform, may clarify some of these issues. In the meantime, we believe that physicians will have to content themselves with reference standards based on population norms, as we do with most other clinical laboratory measurements.
Correspondence: Ernest Beutler, The Scripps Research Institute, 10550 N Torrey Pines Rd, La Jolla, CA 92037; e-mail: email@example.com.