| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letter to the Editor |
Department of Preventive Medicine and Epidemiology Loyola University Medical Center Maywood, IL 60153 E-mail: rcooper{at}lumc.edu
Dear Sir:
We disagree with the conclusions about the uncoupling protein 3 gene (UCP3) as a genetic determinant of susceptibility to obesity in blacks and whites that were put forward by Kimm et al in their report in a recent issue of the Journal (1) and in the accompanying editorial (2). A serious problem exists in genetic epidemiology in relation to multiple statistical testing. Kimm et al studied 8 single-nucleotide polymorphisms (SNPs) and reported that, when one of them (UCP3 exon 5) was used to create a stratum, a significant difference existed for resting metabolic rate (RMR) between blacks and whites (P < 0.02 for the race interaction). However, at least 8 tests would need to have been conducted for that analysis alone, even without consideration of the other possible phenotypes (eg, obesity and RMR). A standard multiple testing correction would require a much smaller P value than was reported, because this was not an a priori hypothesis. It also appears that the significant comparison involving the CC genotype was based on 6 blacks and 15 whites, which does not seem like a very large sample on which to base an inference that genetic susceptibility varies between 2 entire populations. It should further be noted that this SNP had no relation to RMR in whites, which suggests that the finding in blacks is a Type 1 error.
Aside from the deficiencies in the statistical analysis, we are concerned about the inference of causality that is being made. Whereas it is true both that many studies have shown that RMR is lower in blacks than in whites and that a nominally significant difference exists by genotype between the racial groups in this study that ignores multiple comparisons, no evidence is presented by Kimm et al that RMR is related to obesity. In fact, of all studies from the literature, only one shows a causal connection between lower RMR and susceptibility to obesity, even though that is an obvious hypothesis. It is not at all clear to us, therefore, why the authors and the editorialists would conclude that the UCP genotype is a cause of differential racial susceptibility to obesity when there is little evidence it is related to obesity.
On the other hand, it seems to us that there are other, much more plausible explanations for the differences in obesity among the various ethnic or racial groups in the United States. The large social class effect suggests a crucial role for lifestyle. Many groups, notably Native Americans and Hispanics, suffer from epidemic obesity to the same or a greater degree and are not particularly similar to blacks in terms of genetic affinity. Likewise, temporal trends show that, in the 1960s, blacks were not more obese in general than were whites, and in other countries blacks do not have high rates of obesity (3).
A tendency exists in epidemiologic research to conclude that genetic factors are important in racial differences, and a substantial number of reports make this assertion in other areas, such as hypertension (4–7). That literature, however, has been very unreliable, postulating effects that can never be replicated. In fact, it is very difficult to identify a report of causal genetic differences between blacks and whites in which the findings have stood the test of time as reliable or useful. In the case addressed here, it seems to us that much more would be accomplished by attempting to define the underlying causes of obesity before postulating genetic causes of differences. Kimm et al have prospective data on RMR and weight change, for example. It would have been more instructive to know whether a relation exists between RMR and obesity before concluding that a causal connection exists between an SNP in UCP and the differences between blacks and whites in susceptibility to obesity.
In our opinion, race has been overused in epidemiologic research as the basis for causal hypotheses (6, 7). This practice has been a serious distraction and has wasted considerable resources (8). Without a clear-cut demonstration of a relation between the putative at-risk genotype and the phenotype of interest in both groups, causal inferences are not justified.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
