SEARCH

SEARCH BY CITATION

Keywords:

  • epidemiology;
  • mortality risk;
  • telomeres

Summary

  1. Top of page
  2. Summary
  3. Studying racial differences – seek and ye shall find
  4. How do we know a cause from an epiphenomenon?
  5. References

The advent of molecular technology that can be applied across large population samples has added – rather than reduced – complexity in the analysis of the intertwined effects of social history and heritable factors on health outcomes. The report by Hunt et al. in this issue of Aging Cell provides an example of the promises and dilemmas associated with this increased complexity.

Differentials in rates of disease among population subgroups in the USA have been recognized for several centuries. Given the centrality of race to the American historical experience, and our obsession with dichotomous (black–white) social categories, contrasts in health status between persons of European and African descent have received the lion's share of attention (Byrd & Clayton, 2000). The framework within which these observations are collected and explained has evolved over time, moving from the blatant justification of racial subjugation to the contemporary effort to balance the impact of social and genetic effects. Nonetheless, this area of scholarship continues to struggle in its effort to provide an understanding of the complex and intertwined effects of social history and heritable factors. The advent of molecular technology that can be applied across large population samples has added – rather than reduced – complexity in this research program, and the intriguing report by Hunt et al. in this issue of Aging Cell is an example of the promises and dilemmas associated with this increased complexity (2008).

Hunt and colleagues compare leukocyte telomere length in an analysis of two large samples of US blacks and whites. After adjustment for age and body mass index, the authors report that telomeres in blacks are 3.8% longer in the National Heart, Lung, and Blood Institute (NHLBI) Family Heart Study (FHS) and 7.9% longer in the Bogalusa Heart Study (Hunt et al., 2008). Other demographic and physical attributes apparently influence telomere length as well; in addition to age and body mass index, gender was a significant predictor, but only at older ages. The results were similar using restriction enzymes and quantitative polymerase chain reaction, although substantial differences were noted across studies within race groups. For example, the mean length difference between blacks and whites in the FHS is 0.269 kb, which is less than a third of the size of the mean length difference between blacks in the FHS and blacks in the Bogalusa Heart Study (0.919 kb).

These observations might not have advanced beyond the status of curiosities were it not for two additional pieces of information. First, a family-based analysis from the same data indicates that telomere length has high heritability (~70%). Second, a separate report by these investigators proposes that telomere length is predictive of mortality (Kimura et al. in press). Taken together, an interpretation is therefore clearly suggested that telomere length may in some way be implicated in the differential mortality dynamics seen in US blacks and whites. This is further highlighted by the observation that there appears to be a more rapid decrease in telomere length with age among blacks, which narrows the racial gap in length at older ages, and thus suggests a potential relationship to the crossover in black–white mortality observed around age 80.

This set of novel observations on racial patterns adds to the growing knowledge base on the potential relationship between telomeres and health status, but ultimately raises more questions than it answers. Multiple factors have already been shown to influence telomere length, and others may have gone undetected. Any examination of black–white differences quickly becomes complicated by the possibility that unknown or unmeasured exposures account for the apparent ‘racial’ patterns (Kaufman et al., 1997; Kaufman & Cooper, 2008). This ‘residual confounding’ (called ‘omitted variable bias’ in the social sciences) bedevils all comparisons of groups in widely contrasting environments whenever any of these unmeasured factors happens to be correlated with the exposure of interest (Greenland & Morgenstern, 2001).

While telomere length may be heritable, this does not necessarily imply a biologic basis for group disparities. As Feldman and Lewontin pointed out many years ago, heritability may be high, but explain none of the difference between groups (1975). For example, height is a highly heritable trait, and yet this does not imply a genetic explanation to the difference in height between Japanese Americans compared to Japanese, or between succeeding generations of Scandinavians (Feldman & Lewontin, 1975).

However tempting it might be to consider telomere length differences as a contributor to the racial disparity in mortality in the USA, the story is not at all straightforward. It is well established that US blacks have worse health outcomes for most major conditions compared to whites, but Hunt et al. report that telomere lengths are greater in blacks than in whites – a difference in the opposite direction of the existing hypothesis. As noted in a recent review, the longer telomeres seen in women are what ‘might be predicted given the greater longevity of women’ (Demerath et al., 2004). While a race-age-interaction effect could exist, the apparent crossover in all-cause mortality observed between blacks and whites is usually ascribed to differential susceptibility observed by the early death of a much larger proportion of blacks (Johnson, 2000). Moreover, there is still some considerable uncertainty surrounding the evidence linking telomere length and mortality risk in general. For example, although a univariate quantile analysis suggests an association, the relationship becomes unambiguously null once age is included in the models (Kimura et al., in press). One can only conclude that any independent contribution of telomere length to risk of mortality has yet to be firmly established.

Studying racial differences – seek and ye shall find

  1. Top of page
  2. Summary
  3. Studying racial differences – seek and ye shall find
  4. How do we know a cause from an epiphenomenon?
  5. References

The sociologist Gunnar Myrdal (1962) gave a name to half a century of scientific inquiry with the title to his 1944 book: An American Dilemma: The Negro Problem and Modern Democracy. Studies of ‘the Negro problem’ have been no less pervasive in bio-medicine than in sociology, ranging from IQ to muscle fiber distributions to resting metabolic rate, and so on. These studies are framed by the highly consistent pattern of adverse outcomes for nearly all of the key health indicators (Kington & Nickens, 2001). The search for differences between complex processes, however, will inevitably be self-fulfilling. One might by analogy consider the search for genetic predisposition to common disease. In a comparison of cases and controls for any condition, many of the 2–3 million common DNA variants will be different by chance alone. For this reason, the conduct of genetic association studies over the last 10 years has created one of the largest bodies of irreproducible results yet accumulated (Ioannidis et al., 2001). While the search for racial differences has likewise been plagued with false positives, the more fundamental problem has been the lack of any coherent biological significance for the contrasts which have emerged from such a detailed examination of biological processes.

How do we know a cause from an epiphenomenon?

  1. Top of page
  2. Summary
  3. Studying racial differences – seek and ye shall find
  4. How do we know a cause from an epiphenomenon?
  5. References

Racial comparisons are always difficult to interpret because the groups exist under very different social environments in the USA, making it tricky to define any observed biologic difference as a cause or an effect of these disparate exposures. This inability to make any balanced comparison between the groups can be viewed as a form of ‘ecologic confounding’, since any innate differences, if they exist, cannot be distinguished from irrelevant ones with which they are also highly associated (Cooper & Kaufman, 1998). The classic example of ecologic confounding occurs in studies that compare indices of development and disease rates in different countries, for example rates of TV ownership and dental caries. A clear relationship emerges, but it reflects only confounding by overall level of societal wealth. Blacks and whites on average have very different social experiences in the USA, and the indelible mark of social marginalization and exploitation is worse health. Over the last half century, therefore, we have gradually come to accept the notion that the vast majority of health differences between racial/ethnic groups are the result of social processes resulting from differential social status (Kaplan, 1999). Unfortunately, precise measurement of lifetime social status at the individual level is not well approximated by available surrogates in epidemiologic studies, leaving large proportions of unexplained variance whenever models attempt to adjust for this confounding. The consequence is that a large variety of forces get lumped into the error term, resulting in exposure estimates that have no clear utility or interpretability whenever this error term is correlated with the exposure.

The search for causal relationships through the study of black–white differences often founders on this irresolvable problem of residual confounding (Kaufman, 2008). For example, within the field of cardiovascular research there have been many attempts to explain higher hypertension rates among US blacks by searching for metabolic differences. In the 1980s and 1990s, there was substantial interest in cation transport, for example (Adragna et al., 1982). While it became clear that blacks have higher intracellular sodium and lower sodium-lithium countertransport in red cells, these traits appear unrelated to blood pressure except via their correlation with race (Cooper & Borke, 1993). These metabolic differences are therefore an epiphenomenon, a mere curiosity. Studies of racial difference are often predicated on the idea that examples such as these will uncover clues to biological processes influencing health and disease. To date, however, we are unable to think of a single instance when the comparison of metabolic traits between two racial/ethnic groups has successfully led to the identification or understanding of a real physiologic process that was ultimately linked to health status. Given the double risk described above – the abundance of false positives and the presence of strong residual confounding – studies within ethnic groups have remained a far more reliable method to search for new causal exposure–outcome relationships, a notion that is already widely accepted in the world of genetic epidemiology (Serre et al., 2008). We would posit that the difference in telomere length between blacks and whites described in the current report is another such epiphenomenon.

References

  1. Top of page
  2. Summary
  3. Studying racial differences – seek and ye shall find
  4. How do we know a cause from an epiphenomenon?
  5. References