chapter  8
25 Pages

Genetic variation and ethnic variability in disease risk

Spatial and temporal variation in disease risk: the role of ethnicity It is almost trite to note that risk of disease varies by time and place. From John Snow onwards, the perceptive investigation of spatial and tem poral variability in disease risk has helped uncover the aetiology of disease. This prem ise underpins the science of epidemiology. By applying this principle, epidemiological studies have succeeded in identifying the causal factors for a great many diseases with varied aetiologies. As a consequence, this discipline has made a substantial contribution to worldwide public health. However, the ability to transform epidemiological findings into successful public health policy requires more than identifying the underlying aetiology of disease. Development of public health policy is substantially influenced by w hether causal factors are perceived to be intrinsic or extrinsic in nature.At the biological level, the distinction between intrinsic and extrinsic is necessarily blurred by the com plexities of pathophysiology. Even so, the perception of aetiological dichotomy can exert a considerable im pact on how modification of disease risk is approached. A ppropriate m anipulation of extrinsic risk factors, whether environmental or behavioural, offers the promise of disease prevention. By contrast, while disease due to intrinsic factors may be susceptible to treatm ent, it will be much harder to prevent its initial occurrence. Since prevention is usually cheaper and more effective than treatm ent, the perception of w hether disease risk is intrinsic or extrinsic has im portant implications for public health policy.In this context, understanding and in terpreting the aetiological foundation o f ‘ethnic-specific’ disease risk can influence the way strategies are developed to reduce the burden of disease. As a consequence, the public hea lth implication of identifying ethnic differentials in disease risk transcends the etymological argum ents over the m eaning of ‘ethnicity’. Com m unities and individuals are increasingly classified in term s of their ethnic identity - w hether self-defined or adm inistratively imposed. In the epidemiological context, ethnicity is frequently regarded as im portant a clue to aetiology as was Snow’s identification of the company that supplied w ater to the Broad Street pump. If disease distributions vary by ethnicity, as they surely do, it

m atters a great deal how the factors contributing to ethnic-specific risk are perceived. If ethnic-specific risk is considered to be the consequence of intrinsic factors, approaches to prevention are likely to be m uted. This is particularly so when ‘ethnicity’ is equated to ‘genetic’ - as is all too frequently the case. The perception of how ethnic differences influence disease risk, and the extent to which these differences are perceived to be genetic in origin, have im portant implications for both the population and the individual.Characteristics of the external environment are universally regarded as risk factors tha t are unambiguously extrinsic in nature. In this case, the distinction between tem poral variation and spatial variation can be used as a surrogate indicator of the relative im portance of extrinsic factors. Variation in disease risk over time is best in terpreted as a consequence of the tem poral variation of aetiological factors tha t are external to the population. Such extrinsic agents can range from physical factors, such as changing climatic patterns, to socio-demographic changes that influence population density and nu tritional profiles. Except in a ra th e r superficial sense, tem poral variation in disease risk does not reflect changes in the intrinsic properties of the population. This is particularly true for infectious disease.While there are notable differences in the prevalence of infectious disease among ethnic groups, the prim ary reason is attributable to external factors tha t influence the ecology of pathogens and their vectors. Consequently, tem poral variation in risk of infectious disease is attribu ted to changes in the ex ternal environm ent. An exam ple is the dram atic reduction in the burden of infectious diseases since the tu rn of the century, brought about by a general rise in living standards. These striking changes, which occurred throughout the developed world, were not confined to a subset of ethnically defined populations, and there is no evidence th a t in trinsic biological differences among populations played a role. The substantial differences in infectious disease risk tha t rem ain, including the re-emergence of infectious disease as a significant public hea lth th rea t, are due to ecological and economic differentials. Similarly, although less dram atic, the secular changes in risk of non-infectious disease can also be a ttribu ted to the changing distribution of external risk factors. The reduction in cardiovascular mortality essentially occurs at the national level, ra ther than at the level of ethnic groups. D ifferential rates of change are far more likely to be due to aspects of the macro-social environment than differences in genetic heritage.Spatial variability in disease risk can be more problem atic to in terpret. While extrinsic factors resulting from spatial variation of the physical and biological environment contribute to disease risk, intrinsic attributes of the population are also im plicated. Since com m unities tend to be spatially constrained, this includes ethnicity. At a global scale, much of the geographic and ecological variation in disease is frequently perceived to result from ‘ethnic-specific risk distributions’ (Polednak 1989). Indeed, it is not uncommon to find tha t much of the spatial variation in disease risk is simply a ttributed to ethnic differences in risk, as if this constituted an adequate explanation of

aetiology. This is particularly true for non-infectious disease. For example, the epidemiological litera tu re is replete with discussions concerning the increased susceptibility of (US) Black people to hypertension and other cardiovascular disease. For over 50 years, there has been a persistent tendency to attribu te this major public health problem to ‘African genes’ (Cooper and Rotimi 1994). This is despite the pioneering work of Scotch (1963), Akinkugbe and Ojo (1969) and Shaper et al. ( 1969), which clearly implicated environment, ra ther than genes, as the root cause of differential risk of hypertension. In similar fashion, even anthropologists have entered the fray by coining the ‘New World Syndrome’ (Weiss et al. 1984) as an appellation for increased suscep tib ility to a m yriad set of d iseases (not ju s t one) p resum ed to characterise all Amerindians.Unfortunately, such labels provide little insight about the real factors that predispose to disease risk. Even worse, their all too frequent use is likely to divert attention from the underlying complexity of interactions that lead to disease. To believe that the dram atic elevation in diabetes prevalence among some Native Am erican groups is simply a consequence of Am erindian genes can tu rn attention from other, even more im portant, factors. Since Chinese, Asian Indians, African Americans and Micronesians can also display similarly elevated rates of diabetes, it should be clear that the critical risk factors transcend ethnic background and genetic heritage. This is not to gainsay that genes play a role. They surely do, especially at the individual level. It is also im p o rtan t to recognise th a t even qu ite g en e ra l changes in the environment can in teract with underlying variation in genetic susceptibility to cause substantial changes in the risk of non-infectious disease. However, by failing to recognise that ethnicity defines community ecology, as well as genetic heritage, epidemiological studies run the risk of not only focusing on the less relevant risk factor but also on the risk factor that is most resistant to change. Ethnic variability in disease risk: risk differentials in ecology and genetic heritage Although a ttributing variation in disease risk to ethnic differences explains little, it is im portant to recognise that ethnic groups do exhibit characteristic profiles of disease. This suggests more a ttention should be paid to what is presum ed to be encompassed by the label ‘ethnic differences’. Moving beyond the recognition of ethnic differences in disease risk to a considered analysis of the underlying factors that contribute to these differences will help identify the relative contribution of both extrinsic and intrinsic factors. This could lead to advances in both treatm ent and prevention. While the in terpretation of ‘ethnicity’ is fraught with difficulty (Macbeth, C hapter 2), the implicit m eaning of the te rm can range from a defin ition of the m acro-social e n v iro n m en t to an in d ic a tio n o f g e n e tic c o n s titu t io n . W ith in th e epidemiological literature, where ethnic identity is often equated with racial

classification (Polednak 1989), the label is frequently used as a surrogate for ‘genetic distinctiveness’. In reality the situation is much more complex. By and large, it is true th a t ethnic groups tend to differ in th e ir genetic composition (to a greater or lesser degree). These genetic differences can influence the distribution of disease risk. However, with the exception of the classical M endelian disorders, environm ental factors also influence disease risk, and ethnic groups are as variable in their ecology as they are in the genes.In many respects, ethnicity has the greatest relevance for the distribution of non-infectious disease. While the prevalence of certain infectious diseases exhibits associations with ethnicity, this is essentially due to the geographic constraints on the distribution of infectious pathogens and their vectors. This was especially true before the advent of large-scale hum an migrations, when the geographic structure of populations m eant ethnicity was tightly linked to a local ecosystem. Five thousand years ago, the geographic dem arcation of many infectious diseases tended to coincide with ethnic boundaries. Since then, the impact of em pire and conquest has disrupted the demographic integrity of local populations. This uncoupled the formerly tight relationship between individual communities and their local environment. In today’s world, being in Africa ra ther than being an African is the relevant risk factor for con trac ting falciparum m alaria . However, because of its effect on the contem porary p a tte rn of genetic factors th a t influence disease risk, the h is to r ica l c o rre la tio n b e tw een ecolog ical h a b ita t and local g en e tic differentiation still has im portance today.The ecological factors that differentiate ethnic groups are many and varied. There are a num ber of ways in which this ecological variation can lead to the ch a rac te ris tic d is tr ib u tio n of disease risk am ong e th n ic groups. It is tran sp a ren t th a t w henever ethnic groups occupy distinct geographical localities they will d iffer in th e ir im m ed ia te physical and biological environment. Such variation in the external environment not only has a major impact on the risk of infectious disease but also can influence the risk of a m ultitude of non-infectious diseases. More relevant, and certainly more challenging for the epidemiologist, is the im pact of what might be term ed the ‘ethnic macro-social environm ent’, which exerts a major influence on the risk of disease. Ethnic distinctiveness is partly a function of cultural ecology and of social ecology, and each of these aspects of the ethnic macro-social environment can play a critical role in the risk of disease. C ultural ecology, which can influence everything from house style to the availability of basic foodstuffs, can have a significant impact on both infectious and non-infectious disease. The social ecology of the group, which will affect the distribution of high-risk behaviours, not only influences risk of transm ission of infectious disease but is critically im portan t in determ in ing exposure to adverse environm ental factors.To a large extent, it is the in teraction between cultural ecology and social ecology, ra ther than geographic locality, which determ ines the ethnic-specific

distribution of environmental risk factors. It is against the profile of the ethnic macro-social environment tha t any genetic propensity to disease will act. Since every ethn ic group is defined by its ch arac te ris tic m acro-social environment, as well as by its unique genetic history, it follows that ‘ethnicity’ m ust also incorporate a distinctive pattern of disease risk. A unique pattern of disease risk is thus ju st as im portant a m arker of ethnic identity as is outw ard appearance or cu ltu ra l custom . For th is reason, ‘e th n ic ity ’ is frequently used as a critical descriptor when analysing the complex diseases that are such a burden for contem porary societies. However, it needs to be stressed that no single component dom inates this consequence of ethnicity. Consideration of the extensive tabulation by Polednak (1989) indicates that genetic heritage and socio-cultural environment, or a combination of both, can lead to substantial fluctuations in disease risk among ethnic groups. For some diseases socio-cultural factors play the dom inant role [e.g. risk of sexually transm itted diseases (STDs) and hum an immunodeficiency virus (HIV)], in others genetic heritage is the more im portant (e.g. sickle cell disease in African populations and Tay-Sachs disease in Ashkenazi Jews). However, for the vast majority of diseases, both genetic heritage and socio­cultural environment play a role, albeit with differing m agnitudes. The rest of this chapter examines how underlying differences in genetic heritage can influence the ethnic-specific distribution of disease risk. Distribution of genetic variation in human populations: some consequences The relationship between genetic differentiation and ethnicity is one that has to be squarely faced if we are to make sense of the way in which genes may influence ethnic variability in disease risk. Irrespective of how the term ‘ethnicity’ is interpreted, there is an almost universal presum ption that genetic distinctiveness is somehow involved. This is partly because of a tendency to in terpret cultural differences as if they had a biological basis. Nevertheless, this perception has some validity because many of the processes that lead to the establishm ent of culturally distinct traits also lead to genetic differences. At the most fundam ental level, the social perception of ethnic differences between communities tends to impede gene flow between them . C ultural divergence is usually enhanced by population sub-structure and by isolation be tw een po pu la tion s. T he sam e fac to rs inev itab ly re su lt in gen e tic differentiation. Conversely, those demographic and ecological factors that accelerate, or retard , the rate of genetic drift will also influence the degree of cultural divergence. Certainly, this seems to hold for linguistic divergence. It is also likely to be true for many other cultural attributes, even though the tempo of cultural differentiation m aybe very different from the rate at which genetic differences accumulate (Ward et al. 1991).The factors that influence the rate of genetic differentiation include the degree of isolation among populations, the size of founding populations, the

am ount of dem ographic expansion or contraction they experienced, plus the am ount of contact with surrounding populations. However, the interaction between m utation and drift means that genetic variants are not static over time. The most common and widely distributed genetic variants tend to be oldest, while more recent variants are ra re r and have a more restricted d is tr ib u tio n . As T hom pson and N eel (1996) have em p h asised , th e dem ographic processes that result in the form ation of tribal groups ensure that each different group exhibits its own, characteristic, distribution of specific genetic variants. The most common variants will be shared among many groups, although the frequency will vary from one group to another. Less common variants will be restricted to cultural groups within a given region, and the least frequent variants will tend to be restricted to a single cultural entity. More detailed analyses, based on the coalescent model, also em phasise the prim acy of dem ographic processes in shaping the local distribution of genetic variants (Slatkin and Rannala 1997). Thus, otherwise rare variants can display a markedly elevated frequency within a local group, or a geographic region, w ithout needing to invoke the action of natural selection. Since the social correlates of hum an reproductive behaviour only serve to magnify the effects of genetic drift (Austerlitz and Heyer 1998), the effect of cultural divergence on the extent of local genetic differentiation is likely to be more universal and more profound than is usually realised.In term s of hum an history, the most p ertinen t tim e fram e for these processes is unlikely to extend beyond 20,000 years ago. The demographic and ecological processes w ithin this period of hum an history led to the form ation of local populations tha t are recognisably distinct in both the cultural and genetic sense. More ancient historical events, such as those associated with the early expansions out of Africa, will have had more widespread effects such as the association of common polymorphisms with a regionally defined attribu te, such as a common language family. However, it is critical to recognise the degree of dem ographic relativity embodied in the m easure and in terpretation of genetic differentiation among hum an groups. Over a quarter of a century ago, the detailed analysis of Am erindian tribal groups indicated that allelic variation among villages within a single tribe represented a considerable fraction of the genetic diversity observed at the continental level (Neel and Ward 1970). At the global level, analysis of the existing data indicated that 70 per cent to 85 per cent of hum an genetic diversity was contained w ithin ethnic groups, ra th e r than am ong them (Lewontin 1972). A more recent analysis of mtDNA sequence variation has reaffirm ed these conclusions, by dem onstrating a single Am erindian tribe contains appreciable amounts of molecular variability (Ward et al 1991). Thus, irrespective of how ‘ethnicity’ is defined, genetic differentiation within ethnic groups will be at least as im portant an influence on the genetic distribution of disease risk, as the genetic differences among them . However, since it is precisely the cultural consequences of the dem ographic and sociological history of a population th a t defines its ‘e th n ic ity ’, it also follows th a t

populations that are ethnically distinct are also likely to have their own distinct set of monogenic disorders. Ethnic variation in disease risk: monogenic disease Since populations vary genetically, any contribution that genetic factors make towards risk of disease will result in population-specific differentiation of disease risk. This is most obvious for monogenic diseases, which are inherited as M endelian variants. Even before the advent of the molecular revolution, it was recognised that some M endelian disorders were constrained to a specific population group. Perhaps the most clear-cut exam ple is the geographic distribution of cystic fibrosis. While this is the most frequent recessive disorder in Caucasians of N orthern European extraction, it is relatively uncommon in other major population groups. W ith the increased resolution provided by the molecular dissection of the CFTR gene, it is now possible to be more precise about the genetic basis for the ethnic variation in disease risk. The accum ulated data suggest that spatial variation in the incidence of cystic fibrosis is largely attributable to underlying differences in the frequency of a single m utan t allele, the AF508 m utation. This allele, which may have originated as far back as 40,000 years ago, is common in Europe but rare elsewhere. Further, the relative contribution of the AF508 m utation to cystic fibrosis varies considerably within Europe, accounting for 80 per cent of disease chromosomes in Sweden, but only 50 per cent in Spain. More detailed analysis of m utant chromosomes indicates tha t the geographic variation of this m utation is unlikely to be the result of a simple clinal pa tte rn . In particular, analysis of linkage disequilibrium across the AF508 haplotype indicates that the overall genetic affinities among populations do not account for the relative frequency of this m utation w ithin regional populations (B ertranpetit and Calafell 1996). The situation appears far more complex, with each local population displaying a characteristic frequency of the AF508 m utation, and a slightly different mix of the other m utant alleles.One explanation for the observed distribution of cystic fibrosis is that ancient environm ental variation led to differing selective forces that gave rise to local increases in the frequency of the AF508 allele. These local d ifferences m ight th en have been frozen by a subsequent popu lation expansion, perhaps that associated with the introduction of agriculture. Since the CFTR gene is intim ately involved with regulating the flux of chloride ions, a series of epidemics of diarrhoeal disease represents a plausible selective regime. It is thought that m utant forms of the CFTR gene might reduce the amount of intestinal fluid loss and hence increase survival. Intriguingly, recent in vitro studies indicate that Salmonella typhi, the causative agent for typhoid, requires an intact CFTR product to enter cells efficiently (Pier et al. 1998). This provides a putative selective mechanism to account for the geographically constrained distribution of the AF508 m utation.Hence, the ethnic differences in risk of cystic fibrosis observed among

contem porary populations may represent the consequences of past ethnic differences in socio-demographic structure. Different population sizes and variation in the tim e when agriculture was adopted could have resulted in differing intensities of typhoid epidemics and, hence, a differential selective force on the gene. A strikingly sim ilar geographic pa tte rn is observed for the C282Y m utation in the haemochromatosis gene (M erryweather-Clarke et al. 1997). This m utation, which accounts for about 80 per cent of cases in Caucasian populations, appears to have arisen quite recently (about 2,000 years ago) and spread rapidly throughout north-western Europe. Again, it is tem p tin g to specu la te th a t past epidem ics have played a role in the d is tr ib u tio n of th is d isease , th o u g h th e evidence h e re is fa r m ore circum stantial. However, as noted above (p. 91), the demographic processes that led to the form ation of ethnic groups within Europe are also capable of giving rise to the complex pattern of genetic differentiation observed for these two M endelian disorders.The relationship between ethnic identity and risk of genetic disease is perhaps most clearly seen in the instance of religious isolates. For well over a quarter of a century it has been recognised tha t diseases such as Tay-Sachs disease (TSD) occur in high frequency in Ashkenazi Jews, with one in th irty individuals carrying a TSD m utation. This knowledge has been used to establish effective targeted screening program m es in many U nited States communities, in a way tha t would not have been practical if applied to the general population. However, it would be a mistake to presum e that ethnic distinctiveness in risk of genetic disease is confined to only a small handful of monogenic disorders. W ithin Europe, comprehensive investigation of the distribution of M endelian disorders at the population level indicates that Finns are characterised by an extensive set of genetic diseases that are rare, or absent, from other European populations (Peltonen et al. 1995). Conversely, other genetic diseases, which are common elsewhere in Europe, tend to be absent from the Finnish population. This highlights the fact that definition of a genetic isolate can, in some instances, be com m ensurate with perception of national identity. In the case of the Finns, their striking constellation of unique genetic disorders can be attributed to a m ajor demographic expansion that occurred only a few thousand years ago, but which left an indelible trace in the genetic record (Sajantila et al. 1996). Thus, the socio-demographic processes tha t led to the present-day distribution of the cultural attributes that define the contem porary ‘Finnish’ population also resulted in a distinct pa tte rn of genetic diseases that characterises this population.In all the above instances, the increased frequency of M endelian disease within a particular population group is simply a reflection of the unique evolutionary patterns that shaped that group. Past selective differentials are frequently presum ed to underlie m ajor differences in allele frequencies between groups. The attractiveness of the selective argum ent is in large part due to the indisputable im portance of differential selection in shaping the global distribution of the haemoglobinopathies (Flint et al. 1993), still the

most dram atic example of an ethnically constrained pattern of genetic disease. However, the vagaries of demographic history represent an equally likely explanation. Analysis of tribally specific ‘private polymorphisms’ in South Am erican populations provides a model of how genetic differentiation at the trib a l level, followed by dem ographic expansion, can lead to elevated frequencies of a particular allele within a culturally defined group (Thompson and Neel 1996). Thus, the consequences of demographic expansion in a series of relatively isolated tribal populations, characterised by a high degree of local correlation between uniting gam etes, can lead to variable frequencies of genetic disease among populations. In particular, there is no need to invoke selection as a cause for common disease entities, such as cystic fibrosis (Thompson and Neel 1997). This logic applies with even greater force for ra re r m utations of more recent origin. Hence, in general, the particular history that shaped each ethnic group has also led to the ethnic-specific distribution of monogenic disease. C ultural heritage (which embodies genetic history), not natural selection, was probably the more im portant process in determ ining contem porary patterns of monogenic disease among different ethnic groups. Molecular dissection of Mendelian disease: one disease, many lesions W ith the increased resolution afforded by detailed m olecular analysis of individual loci that cause M endelian disease has come the recognition that ‘single gene disorders’ are not caused by a single m utation, as form erly believed. Instead, each individual M endelian disorder is frequently the result of tens, if not hundreds, of individual m utations. L iterally hundreds of m utations are now known to underlie the occurrence of common recessive disorders: over 850 for cystic fibrosis and in excess of 400 for phenylketonuria (PKU) (Scriver et al. 1996). Similarly, a very large num ber of individually distinct m utations are now known to be involved in the dom inantly inherited susceptibility to familial hypercholesterolaem ia [multiple m utations in the low-density lipoprotein (LDL) receptor gene] or fam ilial b reast cancer (multiple m utations in the BRCA1 and BRCA2 genes).There are two im portant consequences to these observations. First, and most obvious, is that each distinct m utation has the potential to result in a s ligh tly d iffe ren t pheno type. W h e th e r ac ting a lone, in concert w ith ‘background modifier genes’ or in combination with environm ental factors, every individual m utation adds variability to the clinical spectrum of disease. Initially, it was an ticipated th a t clinical prognosis could be clarified by determ ining the specific m olecular lesion. However, the detailed analysis of a large num ber of PKU m utations indicates that reality is not so simple. In general, it is impossible to obtain a definitive relationship between a specific m utation and the clinical course of disease (Scriver and W aters 1999). For PKU, the barriers to defining a clear-cut relationship range from the impact

of the environm ent (am ount of dietary phenylalanine), to the modifying effects of complex m ultifactorial traits (physiology of the blood-brain barrier). The com plexity is increased by the varied m olecular consequences of individual m utations (e.g. their influence on protein folding and hence in vivo protein stability). Nevertheless, despite these complications, increased m olecular charac terisa tion of individual m utations th a t underlie each M endelian disorder is starting to refine the spectrum of disease. Eventually, this will have some predictive value for the individual patient, as well as for populations in which a particular m utation has an elevated frequency.The second consequence is that each m utation necessarily has its own evolutionary history. While less obvious, this is arguably more im portant. In essence, the hundreds of individual m utations that collectively give rise to a monogenic disease like cystic fibrosis represent hundreds of independent evolutionary histories. Each genetic variant has its own time depth, resulting in a population distribution tha t is constrained in inverse proportion to the age of the m utation. Relatively recent m utations will be constrained to local population groups, in analogous fashion to private polymorphisms (Thompson and Neel 1996), and will exhibit a very low frequency on a global basis. Such m utations, of which theory predicts a great number, will tend to be restricted to a single ethnic group and may reach an appreciable frequency within that group. Cogent examples are the characteristic ‘founder’ m utations of the BRCA1 an d BRCA2 genes that are found in the French Canadian (Toning al. 1998) and Ashkenazim (H artge et al. 1999) populations respectively. Older m utations, though exhibiting a higher frequency at the global level and likely to have a broad regional distribution, will be fewer in number. Such m utations will be found in many different, though related, ethnic groups, as is the case for the AF508 m utation in cystic fibrosis (B ertranpetit and Calafell 1996). The net result of these multiple independent evolutionary histories is that each d istin g u ish ab le m u ta tio n will have its own popu la tion -specific distribution: recent m utations largely constrained to an individual ethnic group, older m utations shared by many different ethnic groups.The overall consequence is tha t the m utational spectrum of monogenic disease will differ from one ethnic group to another. To the ex ten t that d iffe ren t m u ta tio n s lead to d is tin g u ish ab le pheno types, th e clinical presentation of each monogenic disease will also tend to be ethnic specific. F urther, as a consequence of th e ir un ique evolu tionary history, m ost m utational lesions will reside on a chromosome, or haplotype, tha t is largely constrained to a single ethnic group. Thus, each ethnic-specific m utation will be surrounded by a unique constellation of polymorphic markers. This implies tha t individual ethnic groups will require a different set of markers to screen for the presence of disease alleles.Failure to recognise the practical implications of the individual evolutionary histories of individual m utations will compromise the provision of efficient g en etic services to ind iv idual e th n ic groups. B ecause m ost d e ta ile d inform ation about M endelian disorders comes from populations of European

ancestry, the provision of genetic services to m inority ethnic groups is considerab ly less cost-effective th an it could be. W hile the p rac tica l consequences of these theoretical consequences are still largely unexplored, there are already discouraging examples where screening for genetic disease, such as familial breast cancer (the BRCA1 and BRCA2 genes) is much less effective than originally anticipated (Devilee 1999). This has even turned out to be the case when screening for three ‘founder’ m utations presum ed to exist in elevated frequencies within Ashkenazi Jews (Hartge et al. 1999). It seems clear that, to develop effective screening programs, detailed knowledge about the aggregation of individual m utations within specific ethnic groups will be more im portant than enum erating the total m utational spectrum. Given the growing tendency for genetic medicine to be integrated into modern health care, failure to recognise the ethnic specific distribution of individual m utations will only add to the existing disadvantages experienced by minority groups. Single genes and more complex ethnic diseases: infectious disease The genetic differentiation that correlates with the social identification of ethnic groups can lead to significant associations between ethnicity and genetic response to infectious disease. The most obvious examples are also the oldest. It is well known that the haemoglobinopathies, which confer some degree of protection against m alaria, tend to be associated with specific ethnic groups. This association reflects the historical geography of past ecosystems within which selection operated. Nonetheless, it results in an ethnic specific distribution of susceptibility and resistance to infection. Conversely, high frequencies of an erstw hile protective allele can exert a significantly deleterious effect on populations tha t inhabit environm ents from which m alaria has been eradicated. As a consequence, there are many developed countries, especially those which have substantial populations derived from the African diaspora, which require the establishm ent of public health and community health program m es to address this problem. The spectrum of different evolutionary histories that led to contem porary populations has resulted in ethnic groups that differ in their response to existing pathogens. These ethnic groups may be burdened with the adverse consequences of an ancient evolutionary struggle between pathogen and host.The same evolutionary processes can lead to ethnic-specific differences with respect to the risk of infectious diseases. This includes the response to totally new pathogens. A telling example is how HIV infection is m odulated by allelic variation at the CCR5 gene. The protein produced by this particular chemokine receptor gene, part of a multi-gene family located on the short arm of chrom osom e 3, happens to be the site by w hich the hum an immunodeficiency virus gains entry into certain target cells (Dragic et al. 1996). Deletion of a thirty-two-nucleotide segm ent alters the structure of

the receptor protein and denies HIV entry. As a result, individuals homozygous for the CCR5-A32 allele are essentially resistan t to HIV infection, while infected heterozygotes display a much slower progression to full blown acquired immune deficiency syndrome (AIDS). Initial surveys indicated that, while the CCR5-A32 allele was absent in African and Asian populations, it was most frequent in northern European populations, especially Ashkenazi Jews (M artinson et al. 1997). More recent surveys have confirmed the absence of this allele from African, Asian and Native American populations. They have also identified a wide geographic cline across north-west Europe into Eurasia. Scandinavian populations exhibit the highest frequencies of around 14 per cent, while M editerranean populations have much lower frequencies of around 5 per cent. In central Asian populations, the frequency diminishes even further to 2-3 per cent (Stephens et al. 1998).The high frequency of the CCR5-A32 m utation in north-western European populations, coupled with the d istribu tion of its associated haplotypes, suggests it occurred a relatively short time ago (2,500-4,500 years ago) then spread rapidly across northern Europe into central Asia (Libert et al. 1998, Stephens et al. 1998). Such a scenario has been in terpreted as the footprint of a strong bout of selection that swept across northern Eurasia some 2,000 years ago. One possibility is an ancient plague, now long extinct. Alternatively, as a lready discussed (p. 96), such a d is tribu tio n m ight sim ply be the consequence of the d ifferen tia l dem ographic expansion of subdivided populations.Irrespective of the specific evolutionary processes that caused the variable distribution of the CCR5-A32 m utation, the outcome is that susceptibility to HIV infection varies by ethnic group. This finding is likely to be repeated for HIV as additional gene products tha t influence pathogenesis are identified. Moreover, these observations indicate a general principle. The im pact of infectious disease, both new and old, will tend to vary from one ethnic group to another as a consequence of their different genetic heritages. Overall, ethnic groups will vary not only in their total burden of infectious disease, but also in the spectrum of factors that influences outcome. In part this is due to the correlation between geography and ethnicity. The form er defines the biological ecosystem that constrains the distribution of pathogens and their vectors. The la tte r defines the community macro-social environment of the community that modulates probability of infection. To this la tte r must also be added the way in which the genetic heritage of the com m unity influences susceptibility and resistance. Multifactorial disease: more complex aetiologies M ultifactorial diseases, such as hypertension or non-insulin-dependent diabetes mellitus (NIDDM), are essentially the consequence of interaction between susceptible genotypes and deleterious environm ents. While the strength of the contribution of each component varies by disease, the essential

fea tu re is th a t adverse com ponents in bo th genetic background and surrounding environm ent are required before m anifestation of disease. Hence, no single factor is either uniquely necessary or uniquely sufficient. This is particularly true of genetic factors, because evolutionary pressure for functional homeostasis, for individuals and populations, necessarily imposes constraints on genetic variation. For complex physiological processes that involve m ultiple in teracting com ponents, only so much variation can be compatible with long-term survival. Environm ental variability is subject to no such constraints. This is especially true for those aspects that are modified by hum an in tervention, such as nu tritional profiles. Consequently, it is frequently the case that adverse environment tips the balance inexorably towards disease. In this respect, the extrem es of environm ental variation can dom inate variation in m ultifactorial disease susceptibility, in much the same way as is true for allelic variation in monogenic disease. Both types of variation represent opposite ends of the aetiological spectrum (Ward 1980).An essential feature of m ultifactorial disease is the multiplicity of factors, which combine to define disease risk. On the environmental side of the ledger, the factors that contribute towards such disease as coronary heart disease, or cancer, are many and varied. In addition, the composition of environmental risk ranges from factors which, singly, are all im portant in defining risk (e.g. cigarette smoking and lung cancer), to factors which, by themselves, exert only a minor influence on risk (e.g. alcohol intake and hypertension). In similar fashion, a m ultiplicity of variant alleles combines to define genetic risk. Ju st as is true for environm ental correlates of risk, individual alleles vary in the deg ree to w hich th ey in flu en ce risk . H owever, un like th e case for environmental factors, there is a definite structure to the pa tte rn of genetic influence on disease susceptibility. This regular relationship, term ed the genetic architecture of disease (Sing and Boerwinkle 1987), has been defined by evolutionary imperatives. For every population, the genetic architecture of disease has been shaped by the evolutionary history of individual alleles within th a t po pu la tion . T his has con sid erab le im p lica tio n for how gen e tic differentiation influences the distribution of disease risk among ethnic groups.The basic princip le , il lu s tra ted in F igure 8.1, is due to an inverse relationship between allele frequency and the severity of allelic effect. Alleles with the g reatest im pact will occur infrequently w ithin the population. Conversely, very common alleles will, by them selves, only cause m inor perturbations in the distribution of disease risk. Although genetic architecture is fundam entally a continuum , it is conceptually useful to identify three components: rare M endelian variants, ‘major genes’ and polygenes. Rare M endelian variants, such as the functional variants at the LDL receptor gene, or the BRCA1 gene, exert a prim ary and clinically significant effect on risk. For exam ple, the excessive LDL cholesterol levels caused by functional m utations of the LDL receptor gene are relatively unaffected by genetic background, or by the im m ediate environm ent. Consequently, fam ilial hypercholesterolaem ia due to LDLR variants segregates as an autosomal

dom inant. Such functionally po ten t allelic varian ts are constrained by evolution to be few in num ber and rare in frequency. Hence, although im p ortan t at the individual and fam ily level, ra re M endelian varian ts contribute little to attributable risk at the population level.At the o ther end of the architectural scale are the true polygenes, which are both common in frequency and ubiquitous in occurrence. While their collective impact on disease risk is large, accounting for upwards of 50 per cent of the additive genetic variance, their individual effects are m inute. Hence, they are essentially undetectable for all practical purposes. However, in the aggregate, genetic variation in polygenes can contribute to variability in risk among ethnic groups, especially when genotype and environmental interactions are considered.‘Major genes’ are both m oderate in effect and in frequency. The impact of allelic segregation at individual ‘major gene’ loci is detectable (unlike the s itu a tio n for polygenes). Also, individual alleles at ‘m ajor gen es’ are sufficiently com m on to exert a m easurable im pact on the population-attributable risk (unlike rare M endelian variants). Accordingly, these genes represent the target towards which much of genetic epidemiology is directed. The effect of M endelian segregation of the three common apolipoprotein E alleles, which can account for upwards of 5 per cent of total variation in cho lestero l levels (Sing and Davignon 1985) rep resen ts an early and informative example of such genes. The fact that the same apolipoprotein E alleles also exert a significant effect on susceptibility to A lzheim er’s disease serves notice that genetic architecture is m ultifaceted. Allelic variation at a single locus can influence distribution of risk for more than one disease.In Figure 8.1, the num ber of alleles is indicated by rectangles in each column, and allele frequency by the height of each rectangle. Further, for illustration, allele frequencies have been rescaled for each category of genic effect. In reality, the frequencies of rare M endelian variants are vanishingly small, while allele frequencies for ‘major genes’ and polygenes will tend to be polymorphic. In Figure 8.1, rare M endelian variants are represented by five alleles with substantial effects, one more frequent than the others. Segregation of seventeen alleles defines the cumulative im pact of the ‘major genes’, while th irty-three alleles contribute to the polygenic component. It will be noted that this is an abstraction since, in reality, the num ber of relevant alleles and their frequency varies by disease. Every disease has its own genetic architecture. It should also be stressed that the emphasis is on alleles ra ther than loci. Although many loci will be essentially diallelic, so that the locus effect is mainly defined by segregation of a single allele, other loci will have more than two segregating alleles. In the case of apolipoprotein E, LDL cholesterol levels in most populations are influenced by the segregation of three polymorphic alleles. In general, the insights obtained by molecular dissection of monogenic loci will undoubtedly apply to m ultifactorial disease. Genetic risk will be determ ined by a spectrum of molecular variants at each

102 Health and ethnicity

relevant locus, with every variant differing in term s of its population frequency and m agnitude of effect.The role of environm ental factors, which can range from a modifying one to a dom inant one, represents an added complexity. In the classical model, the jo in t contribution of genes and environment is usually represented by the equation: P = G + E + G x E , with P representing the phenotype (in this case, risk of disease), G genetic factors, E environm ental factors and G X E the interaction between genetic and environmental factors. This simple linear relationship indicates how variability in disease risk is collectively influenced by genetic variation within the population, plus environmental variability a n d the interaction between environm ental and genetic factors. Hence, within a single population, the distribution of disease risk can be extrem ely complex. The conceptual complexity is increased when it is realised that environm ental variation impacts on each component of the genetic architecture. Although somewhat tedious, it would be more insightful if disease risk were represented as a set of th ree of the above equations, one for each genetic category (polygenes, ‘m ajor genes’ and M endelian varian ts). The m ultiplicity of

geno type-env ironm en ta l in terac tio ns w ith in each category also needs emphasis. Not only does each allele represented in Figure 8.1 have its own distribution within the population but also a unique set of environm ental interactions.The complex interactions between the m ultiple factors tha t contribute to risk of disease make identification of individual genetic risk factors a daunting task. The situation is even more difficult when ethnic groups are considered. As we have seen, the unique history of every ethnic group leads to considerable variability in the causal spectrum for monogenic disease. The aetiological complexity of m ultifactorial disease makes the distribution among ethnic groups infinitely more difficult to dissect. In large part, this is because ethnic groups are characterised not only by a unique constellation of alleles, but also by a unique set of environments. Hence, the aetiological underpinnings of m ultifactorial disease are likely to vary much more among ethnic groups than is commonly supposed.In spite of these theoretical considerations, there is a common tendency to attribu te ethnic differences in the risk of m ultifactorial disease to only a sing le , usu ally in tr in s ic , factor. T his has th e u n fo rtu n a te effect of oversim plify ing a com plex m u ltid im en sio n a l p rob lem and reduc ing understanding. The attem pts to explain the difference in hypertension risk between populations of African ancestry and those of European extraction are a case in point. For many years there has been a general, but unjustified, tendency to attribute the epidemiological differences in hypertension observed between U nited States Black people and U nited States W hite people to genetic factors. Differences in hypertension rates between United States Black people and W hite people were first noted in the 1930s. A quarter of a century la ter, the s trik in g m agn itu de of th e d ifference in risk was carefu lly docum ented by a, now classic, series of studies in the rural South (Comstock 1957, McDonough et al. 1964). These data made it abundantly clear that rates of hypertension were up to two to four times greater in U nited States B lack people th a n in th e ir W h ite c o u n te rp a r ts . A b road v a rie ty of environmental factors were implicated, some unique to the Black community, some not. However, the overriding consensus was that much of the excess risk of hypertension in the Black com m unity was due to th e ir genetic distinctiveness.D esp ite some d issen t, the invocation of genetic d ifferences as the underlying cause of this major public health problem rem ains implicit in much of the epidemiological literature. This focus on genetic influence is so pervasive that it almost constitutes bias. H ere, it is relevant to note that, since the first surveys in the U nited States, nearly 70 years ago, the risk ratio has undergone significant change. In the 1950s and 1960s, when national rates of hypertension in the U nited States were extrem ely high, the risk ratio (for women) was as high as 4.4. Thirty years later, when the national prevalence was appreciably lower, the risk ratio had diminished to 1.8. Looking ahead, it appears likely that the risk ratio may increase again, since while

rates of hypertension will continue their decline within the W hite population, the decline may slow, or become reversed, in the Black community. Even though the m agnitude of the risk differential between US Black people and W hite people has fluctuated considerably during the past 70 years, the emphasis on genetic causation has rem ained constant. This is unexpected since, as noted earlier (p. 89), substantial tem poral changes in disease risk implicate extrinsic, ra ther than intrinsic, factors. Most recently, the emphasis on genetic determ inism has acquired new trappings. This is in the form of a hypothesis th a t US Black people are un iquely suscep tib le to risk of hypertension because the ancestral slave population was subject to extrem e selection for salt retention during the infamous ‘middle passage’ (Wilson and G rim 1991). A lthough dressed up in evolutionary principles, and undoubtedly proposed in good faith, this supposition has little credibility. N either the genetic model, nor the historical evidence, holds up under careful scrutiny. However, even though this latest theory has fallen into disrepute, the tendency to attribu te this particular ethnic difference in disease risk to genetic causation still remains.Much of this inappropriate emphasis on genetic causation might never have occurred had more note been taken of data from Africa itself. Even before the first docum ented study of elevated risk of hypertension among US Black people, data existed indicating that Africans in Africa experienced l i t t le h y p e rte n s io n (D on niso n 1929). At th e tim e w hen ex ten siv e epidemiological studies were defining the differential risk of high blood pressure between US Black people and W hite people, a variety of studies within Africa were showing a different picture. The results of those later studies pointed to environment, not genetics, as the more pertinen t factor. In rural Africa, blood pressure was low and hypertension rare, especially in the more traditional communities (Akinkugbe and Ojo 1969). Moreover, when migration to urban centres occurred rates of hypertension increased markedly, as shown by Scotch’s (1963) classic analysis of Zulus. Thirty years later, the risk differential between urban Zulus and rural Zulus rem ained notable, with the transition from a rural to an urban lifestyle being m arked by a decrease in physical activity and an increase in obesity. More recently, yet another analysis of a population that is making the transition between a traditional lifestyle to an urbanised way of life, the Luo, shows the same characteristic risk differential for hypertension (Poulter et a i, 1990).If env ironm ent was so obviously the critica l factor in de te rm in in g differential rates of hypertension within Africa, why was it possible to m aintain the belief that genetic difference was the predom inant cause of the risk differential between US Black people and W hite people? The answer is unclear, though it is likely that the debate will continue unabated for some time to come. However, the recent publication of a critical piece of evidence places the weight of evidence even more squarely in the environmental camp. The data, summarised in Figure 8.2, derive from the first careful comparative analysis of blood pressure and hypertension ra te s am ong a series of

populations of African ancestry, ranging from West Africa to the United States, via the Caribbean. Seven populations of West African origin were extensively surveyed, using the same methodology, with careful a tten tion to ensure comparable evaluation of potential risk factors. The results showed a clear-cut trend in the prevalence of hypertension with the three West African populations exhibiting the lowest rates and the US population the highest. The three Caribbean populations had in term ediate prevalence rates. These resu lts e lim inate any residual concern th a t previous in te rco n tin en ta l comparisons might have been confounded by lack of com parability between d iffe re n t in d ep en d en t surveys. L ocation , no t genes, is th e p rim ary discriminator.The details are even more revealing. First, and most obvious, overall rates of hypertension are clearly associated w ith overall ra tes of obesity [as m easured by body mass index (BMI)]. Here, BMI is used as a surrogate for the degree to which the population has adopted a ‘W esternised’ lifestyle with, among other aspects, increased caloric intake and reduced physical activity. The West African populations with the lowest rates of hypertension are also the leanest. Conversely, the African American community of Maywood, a suburb of Chicago, has twice the overall rate of hypertension and a significant excess of obesity. Overall, it is clear tha t the substantial differences between geographic regions are attributable to the differences in lifestyle factors that m anifest themselves as differences in BMI.Second, and perhaps even more telling, is the distribution within West Africa. It is notable that the difference in hypertension rates between rural

Nigeria (14.5 per cent) and urban Cameroon (19.1 per cent) is an appreciable fraction of the overall difference between West Africa (15.6 per cent) and the Caribbean (25.5 per cent). The ru ra l-u rban divide within West Africa rep resen ts a considerable change in the m acro-social environm ent, as indicated in the difference in the distribution of BMI (22.2 versus 26.1). This results in a 20 per cent increase in hypertension from rural Nigeria to urban Cam eroon. W ithin the C aribbean, the range in BMI values, from 25.7 (Jamaica) to 27.7 (Barbados), is much less, with a m ean not much less than the value for the African American sample (28.9). This suggests a greater degree of sim ilarity in lifestyle factors outside Africa than within Africa. Risk differentials for hypertension reflect this, with a 20 per cent increase in risk within Africa, which equals the 20 per cent increase from the Caribbean to the U nited States and exceeds the 20 per cent risk differential within the Caribbean. The im portance of the macro-social environment is highlighted by the 63 per cent risk differential between West Africa and the Caribbean.Taken together with the earlier studies, these results indicate that adverse environments, ra ther than deleterious genes, are the prim ary cause of the difference in hypertension rates among the populations of African ancestry. However, it also needs to be clarified that an absence of genetic causation to account for differences among ethnic groups does not m ean an absence of ethnic-specific genes that give raise to ethnic-specific risk pathways. While this may seem paradoxical, the in terpretation is clarified by considering the consequences of genetic differentiation for the genetic causation of complex disease. Influence of genetic differentiation on genetic architecture of disease Just as genetic differentiation results in a variable distribution of risk for monogenic disease among ethnic groups, the same is true for more complex m ultifactorial disease. The im pact of population-specific variation in allele frequencies on the genetic architecture of disease is illustrated in Figure 8.3. As a consequence of their distinct evolutionary histories, both populations A and B have a different spectrum of alleles that contributes to disease risk. This is plotted on the ‘species-specific’ architecture illustrated in Figure 8.1. Alleles th a t are present in a population are represen ted by the shaded rectangles, while empty rectangles indicate alleles that are absent from one or o ther of the population. A num ber of consequences are im m ediately apparent. Most obvious is the fact tha t neither of the single populations contains all the relevant risk alleles. For example, instead of the five alleles of substantial effect that segregate as M endelian variants, each population considered alone has only three (60 per cent of the total). The same is true for the ‘major gene’ and polygene categories, though to an increasingly lesser extent since the fraction of missing alleles declines in inverse proportion to allele frequency. More noteworthy, and more cogent for our argum ent, is the

even sm alle r p ro p o rtio n of a lle les th a t are shared betw een the two populations. Since the probability tha t alleles are shared among populations is proportional to their evolutionary age (Thompson and Neel 1996), the rare (and evolutionary young) alleles of substantial effect will tend to be unique to individual populations while the more common polygenes will exhibit a much greater degree of sharing. As indicated in Figure 8.3, within each category of genic effect, there is a general tendency for the more frequent alleles to be shared am ong populations, while ra re r alleles tend to be population specific in their distribution.Figure 8.3 clearly indicates tha t the distribution of risk alleles will differ among populations. Hence, the distribution of genetic risk will differ among ethnic groups both in term s of which loci are involved and in the m agnitude

of effect associated with specific alleles. The main implication for the ethnic d istribu tio n of com plex, m ultifacto ria l disease is th a t the d ifferen tia l distribution of allele frequencies for ‘major genes’ will dom inate the ethnic-specific differences in genetic risk. Although differences in allelic variants that define M endelian traits will be most marked, their overall rarity means they contribute little to differences in population risk. Conversely the majority of alleles at polygenic loci will have an ancient origin and will be widely dispersed among a broad variety of ethnically distinct populations. This leaves little opportunity for substantial differences among ethnic groups, and implies that the additive genetic background will be similar for most ethnic groups. In contrast, ‘major gene’ alleles, which individually contribute substantially to population-attributable risk, will tend to have an interm ediate distribution of m utational ages and hence considerable variability in allelic frequency among ethnic groups (Figure 8.4).W hen evaluating the distribution of disease risk among ethnic groups, it needs to be emphasised that not only will the set of relevant alleles vary, but so will the relevant environments. Moreover, the extent to which im portant environm ental changes have occurred is also likely to differ by ethnic group. Thus, the intersection of genetic differentiation and environm ental change can make an im portant contribution to ethnic differences in disease frequency. W ith genetic d ifferentiation, the genotype by environm ental in teraction will necessarily vary by population . If the environm ent also varies am ong populations, which is usually the case, the interaction between genes and environm ents will be unique to each ethnic group. To the ex ten t th a t ethnically distinct groups have a characteristic environm ental milieu (which

is usually the case), the spectrum of genotype by environment interactions will likewise be increased across ethnic groups. Conclusion While exceedingly preliminary, the available genetic data emphasise that the variable risk of complex genetic disease among different ethnic groups cannot be a ttribu ted to a small num ber of independently acting discrete causes. Hypertension, like many other diseases of complex aetiology, varies appreciably in frequency among ethnic groups. W hen an ethnic group displays an elevated frequency, the public health consequences can be profound, as illustrated by the situation for African Americans in the U nited States. As a group, African Americans are readily characterised in term s of their excessive risk of hypertension and consequent morbidity and mortality. However, it is far less easy to discern the underlying reasons. It is clear th a t to fall back on ‘ethnicity’ as an explanation is at best a profession of woeful ignorance and at worst racist. N either genetic background nor cultural milieu is sufficient to account for the risk of disease. Among populations of African ancestry, elevated frequency of a putative-risk allele does not predict a population-specific risk of disease. This appears true both in the low-risk environment of rura l West Africa and the high-risk environment of urban U nited States. At the level of the individual, presence of the 235T variant exerts a physiological effect in all populations, but the pathophysiological implications are either negligible or irrelevant. Why populations of African ancestry should differ so m arkedly from C au casian popu lations is an im p o rtan t challenge for epidemiology.U ltim ately, it has to be recognised th a t ‘e th n ic ity ’ resu lts from an irreducible mix of biological, social and cultural factors. Similarly, it is the equally complex in teraction betw een genetic factors and environm ental variables that leads to ethnic-specific disease profiles. Just as no single variable can provide an adequate identifier for ‘ethnic affiliation’, neither genes nor environm ent, considered in iso lation , can provide an adequate causal explanation for the differences in disease frequency among ethnic groups. To a ttribu te population differences in disease risk to underlying differences in the frequency of a ‘susceptibility gene’ is as meaningful as attributing differences in IQ to skin colour. It is almost as harmful. The tendency to seize on sim ple, bu t inadequate, explanations tends to forestall serious research into disease causation. It can also result in appropriately stigmatising the group. The way forward is to recognise tha t disease is the result of a complex interaction between genes and environments. This is particularly relevant for diseases that display m arked differences in frequency, or severity, among ethnic groups. A corollary of this principle is that the profile of disease risk that may appear to indelibly characterise an ethnic group is, in fact, no more perm anent than any single cultural attribute that appears to distinguish the group. Even if the genetic heritage of a group is slow to change over time, any change in the physical or macro-social environment can lead to the

relatively rapid disappearance, or appearance, of disease. W itness the rise in the prevalence of diabetes in certain Amerindian and Micronesian populations in the space of less than two generations.In conclusion, although ethnic associations with disease are often used as a descriptive label, this is really a confession of ignorance. Ethnic differences in disease risk result from a complex mix of genetic differentiation and ecological factors. To conclude that a specifiable percentage of the variance in disease risk can be attributable to ethnic factors explains nothing. Use of such labels merely identifies an association, which, if studied further, may ultim ately lead to a deeper understanding of disease aetiology. Taken in isolation, the label is meaningless. Even worse, undue reliance on ‘ethnicity’ as a risk factor not only suppresses understanding but also increases the possibility tha t racist in terpretations of disease risk will prevail.Nonetheless, to disregard the ethnic variability in disease risk is equally misleading and can be ju st as deleterious. As this chapter indicates, ethnic variation in disease risk is a fundam ental characteristic of our species. Fundamentally, it is a consequence of the distinct biological history and cultural history that, collectively, defines every ethnic group. To subscribe to the convenient fiction that the pathophysiology of complex disease, such as hypertension and diabetes, is invariant across ethnic boundaries represents a disservice to each ethnic group on two counts. In the first instance, it obstructs science by constraining our ability to develop a deeper and more comprehensive understanding of the full range of pathophysiology causing hum an disease. Ultim ately, failure to com prehend the gam ut of patho­physiology that leads to disease can only impede effective treatm ent. Second, and at a more im m ediate level, it inhibits our ability to develop appropriate and effective s tra teg ies for detec ting and trea tin g com plex disease in ethnically diverse populations. In the final analysis, a reluctance to recognise the genetic (and socio-ecological) correlates of ethnic diversity in disease risk harm s the very ethnic groups th a t the presum ption of aetiological uniformity is designed to protect. N either science, nor ethics would be well served by such a stance. References A k in k u g b e , O .O . and O jo , O .A . (1 9 6 9 ) A r te r ia l p r e ssu r e in ru ra l and u rb an