Ethnicity and the risk of cancer
Introduction Ever since in ternational data on cancer incidence and m ortality have been available, it has been clear that there are large differences in the risk of specific cancers in d ifferen t populations. M uir (1996) cites the early observations of Hoffman (1915), drawing atten tion to the tenfold difference in m ortality from cancer of the breast between Japanese and British women. If the prim ary variable of in terest is ethnicity, or racial group, the differences observed within the same geographic locality are more meaningful since at least some of the env ironm en tal differences p resen t in in te rn a tio n a l comparisons are reduced or eliminated. There are plenty of examples of such studies from m ulti-ethnic populations in all parts of the world - for example the W hite and Black populations of H arare, Zimbabwe (Bassett et al. 1995), the Chinese, Indian and Malay populations of Singapore (Lee et al. 1988) and, above all, studies from the U nited States (Miller et al. 1996) (Figure 13.1).The simple comparison of age-standardised incidence rates by cancer site may disguise some im portant differences between ethnic groups which vary by age group, or histological sub-type of cancer, pointing out the im portance of different aetiological factors, known or unknown, related in some way to race or ethnicity. Thus, the Black-W hite difference in the incidence of breast cancer in the U nited States varies with age, with higher incidence in premenopausal Black women and higher incidence in W hite women after the menopause (Figure 13.2). The much higher incidence of oesophageal cancer in Black men than in W hite men in the United States is confined to squamous cell carcinoma; for adenocarcinom a the reverse is found. What accounts for ethnic variation in risk? Valid comparisons can only be made using valid data, and, although this seems a banal point to make, there are many instances of uncritical use of poor-quality d a ta in com parative studies. Thus, in particular, in te rn a tio n a l m ortality statistics are of very variable quality, and care should be taken in their selection. Even within a single country, it is possible tha t differential
access to health care and diagnostic services by ethnic group may influence reporting rates of disease, though this is unlikely to be a major problem for cancer. A possible exception is in screening programmes. Screening may result in increases in reported incidence of disease, because of detection of latent cancers that would otherwise never have surfaced clinically. This is a particular problem with screening for cancer of the prostate (Brawley 1997) and for neuroblastom a (Bessho 1996), and it probably accounts too for part of the recorded increases in breast cancer incidence in some countries (Quinn and Allen 1995, G arnzetal. 1997). Certainly, access to and acceptance of screening program m es has been shown to differ by ethnic group in several countries (Seow et al. 1997, Parker et al. 1998).Differences in access to trea tm ent certainly can affect outcome, so that survival rates from cancer are well known to vary by race or ethnicity (Baquet and Ringen 1986). Since m ortality rates are determ ined by both incidence of disease and survival, this is a major consideration if m ortality is being used, as it often is unconsciously, to provide inform ation on cancer risk. From the data in Table 13.1, it is clear that differences of 50 per cent or more in mortality rates between the ethnic groups m ight easily occur as a consequence of survival differentials, ra ther than differences in incidence. From a public health point of view, statistics on num bers of deaths, person-years of life lost, etc., may be the relevant m easure to use. T here is also a considerable literature on the reasons for observed differences in survival between ethnic groups - do they represent variation in stage at presentation, tum our sub-types, host vulnerability, treatm ent received or response to it (Howard et al. 1992)? However, these do not concern us in this chapter, where the focus is on differences in the risk o f disease by race or ethnicity, so we have chosen to concentrate upon statistics on cancer incidence.A rtefact aside, the principal question posed by observed in ter-ethn ic differences in risk is how much is due to variation in exposure (to ‘carcinogens’ or ‘risk factors’) and how much is the resu lt of in heren t differences in susceptibility to such exposures (and hence genetically determ ined). Table 13.1 F ive-year re lative survival rate (%) by site and racial or eth n ic group, U n ited
From an epidemiological point of view, the variable ‘ethnicity’ or ‘race’ defines a constellation of genetic factors, which relate to susceptibility to a given cancer. O f course, there is considerable variation within a given ethnic or racial group (however this is defined), but there are often sufficiently large differences between them to yield distinctive patterns of risk. I f ‘ethnicity’ is the variable of in terest, the first consideration is to elim inate the effect of confounding variables, associated with the risk of disease and differentially distributed by ethnic group. As usual, such ‘confounders’ can be considered at several levels - so-called dem ographic variables such as social class (or occupation, educational level), place of residence, m arital status, and so on, or more defined exposures such as tobacco, alcohol, diet, infection for which the dem ographic variables are themselves proxies.In general, two approaches have been used to exam ine the relative contributions of ethnicity per se (genetic predisposition) and environm ental exposures in determ ining risk.