-
Geographic Ancestry and Mortality from Ischemic Heart Disease: Evidence from the Finnish Population Register
This paper uses the Finnish longitudinal population register to study mortality risks from ischemic heart disease (IHD) for people aged 65–79 years in the period 1981–2004. In spite of substantial reductions in IHD mortality over time, regional differences within Finland have remained practically the same. Having controlled for a number of significant socio-demographic factors, we find geographic ancestry as proxied by people’s birth region and ethnic group to be a strong predictor of IHD mortality. This interrelation is further confirmed by observing death rates in Finns who live abroad. Reasons behind the mortality variation should be sought in circumstances that operate before or at early childhood. In light of recent advances in medical and genetic research, and the population development of Finland, there are strong reasons to believe that patterns observed in these large-scale population data reflect geographic clustering of genetic susceptibility for IHD.
Geographic ancestry, IHD mortality, hereditary factors, geographic clustering, population data, Finland
Introduction
Due to improved medication, prevention and surgical treatment, mortality from ischemic heart disease (IHD) has decreased notably during the past decades in Finland (Figure 1). The reduction corresponds with the development in most other industrialised countries, but from an international perspective IHD mortality in Finland is still relatively high. Of all deaths in people aged 65–79 years in the country, IHD mortality accounts for over 25 per cent in men and roughly 20 per cent in women. This unique disease, which is characterised as heart problems induced by narrowed heart arteries, is hereby the single most important cause of death in these ages (see Table A1 in the Appendix).
Remarkable for Finland is that, in spite of the large reduction in IHD mortality over time, relative differences across regions have remained practically the same. People in Eastern and Northern parts of the country have 10–30 per cent higher risks of dying from IHD than those in the Western parts (Figure 2). The regional variation in all-cause mortality is also heavily driven by IHD mortality (Koskinen and Martelin 1998, 2003).
There were considerable population shifts in terms of internal migration in Finland during the 20th century. Previous demographic studies based on Finnish population register data have found that people’s birth region was a much more decisive determinant of IHD mortality than their current region of residence (Valkonen 1987; Koskinen 1994). These analyses were concerned with ages 35–64 years, and covered the period 1971–1985. Since there has been a rapid decline in IHD mortality rates since then, there are quite few deaths from IHD in the current working-aged population. There is consequently reason to undertake analyses for a later period and to put focus on somewhat higher ages, as we will do in this paper.
In principal, reasons behind regional mortality differentials can be traced to four potential phases of a person’s life course: at conception through genes, in the uterus, during childhood and adolescence, or at adult ages via environmental and behavioural factors (cf. Ben-Shlomo and Kuh 2002; Kuh et al. 2003; Kuh and Ben-Shlomo 2004; Doblhammer 2004; Lynch and Davey Smith 2005). The determinants of IHD have traditionally been sought for in standard risk factors as measured by lifestyle and environment, but genome-wide studies of later date suggest that it is caused also by genetic susceptibility (Watkins and Farrall 2006; WTCCC, 2007).
Data used in standard demographic research contain no information about genetic markers or about health related behaviours. With ordinary population register data it is therefore not possible to explicitly verify genetic influences on health and mortality. The previous demographic studies on regional differences in IHD mortality in Finland treated the role of hereditary factors quite unfairly. Nowadays, a more common opinion still seems to be that the geographic variation cannot be considered primarily attributed to area-specific environmental or behavioural factors (Koskinen and Martelin 2007).
As we argue in the subsequent section of the paper, there are sound reasons to assume that in the specific case of Finland population development may explain why persons’ birth region impacts on their IHD mortality risk. Our aim is hereby to use the large-scale data in the Finnish population register to illustrate that geographic clustering of genetic characteristics may underlie the interrelation between people’s birth region and their IHD mortality risks.
We analyse people aged 65–79 years using data from 1981–2004. An obvious advantage of focusing on older people is that genetic influences on mortality presumably are larger at higher ages than at lower ages (cf. Hjelmborg et al. 2006). Region-of-birth effects also provide a stronger argument for hereditary factors if they can be observed at higher ages, as most moves have been undertaken at young adulthood (cf. Christensen, Johnson, and Waupel 2006).
Background
There is no conclusive historical or archaeological evidence on how Finland was inhabited. One general view is that about 2,000 years ago the country was populated from three main directions: the East (Ural in present Russia), the South (the present Baltic countries), and the Southwest (present Sweden). Human contact across larger spatial areas was thereafter largely prevented by large distances, languages, dialects, and political boundaries, which are circumstances known to reduce probabilities for random mating and induce geographically related genetic clustering (Cann 2001; Bamshad, 2006). Individuals from more proximate regions consequently share more recent common ancestry than those from less proximate regions simply because alleles that vary geographically tend to correlate (Risch, Burchard, Ziv, and Tang 2002; Rosenberg et al. 2005). Such multi-locus allele clustering may explain why specific diseases are geographically concentrated. A number of international genetic studies also demonstrate that, even within national populations, stratification by geographic ancestry can underlie the inheritance of complex genetic diseases (Parra et al. 2003; Helgason et al. 2005; Campbell et al. 2005; Seldin et al. 2006; Price et al. 2008). In Finland, a number of diseases have been found to be concentrated in small spatial areas—presumably caused by a family-history of the disease (Norio 2003).
Due to Finland’s geographic isolation, there has been very little immigration of foreign-born people to the country in modern time. During the past century internal migration has been intense and primarily in one direction: from the East and the North towards the West (Southern and Southwestern Finland). The population shift implicates that the current population in Eastern and Northern parts of the country consist of genetically different and more homogeneous groups than those in Western Finland.
With regard to Finland, a person’s birth region can consequently be used as a rough proxy of her geographic ancestry. These beneficial underpinnings for using large-scale population data to reflect an interrelation between geographic ancestry and cause-specific mortality are in an international perspective quite unusual. Belonging to the Swedish-speaking ethnic minority reflects an additional such dimension. Swedish speakers in Finland, who have markedly lower IHD mortality rates than Finnish speakers, live concentrated on the Southern and Western coastline and have very low internal migration rates. Nowadays, they account for barely six per cent of the total population.
Since susceptibility for IHD among Finns is increasing with the number of parents and grandparents that come from high-risk areas (Juonala et al. 2005), the argument that some latent behavioural factors adopted in childhood or youth, or in-uterus experiences, would exert an irreversible effect on the adult mortality risk (cf. Barker 1999, 2004) is, in our opinion, an inadequate explanation as to why geographic ancestry impacts IHD prevalence. This requires that area-specific health behaviour, also of persons who migrate, would be passed on to subsequent generations.
Numerous genetic mappings and medical research findings, on the other hand, suggest that hereditary factors which influence biological risk factors for IHD follow a geographical pattern in Finland (Pesonen et al. 1986; Guglielmino, Piazza, Menozzi, and Cavalli-Sforza 1990; Lehtimäki et al. 1990; Aalto-Setälä et al. 1991; Virtaranta-Knowles, Sistonen and Nevanlinna 1991; Kittles et al. 1998; Carracedo et al. 2001; Raitio et al. 2001; Jartti et al. 2002; Juonala et al. 2004, 2005; Juonala, Viikari, and Raitakari 2006). The overall conclusion is that genetic predisposal for IHD is higher in Finns who originate in the Eastern parts of the country than in those who come from the Western parts.
Since our study is concerned with higher ages, and the register used contains information since 1970, we cannot explicitly account for circumstances at childhood or adolescence. Similar research conducted on the working-aged population (Saarela and Finnäs 2009), however, finds that variation in cause-specific mortality by geographic ancestry is only marginally reduced when proxies for environmental and behavioural factors at both childhood and adulthood are accounted for, in spite that these socio-economic and social variables have strong effects on mortality. Other register-based studies conducted on adolescents and on people aged 60+ reached similar conclusions (Saarela and Finnäs 2008, 2009).
In light of the previous research findings and the population development of Finland, we therefore consider it very likely that region-of-birth effects primarily reflect geographic clustering of hereditary factors.
Data
The data come from the longitudinal population census file compiled by Statistics Finland. The data consist of individual-level information from the population censuses of 1970, 1975, 1980, 1985, 1990, 1995, and 2000 (Statistics Finland 2008b). For all persons who died during the period 1971–2004, the file was complemented with information about the year of death and main cause of death according to ten main categories. The emphasis here is on mortality by IHD. But for the sake of comparison and contrast, we also report death risks for other cardiovascular diseases, all other diseases, and external causes of death.
The file we have access to contains information on each person’s sex, year of birth, birth region, current region of residence, and several socio-demographic variables that are presented below. These data constitute a five per cent random sample of the total population of Finland, and an identically constructed 50 per cent random sample of the Swedish-speaking ethnic minority in the country. Because the population register includes a variable that determines a person’s unique mother tongue, Swedish speakers can be separated from Finnish speakers.
Our focus is on the role of region—specifically the relative importance of birth region and current region of residence. In accordance with recent studies of regional mortality differentials in Finland (Saarela and Finnäs 2008, 2009), and since we also aim to compare Swedish speakers in Finland with Finnish speakers, we have constructed new variables that combine region and ethnic group. Constructing these new variables was necessary because Swedish speakers live concentrated along the coastline.
The geographic categorisation can be seen from the map in Figure 3. The country has been divided into three main regions: Western Finland, Eastern Finland and Northern Finland. In addition, we separate the coastal area and the Helsinki metropolitan area, which constitute the Swedish speakers’ main settlement area (containing approximately 95 per cent of all Swedish speakers in the country). The Helsinki area needs a separate category as death rates there are known to be relatively high. In these two regions, the population is consequently classified according to whether the persons are Swedish speakers or Finnish speakers, whereas for the other three regions only Finnish speakers are included. Almost 90 per cent of the persons who originate in Eastern and Northern Finland and had moved to Western Finland, had lived in Western Finland for more than 20 years.
The data design implies that we follow individuals over five-year periods. For each individual, we consequently have two potential observation periods. Age at the start of each period is restricted to 65–74 years, implicating that we analyse death risks in the age interval 65–79 years for the cohorts born 1911–1930. The Lexis diagram in Figure 4 illustrates the observation plan.
The strong effects of social class on individuals’ death risks are well known (Pensola 2003). Since the people analysed here have reached the official age of retirement, 65 years, we make use of the longitudinal nature of the data and use information about their socio-economic status at active working age. In Finland, the mean age of actual retirement has for many years been under age 60. To measure socio-economic status accurately for people born 1916–1930 we therefore had to go back to the situation at age 50–54 years. For people born 1911–1915, it was possible to measure socio-economic status at age 55–59 years, because at end-1970 the proportion of retired people under age 60 was very small as compared with the later census years. The variable has the categories (1) Blue-collar worker, (2) White-collar worker, (3) Self-employed, (4) Farmer, and (5) Economically inactive.
An additional indicator of socio-economic position within social class is homeownership. As much as three quarters of all households in Finland own their accommodation, and people in this category have been found to have substantially lower death risks than rent-takers (Saarela and Finnäs 2009). This variable is measured at age 55–59 years.
Marriage is known to be highly selective with regard to health and mortality, and marriage dissolutions—both in terms of divorce and death of a spouse—have strong effects on mortality risks (Martikainen and Valkonen, 1996). Marital status as measured here does not, therefore, simply reflect the current situation. It also aims to account for immediate, and less immediate, effects of marriage dissolution. The variable has four mutually exclusive categories (1) Married or remarried persons, (2) Never married persons, (3) Previously married persons whose marriage ended during the most recent five-year period due to divorce or death of the spouse, and (4) Previously married persons whose marriage ended before that.
The schooling variable refers to the highest level of educational attainment. It is measured at the beginning of each observation period and consists of the categories (1) Primary, (2) Secondary, and (3) Tertiary level of education. As much as 80 per cent of all people in the data have primary (basic) level of education only. Naturally, the variable correlates highly with socio-economic status; it is primarily used to capture any potential effects of social class within socio-economic status.
The variable distributions are presented in Table A2 in the Appendix.
Results
To understand the fundamentals of regional mortality variation in Finland it is necessary to simultaneously observe persons’ current region of residence and birth region. This approach implicates that in practice one must restrict data to regions with Finnish-speaking population only (Western, Eastern, and Northern Finland), as Swedish speakers in Finland live geographically concentrated (along the coastline) and have very low internal migration rates.
A good first apprehension of the relative importance of the region variables can be gained by observing the estimates from main effects models. We therefore begin with presenting the results of models that restrict the data to Western, Eastern and Northern Finland, and for which separate regressions are estimated when including region of residence and birth region, respectively.
These models account for age and time period, and are fitted separately by sex for IHD mortality and each of the other three main causes of death. Social class and social competence, which indirectly affect mortality risks, are accounted for by the variables educational attainment, socio-economic status, homeownership, and marital status. The numbers reported, in Table 1 and Table 2, give risk ratios with 95 per cent confidence intervals. The Wald statistic refers to tests that determine the contribution of each of the two region variables to model improvement. This test statistic has two degrees of freedom and is asymptotically Chi-square distributed.
For men there are evident regional differences in IHD mortality and mortality by external causes, but not in deaths by other cardiovascular diseases and other diseases. For IHD mortality, birth region is also a much more decisive determinant than current region of residence. Similar conclusions apply to women, for whom we, to some extent, can observe even larger regional differences. As an example, IHD mortality is 25 per cent higher for men born in Eastern Finland, and 12 per cent higher for men born in Northern Finland, than for those born in Western Finland. Corresponding numbers for women are 17 and 56 per cent, respectively.
The estimates also reveal that IHD mortality is notably higher for never-married persons and those with marriage dissolutions who have not remarried than for married or remarried persons. The relative differences are generally higher in men than in women.
Since we account for both educational level and socio-economic status, the estimated effects are relatively small and many of the parameters therefore not statistically significant. The estimates still suggest that IHD mortality falls with rising educational level, and people with poor labour market attachment (economically inactive persons) have more than 50 per cent higher IHD mortality risks than blue-collar workers, whereas white-collar workers have approximately 20 per cent lower risks. Self-employed men have an elevated risk of IHD mortality, which presumably reflects stressful work environments and the fact that in Finland, entrepreneurs often come from the lower social classes (Johansson 2000; Saarela 2003). Comparing the sexes, we may note that women’s educational level obviously captures social class better than their socio-economic status, whereas the opposite seems to be the case for men.
In correspondence with previous findings, homeowners have substantially lower mortality risks than rent-takers, which reflect the variable’s great importance as an additional determinant of individuals’ socio-economic position.
Next, we simultaneously include individuals’ birth region and region of residence. The estimation results for the interaction effect between the two variables on the IHD mortality risk are summarised in Table 3 for each sex. As indicated by the confidence intervals, standard errors of the estimated parameters are quite large. It is evident, however, that people born in Eastern and Northern Finland cannot lower their high IHD mortality rates even if they migrate to Western Finland, but their mortality levels are rather somewhat higher than of non-migrants in the source regions. As an illustration, men born and residing in Eastern (Northern) Finland have 25 (14) per cent higher IHD mortality risks than those born and residing in Western Finland, whereas men born in Eastern (Northern) Finland but residing in Western Finland have 38 (19) per cent higher mortality rates. Corresponding numbers for women are 17 (56) and 24 (96) per cent, respectively.
To conclude, we additionally incorporate the coastal area and the Helsinki area into the analysis, which facilitate a comparison between Finnish speakers and Swedish speakers. The results are, for each sex, summarised in Table 4. They show that, within the same regions, Swedish speakers’ IHD mortality risks are approximately ten per cent lower than Finnish speakers’. The relative difference between the two ethnic groups also tends to be somewhat larger in women than in men, similar to the case with the regional mortality differentials within the Finnish-speaking population.
Discussion and conclusions
This paper has illustrated that in Finland people’s birth region is a more decisive determinant of IHD mortality than their current region of residence. This conclusion, for persons aged 65–79 years in the period 1981–2004, corresponds with findings from previous analyses of regional differences in IHD mortality in the country based on working-aged people in the 1970s and early 1980s. We see that mortality levels of people born in Eastern and Northern Finland are notably higher than those of people in Western Finland. Migrants who originate in high-mortality areas do not either have lower mortality rates than people who have remained in the source regions, in spite that they should be integrated into the new environment as most of them migrated several decades ago. Also in conjunction with other studies undertaken, for other age groups, we find no explicit factor that explains the ethnic-group mortality difference in Finland. Swedish speakers have ten per cent lower risks of IHD mortality than Finnish speakers. For the other main causes of death, the difference is not equally wide (not shown). Reasons behind the interrelation between geographic ancestry and IHD mortality must therefore obviously be sought in circumstances that operate before or at early childhood. In previous research, however, variables used to proxy environmental and behavioural factors at childhood and adulthood could not explain the mortality variation by birth region, despite them having strong effects on the overall mortality risk.
The importance of geographic ancestry can be further illustrated by observing IHD mortality levels of Finns who live abroad. During the past 50 years, more than half a million Finns have moved to Sweden. A very large proportion of them also returned to Finland, but the population of Finnish-born people in Sweden was at the end of the 20th century still almost 200,000 persons. As shown by Figure 5, which is based on official vital statistics and the 50 per cent sample of Swedish speakers in Finland used earlier, IHD mortality rates in Finland have consistently been higher than those in Sweden (and in Denmark, as an additional example). Mortality rates of Swedish speakers in Finland, on the other hand, are close to parity with those of the population in Sweden.
With an additional data set (CHESS 2008), we can distinguish Finnish-born persons in Sweden, and observe their IHD mortality rates during the period 1981–2000. Results of these calculations, which are standardised for age and observation year, are summarised by sex in Figure 6 for the age groups 45–64 years and 65–79 years. Similar to the case of internal migrants in Finland, these immigrants have had much time to adapt to a new environment (the lion’s share of all migration abroad has been undertaken by people aged less than 30 years). In the figure, IHD mortality rates of Finnish immigrants in Sweden, Swedish speakers in Finland, and the total population in Sweden, are compared with those in Finland. In the younger age group, Finnish immigrants in Sweden are at levels that are even higher than those in Finland, whereas relative differences are somewhat more equal to Sweden in the older age group. The older immigrants still have more than ten per cent higher death risks than the total population in Sweden (and Swedish speakers in Finland). Only immigrant women aged 65–79 years have death rates that are statistically different from those of the population in Finland.
It should be recognised that population register data in Sweden cannot separate Finnish immigrants with Swedish-speaking ethnic origin because there is no information about a person’s mother tongue. The relative death rates of immigrants as displayed here consequently represent all persons born in Finland, irrespective of ethnic origin. Since the Swedish-speaking Finns have had higher migration rates to Sweden, and lower return migration rates, they constitute as much as approximately one quarter of all Finnish-born persons in Sweden (Rooth and Saarela 2006). If we assume that their IHD mortality rates are the same as in Finland, the relative death rate of immigrants with Finnish-speaking ethnic origin would be even closer to that of the population in Finland, or 0.96 (instead of 0.92) for men aged 65–79 years and 0.78 (instead of 0.75) for women aged 65–79 years.
With the type of large-scale population register data used here, it is of course not strictly possible to statistically verify that mortality variation is due to geographic clustering of genetic factors. In light of recent advances in the knowledge about IHD determinants, we are strongly inclined to lean towards this explanation. Recent genome-wide analysis of the population in Northern Europe also finds that genetic differences between Eastern and Western Finns are of the same magnitude as between Swedes and Britons, and much larger than those between Britons and Germans (Salmela et al. 2008). In correspondence with our illustration of the country-specific IHD mortality rates in Figure 5, the gene mappings indicate that Swedish-speakers in Finland resemble the population in Sweden.
Our research for Finland signals that a life course approach to understanding chronic diseases like IHD should not disregard the role played by hereditary factors. Genetic susceptibility to IHD, and its link to persons’ geographic origin, will no doubt raise additional research interests in the future. One accomplishable approach based on the Finnish population register would be to construct intergenerational data.
Acknowledgment
Thanks are owed to anonymous persons and to CHESS (Stockholm) for seminar comments and permission to use Swedish data on causes of death by country of birth.