Consanguineous marriage continues to be popular in many parts of Asia and Africa 8, although in East Asia a marked general decline in consanguinity has occurred since the mid-20th century. Elsewhere, the picture is mixed with, for example, a partial reduction in consanguinity in South India, but no apparent change in the approximately 50% of marriages contracted between first cousins in Pakistan. In the Middle East, a decline in consanguinity is seen in some Arab countries but there has been an increase in others 2. In Iran, marked inter-ethnic differences in the prevalence of consanguinity make any national trend difficult to determine 9. Cousin marriage has recently become a much more important issue in Europe, with over 10 million migrants from regions with high levels of consanguinity (North Africa, the Middle East, Turkey, Central Asia, and South Asia) now resident across the continent. Equivalent information on the level of consanguinity among migrant communities in the USA is hampered by the fact that first cousin marriage is prohibited or a criminal offence in 31 of the 50 states 2.

A predicted effect of the admixture of individuals from different genetic populations is an increase in average individual genome-wide heterozygosity, h, as previously reported for morphometric and biochemical parameters among younger Israelis 10. Using genomic data, Rudan and colleagues 11 measured h in Croatian populations and demonstrated that levels are increasing through urbanization, isolate break-up and admixture. The observed wide range of variation in h was unlikely to be stochastic since it was closely correlated with individual genetic histories, leading to the conclusion that population substructure, recent inbreeding and population admixture were the underlying mechanisms.

High-density single nucleotide polymorphism (SNP) analysis has been used to monitor the impact of these population changes on individual and community levels of autozygosity - defined as homozygosity in which each allele is derived from the same ancestral gene (that is, is identical by descent). Whole-genome SNP data have demonstrated short runs of homozygosity (ROH) measuring tens of kilobases (kb) and covering up to one-third of the autosomal genome, which are indicative of ancient LD patterns 12. Longer ROH tracts (for example, >500 kb) provide a useful measure of more recent consanguinity. These regions can be used to infer the autozygosity status of genomic regions and, by summing the overall length of ROH regions above a minimum size, they provide a good summary measure of individual genome-wide homozygosity 12. Long genomic segments containing ROH, which reflect inheritance from a recent common ancestor, are common in many populations, including Han Chinese, indigenous Taiwanese, Caucasians and African Americans 13141516171819. Increased urbanization and population admixture in the USA have been associated with a steady decrease in the size and frequency of ROH regions over the course of the 20th century: there has been an estimated 14% decrease in ROH frequency, a 24% decrease in the amount of the genome in ROH, and a 30% reduction in the inbreeding coefficient, F_ID 20. Since Europeans have the smallest among-population variance component, the trend of decreasing autozygosity with younger chronological age in the North American population of European ancestry studied by Nalls et al. 20 has important implications for the populations of other continents, where urbanization may exert even larger effects on genomic organization.

A reduction in autozygosity through outbreeding should decrease the burden of recessive monogenic disorders in communities with a high prevalence of consanguineous marriages 212223. A detailed prospective study of the Pakistani community in Birmingham, UK, indicated a significantly higher prevalence of autosomal recessive disorders among the progeny of consanguineous couples 24. On the basis of these data, it has been calculated that there would be an approximate 7/1,000 increase in autosomal recessive disorders per 0.01 increase in the mean coefficient of inbreeding 25. If this estimate is applied to the reduction in inbreeding coefficient reported by Nalls et al. 20, then one could infer that approximately 1% of the annual births affected with an autosomal recessive disorder have been avoided in the North American population due to the 100-year population trends that have been identified.

Considered on a longer time scale, the reduced numbers of affected homozygotes will, however, result in decreased selection against the causative disease alleles and eventually result in their increased incidence in the gene pool. At the same time, due to isolate break-up, a higher proportion of affected individuals will be compound heterozygotes with different mutations inherited from each parent.

Outbreeding has long been known to influence complex as well as monogenic traits that are influenced by recessive or partially recessive variants 26. How large is this effect likely to be? Recessivity is thought to be a property of multi-enzyme systems, which results in enzymatic safety margins, so that a reduction in the activity of a single allele will generally have little impact on the metabolic pathway 27. The majority of genetic changes in enzymes, which make up 25 to 30% of gene products 28, are expected to show recessive or partially recessive phenotypes, whereas a smaller proportion is expected for non-enzymic proteins. Deleterious mutations are more likely to be recessive, while neutral or small-effect mutations are more often additive 26. A wide variety of complex traits in many organisms are influenced by inbreeding, although it is unclear why the genetic (dominance) effects tend to be unidirectional. Most of the variants affecting human complex traits remain to be identified but a recent study suggested that partially recessive effects of relatively high penetrance may explain a proportion of the genetic predisposition to schizophrenia 16. Therefore, just as the offspring of consanguineous matings may have lower mean health and fitness because of the homozygous expression of detrimental recessive alleles 29303132, similar effects could operate with the more numerous partially recessive variants influencing complex diseases.

In theory, any increase in individual genome-wide heterozygosity would be expected to influence variation in biological traits that show significant variance due to recessive alleles (dominance variance) 3334, such as systolic blood pressure (BP), total cholesterol and low-density lipoprotein cholesterol 3536. Additionally, it has been proposed that these changes may have a significant impact in modifying the epigenetic, epistatic and polygenic pathways that influence complex traits 20.

Rudan and colleagues 29 found a positive linear relationship between the inbreeding coefficient and both systolic and diastolic BP in a Croatian population, with recessive or partially recessive genetic variants accounting for 10 to 15% of the total variation in BP. A decrease in mean inbreeding coefficient (F_ID) of 0.01 was associated with a reduction of approximately 3 mmHg in systolic BP and 2 mmHg in diastolic BP in both sexes 29. Applying these findings to the change in mean inbreeding coefficient reported for the USA during the 20th century 20 would equate to an equivalent decrease of approximately 3 mmHg in systolic BP and approximately 2 mmHg in diastolic BP in both sexes.

The associations between systolic and diastolic blood pressure and coronary heart disease events and stroke have been described in a meta-analysis of 61 cohort (prospective observational) studies 37 and in the largest reported meta-analysis of randomized trials of blood pressure reduction 38. The results indicated that cardiovascular mortality showed a constant proportional change in risk for a specified change in blood pressure (down to a diastolic BP of 70 mmHg). Regression coefficients were estimated from these data, facilitating the prediction of the proportional reduction in disease events for any age and blood pressure difference. The calculated regressions suggested that a decrease in adult diastolic and systolic BP of the magnitude derived from Rudan et al. 29 and Nalls et al. 20 would result in reductions in the incidence of coronary heart disease (and heart failure events) and stroke of approximately 5 to 10% and 10 to 15%, respectively 38.