Twin studies can provide insights into the heritability of psychiatric disorders, and the extent to which within-population variation in psychiatric disorders is influenced by the environment. They can further describe genetic and environmental contributions to overlap between disorders, and provide information on likely gene environment interactions. Previous twin studies have found that 30–70% of liability to major depression is influenced by genetics [1, 2], as is roughly 40% of psychosomatic symptoms , and 50% of alcoholism . Findings of environmental influences that are shared between twins, including family-wide influences, have been of modest magnitude at most, and often absent, in assessments of common mental disorders in adult twins. Heritabilities of over 70% have rarely been reported, leaving room for substantial influences from a combination of environmental factors that are not shared across twins, and for measurement error.
Heritability is an estimate of the causes of variability of a phenotype within a given population. However, a high heritability does not rule out important environmental exposures if these do not vary within the population. For example, if there are important environmental influences on a phenotype, but these exposures do not vary within the specific population, then the phenotype would have a high estimated heritability. Examining populations in diverse environmental settings may increase our understanding of causal influences that are not detectable in populations that are environmentally more homogeneous. For example, the wider range of standards of living experienced in Sri Lanka compared to the West may lead to a greater magnitude of environmental influence on mental disorders. The Sri Lankan environmental context will also allow testing of measurement invariance, for phenotypes that have previously been measured in developed countries.
Past twin studies may not have taken full account of the distribution of environmental risk factors. Most have been confined to developed Western countries, which do not encompass the full range and potency/severity of environmental exposures seen throughout the world [5, 6]. Some have relied on volunteer samples rather than being truly population-based, and such ascertainment may bias or restrict the variation of socio-economic influences sampled  (e.g. underrepresentation of those with few educational qualifications ). It is possible that the experience of being a twin affects the exposures and phenotypes under study [9, 10]. Twins may differ from non-twins for many reasons, especially those twins included in genetic analyses, since the twin pair must be intact: in other words, both twins must still be alive and willing to participate. The surviving twin pairs may be unrepresentative of the general population, especially if being a twin is associated with a higher infant mortality rate, which may be particularly germane to developing countries. Finally, twin studies estimate genetic and environmental effects indirectly, by comparing the similarity of MZ pairs to the similarity of DZ pairs. They can thus provide estimates of environmental influences without having to measure the environment, but the precise nature of these environmental influences is unknown unless the exposures are directly measured .
The twin study we describe aims to overcome these shortcomings. Firstly, it is based in Sri Lanka, a developing country with a per capita gross national income of approximately $1,160, and wide disparities of wealth and environmental exposures. Secondly, we aimed to identify all twins within a defined population, and invite a random sample to participate in the study. Thirdly, we directly measured a range of environmental exposures, both those comparable to populations in developed countries, and exposures specific to this population. Finally, we identified and sampled non-twins in a similar manner, in order to examine how representative exposures and outcomes reported by twins are to the general population.
Sri Lanka has a high life expectancy (74 years), low infant mortality rate (12/1000/year) and high literacy rates (91%) despite its status as a developing country . In contrast to these favourable quality of life indices, the country has a high suicide rate (up to 40/100000/year) , and a high prevalence of alcohol dependence (29/1000) . Consumption of alcohol and deaths from alcoholic liver disease have both risen sharply in the past 20 years and a government document stated "it is no secret that Sri Lanka is listed as one of the countries with (the) highest alcoholism levels" .
The country has been affected by a prolonged civil war with hostilities in the North and East between the government and Tamil separatist forces. Whilst the Colombo district has been relatively spared from this unrest, a proportion of the residents will have had direct experience. Sri Lanka was also affected by the Tsunami of 26th December 2004 with approximately 40,000 deaths. These deaths were predominantly in the South and East of the island, but the Colombo district was affected with the loss of 56 lives. 3,397 homes were damaged to the extent that they could not be used, and partial damage of a further 2,948 homes .
The aim of the present paper is to describe the methods we used to conduct this study, and to provide baseline results on the nature of the samples identified.