Given that the explosion in obesity prevalence over the past 20 years is likely to have taken place against a background of relatively constant population genetic structure,  the question of to what extent obesity is subject to genetic influence is one that merits careful consideration.  Many studies have attempted to resolve the population variance of a specific obesity phenotype into genetic, environmental and unknown (or residual) effects.

In principle, the total observed phenotypic variance,    V_p  may be considered to be due to the sum of genetic variance   (V_g ),  shared environmental variance       (V_c )  and an unknown residual (unshared environmental) variance (V_e ) such that V_p = V_g  + V_c  + V_e . The percentage genetic inheritability of the trait in question is represented by the term V_g /V_p.  Modifications of this simple model to attempt detection of gene–gene and gene–environment interactions and the application of complex multivariate computational modelling in different study populations are reviewed in detail elsewhere (Bouchard et al., 1998).

Twin studies
Twin studies allow separation of genetic and environmental components of variance since monozygotic (MZ)  twins share 100 per cent of their genes whilst non-identical dizygotic (DZ) twins share 50 per cent on average. The fact that there is discordance in the prevalence of obesity between MZ and DZ twins raised together supports the concept of genetic heritability of obesity if it is assumed that twins share exactly the same environmental influences (although it has been suggested that MZ twins may share more environmental influences than DZ twins;  Hebebrand et al.,  2001).  Total genetic variance may then be subdivided into two components; additive variance, which results from the sum of contributions of many alleles at different loci and non-additive effects, which are principally determined by the dominance of one allele over another at the same locus. It follows that an additive model is suggested when intrapair correlations of DZ twins are half that of MZ twins and a non-additive model is suggested when DZ twins have substantially less than half the intrapair correlations of variance of MZ twins.

Comparison of MZ twins raised together with MZ twins raised apart probably represent the ideal study group. However, such study populations are difficult to find and, even then, twins will have shared the same intra-uterine environment (which may be important on the basis of the Barker hypothesis;  Hales and Barker,  1992).  Furthermore,  there may be indirect genetic effects in operation and the effect of environment may be underestimated as certain environmental conditions are likely to be common to both twins (e.g. the general availability of fast foods).

One study (Stunkard et al., 1986) assessed body mass index in a sample of 1974 MZ and 2097 DZ male twin pairs and found concordance in MZ twins to be around 0.8. This was twice as high as that in DZ twins both at age 20 and at 25 year follow up. Others (Fabsitz et al., 1992), however, report an age-specific effect such that only 40 per cent of the genetic factors that influence body weight at the age of 20 continue to do so by the age of 48.

One of the best estimates of obesity heritability, accounting for 67 per cent of variance, is derived from the Virginia cohort of 30 000 twins, their parents, siblings, spouses and children (McLaughlin, 1991). Overall, weighted mean BMI correlations have been calculated to be 0.74 for MZ twins, 0.32 for DZ twins, 0.25 for siblings, 0.19 for parent–offspring pairs, 0.06 for adoptive relatives and 0.12 for spouses (Maes et al., 1997) and the overall relative risk of siblings lies within the approximate range 3–7 (Allison et al., 1996).

The heritability of gene–environment interactions thought capable of leading to obesity has also been demonstrated using MZ twin populations (Bouchard et al., 1998). The principle of this approach lies in the variability of individual response to environmental perturbation (in this case, weight change in response to either overfeeding or to increasing exercise with energy intake held constant). Where the response differs more between than within pairs of MZ twins, it may   be assumed   that genetic factors are responsible.  These   studies have demonstrated   considerably   more   variance   (in some   cases by up to a factor of 3–6 depending on the variable studied)  between rather than within twin pairs for a number   of measures   of body   fat accumulation,    distribution and energy expenditure.  This supports the hypothesis that individual responses to diet and exercise have a substantial genetic component and may go some way to explaining the observation of increasing obesity prevalence     on the background of a relatively constant gene pool.  To what extent there is a genetic basis for the fact that obese people of similar BMI may be variably subject to obesity-related complications such as type 2 diabetes (independent of other risk factors) is unclear.

Adoption studies
Adoption studies rely on the assumption that differences between adopted children   and their adoptive   parents/siblings   are due to genetic   differences   and differences between them and their biological families are due to environmental influences. However, adoption studies are complicated by problems relating to ascertainment of the biological father (false paternity being found in some 8 per cent of individuals in many studies), the effects of selective or late placement of the child and the inherent inability of such studies to assess gene–environment interactions.

Thus there has been considerable heterogeneity in published estimates of the heritability of obesity depending for the large part on the type of study and population used.  Whereas twin studies are thought in general to overestimate the true heritability,  adoption   studies are thought   to underestimate   it.  Thus, estimates for obesity heritability from adoption studies are lower than those from twin studies and range from around 0.2–0.6 with two very large family studies providing estimates of 0.3–0.4 (Maes et al., 1997).

Warden CH and Fisler JS
Katsanis N, Beales PL, Woods MO