The affected relative/sib pair approach has been widely used in linkage analysis. One of the major advantages of this approach is that it is not affected by age-related penetrance (variable age of onset) because all individuals in the analysis are affected. However, there are at least two situations in which this approach alone is not sufficient. First, phenocopies that appear to be identical to a genetic trait but are caused by nongenetic factors may decrease the power of an affected relative/sib-pair analysis, making it less than ideal. Second, phenotypic heterogeneity (where different genes cause different ages of onset) may lead to two affected relative/sibling pairs having different ages of onset due to different genes and, thus, a decrease in the power to detect linkage. Therefore, it is of interest to find an analytic approach that addresses these issues. Restricting the analysis to the phenotype of early age of onset of obesity/overweight (<35 years) attempts to reduce the phenotypic heterogeneity due to differences in age of onset. Another aim of this analytic approach is to "weed out" phenocopies of obesity/overweight that may occur later in life and may be caused by other genetic or environmental factors. The survival analysis residual method attempts to detect a genetic effect on the age of onset of obesity/overweight. This method accounts for cohort, sex, and age-specific risk for obesity/overweight in the sample population, by determining the cumulative incidence in the sample for a particular cohort, sex, and age. For example, the value of the residual (the adjusted affection status) would be 0.8 for an affected person when the cohort, sex- and age-specific cumulative incidence was 0.2 (fewer people had become obese/overweight in their group) and 0.2 if the cumulative incidence was 0.8 (more people had become obese/overweight in their group). The person with the residual of 0.8 could potentially have a genetic risk factor predisposing them to an earlier age of onset than the person with the residual of 0.2. The restriction method would be expected to perform well in the case of an early age of onset phenotype that is etiologically distinct from later age of onset phenotypes. The survival analysis residual method would be expected to be more robust when there is a genetic effect on the variation in the age of onset throughout the life-span. In this case, loss of information would occur by using an age of onset cut-off to restrict the analysis. A strong genetic effect on age of onset throughout the life-span (including ages less than the cut-off age of the restricted analysis) may be expected to produce consistent results between both analytic methods.
In our genome-wide linkage analysis, five regions showed consistent suggestive evidence for linkage (one marker with p < 0.01 and a second contiguous marker at p < 0.05). These regions were chromosome 1 (280–294 cM) and chromosome 16 (56–64 cM) for overweight using the survival analysis residual method and chromosome 13 (102–122 cM), chromosome 17 (127–138 cM), and chromosome 19 (23–47 cM) for obese before age 35.
The results on chromosomes 1 and 16 for overweight using the survival analysis residual method were not replicated in the affected sibling pairs with age of onset of overweight before 35. This region may be responsible for a genetic effect on age of onset of overweight in the middle and later years of life when, in general, a more sedentary lifestyle has been adopted. This may be the case if the region was involved in a gene × environment interaction in which a sedentary lifestyle in addition to the gene was necessary to result in moderate weight gain. Another explanation is that the number of affected sibling pairs was too small to detect linkage, whereas using all the sibling pairs in the survival analysis residual method (a continuous trait) resulted in more power to detect linkage.
The results on chromosomes 13 and 17 for obese before age 35 were not replicated by the survival analysis residual method. This may indicate that chromosomes 13 and 17 are linked with a distinct phenotype of obesity in which the age of onset is before 35 years of age. These regions may even be responsible for a phenotype of childhood or adolescent obesity, although we did not have enough people in these age groups to test this hypothesis.
Another explanation for the inconsistency of the results on chromosomes 13 and 17 could be strong environmental determinants of shifting from normal weight or overweight to obese in the middle and later years of life (age of onset ≥ 35). If a gene(s) plays a role in the age at which one becomes obese, it may be more easily detected at a younger age in which people tend to be more physically active (the environmental exposure of a sedentary lifestyle is absent). As many people become more sedentary in their 30s and beyond, weight gain may occur due to excess energy intake even in the absence of genetic factors, making it more difficult to detect a genetic effect. The stratified analysis of people who were obese at a younger age would not be as affected by this shift in environmental exposure. This approach would perform the linkage analysis on a more homogeneous population of affected individuals. Conversely, although the survival analysis residual method accounts for excess energy intake with increasing age in the population through adjustment of the affection status by the cumulative incidence, all individuals are included in the analysis, potentially creating a more heterogeneous sample with respect to etiology of the phenotype. Hypothetically, a discordant sibling pair in which one sibling became affected at age 40 entirely due to excess energy intake (assume no genetic susceptibility) while the other sibling was unaffected at age 40 and did not have an excess energy intake would tend to "wash out" a true genetic effect on the phenotype in another sibling pair.
There was one region (chromosomes 19 at 22 cM) in which both analytic methods detected evidence for linkage to the phenotype of obesity. Consistency of results from the two analytic methods may have occurred due to a genetic effect on age of onset throughout the lifespan including ages < 35.
Overall, both analytic methods have strengths and weaknesses and the use of both methods together to explore the genetics of the age of onset of a trait may prove to be beneficial in determining a gene that is linked only to an early age of onset phenotype versus one that determines age of onset through all age groups. The survival analysis residual method allows for the entire sample to be analyzed, whereas the stratification method limits one to a smaller sample size, which reduces one's power. Additionally, stratification may lose information from siblings who are not affected at the time of the visit, but may later become affected, whereas the survival analysis residual method uses all siblings and adjusts their affection status by the cumulative incidence of the trait. However, stratification by age may be better able to detect linkage if there is a strong environmental component in ages greater than the stratification cutoff age.