In a unique study of a longitudinal birth cohort born in 1946 and followed through to age 53 years, we observed that the associations of FTO rs9939609 and MC4R rs17782313 on body size varied with age. The changes with age appeared to be biphasic, in that the associations with both variants strengthened during childhood and adolescence, the associations peaked in early adulthood and thereafter weakened with increasing adult age. Associations with FTO rs9939609 were most apparent with BMI, whereas, MC4R rs17782313 was positively associated with childhood and adult height and therefore showed stronger associations with body weight than with BMI.
The main strength of this study was the availability of longitudinal data on weight and height across childhood, adolescence and adulthood up to age 53 years, which allowed a unique exploration of the life course genetic associations with body size. Weight and height were measured at all ages, except at ages 20 and 26 years, when they were self-reported. Despite the reliance on self-reported measures at those ages, the peak associations between FTO rs9939609 and BMI SDS, and between MC4R rs17782313 and weight SDS, were both seen at age 20 years. Differential mis-reporting of body size by genotype seems unlikely, but if heavier individuals were more likely to under-report their weight than others it is possible that the peak associations and the rates of change in associations with age may be downwardly biased. Our statistical models included individuals without complete outcome data and therefore assumed data to be missing at random. This is a reasonable assumption in the sample studied given that the DNA was collected at the most recent contact with participants and hence the sample represents those still alive and in the study at age 53 years. Those with missing genetic information had similar mean body size measures at most ages compared with those included in the analysis.
There have been only a few other studies exploring the variations in these genetic associations over the life course. In older adults, a study of men in the Health Professionals Follow-Up Study reported that, similar to our findings, associations between the same FTO
rs9939609 variant and BMI declined with age. However, in the same report those age effects were not seen in women in the Nurses Health Study (12
). Rather, in those women carriers of the MC4R
rs17782313 C-allele showed greater gains in adult weight and BMI than non-carriers (13
). In both of those populations, heights and weights were self-reported and follow-up only started in adulthood at mean age 56 years in men and 44 years in women (14
). In the NSHD we found no evidence of differences between men and women in the genetic associations with body size across the life course. In a study of 1629 Danish men, born between 1943 and 1977, the FTO
rs9939609 AA genotype was associated with faster weight gain from birth to age 7 years, but not with further weight gain during childhood and adolescence. Subsequently, in contrast to our findings, the AA genotype was associated with a second increase in BMI z-score, from ages 13 or 20 years up to age 35 years (14
). However, in that Danish study, half of the men were selected for the presence of adult obesity (above the 99th percentile), and it is possible that the effect of the FTO
rs9939609 variant on BMI or weight gain might be greater in individuals with obesogenic lifestyles, such as sedentary behaviours (15
). Finally, consistent with our current findings, a study of >7000 UK twins described that the association between the FTO
rs9939609 variant and BMI was not apparent at age 4 years, but became increasingly stronger at ages 7–11 years, in parallel with a rise in the heritability of BMI from 48% at age 4 years to 78% at age 11 years (17
Although our cohort was sampled from a national population, even such representative cohorts will have mean BMI trajectories which differ from cohorts born earlier or later (18
). NSHD cohort members were all born during 1 week in 1946 and experienced post-war rationing during the first 8 years of life, which could have diminished the genetic differences in early childhood weight gain and BMI. The prevalence of overweight or obesity in this cohort remained low during childhood and adolescence until the onset of the obesogenic environment in the 1980s when cohort members were aged 30–40 years old (18
). Our current observed weakening of the genetic associations with FTO
variants across mid-life could therefore reflect the increasing impact of environmental influences on physical activity and food consumption coincident with the onset of the obesity epidemic (18
). However, in a single study where all the participants were born in the same era it is difficult to disentangle age effects from period effects.
Recent studies have found several further common variants associated with BMI and obesity risk (19
). Further longitudinal studies of those variants would help to show whether the biphasic changes with age that we identified with both the FTO
variants are generalizable observations, possibly reflecting age-related trends in the heritability of BMI and obesity risk. Variants in PCSK1
, which encodes an enzyme that converts prohormones into key functional regulators of appetite (22
), demonstrated age-dependent effects on obesity risk in the EPIC-Norfolk Study, being apparent only in younger individuals (aged below the median age 59 years) (24
). However, that analysis was based on cross-sectional data where it is not possible to separate age-effects from birth-cohort effects. The importance of longitudinal data in the detection of age-related effects is highlighted by a recent study of the penetrance of rare deleterious mutations in MC4R
on the development of monogenic obesity (25
). In cross-sectional inter-generational comparisons, the prevalence of obesity among carriers of MC4R
mutations reduced with age, from 79% in children, to 60% in 18–52-year-old adults, and 40% in adults aged >52 years, suggesting that these mutations had lower penetrance among older birth cohorts. In contrast, in their longitudinal analysis, the prevalence of obesity increased from 37% at 20 years to 60% at >40 years, indicating that the penetrance of these mutations increased with age (25
). Further longitudinal population-based studies spanning a wide range of birth years are therefore required to distinguish between the effects of age and period changes in the obesogenic environment.
In conclusion, genetic variants in FTO
showed biphasic changes in their associations with BMI and weight, respectively, strengthening with age up to a peak at age 20 years, and then weakening with increasing adult age. We postulate that these changes may reflect the age-dependent influence of satiety and eating behaviour on weight gain and BMI, as weight gain in adult life may be increasingly determined or modified by other hedonic, psycho-social or environmental influences (26
). Future studies of the aetiology of obesity spanning different age groups may identify age-specific genetic or other determinants of weight gain.