Childhood Body Weight and CVD, Cancer in Older Adulthood
Childhood Body Weight and CVD, Cancer in Older Adulthood
On June 4, 1947, investigators from the Scottish Mental Survey of 1947 attempted to measure the intelligence of every 11-year-old child who was attending school on that day in Scotland; 88% of children responded (n = 70,805). A subgroup of study members—the "36-Day Sample," comprising children born on the first, second, and third days of each month of 1936—were selected by the Scottish Council for Research in Education for further research participation. This subsample was representative of the full sample from the Scottish Mental Survey of 1947 in terms of sex, geographical location, family size, and cognitive score. In present analyses of 5,083 children (2,561 girls) from that period, we excluded those born on the first day of the even-numbered months of 1936 who went on to take part in a more intensive longitudinal survey (the "6-Day Sample"). Our revitalization of the Scottish Mental Survey of 1947 as a cohort study was approved by the Scotland-A Research Ethics Committee, the National Health Service Scotland Privacy Advisory Committee, and the Confidentiality Advisory Group of the Health Research Authority.
For each study member, a head teacher answered a questionnaire (known as the "sociological survey") pertaining to each pupil's physical attributes and socioeconomic circumstances, including physical disability, father's occupational level, number of people in the home, and the number of rooms in the home. Physical disability was denoted by a history of chorea, congenital paralysis, defective vision, deafness, encephalitis, epilepsy, or meningitis. Room occupancy was computed by dividing the number of people living in the dwelling by the number of rooms. Father's or main guardian's occupation was coded into 1 of 5 social class categorizations, ranging from professional (highest prestige) to unskilled. Height (inches) and weight (stone/pounds) were directly measured. Conversion to metric units allowed us to compute BMI using the standard formula (weight (kg)/height (m)).
We electronically traced study members who resided in Scotland using the National Health Service Central Register and located those who had migrated to England and Wales using the Medical Research Information Service Integrated Database and Administration System. For those persons who were not automatically matched, a manual search was undertaken. Morbidity was ascertained through linkage to 2 sources: hospital admissions records (Scotland: 1980–2014; England and Wales: 1997–2013) and cancer registrations (Scotland: 1980–2014; England and Wales: 1984–2013). Irrespective of database, we used the first diagnosis of disease in our analyses, categorizing them according to the International Classification of Diseases, Ninth or Tenth Revisions (see Table 1 and Table 2 for codes). We also grouped malignancies into those known to have an association with cigarette smoking and, by inference, those with no such relationship. In so doing, we attempted to circumvent the problem of an absence of data on smoking status, a potential confounding variable in the present study.
Of the initial 5,083 study participants, we were able to trace 4,826 (95%). Comparing the traced and untraced groups, we found essentially no difference in BMI (16.8 vs. 16.9, respectively; P value for difference = 0.39) or other baseline characteristics. Exclusion of people with missing data for BMI or other covariates resulted in an analytical sample of 4,620 (2,288 women). In the main analyses, we divided the participants into quartiles by BMI and used those in the lowest quartile as the reference group; we also computed morbidity risk for each 1–standard deviation increase in BMI value. We used Cox proportional hazards regression analyses with age in years as the time scale to compute hazard ratios with accompanying 95% confidence intervals in order to estimate the relationship between childhood BMI and later risk of morbidity. Study members were censored at age at hospitalization or age at the end of follow-up period, whichever occurred first. In preliminary analyses conducted separately for men and women, there was no evidence that sex modified the link between BMI and the major causes of morbidity. These data were therefore pooled. Hazard ratios were adjusted first for sex and then for the additional covariates of fathers' occupation, room occupancy, height, and physical disability. All analyses were undertaken using the SPSS for Windows, version 21.0 (IBM Corp., Armonk, New York).
Methods
On June 4, 1947, investigators from the Scottish Mental Survey of 1947 attempted to measure the intelligence of every 11-year-old child who was attending school on that day in Scotland; 88% of children responded (n = 70,805). A subgroup of study members—the "36-Day Sample," comprising children born on the first, second, and third days of each month of 1936—were selected by the Scottish Council for Research in Education for further research participation. This subsample was representative of the full sample from the Scottish Mental Survey of 1947 in terms of sex, geographical location, family size, and cognitive score. In present analyses of 5,083 children (2,561 girls) from that period, we excluded those born on the first day of the even-numbered months of 1936 who went on to take part in a more intensive longitudinal survey (the "6-Day Sample"). Our revitalization of the Scottish Mental Survey of 1947 as a cohort study was approved by the Scotland-A Research Ethics Committee, the National Health Service Scotland Privacy Advisory Committee, and the Confidentiality Advisory Group of the Health Research Authority.
For each study member, a head teacher answered a questionnaire (known as the "sociological survey") pertaining to each pupil's physical attributes and socioeconomic circumstances, including physical disability, father's occupational level, number of people in the home, and the number of rooms in the home. Physical disability was denoted by a history of chorea, congenital paralysis, defective vision, deafness, encephalitis, epilepsy, or meningitis. Room occupancy was computed by dividing the number of people living in the dwelling by the number of rooms. Father's or main guardian's occupation was coded into 1 of 5 social class categorizations, ranging from professional (highest prestige) to unskilled. Height (inches) and weight (stone/pounds) were directly measured. Conversion to metric units allowed us to compute BMI using the standard formula (weight (kg)/height (m)).
Morbidity Ascertainment
We electronically traced study members who resided in Scotland using the National Health Service Central Register and located those who had migrated to England and Wales using the Medical Research Information Service Integrated Database and Administration System. For those persons who were not automatically matched, a manual search was undertaken. Morbidity was ascertained through linkage to 2 sources: hospital admissions records (Scotland: 1980–2014; England and Wales: 1997–2013) and cancer registrations (Scotland: 1980–2014; England and Wales: 1984–2013). Irrespective of database, we used the first diagnosis of disease in our analyses, categorizing them according to the International Classification of Diseases, Ninth or Tenth Revisions (see Table 1 and Table 2 for codes). We also grouped malignancies into those known to have an association with cigarette smoking and, by inference, those with no such relationship. In so doing, we attempted to circumvent the problem of an absence of data on smoking status, a potential confounding variable in the present study.
Statistical Analyses
Of the initial 5,083 study participants, we were able to trace 4,826 (95%). Comparing the traced and untraced groups, we found essentially no difference in BMI (16.8 vs. 16.9, respectively; P value for difference = 0.39) or other baseline characteristics. Exclusion of people with missing data for BMI or other covariates resulted in an analytical sample of 4,620 (2,288 women). In the main analyses, we divided the participants into quartiles by BMI and used those in the lowest quartile as the reference group; we also computed morbidity risk for each 1–standard deviation increase in BMI value. We used Cox proportional hazards regression analyses with age in years as the time scale to compute hazard ratios with accompanying 95% confidence intervals in order to estimate the relationship between childhood BMI and later risk of morbidity. Study members were censored at age at hospitalization or age at the end of follow-up period, whichever occurred first. In preliminary analyses conducted separately for men and women, there was no evidence that sex modified the link between BMI and the major causes of morbidity. These data were therefore pooled. Hazard ratios were adjusted first for sex and then for the additional covariates of fathers' occupation, room occupancy, height, and physical disability. All analyses were undertaken using the SPSS for Windows, version 21.0 (IBM Corp., Armonk, New York).