Height and mortality from cancer among men: prospective observational studyBMJ 1998; 317 doi: https://doi.org/10.1136/bmj.317.7169.1351 (Published 14 November 1998) Cite this as: BMJ 1998;317:1351
- George Davey Smith, professor of clinical epidemiology ()a,
- Martin Shipley, senior lecturer in medical statisticsb,
- David A Leon, reader in epidemiologyc
- aDepartment of Social Medicine, Canynge Hall, Bristol BS8 2PR
- bDepartment of Epidemiology and Public Health, University College London, London WC1E 6BT
- cDepartment of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London WC1E 7HT
- Correspondence to: Professor Davey Smith
- Accepted 16 June 1998
Editorial by Albanes
Experiments in animals have shown that energy intake during the growth period is positively associated with the later incidence of cancer.1 Recently, direct evidence of an association between childhood energy intake and adult mortality from cancer among humans was published in a report from the Boyd Orr cohort study.2 Childhood energy intake was positively associated with mortality from cancers not related to smoking, whereas there was no association between energy intake and mortality from cancers related to smoking.2 This is to be expected as the substantial effects of tobacco would mask any effects of childhood diet on cancers related to smoking.
Height has been used in previous studies as a marker for energy intake in childhood,3 with the limited evidence indicating a positive association for some cancer sites. 3 4 5 In the Boyd Orr study data were not available on smoking behaviour and were limited on adulthood socioeconomic position.2 We therefore analysed the association between height and mortality from cancer in a large cohort of men for whom detailed data on socioeconomic position in adulthood and on smoking behaviour were available.
Subjects, methods, and results
In the Whitehall study of London civil servants, data on employment grade, height, and smoking behaviour were available for 17 378 men aged 40-64 who were examined between 1967 and 1969.4 During follow up until 31 January 1995, 2226 of these men died of cancer: 725 from cancers unrelated to smoking and 1501 from cancers related to smoking. To adjust for the potential confounding effects of other variables proportional hazards analyses were carried out with height as a continuous variable and age (in age bands of five years), employment grade (administrative, professional and executive, clerical, other), and smoking behaviour (cigarette smoker, pipe or cigar smoker, ex-cigarette smoker, and number of cigarettes smoked per day for current cigarette smokers) as covariates.
The table shows relative death rates from cancer by height category and for each increment of 6 inches in height (1 inch is about 2.5 cm). For cancers unrelated to smoking the association with height was positive and strengthened by adjustment for socioeconomic position (indexed by employment grade) and smoking behaviour. For cancers related to smoking the association between height and mortality was negative but not significant. Adjustment for socioeconomic position and smoking behaviour reversed the direction of the association, but it remained small and non-significant. As expected, cigarette smoking was strongly associated with cancers classified as smoking related but not with cancers classified as unrelated to smoking. The age adjusted relative rates for height and the smoking unrelated and smoking related cancers were significantly different (P=0.002). Exclusion of mortality occurring during the first five or first ten years of follow up did not materially alter these findings.
Our findings parallel those relating childhood energy intake to cancer mortality in the Boyd Orr cohort study, in which the positive association between childhood energy intake and subsequent risk of cancer was also confined to cancers unrelated to smoking.2 Most previous studies have either grouped all cancers together or looked only at individual cancers. However, consistent with our results, the physicians health study found a positive association of height with all malignant neoplasms but not with lung cancer.5 In line with extensive animal experimental evidence,1 therefore, our data and those from the Boyd Orr study2 suggest that energy intake during growth may be an important determinant of later risk of developing cancer. Since height serves as only an indirect and comparatively weak proxy measure of dietary intake in childhood, the size of the association found in this study may reflect a much stronger underlying association with directly measured childhood energy intake.
Contributors: The idea for this paper came from a discussion between the authors. GDS wrote the first draft of the manuscript around analyses performed by MS, and all authors contributed to the final draft. GDS is guarantor for the study.
Conflict of interest None declared.