Leg length and risk of cancer in the Boyd Orr cohort

Height is directly associated with mortality from cancer,¹ but the explanation for this association is unclear. Whereas adult height is a marker of nutrition and health throughout childhood, most growth before puberty is due to increases in leg length. Leg length can therefore be used as a marker for exposures that generate the association between adult height and cancer.¹,² We examined the association between prepubertal leg length and mortality from cancer in the Boyd Orr cohort.²,³

Subjects, methods, and results

As childhood height is affected by the timing of puberty we restricted our analysis to study members who were aged 2-8 years when their height and leg length were measured. In all, 1167 of these 1392 subjects (84%) were traced and had information on childhood and adult socioeconomic status, and they form the basis of this analysis. Using Cox’s proportional hazards models we examined mortality from cancer between 1948 and 1997 in relation to internally derived z scores for childhood leg length.² All traced subjects were included in these models up to the time of their death or emigration. To adjust for possible genetic factors associated with overall stature all models also include a term for trunk length z score. The z score (standard deviation score) expresses a child’s measurement as the number of standard deviations its value is from the mean for its age and sex. In these analyses each unit of the z score is approximately equivalent to 3-4 mm (the standard deviation for leg length in these children).

We classed cancers as related to smoking (codes 140, 141, 143-9, 150, 157, 160-3, 188-9 of the international classification of diseases, ninth revision (ICD-9)) and not related to smoking (all other malignant neoplasms, but mainly colorectal, breast, and ovarian cancers). We further divided cancers unrelated to smoking into sex hormone dependent (174 (breast), 179 (uterus), 182-7 (ovary, prostate, and other genital organs)) and non-sex hormone dependent (all other cancers unrelated to smoking).

The analysis showed a significant positive association between childhood leg length and mortality from cancers unrelated to smoking (table). The observed association was most obvious for mortality from sex hormone dependent cancers—the risk of death increased by 129% (95% confidence interval 9% to 383%) for every unit increase in z score for leg length. No significant associations were seen in relation to trunk length (hazard ratio for cancers unrelated to smoking 0.84 (0.58 to 1.22)), and the associations for overall height were in a similar direction but weaker than those for leg length. There was no association with all cause mortality.

Comment

In the literature associations between height and cancer have been most consistently shown for breast and prostate cancer, both of which may be hormonally dependent. In line with this, the associations reported here are strongest in relation to sex hormone dependent cancers. Two possible mechanisms may underlie these associations. Firstly, more rapid growth is associated with an earlier puberty and therefore a longer exposure to adult concentrations of sex hormones. Secondly, childhood diet may influence concentrations of insulin-like growth factor-I,⁴ subsequent growth (as indexed by childhood leg length),² and later risk of cancer.³ Insulin-like growth factor-I may increase the risk of cancer through several mechanisms, including protecting genetically damaged cells from apoptosis (cell suicide), stimulating cell proliferation, and amplifying the production of sex hormones by gonadotrophins.⁴ A role of insulin-like growth factor-I in cancer development is further supported by the results of recent prospective studies showing its strong association with risk of breast and prostate cancer.⁴,⁵ In one study the associations were independent of sex hormone concentrations.⁵ Future studies should determine whether growth hormones are as important in the aetiology of human cancer as they are in animal experiments.

Table.

Hazard ratios for mortality from cancer in relation to one z score^* increase in leg length

Cause of death	No of deaths	Adjusted hazard ratio (95% CI)†	P value for linear trend	Adjusted hazard ratio (95% CI)‡	P value for linear trend
All causes	174	0.96 (0.80 to 1.16)	0.69	1.00 (0.83 to 1.21)	0.98
Cancer:
All cancers¶	53	1.37 (0.94 to 1.99)	0.10	1.43 (0.97 to 2.10)	0.07
Related to smoking§	24	1.02 (0.62 to 1.68)	0.93	1.08 (0.66 to 1.78)	0.75
Not related to smoking	29	1.74 (1.04 to 2.93)	0.04	1.78 (1.02 to 3.01)	0.04
Not dependent on sex hormones	17	1.38 (0.70 to 2.75)**	0.35	1.47 (0.71 to 3.01)	0.30
Dependent on sex hormones‡‡	12	2.29 (1.09 to 4.83)	0.03	2.26 (0.99 to 5.16)	0.05

^*1 unit z score approximates to 3-4 mm. 

^†Adjusted for age at measurement, sex, and trunk length and stratified by survey centre. Standard errors adjusted to account for within family clustering effects. 

^‡Adjusted for age at measurement, sex, trunk length, childhood and adult socioeconomic circumstances and stratified by survey centre. Standard errors adjusted to account for within family clustering effects. 

^¶International classification of diseases, ninth revision (ICD-9) codes 140-209. 

^§ICD-9 codes 140, 141, 143-9, 150, 157, 160-3, 188-9. 

^**Exclusion of five cancers whose primary site was not specified increased adjusted hazard ratio to 1.69 (95% CI 0.81 to 3.51) (P=0.16). 

^‡‡ICD-9 codes 174, 179, 182-7. 

Editorial by Albanes