Body size, as well as diet, has been investigated in relation to the occurrence of the cancer. Height is affected not only by genetic factors, but also by nutritional condition during childhood and adolescent. Some studies (Albanes et al, 1988; Chute et al, 1991; Bostick et al, 1994; Giovannucci et al, 1995; Robsahm and Tretli, 1999; Smith et al, 2000), but not all (Okasha et al, 2000), have reported a positive association between height and colon cancer. Results of studies on body mass index (BMI) in relation to colorectal cancer have been also inconsistent. There are relatively few prospective studies of colorectal cancer among Japanese (Nomura et al, 1985; Kono et al, 1987; Hirayama, 1989; Akiba, 1994), and only Nomura et al (1985) investigated BMI, which was positively associated with an increased risk for colon cancer among Japanese male immigrants to Hawaii. We have therefore conducted a prospective study, among Japanese people, of colorectal cancer in relation to height, BMI, alcohol consumption, and smoking habit.

The 1992 questionnaire sought details of current height, weight, and weight at age 21 years. The intraclass correlation coefficients between self-reported and measured height and weight in a subsample were 0.93 and 0.97 in both sexes, respectively.

Diet was assessed by a semiquantitative food–frequency questionnaire that contained 169 food items, covering average consumption frequencies and serving sizes of selected food items during the previous year. Individual nutrient intake was estimated based upon the frequency of intake and portion size using the Standard Tables of Food Composition in Japan, 5th edition, published by the Science and Technology Agency of Japan. Details including results of validity tests are described elsewhere (Shimizu et al, 1999).

The questions on alcohol use included six types, that is, sake, beer, light beer, shochu (distilled from sweet potatoes, rice, or buckwheat), wine, and hard liquor. For each item, the questionnaire included nine frequency categories (never/less than once a month; once a month; twice or three times a month; once a week; twice or three times a week; four to six times a week; once a day; twice a day; more than four times a day) and the number of cups, glasses, and bottles consumed. The amount of ethanol was calculated in grams using the Standard Tables mentioned above. The correlation coefficients that compare alcohol consumption estimated from the questionnaire with 12 one-day diet diaries at about 1-month intervals over 1 year were 0.72 and 0.64, for men and women, respectively (data not published).

Years of smoking and the number of cigarettes smoked each day were reported. Those who had smoked a total of 20 or more packs of cigarettes in their lifetime were defined as smokers.

Physical activity was based on average hours per week spent performing various kinds of activities; details, including the results of the validity tests, are described elsewhere (Suzuki et al, 1998; Shimizu, 2001).

Participants who did not report their height (564 men, 709 women) were excluded from the study, as were those who reported cancer other than nonmelanoma skin cancer (173 men, 532 women) or colorectal adenoma (281 men, 208 women) at baseline. We could not obtain information about diagnosis for 13 men and 15 women who were known to have died of colorectal cancer; thus, these men and women were also excluded from the analysis. The subjects totalled 29051(13392 men and 15659 women).

There were 198 (105 men, 93 women) with colon and 97 (56 men, 41 women) with rectal cancer, all diagnosed histologically at two major hospitals in Takayama City during the follow-up from 1 January 1993 to 31 December 2000. Compared to the number of colorectal cancer occurred in the city according to the annual reports issued by the prefecture government (Hida Public Health Center, 2000), the two hospitals covered about 90% of colorectal cancers in the city each year. Details of subjects who moved away from the city during the study period were obtained from the residential registers, namely, 629 (4.7%) men and 508 (3.2%) women.

Positive relations between height and colon cancer are seen for men, but were weaker and nonsignificant for women (Table 2). For men, the relative risks (RRs) of colon cancer from the shortest to the middle and tallest height categories were 1.75 (95% confidence interval, CI=1.07–2.85) and 2.13 (95% CI=1.26–3.58), respectively, with a P for trend of 0.004, after controlling for age, BMI, alcohol consumption, smoking, and education. For women, the point estimates of colon cancer risk for the middle and the tallest categories of height were moderately but not significantly elevated. Further adjustments for physical activity did not substantially alter the results (RRs of colon cancer for the tallest height tertile compared with the shortest tertile adjusted for age, BMI, alcohol consumption, smoking, education, and physical activity=2.07 (95% CI=1.21–3.52) and 1.56 (95% CI=0.85–2.88), for men and women, respectively). We further performed adjustments for nutrient intake, such as total energy and dietary fibre. The results were not modified by the adjustments.

Table 3 shows that BMI was positively associated with the risk of colon cancer for men, whereas the association for women was weak, controlling for age, height, alcohol consumption, smoking, and education. There was no significant association between height and the risk of rectal cancer in both sexes. Further adjustment for physical activity did not substantially alter the results. Our additional analysis for BMI calculated using self-reported weight at age 21 years did not demonstrate any associations or trends (data not shown).

The associations between total alcohol intake and colorectal cancer are shown in Table 4. Relative risks of colon cancer for individuals in the highest alcohol consumption compared with abstainers were 2.52 (95% CI=1.00–6.38) and 1.73 (95% CI=0.98–3.05) for men and women, respectively, controlling for age, height, BMI, smoking, and education. However, a significant positive relation with rectal cancer risk was restricted to women. In the analysis by type of beverage, sake consumption was significantly associated with colon cancer risk in men (the RR from the abstainers to the highest sake consumption category=1.91; 95% CI=1.10–3.32).

Six other prospective studies (Albanes et al, 1988; Chute et al, 1991; Bostick et al, 1994; Giovannucci et al, 1995; Robsahm and Tretli, 1999; Smith et al, 2000), but not all, support our finding that being tall significantly elevates the risk of colorectal cancer. One possibility is that taller people may have longer intestines (Hirsch et al, 1956) and have a greater rate of cell division within the tissue (Albanes and Winick, 1988); thus, more colon cells may be at risk. Greater exposure to mitogenic factors, such as growth hormone, insulin, insulin-like growth factors, and sex steroids, could also result in increased cancer risk. Height has been related to elevated risk of several specific cancers, including breast (Swanson et al, 1988) and prostate (Smith et al, 2000). Animal experiments have shown that low-energy diets from early age shorten overall animal length and reduce cancer risk (Kritchevsky, 1995). In humans, wartime food deprivation for pubescent women was associated with lower breast cancer rates than in younger and older cohorts (Tretli and Gaard, 1996). Furthermore, an association has been reported between higher caloric intake during childhood and higher rates of cancer (Frankel et al, 1998). Height is partly determined by total caloric intake during childhood and adolescence; so its association with colon cancer may indicate that total caloric intake during childhood is relevant for a later colon cancer risk.

Evidence has long suggested that people with excess body weight are at higher risk of colorectal cancer (Lew and Garfinkel, 1979; Nomura et al, 1985; Albanes and Taylor, 1990). Our study, as well as others (Garfinkel, 1985; Wu et al, 1987; Russo et al, 1998; Robsahm and Tretli, 1999; Murphy et al, 2000), shows a positive association between the risk of colon cancer of BMI in men but not in women. Other case–control studies (Slattery et al, 1997; Caan et al, 1998) and a cohort study (Ford, 1999) have shown a significant positive association between the risk of colon cancer and BMI in both sexes, although this was weaker among women in the two case–control studies. Central obesity, which may increase colon cancer risk by acting as a tumour-growth promoter or mitogen (Giovannucci, 1995; Bjorntorp, 1991), is more common among males. Thus, BMI may simply be a more accurate indicator of central obesity for men than for women (Murphy et al, 2000). Another possibility is a protective effect of oestrogen.

Our study also shows a significant positive dose–response relation between alcohol consumption and colon cancer risk in both sexes, although not significant in women as found in other studies. In a prospective study in California, the RR of colorectal cancer was 2.42 (95% CI=1.3–4.5) in men who drank more than 30mlday⁻¹ compared to nondaily alcohol drinkers, while the RR among women who consumed more than 30mlday⁻¹ relative to the same group was weaker and not significant (RR=1.45, CI=0.8–2.6) (Wu et al, 1987). Similar results were obtained when the cases of rectal cancer were omitted. Klatsky et al (1988) prospectively studied and found a positive relation between the risk of colon cancer and total alcohol intake in both sexes, although the results in women but not men were statistically significant. On the other hand, a meta-analysis (Longnecker et al, 1990) including both prospective and case–control studies showed that the RR of colorectal cancer was 1.10 (95% CI=1.05–1.14), and the association did not vary according to gender or site within the large bowel. In almost all case–control studies, the use of hospital controls, among which proportions of alcohol-related disease were high, may have biased the results.

There was no association between BMI and rectal cancer risk in both the sexes. Similar finding was obtained in a case–control study reported by Dietz et al (1995), that body weight was positively associated with colon cancer but not with rectal cancer. Positive association between BMI and rectal cancer has been reported in some studies (Phillips and Snowdon, 1985; Russo et al, 1998), but not in others (Dietz et al, 1995; Howe et al, 1997). A significant association between alcohol consumption and rectal cancer risk has been reported (Klatsky et al, 1988; Hirayama, 1989), but in our study, the association was weak in men. We must mention that the number of cases of rectal cancer might be too small to have meaningful analyses.

Recent prospective studies have suggested a positive association between long-term and heavy smoking and colon cancer (Giovannucci et al, 1994; Hsing et al, 1998; Terry et al, 2001). In our present study, no significant association was observed in relation to the quantity of cigarettes smoked except for rectal cancer risk in men.

A major advantage of our study is that it is population based and prospective, thereby minimising recall bias. Incidence rather than mortality can avoid bias because of a progress of medical treatment as well as effect of disease. Furthermore, the data were analysed adjusting for physical activity, nutrition, and years of education.

However, several limitations must be considered. Firstly, height was self-reported and not measured. Although the intraclass correlation coefficient between self-reported height and measured height in a subsample was relatively high, self-reported information may result in misclassification. However, since the height information was collected before the development of cancer, any misclassification would be nondifferential with respect to the disease. Such random misclassification would tend to attenuate RR estimates, and, thus, would not explain the association between height and colon cancer.

Secondly, height was reported at the baseline. Stature has been reported to be at its peak at the end of the third decade of life (Friedlander et al, 1977). Our subjects were 35 years old or over, so their height at the baseline might have been already reduced by ageing or disease and may not have represented their highest. However, age was adjusted for in our analysis, while additional adjustment for pre-existing disease, such as hypertension and diabetes mellitus, did not substantially alter the results.

Another limitation is incomplete detection of colorectal cancers, although details from the two study hospitals appear to cover around 90% of the colorectal cancers in the city reported by the prefectural cancer registry. Some colorectal cancer patients must therefore have been classified as subjects without colorectal cancer. However, it is unlikely that taller or heavier patients were selectively included in the present study. In addition, the study was so large that the effect of misclassification of true cases was minimal.