Our present case-control study, rather large among investigations of diet and CRC to-date, showed that intakes of total energy were significantly positively associated with risk of CRC, whereas inverse associations were seen with intakes of protein, carbohydrate, and dietary fibre. Intakes of total fat, fatty acids, cholesterol and alcohol were unrelated to the risk.
Our study observed a direct association between total energy intake and the risk of CRC, confirming results from several previous studies of other populations [11, 15, 35–37]. Studies of potential mechanisms revealed that revealed the evidence that caloric restriction reduces cancer incidence in rodents [38, 39] as well as colorectal cell proliferation in humans [40–42]. Our study found that CRC risk was positively related to percentage of calories consumed as alcohol but was not related with energy-adjusted total alcohol intake. These results suggest that alcohol, independent of its contribution to energy, may not be associated with CRC risk. Energy intake can be responsible for glycemic overload and a compensatory increase of serum insulin and related insulin growth factor-1 (IGF-1). IGF-1 is a promoter of tumor cell growth in vitro [43, 44], and it may expose colonic and rectal cells to a proliferative stimulus [45, 46]. Diabetes has also been related to increased CRC risk [47–49]. Additionally, higher energy intake, in the absence of compensatory energy expenditure leads to excess body weight, which in-turn, is an established risk factor for CRC (Campbell et al., 2007 CEBP; Campbell et al., 2010, JNCI, both studies used these data).
Williams et al.  found that the risk of CRC moderately decreased with an increase of protein intake (OR = 0.53, 95% CI = 0.37-0.77). Consistent with Williams' results, findings of our study showed that CRC risk was inversely associated with protein intake (OR = 0.85, 95%CI: 0.69-1.00) and percentage of calories from protein intake (OR = 0.76, 95%CI: 0.61-0.96). In all kinds of food, meat, fish, soy and eggs were the main sources of protein intake. In our study population, red meat intake was moderate (about 4 servings/week) with similar proportions (approximately 20%) of proteins derived from red meat, dairy products and the combination of white meat and fish. Meta-analyses of meat consumption and CRC risk have concluded that red meat and processed meat increase the risk of CRC  and that processed meat may be a stronger risk factor than fresh red meat [51, 52]. Thus, non-red meat sources of animal protein may have a beneficial influence . Moreover, several previous studies have consistently found inverse associations with high protein foods (dairy products, white meat, fish and poultry) or with non-red meat protein [14, 53, 54]. In our study, a clearer inverse association with protein may have emerged if non-red meat protein sources were analyzed separately from red meat. A possible explanation for a protective effect of protein is that low intakes of methionine may contribute to DNA methylation abnormalities, which appear to be important in the initiation and progression of colon cancer .
Carbohydrate intake was shown to be inversely related to CRC risk in our study. Compared with participants in the lowest quintile of carbohydrate consumption, those in the highest quintile were 19% less likely to develop CRC. There is limited and inconsistent epidemiologic evidence for a relationship between carbohydrate intake and CRC risk. A study conducted among Chinese in North America found that total carbohydrate consumption was associated with increased risk of CRC , whereas a survival analysis reported high carbohydrate was strongly elevate survival of CRC .
We also observed an inverse association with fibre intake, which was in agreement with several previous studies [15, 53, 58, 59]. In our study, a 16% reduced risk of CRC was observed among participants with higher intake of fibre. A beneficial effect from fibre may arise by several mechanisms, including accelerated the fecal transit time, dilution of colonic contents. .
We found no evidence of any substantial effect of the intake of total fat, saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids or cholesterol on risk of CRC in the present study. These results are consistent with most previous studies [14, 15, 18, 60], but not all . However, our finding again suggests that fat, independent of its contribution to energy, may not be associated with CRC risk. Although there exists a sensible biological rationale for the possible involvement of fat in colorectal carcinogenesis , it appears that if fat is indeed involved, the mechanism must be more complex than that which would be implied by a simple empirical association with daily fat intake per day. It could involve foods or some complex interaction amongst nutrients or other food components. Discussion of such potential biological mechanisms is beyond the scope of the present paper.
Consideration must be given to the potential limitations in the present study that may have influenced the observed associations. Firstly, as in most case-control studies, potential recall and selection biases are possible. Since exposure information was collected after diagnosis, differential recall between cases and controls would bias results; in particular, cases may recall dietary exposures differently from controls because of the presence of illness or symptoms. Controls may have agreed to join this study because of an interest in health and may therefore have healthier dietary and physical activity habits, a pattern that may exaggerate differences with the cases beyond what might have been seen with truly comparable controls.
Secondly, by design, cases and controls had similar sex distributions; however, even though analyses were age-adjusted, it may be noted that cases and controls were not well matched according to age group. Estimates of nutrient intakes from a FFQ are not precise and there is always the potential for measurement error. Although the original FFQ used in this study has been validated [28, 29], this questionnaire requires further evaluation because it was originally developed for the Hawaiian and Californian populations which may be different from people residing in NL and ON. FFQ used in NL has been adapted to include regional foods in NL; however, OFCCR used the original FFQ that has not been adapted. Thus, a sub-study will be necessary to assess the level of agreement between the FFQ used by the OFCCR and the FFQ that was previously developed specifically for Canadian populations. Finally, it is also possible that the 1-2 year referent period on which dietary data were based is insufficient if more remote diet (eg. 5-10 yrs) has a stronger influence on CRC risk.
This study had a number of strengths. The large sample size allowed us to observe associations that might be undetectable in smaller studies. More importantly, the previous findings about the protective effects of macronutrients were confined to a specific study population, which makes it difficult to generalize the results. In this study, we conducted pooled analyses of the population of two Canadian provinces to investigate the associations of total energy, macronutrients, alcohol and CRC risk. Furthermore, nutrient intakes were adjusted for total energy intake. The use of calorie-adjusted values in multivariate models will often overcome the problem of high co-linearity frequently observed between nutritional factors . To the extent that, this adjustment also reduces between-person variation due to over- or underreporting of food intakes . The relationships of total energy, macronutrients, alcohol and CRC risk may differ appreciably by several factors, so we controlled for a wide range of potential confounding factors using multivariate logistic regression models. Additionally, results of the consistent findings in separate strata for total energy, protein, carbohydrate and dietary fibre would argue against multiple comparisons as an explanation for these associations. Although some random misclassification of diet is likely, non-differential misclassification generally tends to bias the risk estimates toward the null.