In Canada, colorectal cancer (CRC) is the second leading cause of death from cancer in men and women combined. In 2011, an estimated 22,200 Canadians will be diagnosed with colorectal cancer and 8,900 will die from it 1 . Genetics research found that high-penetrance mutations account for a small proportion of all CRC cases, with low-penetrance mutations accounting for much of the predisposition, which indicated that 'sporadic' CRC contributed to a large amount of CRC-associated morbidity. 2 . Immigrants rapidly acquire the incidence rates of the host country, suggesting that environmental factors play a crucial role in CRC development 3 4 .

Diet is regarded as one of the most important environmental factors associated with CRC risk 5 6 , particularly when diet is considered in the context of other energy balance indicators such as body size, physical activity, and alcohol or tobacco intake 7 8 9 10 . Several case- control studies have demonstrated a higher risk of colorectal cancer with increased total energy intake 9 11 12 13 14 15 . Thus, excess intake of any of the important energy-supplying macronutrients in the diet (e.g., protein, fat and carbohydrate) could contribute to a higher risk of CRC. However, the question of whether or not individual energy-supplying macronutrients, independent of their contribution to energy intake, are related to CRC risk remains controversial.

A Swiss case-control study claimed that energy intake is directly related to colorectal cancer risk, and that different types of fat may have different roles in colorectal carcinogenesis 16 . Specifically, the risk of CRC increased with total energy intake: consumption of saturated fats appeared a borderline significant relation to CRC risk, while monounsaturated and polyunsaturated fats showed significant inverse trends. In contrast, another case-control study reported that intakes of total fat, saturated fat, and monounsaturated fat were not related to risk of colon cancer, whereas the percent of energy from protein was associated with a risk reduction 14 . Non-red meat sources of animal protein, often derived from low-fat dairy products, fish and poultry sources, have been typically associated with lower risks of CRC 17 18 19 20 . A recent comprehensive report from the World Cancer Research Fund and the American Institute for Cancer Research concluded that total energy intake has no simple relationship with CRC risk, but its effect may be dependent on other factors, such as physical activity. As well, that data were inconsistent for carbohydrates, cholesterol and proteins 21 .

Given the high incidence rate of CRC in Newfoundland and Labrador (NL) and Ontario (ON), it is of great public health importance to identify modifiable risk factors for CRC, including diet. We investigated the associations of total energy intake, macronutrient intake and alcohol in data from a large population based case-control study conducted in NL and ON.

Table 1 shows the distribution of CRC cases and controls according to age, sex, province of residence, BMI and other selected variables. The study sample included 1760 cases (488 from NL and 1272 from ON) and 2481 controls (651 from NL and 1830 from ON) with average response rates of 65.0% and 53.5% in cases and controls, respectively. Among all the participants, cases were slightly younger than controls, and 55.7% of the cases were aged 60 years or older. Cases and controls had similar sex distribution. However, cases were more likely to be obese, to be either physically inactive or extremely physically active, and to have family history of CRC. Controls more often reported regular use of NSAID, higher education and income, and more frequent colon screening procedures.

The mean daily intakes of total energy, macronutrients and alcohol among cases and controls are shown in Table 2. Cases reported significantly higher intakes of total energy, percentage of calories from total fat, percentage of calories from saturated fat and cholesterol (all p < 0.05) than controls. Controls had higher intakes of carbohydrate and total dietary fibre compared with cases (all p < 0.05).

Table 3 gives the ORs and corresponding 95% CI of CRC according to quintile intakes of macronutrients associated food components. After adjusting potential confounders, a significant increased risk of CRC was observed with increasing total energy intake (OR = 1.56 in the highest versus the lowest quintile of intake; 95% CI: 1.21-2.01), which the trend test was also significant (p-trend = 0.02). CRC risk was inversely related to intakes of carbohydrate (OR = 0.81, 95%CI: 0.63-1.00, p-trend = 0.05) and total dietary fibre (OR = 0.84, 95% CI: 0.67-0.99, p-trend = 0.04), both of them shown significant results in trend tests. An inverse association also emerged for intake of protein (OR = 0.85, 95%CI: 0.69-1.00), but the trend test was not significant (p-trend = 0.06). Neither association analysis nor trend test was significant for total fat, alcohol, saturated fatty acids, monounsaturated fatty acids, polyunsaturated fatty acids, cholesterol. We also evaluated percentage of kilocalories from macronutrients in relation to the risk of CRC and obtained similar results (data not shown).

The relationship between intakes of total energy, protein, carbohydrate and dietary fibre with CRC risk was further examined in strata of various covariates. No substantial heterogeneity was observed in separate strata of sex; age(≤ 60, > 60 years); BMI(< 25, ≥ 25 kg/m2); physical activity(< 22.4, ≥ 22.4 METs/week); family history of CRC(no, yes); reported colon screening procedure(no, yes); NSAID use(no, yes); education attainment(lower, higher); household income(lower, higher), and total energy intake(≤ 2109.3, > 2109.3 kcal/day) (data not shown).

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 36 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 41 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 48 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. 14 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 50 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 18 . 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 55 .

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 56 , whereas a survival analysis reported high carbohydrate was strongly elevate survival of CRC 57 .

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. 12 .

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 16 . 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 61 , 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 34 . To the extent that, this adjustment also reduces between-person variation due to over- or underreporting of food intakes 34 . 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.

In conclusion, findings of our large population-based case-control study of CRC conducted in two Canadian provinces provide further evidence that diets high in energy may increase the risk, whereas diets high in protein, fibre, and carbohydrate may reduce the risk of CRC. These results underline the importance of some aspects of total energy and macronutrients and consequently the potential for prevention through dietary changes.

CRC: colorectal cancer; NL: Newfoundland and Labrador; ON: Ontario; NFCCR: Newfoundland Familial Colorectal Cancer Registry; OFCCR: Ontario Familial Colorectal Cancer Registry; C-CFR: Colon-Cooperative Family Registry; ICD: international classification of disease; FHQ: family history questionnaire; PHQ: personal history questionnaire; FFQ: food frequency questionnaire; NSAID: nonsteroidal anti-inflammatory drug; BMI: body mass index; HRT: hormone replacement therapy; METs: metabolic equivalent hours; RDA: recommended dietary allowances; OR: odds ratios; 95%CI: 95% confidence interval.

The authors declare that they have no competing interests.

The work presented here was carried out in collaboration between all authors. ZS, the principal investigator, performed the data analyses and drafted the manuscript; LL conducted the literature search and revision of this article; PPW conceived of the study, participated in its conceptualization, design and coordination. JZ, SB, BR, ED, JRM, PTC, PSP, the co-project investigators, contributed to the execution of this study at various stages. All authors read and approved the final manuscript.