Epidemiological studies of dairy products and colorectal cancer risk

A large number of prospective cohort and case–control studies have investigated the link between dairy foods and CRC risk. One of the problems with assessing the evidence is the considerable variation in how consumption data were collected, with some studies reporting overall dairy product consumption, while others report categories such as milk, butter, cheese and fermented milk products. Generally, results from case–control studies are heterogeneous and do not provide evidence of an association between total intake of dairy products and CRC risk⁽Reference Norat and Riboli¹⁵⁾. Because of the abundance of prospective studies that have investigated the relationship between dairy intake and risk of CRC and because results from case–control studies have been reviewed in detail elsewhere⁽Reference Norat and Riboli¹⁵⁾, the present review will focus mainly on prospective studies. The prospective studies⁽Reference Park, Murphy, Wilkens, Nomura, Henderson and Kolonel^12, Reference Phillips^16–Reference Shin, Li, Shu, Yang, Gao and Zheng³⁹⁾ that have assessed the relationship between dairy foods and CRC risk are summarised in Table 1. Data from these studies point to an inverse association, although generally the relationships are statistically non-significant. However, two recent studies, the Cohort of Swedish Men⁽Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾ and the US Multiethnic Cohort Study⁽Reference Park, Murphy, Wilkens, Nomura, Henderson and Kolonel¹²⁾, have shown significant reduction in risk (54 and 20 %, respectively) with the highest intake of total dairy products.

Table 1 Prospective studies of dairy and/or dietary calcium and vitamin D intake and risk of colorectal cancer (CRC)

RR, relative risk; EPIC, European Prospective Investigation into Cancer and Nutrition.

* 1 ounce serving = 28·35 g serving; 4 ounce serving = 113·4 g serving; 8 oz glass, 237 ml; 1000 kcal = 4184 kJ.

† Adjusted RR for the highest dietary intake compared with the lowest dietary intake.

‡  < 127 v. >373 IU/d ( < 3·2 v. >9·3 μg/d).

§  < 22 v. ≥ 154 IU/d (0·6 v. ≥ 3·9 μg/d).

∥  < 85 v. >280 IU/d ( < 2·1 v. >7·0 μg/d).

¶  < 120 v. >550 IU/d ( < 3·0 v. >13·8 μg/d).

**  < 224 v. >476 IU/d ( < 5·6 v. >11·9 μg/d).

†† 113 v. 367 IU/d ( < 2·8 v. >9·2 μg/d).

‡‡  < 90 v. >240 IU/d ( < 2·3 v. >6·0 μg/d).

§§  < 125 v. ≥ 333 IU/d ( < 3·1 v. >8·3 μg/d).

‖  < 31 v. ≥ 96 IU ( < 0·8 v. >2·4 μg).

Milk is the dairy product that shows the most consistent relationship with CRC risk. Of the cohort studies to date that have investigated this relationship, most showed non-significant decreased risks of CRC with increasing milk intake⁽Reference Park, Murphy, Wilkens, Nomura, Henderson and Kolonel^12, Reference Phillips and Snowdon^17, Reference Ursin, Bjelke, Heuch and Vollset^19, Reference Kampman, Goldbohm, van den Brandt and van't Veer^22, Reference Kearney, Giovannucci, Rimm, Ascherio, Stampfer, Colditz, Wing, Kampman and Willett^23, Reference Singh and Fraser^27, Reference Pietinen, Malila, Virtanen, Hartman, Tangrea, Albanes and Virtamo^29, Reference Jarvinen, Knekt, Hakulinen and Aromaa^30, Reference McCullough, Robertson, Rodriguez, Jacobs, Chao, Carolyn, Calle, Willett and Thun^33, Reference Sanjoaquin, Appleby, Thorogood, Mann and Key^34, Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾. Most studies have considered all types of milk together, but where these have been separated there are differences between low-fat v. whole milk. In the US study on an Adventist population, consumption of skimmed milk, but not whole milk, had a protective effect (relative risk (RR) 0·78 for skimmed milk v. RR 1·04 for whole milk)⁽Reference Singh and Fraser²⁷⁾. A few case–control studies (reviewed by Norat & Riboli⁽Reference Norat and Riboli¹⁵⁾) have shown similar trends although none of the relationships were statistically significant. In a recent case–control study, whole milk consumption was positively associated with cancer of the rectum (OR 1·22; 95 % CI 1·03, 1·44) while skimmed milk consumption was inversely associated with cancers of the colon (OR 0·84; 95 % CI 0·73, 0·97) and rectum (OR 0·76; 95 % CI 0·64, 0·91)⁽Reference Gallus, Bravi, Talamini, Negri, Montella, Ramazzotti, Franceschi, Giacosa and La Vecchia⁴⁰⁾. However, further studies are required to determine any effect of milk fat on CRC risk.

The pooling of data from prospective studies has revealed significant relationships between milk intake and CRC risk. Norat & Riboli⁽Reference Norat and Riboli¹⁵⁾ analysed data from eleven cohorts and showed a RR of 0·80 (95 % CI 0·68, 0·95) for the highest v. the lowest milk intake. Similarly, Cho et al. ⁽Reference Cho, Smith-Warner and Spiegelman¹¹⁾ analysed data from ten cohorts from five countries, including 534 536 individuals, of whom 4992 were diagnosed with CRC at follow-up. A significant protective effect of milk was found; individuals that consumed more than a glass of milk per d ( ≥ 250 g) had a 15 % reduced risk of developing CRC (RR 0·85; 95 % CI 0·78, 0·94) compared with those that consumed < 70 g/d. Each 500 g/d increase in milk intake reduced the risk of CRC by 12 % (RR 0·88; 95 % CI 0·82, 0·95). The inverse association was consistent across studies and sex. The main strength of this meta-analysis is the pooling of individual level prospective data. All included studies used validated diet assessment methods, minimising the possibility of an incorrect recording of the actual intake of dairy products. However, although Cho et al. ⁽Reference Cho, Smith-Warner and Spiegelman¹¹⁾ adjusted the multivariable RR for energy, alcohol, red meat and dietary folate intakes, they did not adjust for other dietary variables related to CRC risk, such as dietary fibre and fruit and vegetable intakes; therefore it is possible that some residual confounding remains.

Data for other dairy products generally do not support a protective effect against CRC. In the meta-analysis by Cho et al. ⁽Reference Cho, Smith-Warner and Spiegelman¹¹⁾ dairy foods such as cottage or ricotta cheese, butter, cream and ice cream (which were measured in at least five out of the ten studies) also showed inverse associations with CRC risk, although results were not statistically significant. Data on fermented milk products such as yoghurt suggest no relationship with CRC risk⁽Reference Kampman, Goldbohm, van den Brandt and van't Veer^22, Reference Kearney, Giovannucci, Rimm, Ascherio, Stampfer, Colditz, Wing, Kampman and Willett^23, Reference Jarvinen, Knekt, Hakulinen and Aromaa^30, Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾. For cheese intake, RR range from 0·68 to 1·35 and in most studies the RR was greater than 1·0, although none of these relationships were statistically significant⁽Reference Phillips and Snowdon^17, Reference Kampman, Goldbohm, van den Brandt and van't Veer^22, Reference Kearney, Giovannucci, Rimm, Ascherio, Stampfer, Colditz, Wing, Kampman and Willett^23, Reference Singh and Fraser^27, Reference Jarvinen, Knekt, Hakulinen and Aromaa^30, Reference Sanjoaquin, Appleby, Thorogood, Mann and Key^34, Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾. Some studies⁽Reference Singh and Fraser^27, Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾ have attempted to separate hard cheese from soft cheeses such as ricotta and cottage cheese, with the latter appearing to have a protective effect; however, the relationships were not significant and it is difficult to draw any reasonable conclusions from such few studies. A recent meta-analysis on three cohort studies⁽Reference Kearney, Giovannucci, Rimm, Ascherio, Stampfer, Colditz, Wing, Kampman and Willett^23, Reference Singh and Fraser^27, Reference Lin, Zhang, Cook, Manson, Lee and Buring³⁷⁾ conducted for the second expert report on diet and the prevention of cancer⁽⁸⁾ showed a RR of 1·14 (95 % CI 0·82, 1·58) per serving of cheese per d, with low heterogeneity, although it is not clear which kind of cheese was included in the analysis. The Panel concluded that there is some (albeit limited) evidence that cheese is a cause of CRC⁽⁸⁾. Cheese is high in saturated fats, which have been shown to increase insulin production and expression of insulin on colonic cells⁽Reference Kiunga, Raju, Sabljic, Bajaj, Good and Bird⁴¹⁾. Saturated fats may also induce the expression of some inflammatory mediators associated with the cancer process⁽Reference Giugliano, Ceriello and Esposito⁴²⁾. However, it is difficult to reconcile this with data on the protective effects of dietary Ca, which is abundant in cheese⁽⁴³⁾.

Although there are inconsistencies between studies, most of the evidence points to a decreased risk of CRC with increasing intake of Ca, both dietary and total (including supplements)⁽Reference Park, Murphy, Wilkens, Nomura, Henderson and Kolonel^12, Reference Martinez, Giovannucci, Colditz, Stampfer, Hunter, Speizer, Wing and Willett^24, Reference Zheng, Anderson, Kushi, Sellers, Greenstein, Hong, Cerhan, Bostick and Folsom^28, Reference Terry, Baron, Bergkvist, Holmberg and Wolk^31, Reference Wu, Willett, Fuchs, Colditz and Giovannucci^32, Reference Flood, Peters, Chatterjee, Lacey, Schairer and Schatzkin^35, Reference Kesse, Boutron-Ruault, Norat, Riboli and Clavel-Chapelon³⁶⁾. Several meta-analyses have quantified the relationship between Ca intake and CRC risk^(8, Reference Cho, Smith-Warner and Spiegelman^11, Reference Bergsma-Kadijk, van't Veer, Kampman and Burema⁴⁴⁾. An early meta-analysis of eight cohort and sixteen case–control studies showed that the highest category of Ca intake was associated with a 14 % decrease in risk of CRC, although there was considerable heterogeneity between studies, which made reaching a conclusion difficult⁽Reference Bergsma-Kadijk, van't Veer, Kampman and Burema⁴⁴⁾. In the meta-analysis by Cho et al. ⁽Reference Cho, Smith-Warner and Spiegelman¹¹⁾ individuals consuming the highest category of dietary Ca had a significantly lower risk of developing CRC (RR 0·86; 95 % CI 0·78, 0·95). A meta-analysis on ten cohort studies conducted for the second expert report on diet and the prevention of cancer⁽⁸⁾ showed a RR of 0·98 (95 % CI 0·95, 1·00) per 200 mg/d, with low heterogeneity. Both meta-analyses suggested a threshold effect of Ca intake; no further protection was observed at intakes >1000 mg/d.

The evidence for Ca suggests that the protective effect of milk or total dairy may be due in part to its Ca content, since dietary Ca can be considered a marker of dairy intake in populations that consume high amounts of milk and dairy products⁽Reference Weinberg, Berner and Groves⁴⁵⁾. Most of the published cohort studies are from dairy-consuming countries such as the USA, Holland, Finland and Sweden, so it is reasonable to assume that a large proportion of dietary Ca comes from dairy products. Recent data from the ongoing Cohort of Swedish Men trial⁽Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾, a high-quality study established in 1997 to study lifestyle–disease interactions, also reported a decrease in risk of CRC (RR 0·68; 95 % CI 0·51, 0·91) in subjects with the highest Ca intake. This study also confirmed previous findings that milk was the individual dairy food with the strongest influence on risk but the analysis provided some evidence that the risk reduction from dairy foods may not be solely due to their high Ca content⁽Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾. First, Ca and total dairy consumption had different effects on risk when the various parts of the colon were considered; for example, Ca reduced the risk for the rectum while total dairy intake reduced the risk for the proximal colon, distal colon and rectum⁽Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾. Second, even though Ca intake was closely associated with total dairy or milk intake, when the analysis was controlled for Ca intake, the influence of milk or dairy intake on risk was diminished but not eliminated⁽Reference Larsson, Bergkvist, Rutegård, Giovannucci and Wolk³⁸⁾. These findings suggest that other milk-associated factors may be partially responsible for the protective effects and these will be reviewed under the section on mechanisms.

Epidemiological studies of calcium and dairy products and adenoma risk

Most case–control studies on colorectal adenoma risk have investigated intake of dietary and supplemental Ca rather than individual dairy products⁽Reference Martinez, McPherson, Annegers and Levin^46–Reference Miller, Keku, Satia, Martin, Galanko and Sandler⁵¹⁾. OR for these studies range from 0·50–0·97 and most do not reach statistical significance⁽Reference Tseng, Murray, Kupper and Sandler^47–Reference Peters, McGlynn, Chatterjee, Gunter, Garcia-Closas, Rothman and Sinha^49, Reference Miller, Keku, Satia, Martin, Galanko and Sandler⁵¹⁾ although some do show a protective effect of Ca⁽Reference Martinez, McPherson, Annegers and Levin^46, Reference Peters, Chatterjee, McGlynn, Schoen, Church, Bresalier, Gaudet, Flood, Schatzkin and Hayes⁵⁰⁾. In some studies the protective effect of Ca was present in men but not in women⁽Reference Tseng, Murray, Kupper and Sandler⁴⁷⁾. In other studies the protective effects was mainly due to Ca supplement use; for example, Peters et al. ⁽Reference Peters, Chatterjee, McGlynn, Schoen, Church, Bresalier, Gaudet, Flood, Schatzkin and Hayes⁵⁰⁾ found a decrease in adenoma risk for subjects taking 1200 mg Ca/d compared with non-users (OR 0·74; 95 % CI 0·58, 0·95), after adjustment for vitamin supplement intake as well as other confounders associated with supplement use. Similarly, high Ca (>900 mg/d) and lower fat intakes ( < 34 % of energy intake) were also reported to protect against colorectal adenomas in a US population, an effect that was not observed with high Ca and higher fat intakes ( ≥ 34 % of energy intake)⁽Reference Miller, Keku, Satia, Martin, Galanko and Sandler⁵¹⁾.

Prospective studies⁽Reference Kesse, Boutron-Ruault, Norat, Riboli and Clavel-Chapelon^36, Reference Kampman, Giovannucci, van't Veer, Rimm, Stampfer, Colditz, Kok and Willett^52–Reference Oh, Willett, Wu, Fuchs and Giovannucci⁵⁶⁾ that have investigated Ca intakes in relation to adenoma risk are listed in Table 2. Martinez et al. ⁽Reference Martinez, Marshall, Sampliner, Wilkinson and Alberts⁵⁴⁾ showed that intake above 1068 mg/d v. intake below 698 mg/d was associated with a significant reduction in adenoma recurrence among 1304 male and female participants in the Wheat Bran Fiber trial ofadenoma recurrence (OR 0·56; 95 %CI 0·39, 0·80; P trend = 0·007) and dietary Ca was found to be more protective than supplemental Ca. Kampman et al. ⁽Reference Kampman, Giovannucci, van't Veer, Rimm, Stampfer, Colditz, Kok and Willett⁵²⁾ used data from the Health Professionals Follow-Up Study and the Nurses' Health Study and showed no relationship between a high total Ca intake or dietary Ca intake and risk of adenoma. Kesse et al. ⁽Reference Kesse, Boutron-Ruault, Norat, Riboli and Clavel-Chapelon³⁶⁾ found a trend of decreasing risk of adenoma with increasing Ca intake, although the RR for the highest v. the lowest quartiles of intake was not significant. An analysis of the baseline data on dietary intake and supplement use from the Polyp Prevention Trial showed that there was no association between Ca or dietary vitamin D intake and adenoma recurrence⁽Reference Hartman, Albert and Snyder⁵⁵⁾, although there were inverse associations between Ca and vitamin D supplementation with both single and multiple adenoma recurrence. Oh et al. ⁽Reference Oh, Willett, Wu, Fuchs and Giovannucci⁵⁶⁾ using data from the Nurses' Health Study showed that higher intakes of total Ca were associated with a reduced risk of distal colorectal adenoma.

Table 2 Prospective studies of dairy and/or calcium and vitamin D intake and risk of colorectal adenoma

RR, relative risk; EPIC, European Prospective Investigation into Cancer and Nutrition.

* 1 US fluid ounce = 29·57 ml; 8 US fluid ounces = 237 ml; 1 ounce serving = 28·35 g serving.

† Adjusted RR for the highest dietary intake compared with the lowest dietary intake.

‡ 118 v. 954 IU/d ( < 3·0 v. >23·9 μg/d).

§ 500 v. 909 IU/d ( < 12·5 v. >22·7 μg/d).

∥ 107 v. 587 IU/d ( < 2·7 v. >14·7 μg/d).

Relatively few studies have reported dairy intakes (as opposed to Ca) in relation to colorectal adenoma risk. A case–control study in a Japanese population has reported a protective effect on colorectal adenomas of high dairy intakes in conjunction with other healthy dietary patterns⁽Reference Mizoue, Yamaji, Tabata, Yamaguchi, Shimizu, Mineshita, Ogawa and Kono⁵⁷⁾. Data from the Health Professionals Follow-Up Study and the Nurses' Health Study showed that fermented dairy products and cheese were not associated with adenoma risk, even after adjusting for saturated fat intake⁽Reference Kampman, Giovannucci, van't Veer, Rimm, Stampfer, Colditz, Kok and Willett⁵²⁾ and baseline data from the Polyp Prevention Trial showed that there was no association between the consumption of low- or high-fat dairy products and adenoma recurrence⁽Reference Hartman, Albert and Snyder⁵⁵⁾. An early systematic review of eleven case–control and two cohort studies did not find an association between dairy intake and colorectal adenoma risk⁽Reference Yoon, Benamouzig, Little, Francois-Collange and Tome⁵⁸⁾.

Vitamin D and colorectal cancer risk

Animal studies have shown that dietary Ca and vitamin D status act synergistically to modify CRC risk⁽Reference Harris and Go⁵⁹⁾. Vitamin D is an important component of dairy foods, and Ca and vitamin D are metabolically related since the absorption of Ca in the gut relies on adequate levels of vitamin D. Several prospective studies that have assessed total vitamin D intake and CRC or colorectal adenoma risk have reported non-significant inverse associations (see Tables 1 and 2), although in some studies total vitamin D intake was unrelated to risk⁽Reference Lin, Zhang, Cook, Manson, Lee and Buring^37, Reference Kampman, Giovannucci, van't Veer, Rimm, Stampfer, Colditz, Kok and Willett⁵²⁾. A few studies have reported a dose–response effect⁽Reference Garland, Shekelle, Barrett Connor, Criqui, Rossof and Paul^18, Reference McCullough, Robertson, Rodriguez, Jacobs, Chao, Carolyn, Calle, Willett and Thun^33, Reference Hartman, Albert and Snyder⁵⁵⁾. However, these findings need to be interpreted with caution since dietary sources do not account for total vitamin D status; UV radiation induces vitamin D₃ production in the skin, making sun exposure an important source of the vitamin⁽Reference Millen and Bodnar⁶⁰⁾.

Intervention trials provide some evidence of benefit of high vitamin D intake on adenoma recurrence, although these studies have generally investigated Ca and vitamin D in combination. Hartman et al. ⁽Reference Hartman, Albert and Snyder⁵⁵⁾ found a reduced risk of adenoma and a significant trend in individuals from the Polyp Prevention Trial. Similarly Oh et al. ⁽Reference Oh, Willett, Wu, Fuchs and Giovannucci⁵⁶⁾ also found a reduced risk of distal adenoma in women from the Nurses' Health Study, with P = 0·07 for trend. Of further interest are studies of genetic variations in the vitamin D receptor, such as the vitamin D receptor Fok1 polymorphism, which have been related to the risk of CRC⁽Reference Wong, Seow, Arakawa, Lee, Yu and Ingles^61, Reference Park, Woo, Nam and Kim⁶²⁾. These studies offer independent evidence (free from the effects of confounding associated with the measurement of dietary intakes) that vitamin D is important for human CRC. There is some evidence of interaction between vitamin D receptor polymorphisms and the consumption of dairy products⁽Reference Murtaugh, Sweeney, Ma, Potter, Caan, Wolff and Slattery⁶³⁾, Ca⁽Reference Peters, Chatterjee, McGlynn, Schoen, Church, Bresalier, Gaudet, Flood, Schatzkin and Hayes^50, Reference Wong, Seow, Arakawa, Lee, Yu and Ingles^61, Reference Slattery, Neuhausen, Hoffman, Caan, Curtin, Ma and Samowitz⁶⁴⁾ and vitamin D⁽Reference Kim, Newcomb, Ulrich, Keener, Bigler, Farin, Bostick and Potter⁶⁵⁾ and the risk of colorectal adenoma or cancer in some, though not all⁽Reference Grau, Baron, Sandler, Haile, Beach, Church and Heber⁶⁶⁾, studies. These important interactions between Ca and vitamin D highlight a limitation in the epidemiological studies that have investigated the dairy–CRC link; most have failed to take into account vitamin D status of the individuals taking part in the studies because of the difficulties in trying to estimate the contribution to vitamin D by sunlight⁽Reference Millen and Bodnar⁶⁰⁾.

Evidence from calcium supplementation trials

Several prospective studies have shown that Ca supplements decrease the risk of CRC⁽Reference Bostick, Potter, Sellers, McKenzie, Kushi and Folsom^21, Reference Kampman, Goldbohm, van den Brandt and van't Veer^22, Reference Wu, Willett, Fuchs, Colditz and Giovannucci^32, Reference McCullough, Robertson, Rodriguez, Jacobs, Chao, Carolyn, Calle, Willett and Thun^33, Reference Lin, Zhang, Cook, Manson, Lee and Buring³⁷⁾. In the pooled analysis of ten cohort studies by Cho et al. ⁽Reference Cho, Smith-Warner and Spiegelman¹¹⁾, the reduction in risk was greater when Ca from supplements was included in the analysis (RR 0·78 (95 % CI 0·69, 0·88) v. RR 0·86 (95 % CI 0·78, 0·95) for Ca from food sources). Three prospective studies showed a decreased risk of adenoma recurrence with Ca supplements⁽Reference Hyman, Baron, Dain, Sandler, Haile, Mandel, Mott and Greenberg^53–Reference Hartman, Albert and Snyder⁵⁵⁾. A meta-analysis⁽⁸⁾ on the studies by Martinez et al. ⁽Reference Martinez, Marshall, Sampliner, Wilkinson and Alberts⁵⁴⁾ and Hartman et al. ⁽Reference Hartman, Albert and Snyder⁵⁵⁾ gave a summary effect estimate of 0·91 (95 % CI 0·85, 0·98) per 200 mg Ca/d.

Randomised controlled trials have been conducted with both adenoma and CRC as endpoints. For colorectal adenoma, two trials showed a decreased risk of adenoma with Ca supplements. In the Calcium Polyp Prevention Study Group, Baron et al. ⁽Reference Baron, Beach and Mandel⁶⁷⁾ showed that supplementation with 1200 mg Ca/d for 4 years in 930 subjects with recently resected adenoma was associated with a 19 % reduction in risk of adenoma recurrence (RR 0·81; 95 % CI 0·67, 0·99). A 5-year follow-up after the end of the intervention showed that subjects in the Ca group still had a significantly lower risk of adenoma than those in the placebo group (31·5 v. 43·2 %; adjusted RR 0·63 (95 % CI 0·46, 0·87); P = 0·005)⁽Reference Grau, Baron and Sandler⁶⁸⁾. However, follow-up for a further 5 years did not show any difference between the groups⁽Reference Grau, Baron and Sandler⁶⁸⁾. Bonithon-Kopp et al. ⁽Reference Bonithon-Kopp, Kronborg, Giacosa, Rath and Faivre⁶⁹⁾ showed that supplementation with 2000 mg Ca/d for 3 years also reduced adenoma recurrence in 665 subjects, although the effect was not significant (OR 0·66; 95 % CI 0·38, 1·17). The results of both the above trials combined showed a significant reduction in adenoma recurrence with Ca supplementation (OR 0·74; 95 % CI 0·58, 0·95)⁽Reference Weingarten, Zalmanovici and Yaphe⁷⁰⁾. In another 3-year intervention trial where 1600 mg Ca was given in conjunction with β-carotene (15 mg), vitamin C (150 mg), vitamin E (75 mg) and Se (101 μg) there was no effect on the growth of pre-existing adenomas, although there was a non-significant reduced risk of new adenoma growth in subjects < 60 years of age (mean difference 2·3 mm; 95 % CI 0·26, 4·36)⁽Reference Hofstad, Almendingen, Vatn, Andersen, Owen, Larsen and Osnes⁷¹⁾.

A recent intervention trial involving 36 282 postmenopausal women participating in the Women's Health Initiative showed that 1000 mg elemental Ca/d (as calcium carbonate) in conjunction with 10 μg (400 IU) vitamin D for 7 years had no effect on the incidence of CRC⁽Reference Wactawski-Wende, Kotchen and Anderson⁷²⁾. Overall this was a good-quality study, loss to follow-up was minimal (data were available on 97 % of living participants) and adherence to the intervention was relatively good (70 % of women took ≥ 50 % of their study medication through the 6th year). Serum concentrations of vitamin D at baseline and additional use of Ca and vitamin D supplements by a subgroup of participants did not modify the effect of the intervention on the outcome. However, baseline Ca and vitamin D intakes (1151 mg/d and 9·1μg (367 IU), respectively) were relatively high in this population. This may explain the lack of effect, particularly since pooled data from prospective studies suggest no further effect on CRC risk of Ca beyond this level of intake⁽Reference Cho, Smith-Warner and Spiegelman¹¹⁾; this is in contrast to intervention trials with adenoma recurrence as an endpoint, in which intakes >1000 mg/d reduced the risk of colorectal adenoma⁽Reference Weingarten, Zalmanovici and Yaphe⁷⁰⁾. Another explanation for the lack of effect could be the duration of the intervention and the short follow-up period; CRC takes 10–20 years to develop so a follow-up period longer than 7 years may be necessary to detect any beneficial effects of Ca supplements on CRC risk.

Potential chemopreventive agents in dairy foods

Although Ca has been the main component in dairy products to be investigated for its protective role in CRC, dairy products contain many potential chemopreventive components such as vitamin D, butyric acid, conjugated linoleic acid (CLA), sphingolipids, and probiotic bacteria in fermented products such as yoghurt⁽Reference Norat and Riboli¹⁵⁾. The following section will present a brief review of the proposed mechanisms and evidence for them.

Ca may provide protection against CRC through two mechanisms. In the colon, Ca sequesters secondary bile acids such as deoxycholic acid (a by-product of fat metabolism)⁽Reference Govers, Termont, Lapre, Kleibeuker, Vonk and van der Meer⁷³⁾ and phospholipids, both of which promote colorectal tumours in animal models, possibly by increasing bacterial production of diacylglycerol. Diacylglycerol is the second messenger for protein kinase C, a key enzyme involved in signal transduction, which stimulates cell proliferation⁽Reference Holt, Moss, Whelan, Guss, Gilman and Lipkin⁷⁴⁾. Ca has also been suggested to act directly on the colonic epithelium to inhibit the proliferation of epithelial cells by inducing differentiation and increasing apoptosis⁽Reference Lipkin and Newmark⁷⁵⁾, both of which are important processes for eliminating cancerous cells in the colon. Ca has been shown to decrease cell proliferation when added in high concentrations to human colonic epithelial cells cultured in vitro ⁽Reference Buset, Lipkin, Winawer, Swaroop and Friedman⁷⁶⁾.

Early studies in human subjects suggested a reduction in colonic epithelial cell proliferation with increasing Ca intake⁽Reference Buset, Lipkin, Winawer, Swaroop and Friedman^76–Reference Lipkin, Friedman, Winawer and Newmark⁷⁸⁾, although these studies had small sample sizes. Subsequent studies showed that Ca supplementation did not reduce epithelial cell proliferation in the rectal mucosa of subjects with adenoma⁽Reference Bostick, Potter, Fosdick, Grambsch, Lampe, Wood, Louis, Ganz and Grandits^79–Reference Alberts, Einspahr and Ritenbaugh⁸¹⁾, ulcerative colitis⁽Reference Bostick, Boldt, Darif, Wood, Overn and Potter⁸²⁾ or hereditary non-polyposis CRC⁽Reference Cats, Kleibeuker, van der Meer, Kuipers, Sluiter, Hardonk, Oremus, Mulder and de Vries⁸³⁾, although Holt et al. ⁽Reference Holt, Atillasoy, Gilman, Guss, Moss, Newmark, Fan, Yang and Lipkin⁸⁴⁾ showed that increasing Ca intake from dairy products in seventy individuals with previously resected colonic adenomas changed colonic biomarkers of cancer risk (proliferative activity of colonic epithelial cells and markers of normal cellular differentiation) towards normal. One study has shown that supplementation of the diet with Ca tablets or low-fat dairy foods lowered colonic epithelial cell proliferation⁽Reference Holt, Wolper, Moss, Yang and Lipkin⁸⁵⁾. Generally, however, the weight of the evidence suggests that Ca supplements do not lower epithelial cell proliferation rates.

Animal studies have shown that dietary Ca and vitamin D status act synergistically to modify CRC risk⁽Reference Harris and Go⁵⁹⁾. Vitamin D may regulate the production of growth factors and cytokines, which are important for normal cell function, and also influence apoptosis and differentiation⁽Reference Harris and Go⁵⁹⁾. These effects act in unison and may inhibit uncontrolled cell growth. Studies in cell lines⁽Reference Shabahang, Buras, Davoodi, Schumaker, Nauta, Uskokovic, Brenner and Evans⁸⁶⁾ and animals⁽Reference Taniyama, Wanibuchi, Salim, Yano, Otani, Nishizawa, Morii and Fukushima⁸⁷⁾ have shown growth inhibition of tumour cells following administration of vitamin D. In animal studies, adequate vitamin D supply has been shown to prevent the hyperproliferation and adenoma formation induced by stress diet (high fat, low Ca, high phosphate and low vitamin D) or carcinogen treatment (azoxymethane or 1,2-dimethylhydrazine). Human studies have shown that administration of vitamin D (in combination with Ca) resulted in decreased proliferative indices and other markers of tumour growth in rectal mucosa of individuals with adenoma⁽Reference Holt, Bresalier, Ma, Liu, Lipkin, Byrd and Yang⁸⁸⁾. Increased levels of the circulating form of vitamin D (25-hydroxyvitamin D₃) in the serum were significantly correlated with several independent indices of cell proliferation⁽Reference Holt, Arber, Halmos, Forde, Kissileff, McGlynn, Moss, Kurihara, Fan, Yang and Lipkin⁸⁹⁾. It is possible that vitamin D may be important in determining the effect of Ca on cell proliferation, and failure to account for vitamin D status may be partly responsible for the discrepancies between Ca supplementation studies with cell proliferation as an endpoint.

Apart from Ca and vitamin D, other potential chemopreventive agents in dairy products have received more limited attention. Dietary butyric acid, found in the lipid fraction of milk and milk products, has been suggested to play a role in colorectal neoplasia. Butyric acid has been shown to inhibit proliferation and induce differentiation in tumour cell lines⁽Reference McBain, Eastman, Nobel and Mueller⁹⁰⁾ and animal studies⁽Reference Velazquez, Zhou, Seto, Jabbar, Choi, Lederer and Rombeau⁹¹⁾. However, it is more likely that these effects are a result of butyric acid production in the colon as a result of fermentation of dietary fibre by colonic microflora⁽Reference Williams, Coxhead and Mathers⁹²⁾, since dietary butyric acid is rapidly absorbed in the small bowel and metabolised in the liver and therefore never reaches the colon⁽Reference Parodi⁹³⁾.

More convincing evidence exists for CLA⁽Reference MacDonald⁹⁴⁾, a naturally occurring fatty acid present in dairy and beef products. CLA is derived from linoleic acid, a PUFA that is converted to CLA by bacteria in the rumen of ruminant animals⁽Reference Maggiora, Bologna and Ceru⁹⁵⁾. Studies in cell lines have shown that CLA inhibits the growth of a variety of tumour cells⁽Reference Maggiora, Bologna and Ceru^95–Reference Bocca, Bozzo, Gabriel and Miglietta⁹⁷⁾. Dietary CLA has also been shown to reduce colon cancer incidence in animals administered a colorectal carcinogen⁽Reference Pak, Ryu, Ha and Park⁹⁸⁾. In a human epidemiological study in which CLA intakes were estimated, high CLA intakes were associated with a reduced risk of CRC; the RR associated with the highest v. the lowest quartile of intake was 0·71 (95 % CI 0·55, 0·91; P for trend = 0·004)⁽Reference Larsson, Bergkvist and Wolk⁹⁹⁾. CLA may protect against cancer by inhibiting cell proliferation, modifying the fluidity of cell membranes, decreasing the production of prostaglandins (inflammatory mediators) and also stimulating the immune response⁽Reference Parodi⁹³⁾.

Bacteria in fermented dairy products such as yoghurt may also have benefits for colonic health. Some strains of Bifidobacteria have been shown to reduce the formation of precursor lesions for CRC (for example, aberrant crypt foci)⁽Reference Rowland, Rumney, Coutts and Lievense^100, Reference Kulkarni and Reddy¹⁰¹⁾. Studies on cell lines have shown that lactic acid bacteria can bind carcinogens present in cooked foods to their cell-wall skeletons⁽Reference Zhang and Ohta¹⁰²⁾ and animal studies have shown the prevention of heterocyclic amine-induced DNA damage in the colon and liver of rats by different lactobacillus strains⁽Reference Zsivkovits, Fekadu, Sontag, Nabinger, Huber, Kundi, Chakraborty, Foissy and Knasmüller¹⁰³⁾. A study in human subjects has shown that the consumption of Lactobacillus acidophilus significantly reduced the excretion of carcinogens that had been ingested from heavily browned or burnt meat cooked at high temperature⁽Reference Lidbeck, Nord, Gustafsson and Rafter¹⁰⁴⁾. Bacteria in yoghurt and other fermented dairy products may also be capable of stimulating the immune system; mice fed milk containing L. bulgaricus or L. casei had increased production of immune cells such as lymphocytes and macrophages⁽Reference Perdigon, Vintini, Alvarez, Medina and Medici¹⁰⁵⁾. However, other studies have shown more ambivalent results; administration of lactic acid bacteria to carcinogen-treated rats had no effect on tumour growth⁽Reference Li and Li¹⁰⁶⁾ and some strains of lactic acid bacteria induced DNA damage and increased the effects of reactive oxygen species-generating chemicals in human colonic cells in vitro ⁽Reference Koller, Marian, Stidl, Nersesyan, Simic, Sontag and Knasmuller¹⁰⁷⁾.

Other components of milk that may have beneficial properties include sphingolipids such as sphingomyelin, which is a potent inhibitor of cell growth and also induces differentiation and apoptosis⁽Reference Merrill, Schmelz, Dillehay, Spiegel, Shayman, Schroeder, Riley, Voss and Wang¹⁰⁸⁾. Dietary sphingomyelin fed in amounts that can be found in dairy products has been shown to inhibit aberrant crypt foci formation and decrease the proportion of malignant tumours in mice⁽Reference Dilehay, Webb, Schmelz and Merrill¹⁰⁹⁾. Another category of sphingolipids known as glycosphyngolipids were also shown to inhibit the formation of aberrant crypts by up to 60 % in mice treated with a colorectal carcinogen⁽Reference Schmelz, Sullards, Dillehay and Merrill¹¹⁰⁾. Other compounds that may have beneficial effects include lactoferrin⁽Reference Tsuda, Sekine, Ushida, Kuhara, Takasuka, Iigo, Han and Moore¹¹¹⁾ and the milk protein casein⁽Reference van Boekel, Goeptar and Alink¹¹²⁾, although the evidence for these is limited.