Dairy Consumption and Risk of Conventional and Serrated Precursors of Colorectal Cancer: A Systematic Review and Meta-Analysis of Observational Studies

Objective The consumption of dairy is associated with decreased risk of colorectal cancer (CRC), but few studies have assessed the relationship between dairy consumption and precursors of CRC. Therefore, we performed the first meta-analysis to further evaluate this association. Methods PubMed, Embase, Scopus, and Web of Science databases were searched through July 2020 for observational studies. Study-specific risk estimates for the highest versus lowest category were pooled using the random-effects and fixed-effects model. The methodological quality of included studies was assessed using the ROBINS-I Scale. Results A total of 12 studies were included (3 cohort studies and 9 case-control studies). Compared with the lowest level consumption, fermented dairy products had a decreased risk of precursors of CRC in both cohort (RR = 0.92 95% CI: 0.87–0.97) and case-control studies (RR = 0.98 95% CI: 0.96–0.99). Total dairy (RR = 0.80 95% CI: 0.68–0.96) and cheese (RR = 0.96 95% CI: 0.93–0.99) consumption was inversely associated with the risk in case-control studies whereas yogurt consumption was inversely associated with the risk in cohort studies (RR = 0.91 95%CI: 0.86–0.96). No significant associations were found for consumption of total milk and non/low-fat milk. For dose-response analyses, evidence of linear association was found in total dairy and yogurt consumption. The risk decreased by 12% for an increment of 200 g/d total dairy consumption (RR = 0.88 95% CI: 0.81–0.95) and decreased by 8% for an increment of 50 g/d yogurt consumption (RR = 0.92 95% CI: 0.85–0.99). Conclusions Fermented dairy products, specifically yogurt and cheese, were significantly associated with decreased risk of conventional and serrated precursors of colorectal cancer.


Introduction
According to GLOBOCAN 2018, colorectal cancer (CRC) is the third most frequently diagnosed cancer and the second leading cause of cancer death. It is estimated that there will be over 1.8 million new CRC cases and 881,000 deaths in 2018 [1]. It is well known that cancer is a complicated disease caused by the interaction of environmental and genetic factors, such as smoking, alcohol, and diet [2]. Various dietary factors have been found to be important protective factors for CRC. For example, the Mediterranean diet might decrease the risk of CRC [3,4]. In an umbrella review of observational studies, intake of total dairy was associated with decreased risk of cardiovascular disease, hypertension, and fatal stroke [5]. e latest report of World Cancer Research Fund (WCRF) and American Institute of Cancer Research (AICR) also reached strong evidence that dairy products intake can help protect against CRC [6].
Adenomas (adenomatous polyps, AP) and a subset of serrated lesions (traditional serrated adenoma (TSA) and sessile serrated adenoma/polyp (SSA/P) are two major subtypes of the precursors of CRC. A previous meta-analysis has shown an increased risk of colorectal adenoma (CRA) with the intake of red and processed meat [7,8]; a decreased risk with increased intake of dietary fiber [9], calcium [10], magnesium [11], selenium, [12] and coffee [13]; and a null association with white meat [14]. However, the relationship between dairy consumption and precursors of CRC is still uncertain. To better understand this relationship, we combined all published epidemiologic studies on the association between dairy intake and precursors of CRC risk and then conducted a dose-response meta-analysis.

Design.
e protocol of this meta-analysis was registered in PROSPERO (CRD42020192846).
is study was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [15]. e completed PRISMA checklist is provided online. PICO format (population, intervention, comparison, outcome) was employed to answer the research question: "Are dairy associated with the occurrence of conventional and serrated precursors of colorectal cancer." Population: adults with conventional adenomas (including tubular adenoma, tubulovillous adenoma, villous adenoma, and adenoma with high-grade dysplasia) and serrated precursors of colorectal cancer (TSA and SSA/P); cases with hyperplastic polyps (HP) were excluded from the analyses. Intervention: total dairy product intake (total dairy, milk, cheese, butter, yogurt, or other dairy products). Comparison: adults without colorectal adenoma or serrated lesion. Outcome: the occurrence of conventional adenomas and serrated lesions.

Search Strategy.
We searched PubMed, Embase, Scopus, and Web of Science database in English until April 2020. e following search terms were used: "dairy" or "dairy products" or "fermented milk products" or "cultured milk products" or "fermented dairy products" or subtypes of dairy products (i.e., "milk" or "yogurt" or "yoghurt" or "cheese" or "hard cheese" or "cottage cheese" or "cheddar" or "butter" or "buttermilk" or "cream" or "ice cream" or "Kefir"; NOT: "milk protein" or "whey protein"; AND: "colorectal adenoma" or "colorectal polyp" or "colorectal lesion" or "colorectal neoplasm" or "colorectal tumor" or "colorectal carcinoma" or "colorectal cancer" without restrictions. Titles and abstracts were screened independently by two reviewers (Liliangzi Guo and Yuting Li) to exclude irrelevant articles. en, the full texts were retrieved to further increase the potentially relevant articles.

Study Selection.
Inclusion criteria: (1) adult partic-ipants>18 years of age; (2) cohort studies, case-control studies, or cross-sectional studies that investigated the association between dairy consumption and risk of conventional adenomas and serrated lesions; (3) diagnosis: colorectal adenomas and serrated lesions that were determined by histology; (4) studies that reported the risk estimates (relative risk (RR), odds ratio (OR), or hazard ratio (HR)) with their corresponding 95% confidence interval (CI) or available original data allowing us to compute the 95% CI; (5) if the published studies reported data for specific subgroups, results for the whole population were considered in this meta-analysis; (6) if the original publications provided several independent studies, they were considered as separate studies in the following data analysis.
Exclusion criteria: (1) animal studies; (2) studies conducted on children, adolescents, or pregnancy women; (3) nonoriginal papers (reviews, editorials, or commentaries); (4) meta-analysis studies; (5) studies that did not provide enough data on dairy products consumption and risk estimates; (6) duplicate reports and abstracts; and (7) studies that investigated the risk of CRC or the recurrence of colorectal adenomas or serrated lesions.

Data Extraction and Quality Assessment.
e data extracted from each study included the name of first author, publication year, study region, study design, sample size, type of colorectal lesion, dairy products categories, the risk estimates with their 95% CIs for each category of exposure variables and adjusted covariates in the multivariable analysis. We extracted the risk estimates that reflected the greatest degree of adjustment for potential confounders.
Dose-response analysis requires the distribution of cases and controls, person-years or noncases, and risk estimates with 95% CIs for at least three quantitative categories of exposure to dairy. e median or mean level of dairy intake for each category was assigned to the corresponding RR for every study. e interval size of the open-ended highest category was assumed as the closest interval while the lowest was considered as zero [9]. Two investigators (Liliangzi Guo and Weiwei Chen) independently extracted the data. Any disagreements were resolved by consulting the third investigator (Shaohui Tang).
Two investigators (Weiwei Chen and Kaiyin He) independently assessed the study quality based on the Risk of Bias in Nonrandomized Studies of Interventions (ROBINS-I) tool which contains seven domains. According to ROBINS-I guidance, the study judged to be at low risk in all seven domains was considered as low risk of bias; the study judged to be at low or moderate risk in all domains was considered as moderate risk of bias; the study judged to be at serious risk or critical risk in at least one domain was considered as serious risk of bias or critical risk of bias, respectively [16]. e definition of the dairy category (Table S1) in this meta-analysis is listed as follows. Total dairy includes total milk and fermented dairy product. Total milk includes non/ low-fat milk and whole milk. e fermented dairy product includes yogurt, cheese, and cottage cheese.

Statistical Analysis.
e results were expressed in terms of RR and 95% CI for the highest versus lowest category of dairy products consumption. e standard error of the logarithmic OR/RR of each study was calculated and taken as the estimated variance of the logarithmic OR/RR. e inverse variance method was adopted (DerSimonian and Laird, 1986) [17]. Dose-response analysis was also conducted and P < 0.05 was considered statistically significant. Cochran's Q-test and I 2 statistics were used to assess the heterogeneity of individual studies. I 2 > 50% and P < 0.1 were considered as statistically significant heterogeneity [18]. Fixed-effects model and random-effects model were adopted as the pooling method. We used the random-effects model to calculate the pooled RRs and 95 CIs when the evidence of heterogeneity was present (>50%). Otherwise, if there was no obvious heterogeneity (<50%), the fixed-effects model was used. Subgroup analyses were performed to explore the source of heterogeneity in this study. e subgroup analyses were performed according to study design, geographic location, patient sex, number of cases, size of adenoma, type of colorectal lesion, study quality, and type of food frequency questionnaire (FFQ) with dairy consumption if data were permitted. Sensitivity analyses were conducted to examine the stability of results by omitting one study at a time. Publication bias was assessed with Begg's test and Egger's test if ≥10 studies are available and it was considered to exist when P < 0.05. e trim and fill method was used to reduce the potential influence of publication bias [19].
Moreover, if ≥3 studies are available, a dose-response meta-analysis was carried out to estimate the trend between different exposure levels of dairy products and precursors of CRC using a random-effects meta-regression. Generalized least-squares trend (GLST) estimation modeling and spline curve modeling (MKspline STATA command) were used to estimate the dose-response relation analysis. Both linear and nonlinear dose-response analyses were performed with data from the included studies. e GLST and variance-weighted least-squares methods require median values for categories of intake levels. When medians and means were not presented, the category mid-point was used. If the dairy consumption was reported in servings/day, we converted it into grams/day as the following standard units: 200 g for total dairy, 200 g for milk (1 glass), 125 g for yogurt (1 cup), and 30 g for cheese [20]. If the cheese consumption was given in slices/day, we considered each slice as 25 g [21]. e doseresponse results were presented for a 200, 200, 50, and 25 g/ day increment for total dairy products, total milk, yogurt, and cheese, respectively. All statistical analyses were performed using STATA, version 12.0 (Stata Corp., College Station, TX, USA). Figure 1 shows the flow diagram of the identification and selection process of the included studies. Of the 1377 potentially relevant articles which initially retrieved, 563 duplicate articles were excluded and 58 articles remained for full-text review after screening the title and abstract. Among them, 46 articles were excluded (16 were review/meta-analysis/guideline articles, 14 did not report relative risk or sufficient data to calculate relative risk, 6 were abstracts and original articles were not available, 5 did not modify risk factor to different dairy products such as yogurt or cheese, 4 reported duplicated analysis from the same data source, and 1 did not report separately results of colorectal adenoma/cancer). In the end, a total of 12 eligible articles were included in this systematic review and meta-analysis: 9 case-control and 3 cohort studies. e 12 included studies, published between 1991 and 2020, had 19957 cases with precursors of CRC. Of the studies, 6 were conducted in Europe, 4 in the United States, 1 in Iran, and 1 in Australia. Tables 1 and 2 show the main characteristics of the included studies.

Dose-Response Meta-Analysis.
Both linear and nonlinear dose-response analyses were performed. e potential nonlinear association was examined using restricted cubic splines with 4 knots fixed at the 5 th , 35 th , 65 th , and 95 th percentiles of the distribution. Combining data from 3 studies [26,30,33], trend meta-analysis showed a significant negative dose-response relationship in total dairy (P −nonlinearity � 0.947) and yogurt (P −nonlinearity � 0.794) consumption from linearity. We found that 200 g/d increment in total dairy consumption could decrease 12% risk of colorectal adenomas and serrated lesions using the fixedeffect model with no heterogeneity (RR � 0.88, 95%CI: 0.81-0.95, P � 0.001; P h � 0.658) (Figure 7(a)). Further, we found the risk of colorectal adenomas and serrated lesions decreased by 8% with an increment of 50 g yogurt using the fixed-effect model (RR � 0.92, 95% CI: 0.85-0.99, P � 0.037; P h � 0.367) (Figure 7(b)). were review/meta-analysis/guideline articles 14 did not report relative risk or sufficient data to calculate relative risk 6 were abstracts and original articles were not available 5 did not modify risk factor to different dairy products such as yogurt or cheese 4 reported duplicated analysis from the same data source 1 did not report separately results of colorectal adenoma/cancer  Journal of Oncology 5  Journal of Oncology   Journal of Oncology

Subgroup and Sensitivity Analyses.
e results of subgroup analyses conducted to explore the sources of heterogeneity were detailed shown in Table S3. Intake of total dairy, fermented dairy products, yogurt, and cheese was significantly associated with a decreased risk of colorectal adenomas and serrated lesions in most subgroup analysis. When stratified by sex, an inverse association was observed for intake of fermented dairy products in both men (SRR � 0.85, 95% CI: 0.77-0.93) and women (SRR � 0.92, 95% CI: 0.87-0.98). When stratified by the size of adenoma, an inverse association was observed for intake of yogurt and cheese in studies with a size of 10 mm or more (SRR � 0.71 95% CI: 0.51-0.99; SRR � 0.96 95% CI: 0.93-0.99) but not among studies with a size of less than 10 mm (SRR � 1.01 95% CI: 0.81-1.25; SRR � 1.08 95% CI: 0.68-1.71). Intake of total milk was not significantly associated with the risk of colorectal adenomas and serrated lesions in subgroup analyses. Too few studies of non/lowfat milk precluded any meaningful subgroup analysis.
Sensitivity analyses were conducted to evaluate the influence of a single study on the overall risk estimate by omitting one study in each turn (Table S4). In the analysis of total milk, the heterogeneity decreased from 56.9% to 0% when omitting the study by Kune et al. In the analysis of yogurt, the heterogeneity decreased from 50.2% to 0% when omitting the study by Karagianni et al.

Discussion
By pooling the 12 observational studies (9 case-control and 3 cohort studies) eventually included, this present metaanalysis was the first study that conclusively indicated an inverse association between fermented dairy products including yogurt and cheese and the risk of conventional adenomas and serrated lesions, of which fermented dairy products intake showed a decreased trend in both cohort and case-control studies but cheese and yogurt consumption display an inverse association only in case-control studies or cohort studies. No significant associations were found for the consumption of total milk and non/low-fat milk. Very few studies have evaluated the association between whole milk and cottage cheese with the risk of conventional adenomas and serrated lesions. e negative associations of total dairy, yogurt, and cheese were not supported by cohort studies which may be explained by many reasons such as the differences in study design, study quality, category of the intake frequency, amounts of dairy intake, and potential confounders. CRC is arising through three major pathways, including adenoma-carcinoma sequence, serrated pathway, and inflammatory pathway [34]. It was estimated that approximately 10-15% of the sporadic CRC were progressed from a serrated polyp and 85-90% were progressed from adenomatous polyps. e subset of serrated polyps includes HP (hyperplastic polyp), TSA, and SSA/P [35]. HP is the most prevalent type of serrated polyps which is the lack of malignant potential [36]. erefore, we excluded the data investigating risk between dairy and HP. No more than 10% of the adenomas will progress to CRC. Compared with nonadvanced adenomas, advanced adenomas (≥10 mm in diameter, villous histology, or high-grade dysplasia, with or without >3 adenomas) are more likely to develop into malignancy [35]. Because high-risk adenoma is more likely to develop into malignancy than low-risk adenoma, the data of high-risk adenoma are more informative for preventive strategies. e result of this meta-analysis showed a negative association between yogurt and cheese intake and adenoma ≥10 mm, suggesting that fermented dairy products may play a protective role against the progression of CRC. However, the result should be interpreted with caution because only a few studies were involved in this meta-analysis.
Dairy foods may decrease the risk of precursors of CRC through several mechanisms. Dairy foods are the main source of calcium in the diet. In the most recent report from the WCRF/AICR, low calcium intake may increase CRC risk. A previous observational study has shown a negative association between supplemental calcium intake and CRC risk [37]. On the one hand, calcium can decrease cell proliferation induced [38].
On the other hand, calcium can have an impact on several intracellular pathways leading to apoptosis in transformed cells and differentiation in normal cells [39,40].
e results of RCT showed that, in normal appearing colorectal mucosa of individuals with a history of adenoma, the expression of APC (adenomatous polyposis coli) and β-catenin is modified by calcium supplementation [41,42]. Moreover, besides calcium, vitamin D, and folate are also micronutrients which might have protective effects [43].
We found that total dairy and fermented dairy products were inversely associated with the risk of the precursors of CRC while total milk was not associated with the risk. ere are several mechanisms proposed to explain how fermented dairy food decreases the risk of precursors of CRC. Since the process of fermentation, fermented dairy foods contain plentiful probiotic such as Lactobacillus and Bifidobacterium [44][45][46]. e microbiota in the human body can form a microenvironment to alter cancer susceptibility and progression [47]. In a case-control study, CRC-associated microbiota was found to change with the degree of malignancy along the adenoma-carcinoma sequence [48]. According to this evidence, the gut microbiome, which can correct microbiota composition, modulate the innate immune system, restore gut barrier function, prevent pathogen colonization, and exert selective cytotoxicity against tumor cells, plays an important role in the development of CRC [49]. Yogurt has been recommended by the Dietary Guideline for Americans [50]. Live microorganisms such as Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus, two lactic acid bacteria used to ferment, may prevent carcinogenesis from initiating [51]. Various studies also have reported a protective effect of probiotics or prebiotics in CRC mice models as reviewed [52]. Furthermore, yogurt has been suggested to induce gastrointestinal hormone secretion [53,54].  Our systematic review and meta-analysis had several strengths. Most of the current meta-analysis focus on the relationship between dietary factors such as milk or dairy products and the risk of CRC. We conducted the first metaanalysis of the association between dairy consumption and the risk of precursors of CRC including the conventional and serrated lesions, based on highest versus lowest analysis, linear and nonlinear dose-response meta-analysis. We included both cohort and case-control studies through a systematic search. All of the included studies used a validated FFQ to assess dairy consumption. Further, a significantly inverse dose-response relationship was observed between total dairy and yogurt consumption and the risk of precursors of CRC, which may strengthen the reliability of the results of our hypothesis. We also carried out a sensitivity analysis to investigate whether a particular study could explain the results.
However, our meta-analysis had several limitations. e main limitation of the study was a small number of included studies and subjects, so further subgroup analyses were not able to perform according to anatomical location and histology type. Dose-response analyses of specific types of dairy products were also limited because a few studies reported the modifiable risk factor with 95% CI to different quantitative categories of exposure to dairy products. Secondly, the observed inverse association between dairy intake and precursors of CRC may be due to unmeasured or potential residual confounding, although the quality of 12 included studies was evaluated to have a moderate risk of bias by ROBINS-I tool.
Although some known confounding factors were adjusted in most of the studies, not all potential confounders were adjusted for in every study. We found that the association between total dairy, fermented dairy products, yogurt, and cheese persisted in most subgroups, with adjustment for confounding factors. irdly, significant heterogeneity was observed among studies. However, there was no evidence of significant heterogeneity found between subgroups analyses. Fourthly, most of the case-control studies may not avoid recall and selection bias, especially dietary recall bias. Lastly, potential publication bias might have influenced the results.

Conclusion
is systematic review and meta-analysis was the first to assess the relationship between different dairy products and the precursor of CRC risk as well as the first to explore a linear association between them. In conclusion, our study suggests an inverse relationship between total dairy, fermented dairy products, yogurt, cheese, and risk of precursors of CRC, though the evidence was limited. However, no harmful effects were found between the intake of total milk and the risk of the precursor of CRC. Knowledge of risk factors associated with the precursors of CRC is important in prevention strategies. More large and precise prospective studies, as well as clinical trials, are needed to further investigate the associations and mechanisms between them. Future studies should also focus on the interactions between gut microbiota and environmental factors and their influences on colorectal carcinogenesis.

Data Availability
No data were used to support this study.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Supplementary Materials
Supplemental Table 1. Definition of dairy products as described in the paper of 12 studies included in the metaanalysis. Supplementary Table 2. Bias risk of each domain of included studies assessed by ROBINS-I. Supplemental Table 3. Subgroup analyses of total dairy, fermented dairy products, total milk, yogurt, and cheese and the risk of colorectal adenomas and serrated lesions. Supplemental Table 4. e sensitivity analysis of the relationship between total milk intake and risk of colorectal adenomas and serrated lesions (Supplementary Materials)