Associations between low‐ and high‐fat dairy intake and recurrence risk in people with stage I–III colorectal cancer differ by sex and primary tumour location

We previously demonstrated that intake of low‐fat dairy, but not high‐fat dairy, was associated with a decreased colorectal cancer (CRC) recurrence risk. These risks, however, may differ by sex, primary tumour location, and disease stage. Combining data from two similar prospective cohort studies of people with stage I–III CRC enabled these subgroup analyses. Participants completed a food frequency questionnaire at diagnosis (n = 2283). We examined associations between low‐ and high‐fat dairy intake and recurrence risk using multivariable Cox proportional hazard models, stratified by sex, and primary tumour location (colon and rectum), and disease stage (I/II and III). Upper quartiles were compared to lower quartiles of intake, and recurrence was defined as a locoregional recurrence and/or metastasis. During a median follow‐up of 5.0 years, 331 recurrences were detected. A higher intake of low‐fat dairy was associated with a reduced risk of recurrence (hazard ratio [HR]: 0.60, 95% confidence interval [CI]: 0.43–0.83), which seemed more pronounced in men (HR: 0.51, 95% CI: 0.34–0.77) than in women (HR: 0.84, 95% CI: 0.47–1.49). A higher intake of high‐fat dairy was associated with an increased risk of recurrence in participants with colon cancer (HR: 1.60, 95% CI: 1.03–2.50), but not rectal cancer (HR: 0.88, 95% CI: 0.54–1.45). No differences in associations were observed between strata of disease stage. Concluding, our findings imply that dietary advice regarding low‐fat dairy intake may be especially important for men with CRC, and that dietary advice regarding high‐fat dairy intake may be specifically important in people with colon cancer.


What's new?
There is strong evidence that dairy intake is associated with a reduced risk of developing colorectal cancer.However, its relation to the risk of colorectal cancer recurrence remains unclear.
Combining data from two prospective cohort studies of patients with stage I-III colorectal cancer, this study found that a higher intake of low-fat dairy was associated with a reduced risk of recurrence in men, where as a higher intake of high-fat dairy was associated with an increased risk of recurrence in people with colon cancer.With a global transition towards more plantbased diets and increasing numbers of colorectal cancer survivors, the findings may contribute to more personalised dietary guidelines.

| INTRODUCTION
There is strong evidence that dairy intake is associated with a reduced risk of developing colorectal cancer (CRC). 1,2However, even though some studies have assessed the intake of different types of dairy and dairy products in relation to CRC-specific mortality, [3][4][5][6][7] no studies thus far other than our previous work 8 have assessed its relation to a more direct indicator of neoplastic growth after diagnosis: the risk of CRC recurrence.We previously observed that a higher pre-diagnostic intake of low-fat dairy, but not high-fat dairy, was associated with a reduced risk of recurrence in people with stage I-III CRC. 8 The current study includes a larger study population with more recurrence data available, which enables us to build on our previous findings by identifying clinically relevant subgroups of individuals who may benefit most from dietary advice regarding dairy intake.
Tumour recurrences develop in 3%-38% of individuals with CRC within 3 years after a resection with curative intent, depending on stage at diagnosis. 9Risk of recurrence has been reported to be higher in men than in women, and in people with rectal cancer compared to people with colon cancer. 9,10Furthermore, the localisation of recurrences also seems to differ per location of the primary tumour, 9,11 and may differ for men and women, 11 implying different underlying mechanisms are involved leading to cancer recurrence.
The association between dairy intake and risk of recurrence may also differ for subgroups of people with CRC based on sex, primary tumour location, and disease stage.First, prior studies have demonstrated that higher intakes of low-fat dairy 7 and calcium from dairy 12 may be associated with a reduced risk of CRC-specific mortality in men, but not in women.Second, colon and rectal cancer differ in terms of embryonical origin, genetic and molecular characteristics, microbiome, oncogenesis, and treatment, 9,13 and may also have different risk factors for recurrence. 9Thirdly, taking into account the wide variation in risk of recurrence across different stages of disease at diagnosis, 9 we aim to study whether dairy intake is similarly related to risk of recurrence for stages I-III of disease.
Based on our previous work, 8 where we observed a reduced risk of recurrence and all-cause mortality with higher intakes of low-fat dairy, but an increased risk of all-cause mortality with higher intakes of high-fat dairy, we assess low-and high-fat dairy separately in the current study.Hypotheses about how dairy components influence risk of recurrence in CRC are largely derived from research on the risk of CRC occurrence.5][16][17][18][19][20] Lactic acid-producing bacteria in fermented dairy have been proposed to inhibit colorectal neoplastic growth. 2,21Furthermore, it has been proposed that dairy intake may affect microbial diversity and associated microbial metabolites in the gut, 22 thereby inhibiting neoplastic growth of the colonocytes. 23th the current study we aim to investigate the association between low-and high-fat dairy intake and risk of CRC recurrence in subgroups of sex, primary tumour location and stage at diagnosis.These subgroup variables are usually readily available in clinical practice and do not require further testing, ultimately enhancing the applicability of our results in daily practice.

| Study population
The initial study population consisted of 2544 adults who were newly diagnosed with stage I-III CRC, from two prospective cohort studies: the COLON study (n = 1945) 24 and the PLCRC-PROTECT study (n = 599). 25Details of the COLON study and overall PLCRC study have been described previously. 24,25Briefly, participants of the COLON study were recruited from 11 hospitals in the Netherlands between August 2010 and February 2020.Recruitment for the PLCRC-PROTECT study started in February 2016 and was ongoing at the time of data analysis in 21 hospitals in the Netherlands.Participants who were recruited before December 2020 were included in the current study population.For the current analyses, we excluded participants with missing data on dietary intake (n = 161), missing data on recurrence (n = 5), no confirmed surgery (n = 91), and those who appeared to have a metastasis before surgery (n = 3) or whose follow-up ended before surgery (n = 1) (Figure 1).The final population for analysis contained 2283 participants.

| Assessment of exposure
In both studies, participants filled out an identical self-administered semi-quantitative food frequency questionnaire (FFQ) of 204 items at diagnosis, 26,27 reflecting on dietary intake in the month prior to the VAN LANEN ET AL. measurement.Building on our previous work, 8 which showed associations between dairy intake at diagnosis, but not 6 months after diagnosis, in relation to risk of recurrence, the current study focuses on pre-diagnostic dairy intake.
Dietary intake of low-fat dairy and high-fat dairy was calculated in grams per day.Low-fat dairy included low-fat or skimmed versions of milk, yoghurt, custard, and soft curd cheese.High-fat dairy included whole-fat versions of milk, yoghurt, custard, soft curd cheese, and all other cheeses, condensed milk, ice cream, whipped cream, and butter.
Ready-made breakfast drinks were not included, because not all ready-made breakfast drinks on the Dutch market contain dairy.
Energy and nutrient intakes were calculated using the online Dutch Food Composition Table (version 2011/3.0). 28Low-and high-fat dairy intakes were adjusted for total energy intake using the energy residual method. 29To improve interpretability, the predicted dairy intake at the median total energy intake was added to individual residuals.
Mean differences between absolute and energy-adjusted intakes were very minimal: energy-adjusted low-and high-fat dairy intakes were on average 3 (SD: 31) and 4 (SD: 36) g/day lower than absolute intakes, respectively.

| Assessment of outcome
Recurrences were defined as a locoregional recurrence and/or metastasis occurring after surgery.Locoregional recurrence was defined as a recurrence in the same segment as the primary tumour, in the lymph nodes of the same segment, or in the draining lymph nodes.For both cohorts, updated recurrence data were requested from the Netherlands Cancer Registry via the Netherlands Comprehensive Cancer Organisation.Follow-up time was calculated starting from date of surgery until date of recurrence, until the date recurrence status was updated, or until end of follow-up (e.g., due to death, occurrence of another primary tumour with metastasis, or moving abroad), whichever came first.If date of surgery was unavailable (n = 2), date of filling out the FFQ was used.

| Assessment of covariates
At diagnosis, participants from both cohorts also filled out a questionnaire on demographics, anthropometrics, cancer family history, and lifestyle habits, including questions about age (years), sex (man/woman), education (low/medium/high), body weight (kg), height (cm), smoking status (current/former/never), and calcium and vitamin D supplement use in the past year (including multivitamins, yes/no).
Physical activity was assessed using the Short QUestionnaire to ASsess Health-enhancing physical activity (SQUASH). 30Moderateto-vigorous physical activity (hours/week) included all activities with a metabolic equivalent value ≥3 according to Ainsworth et al. 31 Clinical data, such as disease stage (I-III), tumor location (colon: caecum to the sigmoid colon; rectum: rectosigmoid junction and rectum) and type of treatment (only surgery, surgery and chemotherapy, surgery and radiotherapy, surgery and chemoradiation) were collected via the Dutch ColoRectal Audit (COLON) 32 or the Netherlands Cancer Registry (PLCRC-PROTECT).To assess possible confounding by other dietary factors previously associated with CRC risk, 2 we also calculated total intake of wholegrains, red meat, processed meat, dietary fibre, and alcohol.Definitions of red and processed meat were as in the Continuous Update Project Expert report of 2018 from the World Cancer Research Fund/American Institute for Cancer Research. 2

| Data analysis
Sex-specific quartiles of dairy intake were constructed.Population characteristics are presented as medians [interquartile range (IQR)] or numbers (percentage).
Cox proportional hazards regression analyses were used to calculate Hazard Ratios (HR) and 95% confidence intervals (95% CI) for the associations between low-and high-fat dairy intake and risk of recurrence, and for the associations between low-and high-fat dairy and total recurrence in strata of sex (man/woman), primary tumour location (colon/rectum), and stage of disease at diagnosis (I-II/III).Log-log  Dairy intake was adjusted for daily total energy intake using the energy residual method. 29To improve interpretability, the predicted dairy intake at the median total energy intake was added to individual residuals.b Model 1 was adjusted for age, sex and total energy intake (as part of the energy residual method).
curves were visually inspected for non-parallelism to check the proportionality assumption for the Cox proportional hazards model.The lowest quartile was used as the reference category in categorical analyses.p trend values were computed over quartiles of intake using the medians of the corresponding quartiles.For continuous analyses, increments of 100 g/day were used.First, a crude model was created, adjusting for age and sex (except for analyses in strata of sex).Then, potential confounders were added to the model when they changed the HR by >10%.The following covariates were considered as confounders based on literature: primary tumour location (except for analyses in strata of primary tumour location), disease stage (except for analyses in strata of disease stage), BMI (continuous), education level (low, medium, high), smoking status (current, former, never), moderateto-vigorous physical activity (continuous), and total dietary intake of wholegrains, dietary fibre, red meat, processed meat and alcohol (g/day).The fully adjusted model included age, sex, disease stage, and total daily intakes of energy, dietary fibre, and alcohol.Simultaneous adjustment for high-fat dairy (in low-fat dairy analyses) and low-fat dairy (in high-fat dairy analyses) did not change conclusions.
We also evaluated restricted cubic splines (RCS) to study linearity of the associations between dairy exposures and CRC recurrence, using the fully adjusted model.For low-fat dairy, the model was observed to fit best with 4 knots based on Akaike's information criterion, and knots were placed at the 5th, 35th, 65th, and 95th percentile.For high-fat dairy, the model was observed to fit best with 3 knots, and knots were placed at the 5th, 50th, and 95th percentile.Graphs were truncated at the 1st and 99th percentile.The median intake of the first quartile of each exposure was used as the reference.
Sensitivity analyses were conducted excluding participants who had a recurrence within 6 months after filling out the FFQ, and those who were diagnosed with CRC before the age of 50 years.As butter has a specifically high fat content within the high-fat dairy category, we also conducted analyses for high-fat dairy intake excluding butter.
Data analyses were performed using R Statistical Software (version 4.0.5).p-values below .05were considered statistically significant.

| RESULTS
Participants were on average 66 years (IQR: 60-72) at CRC diagnosis, and 38% were woman (Table 1).Participants in the highest quartile of low-fat dairy intake (with a median energy-adjusted intake of 378 g/day) were older, less often current smokers, consumed more dietary calcium, and consumed less high-fat dairy and alcohol, compared to participants in the lowest quartile of low-fat dairy intake (with a median energyadjusted intake of 21 g/day).Participants in the highest quartile of highfat dairy intake (with a median energy-adjusted intake of 186 g/day) were older, had a lower level of education, more often had a tumour in the proximal colon, consumed more dietary calcium, and consumed less total energy, dietary fibre, wholegrains, low-fat dairy, processed meat, and alcohol, compared to participants in the lowest quartile of high-fat dairy intake (with a median energy-adjusted intake of 25 g/day).
During a median follow-up time of 5.0 years (IQR: 3.0-7.3),331 recurrences were detected.A higher intake of low-fat dairy was associated with a reduced risk of recurrence in the total population (HR Q4 vs Q1 : 0.60, 95% CI: 0.43-0.83,p for overall association: .005,

(A) (B)
p for overall association: .008p for non−linearity: .88 and high-fat (B) dairy intake in relation to recurrence.Splines were adjusted for age, sex, stage, and total daily intakes of energy, dietary fibre, and alcohol.For low-fat dairy, the model was observed to fit best with 4 knots, and knots were placed at the 5th, 35th, 65th, and 95th percentile.For high-fat dairy, the model was observed to fit best with 3 knots, and knots were placed at the 5th, 50th, and 95th percentile.
In contrast, a higher intake of high-fat dairy tended to be associated with a higher risk of recurrence (HR Q4 vs Q1 : 1.26, 95% CI: 0.91-1.75,p for overall association: .008,p for non-linearity: .88,Table 2, Figure 2B).This association was similar but did not reach statistical significance for subgroups of sex and disease stage (Figure 3).For subgroups of primary tumour location, a higher intake of high-fat dairy was associated with a higher risk of recurrence in participants with colon cancer (HR Q4 vs Q1 : 1.60, 95% CI: 1.03-2.50),while no association with risk of recurrence was observed in participants with rectal cancer (HR Q4 vs Q1 : 0.88, 95% CI: 0.54-1.45).

| DISCUSSION
In the current study, a higher intake of low-fat dairy was associated with a decreased risk of recurrence in people with stage I-III colorectal cancer, which seemed more pronounced in men than in women.In   F I G U R E 3 Forest plot comparing hazard ratios for quartiles of low-and high-fat dairy intake in the total study population and by sex, primary tumour location, and stage.Hazard ratios were adjusted for age, sex (except for analyses in strata of sex), stage (except for analyses in strata of stage), and total daily intakes of energy, dietary fibre, and alcohol.
contrast, a higher intake of high-fat dairy tended to be associated with a higher risk of recurrence, which seemed limited to people with colon cancer.
Low-and high-fat dairy contained similar amounts of calcium in our study, but the saturated fatty acid content was higher in high-fat dairy than in low-fat dairy (weighted average saturated fat content: 9.1 g/100 g vs. 0.6 g/100 g).A previous study in people with metastatic colon cancer observed that replacing 5% of energy from carbohydrates with saturated fat was associated with an increased risk of cancer progression or death (HR: 1.23, 95% CI: 1.04-1.45),and that replacing 10% of energy from carbohydrates with animal fat tended to be associated with an increased risk of cancer progression or death (HR: 1.17, 95% CI: 0.98-1.40). 33However, no significant associations were observed when assessing quartiles of saturated fat or animal fat intake, which was also observed in previous work of the same authors in people with stage III colon cancer. 34In our study, we also observed no statistically significant association between high-fat dairy intake and risk of recurrence when assessing intake in quartiles (HR Q4 vs Q1 : 1.26, 95% CI: 0.91-1.75),but we did detect a statistically significant association with risk of recurrence when assessing continuous intake (HR per 100 g/day : 1.18, 95% CI: 1.06-1.31,p for overall association: .008, Table 2, Figure 2B).Possibly, the difference in saturated fatty acid content among quartiles of high-fat dairy intake is too low to detect an association with risk of recurrence.Furthermore, the abundance of long-chain saturated fatty acids in high-fat dairy may form insoluble complexes with calcium that are excreted in the faeces, [35][36][37] preventing calcium from performing its hypothesised protective action.The calcium-soap forming capacity and inhibition of calcium absorption were observed to be most prominent for long-chain saturated fatty acids, 35,38 which is the predominant fat source in milk. 39In humans, a meta-analysis of different randomised controlled trials also demonstrated that short-term dietary fortification of calcium increases faecal fat content. 40In conclusion, saturated fatty acids in high-fat dairy, at a certain threshold, may be associated with an increased risk of recurrence, possibly by preventing the calcium from performing its protective action.
Nevertheless, our results for low-fat dairy did not change or were even slightly stronger when adjusted for calcium intake from dairy sources (total study population: HR Q4 vs Q1 : 0.53, 95% CI: 0.36-0.77),implying that calcium is at least not the only component in dairy responsible for its relation to risk of recurrence in CRC.Different dairy products have also been described to influence the microbiota composition in the gut. 22Drinking buttermilk has been associated with an increased microbial diversity, and specifically with the presence of the industrial fermentation-related species Leuconostoc mesenteroides and Lactococcus lactis, whereas drinking whole milk has been associated with a decreased diversity. 22Loss of microbial diversity has been associated with an increased risk of chronic inflammatory diseases, 41,42 and may also be involved in CRC progression. 43To increase our understanding of how low-and high-fat dairy influence risk of CRC recurrence, future research could study the calciumrelated pathways and the extent to which calcium soaps are formed in people with CRC upon low-and high-fat dairy consumption.
Inherent to the lack of consensus in literature on how dairy components influence neoplastic growth in CRC, we can only speculate about the biological mechanisms underlying our sex-and tumour location-specific findings.In subgroup analyses, we observed that low-fat dairy was associated with a reduced risk of CRC recurrence in men, whereas this association appeared weaker in women.Previous studies have demonstrated associations between dairy or calcium from dairy and CRC-specific mortality to be more prominent in men than in women. 7,12A possible explanation for these sex-specific findings is that calcium absorption has been observed to decrease with age, especially after menopause. 44,45As the majority of women included in our study was at a reasonable age to be peri-or postmenopausal upon inclusion (median age: 65 [IQR: 59-71] years), 46 a decreased calcium uptake in post-menopausal women could hypothetically explain why the association between low-fat dairy and risk of recurrence was less pronounced in women than in men.Our sample size did not allow for further stratification by calcium supplement use in women.
Furthermore, a higher intake of high-fat dairy seemed to be associated with an increased risk of recurrence in participants with colon cancer, but not in those with rectal cancer.Colon and rectal cancer differ in, among other aspects, genetic and molecular characteristics, oncogenesis, treatment, and possibly also risk factors for recurrence. 9,13sides, as proximal and distal colon cancer also differ in terms of etiological and molecular characteristics, 47 it would be interesting to study the association between high-fat dairy intake and risk of recurrence in proximal and distal colon cancer separately.Although we have merged two datasets of relatively large cohort studies in CRC cases, our sample size did not allow for such analyses.Future investigations should confirm whether high-fat dairy is associated with an increased risk of colon cancer recurrence, and not rectal cancer, if possible further classified into proximal colon and distal colon cancer.
A strength of the current study is the availability of CRC recurrence data, which was retrieved in a standardised manner by specialised data managers from an experienced institute.A recurrence is a direct indicator of neoplastic growth and a common fear of cancer survivors, 48 and therefore a highly relevant outcome measure.
Another strength of this study is its relatively large sample size, that was achieved by merging the datasets of two large prospective cohort studies with similar methods and even identical questionnaires.This enabled us to follow up on previous findings and study associations between dairy and risk of recurrence in subgroups of sex, primary tumour location, and disease stage.Nevertheless, our sample size did not allow for stratification by age, or for further stratification by menopausal status or calcium supplement use in women, which would have been interesting to investigate potential sex-specificity of the association between low-fat dairy and risk of CRC recurrence.
Furthermore, a limitation of the current study is that even though the extensive, 204-item FFQ we used has been successfully validated for fats, cholesterol, vitamin B12 and folate, 27,49 it has not been specifically validated for dairy.][52] In conclusion, our findings imply that dietary advice regarding low-fat dairy intake may be especially important in men with CRC, and that dietary advice regarding high-fat dairy intake may be specifically important in people with colon cancer.Understanding how dairy intake relates to the risk of recurrence in specific subgroups of individuals with CRC may prove especially relevant in a world that is transitioning towards more plant-based diets, 53,54 and where the number of CRC survivors is increasing. 55We recommend future studies investigating associations between dairy intake and CRC prognosis to split total dairy into low-and high-fat dairy, as these seem to be associated to CRC prognosis in opposite directions.Future observational and ultimately intervention studies should confirm our findings before they can be translated to more personalised dietary guidelines for people with CRC who aim to improve their disease prognosis.

1
Newly-diagnosed stage I-III CRC patients that were included in the COLON study between 2010 and 2020 (n = 1945) Missing data on baseline dietary intake (n = 158) Missing recurrence data (n = 3) No (n = 8) surgery or unknown (n = 64) Metastasis before surgery (n = 1) or end of follow-up before surgery (n = 1) Missing data on baseline dietary intake (n = 3) Missing recurrence data (n = 2) No surgery (n = 19) Other tumour with metastasis diagnosed before surgery (n = 2) Total population for analysis (n = 2283) Newly-diagnosed stage I CRC patients that were included in the PLCRC-Flowchart representing participant selection from the COLON study and PLCRC-PROTECT study for the current analyses.TA B L E 1 Baseline characteristics of the study population by quartiles of dairy intake.