4.1. Colorectal Cancer
We found no significant association between red meat or processed meat intake and CRC risk. For poultry, however, high compared with low intake increased CRC risk significantly by 62%, but we found no increased CRC risk per 100 g poultry per day.
In accordance with our result on red meat, several prospective cohort studies representing more than 10 European countries, including Denmark, found no association between red meat intake and CRC risk [16
], but an increased hazard ratio of CRC per 1 serving of red meat per day was seen in two American prospective cohorts [20
]. In meta-analyses covering America, Australia, European and Asian countries, a positive association between red meat intake and CRC risk has been observed [19
]. The association was stronger in Asian and Australian cohorts compared with European and North American cohorts [21
]. The latest meta-analysis performed by the World Cancer Research Fund Continuous Update Project found that red meat intake was positively associated with CRC risk [23
It is difficult to obtain and compare information about actual meat intake (g/day) for different types of meat in different cohorts, which affects the possibility to compare outcomes in studies of effects of high versus low meat intake. For example, high red meat intake in an Asian cohort may be similar in magnitude to low red meat intake in some Western cohorts.
In contrast to our results on processed meat, others found positive associations between processed meat intake and CRC risk in American, Australian, Asian, and European cohorts [16
]. However, no association was found in the Danish Diet, Cancer and Health cohort study [18
Studies on the effects of meat intake from different countries and continents can be difficult to compare because the proportions of the different types of red and processed meat differ significantly between regions. The types of meat—both red and processed meat—constitute different hazards due to their structure and composition. Moreover, certain meat subtypes may be more prevailing in unhealthy diets than others, which can affect the risk estimates. Therefore, analyses on effects of meat subtypes can contribute to our understanding of differences observed in different cohorts and are warranted in future studies.
In a meta-analysis comparing the highest versus lowest red meat intake in Asian and European cohorts, Carr et al. [24
] found that beef intake was associated with an increased risk of CRC in European cohorts but no association was found for pork. In a Danish cohort, no associations were seen for beef or pork intake and colon cancer risk but beef intake was associated with decreased risk and pork intake with increased risk of rectal cancer [18
]. We had too few cases to make subgroup analysis on red meat intake, but from analysis of dietary patterns among the participants [25
], we know that pork constitutes a slightly higher part of their red meat intake than beef/veal, which may have affected our findings.
For poultry intake, our results were in contrast with what others have found. No association between poultry intake and CRC risk was reported by the World Cancer Research Fund Continuous Update Project [23
] or seen in European cohorts [16
]. A decreased CRC risk was associated with 50 g poultry increment per day in a meta-analysis including prospective cohort studies from America, Australia, Europe, and Japan [26
]. Thus, more studies are needed to confirm our findings.
A pronounced difference in meat content in high-meat diets with different healthy eating indices was found by Kappeler et al. [4
]. Thus, comparing groups with low and high meat intake without considering dietary quality and what foods replace the meat will simultaneously be a comparison of healthy and unhealthy diets. Therefore, we analysed our data by looking at the effects of meat intake stratified by DGCS to reduce the confounding from dietary quality. However, when stratified by DGC, we found no statistically significant differences in the associations between meat intake and CRC risk in low-compliers and high-compliers.
Norat et al. [16
] found that the CRC risk associated with high intakes of red and processed meat was more pronounced in participants from a European cohort including Denmark with low and medium fibre intake (≤26–28 g/day) compared with those with high fibre intake (>26–28 g/day). Others have found that in two US cohorts, an increase in total fibre, cereal fibre, or whole-grain intake of 5 g per day reduced CRC risk by 7–25%, while fibres from fruit and vegetables did not have such effect [27
]. From dietary pattern analyses of our participants’ diet, we know that those who comply well with dietary guidelines had both a high whole-grain intake and total fibre intake, but it apparently did not influence the CRC risk associated with meat intake.
4.2. All-Cause Mortality
We found no significant associations between red meat, processed meat, and poultry intake and all-cause mortality.
Similar results were found for red meat in a large American cohort [4
] but not in another American cohort [6
], and not in European cohorts [6
]. Three meta-analyses showed no associations between red meat intake and all-cause mortality risk [5
], while one meta-analysis showed that each additional intake of 100 g red meat/day was positively associated with all-cause mortality [30
In contrast to our results, in a European cohort including Denmark, intake of processed meat was positively associated with all-cause mortality [28
], which was also the result of four meta-analyses [5
In a recent meta-analysis, Han et al. found a small, positive association between red and processed meat intake and cancer mortality, but the evidence was rated to be of low certainty [31
White meat (including chicken, turkey, and rabbit) intake was not associated with all-cause mortality in meta-analyses of prospective cohort studies [5
]. Likewise, no association was found between poultry intake and CRC mortality in a dose-response meta-analysis of prospective cohort studies [26
], and no association was found between poultry intake and cancer mortality in a meta-analysis of prospective cohort studies [32
In our study, a diet composition that did not comply well with the official, quantitative Danish dietary guidelines (independent of meat content) was significantly associated with mortality risk in the least adjusted model (adjusted by sex and age) (HR 1.66; 95%CI 1.32–2.10). However, DGCS was not significantly associated with mortality risk in the multivariate model (adjusted by sex, age, educational attainment, ethnicity, smoking, physical activity, alcohol, BMI, and total energy intake), and p for trend showed no significant effect of DGCS.
Kappeler et al. [4
] found a 27% decreased mortality risk among Americans with the top third Healthy Eating Index score (developed by the US Department of Agriculture) compared with the bottom third Healthy Eating Index score. Unfortunately, these authors did not estimate the all-cause mortality risk in participants with different meat intake stratified by dietary quality.
4.3. Strengths and Limitations
Our study had several strengths. The studied outcomes were based on national registers with high validity and completeness, and we included complete information on migration and death ensuring complete follow-up of the study cohort. The linkage also enabled us to include only incident cases of disease and to minimise the risk of reverse causality as we excluded those with disease before baseline. The study included comprehensive information on dietary components, which made it possible to evaluate if associations differed with DGC. The diet registration for each participant included seven days including weekend days, and the data collection process in the study population covered all seasons to allow for seasonal variations in dietary data on study population basis.
However, the study also had limitations. The dietary surveys were representative regarding gender and age. However, in the latest surveys, participants with short education were under-represented, which may limit the generalisability of the findings. In addition, the study only included one dietary registration for each individual. Therefore, it was assumed that the diet composition did not change during follow-up, but if the population had large variations in DGC during follow-up, this would influence the estimated associations. Finally, as mentioned previously, the size of the study population affected the power to identify statistically significant associations, especially in analyses on interactions between meat intake and DGC, where the numbers of participants in groups were low.
In the analyses, BMI was included as a confounder, as is common practice in similar studies. However, it is likely that BMI is a mediator instead of a confounder in the presented associations, implying that the presented results have been over-adjusted. Analyses without BMI in the model (data not shown) showed that inclusion of BMI only mildly attenuated the estimates, and that the results on low DGCS did not become significant when BMI was not included in the model.
We did not find statistically significant associations between meat intake and CRC or mortality risk. However, the ability to reach statistically significant results is influenced by many factors. For example, since the study population was 15–75 years at baseline, a large proportion of the population was too young to be at real risk of developing CRC. This is why we only studied CRC risk among individuals aged 50 years and older. Thus, the number of outcomes could be an explanation why the associations between meat intake and the CRC risk was non-significant. Similarly, a large proportion of the population were too young to be at an appreciable mortality risk.
Analyses of dietary patterns in our cohort showed that a low dietary content of one type of meat, e.g., poultry, was associated with a high dietary content of other types of meat, e.g., red meat [25
]. Thus, dietary content of meat types could be confounders. Before we made the estimates of associations between meat intake and disease risk, it was not known to us exactly which types of dietary meat content were associated, and, therefore, we did not include different types of meat in the same analyses. However, in future analyses, it may be appropriate to take dietary content of other types of meat or other replacement foods into consideration.
Another limitation was that the size of the study population restricted our opportunity to study differences between those with very low and those with a very high meat intake. In analyses of CRC risk, we were only able to divide the population’s meat intake into two groups instead of quartiles. This introduced some arbitrariness around cut-off values of meat intake since we split the population into two groups without having a meaningful difference for meat consumed around the median. However, in the interpretation of results, we also focused on estimates of associations with meat intake on a continuous scale, which did not suffer from this limitation.