Non-communicable diseases such as CVD, stroke and cancer caused 71 % of deaths globally in 2018, ranging from 37 % in low-income countries to 88 % in high-income countries(1). These diseases are partly attributable to diet and lifestyle(Reference Forouzanfar, Afshin and Alexander2). Thus, primary prevention through shifting population diets towards more healthy habits is a global priority.
An unhealthy diet including red and processed meat, as well as saturated fat from meat, is among the risk factors associated with the leading causes of death(Reference Forouzanfar, Afshin and Alexander2), and the International Agency for Research on Cancer has classified consumption of processed meat as carcinogenic to humans(Reference Bouvard, Loomis and Guyton3,4) . The biological mechanisms underlying the association between red and processed meat and mortality are not yet fully established, but may include added nitrite and nitrate salts in processed red meat and the natural content of haem iron in red meat(Reference Pan, Sun and Bernstein5–Reference Etemadi, Sinha and Ward9).
Recent studies have explored associations between intakes of red meat, poultry and fish with all-cause mortality(Reference Zhong, Van Horn and Greenland10), and positive associations between processed meat intake and risk of premature mortality are consistently observed(Reference Zeraatkar, Han and Guyatt11). However, because dietary guidelines recommend that individuals consume more healthy foods while maintaining a stable weight, following such guidelines requires isoenergetic replacement of less healthy energy-containing foods. Few studies have directly modelled the health effects of food substitutions within habitual dietary intakes that entail following dietary recommendations to reduce red and processed meat intake(Reference Pan, Sun and Bernstein5,Reference Van den Brandt6,Reference Etemadi, Sinha and Ward9,Reference Farvid, Sepanlou and Willett12) . It is likely that a lower total red meat intake would be accompanied by an increase in other meat products, which may have independent health effects. Consuming poultry in place of total red meat has been suggested to be associated with lower risk of all-cause mortality, but the associations for unprocessed red meat or for substitutions with fish are less clear, particularly in European populations(Reference Pan, Sun and Bernstein5,Reference Van den Brandt6,Reference Etemadi, Sinha and Ward9,Reference Farvid, Sepanlou and Willett12) . Better understanding of the health effects of lowering total red meat, unprocessed and processed meat consumption while concomitantly increasing consumption of other meat products is necessary when updating national dietary guidelines. Therefore, we aimed to estimate the association of substitution of 150 g/week red meat with 150 g/week of poultry or fish on the outcomes total mortality and deaths due to cancer or CVD. We also investigated substitution of unprocessed and processed red meat separately, since intake of processed red meat has more consistently been associated with a higher risk of premature mortality than unprocessed red meat in studies that did not specify substitutions.
Methods
Study design and population
We used data from the Diet, Cancer and Health study. This cohort was initiated between December 1993 and May 1997, when 79 729 women and 80 996 men were invited to participate, of whom 57 053 consented (Supplementary Fig. 1). Participants were recruited from the greater Copenhagen and Aarhus areas. Inclusion criteria for participation were age between 50 and 64 years, born in Denmark and not previously registered in the Danish Cancer Registry. Informed consent was obtained from all participants included in the study. The Diet, Cancer and Health study was approved by the Danish Data Protection Agency and the regional ethical committees on human studies in Aarhus and Copenhagen. A detailed description of the Diet, Cancer and Health study has been published elsewhere(Reference Tjønneland, Olsen and Boll13).
Exposure assessment
All participants filled in a validated 192-item semi-quantitative FFQ before visiting one of the two study centres in Copenhagen and Aarhus. Development and description of the FFQ have been published elsewhere(Reference Overvad, Tjønneland and Haraldsdóttir14–16). Participants were asked to report their average dietary consumption over the past 12 months, in twelve response categories ranging from ‘never’ to ‘eight times or more per day’. The daily intake of foods was calculated for each participant by using a software programme developed for the Diet Cancer and Health study, FoodCalc, using sex-specific portion sizes(Reference Tjønneland, Overvad and Haraldsdóttir15,16) . In total, sixty-three questions covered the intake of meat, poultry and fish and dishes with meat as an ingredient. For the present study, meat was subdivided into total red meat, consisting of both unprocessed and processed red meat. Unprocessed red meat consisted of cooked and uncooked red meat, including beef, veal, pork, lamb and offal. Processed red meat consisted of all red meat that had undergone processing such as smoking, salting or curing, including sausages, salami, smoked or cooked ham, bacon and liver pate. Poultry included chicken (meat and skin) and turkey. Fish included all types of fish and shellfish, such as raw fish, cold cuts, roe, shellfish and prepared/cooked fish. The FFQ was checked for reading errors and missing information at the visit to the clinic; no missing data were accepted.
Covariates
After handing in the FFQ, all participants visited a study centre where they filled in a lifestyle questionnaire containing questions on health status, social factors and lifestyle habits. The potential confounders of the present study were chosen a priori, based on directed acyclic graphs created after a literature review (Supplementary Fig. 2–4). Information about educational level; time spent weekly in leisure time physical activities such as walking, jogging and gardening during summer and winter seasons; smoking habits; hypertension, diabetes and hypercholesterolaemia was retrieved from the lifestyle questionnaire. Answers to questions on whether a doctor had ever diagnosed hypertension, diabetes or high cholesterol were combined with answers to questions on medication and was used to determine the participants’ history of hypertension, diabetes or hypercholesterolaemia. At the visit to the study centre, height and weight were recorded by trained health professionals. Height was measured without shoes, and weight was recorded to the nearest 100 g using a digital scale. Information on alcohol consumption, fruit and vegetable intake and total energy intake was obtained from the FFQ.
Case ascertainment
The outcome of the present study was all-cause and cause-specific mortality. Cause-specific mortality (cancer and CVD) was defined from the International Classification of Diseases 10th revision, codes C00-C97 for cancer mortality and codes I00-I99 for CVD mortality. Total mortality and deaths due to cancer or CVD were obtained by record linkage to the Danish National Patient Register and the Cause of Death Register. Study end dates were 13 July 2015 for total mortality and 30 December 2013 for cause-specific mortality.
Exclusions
Participants diagnosed with cancer but not registered in the Danish Cancer Registry before baseline (due to processing delay) were subsequently excluded, as were those with a previous CVD diagnosis (myocardial infarction or stroke). Furthermore, participants with missing information on covariates were excluded.
Statistical methods
Baseline characteristics of the participants were summarised using medians and 25th and 75th percentiles for continuous variables and proportions for categorical variables.
The participants were followed up from the date of entry into the Diet, Cancer and Health study until the date of death, emigration, loss to follow-up or the relevant outcome end date, whichever occurred first. The association between substitution of red meat with poultry or fish and death was investigated by Cox proportional hazards models, in which hazard ratios (HR) and corresponding 95 % CI were calculated. The participant’s age was included as the underlying timescale. The observation time was calculated as the time between study entry and end of follow-up for each participant. The analysis was carried out for total mortality, deaths due to cancer and deaths due to CVD.
Intakes of 150 g/week, reflecting a usual serving size of meat or fish in Denmark, were investigated for each meat item. To construct substitution models, a sum-variable including each participant’s total intake in servings of processed and unprocessed red meat, poultry and fish was entered into the Cox regression model in addition to each of these food groups separately, except for the food group to be replaced. For example, to investigate substitution of total red meat with poultry or fish, we held the total intake of unprocessed and processed red meat, poultry and fish constant using the sum-variable and added the specific variables for poultry and fish into the Cox regression model leaving out the specific variable for total red meat. The calculated HR may be interpreted as the HR for substitution of the omitted variable. The calculated HR in this example reflect substitution of 150 g/week total red meat with 150 g/week of poultry or fish. This method has been described previously(Reference Ibsen, Warberg and Würtz17,Reference Würtz, Hansen and Tjønneland18) .
All analyses were stratified by date of enrolment (quartiles) to allow for differences in the baseline hazards and adjusted for total energy (continuous, kJ/week) and age (model 1). In model 1a, the following risk factors were additionally included: sex and tertiary educational level (categorical: none, short, medium or long) as a measure of socio-economic status, BMI (categorical: <18·5 kg/m2, 18·5–25 kg/m2, <25–30 kg/m2, >30 kg/m2), alcohol consumption (abstainer, ≤0·5 units, >0·5–1 units, >1–2 units, >2–4 units, ≥5 units), physical activity (<30 min/d, ≥30 min/d) and smoking (never, former, current < 15 g tobacco/d, current 15–25 g tobacco/d, current > 25 g tobacco/d). In model 2, potential intermediary lifestyle diseases were additionally included: hypertension (yes, no, do not know), hypercholesterolaemia (yes, no, do not know) and diabetes (yes, no, do not know). Model 3 included, besides the risk factors in model 1a, fruit intake (continuous, g/week) and vegetable intake (continuous, g/week).
To investigate possible effect modification, we stratified models of substitution analyses by sex or BMI (categorised, <18·5 kg/m2, 18·5–25 kg/m2, <25–30 kg/m2, >30 kg/m2).
The proportional hazards assumption was investigated using Schoenfelds Residuals Test and diagnostic plots. No deviation from proportionality was detected. All analyses were performed using Stata 15.0 (StataCorp LLC).
Results
After excluding participants with cancer (n 569) or CVD (n 1443) before baseline, missing lifestyle questionnaires (n 42) and missing data for covariates (n 770), 54 229 participants were available for the analyses. During a mean follow-up time of 16·1 years, 8840 deaths occurred. Among cause-specific deaths, 4567 were due to cancer and 1816 due to CVD.
Table 1 presents baseline characteristics of the entire cohort and the participants who died during follow-up. Those who died during follow-up had a higher intake of red meat (unprocessed and processed) and a lower intake of poultry and fish than the cohort as a whole. Furthermore, they were more likely to be men, older, overweight or obese, more physically active, and current smokers and have a higher alcohol consumption, and lower consumption of fruit and vegetables than the cohort as a whole. They were also more likely to have a history of hypertension, hypercholesterolaemia and diabetes.
Baseline characteristics for the entire diet, cancer and health cohort, and those participants who died during follow-up
(Numbers and percentages; mean values and standard deviations; median and 25th and 75th percentiles)
Table 2 shows the associations for substitution of total red meat, unprocessed red meat or processed red meat with poultry or fish. In the adjusted model (model 1a), substitution of 150 g/week total red meat with 150 g/week poultry or fish was associated with a lower total mortality (HR 0·97 (95 % CI 0·94, 1·00), HR 0·98 (95 % CI 0·96, 1·00). After additional adjustment for fruit and vegetables intakes, the associations were slightly weaker (model 3). The association with total mortality was only statistically significant for substitution of processed red meat with poultry (HR 0·95 (95 % CI 0·92, 0·98)) or fish (HR 0·96 (95 % CI 0·94, 0·98)). The association with death due to cancer when substituting red meat with poultry or fish showed a lower risk, but was only statistically significant for substitution of processed red meat with poultry (HR 0·93 (95 % CI 0·90, 0·97)) or fish (HR 0·93 (95 % CI 0·90, 0·96)). For deaths due to CVD, we found no associations for the investigated substitutions. For example, when substituting 150 g/week total red meat with poultry, it showed a HR 1·03 (95 % CI 0·98, 1·09) and with fish, it showed a HR 1·00 (95 % CI 0·96, 1·04). The pattern of associations was similar after additional adjustment for hypertension, hypercholesterolaemia and diabetes (model 2).
Substitution of total red meat, unprocessed red meat and processed red meat with poultry or fish and risk of death, per 150 g/week
(Hazard ratios (HR) and 95 % confidence intervals)
* Adjusted for age, total energy and date of enrolment.
† Further adjusted for sex, educational level, BMI, alcohol consumption, physical activity and smoking.
‡ Further adjusted for hypertension, hypercholesteraemia and diabetes.
§ Adjusted for sex, educational level, BMI, alcohol consumption, physical activity, smoking (model 1a) and for fruit consumption and vegetable consumption.
Results of analyses stratified by sex, and by BMI group, were similar to the main results (data not shown).
Discussion
In this follow-up study, we found an inverse association with total mortality when replacing total red meat and processed red meat with poultry or fish. Statistically significant inverse associations for processed red meat were only found for replacement with poultry or fish for total mortality and death due to cancer. However, inverse associations were found when replacing total red meat with poultry or fish for both total mortality and death due to cancer. The patterns of associations did not differ by sex or BMI.
Our study has some limitations and some strengths. Due to the very low proportion of participants lost to follow-up, it is unlikely that selection bias has influenced our results. Information about habitual diet was obtained from a self-administered FFQ, the data from which are likely to be affected by random as well as systematic measurement error. Random error is likely to bias our estimates towards the null. Systematic errors, such as underreporting of red meat intake according to sex, could also bias our results towards the null. However, at the time of dietary intake assessment, the major focus for healthy diets was lowering saturated fat intake and meat as such was considered neither particularly healthy nor unhealthy in the Danish population. The Danish dietary guidelines from 1994 recommend choosing low-fat meat products and eating fish often. Thus, in our study, while differential misreporting of foods containing saturated fat is possible due to social desirability bias, it is unlikely that this would influence self-report of total meat intake. We modelled substitutions of red meat, fish and poultry with and without adjustment for fruit and vegetables. When adjusting for other dietary components in model 3, we emulate the pure association between choosing fish or poultry instead of red meat. However, it is possible that dishes including fish or poultry differ in vegetable content compared with red meat dishes, and thus it may be that a greater overall health benefit could be achieved by shifting dietary patterns from red meat dishes to fish- or poultry-based dishes. We therefore also presented the results in model 1a. We used directed acyclic graphs to illustrate our assumptions about the relationships between potential confounders and to guide our choice for adjustment (Supplementary Fig. 2–4). However, residual confounding cannot be ruled out. Our models assessing all-cause mortality are not affected by competing risks. However, the analyses of cause-specific death should be interpreted as the instantaneous risk of the particular cause of death, given survival from all other causes of death.
We compared participants who had an identical total energy intake and an identical total intake of red meat, poultry and fish, but for whom the intake of each of the three meat types differed. Controlling for the total amount of red meat, poultry and fish and at the same time including two of the three meat types in the statistical models allowed us to specify the substitutions between these three meat types. The substitution method is a relatively new investigation method in studies of meat intake and health outcomes, and Pan et al.’s(Reference Pan, Sun and Bernstein5) study is the earliest paper that we identified to use this method in relation to mortality. When using this method, it is possible to investigate the effect of modelled substitutions of intake of specific foods, mimicking changes in dietary composition. This is likely to be more useful for public health strategies to improve health than targeting changes in total energy intake, unless physical activity or body weight is similarly changed(Reference Willett19). For this reason, we chose to model substitutions of 150 g/week, corresponding to one serving size, thus facilitating the translation of our results into dietary recommendations that would be feasible in this population. The magnitudes of the associations we determined were modest, but would be consistent with a meaningful impact on preventing premature mortality in the population, upon shifting dietary habits as recommended by dietary guidelines were the effect causal.
A limitation of our study was that we investigated all deaths due to cancer and all deaths due to CVD. It is likely that diet affects risk of cancer and CVD mortality differently depending on the site of the cancer or the specific vascular disease. However, by investigating all deaths due to these causes, we reduced potential misclassification in the outcome that may arise due to errors in death certificate completion. Further, the associations for total mortality and total cancer and total CVD are of major importance from a public health point of view.
Four previous studies have investigated the association of substitution of red meat with poultry and fish with the risk of death. Pan et al. found that substitution with one serving per day (85 g/day) total red meat with one serving poultry per day was associated with statistically significant lower risk for total mortality (HR 0·86 (95 % CI 0·82, 0·91)) and a statistically significant lower risk when substituting with fish (HR 0·93 (95 % CI 0·90, 0·97))(Reference Pan, Sun and Bernstein5). These findings from a US cohort are in agreement with the findings in our study. In another US cohort, Etemadi et al. found that for each 20 g/1000 kcal increase in daily intake of white meat (including both unprocessed and processed white meat and fish) and same decrease in processed red meat and unprocessed red meat, the HR for all-cause mortality was 0·92 (95 % CI 0·91, 0·93), the HR for deaths due to cancer was 0·93 (95 % CI 0·92, 0·94) and 0·92 (95 % CI 0·91, 0·94) for deaths due to CVD(Reference Etemadi, Sinha and Ward9). Although the substituted amounts were different, the association for total mortality and deaths due to cancer points in the same direction as the findings in our study. Van den Brandt found that substitution of 50 g/d processed meat with 50 g/d poultry was associated with a lower risk of total mortality (HR 0·81 (95 % CI 0·63, 1·05)), a lower risk of deaths due to cancer (HR 0·81 (95 % CI 0·61, 1·07)) and a lower risk of deaths due to CVD (HR 0·76 (95 % CI 0·56, 1·05)) in a Dutch cohort, albeit with statistical uncertainty(Reference Van den Brandt6). In contrast, our results indicate that substitution of processed meat for poultry is primarily associated with total mortality and death due to cancer. Finally, in a Middle Eastern cohort, Farvid et al. found that substituting one serving/d of fish for one serving/d of total red meat was associated with a lower risk of all-cause mortality (HR 0·63 (95 % CI 0·47, 0·84))(Reference Farvid, Sepanlou and Willett12), but the mean consumption of total red meat in the Iranian study population was only 0·19 serving/d. The finding in Farvid et al. is in agreement with our findings, taken the cultural context and substitution amount in consideration.
Multiple mechanisms might explain the observed findings. Red meat is a source of saturated fat and haem Fe, and processed red meat additionally typically contains nitrate, nitrite and Na salts. In the gastrointestinal tract, nitrite reacts with amines to form nitrosamines, which are carcinogenic to humans(Reference Bouvard, Loomis and Guyton3). The formation of these redox active compounds is promoted by haem Fe(Reference Bouvard, Loomis and Guyton3). Haem iron and nitrite are also pro-oxidants that can promote oxidative damage and inflammation in vital organs and can lead to arteriosclerosis(Reference Corpet8,Reference Halliwell and Gutteridge20) . Furthermore, replacement of saturated fat with PUFA has showed significant advantages in CVD development, and sodium consumption contributes to hypertension, which is a major risk factor for CVD and CVD death(Reference Piepoli, Hoes and Agewall21–Reference Li, Hruby and Bernstein24). Poultry and fish are generally not preserved using nitrite-based salts and are often lower-fat alternatives to red meat. Fatty fish contains n-3 PUFA, which are beneficial to cardiometabolic health and might have a beneficial effect in cancer prevention and treatment(Reference Farvid, Ding and Pan23,Reference Mozaffarian, Micha and Wallace25,Reference D’Archivio, Scazzocchio and Vari26) . Substitution of red and processed meat with these alternatives may thus have a protective effect on incidence of diseases and thus premature death due to cancer or CVD.
In conclusion, replacing total red meat or processed red meat with poultry or fish was associated with a lower risk of total mortality and deaths due to cancer. The strength of the associations was modest, but our findings may have an important impact in the prevention of premature mortality, if entire populations shifted their dietary habits.
Acknowledgements
The authors thank the Danish Cancer Society and the staff at the Diet, Cancer and Health study for the collection and administration of data.
The primary data collection was funded by the Danish Cancer Society. The funding agencies had no role in the design, analysis or writing of this article. The Diet, Cancer and Health study was approved by the Danish Data Protection Agency and the regional ethical committees on human studies in Aarhus and Copenhagen.
The authors’ contributions are as follows: T. B. N., C. C. D., A. M. L. W., A. T. and K. O. contributed to the study design; A. T. and K. O. collected the data; T. B. N. performed the statistical analyses under guidance of C. C. D. and A. M. L. W.; T. B. N. wrote the first manuscript draft. All authors were involved in the interpretation of the data and critical revision of the manuscript; T. B. N. had primary responsibility for the final content. All the authors read and approved the final version of the manuscript.
The authors declares that there is no conflict of interest.
Supplementary material
For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114521001252
