Effect of Lower Versus Higher Red Meat Intake on Cardiometabolic and Cancer Outcomes

Abstract

Observational studies have reported that intake of red meat is associated with cardiometabolic disease and cancer (18). Dietary guidelines from the United States, United Kingdom, and the World Cancer Fund/American Institute for Cancer Research recommend limiting intake of red and processed meat (810). Such recommendations are primarily based on observational studies that are at high risk for confounding. Randomized trials generally provide higher-certainty evidence supporting causal relationships (11, 12). The few systematic reviews of trials addressing red meat consumption have evaluated only surrogate outcomes, such as blood pressure and lipid levels (1315). In this systematic review of randomized trials, we investigate the effect of lower versus higher red meat intake on the incidence of major cardiometabolic and cancer outcomes. The review was performed by the Nutritional Recommendations (NutriRECS) working group as part of a new initiative to develop trustworthy guideline recommendations in nutrition (16). In addition to this review, we performed 4 parallel systematic reviews that focused on observational studies addressing the effect of red and processed meat consumption on cardiometabolic and cancer outcomes (1719), and a review of health-related values and preferences related to meat consumption (20). These reviews were used to underpin guideline recommendations for consumption of red and processed meats (21). Methods We registered the systematic review protocol in PROSPERO (CRD42017074074) on 10 August 2017 (22). Data Source and Searches We searched MEDLINE, EMBASE, CENTRAL (Cochrane Central Register of Controlled Trials), CINAHL (Cumulative Index to Nursing and Allied Health Literature), and the Web of Science from inception until July 2018, and MEDLINE from inception through to April 2019, with no restrictions on language or date of publication (Section I of the Supplement). We also searched ProQuest Dissertations and Theses Global (1989 to 2018); trial registries, including ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform Search Portal, to April 2019; and bibliographies of eligible studies and relevant systematic reviews. Supplement. Supplementary Material Study Selection We included English-language and nonEnglish-language reports of randomized trials of adults allocated to consume diets that included varying quantities of unprocessed red meat (measured as servings or times/week, or as g/d) or processed meat (meat preserved by smoking, curing, salting, or adding preservatives) for 6 months or more (23). Eligible trials compared diets lower in red or processed meat with diets higher in red or processed meat that differed by a gradient of at least 1 serving per week (Table 1). If a trial reported more than 2 study groups (24, 25), we used the groups with the largest gradient in red meat intake or combined groups if red meat intake was equal. Studies in which more than 20% of the participants were pregnant or had cancer or a chronic health condition, other than cardiometabolic diseases, were excluded. Table 1. Study Characteristics Outcomes of interest, which were determined a priori and in consultation with the guideline panel, were all-cause mortality, cardiovascular mortality, adverse cardiometabolic events and major morbidity, cancer mortality and incidence, quality of life, and surrogate outcomes (weight, body mass index, blood lipid levels, blood pressure, and hemoglobin level) (22). Pairs of reviewers screened titles and abstracts for initial eligibility and reviewed the full text of potentially eligible studies, independently and in duplicate. Reviewers resolved disagreements by discussion and third-party adjudication if needed. Data Extraction and Quality Assessment Using standardized, piloted forms, pairs of reviewers conducted calibration exercises and independently extracted information on study design, participant characteristics, interventions, comparators, and outcomes of interest and resolved disagreement by discussion or, if necessary, third-party adjudication. When details related to methods or results were unavailable or unclear, we contacted study authors for additional information. Reviewers, independently and in duplicate, assessed the risk of bias of eligible trials by using a modified version of the Cochrane Collaboration s risk of bias instrument for randomized trials (2628). The modified version categorizes risk of bias as definitely low, probably low, probably high, or definitely high for each of the following domains: sequence generation, allocation sequence concealment, blinding, missing participant outcome data, selective outcome reporting, and other bias (for example, prematurely terminated studies). We resolved any disagreements by discussion or, if necessary, third-party adjudication. We collapsed ratings of probably low and definitely low into low risk of bias and ratings of probably high and definitely high into high risk of bias. Among the 8 risk of bias domains, we considered a study to be at high risk of bias if, at the outcome level, 2 or more domains were at high risk of bias (Section I of the Supplement). Data Synthesis and Analysis We reported risk ratios (RRs), hazard ratios (HRs), and mean differences (MDs) with their 95% CIs for the lowest versus highest category of red meat intake, at the last reported time point. We used the HartungKnappSidikJonkman approach to pool data (29, 30). To calculate absolute risk differences, we multiplied the effect estimate for each outcome with the population risk estimates from the Emerging Risk Factors Collaboration study for cardiometabolic outcomes (31) or from GLOBOCAN for cancer outcomes (32, 33) and, when this was not available, the control group estimate from the largest study (Section I of the Supplement). We investigated heterogeneity by using the Cochran Q test and the I 2 statistic (34). We used R Project, version 3.3.0 (R Foundation for Statistical Computing), for all analyses. To rate the certainty of the evidence for each outcome, we used the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) approach (11, 3539). Reviewers, independently and in duplicate, assessed the certainty of evidence for each outcome, and resolved disagreements by discussion. Role of the Funding Source This systematic review was conducted without financial support. Results Study Selection Electronic searches yielded 13190 unique articles (Appendix Figure). Of these, 24 articles (24, 25, 4062) reporting on 12 unique randomized trials met eligibility criteria. In 2 instances, authors provided clarification about study characteristics or outcomes: Turner-McGrievy and colleagues (24) clarified the aggregated change in weight for vegan/vegetarian and semi-vegetarian/omnivorous groups, and Griffin and associates (44) clarified reported effect estimates. Appendix Figure. Evidence search and selection. Study Characteristics Trials ranged in size from 32 to 48835 participants (Table 1). The mean age of participants ranged from 22.4 to 70.9 years. The largest study, the Women s Health Initiative (WHI), enrolled postmenopausal women (45). Five trials, including the WHI, enrolled overweight and obese participants (24, 25, 41, 45, 59, 60); 5 focused on participants with medical conditions, such as diabetes or hypercholesterolemia (42, 43, 57, 58, 61); and 1 enrolled older (>64 years) healthy individuals (41). Only 1 trial explicitly reported participants consumption of both unprocessed red meat and processed meat (62). All trials used parallel designs, except for a small crossover trial in patients with hypercholesterolemia (57). Intervention and control diets varied widely. The primary protein intake in the low red meat group was from plant sources in 4 trials (40, 60, 58, 61); from animal protein sources in 5 trials (25, 43, 44, 57, 59); and from a mix of plant and animal protein in 3 trials (24, 41, 42). The largest trial, the WHI trial, compared a low-fat dietary intervention aimed at reducing total dietary fat to 20% with a usual diet group given diet and health-related materials (4556). The duration of interventions ranged from 6 months (24, 25, 41, 59) to 12 years (51). Risk of Bias Trials were most often rated as high risk of bias for lack of blinding (not possible for participants) and missing outcome data overall (Supplement Table 1). However, some trials were rated as low risk of bias for specific outcomes (all-cause mortality, cardiovascular disease, type 2 diabetes, adenocarcinoma) because there were either more outcome events than missing data for dichotomous outcomes or there were less than 10% missing data for continuous outcomes. Selective reporting bias was detected in 4 trials (40, 42, 44, 57). Other biases included a nonpaired analysis of data from a crossover trial (57) and early termination for benefit in the Lyon Diet Heart Study (42). Outcomes None of the trials reported on a combination of fatal and nonfatal myocardial infarction, fatal infarction, nonfatal coronary heart disease, prostate cancer, and satisfaction with diet. Only 2 trials, the Lyon Diet Heart Study and the WHI trial (42, 54), addressed all-cause mortality and other patient-important, major morbid cardiovascular outcomes. The Lyon Diet Heart Study reported an implausibly large treatment effect, potentially due to stopping the trial early for benefit, and had a sample size (605 participants) more than 80 times smaller than the WHI trial (48835 participants); for this reason the 2 trials were not pooled (63). Results presented below and in Table 2 regarding all-cause mortality and cardiovascular outcomes are based on the WHI trial results. Results of the Lyon Diet Heart Study are presented in Section II of the Supplement. Table 2. Summary of Findings for Lower Intake of Red Meat* and Mortality Outcomes All-Cause Mortality and Cardiometabolic Outcomes Low-certainty evidence from the W