Association of changes in red meat consumption with total and cause specific mortality among US women and men: two prospective cohort studies
BMJ 2019; 365 doi: https://doi.org/10.1136/bmj.l2110 (Published 12 June 2019) Cite this as: BMJ 2019;365:l2110All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
The study by Zheng and colleagues is a textbook example of ignoring multiple testing and multiple modeling with the use of a statistical significance threshold of 0.05 (1). The authors did not discuss issues with multiplicity in their study limitations.
For instance, I counted 585 confidence intervals from the Tables and Figures. Additionally, I counted 2 cohorts, 3 exposures, and 25 or 26 adjusted covariates (selected from a set of n covariates) for all-cause mortality from Table 3. One can specify 100,663,296 or 201,326,592 Cox regression models.
Indeed, it would be interesting to see how hazard ratios and confidence intervals would differ between these hundreds of millions of models. Further, I counted 8 outcomes from Supplemental Figure 1 which amplifies the issues with multiple modeling.
References
1. Zheng Y, Li Y, Satija A, et al. Association of changes in red meat consumption with total and cause specific mortality among US women and men: two prospective cohort studies. BMJ 2019;365:I2110.
Competing interests: I like statistics.
Our study found that increasing red meat (beef, pork, lamb) intake by 0.5 serving or more per day over time was associated with increased risk of mortality compared to those who did not appreciably change their eating habits. We also found that the association was stronger for processed meats such as sausages, bacon, and hot dogs than unprocessed red meat, which is primarily beef in the US.
Competing interests: No competing interests
Because the risk of chronic diseases and mortality in a population is determined by multiple diet and lifestyle factors, changes in one factor do not necessarily correspond to the secular trends of diseases. Although there has been some improvement in overall diet quality such as a modest reduction in red meat consumption in the US population, other risk factors especially obesity have increased substantially in the past decades. Our analyses indicate that increasing red meat intake may contribute to higher mortality independent of other risk factors. Our previous analyses have shown that other lifestyle factors such as maintaining a healthy weight, being physically active, and avoiding smoking are also important in promoting longevity (1).
Reference:
1. Li Y, Pan A, Wang DD, Liu X, Dhana K, Franco OH, Kaptoge S, Di Angelantonio E, Stampfer M, Willett WC, Hu FB. Impact of Healthy Lifestyle Factors on Life Expectancies in the US Population. Circulation. 2018 Jul 24;138(4):345-355.
Competing interests: No competing interests
We thank Henderson for the thoughtful comment on potential conscientiousness bias. To address this comment, we conducted a preliminary analysis and did not observe a significant association of changes in red meat consumption with injury or accidental deaths, either in women or in men (either p values for trend >0.05). In the pooled results of women and men, the overall 8-year HRs and 95%CIs from model 2 for injury or accidental deaths in the groups of decreased red meat consumption (">0.5 serving /d" group and "0.15-0.5 serving /d" group), no change group, and the groups of increased red meat consumption (">0.5 serving /d" group and "0.15-0.5 serving /d" group) were 0.87 (0.60-1.28), 1.14 (0.84-1.54), 1, 0.93 (0.67-1.30), and 0.92 (0.64-1.33), respectively (p for trend= 0.94). Of note, because excess body weight has been associated with an increased risk of injury-related deaths (1), one cannot assume that these deaths are completely unrelated to diet.
The second comment is about the age difference in our study groups. In Table 2 of the paper, we presented the age distribution across our study groups, and we noted in the table legend that “values are standardized to the age distribution of the study population except for age.” The age difference across groups was between those who decreased red meat consumption and those maintained red meat consumption, and such difference was small (1.2 years). In our previous analyses on dietary fat and mortality, the mean difference at baseline age was 2.3 years across quintiles of saturated fat intake (2), although the updated age difference across multiple 4-year follow-up cycles over 32 years was much larger. All our analyses were adjusted for age in years.
Reference:
1. Jia T, Tynelius P, Rasmussen F.U-shaped association of body mass index in early adulthood with unintentional mortality from injuries: a cohort study of Swedish men with 35 years of follow-up. Int J Obes (Lond). 2016 May;40(5):809-14.
2. Wang DD, Li Y, Chiuve SE, et al. Association of Specific Dietary Fats With Total and Cause-Specific Mortality. JAMA Intern Med. 2016;176(8):1134–1145.
Competing interests: No competing interests
Interestingly, we can suppose that red meat consumption is associated with a higher rate of deaths. What about the consumption frequency or type of meat in this pattern?
Competing interests: No competing interests
We are passing from the pharmaceutical to the nutraceutical era--from a health system that turns around the care of very sick patients to a new but very ancient system where the main goals are disease prevention and promotion of good habits for living healthy, wealthy and wise. Three parameters determine the biological value of a food-- its protein composition, its essential amino acid content, which forms the primary structure of its proteins, and the digestibility of these nitrogen compounds--that is, the ability of the organism to assimilate these structural biomolecules. The foods of greatest biological value to the human species are milk, meat and eggs. For these mentioned above reasons it is interesting for me not only to read the abstract, but also the content of the article published in this journal BMJ 2019;365:l2110, and written by several well-known and prestigious researchers from China and the US.
The authors begin with a background related to the assumption that red meat consumption, especially industrially processed meats, is associated with an increased risk of several chronic diseases. They obtained that information from exhaustive revision of systematic reviews and meta-analysis of prospective studies. This is followed by detailed information about the methods used for analyzing changes in food consumption over eight years and studying its association with disease and mortality. The people studied by the authors were samples drawn from a good representation of the North American population. Participants were US women from the Nurses’ Health Study and US men from the Health Professionals follow-up study. The authors have used multivariable models for analyzing and interpreting data. Then, in the discussion section they wrote on possible explanations of the results and the implications of their analysis and observations. They recognize that the adverse effect of red meat consumption on the risk of death may be attributable to a combination of factors. Although they have used statistics as a way of making good sense of data in a systematic way and they have used the appropriate model for interpreting results and making the corresponding decisions, the conclusion in the last paragraph could be used for commercial purposes by unscrupulous people who use untrustworthy statistics for promoting harmful unhealthy lifestyles.
A lot of other genetic, physical, and lifestyle factors were not considered by authors, but these factors are undoubtedly of great importance in the etiology and pathogenesis of chronic disease. I think this mistake of judgment may be because the data’s matrix used doesn’t have the necessary variables for making a complete analysis. Taking into account that meat consumption is almost always accompanied by the intake of bad quality foods such as junk food, the problem may be not due to meat consumption but to, say, excessive ingestion of foods rich in carbohydrates or other fuel biomolecules.
On the basis of the style of the discussion section, it seems to me that the conclusion is drawn from biased estimations, and this is very dangerous for the health sector. In making conclusions we must act with caution and integrity. We should be cautious about the extent to which we can generalize from our always limited knowledge and experience.
Competing interests: No competing interests
One possible issue with the HNS/HPFS studies is the exposure of the participants to the opinions of the researchers. Medical personnel in the USA can hardly be unfamiliar with the idea that meat causes disease, which is communicated to them in various media, including from authorities such as the AHA and ADA, and mostly based on the earlier epidemiology of the Harvard Chan group. They are less likely to be made aware of research from other countries with different conclusions.
This creates an exceptional risk of conscientiousness bias, as the subjects and researchers are caught in a feedback loop where the health-conscious are more likely to take the advice of the researchers. This is indicated by lack of an association in never-smokers and those who overall eat a "healthy diet" between higher meat intake and mortality.
One way to test for this effect is to publish the associations for causes of death that relate to conscientiousness but are not plausibly related to diet, such as vehicular accidents, violence, suicide and legal executions, a category available to the researchers. Given the scale of mortality due to the opioid epidemic in the US and the exposure of medical professionals to opioids, accidental overdoses are another potential test of value. Studies of this sort adjust for alcohol intake (even though this is known to be under-reported, and may not be reported at all when use is problematic) but do not yet include opioids.
There is, as far as I can find, only one study looking at the correlation between red meat and accidental/violent death, the 2009 NIH-AARP study.[1] In this study, the adjusted HR association between red meat (highest vs lowest quintile) and accidental/violent death in men was 1.26 (1.04-1.54), similar to the HR for CVD mortality, 1.27 (1.20-1.35) and cancer mortality, 1.22 (1.16-1.29). There was no association for women, perhaps because women are more likely to be killed in accidents caused by men, so that conscientiousness is a lesser factor in this outcome.
A further issue is the absence of baseline data and raw numbers. The first table that appears with subject characteristics (table 2) is already age-adjusted. Yet we know from enquiries made by my colleagues and I in 2017 that age differences between quintiles in Harvard analyses of the NHS?HPFS data can be very large as an artifact of the cumulative survey method unique to these studies,, for example: "participants in the lowest quintile of SFA intake were much older than those in the highest quintile (mean difference, approximately 15 years), resulting in a strong confounding by age."[2] This kind of information should be supplied in a transparent manner when an epidemiological paper is first published.
[1] Sinha R, Cross AJ, Graubard BI, Leitzmann MF, Schatzkin A. Meat intake and mortality: a prospective study of over half a million people. Arch Intern Med. 2009;169(6):562–571. doi:10.1001/archinternmed.2009.6
[2] Wang DD, Willett WC, Hu FB. Lingering Questions Concerning Specific Dietary Fats and Mortality—Reply. JAMA Intern Med. 2017;177(4):597–598. doi:10.1001/jamainternmed.2017.0043
Competing interests: No competing interests
The Introduction of the paper states:
1. "A large body of evidence has shown that higher red meat consumption, especially processed red meat, is associated with an increased risk of type 2 diabetes, cardiovascular disease, certain types of cancer, including colorectal cancer, and mortality.
2. "The average consumption of red meat in the United States has decreased in recent decades..."
The Conclusion states: "Increases in red meat consumption, especially processed meat, were associated with higher overall mortality rates."
Therefore, decreases in red meat consumption should cause a decrease in the prevalence of these diseases. But we have had a decline in the quantity of red meat consumption, while concurrently witnessing a huge increase in the prevalence of these diseases.
The hazard ratios in this epidemiological study are too low to be of significance, and the author's conclusions are clearly countered by what they state in the introduction. It is my opinion that this paper is of no value.
Competing interests: No competing interests
Response to Yan Zheng
I thank Dr Zheng for providing data on accidental death - I hope this becomes a general practice. While obesity can increase injury, perhaps leading to increased death, protein is the most satiating macronutrient and thus the foods in this analysis are those least likely to lead to weight gain.
I admit I did not read table 2 properly, hence my question about age-adjustment. However, in the dietary fat analysis mentioned, if fat intakes changed across multiple 4-year follow-up cycles over 32 years, sufficiently to skew the age difference from 2.3 to 15 years, it is hard to see what relationship the baseline characteristics of the quintiles had to the characteristics of the people in those same quintiles at the 32 year mark - surely many of them were in fact different people?
A further point relates to the framing of the question. Although there was no benefit from red meat reduction per se, further modelling found a benefit from replacing red meat with nuts. However, nuts are independently associated with benefit.
"The pooled multivariate hazard ratios for death among participants who ate nuts were 0.93, 0.89, 0.87, 0.85 and 0.80 for those who consumed nuts less than once per week, once per week, two to four times per week, five or six times per week, and seven or more times per week, respectively (p < 0.001 for trend)."[1]
So, what happened to people who kept red meat consumption stable but increased intake of nuts? This was also a possibility. Or, who kept red meat stable but increased any other "beneficial" food? The question presumes that replacing red meat will produce benefit one way or another, but given that there was no benefit associated with red meat reduction per se, the reduction in mortality associated with the replacement of other, "discretionary" or ultra-processed, foods may well have been large enough to be indicative of a causal relationship.[2,3]
[1] Y. Bao et al. Association of nut consumption with total and cause-specific mortality. N Engl J Med Vol. 369; 2013 2001 2011 Data from the “Nurses’ Health Study” and the “Health Professionals Follow-up Study”. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4121755/
[2] Collin LJ, Judd S, Safford M, Vaccarino V, Welsh JA. Association of Sugary Beverage Consumption With Mortality Risk in US Adults: A Secondary Analysis of Data From the REGARDS Study. JAMA Netw Open. 2019;2(5):e193121. Published 2019 May 17. doi:10.1001/jamanetworkopen.2019.3121
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6537924/
[3] Rico-Campà A, Martínez-González MA, Alvarez-Alvarez I, et al. Association between consumption of ultra-processed foods and all cause mortality: SUN prospective cohort study. BMJ. 2019;365:l1949. Published 2019 May 29. doi:10.1136/bmj.l1949
https://www.bmj.com/content/365/bmj.l1949
Competing interests: I am an employee of PreKure, an online health coach certification service.