Red meat is in the news again.
Last week, the Annals of Internal Medicine published 5 new reviews from the Nutritional Recommendations (NutriRECS) international consortium looking at the impact of red and processed meat consumption on many health outcomes — cancer mortality, cancer risk, diabetes, heart disease, and more.
What caused a stir are the specific recommendations made by this group. In their dietary guideline recommendations the panel suggested “ ... that adults continue current unprocessed red meat consumption (weak recommendation, low-certainty evidence). Similarly, the panel suggests adults continue current processed meat consumption (weak recommendation, low-certainty evidence)”.
These reviews analyzed dozens of studies covering millions of participants. The results, including an editorial and dietary guideline recommendations, span over 300 pages of content! With so much information to process, and the ensuing media frenzy that surrounded these papers, you may have been left a bit confused as to what to think about the health impacts of red meat consumption. Luckily, we’re no strangers to this complex and divisive topic.
We’ll walk you through this latest series of studies and break down what practical implications they may have.
These studies do not introduce new evidence about red meat. Rather, they are summarizing existing studies (of which there are a lot).
A quick note before digging into the studies. When referring to ‘red meat’, we mean items like beef, veal, pork, and lamb. The term ‘processed meat’ refers to meats that have been processed via methods such as salting, curing, or smoking.
Listed below is what each of the five studies looked at.
A systematic review of randomized trials looking at the effects of higher vs lower red meat intake on the incidence of cardiometabolic and cancer outcomes.
A systematic review and meta-analysis of observational cohort studies looking at the association of dietary patterns consisting of different red and processed meat intakes on all-cause mortality, cardiometabolic outcomes, and cancer incidence and mortality.
A systematic review and meta-analysis of observational cohort studies looking at the association of red and processed meat intakes on cardiometabolic outcomes.
A systematic review and meta-analysis of observational cohort studies looking at the association of red and processed meat intakes on cancer mortality and incidence.
A mixed-methods systematic review looking at values and preferences regarding meat consumption.
In this discussion, we’ll be focusing on what was found in the first four of these studies.
Importantly, the central goal of these studies was to focus on a narrow question: the effect of red meat and processed meat on human health. Hence, the following were intentionally not considered.
Environmental impacts of red meat consumption
Ethical considerations of red meat consumption
Animal and mechanistic data (e.g., cell studies) that may help explain the effects of red meat on health
This doesn’t mean these points aren’t important, it just means one study can’t possibly cover all aspects of this question.
One additional note. While these studies did not receive industry funding, one of the authors disclosed that his other research had received funding from AgriLife Research, which does have financial ties to the beef industry.
Five new studies summarized the existing evidence on red and processed meat on health. They looked at RCTs and observational studies investigating how red meat consumption patterns affected all-cause mortality, cardiovascular disease incidence and mortality, and cancer incidence and mortality. They did not look at evidence of potential environmental impacts, ethical considerations, or non-human studies.
Before we dive into the findings in the below tables, there are two concepts we should cover first — confidence intervals and GRADE scores. Understanding both will help you better interpret the study results.
How sure can we be about the results of these studies? That’s what confidence intervals are all about.
In the results tables, we will see below, the average risk difference per 1000 people with a 95% CI (confidence interval) when lower amounts of red meat are consumed is reported for each outcome. For example, let’s say lower red meat consumption leads to a decreased risk in type 2 diabetes of −10 fewer cases on average with a 95% CI of −15 to −5. A 95% CI indicates that in the long-run, if studies were repeated several times, 95% of the confidence intervals would include the true value of the effect. But we are unsure whether the range given by the confidence interval reported (−15 to −5) is one of the intervals that includes the true value. In fact, it may be the few intervals (5%) that do not contain the true effect.
The CI can inform you about the degree of uncertainty in the result. In our above example, there is a bit of uncertainty. Yet if the results came back as −10 with a 95% CI of −90 to +70 it would indicate a huge amount of uncertainty. Keep this in mind when reading the results.
Confidence intervals (almost always set at 95%) are a mathematical way of saying "if we repeat this study over and over, 95% of the intervals will contain the true value, but we don't know if this interval is one of the 95%". A narrow Cl indicates a greater degree of certainty in the result (say a −15 to −5 interval) whereas a large CI indicates less certainty (a −90 to +70 interval, for example).
The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) tool is used for evaluating research quality and the strength of recommendations that stem from this research. The tables below show the GRADE certainty of the evidence score, which has four possible ratings: high, moderate, low, and very low.
This score tells you the likelihood that the effect seen in studies will be substantially different from the real world. Let’s take the results of our above fictional example — 10 fewer type 2 diabetes cases with lower red meat consumption. If the rating was ‘high’, the real-world effect is likely to be close to what the research shows. If the rating was ‘very low’, the effects seen in the real world may be substantially different than our findings.
A GRADE certainty of evidence score tells you the likelihood that the effect seen in studies will be very similar or substantially different from the real world. Possible scores are high, moderate, low, and very low.
The first study examined randomized controlled trials (RCTs) available on the topic. Though 12 trials were initially found, only one was determined to meet the authors inclusion criteria (these criteria have been critisized as too stringent by some) — the Women's Health Initiative study which enrolled 48,000 postmenopausal women. Because only one trial was included, a proper meta-analysis could not be conducted. Thus, the results in the table below are just the effects reported in this single trial.
The second study reported on overall dietary patterns of red and processed meat and their association with cardiometabolic and cancer outcomes. The researchers looked at prospective cohort studies with 1000 or more participants and included trials that lasted anywhere from 2 to 34 years. Overall, the researchers identified 105 eligible studies with 6,035,000 total participants.
The third study reported on overall consumption of red and processed meat and its association with all-cause mortality and cardiometabolic outcomes. Importantly, these trials allowed the researchers to separately analyze the effects of unprocessed and processed red meat intake. Overall, the researchers identified 62 eligible studies with 4,200,000 total participants.
The fourth study reported on the reduction of red and processed meat and the association with cancer mortality and incidence. These trials also allowed the researchers to separately analyze the effects of unprocessed and processed red meat intake. Overall, the researchers identified 118 eligible studies with 6,100,000 total participants.
From these studies, the NutriRECS group concluded the following.
“The panel suggests that adults continue current unprocessed red meat consumption (weak recommendation, low-certainty evidence). Similarly, the panel suggests adults continue current processed meat consumption (weak recommendation, low-certainty evidence)”.
The group further explained that their recommendations indicate that they “believed that for the majority of individuals, the desirable effects (a potential lowered risk for cancer and cardiometabolic outcomes) associated with reducing meat consumption probably do not outweigh the undesirable effects (impact on quality of life, burden of modifying cultural and personal meal preparation and eating habits). The weak recommendation reflects the panel's awareness that values and preferences differ widely, and that as a result, a minority of … (individuals fully informed about the risks would) choose to reduce meat consumption”.
Importantly, the group added that, while their recommendations differ from that of other international guidelines, it could be argued that “ ... neither do they seriously challenge those recommendations”.
Another key aspect was that the group did not state that people should increase their red and processed meat intake, nor did they state that it didn’t have the potential to cause some amount of harm. Instead, they are highlighting uncertainty in the evidence about the “causal relationships between red meat consumption and major cardiometabolic and cancer outcomes”.
There is some evidence, of lower certainty, that high red and processed meat consumption can increase the risk of various cardiometabolic diseases and cancers. However, there is no rock-solid evidence that they will cause harm. If you wish to take a more cautious approach, then cut back on your intake to 3 servings per week or fewer.
Red meat consumption is just a small part of the overall lifestyle factors that will influence your health. Don't get so caught up in the details that you end up missing the bigger picture that overall eating pattern and physical activity levels will play.
The current round of red meat studies and the ensuing disagreements that have played out between various health organizations is not so much about the actual research, but about how that research should be interpreted and applied in terms of public health messaging. This is a debate about many topics in nutrition, not just red meat.
Beneath the discussion of whether red meat is healthy is a heated debate between methodologists about how reliable certain research designs are for inferring cause and effect. The vast majority of nutrition research includes observational studies such as cohort and case-control studies. Studies of this sort tend to be cheaper than clinical trials, and they attempt to investigate phenomena that are usually difficult to study experimentally. But they are also vulnerable to the effects of confounding, making it more difficult to infer a cause and effect relationship.
Randomized controlled trials deal with these problems via random assignment, making the distribution of all potential confounding variables, random. This reduces systematic variation between the groups being studied and makes them comparable, allowing trialists to infer a cause and effect relationship about the dietary exposure that’s being manipulated.
Unfortunately, randomized controlled trials are also very expensive, tend to be short, and are not possible to conduct when answering certain questions. For example, how plausible would it be to randomize participants to a group and tell them to eat meat for 10 years, while telling the other group to completely avoid eating meat for 10 years? It’s not very likely, which is why observational studies tend to dominate the nutrition field.
Researchers who conduct observational studies try to make the groups they are studying as comparable as possible with several statistical and graphical methods. In fact, there has been an explosion in research within this area in the past 50 years, where sophisticated statistical methods have been developed to strengthen the ability of observational studies to conclude cause and effect relationships.
However, for most observational studies, it is impossible for typical methods to capture and adjust for all potential confounding variables, especially in nutrition, where dietary exposures are connected to several variables. Thus, there is always some “residual confounding”.
Nutrition science presently relies a lot on observational studies, which can make it difficult when attempting to establish a cause and effect relationship. While RCTs would be preferred, they can be prohibitively expensive and, in some cases, are not feasible to conduct for certain research questions. This is the catch-22 of nutrition research that can lead to much debate.
As a result, most researchers who critically appraise evidence consider RCTs to be the most reliable form of evidence and the tools they use to assess the evidence, also reflect this.
For example, the GRADE tool automatically gives randomized controlled trials a “high-quality rating” assuming no serious biases and the vast majority of observational studies are given a “low-quality rating” from the beginning. The exception to this is when the observational studies produce large associations and/or dose-response relationships. However, most effects in nutrition are likely to be small, making it next-to-impossible for the average observational study to receive a high-quality rating.
So even if an observational study was well done, had high quality measurements, and sophisticated design/analytical methods to strengthen any claims to causality, it would still be given a low rating on the GRADE scale.
Many nutrition researchers believe that applying GRADE to nutrition studies in this manner is inappropriate, since GRADE has primarily been used for simple interventions (such as pharmaceuticals and medical devices), given that it is rarely feasible to conduct a randomized controlled trial, and because GRADE automatically penalizes studies for being observational. Some researchers have proposed other rating systems to use for assessing the overall evidence in nutrition, that do not automatically penalize research for being observational. However, these methods have not been adopted and are still fairly new. Many in the meta-research community believe that all research, including nutrition research, should be held to the same standard, and therefore, GRADE is an appropriate way to assess the evidence. These sentiments were echoed in a recent interview by Dr. Gordon Guyatt, one of the red meat study authors and developer of the GRADE guidelines. “Why have one set of rules for judging nutrition and another set of rules for some other area?” he asked.
Let’s say lower red meat consumption led to 10 fewer type 2 diabetes cases for every 1000 people who reduced their consumption. On the population level, assuming everyone reduces their red and processed meat intake, a 1% reduction in diabetes cases may translate into tens of thousands fewer T2D cases. On an individual level, we can’t say if you will be one of those 10 cases out of 1000.
This is an issue not new to nutrition research. At the heart of this and many other nutrition debates is this question: How do we obtain practical takeaways about the consumption of specific foods when all we have are lower-quality studies that give us unsure data?
Could we do higher-quality clinical trials examining the question of red meat on health? Absolutely, but the cost of such studies would likely be high and even then, they could only tell us so much. Rigorous and long-term trials tend to rack up quite an expensive bill and funding opportunities for nutrition research are highly competitive and can be somewhat limited.
Rather than using only randomized controlled trials to investigate dietary exposures, it seems sensible to combine evidence from several lines to inform policy, as it’s been done in the past. The consistency of the results between these different lines could strengthen any claims to causality. This is what has traditionally informed the messaging of various dietary guidelines.
Unfortunately, much of nutrition research has been inconsistent. Many observational studies conflict with the results of randomized trials, but also with other observational studies. Some meta-researchers have made this widely known by showing how sensitive nutrition results are to small changes in the analysis plan and by meta-analyzing the results of several observational studies and showing how these results shrink.
Although meta-analyses are often seen as one of the highest forms of evidence, they can often be inappropriate. For example, in the meta-analyses recently published, the researchers pooled nearly a hundred observational studies to compute a summary effect. However, when the individual cohort studies are vastly different in their designs, the exposures, the characteristics of participants, and in the statistical methods used, such summary effects are practically uninterpretable. In fact, many of the results showed notable amounts of heterogeneity, which is the variance between study results as a result of systematic reasons, rather than random error. Rather than average the effects of studies that are vastly different, it may be more informative to qualitatively assess them and describe the results.
Much of observational research relies on memory-based methods such as food frequency questionnaires and 24-hour recalls to assess food intake by participants. These methods have been heavily criticized since they have often been shown to be unreliable and prone to serious systematic errors, particularly when trying to ascertain consumption patterns of single foods such as red meat or eggs. Several methodologists have proposed using statistical methods and supplementary measures, such as biomarkers to improve measurement in observational studies, and other methods such as food records.
In an interview with Examine.com, nutrition researcher Dr. Christopher Gardner of Stanford University elaborated that, “yes, we can make nutritional epidemiology more reliable” and offered the following examples of strategies being used to move the field forward.
Adding photographs with cell phones of the food eaten
Apps created to more easily log eating episodes in the moment, more accurately
Statistical analysis to help show which questions about foods are the most effective at capturing eating patterns so that the number of questions can be decreased and the time burden can be lowered
Confirming biomarkers, such as omega-3’s in red blood cells, sodium in urine, fiber derivatives in stool samples
However, these have yet to be widely adopted and some require more validation.
Nutrition methodology has room to improve, but nutrition epidemiology still has a place when RCTs can’t be done. In the absence of RCTs, we have to make decisions on a personal level based on costs and benefits of including these things in our diet as individuals, while on a population level, it’s hard to ascertain how much such guidelines would impact public health, especially due to a lack of high-quality evidence.
Even if we knew exactly how much red meat everyone consumed, it still may not be enough information to allow us to provide concrete recommendations about its consumption and health effects. A host of other factors — foods consumed instead of red meat, genetics, exercise routines — are all pieces of the overall health puzzle.
Newer guidelines have moved away from setting specific nutrient recommendations (e.g., specific daily limits for red meat, saturated fat, salt, etc.) to focusing more on the overall dietary pattern (e.g., consume more vegetables and fruit, fewer sugar-sweetened foods and beverages). This has the advantage of more closely aligning with how people actually go about consuming food.
- Can creatine cause cancer?
- Are nitrates from beetroot and processed meats the same thing?
- Scientists found that red meat causes cancer ... or did they?
- Does red meat cause cancer?
- Is processed meat bad for me?
- Do muscle building supplements cause testicular cancer?
- How can I make red meat healthier?
- Do MCTs or CLA help with appetite reduction?
- Can eating too much protein be bad for you?
- What beneficial compounds are primarily found in animal products?
- Put down the apple and have some cheddar
- Red and Processed Meat Consumption and Risk for All-Cause Mortality and Cardiometabolic Outcomes: A Systematic Review and Meta-analysis of Cohort Studies. Ann Intern Med. (2019) Zeraatkar D, et al.
- Reduction of Red and Processed Meat Intake and Cancer Mortality and Incidence: A Systematic Review and Meta-analysis of Cohort Studies. Ann Intern Med. (2019) Han MA, et al.
- Patterns of Red and Processed Meat Consumption and Risk for Cardiometabolic and Cancer Outcomes: A Systematic Review and Meta-analysis of Cohort Studies. Ann Intern Med. (2019) Vernooij RWM, et al.
- Health-Related Values and Preferences Regarding Meat Consumption: A Mixed-Methods Systematic Review. Ann Intern Med. (2019) Valli C, et al.
- Effect of Lower Versus Higher Red Meat Intake on Cardiometabolic and Cancer Outcomes: A Systematic Review of Randomized Trials. Ann Intern Med. (2019) Zeraatkar D, et al.
- Unprocessed Red Meat and Processed Meat Consumption: Dietary Guideline Recommendations From the Nutritional Recommendations (NutriRECS) Consortium. Ann Intern Med. (2019) Johnston BC, et al.
- Meat Consumption and Health: Food for Thought. Ann Intern Med. (2019) Carroll AE, Doherty TS.
- Evidence based medicine: what it is and what it isn't. BMJ. (1996) Sackett DL, et al.
- Hierarchies of evidence applied to lifestyle Medicine (HEALM): introduction of a strength-of-evidence approach based on a methodological systematic review. BMC Med Res Methodol. (2019) Katz DL, et al.
- Assessment of vibration of effects due to model specification can demonstrate the instability of observational associations. J Clin Epidemiol. (2015) Patel CJ, Burford B, Ioannidis JP.
- Is everything we eat associated with cancer? A systematic cookbook review. Am J Clin Nutr. (2013) Schoenfeld JD, Ioannidis JP.
- Validity of the 24-hour dietary recall. J Am Diet Assoc. (1985) Karvetti RL, Knuts LR.
- A toolkit for measurement error correction, with a focus on nutritional epidemiology. Stat Med. (2014) Keogh RH, White IR.
- Using surrogate biomarkers to improve measurement error models in nutritional epidemiology. Stat Med. (2013) Keogh RH, White IR, Rodwell SA.