{"id":9747,"date":"2023-10-23T10:45:36","date_gmt":"2023-10-23T09:45:36","guid":{"rendered":"https:\/\/www.zoeharcombe.com\/?p=9747"},"modified":"2023-10-23T18:08:02","modified_gmt":"2023-10-23T17:08:02","slug":"red-meat-type-2-diabetes","status":"publish","type":"post","link":"https:\/\/www.zoeharcombe.com\/2023\/10\/red-meat-type-2-diabetes\/","title":{"rendered":"Red meat & Type 2 Diabetes"},"content":{"rendered":"
<\/p>\n
Executive summary<\/strong><\/p>\n * This week’s study made media headlines worldwide and generated much interest on Twitter.<\/p>\n * It claimed that having examined data from 216,695 people, red meat intake (total, processed and unprocessed) increased the risk of developing Type 2 Diabetes (T2D).<\/p>\n * This makes no sense from the outset, as diabetes is a glucose handling issue and meat contains no glucose. It would be more obvious to suspect the buns, fries and fizzy drinks that hot dogs and burger are consumed with.<\/p>\n * I reviewed the paper and found 14 issues. There may be more:<\/p>\n Issue 1 \u2013 the inaccuracy of Food Frequency Questionnaires (upon which this study was based).<\/p>\n Issue 2 \u2013 the reported intakes were changed (the researchers \u2018calibrated\u2019 the reported intakes, which increased risk ratios.)<\/p>\n Issue 3 \u2013 the definition of red meat included sandwiches and lasagna.<\/p>\n Issue 4 \u2013 the serving sizes have changed since the original Food Frequency Questionnaires.<\/p>\n Issue 5 \u2013 the intakes used to compare people have become more extreme.<\/p>\n Issue 6 \u2013 the study claimed that women consume more red meat than men; that would be a first.<\/p>\n Issue 7 \u2013 total red meat was claimed to have a higher risk than both processed red meat and unprocessed red meat. Total red meat is the sum of the other two. It can\u2019t be worse than both.<\/p>\n Issue 8 \u2013 the healthy person confounder. The red meat eater had a higher BMI and was more likely to smoke and less likely to exercise. We can\u2019t adjust for a completely different person.<\/p>\n Issue 9 \u2013 the reported calorie intake was absurd.<\/p>\n Issue 10 \u2013 the characteristics table reported all food intake except the relevant ones \u2013 sugar and grains.<\/p>\n Issue 11 \u2013 the headline claims did not adjust for the higher BMI.<\/p>\n Issue 12 \u2013 even if there were no issues 1-11, the study could only suggest association not causation.<\/p>\n Issue 13 \u2013 the relative risk numbers grabbed the headlines; the absolute risk differences were a fraction of one per cent.<\/p>\n Issue 14 \u2013 the plausible mechanisms proposed applied far more sensibly to the bun, fries and fizzy drink (which were ignored) than to the burger.<\/p>\n The bottom line can be summed up by surgeon captain Peter Cleave. \u201cFor a modern disease to be related to an old fashioned food is one of the most ludicrous things I have ever heard in my life<\/em>.\u201d<\/p>\n Introduction<\/strong><\/p>\n I lost count of the number of emails I received asking me to look at this week\u2019s topic. On October 19th, 2023, there were headlines across the pond claiming, \u201cEating red meat twice a week may increase type 2 diabetes risk, study finds<\/em>\u201d (Ref 1).<\/p>\n This makes no sense. Diabetes is essentially the inability to handle glucose. Meat contains no glucose. Carbohydrates contain glucose. My immediate thought was \u2013 don\u2019t blame the burger for what the bun, fries and fizzy drink did. It\u2019s also the latest paper from the Harvard epidemiological paper production factory. All their papers promote plants and condemn animal foods. This is just their latest attack on red meat.<\/p>\n The paper generating this week\u2019s headlines was called \u201cRed meat intake and risk of type 2 diabetes in a prospective cohort study of United States females and males<\/em>\u201d (Ref 2). The article was published in the American Journal of Clinical Nutrition. The lead author was Gu. Other authors included Frank Hu and Walter Willett. It used the standard US population studies \u2013 the Nurses\u2019 Health Study (women) and the Health Professionals Follow Up Study (men). It used both the Nurses\u2019 Health Study I and the Nurses\u2019 Health Study II.<\/p>\n I checked back for previous posts on my site on this topic and found one from 2011 (Ref 3). It emanated from the Harvard epidemiological paper production factory. It generated media headlines “Two slices of bacon a day increases diabetes threat by 50<\/em>\u201d (Ref 4). The article was called “Red meat consumption and risk of type 2 diabetes: 3 cohorts of US adults and an updated meta-analysis<\/em>” (Ref 5). The article was published in the American Journal of Clinical Nutrition. The lead author was Pan. Other authors included Frank Hu and Walter Willett. It used the standard US population studies \u2013 the Nurses\u2019 Health Study (women) and the Health Professionals Follow Up Study (men). It used both the Nurses\u2019 Health Study I and the Nurses\u2019 Health Study II.<\/p>\n Spot the similarities; twelve years apart.<\/p>\n<\/p>\n <\/a><\/p>\n \n The study<\/strong><\/p>\n Epidemiological studies, also called population studies, start with a large population of people. They gather lots of information on them at baseline and then follow them for a number of years, continuing to gather information. Researchers review the data to look for patterns. E.g. did the smokers end up with more lung cancer? These patterns, called associations, are then supposed to be tested in randomised controlled trials (RCTs). This doesn\u2019t happen nowadays. RCTs are too expensive and take too long. Researchers just publish associations and they\u2019re happy for people to infer that A causes B. The clear inference from this study, and the media headlines, is \u2018eat red meat and get type 2 diabetes (T2D).\u2019<\/p>\n Let\u2019s go through what was done and reveal the issues along the way.<\/p>\n To capture the facts up front. The Gu et al<\/em> 2023 study used data from 216,695 US people:<\/p>\n \u2013 84,315 women in the Nurses\u2019 Health Study I (NHS I) (started 1980);<\/p>\n \u2013 42,163 men in the Health Professionals Follow Up Study (HPFS) (started 1986);<\/p>\n \u2013 90,217 women in the Nurses\u2019 Health Study II (NHS II) (started 1991).<\/p>\n They were followed up for a total of 5,483,981 person years. This will be useful for calculating absolute risk later on.<\/p>\n The issues<\/strong><\/p>\n Issue 1 \u2013 the inaccuracy of Food Frequency Questionnaires.<\/u><\/p>\n At baseline, participants in the NHS I, the NHS II and the HPFS would have completed a Food Frequency Questionnaire (FFQ). The 1980 FFQ used in the NHS included 61 items. In 1984 and thereafter the FFQ included 120 items. On the FFQs participants were asked to report their average intake of each food or drink over the past 12 months. This is known to be inaccurate to start with (Ref 6).<\/p>\n This reference links to the original 61 item FFQ (Ref 7). You can see the meat questions on P4. Participants were asked \u201cHow often did you eat the following over the past 12 months…?<\/em>\u201d (How accurately could you answer these questions for the past year?)<\/p>\n The amounts were in 9 categories: 6+ per day; 4-6 per day; 2-3 per day; 1 per day; 5-6 per week; 2-4 per week; 1 per week; 1-3 per month; almost never.<\/p>\n The meats were:<\/p>\n i) Chicken without skin (6-8oz);<\/p>\n ii) Chicken with skin (6-8oz);<\/p>\n iii) Hamburgers (1);<\/p>\n iv) Hot dogs (1);<\/p>\n v) Processed meats (sausage, salami, bologna etc) (piece or slice);<\/p>\n vi) Bacon (2 slices);<\/p>\n vii) Beef, pork or lamb as a sandwich or mixed dish (stew, casserole, lasagne etc); *<\/p>\n viii) Beef, pork or lamb as a main dish (steak, roast, ham etc. 6-8oz).<\/p>\n * For (vii), note that no serving size was given and meat sandwiches and meat lasagna (carbs) were considered meat.<\/p>\n Issue 2 \u2013 the reported intakes were \u2018calibrated.\u2019 <\/u><\/p>\n The authors proved that they know that FFQs are inaccurate because they changed the results. They called it calibration: “we calibrated self-reported red meat intake with weighed diet records for the first time<\/em>.”<\/p>\n We need to introduce two other (smaller) population studies at this point \u2013 the Women\u2019s Lifestyle Validation Study (WLVS) and the Men\u2019s Lifestyle Validation Study (MLVS). Participants of WLVS and MLVS were recruited as subsets of NHS, NHS II, and HPFS participants and members of a Boston-area health plan. Among the WLVS and MLVS participants, there were 1,207 people who provided 7-day weighed diet records and FFQs. The weighed diet records are more accurate, albeit over a much shorter period.<\/p>\n Gu et al<\/em> cited two papers as evidence for how to correct errors in FFQs using 7-day diet records (Ref 8). They then adjusted the intakes for all 216,695 participants based on this. The paper reported \u201cStronger associations between red meat intakes and T2D risk were observed after calibrating dietary exposures. Before the calibration, a 1-serving intake increment in total red meat was associated with a 28% higher risk of T2D; after the calibration, this was 47%.\u201d<\/em><\/p>\n This calibration exercise thus increased claimed risks (Ref 9).<\/p>\n Issue 3 \u2013 the definition of red meat included sandwiches and lasagna.<\/u><\/p>\n The above categories of red meat (from the 1980 FFQ) continue to be the definitions of total red meat, processed red meat and unprocessed red meat. The Gu et al<\/em> (2023) paper reported that total red meat intake was the sum of serving intakes of processed and unprocessed red meats. Processed red meats included beef or pork hot dogs; bacon; processed meat sandwiches; and other processed meats such as sausage. Unprocessed red meats included lean or extra lean hamburger; regular hamburger; beef, pork, or lamb as a sandwich or mixed dish; pork as a main dish; and beef or lamb as a main dish. The sandwich\/lasagna (carbohydrate) confounder can thus occur in both processed and unprocessed red meat categories.<\/p>\n Chicken \u2013 called poultry in the paper \u2013 was not included as red meat. Poultry intake was not reviewed as a risk factor in the paper.<\/p>\n Issue 4 \u2013 the serving sizes have changed.<\/u><\/p>\n The serving sizes used were identical in the 2011 and 2023 papers. \u201cOne serving of unprocessed red meat equals 85 g of pork, beef, or lamb; one serving of processed red meat equals 28 g of bacon or 45 g of hot dog, sausage, salami, bologna, or other processed red meats<\/em>.\u201d<\/p>\n The first nurses recruited were asked how often they consumed 2 slices of bacon. The bacon serving size is now 1 slice of bacon (28g\/1 oz). The original 1 hot dog would equate to the 45g of hot dog used today. Did participants realise that a slice of salami meant 45g? The FFQs asked about 6-8oz intake of chicken, beef, pork, lamb etc. A portion is now defined as 85g, which is 3 oz. How has this been adapted? If a 1980 nurse ate 5oz of beef every day but never ate 6-8oz of beef, did she translate her consumption into 6-8oz equivalents?<\/p>\n Issue 5 \u2013 the intakes used have changed.<\/u><\/p>\n The 2011 paper reported risk ratios for a one serving a day increase. The Pan et al<\/em> study reported \u201cThe pooled HRs (95% CIs) for a one serving\/d increase in unprocessed, processed, and total red meat consumption were 1.12 (1.08, 1.16), 1.32 (1.25, 1.40), and 1.14 (1.10, 1.18), respectively<\/em>.\u201d i.e. they claimed that there was a 12% higher risk in developing T2D for 1 serving a day more unprocessed red meat, a 32% higher risk of T2D for 1 extra serving a day of processed red meat and a 14% higher risk of T2D for 1 extra serving a day of all red meat.<\/p>\n The 2023 paper reported intakes differently. This paper put participants into five similar sized groups (called Quintiles). They compared people in the lowest intake group (Q1) with people in the highest intake group (Q5). Women in the NHS I, for example, consumed 0.46 servings a day of all red meat in Q1 and 2.62 servings a day in Q5. Men in the HPFS consumed 0.24 servings a day of all red meat in Q1 and 2.42 servings a day in Q5. We have another issue\u2026<\/p>\n Issue 6 \u2013 men vs women red meat consumption.<\/u><\/p>\n This study claims that women consume more red meat than men in servings per day. (This held for NHS I, but not NHS II). As far as I recall, this would be the first study to assert this.<\/p>\n Comparing the top and bottom groups, the 2023 paper claimed, \u201cComparing the highest to the lowest quintiles, hazard ratios (HR) were 1.62 (95% confidence interval [CI]: 1.53, 1.71) for total red meat, 1.51 (95% CI: 1.44, 1.58) for processed red meat, and 1.40 (95% CI: 1.33, 1.47) for unprocessed red meat<\/em>.\u201d Now 62%, 51% and 40% higher risks for developing T2D are being claimed for total red meat, processed red meat and unprocessed red meat. We have another issue\u2026<\/p>\n Issue 7 \u2013 processed red meat vs unprocessed red meat.<\/u><\/p>\n In other studies I have reviewed, (including the 2011 one) processed red meat has the highest claimed risk; unprocessed red meat has the lowest risk and total red meat is somewhere in between. This study claims that total red meat has a higher risk than both processed and unprocessed red meat. Total red meat is the sum of the other two. It can\u2019t be worse than both (Ref 10).<\/p>\n I found a fundamental error in the 2011 paper (see the original post), which was acknowledged and corrected by the Harvard experts. Is there another fundamental error in this paper?<\/p>\n Issue 8 \u2013 the healthy person confounder.<\/u><\/p>\n The characteristics table is the place to start in any population paper. This is usually Table 1 in the paper. It tells us the baseline characteristics of people in the study. The characteristics table is categorised by the subject of interest \u2013 in this study, that\u2019s red meat intake. Hence the 216,695 people in this study were split into columns by red meat intake.<\/p>\n Table 1 retained the original data from each population study so the 84,315 women from NHS I were separated from the 90,217 women from NHS II and the 42,163 men from HPFS. Each of these three groups was split into a further three groups \u2013 the lowest intake of red meat (Q1), the middle intake of red meat (Q3) and the top intake of red meat (Q5). The top and bottom groups were compared with each other, as we saw above.<\/p>\n When you look at the characteristics table, you can immediately see differences in people in the top and bottom groups. The red meat intake is never<\/em> the only difference. In this table, the highest red meat consumers had higher BMIs, they were less physically active, they were more likely to be current smokers and they were less likely to take multivitamins. There is always<\/em> a healthy person confounder. The burger\/hot dog consumer is less healthy than the quinoa\/kumquat consumer in many ways \u2013 not just red meat.<\/p>\n Issue 9 \u2013 the reported calorie intake.<\/u><\/p>\n The striking thing about this characteristics table was the energy intake. One look at Table 1 and a peer-reviewer should have rejected the paper. The average (mean) total calories were as follows:<\/p>\n <\/p>\n We are being asked to believe that (take NHS I first), women in Q1 (the lowest red meat intake group) averaged just 1,202 calories a day. The 2,032 calories a day for Q5 seems far more accurate for nurses (on their feet all day) in 1980. Take the HPFS, do we really think that male health professionals were averaging 1,684 calories a day in 1986? I suggest that people in the bottom groups did not complete the questionnaire properly.<\/p>\n (The study excluded women with an energy intake below 500 or above 3,500 calories and men with an energy intake below 800 or above 4,200 calories. It is common to exclude extremes as being clearly unreliable. However, the bottom group \u2013 and the middle group \u2013 still had calorie intakes that were literally unbelievable.)<\/p>\n Calorie intake was then adjusted for, but it was so obviously wrong to start with, any adjustment would have been wrong. The paper should be ignored for this alone.<\/p>\n<\/p>\n