“White meat as bad as red meat for cholesterol”

Introduction

I spotted this week’s story on twitter. A couple of plant-diet advocates were tweeting that white meat is as bad as red meat for cholesterol. The NBC headline was “When it comes to cholesterol, is chicken as bad as red meat?” (Ref 1). The story was covered more in the US than elsewhere, but then it was a US study.

The news coverage was generated by an article in the American Journal of Clinical Nutrition (AJCN). It was called “Effects of red meat, white meat, and nonmeat protein sources on atherogenic lipoprotein measures in the context of low compared with high saturated fat intake: a randomized controlled trial” and it’s on open view (Ref 2). The senior author was Professor Ronald Krauss.

When I had the privilege of being on one of the SwissRe/BMJ panels in Zurich in June 2018, we were asked to open with a couple of minutes about evidence that we trusted. I shared that randomised controlled trials (and meta-analyses thereof) are the gold standard for evidence, but that these are impossible in the field of nutrition. This is because it is impossible to change one thing alone. The trial with the least change would be one involving different oils. In an ‘oil-only’ trial, macronutrients (fat, carb and protein) could be kept identical in the intervention and control. However, even changing one oil for another would change all three fats (saturated, monounsaturated and polyunsaturated) and fat-soluble vitamins (there are no water-soluble vitamins or minerals of any consequence in oils) and thus how would we know what made any difference?

This week’s note will show why randomised controlled trials (RCT) in nutrition are fundamentally flawed. It will also show that this study had an additional flaw, which should have been obvious from the outset. Finally, it will show that – as usual – the headlines and study conclusions were misleading. Neither red, nor white, meat had any effect on cholesterol in this study. Non-meat did, but there’s an easy explanation for this…


The study

The objective of this study was to examine the effect of red meat, white meat and non-meat sources of protein on cholesterol and lipoproteins. The protein intakes were intended to be the same (although much higher than baseline), so that protein itself wasn’t a confounder. As a further dimension, ‘high’ vs. low saturated fat intake was also tested (Note 3). The trial has an anacronym: APPROACH – this stands for Animal and Plant PROtein and Cardiovascular Health.

The dietary interventions were undertaken on generally healthy men and women, aged 21-65 years, with BMIs in the range 20-35 (i.e. from normal weight to obese). The participants were randomly assigned at the outset to either high saturated fat (62 people) or low saturated fat (SFA) (51 people) (Note 4).

Within the high or low SFA arms, participants were allocated to the red meat, white meat or non-meat diets in random order. Each diet was undertaken for 4 weeks at a time. There was a 2-7 week ‘wash-out’ period in-between each diet period, so that effects from the previous diet didn’t carry forward. So, this was a crossover trial, within a randomised controlled trial. This would have been necessary because the numbers of people may have been too small otherwise to achieve statistically significant results. It should be noted that the design and execution of the trial was excellent, which makes the (plant sterol) flaw, which we’ll explain below, even more surprising.

The main outcomes of interest were what we call surrogate end points – markers, rather than real health outcomes. The markers for this study were chosen as LDL-Cholesterol, apolipoprotein B (apoB) (Note 5), the number of small and medium LDL-Cholesterol particles added together (don’t get me started!) and the ratio of total to HDL-Cholesterol. Personally, I care about none of these.

The diets

Table 1 in the paper reported the macronutrient and fat composition of the different diets. I have extracted the relevant numbers below:

 

(*) Note – the individual fats don’t add up to the total fat number. This is often the case with nutritional information; it’s not an exact science. (MUFA = monounsaturated fat and PUFA = polyunsaturated fat).

You can see from the table above that protein was substantially increased from the baseline level of 14% to 23-26% for the trial diets. You can also see that carbohydrate varied from 39% to 46% between diets and fat accordingly varied from 31% to 35%. All three macronutrients changed, therefore. (The high SFA diet macronutrient composition changed the least.)

The supplemental file contained sample menus for all six variants: red meat, white meat and non-meat for the high SFA group and red meat, white meat and non-meat for the low SFA group. (I’ve shared this menu table as Appendix 1 at the end of this note). The main differences between the high SFA and low SFA diets were achieved by manipulating dairy products e.g. the high SFA diets contained cheese and whole or semi-skimmed milk and the low SFA diets contained non-fat cottage cheese and skimmed milk. However, these were by no means the only changes.

The meat/non-meat foods given to the participants were: mostly beef, with a bit of pork, for the red meat eaters; chicken and turkey for the white meat eaters; and legumes, nuts, grains and isoflavone free soy products for the non-meat eaters. The isoflavone free bit means that soy products without phytoestrogens (plant oestrogens) were chosen. The fundamental flaw in this study is that the non-meat foods contain plant sterols (plant cholesterol) (Ref 6). Plant sterols are known to lower blood cholesterol levels (although lowering human cholesterol with plant sterols makes actual outcomes – heart disease and cancer – worse) (Ref 7). (I don’t know why the researchers didn’t think about the plant sterol confounder. Or maybe they did and knew they were about to embark on a study where non-meat would be guaranteed to lower cholesterol more than diets without added plant sterols).

The inclusion of foods that contain plant sterols/lower cholesterol was not limited to the red/white meat diets vs the non-meat diet. This confounder was replicated in the high vs low SFA comparison diets too. Appendix 1 has the sample menus included in the supplemental material for the study. As an example, the high SFA/red meat breakfast cereal was All Bran. The low SFA/red meat cereal was oatmeal, which is known to have a significant cholesterol lowering effect. The grain for the high SFA/red meat dinner was brown rice. The grain for the low SFA/red meat dinner was barley, which is known to have a significant cholesterol lowering effect.

The high vs low SFA diet had another built-in bias in that the swap was achieved by increasing monounsaturated fat to replace SFA. Nuts and oils were among the foods used to increase the monounsaturated fat content of the low SFA diets and both these contain plant sterols and thus both lower cholesterol levels (Ref 8).


The two critical flaws with this trial, therefore, are:

1) That multiples of things changed: macronutrients; micronutrient; all three fat; vitamins; minerals; and all foods assigned to each diet. That’s an issue with any nutritional RCT.

2) That foods that lower blood cholesterol levels have been included and this fundamentally undermined any results. That’s an issue with this RCT particularly.

The results

Just before we come to the results, it’s often useful to share non-results. There were no significant effects of dietary protein source on blood pressure, fasting glucose, or endothelial activity. There were no significant effects on triglycerides. There was a significant fall in HDL-C, if you care about that measure.

Time for another table – this time to extract the cholesterol and lipoprotein outcome measures:

 

(*) Note – Total cholesterol was not a primary outcome of interest, but it is worth noting, nonetheless.

(**) Note – Of the four primary outcome measures of interest – LDL-Cholesterol, apolipoprotein B, the number of small and medium LDL-Cholesterol particles added together, and the ratio of total to HDL-Cholesterol – the only two that achieved any (statistically) significant difference were LDL-Cholesterol and apoB (highlighted in grey). The following is critical to note:

LDL-Cholesterol and apoB were (statistically) significantly higher in the red meat group compared to the non-meat group and in the white meat group compared to the non-meat group. You can see from Table 2 that LDL-Cholesterol and apoB were lower in every one of the six diets when compared to baseline. Almost doubling protein intake and giving people red or white meat twice a day, did not increase total or LDL-Cholesterol or apoB. All of these fell in all diets.

The headline should therefore have been “Study shows that almost doubling protein intake and having red or white meat twice a day makes no difference to cholesterol.”

Summary

* A study published on 4th June in the American Journal of Clinical Nutrition generated media headlines that ‘white meat is as bad as red meat for cholesterol.’

* The study was a randomised controlled trial, which had first randomly allocated 62 people to a ‘high’ saturated fat group and 51 people to a low saturated fat group. Within each group, people undertook a red meat diet for 4 weeks and a white meat diet for 4 weeks and a non-meat diet for 4 weeks. All participants tried all three diets in what is known as a crossover trial. This is a good design for a small study.

* The study concluded: “The findings are in keeping with recommendations promoting diets with a high proportion of plant-based food but, based on lipid and lipoprotein effects, do not provide evidence for choosing white over red meat.” The media reported this as “When it comes to cholesterol, is chicken as bad as red meat?”

* The fundamental flaw with all nutritional interventions is that it is impossible to change one thing alone. In this study, all macronutrients (carb, fat and protein) changed; all micronutrients (vitamins and minerals) will have changed; and all fats (saturated, monounsaturated and polyunsaturated) changed (polyunsaturated fat changed little).

* There was an even more fundamental flaw with this study in that the outcome of interest was cholesterol and non-meat protein sources contain plant sterols, which are known to lower cholesterol (although this has been shown to worsen heart disease/cancer). Meat protein sources contain no plant sterols, so this study was always going to create a difference between meat (red or white) and non-meat.

* The most interesting outcome of the study was that every intervention diet lowered cholesterol (Total Cholesterol and LDL-Cholesterol and HDL-Cholesterol). All interventions almost doubled protein intake from 14% at baseline to 23-26%. Whether in the ‘high’ or low saturated fat group, participants consumed red meat twice a day, or white meat twice a day, or non-meat twice a day. Total, LDL and HDL-Cholesterol fell in all six variants of the diet. (The falls were not statistically significant for red or white meat).

* Because of plant sterols, cholesterol measures in the non-meat diets fell more than in the red or white meat diets. The paper thus judged red and white meat relative to non-meat. Had the paper reported the actual results for red and white meat, there would have been a very different headline.

* The headline should have been: “Study shows that almost doubling protein intake and having red or white meat twice a day makes no difference to cholesterol.”

Appendix 1
Supplemental Table 2: Sample menus for study diets, 10460 kJ

 

High Saturated fat diets

Meal Red meat White meat Non-Meat
Breakfast All Bran Cereal
Egg and cheddar cheese
Low-fat milk
wheat bread
Fresh apple
Coffee
Egg sub/cheddar cheese/oil
Wheat bread
Whole milk
Fresh apple with skin
Coffee
Silver dollar pancakes
Syrup
Egg
Smart bacon
Low-fat milk
Banana
Coffee
Snack Baby carrots, raw
Celery, raw
Cottage cheese, low-fat
Cottage cheese, low-fat Macadamia nuts
Cottage cheese, full fat
Lunch Spinach soup
Meatloaf & Mushroom gravy
Mashed potatoes
Broccoli and cheese soup
Teriyaki chicken thighs
Green beans
Brown rice
Soy split pea soup
Rye crispbread crackers
Seitan fusion sauté
Brown rice
Swiss cheese
Low-fat milk
Dinner Beef stew
Brown rice
Tomato and cucumber salad
Whole milk
Tomato soup
Chicken Thigh Cacciatore
Boiled potato with butter
Spinach salad with cucumbers and red pepper
Vegetarian chili
Couscous
Mori-Nu silken lite firm tofu
Whole milk
 

Low Saturated fat diets

Breakfast Quaker Instant Oatmeal Cereal
Egg
Skim milk
Wheat bread
Coffee
Egg
Wheat bread
Fresh pear
Skim milk
Coffee
Silver dollar pancakes
Syrup
Egg
Smart bacon
Skim milk
Avocado & salsa
Coffee
Snack Raisins
Nonfat cottage cheese
Nonfat cottage cheese Nonfat cottage cheese
Lunch Meatloaf and mushroom gravy
Orzo
Skim milk
Teriyaki chicken thighs
Brown rice/oil
Macadamia nuts
Cooked carrots
Skim milk
Soy split pea soup
Saltine crackers
Seitan fusion saute
Rice
Skim milk
Dinner Beef stew
Barley
Tomato, bell pepper and avocado salad
Macadamia nuts
Skim milk
Chicken cacciatore
Baked potato
Spinach salad with cucumbers, avocado, red pepper
Green beans
Skim milk
Vegetarian chili
Couscous
Mori-Nu silken lite firm tofu
Celery sticks & almond butter
Skim milk

 

References

Ref 1: https://www.nbcnews.com/health/health-news
Ref 2: https://academic.oup.com/ajcn/advance-article/doi/10.1093/ajcn/nqz035/5494812
Note 3: It is interesting to note that the so-called “high” saturated fat wasn’t a diet high in saturated fat. This baseline intake of SFA was 13%. The three ‘high’ SFA diets contained 13-14% SFA – essentially the same as that in the baseline diet. This should have been called baseline SFA, therefore. I’ll drop the inverted commas from this point on in the note.
Note 4: The abstract is wrong – it says 61 people were in the high SFA group and 52 in the low SFA group. I’ve let the lead author know and he’s thanked me for spotting this.
Note 5: This is another lipid marker sometimes tested for. It tends to reflect LDL-Cholesterol levels. Some people think this may be a better marker than LDL-C. I can’t help thinking this is just another marker to start demonising, to keep the cholesterol hypothesis alive.
Ref 6: https://lpi.oregonstate.edu/mic/dietary-factors/phytochemicals/phytosterols
Ref 7: https://thescipub.com/abstract/10.3844/ojbsci.2014.167.169
Ref 8: https://nutritiondata.self.com/facts/fats-and-oils/509/2
https://nutritiondata.self.com/facts/nut-and-seed-products/3123/2

 

4 thoughts on ““White meat as bad as red meat for cholesterol”

  • avatar
    June 13, 2019 at 8:42 am
    Permalink

    Hi Zoe
    What impresses me most about this study is the miniscule difference that varying protein and fat intakes had on the lipid profiles. I am a GP and to me these lipid numbers all look normal and completely uninteresting. LDL varying between 2.22 and 2.62? Maybe statistical significance but these numbers are all of the same magnitude; I see no clinical significance.
    My interpretation: in diets characterised by at least 40% energy as carbs (ie 200g of carbs in a 2000kCal diet), doubling protein and varying saturated fat intake between 10 and 13% made SFA (and I don’t mean saturated fatty acids) difference to the lipids.
    What that tells us: if you are eating lots of carbs it really doesn’t matter much how much protein and fat you eat.
    Which we all knew anyway. I’m sure Ron Krauss knew this too – he is after all the coauthor of the meta-analyses of prospective cohort studies that show no association between saturated fat and heart disease.
    Which makes me wonder why these authors bothered.

    PS – the sample diets look vile. All Bran cereal? Seitan saute? “Smart Bacon”? Skim milk? Pass the emoji for “spew.”

    Reply
  • avatar
    June 12, 2019 at 11:25 am
    Permalink

    Did you see this horrific headline in the online Guardian this morning?
    https://www.theguardian.com/environment/2019/jun/12/most-meat-in-2040-will-not-come-from-slaughtered-animals-report

    I shudder to think how fat, depressed, anxious and generally unwell “passionate meat eaters” (the Guardian’s words) will be in 2040 if they are eating soya beans and sunflower oil coloured with beetroot and thinking that because it tastes more or less the same as meat it will have the same nutritional effect.

    Reply
    • avatar
      June 12, 2019 at 1:53 pm
      Permalink

      Hi Jane
      No I missed that – many thanks for it. Must admit, I’ve largely ignored The Guardian for some time – since it has gone anti-cholesterol, anti-fat, anti-meat, anti-all sorts. Probably well-intentioned, but completely misguided. Attack meat, promote fake meat – big biz will be happy!
      Best wishes – Zoe

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.