Calorie Restriction with Optimum Nutrition
Home • Forum • What's New • What to Eat • Short/Long-Term Benefits of CR • Books • Links, References & Resources
Supplements • 2000 Poll: Graphics • Poll: Statistics • Theories of Aging • CR Mimetics • Science and Research • CRONies

The Albatross -- Part Two: You Want More Protein

Go To >> Introduction | Part One: More Fat | Part Two: More Protein | Miscellaneous Observations | References

Let's start with basic, physiological need: you need protein in your diet, or you die. The same cannot, evidently, be said of CHO. But how much protein?

Well, if your goal is to maintain LBM [lean body mass] as much as possible (which is not just an ego thing, or even a quality of life issue (old folks who can't open the jam jar or get into shopping malls), but is a real survival issue, affecting your odds of falling and breaking a hip, saving yourself from a fall, and dying because you didn't CLOSE the jam jar and got some evil thing in the stuff), it would appear tat the answer is, quite a bit. I suspect we've all seen the fruits of Sherm's ingenuity by now, but just in case, and for posterity :) :

This data comes from Modern Nutrition in Health and Disease, 8th ed., 1994, Ch. 1, pp. 24-29. The faulty reasoning below is my own.[Fortunately, there isn't any :) -MR].

the book says:

Nitrogen balance is also affected by energy intake. ... N balance becomes progressively more negative as energy intake is reduced below the needs of the body. ... A direct relationship exists between energy intake and N balance from negative at low-energy levels to positive at excessive intakes of energy.

A table then shows measured requirements varying by a factor of two over a calorie intake range from 57 cal/kg (RDA would be 0.5g/kg) to 40 cal/kg (RDA would be 1.02g/kg).

Now I personally weigh about 57 kg and eat about 1750 cal/day. This is not particularly low for a CR diet, yet at 31 cal/kg is way off the chart. Fortunately the data were fairly linear at that end, so I extended the table with a ruler and came up with a requirement of 1.5g/kg! That is near double the US RDA of 0.8g/kg and would put my daily requirement for protein at 87g, which is about what I eat.

For your own use, try this formula:

RDA protein (g/day) = 3.22 BWInKg - .055 totalCalsPerDay

So if you weigh 70 kg and eat 1800 cal/day, your protein RDA is 3.22 * 70 - 0.055 * 1800 = 126 g/day. The above formula is a linear fit to the low end of the data in Table 1-6, pg 24...

However, I still think this is an underestimate, because:

(1) Protein requirements are higher for lean body mass than for fat, and

(2) these experiments would have been done on people of normal body fat content, and

(3) CR people have lower-than-normal body fat.

If we assume that normal people have 20% body fat, the above formula can be rewritten in terms of lean body weight (LBW):

RDA protein (g/day) = 4.03 LBWInKg - .055 totalCalsPerDay

Now if a CR person has 10% body fat, then a 70kg 1800 cal/day person requires 4.03 * (0.9*70) - (0.055 * 1800) = 155 g/day. That would be 34% of calories from protein, which seems very high. My own requirement by this formula would then be 108g/day, meaning that 25% of my calories should come from protein.

[PW added]:

In animals on CR, protein turnover it more rapid. During protein turnover (catabolism followed by anabolism) not all amino acids (nitrogen) is reclaimed. Some catabolic products are not reclaimable and are excreted in one form or another, thus increasing nitrogen (protein) loss.

[Walford points out the limitations of these data, which are relevant, but not IMHO of very little likely significance]:
> none of the material in Modern Nutrition in Health and Disease was
> derived from long-term CR studies in either rodents or humans, and may not
> be directly applicable since CR is (probably) a different metabolic state
> than either controls, short-term fasting, or starvation.


So it would appear that, simply because you re eating less FOOD, you're gonna need a surprisingly large amount of protein. In the face of such estimates, Sears' recommendations for AVERAGE people(which NEVER exceed 1.0 g/POUND of LBM, even for elite athletes) seem really quite tame. Indeed, for a person of av g weight, body fat, and activity level, his suggestions come out rather close to the RDA. High protein, my ass: Sears recommends an excessively LOW protein diet for CR!

But let's move on to the Zone issue per se. The most controversial, and most definitive, Zone hypothesis is that optimal health is to be achieved within the context of a CR diet in which the protein:carbohydrate ratio is held constant at a figure between 0.6 and 1.0, with 0.75 optimal for most. The reason for this lies in the effects of such a ratio of macronutrients on hormonal response to a meal (most notably the ratio glucagon: insulin), and resulting effects on a variety of biochemical parameters, most notably eicosanoid synthesis. Let's start with the hormonal effects of macronutrients, individually and in mixed meals, and work our way to eicosanoids.

Now, it is the Zone line that protein has a much lower insulin-stimulating effect than does CHO. Immediately, numerous folks with reasonable memories will recall fairly recent posts from Doug Y and Sherm, which seem to have been interpreted as showing that protein had at least equal, and sometimes greater, insulin-stimulative powers as does CHO. Actually, the data cited by DY and MS have been partly misread, and partly over-extrapolated (this latter being in part the fault of the abstract, as distinct from the full-text, of DY's studies). They, along with other, related studies, actually help prove Sears point, FOR HEALTHY INDIVIDUALS, though perhaps not for NIDDM folks. Let's dig into 'em.

First, the insulin index paper (63). This paper did NOT show that protein has a high insulin response, but the reverse. Total carbohydrate (r =0.39, P < 0.05, n = 36) and sugar (r = 0.36, P < 0.05, n = 36) contents were POSITIVELY related to the mean insulin scores, whereas fat (r = -0.27, NS, n =36) and protein (r = -0.24, NS, n = 38 ) contents were NEGATIVELY related [my emphasis]. What caused the confusion was the statement that protein-rich foods ... elicited insulin responses that were disproportionately higher than their glycemic responses. This, however, is what one would expect: as noted above, protein does cause SOME insulin response, whereas the higher the protein content of a food, the lower its glycemic response will be -- the extreme case being a slab of lean beef, which has essentially zero carb (although it still elicits a small glycemic response), but a nonzero insulin response. This doesn't mean the insulin response PER CALORIE -- i.e. the insulin index's absolute value -- is high, however. In fact, the II of the protein foods, on average, tested lower than the averages for most other food categories tested (fruit, bakery products, confectionary) except (counter-intuitively) breakfast cereals, and CHO-rich foods, which were a little lower in II -- but this is reversed by the removal of baked beans, which were mysteriously put in the high-protein category despite obviously not belonging there, having but 12.17 g protein for 52 g CHO (including 12.7 g fiber), as per DWIDP, NOT INCLUDING the tomato sauce which the paper says they were in! The beans thus, not surprisingly had by far the highest II of all the foods in the protein category: 20106, over TWICE its nearest food in the category (fish, at 9350!) and nearly THREEFOLD the category average excluding it (7453). This latter figure makes protein foods tied for lowest II as a category.

And, indeed, previous research found that a 50g portion of protein from beef, tested in normals, yields an insulin AUC that is only 21% (64) to 28%(65) of that of 50g of carb as glucose . (At a guess, the difference here might be due to the use of lean beef by (65) vs. (64)'s standard beef: fat slows gastric emptying).

Now, some of you are remembering Doug Y's posts of abstracts (66, 67) reporting that many protein sources, including beef, had insulin responses nearly identical to that of glucose, and created a heady synergistic insulin response when combined with the latter. True, but there are two reasons why these reports do not contradict the Zone in healthy folks. First, these latter reports were on NIDDM patients. No, I am not just waving my hands: (66 and 67) were reports from the SAME group that issued (65), above (and several similar reports), using the same methods. And (65), which did test CHO combined with pro, did NOT report a synergistic effect in NORMALS, but rather a summed effect which was non-significantly LOWER than the sum of the insulin responses of the beef and the glucose alone. Likewise, (64) actually compared healthy folk and NIDDM and reported a dramatically higher insulin response to protein in the latter. So, from these papers, it would appear that, in normals, replacing CHO with protein will lower insulin response to a meal, while simultaneously increasing glucagon.

But it goes further. (68 ) reported that, if CHO is held constant at 58 g, increasing the protein dose beyond an initial 10 g all the way up to 49.9 g does not increase the insulin response! This finding is highly counterintuitive: one might expect continuous additive increases; but it is replicated in (69), an experiment done by the same authors as (65, 67, 68 ) with a very similar experimental design as (68 ) (constant CHO, increasing PRO), precisely because it was designed to test another of (68 )'s findings (to which we shall return). (69) found that there were ONLY further increases in insulin response when (wait for it!) the P:C was elevated to 1:1 -- that's right, the extreme end of the Zone-favored P:C ratio (0.6-1.0, with 0.75 considered optimal for most folks). (68 ) never got up this high (its highest P:C was 0.86), so the two are consistent in this regard.

Thus, on the one hand, if CHO is held constant, then as protein is added into the meal, P:C climbs, glucagon levels go up (69), but insulin levels stay steady, WITHIN THE CONFINES OF THE ZONE. And, on the other hand, REPLACING CHO with protein will actively LOWER insulin, as the summed effect, in normals, is less than additive -- within the Zone -- and the insulin response of protein is lower than that of carb to begin with. So, in an isocaloric diet, you get both a significant increase in glucagon, and a significant reduction in insulin, resulting in a VERY significant elevation in glucagon:insulin, when you raise P:C by replacing CHO with protein -- within the Zone. Beyond 1.0 P:C, you plateau (glucagon and insulin are both elevated) or perhaps begin slowly losing ground.

Now let's go back to fat: since fat slows the release of carb into the blood, and thus blunts insulin response t a meal (as per the insulin index, above, and as expected from fat's effects on the GI), replacing some carb with fat (or lowering carb while holding fat steady) FURTHER lowers insulin response, further elevates the glucagon:insulin ratio. See how this all fits together?

Now, I'm sure we all agree that lowering insulin levels is a good thing, to avoid the known consequences of constantly high insulin levels: slow development of insulin resistance, and thus hyperinsulinemia, high blood pressure, dyslipidemia (due in large part to insulin's stimulatory effects on HMGCoA reductase (70)), and serious risk of CHD. But why do we care about this glucagon: insulin ratio so damned much? Well, for one thing, insulin isn't the ONLY hormone with effects on HMGCoA Red: while insulin stimulates it, glucagon inhibits it (70). So the higher g:i gets, the more powerfully cholesterol synthesis is inhibited. But beyond this -- and here is Sears real focus -- the glucagon:insulin ratio is one of the most important diet-dependent modulators of eicosanoid synthesis.

But first, a little background. All eicosanoids are produced from long-chain PUFA, which come either direct from the diet, or are synthesized endogenously. This latter is done by taking short-chain PUFA (ALA and LA) and adding more double bonds in, using desaturase enzymes (and elongation enzymes, but it's the desaturases which are rate-limiting). The desaturation of LA goes LA [delta-6 desaturase]--> GLA ---> DGLA, which is then either used to make series 1 eicosanoids, like PGE1, or is acted on by delta-5 desaturase, forming AA and thence series 2 eicosanoids. ALA also uses these same enzymes, using d6d to turn ALA into SDA, and d5d to make EPA from ETAn3; from EPA, series 3 eicosanoids are made. Broadly speaking, series 1 eicosanoids are good for one in the long term, and series 2 are bad : series 1 are anti-inflammatory, hypotensive, immunostimulative, antithrombotic, etc, while series 2 eicosanoids are the reverse. ( The exception is PGI2, prostacyclin, which is antithrombotic -- but as we ll see below, it's not a real exception in terms of the practical implications). Series 3 LEAN toward the good : they are certainly dramatically less inflammatory etc than series 2. Accessible details on the role of eicosanoids in heart disease, cancer, and the whole mishmash can be found in _The Zone_ and Simopoulos _The Omega Diet_, but the former ignores the role of series 3, and the latter, that of series 2.

Bottom line: if you can lower the d5d:d6d ratio, you will produce more series 1 and less series 2 eicosanoids, thus improving your odds of long-term health.

Now, how do macronutrients -- and their effects on the glucacon:insulin ratio -- affect these enzymes? Hard human studies have not been done on this issue DIRECTLY (but see below). However, extensive rodent data are available (reviewed in 50-52). What these data show is that insulin (which is stimulated by both CHO and protein, though much more by CHO) stimulates, while glucagon (released by protein exclusively) inhibits, the desaturation enzymes, so that higher P:C (& thus an more-than-linearly higher glucagon:insulin, as discussed above) results in lower EFA desaturation by delta-5 desaturase. This effect is reinforcedby the fact that glucagon inhibits insulin release.

If the hormonal effects were the only factor involved, then the same would hold for delta-6 desaturase. HOWEVER, the presence of protein ITSELF, sans mediator, potently *activates* d6d, sufficiently (in combination, presumably, with its effects on insulin) to overcome the inhibitory effects of glucagon, an effect not seen with d5d. This fact is offhandedly explained as being due to stimulation of protein synthesis, as if this self-evidently explained the discrepancy. It does not, at least not to my educated layperson's ears. But the end result is clear: the higher one s P:C, the more d6d is activated, and the more d5d is inhibited. In other words, higher P:C means more series 1 good eicosonoids and less series 2 bad ones. It also means less-peroxidizable PUFA (DGLA vs. AA) will be available for mt membrane incorporation.

But as you increase protein, you also increase glucagon relative to insulin, which inhibits both d6d and d5d. So beyond a certain point, we expect that we will see diminishing/negative/plateau returns on increasing P:C. And where is the Goldilocks point for this effect? Glad you asked. A rodent study (53) found tht the protein activatoin of d6d increases with isocaloric P:C, with the peak activity reached at a ratio of 0.64,after which it leveled off. Unfortunately, th next data point is P:C of ~1.0, so we don't know if an intermediate ratio (say, 0.75 :) ) might have produced a higher peak, from which subsequent values slowly fell. But clearly, a zonish P:C is the best range; at the very least, no additional gains are to be made after P:C 1.0, and gains ARE to be made up to at least 0.64.

But, you say, this is rodent data. Our hormonal metabolisms are different. How do we know that the ratio will be at all similar in humans?

As noted above, the direct dietary effects on desaturases have not been characterized, AFAIK. However, there is a fair amount of research comparing the desaturase activity in healthy humans, insulin resistant persons/type II diabetics, (who have unusually HIGH insulin levels) and type I diabetics (who produce NO insulin themselves and have to shoot up). All of this ends up giving us a consistent picture of the role of insulin (but not glucagon, as yet, AFAIK) on desaturases in humans. Here are the data:

(1) Type I diabetics fed DGLA show the same rise in TG and phospholipid DGLA as do normals -- but their AA levels stay at baseline until they shoot up, whereupon their AA levels climb to the normal range. The effect of insulin and the data from the literature of animal studies suggest insulin dependence of delta 5 desaturase in humans. (71)

(2) Obese children have the same LA levels as normals, but higher levels of fasting immunoreactive insulin, and of all downstream metabolites, and in particular higher AA. We conclude that the significantly higher values of n-6 long-chain polyunsaturated fatty acids (LCP) in plasma lipids of obese children than in age-matched controls may be caused by an enhanced activity of delta 6-desaturation, and we speculate that elevated fasting immunoreactive insulin seen in obese children...may stimulate synthesis of n-6 LCP fatty acids. (72)

(3) Insulin action ... correlated with composite measures of membrane unsaturation (% C20-22 polyunsaturated fatty acids ... ), unsaturation index ... , a number of individual fatty acids and with delta5 desaturase activity ... The results demonstrate that delta5 desaturase activity is independently related to both insulin resistance (73).

(4) Insulin positively correlated with increased C20:4n6/C18:3n6 (index of delta 5-desaturase) (p < 0.05) and C20:4n6/C18:2n6 (index of overall n6 pathway activity) (p < 0.01) in serum, and the n5 pathway in platelets (p < 0.01), but there was no correlation for insulin with platelet C18:3n6/C18:2n6 (index of delta 6-desaturase activity). (74).

What about direct measurements of the results of P:C on eicosanoid synthesis in humans? Well, we don't have them, although granted what has been covered to date, it d be pretty damned surprising if things didn't work out similarly to what's seen in rodents. However, early on in his work with atheletes and others, Sears (54) developed a qualitative battery of tests with which to test out different P:C for effects on eicosanoids. These are mocked, without explaining their interdependence and underlying logic, in _Beyond_. Most of tests are based on the known effects of eicos themselves (eg. vasodilation, keratin synthesis, peristalsis); others are based on the known effects of insulin and glucgon themselves on balancing blood sugar etc (see below), from which we can partially extrapolate to eicosanoid synthesis as per the above; yet others appear to be purely Sears intuition that you should feel good in the Zone. And what he found (apparently in utter ignorance of (53), to which he never refers or includes in his bibliographies) was that the Goldilocks point -- the P:C where people entered the Zone -- was between 0.6 and 1.0, and for most folks, at 0.75. It was on the basis of THESE data tjt Sears set the Zone goldilocks point here; again, he seems ignorant of the op cit rodent data.

This may be a coinkidink [coincidence], but it's pretty damned striking if so.

And, strangely, the synthesis of the only good series 2 eicosanoid -- PGI2 -- is actually inhibited by insulin (56). So you re not losing out on even THIS by Zoning.

High-protein diets are widely believed to increase risk of heart disease. This, I believe, is entirely due to the confounding variable that those who get the most protein (especially in the US, which has unusually fatty meat) tend to also get the most saturated fat and long-chain n6. When we isolate out this factor, we find that women who eat more protein have a LOWER risk of ischemic heart disease (48 ). Likewise, I find that, just as replacing CHO with MUFA improves CVD risk factors, so it is with protein (49). This is to be expected, at least from a cholesterol POV, because of the known effects of insulin and glucagon on VLDL synthesis, CETP, and HMGCoA reductase (30, 70); also, the P:C's effects on vasodilation, thrombosis, hypertension, etc; and perhaps because more protein may lower Hcy, counterintuitively REFERENCE. Hence the much greater risk of CVD mortality in Syndrome X and diabetes, in which insulin is elevated, and the benefits of the P:C largely nullified.

Now, Walford's third objection on the Zone is as follows:
> Third: the Biospherians were on 10-14% calories from protein, 10% fat, the
> rest carbohydrates, including lots of bananas (high GI index) and sweet
> potatoes; had BMI's averaging around 20 (i.e., good but higher than a
> lot of persons on the Internet CR list) -- but had health risk
> factors (blood sugar, lipids, etc) somewhat better than what seems
> the averages in the Internet group.

I've already addressed everything here except glucose. PH [author's initials] raises the same objection:
> What I find curious is blood sugar levels being reported on this list
> (Zoners and some non-Zoners alike) that are not nearly as 'good' as
> those uniformly found in Walford's half-dozen Biospherians despite the
> latter's high banana intake.

And, curiously, they ALSO aren't as good as Markovik's data, which (again) used Zone macronutrients, but (crucially) did NOT balance P:C at every meal. This has caused much angst amongst we Zone types, myself included. But in the course of looking at the data on the impact on insulin secretion from adding protein to carb, above (68, 69), as well as reading a paper on GI (80), I came across the answer to this conundrum -- and it FURTHER emphasizes the superiority of the Zone.

Let's start at the beginning. (68 ) reported that, as more protein was added to a fixed am t of CHO, the initial glucose peak was blunted. This makes sense, because (as noted in the discussion of this paper and related stuff, above) adding protein does cause a mild increase in the insulin response to a meal -- a less-than-fully additive one. Yet it does NOT, itself, significantly increase glucose levels. So the extra insulin will drive a bit more sugar into the muscles, blunting the blood glucose response.

Then, as the initial postprandial glucose and insulin surge wanes and the Protein's glucagon-stimulating effect comes into play, glucose levels would be expected to rise again -- and so they do, whereas in the CHO-only meal they continue to fall, more and more slowly.

So the picture is that, DOSE-DEPENDENTLY, higher and higher P:C (at least up to 0.86) result in smoother and smoother glucose responses: less extreme peaks AND nadirs. Likewise, the glucose AUC (area under the curve) was dose-dependently reduced by added protein .

In short, the blood glucose response to adding increasing protein to a meal looks pretty darned close to lowering its GI.

Now, this would seem to offer a pretty attractive explanation for why we Zoners have higher FBG than the Biospherans (although their intense physical labor is clearly also a factor, this had begun before closure, and so doesn't explain the DROP, though it may explain a significant amount of the ABSOLUTE NUMBERS). This may seem an unreasonable speculation from the similarity of their 2h glucose curves, but work by Wolever (83) has shown that eating a low-GI supper leads to greater insulin sensitivity at breakfast. And another trial (80) reported that the increase in insulin sensitivity at lunch after a low-GI breakfast was accompanied by increased plasma glucose just before the second meal, and that meals which had just slightly higher GIs, and did not result in higher 4h glucose levels, did not improve insulin sensitivity. So the finding or higher FBG in Zoners is quite consistent with what's been found in low-GI meals -- and, indeed suggests a benefit in sustained insulin sensitivity.

IIRC, those Zoners who have both FBG and F insulin have seen no decrease in the former but significant decreases in the latter, consistent with the above.

But what folks are worried about is AGE. So take pyridoxamine! No, no, htat's a joke. Seriously: the conventional wisdom in the group is that FBG is the key number, since we don't spend much time in the immediate postprandial state, and since it gives us a baseline level onto which the postprandial spike is added. But glycation happens dose-dependently. According to Brand-Miller, in her popular book GET REFERENCE, lowr GI foods do reduce, not just the 2h curve, but the TOTAL GLUCOSE EXPOSURE from the meal, ever, in spite of the slight elevation in the late period. And, in addition, it SEEMS TO ME that glycatoin must proceed more than linearly with glucose concentration, so that doubling blood glucose must more than double glycation -- just because of sheer kinetics. If so, It seems to me that one is in much more danger from the massive postprandial glucose spike than one is from slightly elevated FBG, even if the latter lasts longer. And blood glucose over the 2h postprandial period is both lowered in total amount (AUC), and made smoother and less spiky. As in, 50 g glucose alone yields an AUC of 1700 mg*min/dL abpve baseline, vs. 0.35 for the protein-added group. It therefore seems to me that a higher P:C would significantly reduce glycation.

The best evidence for this would be to test glycation on high- and low-protein, isocaloric diets in which P:C is held constant at every meal; I don't know that this has been done. The NEXT best thing, however, would be to look at the effects on glycation from high-and low- GI meals, due t o the similar effects on postprandial glucose curves. Several studies have shown that low GI meals do reduce glycation in rodents; however, they were relatively short-term studies. The only lifetime study comaparing different CHO sources for effects on AGE(82) found (surprisingly) that, whereas CR inhibited glycation, different CHO sources (from high-GI glucose down to low-GI fructose) does not. This at first might seem a point against my hypothesis.

However, remember that (68 ) used a CONSTANT amount of glucose, with increasing protein dosage, to achieve their results. As per the Zone, one isn't doing that, but LOWERING CHO while holding protein constant (Sears purists) or actually elevating protein (per Sherm's extrapolation). And further remember that youre actually to use low-GI CHO on the Zone in ADDITION to the higher protein intake, and getting the GI-lowering addition of a relatively high %fat. So you re looking at both lower total POSSIBLE glycemic exposure, and considerable blunting of the postprandial glucose spike due to (a) higher P:C, (b) lower CHO GI, and (c) GI-lowering mono fat, relative to a Walfordian mode.

And remember that there may be tradeoffs, glycation-wise, in the lower-GI carbs vs. higher ones: galactose and fructose, IF they circulate as such systemically, are more potent glycators, unit er unit, than is glucose. This may explain the lack of long-term effect on AGE of lower GI sugars: they lead to significantly lower glucose, but somewhat higher levels of a more glycating sugar. Combining the GI effect with the lower absolute CHO exposure from all sources, AND the effects of higher P:C and fat:CHO ratios on the Zone, might be expected to tip the balance.

PLUS, remember the insulin-sensitizing effect mentioned above. This wouldn't play into the results of (68 ), as CR animals tend to gorge all at once, so they have a 24h wait for their next meal; there's no guarantee that insulin sensitivity is still significantly higher THAT much time after a low-GI meal, whereas timing meals at 4h intervals (as I do) clearly does improve glucose tolerance, per (80).

All that said: many Zoners have failed to see reduced HbA1C. So maybe I'm full of shit :). OTOH [on the other hand], many classical Walfordians aren't seeing much progress here, either.

Now, what about the really interesting issue: aging per se? Again, the evidence appears to be that higher-protein diets decelerate aging. (Some of you just thought, But wasn't that recently disproved? Read on!).

As part of his eat more protein post, op cit, Sherm provided these quotes from The Retardation of Aging and Disease by Dietary:

> Restriction, by Weindruch and Walford:
> The average LS [lifespan] of rats on DR increased as the amount
> of protein in the diet increased suggesting that the
> effects of DR on LS may be enhanced by diets high in protein
> content. (pg 54) ... epithelial ... adrenal and thyroid
> tumor morbidity were ... inversely related to dietary protein
> ... In short, both LS and tumor data suggest that high
> protein diets accentuated the benefits of DR in Ross' colony.
> (pg 81) [the extremely restricted rats that did best were on
> a 51% calories-from-protein diet!]

Walford replied:
> First: looking at Table 2.4 in the Weindruch/Walford book, study 7, one
> observes that the shortest LS's were at 8% casein; but there wasn't much
> difference between 21% casein and 51%. In fact, 21% gave the longest
> MLS, better than both 30% and 51%.
> Looking also at Study 9, 10% casein was shortest, and 51% better than --
> but not much better --than 20%. With regard to tumor incidence, the stated
> comparison is between the 51% casein and the 8% casein. So the higher
> protein seems better but only when compared with a very low protein
> intake. The intermediate ranges might be even better.

At the time, I was not in possession of a copy of _The Retardation_, but WAS in possession of the only paper of Ross which I could find in which various %proteins were tested for effects on LS, tumor incidence, and malignancy (77). I responded:


But that isn't what Ross' data shows. In fact, to my surprise, it shows what we are constantly told that the data DON'T show: that MACRONUTRIENT COMPOSITION CAN INFLUENCE MAX LS, and that the higher the protein, the longer the LS.

Turning to their [77] table 2, we find that there is improved max LS in BOTH the Cr AND the ad lib groups, correlated positively with protein intake.

> >>>DAYS

> >>>


> >>>

> >>> 10% 22% 51% 10% 22% 51%

> >>> > >>>900 8 2 6 145 116 149

> >>>1000 0 0 1 93 85 122

> >>>1200 0 0 0 48 61 91

> >>>1300 0 0 0 5 31 59

> >>>1400 0 0 0 0 11 24

> >>>1500 0 0 0 0 1 10

> >>>1600 0 0 0 0 0 3

Lest anyone object that what we're really seeing in the high-pro CRONies is just adequate-protein diet, note that the same results were observed in the ALers [those who follow ad lib diet].

Further, Ross' table 10, summarizing all tumor data, shows that, again for both ALers and CRONies, the highest-protein groups had the lowest ratios of malignant to benign tumors, and the lowest AGE-SPECIFIC tumor rates of all isocaloric groups: that is, the low-protein group had fewer cancers only because they died sooner -- thir rate of tumor formation was higher.

This is only one study, but they used a lot of animals (1600), and I know of no contradictory evidence. Even [78] tends broadly to support these conclusions.


[After posting the above, but BEFORE Walford's recent proxy post to the list on protein, I wrote the following]:


Again, at the time, I was not in possession of a copy of _The Retardation_, and thought there must simply be a mistake of some kind in same. I just got a copy out on interlibrary loan, however, and what I actually find is that Walford is referring to ANOTHER study, also by Ross's group (79). I haven t seen this study, but will assume that the book accurately summarizes the data. And, yes, in this study *a* group of 21% casein rats did outlive *a* group of 51% rats, albeit not by much: 49 mo vs. 46. Now, right off the bat, there's a question of the power of this result, since there were more than half again as many rats in the 21% group as in the 51%, which can result in more absolute numbers of animals reaching a higher max LS just by default; a similar problem exists with the lower TOTAL population of this colony compared to (77) s, which makes the latter a more important study to look at. But sheer population numbers are the LEAST of the problems with this study.

It would appear that the Grand Old Man wasn't reading the tables as carefully as he might when he wrote his reply above, for two reasons. First, from the table itself, I note that the 21% casein CR group in THIS study (79) were only fed 16 Cal/day -- vs. 25 Cal for the 51% casein CR rats! The average AL rodent in this study ate 75.2 Cal/day, so the 51% casein group were 66% restricted -- but the 21% casein animals were *78%* restricted! By contrast, (77)'s diets were isocaloric.

And then, I see that the table's notes indicate, Groups 7g and 7h [the 21% casein CR and AL groups in (79), respectively] ate diets enriched in vitamins and minerals relative to the other groups in the study. Now, whether this means the 21%-protein animals diets were enriched just enough to make the 21% casein CR diet isonutrient with less severely restricted 30% and 51% casein CR animals, or were higher nutrition in an absolute sense, is unclear from the table, as BOTH the AL and CR med-pro groups diets were so enriched. Having looked at the paper, I'm still unclear what the story is.

But whatever the case, we see that the 21%-protein animals were 12% more restricted, but they only lived 6% longer. Further, the AV G LS of the two groups in question were the same! When you extend max LS but DON T extend av g, it means that more critters are dying off early (relatively speaking) under the influence of the regime. This is hardly a ringing endorsement of the 21%-protein regime, especially if they were more (micronutrient) ON as well as more CR.


By strange coincidence, Walford's second analysis of (79) etc , which confirms its obscurity and the difficulty in using it to make meaningful comparisons, came just DAYS after I wrote the above. Importantly, W notes that, in the more easily-interpreted, better-controlled (77), while max LS was greater in the high-protein group, mortality-rate doubling time (a more accurate measure of aging per se, many agree -- see Austad) was better in the 21% casein group, as per data from Finch, _Longevity, Senescence and the Genome_, Table 10.1 on page 508 (thanks to Sherm for providing this reference: I don't have Finch to hand).

For graphic illustration of the relationship between a survival curve and the slope of a line of age-specific mortality, and thus MRDT, see:,5716,4555+asmbly%5Fid,0

Well, that SOUNDS bad, I agree. However, aside from the inherent counter-intuitiveness of a discrepancy between max LS and MRDT -- between the longevity of the longest-lived, and the rate of aging, which is the very factor which determines same -- there is actually some reason to question Finch's math in calculating MRDT. Again, I am not just hand-waving. Here is Aubrey de Grey on the subject:

The main quarrel that I have with the usual way of calculating MRDT (as set out by Finch) is that it considers the whole life table from puberty to the death of the oldest member of the population. This has problems at both ends. At the young end there are many known examples of early mortality of a substantial minority of a population, resulting from environmental factors which may be altered by an intervention such as CR but do not intuitively have much to do with the rate of aging. At the old end there is the well-known deviation from the Gompertz curve, the deceleration of mortality, whose magnitude is -- according to some models -- highly dependent on the population's degree of heterogeneity (either genetic or due to environmental influences during development) for characteristics that affect the rate of aging.[ST_rn=ps]

To eliminate the prejudice of the excessive early mortality (seen, NB, in the 22% casein group (which, of course, is not exactly a big score in favor of this regimen)) and the effects of genetic heterogenity, Aubrey's calculation methods (personal communication), after much math, boils down to a formula which ought to be surprisingly simple:

But the slope is (as above) 0.5 divided by the difference between the age at which 1/4 are dead and the age at which 3/4 are dead. So, that difference is [2.27*] the MRDT itself, not even any scaling factor needed.

Now, this is actually made a bit more complicated, because Ross only reports his data every 100 days. Thus, I made the assumption that animals died at an exactly even rate in these 100 day periods, so that (e.g.) if there are 197 animals alive on day 700, and 176 on day 800, and I want to know when 187.5 animals are still alive (the 3/4 point), then I calculate that this point is reached at day 700 + {100 * ([197-187.5]/[197-176]) } = 745.2. And so forth.

Quoth Da Man [Aubrey de Grey]:

I bet that this way of calculating MRDT gives numbers that correlate exactly with max LS in the Ross+Bras study.

He wins his bet!

I pumped the data for (77) thru , with the above assumption, and found an MRDT for the 51% protein group of 403.7 days, versus 398.4 for the 21% protein group . So by THIS method, the Gompertz slope -- the rate of acceleration of mortality -- is ever-so-slightly HIGHER in the 21% casein group -- or, IOW, it takes just a little bit LONGER for the 51% casein group's age-specific mortality rate to double.

Or, to simplify further: the high-protein group was aging more slowly. But because the 22% casein group experienced greater early mortality, the illusion was cast that they had a slower rate of aging later on.

And, in fact, (77) actually PROVIDED the age-specific mortality rate for each 100 day period -- and it's consistently lower, after the early period before 400 days, in the 51% protein group.

And how could this happen? Here's ANOTHER of my little theories at work. Remember what has come before: higher P:C, up to a point, somewhat increases the activity of d6d, and inhibits the activity of d5d. The result to be expected of this, in terms of PUFA available for MIM incorporation, would be that higher P:C leads to somewhat more medium-chain PUFA, but considerably less long-chain PUFA: eg. somewhat more SDA (18:4n3) but much less EPA (20:5n3) and (I ASSUME) even less DHA (because of the extra desaturase step required, which (at the very least) is rate-limiting, and (I expect) should also be inhibited by glucagon and stimulated by insulin). As has been discussed on a recent thread, it is my hypothesis that aging is accelerated by the incorporation of more long-chain PUFA in the mt inner membrane. So the above scenario suggests that a higher P:C, within the Zone, will actually slow aging. and if, as I do, one restricts one's PUFA intake to the ACTUAL EFAs (LA and ALA), the total long-chain PUFA produced will be reduced further, due to competition at BOTH rate-limiting steps for the desaturase enzmes, without keeping DHA out of brain cell membranes. These rodent data thus happen to tend to reinforce my pet hypothesis: although the P:C was too high to be optimal in the longevous 51% casein group, it was also too low in the shorter-lived 22% casein group -- and the rodent data suggest that the declining slope, above the goldilocks point, must be much lower than the slope leading up to it from a desaturase POV.

Note that this is, in fact, observed in CR animals (84): aging is associated with greater accumulation of long-chain PUFA in MIM, which is retarded by CR. Higher P:C is thus reinforcing a known action (and perhaps a causal one) of CR.

And then there's the human evidence. IF Sears review of the evidence is accurate -- and my experience has been that it generally is, and abstracts to be cited tend to suggest that he is here too [although I positively BEG EVERY PERSON ON THIS LIST to see if their local Universities carry any of these journals, and to dig up the papers he cites (85-88 ) which my U doesn't carry (I beg you! I d pay you for copies! You d be a hero f the Revolution!)] -- then THE OKINAWANS ARE IN THE ZONE.

In Sears latest book (89) , I find the following statements, evidently derived from the above papers:

-Chart showing that the Okinawans eat 100 g of soy protein per day (p. 12). Same statement made on pp. 25 and 231. I believe that Sears is wrong, here: I think he has confused intake of soy PROTEIN SOURCES with soy protein itself. An article I read (90) suggests that the average Japanese person eats ~34 g of soy foods (IIRC) which provide 7.7g pf pprotein (I'm quite sure of this datum). Sears suggests that the average is 40 g.

-Statement that they eat twice as much fish as mainland Japan (p. 12).

-Statement that they eat 2-3 times more vegetables than mainland Japan (ibid).

-Statement that their diet is ~30% fat (p. 235).

Statement that the most important factor for the Okinawan longevity is that they consume 20 to 40 percent fewer calories than the Japanese. (p. 12)

In the abstracts to (85-88 ), I find statements which are certainly consistent with the above, although they don't give numbers (PLEASE get the papers, folks!!); more importantly, however, what they DO say is that the incidence of longevity within the Japanese population is correlated with more protein and fat in the diet -- i.e. the more Zonish the diet, the higher the incidence of centenarianism:

They took rice or potato as carbohydrate with abundant vegetables and vegetable protein or fish protein. (87)

Nutrient intakes in 94 Japanese centenarians investigated between 1972 and 1973 showed a higher proportion of animal protein to total proteins than in contemporary average Japanese. 2. High intakes of milk and fats and oils had favorable effects on 10-year (1976-1986) survivorship in 422 urban residents aged 69-71. The survivors revealed a longitudinal increase in intakes of animal foods such as eggs, milk, fish and meat over the 10 years. 3. Nutrient intakes were compared ... between a sample from Okinawa Prefecture where life expectancies at birth and 65 were the longest in Japan, and a sample from Akita Prefecture where the life expectancies were much shorter. .. the proportion of energy from proteins and fats were significantly higher in the former than in the latter. (86)

we studied geographic distribution of centenarians in Japan, ...By prefecture, the highest proportion lived in Okinawa (133.8 ), whereas the fewest were found in Akita (8.9) [see above on diet comparisons between the 2] .... correlation coefficients between proportion of centenarians and protein (% of energy) was positively significant, while intake of total energy was negatively significant. (85)

The energy intake of the Okinawan centenarians living at home was about 1,100 kcal/day for both sexes, which was similar to that of centenarians throughout Japan. (88 )

This reinforces my belief above: 100 g of soy protein would barely leave any room for anything else in an 1100 Cal diet -- let alone double the large Japanese national average in fish!

Now, if, indeed, the Okinawan centenarians eat an 1100 Cal diet; if it contains a great deal of protein, from fish and soy, and greater protein intake is associated with greater longevity; if it is 30% fat; if the carbs are fairly low-GI, high-nutrient ones, at least compared to the average Japanese person (who certainly eats a lot of rice and not enough veggies) -- Indeed, there really isn't much room for rice or potatoes; then their macronutrient ratios are must be pretty damned Zonish. If we really can take Sears 100 g soy protein figure seriously, then the numbers work out to be P:C:F of > 36: <34: 30. But even if not, it would seem that the Okinawans are in the Zone -- and Zonishness of diet correlates with % of centenarians in the Japanese population.

PREVIOUS: Part One: More Fat | NEXT: Miscellaneous Observations

Top of page

Home | What's New | CR Science and Research | CRONies | Links + References | 2000 Poll: By The Numbers Webmastering+Admin

Last modified: Saturday, February 15, 2020 11:12 AM