High-Protein Diets

00 Abstract

The adult protein Recommended Dietary Allowance of 0.80 g·kg⁻¹·d⁻¹ is a nitrogen-balance estimate of the intake that covers the estimated average requirement of a reference adult plus two standard deviations (Rand et al., 2003; Trumbo et al., 2002). It is not an optimum for lean-mass retention under energy deficit, for older adults with anabolic resistance, or for people who train. Human dose–response infusions after resistance exercise saturate mixed-muscle protein synthesis near 20 g of high-quality protein in young men performing isolated-leg work (Moore et al., 2009), while whole-body sessions and older muscle shift that ceiling upward (Macnaughton et al., 2016; Moore et al., 2015). A meta-regression of protein supplementation during resistance training placed a breakpoint for fat-free mass near 1.62 g·kg⁻¹·d⁻¹ (Morton et al., 2018). Those are not the same claim.

In energy restriction, higher-protein diets have, in human meta-analysis and in a tightly controlled training trial, favoured fat loss and lean-mass retention relative to lower-protein comparisons at similar energy (Wycherley et al., 2012; Longland et al., 2016). In healthy kidneys, randomised comparisons have not shown a clinically meaningful deterioration of glomerular filtration rate attributable to higher versus lower protein (Devries et al., 2018); they have shown reversible haemodynamic hyperfiltration (Juraschek et al., 2013). In chronic kidney disease the question changes species: guideline instruments restrict protein in metabolically stable non-dialysis CKD (Ikizler et al., 2020). Prospective cohorts that appear to indict “protein” usually indict a food — processed meat more than fish, red meat more than legumes (Pan et al., 2012; Zhong et al., 2020; Naghshi et al., 2020). The mechanistic argument that chronic high protein shortens life by driving mTOR and IGF-1 is a real biochemistry (Saxton and Sabatini, 2017) that has not been converted into a protein-controlled human survival trial. Mouse geometric-framework work finds the opposite macronutrient optimum from the gym (Solon-Biet et al., 2014). Both results can be true of their own systems. They are not interchangeable.

The phrase “high-protein diet” is not a scientific exposure until it is restated in four units. Absolute intake is grams per day. Relative intake is grams per kilogram of body mass per day — and the kilogram must be named as actual body mass, fat-free mass, or a reference body mass, because the same 120 g·d⁻¹ is 2.0 g·kg⁻¹ in a 60 kg adult and 1.2 g·kg⁻¹ in a 100 kg adult. Energy share is the percentage of metabolisable energy supplied by protein, using the Atwater factor of 17 kJ·g⁻¹ (4 kcal·g⁻¹). Energy context is whether those numbers were recorded in deficit, balance, or surplus, because protein that is eaten while glycogen and fat are being oxidised is not the same physiological event as protein eaten while all three fuels are being stored.

Those four numbers are not interchangeable restatements. Hold energy and body mass constant and raise protein: the percentage of energy rises and carbohydrate or fat must fall. Hold grams per day constant and lose 15 kg: grams per kilogram rise without any dietary change. Hold grams per kilogram constant and compare a 1,600 kcal reducing diet with a 3,200 kcal surplus: the same relative protein is a different fraction of energy and a different competitor for other macronutrients. The Institute of Medicine Acceptable Macronutrient Distribution Range for protein is 10–35% of energy (Trumbo et al., 2002). A 70 kg adult eating 0.80 g·kg⁻¹·d⁻¹ on 2,400 kcal is at about 9% of energy — below the AMDR floor — while the same adult eating 1.6 g·kg⁻¹·d⁻¹ on a 1,600 kcal deficit is at about 28% of energy. Both are ordinary numbers in the literature. Calling both, or neither, “high protein” without the four-unit statement is the first analytical failure this article is written to prevent.

A working vocabulary used in this document, and not treated as a biological law: intakes near the RDA (~0.8 g·kg⁻¹·d⁻¹, often ~10% of energy at typical adult energy intakes) are requirement-range; 1.2–1.6 g·kg⁻¹·d⁻¹ in weight-stable or training adults are elevated; intakes at or above ~2.0 g·kg⁻¹·d⁻¹, or protein shares approaching the upper AMDR in energy restriction, are high in the sense the trials and position stands actually used (Phillips, 2016; Jäger et al., 2017; Morton et al., 2018). The labels are operational. They are not a recommendation.

02 How the requirement was built

Nineteenth-century physiologists treated protein as the tissue nutrient and guessed high. Carl Voit’s 118 g·d⁻¹ for a 70 kg labourer, and the Atwater-era American figures that followed it, were observations of what working men ate, not measurements of what they needed. The modern requirement is a nitrogen-balance construction. Nitrogen in (diet) minus nitrogen out (urine, faeces, skin, miscellaneous) is set to zero, and the intake that achieves zero in a group of healthy adults is taken as the estimated average requirement. Rand, Pellett and Young meta-analysed short-term nitrogen-balance studies in healthy adults and estimated a median requirement of 0.65 g·kg⁻¹·d⁻¹ and a 97.5th-percentile recommended intake of 0.83 g·kg⁻¹·d⁻¹ (Rand et al., 2003). The Institute of Medicine rounded that recommendation to 0.80 g·kg⁻¹·d⁻¹ and set the AMDR at 10–35% of energy (Trumbo et al., 2002). That is the origin of the number that is still printed on every American food label as if it were an optimum.

It is not. Nitrogen balance underestimates the intake that preserves function when the endpoint is lean mass, strength, or recovery rather than zero nitrogen. The method is sensitive to energy intake, to the adaptive downregulation of amino-acid oxidation as intake falls, and to incomplete collection of miscellaneous losses (Millward, 1999, 2001). Indicator amino-acid oxidation, which estimates requirement from the breakpoint at which a labelled indispensable amino acid stops being oxidised because it is being incorporated, has repeatedly returned higher estimates than classical balance for several indispensable amino acids and, in some applications, for total protein (Elango et al., 2008; Courtney-Martin et al., 2016). The two methods do not always agree. Neither is a hypertrophy prescription. Phillips’s review of intakes “beyond the RDA” is the cleanest statement of the gap: the RDA is a deficiency floor derived from one method in one population, and the intakes that support muscle in trained or older adults sit above it (Phillips, 2016).

03 Digestion, amino acids, nitrogen

Dietary protein is hydrolysed in the stomach and small intestine to peptides and free amino acids. Absorption is efficient for most mixed meals; the nutritional difference among foods is less whether nitrogen crosses the gut than which amino acids cross and how fast. Boirie and colleagues distinguished “fast” whey from “slow” casein in human tracer studies: whey produced a rapid, high, transient rise in aminoacidaemia and oxidation, casein a prolonged, lower plateau and better net postprandial protein accretion under the conditions of that experiment (Boirie et al., 1997). The finding is kinetics, not a ranking of foods for every endpoint. A large and amino-acid-specific fraction of the meal is extracted by the splanchnic bed — gut and liver — before skeletal muscle sees it. Volpi’s elderly-adult work is the human demonstration that first-pass extraction is higher with age and that essential amino acids, not the non-essential fraction, drive the peripheral anabolic response (Volpi et al., 1999, 2003).

Amino acids that escape first-pass use enter the free pool, charge tRNA, and are used for protein synthesis, or are transaminated and oxidised. The carbon skeletons feed glucose (glucogenic) or ketone (ketogenic) pathways; the nitrogen is disposed of as urea. There is no storage protein analogous to glycogen or triglyceride. What is commonly called a protein store is the slow turnover of lean tissue. Nitrogen balance is therefore a rate, not a reservoir: the adult who is “in balance” is replacing what is being broken down, not filling a tank. Diet-induced thermogenesis is higher for protein than for carbohydrate or fat. In human respiration-chamber work, Westerterp measured a thermic effect on the order of 20–30% of protein energy, against roughly 5–10% for carbohydrate and 0–3% for fat (Westerterp, 2004; Westerterp et al., 1999). Halton and Hu’s critical review reached the same qualitative conclusion and added the satiety literature (Halton and Hu, 2004). Those are established physiological costs of processing protein. They are not, by themselves, a weight-loss drug.

04 Satiety is not a single hormone

Protein is the most satiating macronutrient in acute human meal studies, with supporting evidence from ghrelin suppression after a high-protein breakfast (Blom et al., 2006) and from 24-hour chamber work in which a high-protein condition raised diet-induced thermogenesis and satiety relative to high-fat (Westerterp-Plantenga et al., 1999, 2009; Lejeune et al., 2006). Leidy’s reviews treat higher protein as a dietary strategy that can reduce daily energy intake and help weight-loss maintenance, with the important qualifier that the effect is context-dependent and not large enough to override an energy surplus (Leidy et al., 2014, 2015). Simpson and Raubenheimer’s protein-leverage hypothesis is a different claim: animals, including humans, prioritise a protein target and will over-eat energy on a low-protein diet to reach it (Simpson and Raubenheimer, 2005; Raubenheimer and Simpson, 2019). Gosby and colleagues reported human evidence consistent with leverage under controlled conditions (Gosby et al., 2014). The hypothesis is plausible and mechanistically tidy. It is not established as the dominant explanation of population obesity, and it does not license the reverse inference that any high-protein diet is automatically hypocaloric.

Skeletal muscle protein synthesis (MPS) can be measured in humans with stable-isotope tracers over hours. After resistance exercise, mixed-muscle MPS in young men rose with ingested egg protein and plateaued at 20 g; 40 g produced no further mixed-muscle synthesis and increased amino-acid oxidation (Moore et al., 2009). Witard and colleagues, using whey and a myofibrillar-specific measurement, likewise found 20 g sufficient after unilateral exercise, with 40 g adding only a modest further increment (Witard et al., 2014). Macnaughton then changed the session: after whole-body resistance exercise, 40 g of whey raised myofibrillar MPS more than 20 g (Macnaughton et al., 2016). The ceiling is not a constant. It moves with how much muscle was recruited, with the protein’s amino-acid pattern, and with the age of the muscle.

Protein quality appears in the acute record as a difference among isolates. Tang and colleagues reported that, after resistance exercise, whey hydrolysate stimulated mixed-muscle MPS more than soy, and soy more than casein, in young men (Tang et al., 2009). Wilkinson and colleagues found that fat-free milk supported greater net muscle protein balance after exercise than an isonitrogenous soy drink (Wilkinson et al., 2007). Yang and colleagues later showed that older men required more whey than young men to saturate myofibrillar MPS, and that soy was a weaker stimulus than whey at equal doses (Yang et al., 2012a, 2012b). These are acute, isolate, tracer studies. They are established as descriptions of the hours after a drink. They are not food-pattern epidemiology, and they are not hypertrophy trials.

06 Hypertrophy, training, recovery

The conversion of an acute MPS pulse into months of fibre growth is not automatic. Morton, Phillips and colleagues meta-analysed protein supplementation during prolonged resistance training. Supplementation increased fat-free-mass gains relative to control; a two-phase regression placed a breakpoint at approximately 1.62 g·kg⁻¹·d⁻¹, beyond which further protein did not further increase fat-free mass (Morton et al., 2018). The mean fat-free-mass increment attributable to supplementation was small in absolute terms (~0.3 kg). The result is strongly supported for trained and untrained adults in the trial set that entered the meta-analysis. It is not a demonstration that 1.62 g·kg⁻¹ is a universal optimum, and it is not a demonstration that 2.2 g·kg⁻¹ is harmful. It is a statement about diminishing returns for fat-free mass under the training programmes studied.

The International Society of Sports Nutrition position stand places a daily range of 1.4–2.0 g·kg⁻¹ for exercising individuals, with higher intakes considered in energy restriction (Jäger et al., 2017; Aragon et al., 2017). Helms and colleagues, reviewing protein during caloric restriction in lean resistance-trained athletes, argued for the upper part of that range when the goal is to hold lean mass at low body fat (Helms et al., 2014). Those are position-stand and narrative-review inferences from a thin trial base in very lean athletes. They are not RDA revisions.

Recovery is a narrower question. Pre-sleep casein has been shown, in human tracer studies, to raise overnight myofibrillar protein synthesis and to be incorporated into muscle overnight (Trommelen et al., 2016, 2018). That is an acute recovery mechanism. It is not, by itself, a year of added muscle. Areta and colleagues distributed 80 g of whey across twelve hours as 8×10 g, 4×20 g, or 2×40 g after resistance exercise; the 20 g every three hours condition produced the highest integrated myofibrillar MPS (Areta et al., 2013). Again: hours, not months.

07 Leucine, meals, anabolic resistance

Leucine is the indispensable amino acid most clearly implicated in activating mTORC1 in muscle. Churchward-Venne and colleagues showed that adding leucine to a low-protein mixed drink could raise myofibrillar MPS toward the level of a larger protein dose in young men (Churchward-Venne et al., 2014). In older adults, Katsanos and colleagues found that a higher leucine proportion was required to stimulate MPS (Katsanos et al., 2006). Cuthbertson and colleagues described the underlying physiology as anabolic resistance: older muscle showed smaller MPS and anabolic-signalling responses to the same amino-acid dose (Cuthbertson et al., 2005). Drummond and colleagues reported that the anabolic response to exercise plus essential amino acids is delayed in older muscle (Drummond et al., 2008). Moore and colleagues quantified the meal-level consequence: the protein intake required to maximise myofibrillar MPS was about 0.24 g·kg⁻¹ in young men and about 0.40 g·kg⁻¹ in older men (Moore et al., 2015).

From those acute data a per-meal heuristic was built. Paddon-Jones argued for distributing protein across meals rather than concentrating it at dinner (Paddon-Jones et al., 2004). Loenneke and colleagues, in a cross-sectional analysis, associated more frequent meals containing 30–45 g of protein with greater lean mass and strength (Loenneke et al., 2016). Schoenfeld and Aragon reviewed the single-meal ceiling argument and concluded that a per-meal dose on the order of 0.4 g·kg⁻¹, repeated across the day, is a defensible reading of the acute literature if the daily target is near 1.6 g·kg⁻¹ (Schoenfeld and Aragon, 2018). Kim and colleagues later reported that, in older adults, daily distribution pattern did not change the anabolic response or lean-mass outcome when total protein was adequate (Kim et al., 2018). The distribution claim is therefore emerging: biologically tidy, supported by some acute and observational work, not confirmed as a necessary condition once the daily total is high enough.

08 Dieting and body composition

Energy deficit is the setting in which higher protein has its clearest human composition evidence. Wycherley and colleagues meta-analysed energy-restricted high-protein, low-fat diets against standard-protein, low-fat diets. The higher-protein arms lost more weight and more fat mass, retained more fat-free mass, and reduced triglycerides more (Wycherley et al., 2012). Krieger and colleagues, in an earlier meta-regression of diet composition, likewise found that higher protein (and lower carbohydrate) associated with more fat-free-mass retention during weight loss (Krieger et al., 2006). Santesso and colleagues’ systematic review of higher- versus lower-protein diets on health outcomes found modest improvements in weight, BMI, and some cardiometabolic markers, with heterogeneity (Santesso et al., 2012). These are human trial syntheses. They are strongly supported for short-to-medium-term composition under energy restriction. They are not demonstrations that a high-protein diet outperforms a calorie-matched Mediterranean pattern on ten-year coronary events.

Longland and colleagues ran the cleanest single trial on the interaction with training: during a marked energy deficit with intense exercise, 2.4 g·kg⁻¹·d⁻¹ produced lean-mass gain and greater fat loss, whereas 1.2 g·kg⁻¹·d⁻¹ produced lean-mass loss (Longland et al., 2016). Skov and colleagues, in an ad libitum fat-reduced trial, found that a higher-protein condition produced greater weight loss than a higher-carbohydrate comparison — an intake effect mediated partly by spontaneous energy reduction (Skov et al., 1999). Bray and colleagues reversed the sign of energy: in supervised overfeeding, protein content determined the change in lean mass more than the change in fat mass; fat gain tracked energy surplus (Bray et al., 2012). The pair of results is the energy-context rule in experimental form. Protein is not a calorie-free anabolic. In surplus it builds lean tissue and does not prevent fat gain. In deficit, with training, it is one of the few dietary levers that consistently changes the composition of the loss.

Once the RDA is recognised as a nitrogen-balance floor (section 02), the remaining question is what “optimal” would even mean. Optimal for zero nitrogen is not optimal for fat-free mass during a cut, not optimal for gait speed at eighty, and not optimal for a powerlifter’s recovery. Wolfe reviewed dietary protein as a determinant of muscle mass and function and argued that intakes above the RDA are justified when the endpoint is muscle, particularly in older adults (Wolfe, 2012). Phillips made the same argument from the training literature (Phillips, 2016). Neither paper replaces the DRI process. Both papers refuse to treat 0.80 g·kg⁻¹ as a target rather than a minimum.

An upper safe intake has not been set as a DRI tolerable upper intake level. The AMDR cap of 35% of energy is the nearest institutional ceiling (Trumbo et al., 2002). Antonio and colleagues fed resistance-trained men 3.4 g·kg⁻¹·d⁻¹ in one trial and, in a one-year crossover, very high protein intakes with no adverse change in the clinical chemistry they measured, including renal and hepatic panels (Antonio et al., 2015, 2016). Those are small, self-selected, trained, mostly male samples, with food logs rather than metabolic-ward control. They are evidence against an immediate toxicology signal at intakes far above the AMDR when expressed in g·kg⁻¹. They are not a licence, and they are not a longevity study. Cuenca-Sánchez and colleagues reviewed the satiety, kidney, and bone controversies and concluded that the popular harms are overstated in healthy people and under-specified as to source and duration (Cuenca-Sánchez et al., 2015). That review is a map of the arguments, not a new trial.

10 Older adults and sarcopenia

Sarcopenia is now a defined syndrome — low muscle strength, with low muscle quantity or quality, and poor physical performance as a severity marker — not a synonym for aging (Cruz-Jentoft et al., 2019). The Health ABC cohort found that older community-dwelling adults in the highest quintile of energy-adjusted protein intake lost less lean mass over three years than those in the lowest quintile (Houston et al., 2008). That is prospective observational evidence. It is strongly supported as an association. It is not a randomised demonstration that raising protein prevents sarcopenia, and it cannot fully separate protein from the dietary pattern that carried it.

The PROT-AGE study group recommended 1.0–1.2 g·kg⁻¹·d⁻¹ for healthy older adults, 1.2–1.5 g·kg⁻¹ in illness or malnutrition, and higher still in severe illness or injury, with per-meal protein of 25–30 g (Bauer et al., 2013). The ESPEN Expert Group made a congruent recommendation and added resistance exercise as a co-prescription for function (Deutz et al., 2014). Gregorio and colleagues reported that among post-menopausal women, protein intakes closer to 1.2 g·kg⁻¹ associated with better physical performance than lower intakes (Gregorio et al., 2014). These documents are expert syntheses, not new RCTs. They exist because the RDA sample was not older muscle with anabolic resistance (Cuthbertson et al., 2005; Moore et al., 2015). The gap between 0.8 and 1.2 g·kg⁻¹ in older adults is the most institutionally recognised “beyond RDA” claim in clinical nutrition. It remains a recommendation from societies, not a DRI revision.

Aging and sarcopenia evidence table. Position papers are not trials. Cohort associations are not causal demonstrations.

11 Athletes and the upper range

Athletes are not a separate species of nitrogen metabolism. They are people who impose larger contractile and energetic loads and who often restrict energy while trying to hold lean mass. The ISSN range of 1.4–2.0 g·kg⁻¹ (Jäger et al., 2017) and the Morton breakpoint near 1.6 g·kg⁻¹ for fat-free-mass accretion (Morton et al., 2018) are the two most cited quantitative summaries. Helms’s review of lean, trained athletes in a deficit is the document that pushes toward the top of the range (Helms et al., 2014). Antonio’s very-high-protein series is the empirical ceiling-test in trained men (Antonio et al., 2015, 2016). None of these papers measured championship performance as a primary endpoint. “Recovery” in this literature usually means overnight MPS, next-day function, or fat-free mass, not a race time.

Sports and training matrix. Acute MPS rows are not interchangeable with fat-free-mass rows.

12 Bone is not dissolved by steak

The acid-ash hypothesis held that high-protein diets, especially from animal foods, generate a metabolic acid load that is buffered by bone, wasting calcium in urine and, over years, bone. Urinary calcium does rise with protein. Kerstetter and colleagues showed that the rise is accompanied by increased intestinal calcium absorption, so the net calcium economy is not a simple drain (Kerstetter et al., 2003, 2011). Darling and colleagues’ systematic review and meta-analysis found little evidence that higher protein harms bone density or fracture, and some evidence of benefit at the lumbar spine (Darling et al., 2009). Shams-White and colleagues, for the National Osteoporosis Foundation, likewise found that higher protein was associated with higher bone mineral density and a possible reduction in hip fracture risk, with the usual observational caveats (Shams-White et al., 2017). Fenton and colleagues’ systematic causal assessment of dietary acid load and bone disease did not support the acid-ash account of osteoporosis (Fenton et al., 2011). Calvez and colleagues reviewed protein, calcium balance, and health consequences and did not find a coherent skeletal indictment of higher protein in healthy adults (Calvez et al., 2012).

The bone claim against high-protein diets is therefore not established as a chronic harm in healthy adults. The remaining uncertainty is the usual one: very high protein with inadequate calcium, or protein as a marker for a dietary pattern that is also low in fruit, vegetables, and dairy, is a different exposure from protein in a mixed diet that meets calcium and alkali needs. That is source and pattern again, not grams.

A protein meal raises glomerular filtration rate. That is haemodynamic hyperfiltration, visible in human feeding trials, including the OmniHeart protein arm (Juraschek et al., 2013). Hyperfiltration is not the same event as a falling estimated GFR over years, and it is not the same event as a hard renal endpoint. Martin, Armstrong and Rodriguez reviewed dietary protein and renal function and did not find evidence that high-protein diets damage healthy kidneys (Martin et al., 2005). Poortmans and Dellalieux asked the same question in athletes and did not find a renal-risk signal from regular high-protein diets in that population (Poortmans and Dellalieux, 2000). Friedman reviewed potential kidney effects and distinguished healthy kidneys from diseased ones (Friedman, 2004).

Knight and colleagues, in the Nurses’ Health Study, found that higher protein intake was associated with faster GFR decline only in women who already had mild renal insufficiency; women with normal baseline function did not show that association (Knight et al., 2003). Devries and colleagues meta-analysed randomised comparisons of higher- versus lower-protein diets in healthy adults and found that changes in GFR did not differ (Devries et al., 2018). Antonio’s one-year high-protein crossover in trained men reported no harmful change in the renal panel they measured (Antonio et al., 2016). The healthy-kidney harm claim is therefore not established. The remaining honest reservation is duration and endpoint: few trials last a decade, and few measure hard renal outcomes rather than creatinine-based GFR.

14 Chronic kidney disease is a different exposure

Once nephron number is reduced, protein load is no longer a gym controversy. The Modification of Diet in Renal Disease trial tested protein restriction and blood-pressure control on CKD progression; the results were mixed and modest on GFR slope, and they have been re-analysed ever since (Klahr et al., 1994; Levey et al., 1999). Cochrane reviews of low-protein diets in non-diabetic CKD have reported uncertain effects on death and a possible delay of end-stage disease, with low-to-moderate certainty and adherence problems (Fouque and Laville, 2009; Hahn et al., 2018, 2020). The KDOQI 2020 nutrition guideline recommends protein restriction in metabolically stable adults with CKD 3–5 not on dialysis — 0.55–0.60 g·kg⁻¹·d⁻¹, or a very-low-protein diet plus ketoacid analogues under specialised care — and higher protein once dialysis begins (Ikizler et al., 2020). Those numbers are guideline parameters, not a lifestyle prescription for people without CKD.

The analytical error this section exists to prevent is the transfer of the CKD restriction into the healthy-adult argument, and the reverse transfer of gym safety data into the nephrology clinic. They are not the same patients, not the same residual nephron mass, and not the same endpoints.

Renal evidence table. Hyperfiltration, eGFR slope, and dialysis are three different outcomes.

15 Liver and nitrogen disposal

The liver is the organ that converts leftover amino-nitrogen to urea. A high-protein meal increases ureagenesis; that is physiology, not hepatitis. In people with established cirrhosis and a history of hepatic encephalopathy, protein handling is a clinical problem of a different order, and this article does not treat that problem as a general-population finding. In healthy adults, including Antonio’s very-high-protein trained men, routine hepatic enzymes have not announced a toxicology signal (Antonio et al., 2016). Absence of a signal in small athletic samples is not a proof of lifelong hepatic indifference. It is the evidence that exists.

16 Glucose metabolism

Protein is insulinotropic and, in people with type 2 diabetes, can flatten postprandial glucose when it replaces carbohydrate. Gannon and Nuttall reported that raising protein and lowering carbohydrate improved 24-hour glucose in human feeding studies of type 2 diabetes (Gannon and Nuttall, 2003). OmniHeart’s protein arm, relative to a high-carbohydrate arm, improved blood pressure and some lipid measures; insulin-sensitivity results in later OmniHeart analyses were not a simple protein victory (Appel et al., 2005; Furtado et al., 2008; Gadgil et al., 2013). The substitution rule is mandatory: protein that replaces refined carbohydrate is not protein that replaces unsaturated fat, and neither is protein added on top of an unchanged diet. Claims that high-protein diets “cause insulin resistance” in healthy humans are not established by the controlled-feeding literature at ordinary elevated intakes. Claims that they are a diabetes therapy are also not established beyond short feeding studies and weight-loss trials in which energy deficit does most of the work.

17 Cardiovascular outcomes

Randomised evidence that isolates protein quantity and follows hard cardiovascular events does not exist at meaningful duration. What exists is substitution feeding and source epidemiology. OmniHeart found that replacing carbohydrate with protein (or with monounsaturated fat) lowered blood pressure and improved lipids relative to a high-carbohydrate DASH-like diet (Appel et al., 2005; Furtado et al., 2008). Santesso’s higher- versus lower-protein systematic review found modest cardiometabolic movement without a hard-event result (Santesso et al., 2012). Bernstein and colleagues, in the Nurses’ Health Study, reported that the protein source predicted coronary disease: higher red-meat protein associated with more CHD, higher nut, fish, and poultry protein with less, when substitution was modelled (Bernstein et al., 2010). Zhong and colleagues associated processed meat and unprocessed red meat, but not poultry or fish, with incident cardiovascular disease in a large US cohort (Zhong et al., 2020).

Low-carbohydrate scores that are high in animal protein and animal fat have a different observational record from plant-forward low-carbohydrate scores. Halton and colleagues found that a low-carbohydrate-diet score was not associated with higher CHD in women when vegetable sources were emphasised (Halton et al., 2006). Seidelmann and colleagues reported a U-shaped association of carbohydrate intake with mortality, with both very low and very high carbohydrate associated with higher death, and with plant-derived low-carbohydrate patterns faring better than animal-derived ones (Seidelmann et al., 2018). The PURE study reported that higher carbohydrate and lower fat associated with higher mortality across 18 countries — a result that is not a protein trial and that remains contested on residual confounding and diet-assessment grounds (Dehghan et al., 2017). None of these papers is a randomised test of 1.6 versus 0.8 g·kg⁻¹ on myocardial infarction.

18 Mortality

Song and colleagues, in two large US cohorts, found that higher animal-protein intake associated with higher cardiovascular mortality and that higher plant-protein intake associated with lower all-cause and cardiovascular mortality; substitution models drove the interpretation (Song et al., 2016). Naghshi and colleagues’ dose–response meta-analysis of prospective studies found that higher total protein and higher plant protein associated with lower all-cause mortality, while animal protein was not significantly associated with mortality in the main analysis (Naghshi et al., 2020). Chen and colleagues, in the Rotterdam Study, reported source-specific associations that again refused to treat “protein” as one exposure (Chen et al., 2020). Levine and colleagues reported that, among adults aged 50–65 in NHANES, higher protein associated with higher IGF-1 and with higher cancer and overall mortality, an association that inverted after age 65 (Levine et al., 2014). That paper is the most cited exhibit for a high-protein longevity harm. It is observational, age-split, and entangled with source; its mouse experiments are mouse experiments. It is emerging as a hypothesis generator and not established as a causal human aging result.

Every number in the first four parts was a quantity of protein or of an amino acid. A person does not eat a quantity of protein. They eat beef, milk, eggs, lentils, a whey shake, or a slice of ham, and the protein arrives inside a matrix that the gram does not describe. Two foods matched to the same grams of protein can differ in amino-acid pattern, in digestibility, and in everything else they carry — the heme iron and saturated fat of red meat, the curing salts of a processed product, the fibre and isoflavones of a legume. The food carries a matrix; the gram does not. Source is therefore a fifth axis, independent of the four units of intake, and it is the axis on which most of the chronic-disease argument actually turns.

The amino-acid axis is measurable. Animal proteins supply the nine indispensable amino acids in proportions close to human requirement and are highly digestible; most single plant proteins are lower in one or more indispensable amino acids — lysine in cereals, methionine in legumes — and their digestibility is reduced by fibre and antinutrients. That difference is visible in the acute muscle-protein-synthesis literature that Part Two graded. Whey, a fast leucine-rich dairy fraction, is absorbed more rapidly than casein and drives a sharper postprandial rise in whole-body protein synthesis (Boirie et al., 1997); at a matched dose it raises myofibrillar synthesis more than casein or soy (Tang et al., 2009), and fluid skim milk supported greater muscle protein accretion than a soy drink after resistance exercise (Wilkinson et al., 2007). Katsanos and colleagues showed that a high proportion of leucine is required to drive synthesis in older muscle (Katsanos et al., 2006). Those are strongly supported acute mechanistic facts. They are not a ranking of foods for health, and the leucine advantage of whey does not travel to a cardiovascular endpoint.

20 Source-specific epidemiology

When protein is followed as a food rather than as a nutrient, the associations separate by source in a consistent direction. In the Nurses’ Health Study, Bernstein and colleagues modelled substitution and found that protein from red meat associated with more coronary heart disease, while protein from nuts, fish, and poultry associated with less (Bernstein et al., 2010). Pan and colleagues, across two large US cohorts, associated red-meat intake — processed more strongly than unprocessed — with higher total mortality (Pan et al., 2012), and Zheng and colleagues later reported that increasing red-meat consumption over time associated with higher subsequent mortality (Zheng et al., 2019). Zhong and colleagues separated the sources directly: processed meat and unprocessed red meat, but not poultry or fish, associated with incident cardiovascular disease and death (Zhong et al., 2020). Eggs, treated as their own exposure across three cohorts, were not associated with cardiovascular risk at moderate intake (Drouin-Chartier et al., 2020).

The plant side of the same ledger runs the other way, though Part Four’s mortality section already recorded the endpoint: higher plant-protein intake tracked lower all-cause and cardiovascular mortality across the same cohorts and meta-analysis that found animal protein flat or adverse (Song et al., 2016; Naghshi et al., 2020). What those studies leave for a source section is the non-mortality, food-specific signal. Soy has its own controlled record: it modestly lowers LDL cholesterol in the FDA-reviewed trial set (Blanco Mejia et al., 2019) and carries no demonstrated adult harm in a large clinical and epidemiologic review (Messina, 2016). On cancer, Farvid and colleagues associated red and processed meat — not protein in the abstract — with higher breast-cancer incidence (Farvid et al., 2018). The signal keeps attaching to the food.

Two cautions hold this section together. First, every one of these is an observational substitution model; the effect is the estimated consequence of exchanging one source for another at constant energy, and residual confounding by fibre, processing, and the health behaviours that travel with meat avoidance is not removed by adjustment. Second, none of these studies isolates protein quantity. They are the strongest evidence that source matters and among the weakest possible evidence about whether 1.6 versus 0.8 g·kg⁻¹ matters when source is held constant.

Source epidemiology matrix. Every row is a substitution association at constant energy, not a test of protein quantity.

21 IGF-1, mTOR, and the longevity argument

The strongest theoretical case against high protein is not renal and not cardiovascular. It is the growth-signalling argument: dietary protein, and leucine in particular, activates the mechanistic target of rapamycin (mTOR) and raises insulin-like growth factor 1 (IGF-1), and reduced nutrient-sensing signalling is one of the more reproducible ways to extend lifespan in model organisms. Saxton and Sabatini’s review sets out mTOR as the central controller of growth and metabolism (Saxton and Sabatini, 2017). The inference under examination is that a pathway which promotes growth must, chronically driven, shorten life.

Two results are used to carry that inference into humans, and neither closes it. Levine and colleagues reported that, among NHANES adults aged 50–65, higher protein intake associated with higher IGF-1 and with higher cancer and overall mortality — an association that reversed after age 65, where higher protein tracked lower mortality (Levine et al., 2014). Part Four already graded that paper as observational, age-split, and entangled with animal source; what it contributes to the mechanism argument is the mouse tumour-xenograft work in the same report, which shows IGF-1-dependent tumour growth in animals, not shortened life in people. A mouse study published in the same issue, Solon-Biet and colleagues, found that low-protein high-carbohydrate diets extended median lifespan and improved cardiometabolic markers in mice fed ad libitum — a result about macronutrient ratio in a caged rodent, not a human protein prescription (Solon-Biet et al., 2014).

The honest reading distinguishes three claims that the slogan fuses. That protein raises mTOR and IGF-1 acutely is established biochemistry. That lower lifelong nutrient-sensing signalling extends lifespan in mice and other short-lived models is strongly supported in those organisms. That eating more protein shortens human life is emerging at most — a hypothesis generated by age-stratified observational data and animal work, confounded by source, and directly opposed by the same literature’s finding that higher protein protects older adults. A kinase that builds muscle in a seventy-year-old is being read as a death certificate on the strength of a mouse and an age-stratified survey. The reading does not hold.

The same fusion drives the high-protein low-carbohydrate diet, where raising protein cannot be separated from cutting carbohydrate. As a short-horizon tool the record is real: Skov and colleagues found that a higher-protein ad libitum diet produced more weight loss than a higher-carbohydrate one at matched fat (Skov et al., 1999), and Wycherley and colleagues’ meta-analysis of energy-restricted high-protein low-fat diets found modestly greater loss of weight and fat with better retention of lean mass (Wycherley et al., 2012). What that record cannot settle is the long horizon, because a low-carbohydrate pattern is defined by its replacement — Part Four found that animal-based and plant-based low-carbohydrate scores diverge, the plant-forward version faring better. It is a weight and glycaemic instrument with a short evidence horizon, not a demonstrated longevity strategy in either direction.

22 The strongest case, and the strongest caution

Put the two kinds of evidence side by side. The strongest case for benefit is short-to-medium-term and mechanistically clean: under an energy deficit, higher protein preserves lean mass and favours fat loss (Longland et al., 2016; Wycherley et al., 2012), and combined with resistance training it produces measurably greater gains in lean mass and strength up to roughly 1.6 g·kg⁻¹·d⁻¹, above which the average marginal return flattens (Morton et al., 2018). In older adults, who face anabolic resistance and sarcopenia, intakes above the RDA are associated with better lean mass and physical function, and expert groups converge on roughly 1.0–1.2 g·kg⁻¹·d⁻¹ as a floor for that population (Bauer et al., 2013; Deutz et al., 2014; Phillips et al., 2016). None of this requires a supplement or a particular source; it requires enough total protein, distributed across meals, in someone who trains.

The strongest case for caution is long-horizon and source-specific. It is not a demonstrated renal toxin in healthy kidneys, which Part Four graded not established (Devries et al., 2018), and it is not a proven human longevity harm from protein quantity. It is the consistent observational association of processed and unprocessed red meat with cardiovascular disease, some cancers, and mortality (Pan et al., 2012; Zhong et al., 2020; Farvid et al., 2018), set against the inverse association of plant protein (Song et al., 2016; Naghshi et al., 2020) — and, separately and firmly, the guideline restriction of protein load in established chronic kidney disease (Ikizler et al., 2020). The caution attaches to the food and to the patient, not to the gram.

Peer-reviewed identifiers were verified against NCBI PubMed records, not reproduced from memory. In-prose citations are author–year; the numbered list is sorted by first-author surname. The study design is named in the sentence that reports each finding. Animal and cell results are not restated as human outcomes, and the mouse macronutrient-geometry and tumour-xenograft experiments are labelled as such wherever the longevity argument leans on them. Epidemiological associations are reported as substitution models at constant energy, not as demonstrated causes, and residual confounding by processing, fibre, and the health behaviours that travel with meat avoidance is named rather than adjusted away. Guideline text — the KDOQI renal recommendation and the Institute of Medicine reference intakes — is quoted as recommendation language, not as trial evidence. Where two exposures share a name — protein the nutrient and the food that carries it, the healthy kidney and the diseased one, acute synthesis and lifespan — they are kept apart. Only peer-reviewed literature and named guideline and agency sources are cited; general reference databases were used to locate that literature, never as evidence in their own right.

24 Scope relative to sibling articles

This is the high-protein-diet article in the SBL-41 series. It is not the amino-acid nutrition framework article, which orients requirements at the level of individual amino acids, and it is not the protein-supplements article, which treats isolated powders as products. It is not a sports-nutrition fuelling guide, a weight-loss-diet manual, or a clinical nutrition-support protocol. Chronic kidney disease appears only to forbid the transfer of its protein restriction onto healthy adults, and the reverse transfer of athletic-safety data into the nephrology clinic; this document is not a nephrology reference.

25 References

99 peer-reviewed records below were verified against NCBI. In-prose citations use author–year; where a first author appears twice in the same year, the text distinguishes the two by subject.