The energy-balance protein target (~1.6 g/kgBM/day, Protein and Resistance Training for Muscle and Strength) answers how much protein to BUILD muscle. This page answers a different decision in a different stratum: how much protein to RETAIN fat-free mass while in a caloric deficit — the cutting athlete, the dieter, the person on a GLP-1 drug. The single dedicated source is Refalo 2025, an updated systematic review with Bayesian meta-regression (29 studies; nonobese, energy-restricted, resistance-trained adults) (Refalo et al., 2025). Two things bound everything below: the authors call their own findings exploratory (not confirmatory), and the effect of protein is small next to the resistance-training stimulus.
A second, independent source now backs the core direction from a different design and a different
population: Wycherley 2012, a gold meta-analysis of 24 isocaloric high-protein-vs-standard-protein
RCTs (1063 overweight/obese adults, structured exercise EXCLUDED) (Wycherley et al., 2012).
Where Refalo is a dose-response meta-regression in resistance-trained non-obese dieters, Wycherley is
a direct HP-vs-SP contrast MA in the sedentary overweight/obese dieter — so the page now spans two
strata, and the FFM-preservation direction is reached by two non-overlapping labs (see the Wycherley
section and its parameter table below). This is why confidence: is medium, not low — but the
support is for the direction, not for any specific target number.
A third gold source extends the AGE stratum: Kim 2016, a SR+MA of 20 HP-vs-normal-protein RCTs in adults aged 50 and older (Kim et al., 2016). It closes the older-adult gap Wycherley flagged, and finds the lean-mass-sparing magnitude is essentially identical in older adults to Wycherley’s all-ages pool — so the contrast benefit does not appear age-modified at this modest dose. Kim is NOT an independent (type-E) leg, though: it explicitly benchmarks against Wycherley and shares trials with it (see the Kim section below).
The target RISES under a deficit — but read it as a region, and the caveats first
(Refalo et al., 2025) Refalo’s exploratory recommendations (their Table 3, their word): «increasing protein intake up to 1.9 g/kgBM/day or 2.5 g/kgFFM/day, on average, is associated with less FFM loss in the present analysis (representing the protein intake levels at which the trend line of our meta-regression crosses an ES of zero). Furthermore, intakes above these values, up to the highest analyzed intakes of 3.2 g/kgBM/day and 4.2 g/kgFFM/day, are linearly associated with larger FFM gain».
- ~1.9 g/kgBM (or ~2.5 g/kgFFM) is the intake at which the trend line crosses ES = 0 — i.e. where average FFM change flips from loss to hold. It is HIGHER than the ~1.6 g/kgBM energy-balance target, which matches the mechanistic premise that a deficit raises the requirement.
- These are not point optima — they are ES-zero crossings on a monotone line with no observed peak, so above them more protein is (linearly) associated with FFM gain up to the highest analyzed 3.2/4.2. Read as a floor-and-direction, not a target -> The Underivable Optimum.
- The whole thing is on a caloric budget: more protein necessarily displaces carbohydrate or fat, with potential costs to RT performance, sex-hormone levels, and adherence — weighed at Layer 3, not netted here.
The shape: LINEAR, no knee — consistent with Morton on no significant knee only, across two populations
(Refalo et al., 2025) Bayes-Factor model comparison found the linear model (no additional predictors) beat quadratic and cubic: «there was a 97% probability that FFM change becomes more positive as daily protein intake (g/kgBW) increases» (per BM, b=0.07, 95% HDI -0.01 to 0.14), rising to 99% per FFM (below). No knee, no plateau over the analyzed range.
This is a second, same-lineage instance of a no-significant-knee protein curve (Helms co-authors both papers, and Refalo cites Morton). The parameter table below shows the shared ground is narrow — no statistically-significant knee — and NOT a 1.6-vs-1.9, deeper-is-more contrast — the two numbers describe different populations (energy balance vs deficit), so they are a stratum distinction, not a tension.
| Parameter | Morton 2018 — quoted + locus | Refalo 2025 — quoted + locus | Same quantity? |
|---|---|---|---|
| Population | «healthy and not energy-restricted» RT adults (energy balance/surplus) [Morton chunk 01] | nonobese, ENERGY-RESTRICTED (fat-mass loss) RT adults, BF% <=27.8%M/<=39.7%F [Refalo chunk 01] | NO — opposite energy state |
| Curve shape | biphasic knee «1.62 g/kg/day, p=0.079», non-significant; on baseline axis «linear regressions explained significantly more variance than biphasic» [Morton chunk 02] | «linear model, without additional predictors, provided the best fit»; «97% probability that FFM change becomes more positive» [Refalo chunk 01] | YES only on no statistically-significant knee — on the total-intake axis Morton’s biphasic fit beat linear (p=0.079, ns) whereas Refalo’s linear won; Morton’s linear-wins result is a different regression (baseline intake; young-adult slope −1.54) |
| Denominator tested | total body mass only (per-FFM NOT tested) | per-FFM «99 versus 97%» prob; «entirety of the 95% HDI was positive» (0.01-0.12 vs -0.01-0.14) [Refalo chunk 01] | NO — Morton did not compare denominators |
| Target number | ~1.6 g/kgBM plateau (energy balance) | ~1.9 g/kgBM / 2.5 g/kgFFM ES-zero crossing (deficit) | NO — different construct AND population |
- The SHAPE row is same-lineage consistency, NOT type-E — Morton (energy balance) and Refalo (energy deficit) share only no statistically-significant knee, and they are not independent: Eric Helms co-authors both, and Refalo cites Morton (its ref 39). Read it as refinement/attribution within one research lineage. (corrected 2026-10-06: no shared author + the E-independent token -> same-lineage, not E; self-critique)
- The per-FFM denominator point is NOT independent. Refalo’s per-FFM finding rests on Helms 2014’s speculation — «it may be worthwhile to prescribe protein intake based on FFM versus total body mass in athletic populations» — and Helms is a co-author of Refalo. So per-FFM-is-better is one lineage’s idea tested by that lineage, not a cross-lab convergence. Refalo is likewise NOT independent of Helms 2014 generally (it is an update of it, same author). (inferred from Refalo et al., 2025)
Why per-FFM is the better denominator here
(Refalo et al., 2025) Only the per-FFM slope’s 95% HDI excludes zero (0.01 to 0.12), while the per-BM slope’s includes it (-0.01 to 0.14); direction probability 99% vs 97%. Scaling to FFM avoids a target «too low for lean athletes» (Helms’ rationale). Decision consequence: where FFM can be reasonably estimated, prefer the FFM-based target (2.5 g/kgFFM up to 4.2 for gain). This also means the number is undefined until the denominator is named — the same trap as the energy-balance page’s g/kg-total ambiguity, sharper here because per-FFM is the recommended scaling.
Who benefits more per gram — the BF% ordering (leaner may benefit more; suggestive, not an established moderator)
(Refalo et al., 2025) «for each additional gram of protein consumed per kilogram of BM and FFM, FFM change increased by 0.14 and 0.12 in the Low BF% category, 0.10 and 0.07 in the Medium BF% category, and 0.04 and 0.03 in the High BF% category». Mechanism (Elia’s early-starvation review): «lean individuals have 2-fold to 3-fold higher rates of urinary nitrogen excretion, leucine oxidation, and contributions of protein to glucose production compared with individuals with obesity» — leaner people burn more body protein for energy under restriction, so protein may defend them more. Protein may therefore inversely scale with BF% (suggestive; see the moderator caveat below).
- This is a REASONING TOOL toward higher-BF people, NOT obese evidence. The whole sample is nonobese (BF% <=27.8%M/<=39.7%F). The BF% gradient and the Elia mechanism point to a lower per-gram benefit as adiposity rises, but the obese are off-support — do not read this as the obese need less protein, read it as the obese are unstudied here -> same gap as Protein and Resistance Training for Muscle and Strength.
- The Low-BF% cell has only 10 observations, so its slope is the most uncertain. More broadly, the BF% gradient is a suggestive ordering, not an established moderator: every BF% subgroup slope’s HDI includes zero (Table 2), «neither model fit was improved by including addi- tional predictors such as intervention duration, energy deficit magnitude, baseline BF%, and participant sex», and the authors call the relationship «somewhat tentative». (corrected 2026-10-06: the direction is firmer than the magnitude -> suggestive ordering; self-critique)
Sex, duration, deficit magnitude
- Sex (tentative): «the probability of a linear dose-response was stronger for male participants than female participants, with the difference more pronounced with protein expressed relative to FFM (99 versus 69%) than BM (98 versus 81%)». Far fewer female observations — hold as a signal, not a finding.
- Duration: >4 weeks stronger than <=4 weeks — short interventions are confounded by fluid shifts (which also feed the FFM-is-not-muscle caveat below).
- Deficit magnitude did NOT clearly moderate the protein->FFM slope — «The relationship between dietary pro- tein intake and FFM changes were similar between smaller (#300 kcal/ d) and larger (.300 kcal/d) energy deficits.» (a median split; the authors note a <=500/>500 cut might differ). That is a statement about the protein dose-response, not that deficit size leaves FFM unaffected (corrected 2026-10-06: deficit did not moderate -> did not moderate the protein slope, Refalo chunk 01). BUT «Murphy and colleagues recently suggested that an energy deficit of ;500 kcal per day prevented gains in FFM» — so «individuals seeking to avoid FFM losses at all costs should likely implement modest energy restriction». A large deficit is itself likely a lever on FFM independent of protein.
The load-bearing caveats — do not ship a clean number
- EXPLORATORY, not confirmatory — the authors’ own verdict: «the heterogeneity between studies included in our meta-regression renders the findings exploratory rather than confirmatory» (study-level I2 83-85%). It is a meta-regression of mostly non-controlled dose-response, not a meta-analysis of higher-vs-lower RCTs (only 5 such comparisons existed — «insufficient»). Treat the numbers as hypotheses for practice, not established targets.
- The effect is SMALL vs the RT stimulus. Lowest-to-highest analyzed intake is «only a “small” ES difference»; a robust RT stimulus is likely «more important for retaining FFM during energy restriction than dietary protein intake» (the sentence names RT only; Refalo’s «likely» is split from the span by a table insert in the chunk). Deficit size enters via Helms’ speculation, reported by Refalo, that «energy deficit magnitude may be more important for FFM retention than pro- tein intake», and Refalo’s closing position is that which lever matters most «is likely context-dependent». (corrected 2026-10-06: RT + deficit stated flatly as more important -> RT likely, deficit via Helms, ranking context-dependent; self-critique) Worked case: «Longland and colleagues (30) observed no mean loss of lean body mass (+0.1 6 1.0 kg)» at just 1.2 g/kgBM with hard RT, while the 2.4-g/kgBM comparator «gained significantly more lean body mass (+1.2 6 1.0 kg), the difference (1.1 kg) was modest». Get the training right first; protein is the adjunct -> Protein and Resistance Training for Muscle and Strength.
- FFM is a SURROGATE, and FFM != skeletal muscle — «changes in FFM do not directly equate to changes in skeletal muscle mass and may be affected by fluid alterations» -> Surrogate Outcomes. No patient-important endpoint (function, strength, mortality) is measured. Same boundary as the energy-balance protein page and GLP-1 and Lean Mass.
- Single data extractor (M.C.R.) — «may increase the potential for error in data extraction».
- Self-reported intake — the accuracy of the reported protein intakes is «in question» -> Measurement Error in Dietary Assessment (the same instrument problem flattens dose-response and smears any threshold).
- Mild COI — the authors are tied to the evidence-based-fitness industry (Renaissance Periodization, MASS Research Review, 3D Muscle Journey); the method is rigorous, but noted.
The independent contrast MA (Wycherley 2012) — same direction, different design, different stratum
(Wycherley et al., 2012) Wycherley pooled 24 RCTs that each randomized energy-restricted, ISOCALORIC HP vs SP low-fat diets (HP ~1.25 vs SP ~0.72 g/kg/d achieved; both <=30% fat). Pooled weighted mean differences (HP minus SP; negative = greater effect with HP):
- FFM: +0.43 kg (95% CI 0.09, 0.78) — HP spares fat-free mass. The direction Refalo reaches by a monotone regression slope, Wycherley reaches by a between-arm contrast.
- Fat mass: -0.87 kg (95% CI -1.26, -0.48); body weight: -0.79 kg (95% CI -1.50, -0.08).
- Triglycerides: -0.23 mmol/L (95% CI -0.33, -0.12), homogeneous — but total/LDL/HDL cholesterol, systolic + diastolic BP, fasting glucose and insulin were all NULL (P >= 0.20). The cardiometabolic benefit is confined to a surrogate (TG), and «whether triglyceride reductions translate to reduced cardiovascular disease risk remains unclear» -> Surrogate Outcomes.
- REE: +595.5 kJ/d (95% CI 67.0, 1124.1) — but 4 studies only, mean 6 wk, and a tissue-model estimate is ~24 kJ/d; treat as a mechanism hint, not a firm effect.
The authors’ own verdict is «modest benefits» — the magnitudes are small, reinforcing this page’s standing theme that protein is the adjunct, not the main lever, for FFM under a deficit.
Type-E: the FFM-preservation direction is reached by two non-overlapping labs across two populations
Wycherley (Adelaide/CSIRO — Wycherley/Moran/Clifton/Noakes/Brinkworth) shares no author with Refalo
(Refalo/Trexler/Helms) or Morton/Phillips, and it pools a different, largely non-overlapping trial
set in a different population (sedentary overweight/obese vs resistance-trained non-obese). The
convergence is on the DIRECTION only — the metrics differ, so this is not a pooled magnitude. [E-independent]
| Parameter | Refalo 2025 — quoted + locus | Wycherley 2012 — quoted + locus | Same quantity? |
|---|---|---|---|
| Design | Bayesian meta-regression of 29 dose-response studies [Refalo chunk 01] | meta-analysis of 24 isocaloric HP-vs-SP RCTs (Wycherley et al., 2012) | NO — continuous slope vs two-arm contrast |
| Population | «nonobese», resistance-trained, energy-restricted [Refalo chunk 01] | overweight/obese, structured exercise EXCLUDED (Wycherley et al., 2012) | NO — opposite training state |
| Protein levels | continuous 0.8-3.2 g/kgBM; ES-zero ~1.9 | achieved «1.25 6 0.17» (HP) vs «0.72 6 0.09» (SP) g/kg/d (Wycherley et al., 2012) | NO — continuous vs a low-range contrast |
| FFM direction under deficit | «97% probability that FFM change becomes more positive as daily protein intake (g/kgBW) increases» [Refalo chunk 01] | FFM WMD «0.43 kg; 95% CI: 0.09, 0.78» favouring HP (Wycherley et al., 2012) | YES on DIRECTION only — a positive-slope probability and a kg-WMD are not the same parameter, but both say higher protein preserves FFM under energy restriction |
Only the last row is the E-claim. A third line points the same way inside Wycherley’s own
discussion — Krieger’s 2006 meta-regression found FFM retention «tended to increase with each
successive quartile of dietary protein intake» and «protein intakes .1.05 g·kg21·d21 may improve FFM
retention» (secondary citation, via Wycherley; Krieger not directly held, so not a sources: entry). Krieger is
also a Morton 2018 co-author, so this line is not independent of the Morton lineage.
Type-F: Wycherley broadens the stratum and lowers the demonstrated dose
(inferred from Wycherley et al., 2012) Two refinements the composite makes that neither source alone does:
- Stratum extension. Refalo’s evidence is resistance-trained non-obese only; Wycherley’s is the sedentary overweight/obese dieter with exercise excluded — so FFM-sparing by protein under a deficit is now evidenced in a population where the RT stimulus (likely the bigger lever on this page) is absent. It holds even without training, though the magnitude is small.
- Lower demonstrated dose. Wycherley’s «high protein» arm is only ~1.25 g/kg achieved — BELOW Refalo’s ~1.9 g/kg ES-zero crossing and below the ~1.6 g/kg energy-balance target — yet the FFM/FM benefit is still resolvable. So a modest protein increase (roughly 1.2-1.3 vs 0.7-0.8 g/kg) already buys a measurable, if small, FFM/FM effect; you do not need to reach ~1.9 g/kg to see any benefit. This is consistent with the monotone-no-knee shape: benefit accrues gradually from low intakes, not as a step at a threshold -> The Underivable Optimum.
The measurement-error confound bites the contrast itself
(Wycherley et al., 2012) Wycherley flags that the weight/FM differences «may have occurred because of differences in energy intake that are beyond the sensitivity of food record analysis», and that an HP diet «is further removed than an SP diet from a usual dietary intake it may have more limited food choices that are easier to catalog and report with accuracy». So isocaloric-prescribed is not isocaloric-achieved: a DIFFERENTIAL misreporting (the SP arm covertly eating more, or reporting worse) could inflate the apparent HP advantage. This is the same instrument problem that flattens dose-response elsewhere, here acting on a between-arm contrast -> Measurement Error in Dietary Assessment.
Duration, long-term, and the older-adult gap
- Duration: subgroup analysis (<12 vs >=12 wk) showed no significant between-subgroup difference on any outcome; the FFM benefit reached significance only in the longer-duration studies. Long-term (>=12 mo) efficacy «remains largely unknown» — the one 52-wk trial could not be pooled.
- Clinical-relevance bridge — now grounded in its primary source. The extra ~0.79 kg weight loss is small, but Wycherley invokes DPP to argue it still matters, and the primary DPP mediation analysis (Hamman 2006) grounds the figure directly: «Weight loss was the dominant predictor of reduced diabetes incidence (hazard ratio per 5-kg weight loss 0.42 [95% CI 0.35–0.51]; P <0.0001). For every kilogram of weight loss, there was a 16% reduction in risk, adjusted for changes in diet and activity.» (Hamman et al., 2006) So a protein-driven increment in weight loss maps onto a real weight -> diabetes-incidence dose-response — but the link is associational (in DPP the weight change was not itself randomized, only arm assignment was) and lands on a diagnosis endpoint, not a hard CV/mortality outcome, so it inherits the parent DPP’s surrogate-rung caveat -> Does Weight Loss Reduce Cardiovascular Events, Lifestyle vs Metformin for Diabetes Prevention (where Hamman is fully woven).
- Older adults are unstudied within Wycherley’s own pool (sarcopenia + anabolic resistance flagged as a gap) — now closed by Kim 2016, whose age-restricted MA finds the same FFM-sparing direction and magnitude in adults >=50 (see the Kim section below).
The older-adult stratum (Kim 2016) — same direction, same magnitude, NOT an independent leg
(Kim et al., 2016) Kim 2016 is a gold SR+MA of 20 energy-restricted HP-vs-normal-protein (NP) RCTs, restricted to adults aged 50 and older (mostly overweight/obese, BMI ~30-36; DXA body composition; durations 8-104 wk). It classifies HP two ways — >=25% of energy OR >=1.0 g/kg/d — a modest low-range dose contrast (HP achieved ~1.0-1.5, NP ~0.6-1.0 g/kg/d), close to Wycherley’s. Pooled weighted mean differences (HP minus NP; DXA):
- Lean mass — HP retains MORE: by %energy, WMD +0.45 kg (95% CI 0.20, 0.71; I2 34%); by g/kg/d, WMD +0.83 kg (95% CI 0.47, 1.19; I2 0%). Both exclude zero. The g/kg/d contrast gives the larger, fully homogeneous effect — Kim argues the quantity metric (g/kg/d) is preferable to %energy because it tracks amino-acid availability.
- Fat mass — HP loses MORE: by %energy, WMD -0.57 kg (95% CI -0.98, -0.15; I2 0%), significant; by g/kg/d, WMD -0.53 kg (95% CI -1.08, 0.03; I2 0%), a trend (CI touches zero).
- Body mass — NO difference: %energy WMD -0.54 kg (95% CI -1.30, 0.23); g/kg/d WMD -0.06 kg (95% CI -0.66, 0.53). As in Wycherley, higher protein reshapes composition of the loss, not its total.
- Verdict: «men and women aged 50 years and older better retain lean mass while los- ing fat mass when they consume energy-restricted higher-protein rather than normal-protein diets», and the review «strengthens the sci- entific foundation for older overweight and obese adults to consume protein intakes 1.0 g/kg/d to help preserve lean mass».
Type-F: the stratum extends to older adults with the SAME magnitude — no age modification shown
Kim itself benchmarks its result against Wycherley (its ref 20): «older adults who consumed higher protein diets (expressed as a percentage of energy in-take) during energy restriction preserved about the same amount of lean mass (WMD 0.45 kg; 95%CI¼0.20–0.71 kg) as participants older than 18 years (WMD 0.43 kg; 95%CI¼0.09–0.78 kg) in a previ-ous analysis». The two lean-mass WMDs are the same quantity (DXA fat-free/lean mass, kg, HP-minus-NP, under energy restriction), and they nearly coincide.
| Parameter | Wycherley 2012 — quoted + locus | Kim 2016 — quoted + locus | Same quantity? |
|---|---|---|---|
| Design | meta-analysis of 24 isocaloric HP-vs-SP RCTs (Wycherley et al., 2012) | SR+MA of 20 HP-vs-NP RCTs, DXA body comp (Kim et al., 2016) | YES — both HP-vs-NP contrast MAs under energy restriction |
| Age stratum | adults >18, unrestricted (Wycherley et al., 2012) | «adults 50 years and older» only (Kim et al., 2016) | NO — Kim is age-restricted; that IS the extension |
| Dose contrast | HP «1.25 6 0.17» vs SP «0.72 6 0.09» g/kg/d achieved (Wycherley et al., 2012) | HP >=1.0 (or >=25%E) vs NP <1.0 (or <25%E) g/kg/d (Kim et al., 2016) | YES roughly — both modest low-range contrasts, not high-dose |
| Lean/FFM WMD (%energy) | «0.43 kg; 95% CI: 0.09, 0.78» favouring HP (Wycherley et al., 2012) | «0.45 kg; 95%CI¼0.20–0.71» favouring HP (Kim et al., 2016) | YES — same DXA lean-mass kg contrast; the F-claim |
The composite beats either alone: FFM-sparing by a modest protein increase under a deficit is now evidenced both across the adult range and specifically in adults >=50, at essentially the same magnitude (~0.4-0.45 kg by %energy). This is a between-analysis comparison of two separate pooled estimates (Kim’s own), NOT a formal within-study age-interaction test — so read it as no age signal apparent at this dose, not as a proven null interaction. Despite the mechanistic expectation that anabolic resistance would make older adults respond differently, the contrast does not visibly amplify with age — and this is not evidence that older adults need less protein in absolute terms (they lose more lean mass at baseline, so the same relative sparing matters more) -> Protein Intake for Older Adults, Anabolic Resistance.
Why this is NOT type-E (independence fails despite a clean author-diff)
(inferred from Kim et al., 2016)
Kim (Campbell/Purdue — Kim/O’Connor/Sands/Slebodnik/Campbell) shares no author with Wycherley
(Adelaide/CSIRO — Wycherley/Moran/Clifton/Noakes/Brinkworth), so the cheap author-diff test passes. But
independence still fails on two counts, so this is corroboration + stratum-extension (F), not
[E-independent]:
- Kim explicitly cites and benchmarks Wycherley (its ref 20) — the convergence is self-stated, so a RAG over Kim alone reproduces it. The strict-E rule bars a source that restates an earlier one it cites.
- Trial-set overlap. Matching author + year across Kim’s trial table and Wycherley’s Table 4 trial list, about 7 of Kim’s ~20 trials are also Wycherley trials (Belobrajdic 2010, Campbell 2010, Farnsworth 2003, Layman 2003, Leidy 2007, Parker 2002, Wycherley 2010) — so the two MAs are not independent samples of trials. (Kim et al., 2016) (Wycherley et al., 2012) (corrected 2026-10-06: a speculative Adelaide/CSIRO list incl. trials not in Wycherley + Wycherley’s list is not held -> the verified shared trials; self-critique)
The surrogate + scope caveats persist (unchanged direction, honest ceiling)
- Lean mass is a DXA surrogate, not skeletal muscle. «While changes in lean mass measured using DXA are often attributed to changes in skeletal muscle mass, this inference should be made with caution» -> Surrogate Outcomes. Same boundary as the Refalo/Wycherley sections above.
- Soft tissue only — no function, bone, or health endpoint. The review is «limited to changes in soft tissue body composition (lean mass and fat mass) and does not address the potential impact of pro-tein intake on energy restriction–induced changes in skeletal bone or indices of health and functional well-being». The loop stays open at the patient-important level.
- Named gap: «New RCTs are needed to directly assess the effects of higher protein vs normal protein intakes on energy restriction–induced changes in body composition (including skeletal muscle mass) … especially older adults with sarcopenic obesity» — the sarcopenic-obesity RCT is unstudied.
- Modest dose contrast — the demonstrated benefit lands at ~1.0-1.5 vs ~0.6-1.0 g/kg/d, so like Wycherley it does not speak to whether pushing toward Refalo’s ~1.9 g/kg buys more in this population.
Decision relevance
- Cutting athlete / dieter who must protect lean mass (nonobese, resistance-trained): aim ~1.9 g/kgBM/day (or ~2.5 g/kgFFM/day if FFM is estimable), biasing UP toward 2.5-3.2 g/kgBM / 3.2-4.2 g/kgFFM when FFM loss would be genuinely costly (weight-class combat/strength athletes, bodybuilders). This is above the ~1.6 energy-balance target — the deficit raises the requirement.
- The RT stimulus and a modest deficit come FIRST. A well-programmed resistance-training stimulus and a ~modest (not ~500 kcal-plus) deficit likely do as much or more for FFM retention than the protein dial — Refalo calls the ranking «likely context-dependent» (Refalo et al., 2025). Do not let the protein number substitute for the training. The RT lever is quantified for the obese-elderly dieter on Resistance Training During Caloric Restriction (RT preserves ~0.82 kg lean mass under CR, RMD 0.819 kg [0.364-1.273]) — and Sardeli notes the single included trial with higher protein (30% vs 15% of energy) showed lower LBM loss than the others (RMD 1.3 kg vs a main effect the authors give there as RMD 0.6 kg, which does not match the 0.82 kg pooled headline) (Sardeli et al., 2018). That is one trial set against the rest, not a tested protein x RT interaction, so RT and protein stack as complementary levers is a single-trial signal (corrected 2026-10-06: stack as complementary levers -> single-trial signal, Sardeli chunk 01).
- Sedentary overweight/obese dieter (no training) [@wycherley2012]: raising protein from ~0.8 to ~1.2-1.3 g/kg/d within an isocaloric low-fat deficit still buys a small FFM/FM benefit (FFM +0.43 kg, FM -0.87 kg, weight -0.79 kg) — the direction holds even without resistance training, though the magnitude is modest and the cardiometabolic upside is limited to triglycerides (a surrogate), with lipids/BP/glucose unchanged. Treat this as a low-cost adjunct to the deficit, not a large lever.
- Older adult (>=50) losing weight [@kim2016]: raise protein to >=1.0 g/kg/d (above the 0.8 RDA) within the deficit — older dieters retain more lean mass and lose more fat at the same total weight loss: ~0.45 kg lean by %energy (vs 0.43 kg in Wycherley’s all-ages pool); ~0.83 kg by g/kg/d, which Kim calls «comparable» to 0.60 kg in younger adults (Krieger, via Kim). (corrected 2026-10-06: essentially identical for both metrics -> per-metric comparison; self-critique) This matters more for them (lower baseline lean mass, more likely to lose lean mass during weight loss), even though the contrast effect is not age-amplified. Prescribe on g/kg/d rather than %energy where possible. Add a resistance stimulus if feasible (likely the bigger lever, though most trials here were diet-only) -> Protein Intake for Older Adults.
- Leaner -> plausibly higher target (a suggestive BF% ordering in Refalo, not an established moderator); the specific raised target (~1.9 g/kg) and the per-gram BF% slope remain off-support for the obese. But does more protein help at all under a deficit in the overweight/obese? is now answered yes, modestly by Wycherley (a low-dose contrast in that population) — so the obese are no longer wholly unstudied on the DIRECTION, only on the dose-response and the target. The per-FFM denominator only sharpens that the number is undefined until the kg is named.
- On a GLP-1 drug: this is the protein side of the muscle-preservation strategy — but appetite suppression makes hitting a raised protein target harder exactly where it matters -> GLP-1 and Lean Mass.
- Open loop / medium confidence on the DIRECTION only: the FFM-preservation direction now rests on three gold sources across strata — a dose-response meta-regression (Refalo, RT non-obese), a direct-contrast RCT MA (Wycherley, all-age sedentary overweight/obese), and an age-restricted MA (Kim, >=50) — with two independent labs (Refalo, Wycherley) plus a same-direction, same-magnitude age extension (Kim, not an independent leg). Hence medium, not low. But all three terminate on surrogates (FFM/lean mass, TG), the effect sizes are small (sub-kg), and nothing here grades a chosen protein intake against a realized function or health outcome — the loop stays open. No target number earns above low: the ~1.6 and ~1.9 g/kg regions are still poorly identified, and both contrast MAs (Wycherley, Kim) show benefit at a LOWER achieved dose (~1.0-1.25 g/kg). Treat as a well-reasoned prior for practice, revisable on controlled higher-vs-lower RCTs powered on function — including the sarcopenic-obesity RCT Kim names as unstudied.