The standing objection to a higher-protein target (the ~1.6 g/kg muscle number, the older-adult maintenance number, the deficit-stratum number) is that it will damage the kidneys. This page holds the direct evidence on that objection, and the answer for people with normal kidneys is a null: higher protein raises the glomerular filtration rate (GFR) a little, but there is no evidence that rise reflects kidney harm — it is read as a benign adaptive response. The finding is a route-(c) contraindication check — and its whole value is in getting the stratum boundary right, because the answer flips for established kidney disease.
The scope firewall, stated first. Devries studied adults without chronic kidney disease (CKD) — healthy, obese, type-2-diabetic, or hypertensive people with normal kidney function. Trials of people with CKD, type-1 diabetes, proteinuria, or prior kidney impairment were excluded (Devries et al., 2018). So this is no evidence of harm in the healthy — not protein is safe for everyone. In CKD, protein restriction is standard care; that is a different question on a different stratum and this page does not speak to it.
The finding — higher protein does not change kidney function in healthy adults
Devries pooled 28 RCTs (1358 subjects; 1975-2016) comparing higher-protein (HP: mean 1.81 ± 0.60 g/kg/d) with normal-or-lower-protein (NLP: 0.93 ± 0.51 g/kg/d) diets, in adults without CKD. The headline: «the results of the current meta-analysis suggest a nonexistent or trivial effect of HP consumption on GFR in individuals with normal kidney function.»
The number that matters is the change-from-baseline analysis, and it is null:
| Analysis | Effect (SMD, 95% CI) | n (studies) | GRADE |
|---|---|---|---|
| Change in GFR (pre→post) | 0.11 (−0.05, 0.27); P=0.16 — no difference | 1547 (19 RCTs) | Very low |
| Post-only GFR | 0.19 higher (0.07, 0.32); P=0.002 — trivial | 1688 (31 RCTs) | Low |
The two rows disagree, and that disagreement is the paper’s methodological point. The post-only comparison (HP-group’s ending GFR vs NLP-group’s ending GFR) finds HP a trivial amount higher; the change-from-baseline comparison finds no difference. Devries argues the post-only figure is the artifact: baseline GFR differed by >5% between arms in 6 of 19 studies despite randomization, so comparing only ending values imports that baseline gap — «These findings suggest that the post-only analysis exaggerated an effect of an HP diet on GFR.» The change analysis, which nets out the baseline, is the more valid estimate, and it is flat.
Supporting lines all point the same way (Devries et al., 2018):
- Dose gives no gradient on the outcome that counts. Protein dose correlated with post-only GFR (r=0.332, P=0.03) but not with the change in GFR (r=0.184, P=0.33); a biphasic fit was non-significant. So more protein does not drive a larger GFR change.
- The transient signal fades with time. In energy balance the small change-in-GFR effect sat almost entirely in the first 8 wk (SMD 0.30); at 8-24 wk and >24 wk it was gone (SMD −0.07 and 0.00) — a short-lived adaptation, not a progressive decline.
- Kidney-damage markers do not move. Of 8 trials reporting albumin excretion (a staging marker of kidney damage), «only 1 trial found that an HP diet increased the albumin excretion rate».
Why the GFR rise is not harm — kidney functional reserve
The worry traces to Brenner (1982): a higher protein load raises GFR (glomerular hyperfiltration), and hyperfiltration is thought to drive glomerular damage and eventual failure. Devries’s reframe is that this conflates two different hyperfiltrations — the animal-model kind that precedes damage is at the single-nephron level (surviving nephrons overworking after nephron loss), whereas protein feeding raises filtration at the whole-kidney level: «However, in these situations, the increase in GFR occurred at the single-nephron level, whereas in humans in response to protein-feeding or other stimuli, such as pregnancy or nephrectomy, glomerular hyperfiltration occurs at the whole- kidney level as a result of increased kidney blood flow».
That whole-kidney rise is «kidney functional reserve … a normal adaptive function of the kidney to increase solute clearance in response to an increase in solute load», and «this adaptive response does not represent a risk factor for the development of CKD».
The physiological analogues make the point: GFR rises ~65% in pregnancy without raising kidney-disease risk, and a single kidney after nephrectomy hyperfilters yet stays normal for >20 y. So protein raises GFR is a true observation that is not evidence of harm — and Devries is explicit that the raised GFR «alone is not evidence that the risk of CKD is modified».
Scope — how far the null reaches, and where it stops
The included populations were healthy, obese, hypertensive, and type-2-diabetic adults with normal kidney function. The type-2-diabetes subgroup matters because those individuals carry elevated CKD risk, yet «there was no difference in kidney function with the consumption of an HP diet» (pre/post SMD −0.11 [−0.35, 0.14]; post-only 0.11 [−0.13, 0.34]). So the null extends to an at-risk-but-not-yet-diseased stratum — Devries notes the small subgroup n and flags caution.
Where it stops (the stratum flip): established CKD, type-1 diabetes, proteinuria, or prior kidney impairment — none studied here, and the standing clinical rule there is protein restriction, not liberalization. The specific cutpoint is a clinical convention, not from Devries: PROT-AGE names severe kidney disease (eGFR <30, not on dialysis) as the exception where protein should be limited -> Protein Intake for Older Adults (where that boundary is source-attributed to Bauer). The route-(c) contraindication is real and narrow — it is a flip for a defined diseased stratum, not a general caution that erodes the healthy-adult target. For the drug side of the diseased stratum see Semaglutide and Kidney Outcomes in Chronic Kidney Disease.
The second source (Van Elswyk 2018) — same direction, different lineage, and four new arms
A second 2018 systematic review reaches the same read on the healthy kidney from a different research group and a different study set, and covers arms Devries’s meta-analysis did not. Van Elswyk narratively reviewed 26 studies (18 RCTs + 8 observational) of healthy adults above the US RDA but within the AMDR, requiring normal GFR (>=90) and energy balance and excluding weight-loss designs. Its GFR read is the same benign one: of 13 RCTs, «Most (n = 8), but not all (n = 5), RCTs reported significantly higher GFRs in response to increased protein intake, and all rates were consistent with normal kidney function in healthy adults.» — no pooled effect (this is a vote-count narrative synthesis, not a meta-analysis), the highest GFRs sitting in short/young/small studies the authors attribute chiefly to subject age. (Van Elswyk et al., 2018)
Parameter table — is this the same quantity as Devries? (BLOCKING precondition; the fourth column is the point.)
| Parameter | Devries 2018 (SR-MA) | Van Elswyk 2018 (SR, narrative) | Same quantity? |
|---|---|---|---|
| Synthesis mode | pooled SMD (meta-analysis) | vote-count + range (no pooling) | NO — different mode; compare DIRECTION, not a shared effect size |
| Population | adults w/o CKD (healthy/obese/T2D/HTN) | healthy adults, GFR>=90 + energy balance required, weight-loss excluded | overlapping; Van Elswyk narrower |
| Exposure contrast | HP (>=1.5 g/kg / >=20%E / >=100 g/d) vs NLP | >RDA (>=20%E / >=10% more) vs RDA (0.8/10-15%E) | YES — both «higher vs habitual/RDA» |
| GFR direction | «nonexistent or trivial effect … on GFR» | «all rates … consistent with normal kidney function» | YES — modest rise, stays normal |
| Kidney-damage marker | albumin: 1/8 trials raised AER | blood markers «little or no effect» | YES — markers ~unchanged |
The direction is commensurable (fourth column YES on the load-bearing rows); the effect-size cell is NO, so this is not a shared magnitude to pool — it is a same-direction arrival from a different dataset.
Independence — method/lineage-independent, but funding-CORRELATED (not a clean type-E). The two author
lists are disjoint — Devries/Sithamparapillai/Brimble/Banfield/Morton/Phillips (Phillips/McMaster academic)
vs Van Elswyk/Weatherford/McNeill (industry consulting + NCBA) — a different lineage and a different study
set (Van Elswyk adds 8 observational studies incl. two long cohorts; requires energy balance and GFR>=90;
excludes weight-loss), so the concordance on GFR rise is benign in healthy kidneys is a real robustness
gain on the direction. But it is NOT a bias-independent type-E second opinion: both sources carry
beef/dairy-industry funding, so the one channel most able to bias the interpretation — commercial interest
pushing toward removing a barrier to eating more protein — runs the same way in both. Independence of
method and data, correlated on incentive. No [E-independent] tag; the robustness is real but bounded.
Type-F refinement — four arms Devries’s GFR-only meta-analysis did not cover (Van Elswyk et al., 2018):
- Blood pressure — no harm, plausibly a benefit. «Evidence reported here suggests that protein intake above the US RDA has no adverse effect on blood pressure»; the two largest RCTs (DASH, OmniHeart) found higher-protein arms LOWERED BP in mild hypertensives, and the broader evidence base shows an inverse protein-BP relation (esp. when protein replaces carbohydrate), «regardless of source» (animal ≈ plant).
- Kidney stones — Van Elswyk read it as insufficient; a gold incidence MA now grades it positive.
Van Elswyk: «The evidence from the current review is limited and inconsistent with regard to the role of
protein intake and the risk of kidney stones» (a Norway evidence-based review likewise: «inconclusive»)
— the insufficient-evidence state, not no effect.
- [2026-09-23, Lin 2020] refinement — the incidence sub-gap is no longer insufficient. A gold SR+MA of incident nephrolithiasis pools animal protein RR 1.1 (1.02-1.19) (borderline-significant, no heterogeneity) and total meat RR 1.24 (1.12-1.39) as risk factors for first stones in the general population (Lin et al., 2020). So on the general- population incidence stratum, higher animal protein is now a graded (if modest and borderline) positive association, not merely «insufficient» — a type-F upgrade of Van Elswyk’s read. The mechanism matches the recurrence one (animal protein -> lower urine citrate/pH, higher uric acid). Still a highest-vs-lowest contrast with no dose-response, so read it as direction + rough size, not a threshold.
- Stratum split, not a contradiction: the incidence association (Lin, general population) sits beside a strong, stratum-conditional recurrence lever — in established stone-formers (esp. hyperuricosuria) the EAU restricts excess animal protein at LE 1b/Strong via specific urinary changes (hypocitraturia, low pH, hyperoxaluria, hyperuricosuria). Different outcomes (incidence vs recurrence) on different strata, concordant in direction, not opposed — the stone-forming lever is stratum-conditional because its mechanism (existing urinary abnormalities) is most active in stone-formers, not a formally-tested effect-modification interaction. See Kidney Stone Recurrence Prevention for the full incidence-vs-recurrence distinction.
- Blood markers of kidney function — little/no effect. No pattern of abnormality across 10 studies; the lone exception was BUN just past the UL at 3.6 g/kg (4x RDA) in 5 young men, which the authors flag as possibly «real or an artifact» given the tiny sample.
- Plant vs animal source — undetermined. «Data in the current review are insufficient to determine if increased protein intake from a particular source, i.e., plant or animal, influences kidney health outcomes.» — the source question stays open on both pages.
Two long cohorts push past the weeks-to-months surrogate (narrows, does not close, the G-gap). Van Elswyk includes the longest-horizon evidence held anywhere on this page — still observational, still eGFR / incidence rather than a randomized hard outcome, but 15-21 years rather than weeks (Van Elswyk et al., 2018):
- Herber-Gast 2016 (Doetinchem, n=3798, 15 y): total, vegetable, and animal protein «not associated with changes in eGFR over time»; increased low-fat dairy associated with less eGFR decline (P=0.0004).
- Rebholz 2015 (ARIC, n=15792, 21 y): «Neither total (P = 0.40) nor animal protein (P = 0.10) were associated with increased risk of kidney disease (15.6% during follow-up); vegetable protein was associated (P < 0.004) with an 11% decrease in kidney disease risk» — higher dietary acid load carried an 8% increased risk.
So total and animal protein are null for long-horizon eGFR decline and kidney-disease incidence in two large cohorts, and the only protective protein signals are for vegetable protein and low-fat dairy — consistent with the healthy-kidney null, and if anything nudging the source question toward plant/dairy.
Certainty, lineage, and conflicts — read the null honestly
-
The evidence is low-certainty, and the outcome is a surrogate. Devries’s own GRADE rating is «low to very low» — chiefly for unclear risk of bias (only 4 of 28 trials reported their randomization method) and unexplained heterogeneity. And GFR is a surrogate: the trials show no adverse move in a filtration marker over weeks-to-months, not an absence of hard kidney outcomes (kidney failure, need for dialysis) over years. No protein→hard-kidney-outcome RCT exists — that is the standing G-gap the Protein Intake for Older Adults page also names. The direction of the null is robust (change analysis, dose, duration, albumin, and mechanism all agree); its certainty is low. (Devries et al., 2018)
- [2026-09-17, Van Elswyk] the G-gap NARROWS but does not close. Van Elswyk adds two long prospective cohorts — Herber-Gast (15 y, eGFR decline) and Rebholz (21 y, kidney-disease incidence) — both null for total/animal protein. These are the closest thing held to a hard kidney outcome: long-horizon and on incidence/decline rather than a weeks-to-months filtration marker, but still observational, not the missing RCT. So the gap is now no RCT on hard kidney outcomes, with long-horizon observational incidence data present and null — do not read the cohorts as the randomized outcome trial that is still absent. (Van Elswyk et al., 2018)
-
This is NOT independent corroboration of the protein cluster — same lineage. Devries shares authors Morton and Phillips with Protein and Resistance Training for Muscle and Strength (Morton 2018), and Phillips also co-authors the PROT-AGE consensus that made the parallel renal- safety claim on Protein Intake for Older Adults (Bauer 2013). So the protein-safety position across these pages is one research school (Phillips / McMaster), reached by a coherent programme — a type-F extension that adds the kidney dimension and upgrades Bauer’s directional consensus to pooled RCT evidence, not a type-E independent second opinion. Do not count it as robustness-by-independence.
-
Commercial interest, disclosed. The senior author «reports having received research funding, travel support. and honoraria from the US National Dairy Council, Dairy Farmers of Canada, and the US National Beef Cattlemen’s Association» — protein/meat/dairy industry funding on a finding that removes a barrier to eating more protein. By symmetric standards this discounts the framing and cautions against over- reading the null as proven safe; it does not by itself overturn a result concordant with independent guidance (below). (Devries et al., 2018)
- [2026-09-17, Van Elswyk] the second source’s COI is STRONGER, and it does not remove the concern — it repeats it. All three Van Elswyk authors are beef-industry-paid (MEVE/CAW paid consultants; SHM is «the Executive Director of Human Nutrition Research» at the NCBA), and the funding was not arm’s-length: «Financial and material support for the writing of this manuscript has been provided by the Beef Checkoff», and «Technical experts employed by the funding source participated in the conception and design of the manuscript and provided critical review». So the two concordant sources share a commercial-bias direction — this is why the second source is not counted as bias-independent (above): it discounts the interpretive framing on both, and leaves the underlying renal-marker data as the less bias-exposed part. (Van Elswyk et al., 2018)
-
Confidence decision — held at
medium, not bumped[2026-09-17]. A second review from a different lineage and dataset, plus two long cohorts null for total/animal protein, genuinely raise the robustness of the direction. But three ceilings on certainty are unchanged: the shared beef-funding channel (correlated, not independent, bias), the individual-study quality (Van Elswyk: «All included studies were of moderate to high risk of bias» and, but for the two cohorts, «<6 mo»), and the still-missing RCT on hard kidney outcomes (GFR/eGFR remains a surrogate). Robustness up, certainty ceiling unchanged →mediumholds. -
Concordant with guidance (the guidance-null, defeated by agreement-with-warrant, not divergence). The null «are in line with statements from the WHO … and Institute of Medicine … on protein intake and kidney function» — bodies outside the Phillips lineage that already dismiss a causal protein→kidney- decline role. Devries’s contribution over guidance is the warrant: the change-from-baseline reanalysis and the functional-reserve mechanism, not a new direction. (Devries et al., 2018)
Decision relevance
- For a healthy adult, kidney worry is not a reason to hold protein below the muscle/older-adult target. The common objection to reaching ~1.6 g/kg (RT), ~1.0-1.2 g/kg (older maintenance), or the higher deficit-stratum figure does not survive the evidence for normal kidneys — a modest overshoot is low-harm on the kidney axis, which is exactly what lets The Estimate-to-Action Gap bias the target up where undershoot is costly. This page is the source for the «healthy kidneys» clause those pages lean on.
- The one hard flip is established kidney disease. For diagnosed CKD (PROT-AGE draws the line at severe disease, eGFR <30, not on dialysis — a clinical convention, see Protein Intake for Older Adults), protein is restricted, not liberalized — a genuine route-(c) contraindication, not a soft caution. Screening for it is a prescriber act outside this wiki’s scope; the decision-relevant fact is that the boundary exists and is narrow.
- Don’t oversell it. The correct claim is no evidence of GFR harm in the healthy over the studied horizon, on low-certainty surrogate evidence from one research lineage with commercial funding, but concordant with independent guidance. That is enough to defuse the objection; it is not enough to say proven safe at any dose indefinitely.
Limits
- Surrogate, not outcome — GFR and albumin excretion over weeks-to-months, not hard kidney outcomes over years. No protein→kidney-outcome RCT is held (the G-gap); two long observational cohorts (Herber-Gast 15 y, Rebholz 21 y, both null for total/animal protein) now sit closer to a hard outcome but are not the missing randomized trial. (Van Elswyk et al., 2018)
- Low-to-very-low GRADE — unclear randomization/allocation reporting in most trials; the primary change analysis is «Very low». Van Elswyk agrees on the individual-study quality: «All included studies were of moderate to high risk of bias» and, but for its two cohorts, «<6 mo». (Van Elswyk et al., 2018)
- Two sources, correlated funding — the Phillips/McMaster meta-analysis (dairy/beef-council funding) and the NCBA/beef-funded Van Elswyk review agree from different lineages and datasets, but share a commercial-bias direction; robustness of direction, not bias-independence.
- CKD and the seriously kidney-impaired are unstudied here — the null does not transport to them, and the standing rule for them runs the other way.