Caloric Restriction (CR)

Verdict: Strong (in mice) / Suggestive (in humans, biomarkers only) Last reviewed: 2026-04-24 Triangulated against anchor: Caloric restriction (this page is the canonical anchor for "Strong" in mice)

TL;DR

In mice, CR is the single most replicated lifespan intervention in biology — Strong. In humans, the largest RCT (CALERIE-2) showed slowing of one DNA-methylation pace-of-aging clock and improvements in cardiometabolic biomarkers, but no human lifespan data exists and the rhesus monkey results from NIA vs Wisconsin diverged for diet-quality reasons. Human verdict: Suggestive on biomarkers; sustained CR remains hard to do safely in lean adults.

What it is

A sustained reduction in caloric intake — typically 10-30% below ad libitum — without malnutrition (adequate protein, micronutrients, essential fatty acids). Distinct from short-term fasting, time-restricted eating, and protein restriction (each gets its own page). The CR literature uses "ad libitum" controls in animals and self-reported intake in humans, which complicates inference.

Proposed mechanism

Multiple converging mechanisms: reduced mTOR signaling, increased autophagy, improved insulin/IGF-1 sensitivity, reduced reactive oxygen species, altered metabolic substrate use (more lipid oxidation), and shifts in neuroendocrine signaling. CR is best understood as a systemic intervention that engages most aging hallmarks simultaneously.

Confidence: Established for the cellular and metabolic mechanisms in mice.

Evidence ladder

Invertebrate (T5)

CR / dietary restriction extends lifespan in S. cerevisiae, C. elegans, D. melanogaster. Effect sizes 30-100% depending on organism and protocol. The most reproducible single intervention in invertebrate aging research.

Mouse / rat (T3 — Strong, with extensive replication)

This is where CR earns its status as the ceiling intervention in mammalian aging.

• Decades of replication across hundreds of studies, multiple strains (C57BL/6, BALB/c, UM-HET3, F344 rats, etc.), multiple labs, multiple decades.
• Effect size: Median lifespan extension typically 20-40% under 30-40% CR; max lifespan extension 10-30%. Larger than rapamycin's effect.
• ITP not directly relevant — CR is a paradigm against which pharmacological interventions are compared, not a drug to test.
• Dose-response established: Mortality decreases monotonically with CR magnitude up to ~50% restriction; below that, malnutrition harms outweigh CR benefits.
• Strain dependence: Most strains respond; magnitude varies. Some strains (NZO/HlLtJ, others) show smaller or null responses — important caveat against universalist claims.
• Sex dependence: Both sexes respond; magnitude differs.
• Late-life initiation: Even CR initiated in middle-age extends lifespan, though magnitude is reduced vs. early-life onset.

NHP (T4 — instructive divergence)

Two long-running rhesus monkey studies famously disagreed:

• Wisconsin study (Colman 2009, Colman 2014) — clear lifespan extension under CR, with reduced age-related disease incidence.
• NIA study (Mattison 2017) — no significant lifespan extension, though healthspan improvements were real.

The reconciliation: the control diets differed. NIA's control diet was healthier (lower sucrose, naturally sourced). Wisconsin's controls were eating a sweeter, more refined diet — closer to the modern Western diet that CR is implicitly compared against. This is one of the most important confound stories in aging research and is the reason CR's translation to already-healthy humans is uncertain.

Human (T2 — biomarkers only; no lifespan data)

• CALERIE-2 (Ravussin 2015 onward) — 2-year RCT, ~220 healthy non-obese adults, ~12% mean CR achieved (target was 25%; participants undershot). The single most rigorous human CR trial.
• Outcomes shown:
  • Cardiometabolic improvements (LDL, blood pressure, insulin sensitivity)
  • Improved liver function, skeletal muscle quality, immune health
  • DunedinPACE clock slowed by 2-3%, translating to ~10-15% lower projected mortality risk (Waziry 2023)
  • PhenoAge and GrimAge clocks: not significantly affected (an inconsistent signal across clocks)
  • Reduced cellular senescence biomarkers (2024 plasma analysis)
  • Glycomic biological age (GlycAge) decreased in pilot analysis (2025)
  • Telomere length: mixed — accelerated attrition Year 1, different pattern with sustained CR (Hastings 2024)
• Adverse effects: sustained ~12% CR is difficult; bone density loss in some participants; mood and psychosocial concerns documented; not appropriate for lean elderly.
• No mortality endpoint in any human CR RCT — and given trial duration requirements, may never exist at scale.
• Okinawan / observational — historical Okinawan diet (~20% caloric deficit + plant-rich) associates with longevity; classic dietary-cohort confounding makes this T2 at best.

Confounds

• Control diet adequacy is the single most important confound (Wisconsin vs NIA). When the comparison diet is unhealthy, CR looks better.
• Self-report in humans — measured intake in CALERIE was lower than reported; participants struggled to maintain target.
• Heterogeneity of "CR" — short-term fasting, alternate-day fasting, TRE, and continuous CR are mechanistically distinct; pooling them muddles the evidence.
• Strain effects in mice — universalist "CR works" claims are slightly overstated; some strains don't respond.
• Generalizability to already-lean humans — most strong CR data comes from mice on standard chow that may itself be over-rich.

Conflict of interest scan

• CALERIE-2 was NIA-funded — no discount.
• The supplement / commercial-CR-tools market exists but is small relative to NMN/resveratrol/rapamycin commercialization.
• Net: minimal COI concerns for the rigorous evidence base.

Human translation

The honest decomposition:

1. In mice, CR is Strong. This is not in dispute. The mouse evidence is the gold standard against which other interventions are compared.
2. In humans, sustainable moderate CR (~12% achievable in CALERIE) appears to slow at least one validated pace-of-aging clock and improves cardiometabolic markers. This is real. It is not the same as "CR extends human lifespan" — that claim is unsupported by direct evidence.
3. The translation is gated by adherence and population. Lean, healthy adults rarely sustain meaningful CR. Already-overweight populations doing weight-loss may capture CR benefits as a side-effect of weight normalization without "CR" being a useful framing.
4. The Wisconsin/NIA NHP divergence suggests that "CR vs ad libitum" matters less than "healthy diet vs unhealthy diet" — which is a less radical implication than the longevity community typically frames.

Calibrated verdict

Strong (in mice) / Suggestive (in humans). This page is the anchor for Strong in mice — every other mouse intervention's "Strong" claim must be weaker than CR's because nothing has CR's depth of replication.

The split human verdict is unusual but methodologically defensible: the methodology says the verdict reflects evidence on the species-specific outcome. CR's mouse evidence is Strong; CR's human evidence is at most Suggestive (biomarkers, no mortality, adherence problems). Future intervention pages may inherit this dual-verdict structure when the species evidence base differs significantly.

Compared to rapamycin (Probable), CR has more decades of replication, more strain breadth, and a larger effect size in mice — hence Strong vs Probable. In humans, both are at the surrogate-endpoint stage, but rapamycin has at least one functionally meaningful endpoint (vaccine response) where CR has biomarker movement of less directly clinical relevance.

Compared to exercise (Strong, in humans), CR is the mirror image: exercise has Strong human evidence and weaker direct lifespan-RCT data; CR has Strong mouse evidence and only biomarker-tier human evidence.

Confidence interval on verdict

Mouse verdict (Strong):

• Could only be downgraded by a methodological reckoning of the entire CR literature (e.g., systematic finding that the strain/diet effects make CR's apparent lifespan benefit a control-diet artifact). Possible but unlikely.
• Will not be upgraded — Strong is the ceiling.

Human verdict (Suggestive):

• Could move to Probable if a long-duration follow-up of CALERIE participants shows a hard-endpoint divergence (e.g., reduced incidence of major age-related diseases over 10-20 years). Pre-registered, in progress.
• Could move to Mostly hype if subsequent biomarker data fails to track outcome data (e.g., DunedinPACE slowing doesn't predict mortality reduction in independent cohorts).
• Most likely 2-year trajectory: stays at Suggestive on biomarkers; the long-tail follow-up data is the decisive future input.

Open questions

• Q: Does the DunedinPACE slowing in CALERIE-2 translate into hard-endpoint differences (mortality, age-related disease incidence) on long-term follow-up?
• Q: Is the Wisconsin-vs-NIA divergence fully explained by control-diet quality, or are there residual strain or protocol effects?
• Q: For lean adults, does CR provide benefit beyond what is captured by switching from a Western diet to a Mediterranean / whole-food diet at the same caloric intake?
• Q: Is there a "minimum effective CR" threshold below which biomarker effects vanish? CALERIE achieved ~12% — what does ~5-8% look like?

Sources

• Colman et al. 2014, Nat Commun — CR reduces age-related and all-cause mortality in rhesus monkeys (Wisconsin)
• Mattison et al. 2017, Nat Commun — CR improves health and survival of rhesus monkeys (NIA)
• Ravussin et al. 2015 — CALERIE-2 design and 2-year outcomes
• Waziry et al. 2023, Nature Aging — CR slows DunedinPACE in CALERIE
• Hastings et al. 2024, Aging Cell — Long-term CR effect on telomere length
• 2024 — CR reduces biomarkers of cellular senescence in CALERIE plasma
• 2025 — 2-year CR intervention reduces glycomic biological age biomarkers (medRxiv → npj Aging)
• CALERIE Phase 2 highlights — physiological, psychological, behavioral outcomes

Produced under methodology locked 2026-04-24. Anchor for the Strong (mice) band; Suggestive (humans).