The central-fat analog of the BMI mortality curve, from the first dose-response systematic review and meta-analysis of the question: 72 prospective cohort studies, 2 528 297 participants, 150 164 deaths (Jayedi 2020, BMJ). It quantifies waist circumference, hip circumference, thigh circumference, waist-to-hip ratio (WHR), waist-to-height ratio (WHtR), waist-to-thigh ratio, body adiposity index (BAI) and A body shape index (ABSI) against all-cause mortality, and — the reason it is a distinct page from BMI and All-Cause Mortality — asks the shape and the BMI-independence of each. (Jayedi et al., 2020)
The per-measure associations — central fat is harmful, hip and thigh are protective
Summary hazard ratios per the stated increment (maximally adjusted models):
| Measure | Increment | HR (95% CI) | n studies | I2 |
|---|---|---|---|---|
| Waist circumference | 10 cm | 1.11 (1.08-1.13) | 50 | 88% |
| Waist-to-hip ratio | 0.1 unit | 1.20 (1.15-1.25) | 31 | 90% |
| Waist-to-height ratio | 0.1 unit | 1.24 (1.12-1.36) | 11 | 94% |
| Waist-to-thigh ratio | 0.1 unit | 1.21 (1.03-1.39) | 2 | 97% |
| A body shape index | 0.005 unit | 1.15 (1.10-1.20) | 9 | 87% |
| Body adiposity index | 10% | 1.17 (1.00-1.33) | 4 | 75% |
| Hip circumference | 10 cm | 0.90 (0.81-0.99) | 9 | 95% |
| Thigh circumference | 5 cm | 0.82 (0.75-0.89) | 3 | 54% |
The waist-based measures all carry significant excess mortality; larger hips and thighs carry lower mortality (see the gluteofemoral section below). Heterogeneity is high throughout (I2 mostly 85-97%) — a persistent caveat on every point estimate here.
The load-bearing finding — central fat predicts mortality independent of BMI
«Positive associations persisted after accounting for body mass index» (Jayedi et al., 2020) — and for waist circumference the association does not merely survive BMI adjustment, it strengthens: HR 1.11 overall rises to 1.17 (1.13-1.22, n=15) in the studies that adjusted for BMI, and to 1.19 (1.14-1.24) in those adjusting for BMI + physical activity + smoking + alcohol together. (Jayedi et al., 2020) WHR shows the same direction (1.20 -> 1.26 with BMI adjustment). This is the decision-relevant delta over the BMI curve: two people at the same BMI carry different mortality risk according to where the fat sits, and the waist measure carries information BMI does not.
Jayedi names the mechanism for why central fat is not redundant with BMI, and specifically why the low end behaves differently: «being underweight by body mass index reflects lower beneficial lean mass and detrimental visceral fat mass. In contrast … smaller waist circumference might reflect lower detrimental visceral fat mass and does not necessarily reflect lower lean body mass» (Jayedi et al., 2020). Low BMI conflates frailty (low lean mass) with leanness; low waist does not — which is exactly why the central-fat curve is less U-shaped than the BMI curve (next section).
Most of the signal runs through the cardiometabolic intermediates. Adjusting for blood pressure, serum cholesterol and type 2 diabetes attenuates WHR from 1.23 (no adjustment) to 1.07 (0.99-1.15) and waist circumference from 1.11 to 1.06 (0.99-1.13) — near-null once the intermediates are controlled. (Jayedi et al., 2020) The effect is largely mediated by BP/lipids/glycaemia rather than acting through an independent channel — the same mediation Waist-to-Hip Ratio and Cardiovascular Risk reads in the ERFC CVD data. Adjusting for a mediator is not a confounder correction, so this bounds the mechanism, not the association’s reality.
The shape — central fat is LESS U-shaped than BMI, and the protective lower arm largely vanishes
This is the sharpest contrast with BMI and All-Cause Mortality. The BMI-mortality curve is a J with a substantial protective-looking lower arm (the obesity paradox — overweight appearing protective until bias is stripped) -> The U-Shaped Association Artifact. The central-fat curves have little or no protective lower arm:
- Waist-to-hip ratio — monotone, no knee. «Overall the relations were monotonic with little evidence of sharp changes at particular cut-off points» (Jayedi et al., 2020); men and women alike. No threshold located over the studied range — read as the dose-response prior says (no knee is shown, not none exists).
- A body shape index — monotone positive across the studied range without a nadir arm. (Jayedi et al., 2020)
- Waist circumference — nearly J, but the protective arm is negligible. In men the nadir is at WC 90 cm: «lowest at a waist circumference of 90 cm (hazard ratio90cm 0.96, 95% confidence interval 0.94 to 0.98), and then increased sharply and linearly» (Jayedi et al., 2020); in women risk is flat across WC 60-80 cm (HR at 80 cm 1.01, 0.99-1.03) then rises sharply and linearly. (Jayedi et al., 2020) The lower arm spans only ~0.96-1.01 — a threshold-then-rise, not a benefit-then-harm curve. The nadirs (WC 90 cm men / 80 cm women; WHtR 0.50) coincide with the conventional clinical cut-points for abdominal adiposity.
- Waist-to-height ratio — J, nadir 0.50, then a sharp linear increase (men HR0.50 0.96, women 0.90). (Jayedi et al., 2020)
- Body adiposity index — the one genuine U (nadir at BAI 30%, HR 0.89, 0.84-0.95), but on only 4 studies and BAI carries a large total-fat component, so it is the weakest and least waist-like of the set. (Jayedi et al., 2020)
Artifact reading (weak checks only). Jayedi ran the exclude-early-follow-up / sick-baseline family of checks — restriction to never smokers, to healthy participants, to healthy never smokers, to studies with >10 years’ follow-up — and the waist/WHR associations «were approximately similar to those of the main analyses». (Jayedi et al., 2020) Per The U-Shaped Association Artifact these are the weak checks (the alcohol U survived them and was still artifact); no Mendelian randomization and no referent-correction were run, so the residual arms are not strongly adjudicated. But the key point is structural: central fat has far less protective lower arm to defend than BMI does — the frailty/low-lean-mass confound that manufactures BMI’s obesity paradox does not load the low-waist end (mechanism above), so the shape is closer to monotone by construction. The kickoff hypothesis — a waist-based measure strips the BMI U — holds for WHR/ABSI/WC.
Hip and thigh are protective — the gluteofemoral / muscle signal
Larger hip circumference (HR 0.90 per 10 cm) and thigh circumference (0.82 per 5 cm) carry lower mortality, independent of BMI and opposite in sign to waist. This is the same waist-harmful / hip- protective split Waist-to-Hip Ratio and Cardiovascular Risk found for MI (why the ratio outperforms either alone), now on the mortality endpoint: gluteofemoral fat is metabolically protective and thigh circumference partly indexes skeletal muscle mass -> Low Muscle Mass and Mortality. The thigh estimate rests on only 3 studies, the hip on 9 — directional, not precise. (Jayedi et al., 2020)
Age attenuation — the signal weakens in the old
The association became non-significant for waist circumference or weakened substantially for WHR in participants aged 60 and older (Jayedi et al., 2020): WC 1.03 (0.98-1.08, n=14) and WHR 1.09 (1.01-1.17, n=8) over 60. This mirrors the older-adult attenuation on the BMI curve (competing risk + reverse causation rise with age) — condition on age before reading the central-fat estimate, and treat the old-age arm as the reverse-causation-suspect one.
Decision relevance
- Measure waist (or WHtR) alongside BMI, not instead of it. Jayedi’s own conclusion: «measures of central adiposity could be used as a supplementary approach, in combination with body mass index, to determine the risk of premature death» (Jayedi et al., 2020) — a third position on the which-measure debate, distinct from INTERHEART’s switch to WHR and ERFC’s BMI is adequate stances: on the mortality endpoint the waist measure adds BMI-independent information, so hold both -> BMI vs Abdominal-Adiposity Markers - Which Predicts CVD. WHtR carries a memorable action rule (keep waist below half your height; nadir 0.50).
- Central fat is a big-rock lever, and a normal-BMI person is not exempt. A person at healthy BMI but high waist (normal-weight central obesity) carries the excess risk the BMI cut-point misses — the BMI-independence finding says so directly -> Layer 1 - Ranking Interventions for a Stratum, Menopause and the Shifting Levers.
- This is association, not the benefit of shrinking a waist. A dose-response across people does not establish that reducing an individual’s waist moves them down the curve; almost all included studies measured baseline anthropometry only and «potential changes during the follow-up period were not considered». (Jayedi et al., 2020)
Limits
- Weak checks only — no MR, no referent-correction. The reverse-causation controls are the exclude- sick-baseline family (the weak route); the strong instruments the The U-Shaped Association Artifact decision rule asks for were not run. So the causal reading is directionally supported, not established — though there is little protective arm to adjudicate.
- Aggregate-data dose-response — shapes are coarse. «some subtle issues such as identification of thresholds and the exact shape of the curves cannot be reliably observed in such data» (Jayedi et al., 2020); Jayedi calls for an individual-participant-data MA carrying BMI and the central measures to settle incremental value. The thin measures (waist-to-thigh n=2, thigh n=3, BAI n=4) cannot support strong shape claims.
- High heterogeneity (I2 85-97%) across almost every measure; follow-up duration, study quality, smoking adjustment and assessment method were named heterogeneity sources.
- Coherence, not validity (R1). Internally sound and source-faithful; no operation here grades the measure against a realized outcome.