The joined disagreement. INTERHEART (case-control, acute MI) found waist-to-hip ratio three times more strongly related to MI than BMI, and recommended switching the principal clinical adiposity measure from BMI to WHR. ERFC 2011 (pooled prospective cohorts, incident CVD) found BMI, waist circumference and WHR have similar association strength, and explicitly names and refutes the INTERHEART recommendation. Same contrast (which adiposity measure associates most strongly with a coronary/CVD event), opposite verdicts, and ERFC cites INTERHEART as the report it overturns — a genuine type-D clash, not two different questions. — this framing (what is joined, where each design’s answer holds) is the wiki’s; the source-attributed claims sit below.
The parameter table — is this a real joined issue?
The commensurable quantity is the within-study relative ranking of the adiposity measures (does WHR beat BMI in the same dataset), NOT the absolute per-SD magnitudes — those use different adjustment sets, SD sizes and referents across the two studies and are not directly comparable. On the one cell that IS directly matched (BMI per 5 kg/m2, which ERFC itself compares head-to-head), the case-control under-states BMI relative to the cohort.
| Parameter | INTERHEART (Yusuf 2005) | ERFC 2011 (Wormser et al.) | Same quantity? |
|---|---|---|---|
| Contrast tested | WHR vs BMI, strength of relation to acute MI | BMI vs WC vs WHR, strength of relation to incident CVD (CHD + ischaemic stroke) | YES — both: which adiposity measure relates most strongly to a coronary/CVD event |
| Design | retrospective case-control, anthropometry measured concurrent with the acute MI; 27 098 people, 52 countries | prospective IPD-pooled cohorts, baseline anthropometry, serial-measurement regression-dilution correction; 58 studies, 221 934 people | NO — the crux (reverse causation / selection present vs largely removed) |
| Relative-strength verdict | WHR ~3x more strongly related than BMI; WHR the strongest of all measures | measures similar: per-SD CVD HR BMI 1.23, WC 1.27, WHR 1.25 | Opposed verdicts on the SAME contrast |
| Matched cell: BMI per 5 kg/m2 | OR for MI 1.12 | HR for CHD 1.32 | YES — same unit; case-control under-states BMI (ERFC’s own comparison) |
| Outcome | first acute MI (coronary) | incident CVD; CHD reported separately | ~same for the coronary leg; ERFC broader |
| Population | all-ethnicities, incl. South Asian / Arab / African | 90% European descent, developed countries | NO — part of the resolution, not the tension |
The fourth column is YES on the contrast and on the matched BMI cell; the NO on design is why the
verdicts differ, and the NO on population is what confines the residual live disagreement. Issue
joined -> file the tension.
View A — the abdominal marker beats BMI (Yusuf, INTERHEART 2005)
«The INTERHEART study clearly indicates that of the various anthropometric measures commonly used, waist-to-hip ratio shows the strongest relation with the risk of myocardial infarction worldwide» (Yusuf et al., 2005). Per 1 SD (BMI/height- adjusted) WHR 1.37 vs BMI 1.10; BMI’s MI signal collapsed after adjusting for WHR and vanished after full risk-factor adjustment, while WHR held. BMI was weakest in all 8 ethnic groups and non-significant in South Asians, Arabs and mixed-race Africans. Full extraction: Waist-to-Hip Ratio and Cardiovascular Risk.
View B — the measures are similar; the superiority claim is a design artifact (ERFC 2011)
«BMI, waist circumference, and waist-to-hip ratio each have a similar strength of association with cardiovascular disease risk» (Collaboration, 2011), and «Our findings reliably refute previous recommendations to adopt baseline waist-to-hip ratio instead of BMI as the principal clinical measure of adiposity» [EXTRACTED (Emerging Risk Factors Collaboration - BMI Abdominal Adiposity Cardiovascular 2011) chunk 01]. The named mechanism for the discrepancy is design bias: «the greater susceptibility of retrospective studies of acute myocardial infarction to some biases (eg, selection biases, reverse causality)» than prospective studies (Collaboration, 2011). ERFC adds two facts INTERHEART could not: BMI is far more reproducible (regression dilution ratio 0.95 vs WHR 0.63), and no adiposity measure adds prediction over BP + diabetes + lipids (C-index changes -0.0001 / -0.0001 / +0.0008).
Hidden insight — a marker’s superiority can be manufactured by the measurement design
The abdominal-marker advantage is exactly what a concurrent case-control would produce even if the measures were equal: acute illness before an MI strips skeletal muscle, lowering hip circumference and raising WHR in cases (reverse causation), inflating the WHR gradient; the same illness can lower weight, deflating BMI. Prospective baseline measurement removes this, and the ranking flattens. So the WHR-over- BMI ordering is a design artifact for prediction purposes — a worked instance of the reverse- causation hazard in The U-Shaped Association Artifact, here manufacturing not a U but a spurious between-marker ordering. Note this does not touch aetiology: abdominal adiposity remains a real, modifiable CVD determinant acting through BP / lipids / diabetes; the artifact is in the marker ranking, not in whether fat matters. — the reverse-causation-on-hip mechanism is the wiki’s reading of ERFC’s stated “selection biases, reverse causality”; the muscle-loss-raises-WHR step is mechanistic inference, not an ERFC claim.
Resolution — scope-split, not a flat winner
- Developed-country prospective CVD prediction, conventional risk factors known: ERFC wins on design. BMI is an adequate, more-reproducible clinical adiposity measure; switching to WHR buys no discrimination. This is the higher-design test and it explicitly refutes the case-control claim.
- Non-European populations: unresolved. ERFC was 90% European descent and states more data are needed elsewhere, so INTERHEART’s ethnic-heterogeneity signal (BMI carries no MI signal in South Asians / Arabs / mixed-race Africans) is untested, not refuted — the live residual question.
- Where lipids are unmeasured (resource-poor settings): a simple adiposity measure is a modest substitute (BMI+WHR ~ a quarter of the predictive gain of total+HDL cholesterol), and here the marker-vs-marker question INTERHEART asked is again relevant.
- Coherence, not validity (R1): ERFC and INTERHEART are both internally sound; neither is graded against a realized outcome here. The resolution is which design answers which stratum’s question.