# Waist Circumference Task Card Task id: `waist` Evidence status: grounded Evidence reviewed: 2026-07-29 ## Target Waist circumference. Continuous-scalar regression. The default research stage is `00_00_visit`, but the Task can be configured for another available stage. Implementation facts: - target field: `waist_circumference` - target dataset: `anthropometrics` - target source: None - default research stage: `00_00_visit` - valid range: 50.0-200.0 cm - primary metric: R2 ## Why This Matters Waist circumference is a simple abdominal-adiposity measure and a clinically legible cardiometabolic-risk marker. It can capture central adiposity better than BMI alone, but it remains strongly tied to body size. For PhenoBench, waist is useful as a low-cost body-shape target and as a comparator to DXA VAT. ## Baselines And Ceilings The required floor is a no-modality-feature baseline: `NoFeaturesPredictor` contributes no modality features, and `DemographicFloorStrategy` fits age, sex, and BMI. The task code records a high floor around R2 0.74 because BMI is mechanically close to waist. Natural comparators include BMI-only/anthropometric models, VAT, waist-to-hip ratio, activity/gait features, and metabolic feature sets. The interesting number is lift over the anthropometric floor, not raw R2. ## Benchmark-Track Evidence literature review and literature search support a BMI-independent Nightingale signature most strongly for VAT, not waist itself. Waist remains a useful body-shape probe when every direct anthropometric reconstruction field is excluded. The proposed V1/V2 track uses only age, sex, BMI, and the fixed Nightingale panel. ## Caveats - BMI can almost reconstruct waist; raw performance can be misleading. - Measurement protocol, clothing, posture, and operator variation can add noise. - Waist is not the same as visceral fat or cardiometabolic event risk. - Same-anthropometrics predictors can leak target-adjacent measurements. ## Metrics Primary metric is R2. Secondary MAE in cm is useful. Audit views should include demographic-floor delta R2 and matched comparison to `vat`. ## References - Source material: `docs/task_cards/source_material/foundation-model-surrogate-targets-v1.md` - Waist-circumference cardiovascular-risk context from the source material's INTERHEART reference. - Shilo et al. (2026), heterogeneity of insulin-resistance surrogates, *medRxiv*, . Local copy in **HPP documentation**: `assets/papers/hpp/shilo_2026_heterogeneity_of_insulin_resistance_surr/`. - Neeland et al. (2019), replicated NMR signature of VAT, . ## Open Questions - Should waist be mainly a comparator for VAT rather than a headline target? - Which anthropometric predictors should be excluded to avoid trivial reconstruction? - Should sex-specific or ethnicity-specific waist thresholds be added for classification variants?