CT Body Composition as Predictor of Exercise Therapy Outcome in Peripheral Arterial Disease

Overview

Supervised exercise therapy (SET) is the recommended first treatment for patients with leg artery disease (peripheral arterial disease, PAD) causing pain when walking. However, approximately 40% of patients do not benefit meaningfully and go on to require a procedure to open the blocked arteries within three months. This study investigates whether body composition measurements - specifically the quality of muscle and the amount of belly fat - taken from a CT scan already performed as part of routine care, can identify before treatment begins which patients are unlikely to respond to exercise therapy. If confirmed, this approach would allow doctors to use information from a scan patients are already having, with no additional tests, to better match patients to the right treatment from the start.

BACKGROUND Peripheral arterial disease (PAD) affects approximately 236 million adults worldwide. Intermittent claudication - exertional lower limb ischaemic pain that resolves with rest - is its most prevalent symptomatic form. Three 2024 international guidelines (ACC/AHA, ESC, ESVS) unanimously recommend supervised exercise therapy (SET) as first-line treatment before revascularisation. Two Cochrane systematic reviews confirm that SET significantly improves maximum walking distance and quality of life compared with home-based exercise and walking advice alone. Despite population-level efficacy, individual response is highly heterogeneous: approximately 40% of patients fail to achieve meaningful functional improvement or require revascularisation within three months of initiating SET. Conventional predictors - ankle-brachial index and baseline claudication distance - do not reliably predict the magnitude of SET response. No validated pre-therapy metabolic tool exists to identify patients who lack the biological substrate for aerobic adaptation. RATIONALE Chronic lower-limb ischaemia causes progressive skeletal muscle remodelling - including intramuscular lipid infiltration, fibre-type shift toward type IIx fibres, and mitochondrial dysfunction - that directly constrains the aerobic adaptation capacity upon which SET depends. CT-derived skeletal muscle attenuation (Hounsfield units, HU) quantifies intramuscular lipid with excellent reproducibility (coefficient of variation \<1%). Single-slice CT analysis at the L3 vertebral level provides whole-body body composition estimates correlated with DXA (r = 0.86-0.94) and is a validated systemic metabolic phenotyping tool - it reflects whole-body metabolic reserve, not ischaemia-specific lower limb myopathology. Visceral adipose tissue (VAT) is an independent pro-inflammatory and endocrine compartment that impairs aerobic exercise adaptation through systemic mechanisms. CT angiography (CTA) is routinely performed in symptomatic PAD patients being evaluated for revascularisation; L3 body composition extraction from existing CTA incurs zero additional patient burden. The research group has previously demonstrated that L3 CT body composition parameters predict overall survival after transcatheter aortic valve implantation using the same methodology (Pekar et al., Sci Rep 2024; Kantor et al., Cor Vasa 2025). STUDY AIMS Primary: To determine whether L3 CT-derived skeletal muscle density and VAT area, extracted opportunistically from existing diagnostic CTA, independently predict composite SET treatment success at three months in patients with Fontaine IIa-IIb intermittent claudication. Three co-primary hypotheses: H1: Lower skeletal muscle density (HU) independently predicts SET treatment failure. H2: Higher VAT area (cm²) independently predicts SET treatment failure. H3: The sarcopenic obesity phenotype (co-occurrence of low muscle density and elevated VAT) is associated with the lowest SET treatment success rates across four body composition phenotype groups. STUDY DESIGN Prospective, single-centre observational cohort study. February 2023 - December 2026. Complex Cardiovascular Centre, Hospital AGEL Trinec-Podlesi, Trinec, Czech Republic. Ethics approval: EK 314/22. Reported per STROBE and TRIPOD guidelines; SPIRIT 2013 checklist provided as supplementary file. INTERVENTION (standard clinical care, not study-allocated) 36 supervised exercise sessions (3×/week × 12 weeks), per 2019 AHA Scientific Statement on optimal exercise programmes for PAD. Each session: 60 minutes (45 min structured exercise + 15 min patient education). Treadmill exercise at 3.2 km/h, walking to claudication level 3 on a four-point scale, rest to level 1, five to six cycles per session. Progression: speed +0.3 km/h every two weeks (weeks 1-4, maximum 4.0 km/h); grade +1% every two weeks (weeks 5-8, maximum 3%). Resistance training from week 5 (2 sets × 10 repetitions progressing to 5 sets at \~70% one-repetition maximum, major lower extremity muscle groups). Adequate adherence defined as ≥30 of 36 sessions. CT BODY COMPOSITION L3 axial CT slice; AutoMATiCA U-Net neural network segmentation; \~350 ms/scan. * Skeletal muscle: -29 to +150 HU within muscle fascia mask * VAT: -150 to -50 HU within peritoneal cavity mask * SAT: -190 to -30 HU exterior to body wall * IMAT: -190 to -30 HU within muscle fascia, exterior to individual muscle boundaries Baseline (T0) from existing diagnostic CTA; repeat CT at 3 months (T1) for exploratory analysis. PRIMARY OUTCOME Composite at 3 months: (a) functional improvement ≥50 m absolute or ≥50% relative MWD gain on standardised treadmill (3.2 km/h); AND (b) revascularisation-free status (no endovascular or surgical lower extremity revascularisation). Treatment success = both criteria met. Revascularisation events adjudicated by vascular surgeon blinded to body composition data. STATISTICAL ANALYSIS Multivariable logistic regression; predictors z-score standardised; bootstrap internal validation (1,000 samples); optimism-corrected AUC, calibration slope, calibration-in-the-large. Bonferroni-corrected p \< 0.017 per co-primary hypothesis. Fine-Gray subdistribution hazard model for competing risks (death). Multiple imputation by chained equations (MICE; 20 imputed datasets). Pre-specified sensitivity analyses (n=13) including composite threshold alternatives, component-separate analyses, three-predictor model, alternative HU and VAT thresholds, adherent subgroup analysis. SAMPLE SIZE Target: 40 treatment failures = 10 events per variable for four-predictor model; yields AUC confidence interval ±0.10. Assuming 40% failure rate and 22% attrition: 128 enrolled → 100 evaluable → 40 failures. Supplementary power: 80% to detect OR ≥2.0 per 10 HU decrease in muscle density (α = 0.05, two-tailed). DERIVATION STATUS This is an explicit derivation cohort. AUC precision of ±0.10 at 40 events is pre-stated; conclusions will be confined to predictor association direction and magnitude. External multicentre validation is the pre-specified mandatory next step.