Abstract Background: Cancer-related sarcopenia, defined by progressive loss of skeletal muscle mass and function, worsens outcomes but lacks a single universally accepted diagnostic criteria. Current assessments are time-consuming. We aimed to identify a plasma proteomic signature of sarcopenia and uncover soluble mediators involved in muscle decline. Methods: Two MATCH-R cohorts (NCT2517892) were analyzed: advanced cancers treated with immunotherapy (training) and metastatic castration-resistant prostate cancer (mCRPC) treated with enzalutamide (validation). An external validation cohort came from TRACERx (NCT01888601, resected and relapsed NSCLC). In MATCH-R, skeletal muscle index (SMI) at L3 level was measured on CT/PET within 42 days of blood draw; ECOG PS was used as a surrogate for muscle function. Plasma proteomics was performed using Olink Explore 1536/3072. Bulk and, for selected cases, single-cell RNA-seq were available for paired biopsies. In TRACERx, skeletal muscle area (SKM) was quantified via an automated deep-learning pipeline (PMID: 37045997). An XGBoost classifier was trained on high-contrast cases (low sarcopenia, LS: high SMI, ECOG 0; high sarcopenia, HS: low SMI, ECOG ≥2) using neuromuscular-related proteins enriched in LS. The model generated sarcopenia probability (SP, 0-100%) applied to all cohorts. Results: The training cohort included 99 patients (36 high-contrast: 21 HS, 15 LS). Using a 50% SP cutoff, the model showed an accuracy of 0.889 in 18 high-contrast validation cases. SP correlated with ECOG PS when available (p 0.001) and SMI in training and mCRPC cohorts ( training: ρ = -0.39, p = 0.004 for male and ρ -0.42 for female; mCRPC: ρ = -0.41, p = 0.008), with SKM in TRACERx baseline (ρ = -0.29, p 0.01 and ρ -0.24 p 0.07 for male and female), recurrence ρ = - 0.42, p 0.002 and ρ -0.36 p 0.04 for male and female). A change in SP was associated with a corresponding changement in SKM in paired samples (ρ -0.32 , p=0.006). SP 50% was associated with poorer survival across datasets (training: OS 5 vs 25.8 months, p 0.0001; mCRPC: 8.9 vs 21.9 months, p 0.0001; TRACERx baseline: DFS 10.7 vs 20.2 months, p = 0.007; OS 25 vs 48 months, p 0.001; TRACERx recurrence: OS 44 vs 30.5, p =0.043). Transcriptomic analyses in both training and mCRPC cohorts showed convergent upregulation of inflammatory pathways and suppression of muscle-related programs in patients with high SP. Across all cohorts, amongst others, plasma IGFBP1, IGFBP2, and IL6 were consistently higher in sarcopenic patients. Functional assays showed that IGFBP1/2 ( 1µg/mL) markedly impaired human myoblast differentiation (reduced fusion index and decreased MHC/β-actin expression) Conclusions: Plasma proteomics offers a scalable, imaging-free diagnostic for sarcopenia and identifies IGFBP1/2 as actionable drivers of cancer-associated muscle dysfunction Citation Format: Filippo Gustavo Dall'Olio, Wael Salem Zrafi, Xinran Song, Littisha Lawrance, Fei Chen, Pierre Busson, Catherine Brenner, Rebecca Ibrahim, Marie Guinhut, Caroline Even, Nathalie Lassau, Diana Cardenas-Braz, Fabrice Barlesi, Yohann Loriot, Fabrice Andre, Mariam Jamal-Hanjani, Antoine Italiano, Yegor Vassetzky, Benjamin Besse. High-throughput plasma proteomics enables sarcopenia stratification and identifies the IGFBP axis as a key mediator of muscle impairment in cancer patients abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7677.
Dall'Olio et al. (Fri,) studied this question.