The validation of low-dose CT scans from the [18F]-FDG PET-CT scan to assess skeletal muscle mass in comparison with diagnostic neck CT scans

**Background:** Radiologically defined sarcopenia (low skeletal muscle index, SMI) is an emerging biomarker for adverse clinical outcomes in head and neck cancer (HNC) patients. SMI measurements at the third cervical vertebra (C3) have been validated on diagnostic neck CT scans but not on low-dose (LD) CT scans from [18F]-FDG PET-CT. This gap prevents SMI analysis in HNC patients who have PET-CT but no diagnostic neck CT. The study aimed to determine whether SMI and sarcopenia diagnosis based on LD CT are comparable to those from diagnostic neck CT.

**Methods:** This retrospective cohort study used prospectively gathered data from the OncoLifeS data-biobank at University Medical Center Groningen. Inclusion criteria: patients with pathologically proven mucosal or cutaneous HNC, with paired diagnostic neck CT and LD CT from [18F]-FDG PET-CT within ≤30 days. Between August 2011 and March 2019, 272 patients were identified; 39 were excluded (low signal-to-noise ratio n=17, severe kyphosis n=7, anatomy not imaged n=7, muscle infiltration n=5, technical issues n=3), yielding 233 patients. Skeletal muscle was delineated semi-automatically at C3 using Hounsfield unit thresholds of −29 to +150. CSA at L3 was calculated using the Swartz equation, then normalized for height to obtain SMI (cm²/m²). Two sets of sex-specific SMI cut-off values (van Rijn-Dekker et al., Zwart et al.) defined sarcopenia. Inter-observer reliability was assessed with ICC. Agreement was evaluated using Bland-Altman plots and Cohen's Kappa.

**Key Results:** Mean age was 66.2±12.8 years; 74.2% were male; median BMI was 24.3 (IQR 7.2). Most common tumor sites: oropharynx (39.9%), larynx (34.3%), hypopharynx (9.9%). Inter-observer reliability was excellent for both diagnostic CT (ICC=0.992, 95% CI 0.980–0.997, p<0.001) and LD CT (ICC=0.990, 95% CI 0.975–0.996, p<0.001). Mean SMI was 43.27±7.71 cm²/m² for diagnostic CT and 43.08±7.77 cm²/m² for LD CT. Bland-Altman analysis showed mean ΔSMI = −0.19 cm²/m² with limits of agreement from −1.78 to 2.16 cm²/m²; 10 (4.29%) outliers were outside limits. No proportional bias was found (p=0.405). Using Zwart et al. cut-offs, sarcopenia prevalence was 60.5% (diagnostic) vs 57.9% (LD); agreement was almost perfect (κ=0.929, p<0.001). Using van Rijn-Dekker et al. cut-offs, prevalence was 42.1% vs 40.3%; agreement was also almost perfect (κ=0.911, p<0.001). Arm positioning did not substantially affect agreement: for arms down (n=200), κ=0.917 (Zwart) and κ=0.906 (van Rijn-Dekker); for arms up (n=31), κ=1.000 and κ=0.936, respectively.

**Clinical Implications:** This study validates that SMI measurements from LD CT of [18F]-FDG PET-CT are reliable and interchangeable with diagnostic neck CT for assessing sarcopenia in HNC patients. The findings enable sarcopenia assessment in patients who only have PET-CT imaging, expanding the population that can be screened for malnutrition risk and adverse outcomes. The robustness across different CT acquisition parameters, arm positions, and observers supports clinical implementation. Future research should focus on external validation of SMI cut-offs, incorporating muscle function measures, and interventional studies to preserve muscle mass during treatment.