Tumour proximity to major neurovascular structures complicates resection of deep soft tissue sarcomas (STS), yet its independent impact on overall survival (OS) remains unclear. This study assessed whether anatomical proximity, measured by dichotomised distance and the Fujiwara classification, predicts oncologic outcomes after curative-intent surgery.

Patients with high-grade extremity or truncal STS treated between 2004 and 2020 were retrospectively analysed. Minimal tumour–vessel distance (> 1 cm vs. ≤ 1 cm) and Fujiwara type (I–IV) were determined from preoperative MRI. OS was evaluated using Kaplan–Meier and Cox regression. Multivariable Cox models were pre-specified to include established prognostic confounders (age, tumour size, histologic grade, and resection status), irrespective of univariate significance. Fujiwara types III and IV were pooled to improve estimate stability.

A total of 113 patients met inclusion criteria (median follow-up 84 months); 47 deaths occurred during follow-up. Tumour proximity ≤ 1 cm was associated with significantly reduced OS (42 vs. 107 months; p = 0.003). The Fujiwara classification was associated with significant survival differences across categories (p < 0.001).

In multivariable analysis (Model A), proximity ≤1 cm remained independently associated with OS (HR 3.04; 95% CI 1.38–6.67; p = 0.006), together with age ≥65 years (HR 3.42; p = 0.007) and R2 resection (HR 9.67; p = 0.002). In Model B, pooled Fujiwara type III/IV remained independently associated with impaired OS (HR 3.62; 95% CI 1.47–8.94; p = 0.005), alongside age ≥65 years (HR 3.07; p = 0.015) and R2 resection (HR 7.32; p = 0.007). Higher Fujiwara types correlated with increased local recurrence (p = 0.011), while distant metastasis rates were similar across groups.

Tumour proximity to neurovascular structures is independently associated with OS in high-grade deep STS. The Fujiwara classification remained prognostically relevant after adjustment for established risk factors including resection status. These findings suggest that anatomical tumour–vessel relationships provide complementary prognostic information beyond conventional clinicopathologic variables. Prospective multicentre validation is warranted.