Rule-derived preoperative inflammatory-lymphocyte recovery phenotype and pathological response in resected osteosarcoma: a retrospective cohort study.

Background: Pathological response after neoadjuvant chemotherapy is a key treatment milestone in osteosarcoma. We evaluated whether a rule-derived preoperative peripheral inflammatory-lymphocyte recovery phenotype constructed from routinely collected laboratory measurements was associated with poor pathological response. Methods: This single-center retrospective cohort included patients with osteosarcoma who completed neoadjuvant chemotherapy, underwent definitive surgery, had pathological response assessed, and had valid baseline (T0) and preoperative (T3) laboratory measurements. A failure score combined standardized myeloid-inflammatory burden and lymphocyte-nutritional reserve. Patients were classified as favorable recovery, intermediate recovery phenotype, or persistent recovery failure using the primary T0-T3 rule. The primary outcome was tumor necrosis <90%. The primary association was estimated using parsimonious Firth logistic regression, with modified-Poisson risk ratios and standardized absolute risks. Model comparisons used nested full-pipeline repeated cross-validation. Event-related analyses were exploratory. Results: Among 162 patients, 88 (54.3%) had poor pathological response; 65 were classified as favorable recovery, 45 as intermediate, and 52 as persistent recovery failure. Persistent failure was associated with poor pathological response versus favorable recovery (Firth OR, 5.72; 95% CI, 2.42-13.53; P<0.001; adjusted RR, 2.10; 95% CI, 1.46-3.01). Standardized risks were 37.2% for favorable recovery and 75.7% for persistent failure, corresponding to a risk difference of 38.6 percentage points (95% CI, 19.7-52.7). In nested full-pipeline cross-validation, the phenotype model had an AUC of 0.641 and did not outperform the continuous T3 score (AUC, 0.665); its out-of-fold calibration slope was 0.587 (95% CI, 0.220-0.953). Exploratory associations were observed for event-free survival (EFS) and pulmonary metastasis, whereas overall survival (OS) was immature with 24 deaths. Conclusion: A rule-derived preoperative peripheral inflammatory-lymphocyte recovery phenotype was associated with pathological response among patients who completed neoadjuvant chemotherapy and underwent definitive surgery for osteosarcoma. The association was robust across alternative definitions and sensitivity analyses; however, the categorical phenotype did not outperform the continuous preoperative T3 score, and out-of-fold calibration indicated residual overfitting. Event-related findings and biological interpretation require external validation.