Purpose: This study aims to identify accurate and universal prognosis prediction tools for NB (neuroblastoma), and to explore potential prognostic factors further. This will assist clinicians in accurately assessing the prognosis of affected children, ultimately improving the survival rate and quality of life for NB patients.
Method: Collected basic clinical data of children diagnosed with NB in our center (January 2015 - June 2024) and at 17 registration areas (2000-2020) in the surveillance, epidemiology, and end results database. Based on the above multi-center clinical data and 386 machine-learning algorithms, a prognostic nomogram was created to identify the optimal prediction model and potential prognostic predictors.
Results: A total of 1214 NB children were split into a training set (1000 cases) and a validation set (214 cases). Multivariate Cox regression identified age, tumor size, distant metastasis, chemotherapy, and primary lesion in the adrenal gland as independent risk factors for the prognosis of NB (P < 0.05). The average area under the curves of two data sets for the nomogram using RSF (random survival forests) achieved 0.86, 0.91, 0.92, and 0.93 in predicting 1/3/5/10-year survival rates, respectively. The RSF-based nomogram score correlated significantly with tumor size before treatment (R = 0.51), NSE (neuron-specific enolase) before treatment (R = 0.41), postoperative inflammatory markers [platelet-to-lymphocyte ratio (R = 0.26), neutrophil-lymphocyte ratio (R = 0.24), and systemic immune inflammation index (R = 0.21)], RBC (red blood cell count) before discharge (R = -0.28), and Hb (hemoglobin) before discharge (R = -0.38) (P < 0.05).
Conclusion: The RSF-based nomogram model shows superior predictive ability for NB prognosis compared to similar models. Its use of accessible indicators makes it suitable for resource-poor areas and families with limited finances. Tumor size, NSE, inflammatory markers, RBC, and Hb may be potential prognostic predictors for evaluating NB prognosis.
Keywords: neuroblastoma, children, prognosis, machine-learning, nomogram
Anahtar Kelimeler: