《中国医学装备》杂志社

目的：融合正电子发射计算机体层扫描(PET/CT)代谢参数构建临床病理特征的Ⅲ期非小细胞肺癌(NSCLC)多维度预后模型，并验证其在个体化治疗决策中的临床应用价值。方法：回顾性纳入2019年1月至2023年12月期间保定市第一中心医院接受规范化治疗的150例Ⅲ期NSCLC患者的病例资料，采集所有患者PET/CT代谢参数最大标准化摄取值(SUV_max)、代谢肿瘤体积(MTV)、总病灶糖酵解(TLG)以及表皮生长因子受体(EGFR)突变、中性粒细胞与淋巴细胞比率(NLR)、细胞角蛋白19片段(CYFRA21-1)等临床病理特征、临床评分及病理分型进行分析。按7∶3比例将150例患者分为训练集(105例)和验证集(45例)。采用最小绝对收缩和选择算子(LASSO)回归筛选关键变量，构建Cox比例风险模型，通过原始数据集Bootstrap重采样(1 000次)进行内部验证，并计算一致性指数(C-index)、校准曲线，采用决策曲线分析(DCA)评估模型性能。结果：多因素分析显示，全身代谢肿瘤体积(MTV_wb)每增加1 cm³，死亡风险升高0.9%(HR=1.009,95%CI:1.003～1.015,P=0.006);NLR≥3.5(HR=2.107,95%CI:1.424～3.119,P=0.001)、CYFRA 21～1≥5.0 ng/ml(HR=1.735,95%CI:1.161～2.594,P=0.012)为独立危险因素，EGFR突变阳性患者死亡风险降低40.1%(HR=0.599,95%CI:0.387～0.928,P=0.022)。模型在训练集和验证集的C-index分别为0.813(95%CI:0.761～0.865)和0.795(95%CI:0.730～0.860)，预测模型校准曲线斜率接近1(0.931～1.018)，预测模型1年生存预测误差≤1.52%。决策曲线分析显示，当阈值概率为0.2～0.6时，模型净获益值最高达0.469。全身总病灶糖酵解(TLGwb)、程序性死亡配体-1(PD-L1)表达及性别均与生存无相关性(P>0.05)。结论：本研究构建的预后模型通过整合代谢负荷与宿主炎性指标，显著提升了Ⅲ期NSCLC生存预测精度，为个体化随访周期制定及治疗策略优化提供了依据。

Objective: To construct a multidimensional prognostic model for stage Ⅲ non-small cell lung cancer(NSCLC) by integrating metabolic parameters of positron emission tomography/computed tomography(PET/CT) with clinicopathological features, and to validate its clinical application value in individualized treatment decision-making. Methods: A retrospective analysis was conducted on case data of 150 patients with stage Ⅲ NSCLC who received standardized treatment at the Baoding NO.1 Central Hospital, Hebei Province, between January 2019 and December 2023. The PET/CT metabolic parameters [maximum standardized uptake value(SUV_max), metabolic tumor volume(MTV), total lesion glycolysis(TLG)], and clinicopathological features [epidermal growth factor receptor(EGFR) mutation, neutrophil-to-lymphocyte ratio(NLR), cytokeratin 19 fragment(CYFRA21-1) ], clinical scores, and pathological subtypes of all patients were collected. Patients were randomly divided into a training set(n=105) and a validation set(n=45) as a 7:3 ratio. Key variables were screened using the least absolute shrinkage and selection operator(LASSO) regression. A Cox proportional hazards model was subsequently constructed. Internal validation was performed via Bootstrap resampling(1,000 iterations) from original data set. Model performance was assessed by calculating the concordance index(C-index), calibration curves, and decision curve analysis(DCA). Results: Multivariate analysis identified the following independent risk factors: death's risk increased 0.9% with increasing per 1 cm3 whole-body metabolic tumor volume(MTV_wb) [hazard ratio(HR)=1.009, 95% confidence interval(CI): 1.003-1.015, P=0.006], NLR ≥3.5(HR=2.107, 95%CI: 1.424-3.119, P=0.001), and CYFRA21-1 level ≥5.0 ng/ml(HR=1.735, 95%CI: 1.161-2.594, P=0.012). Conversely, patients with positive EGFR mutations demonstrated a 40.1% reduction in mortality risk(HR=0.599, 95%CI: 0.387-0.928, P=0.022). The C-index values of model were respectively 0.813(95%CI: 0.761-0.865) in the training set and 0.795(95%CI: 0.730-0.860) in the validation set. The slope of the calibration curve for the predictive model was close to 1(range: 0.931-1.018). The prediction error of 1-year survival was ≤1.52%. Decision curve analysis indicated that the highest net benefit value of the model reached to 0.469 when the range of threshold probability was 0.2-0.6. The expression of whole-body total lesion glycolysis(TLG_wb) and programmed death-ligand 1(PD-L1) did not appear correlation with gender and survival(P>0.05). Conclusion: The prognostic model that is constructed by this study, which integrates metabolic burden and host inflammatory markers, can significantly enhance predictive accuracy for survival of patients at stage Ⅲ NSCLC. It can provide valuable basis for formulating individualized follow-up period and optimizing therapeutic strategies.