Scholarlink

Introduction <p>Neoadjuvant immunotherapy has demonstrated promising efficacy in resectable lung cancer; however, substantial interindividual variability in the treatment response remains. Although programmed death-ligand 1 (PD-L1) expression and tumor mutational burden (TMB) are established biomarkers of immunotherapy response, the performance of predictive models that integrate these biomarkers remains unclear.</p> Aim <p>To develop and validate predictive models integrating PD-L1 expression and TMB for estimating pathological complete response (pCR) following neoadjuvant immunotherapy in patients with resectable lung cancer and to compare the performance of a nomogram and a random forest model.</p> Method <p>This retrospective cohort study included 232 patients who underwent surgical resection following neoadjuvant immunotherapy as the model development cohort and were randomly assigned to the training and internal validation sets (7:3). A temporally independent validation cohort of 115 consecutive patients treated at the same institution during a later study period was used to assess the generalizability of the model. Univariate and multivariable logistic regression analyses were performed to identify independent predictors of pCR. A nomogram and random forest model were developed and evaluated in terms of discrimination, calibration, and classification performance.</p> Results <p>Clinical stage II (odds ratio [OR] 0.54, 95% confidence interval [CI] 0.31–0.94), absence of lymph node metastasis (OR 0.35, 95% CI 0.16–0.77), receipt of immunotherapy combined with chemotherapy (OR 4.00, 95% CI 1.54–10.39), higher PD-L1 expression (OR 1.62, 95% CI 1.03–2.54), and higher TMB (OR 3.07, 95% CI 1.42–6.65) were independently associated with pCR (all <i>P</i> &lt; 0.05). In the random forest (RF) model, PD-L1 expression and TMB were the most influential predictors. The RF model demonstrated superior discrimination compared with the nomogram in both the training cohort (area under the receiver operating characteristic curve [AUC] 0.866 vs. 0.782; <i>P</i> &lt; 0.001) and the internal validation cohort (AUC 0.769 vs. 0.669; <i>P</i> &lt; 0.001), with better overall classification performance. In the temporally independent validation cohort, the RF model achieved an AUC of 0.757 (95% CI 0.682–0.902), a sensitivity of 82.6%, and a specificity of 76.2%, although the calibration was suboptimal.</p> Conclusion <p>PD-L1 expression and TMB were independently associated with pathological response to neoadjuvant immunotherapy in patients with resectable lung cancer. The predictive model incorporating these biomarkers demonstrated acceptable performance, with the RF model outperforming the nomogram. However, the model should be considered exploratory and requires prospective multicenter external validation before routine clinical application.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development and validation of a PD-L1 and tumor mutational burden-based predictive model for pathological complete response after neoadjuvant immunotherapy in resectable lung cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月29日

查看全文
  • Xiao Wu,
  • Duo Jiang,
  • Bin Li

摘要

Introduction

Neoadjuvant immunotherapy has demonstrated promising efficacy in resectable lung cancer; however, substantial interindividual variability in the treatment response remains. Although programmed death-ligand 1 (PD-L1) expression and tumor mutational burden (TMB) are established biomarkers of immunotherapy response, the performance of predictive models that integrate these biomarkers remains unclear.

Aim

To develop and validate predictive models integrating PD-L1 expression and TMB for estimating pathological complete response (pCR) following neoadjuvant immunotherapy in patients with resectable lung cancer and to compare the performance of a nomogram and a random forest model.

Method

This retrospective cohort study included 232 patients who underwent surgical resection following neoadjuvant immunotherapy as the model development cohort and were randomly assigned to the training and internal validation sets (7:3). A temporally independent validation cohort of 115 consecutive patients treated at the same institution during a later study period was used to assess the generalizability of the model. Univariate and multivariable logistic regression analyses were performed to identify independent predictors of pCR. A nomogram and random forest model were developed and evaluated in terms of discrimination, calibration, and classification performance.

Results

Clinical stage II (odds ratio [OR] 0.54, 95% confidence interval [CI] 0.31–0.94), absence of lymph node metastasis (OR 0.35, 95% CI 0.16–0.77), receipt of immunotherapy combined with chemotherapy (OR 4.00, 95% CI 1.54–10.39), higher PD-L1 expression (OR 1.62, 95% CI 1.03–2.54), and higher TMB (OR 3.07, 95% CI 1.42–6.65) were independently associated with pCR (all P < 0.05). In the random forest (RF) model, PD-L1 expression and TMB were the most influential predictors. The RF model demonstrated superior discrimination compared with the nomogram in both the training cohort (area under the receiver operating characteristic curve [AUC] 0.866 vs. 0.782; P < 0.001) and the internal validation cohort (AUC 0.769 vs. 0.669; P < 0.001), with better overall classification performance. In the temporally independent validation cohort, the RF model achieved an AUC of 0.757 (95% CI 0.682–0.902), a sensitivity of 82.6%, and a specificity of 76.2%, although the calibration was suboptimal.

Conclusion

PD-L1 expression and TMB were independently associated with pathological response to neoadjuvant immunotherapy in patients with resectable lung cancer. The predictive model incorporating these biomarkers demonstrated acceptable performance, with the RF model outperforming the nomogram. However, the model should be considered exploratory and requires prospective multicenter external validation before routine clinical application.