Predictive Nomogram for Hyperprogressive Disease During Anti-PD-1/PD-L1 Treatment in Patients With Advanced Non-Small Cell Lung Cancer

Background: Immune-checkpoint inhibitors treatment (ICIs) like anti-PD-1/PD-L1 checkpoint inhibitors has been considered as a standard treatment and produced significant effects in patients with non-small cell lung cancer (NSCLC). However, various studies have reported that anti-PD-1/PD-L1 treatment may lead to the rapid development of tumors called hyperprogressive disease (HPD). It is urgent to determine biomarkers and develop a nomogram for HPD prediction in patients with NSCLC before anti-PD-1/PD-L1 treatment. Methods: This retrospective cohort study included 176 cases (collected from January 1, 2016 to September 31, 2019) for establishing a model of HPD prediction, and 37 cases (collected from October 1, 2019 to April 31, 2020) for validation in the patients with advanced NSCLC treated with PD-1/PD-L1 inhibitors in Sun Yat-sen University Cancer Center. HPD was defined as tumor growth rate (TGR, ≥ 2), or tumor growth kinetics (TGK, ≥ 2) or time to treatment failure (TTF, ≤ 2 months). Univariate and multivariate logistic regression were used to estimate the specified factors associated with HPD. Then the nomogram was developed and validated. Finding: The anti-PD-1/PD-L1 therapy resulted in 9.66% (17/176) incidences of the HPD phenomenon in advanced NSCLC. The overall survival (OS) and progression free survival (PFS) in patients with HPD were significantly shorter than those patients without HPD (OS: 8.82 months [95%CI, 4.91-12.74 months] vs. 15.74 months [95%CI, 13.95-17.53 months], P<0.01; PFS: 2.29 months [95%CI, 1.90-2.69 months] vs. 11.06 months [9.12-12.85 months], P<0.001), respectively. The HPD prediction nomogram included APTT (P<0.01), CD4+CD25+ cells (Treg cells) (P<0.01), the presence of liver metastasis (P<0.05), and more than two metastatic sites (P<0.05). Then patients could be divided into two groups by “HPD score” calculated by the nomogram: “HPD score” = 0.8225× presence of liver metastasis + 0.3954 × more than two metastatic sites + 0.2140 × APTT+0.0873 × Treg cells – 11.2476, with the cut-off point was -1.37 (patients with “HPD score” ≥ -1.37 have high risk of HPD, while patients with “HPD score” < -1.37 have low risk of HPD). The C-index was 0.845, while the area under the curve (AUC) was 0.830 (95% confidence interval [CI]: 0.719-0.950, sensitivity 75.00%, specificity 91.70%). The calibration plot of HPD probability showed an optimal agreement between the actual observation and prediction by nomogram. The good discrimination [AUC, 0.911, 95% CI (0.745-0.950), sensitivity 79.60%, specificity 87.50%] was still exited in the validation cohort. The decision curve analysis showed that the nomogram was more beneficial to patients than the reported biomarkers (including PD-L1, number of metastases, age ≥ 65). Interpretation: The nomogram was constructed with the presence of liver metastasis, more than two metastatic sites, long APTT and high level of Treg cells, which could be used to predict HPD risk. Thus NSCLC patients could receive reasonable management before anti-PD-1/PD-L1 therapy. Funding Statement: This study was funded by National Natural Science Foundation of China (No. 81473233); Science and Technology Program of Guangzhou (No. 201504010038, 201604020079, 201601010008). Declaration of Interests: The authors declare no conflict of interest. Ethics Approval Statement: Written informed consents were exempted because of retrospective analysis. The Institute Research Ethics Committee of the Sun Yat-Sen University Cancer Center, Guangzhou, China approved this study (IRB No. YB2020-006-01).