Pan-immune-inflammation and its dynamics: predictors of survival and immune-related adverse events in patients with advanced NSCLC receiving immunotherapy

The study retrospectively enrolled patients with aNSCLC who were consecutively treated at the Affiliated Hospital of Nantong University, China, between January 2019 and December 2022. All patients had undergone immunotherapy with nivolumab 3 mg/kg every 2 weeks or pembrolizumab 200 mg every 3 weeks until unacceptable toxicity, or death from any cause. Patients who only received other antitumor therapies, such as chemotherapy, targeted therapy, or radiotherapy, were excluded from the study. Only patients with a minimum follow-up of 12 months (until December 2022) were enrolled, owing to the high probability of developing irAEs within the 1– 6 months of immunotherapy [30] and the need for sufficient time for survival assessment.

Data on the following patients’ characteristics were extracted: age, sex, smoking history, histology, cancer stage, and the number of metastatic sites. In addition, laboratory data such as complete blood cell counts (e.g., neutrophils, lymphocytes, monocytes, and platelets) at baseline (within 7 days before the administration of immunotherapy, defined as T1) and at the endpoint (3–4 weeks after the first dose, defined as T2) were extracted from the electronic medical records. PIV was calculated as (neutrophil count × platelet count × monocyte count)/lymphocyte count. SII was calculated as (neutrophil count × platelet count)/lymphocyte count. NLR was calculated as neutrophil count /lymphocyte, and PLR was calculated as platelet count/lymphocyte count. MLR was calculated as monocyte count/lymphocyte count and d-NLR as neutrophil count/(leucocytes count – neutrophil count). The cutoff values for PIV, SII, NLR, PLR, MLR, and d-NLR at baseline as dichotomous variables were intercepted according to the receiver-operating characteristic curve (PIV < 288.1; SII < 784.1; NLR < 3.163; PLR < 0.420; dNLR < 2.316). The dynamic change in PIV was calculated by subtracting PIV (T1) from the absolute PIV (T2). The cutoff values for the dynamic changes in PIV as dichotomous variables were determined using the X-title software (PIV dynamics < 608.2). PFS was determined as the duration from the date of treatment initiation to the date of disease progression or patient death from any cause. OS was defined as the duration from the date of treatment initiation to the date of death from any cause. Tumor responses were assessed using the Response Evaluation Criteria in Solid Tumors guidelines (version 1.1) every 10 ± 2 weeks. This study was approved by the board/ethics committee of the Affiliated Hospital of Nantong University, and exception to the requirement of informed consent was approved (Ethic Number: 2018-K020).

Variability was compared using Fisher’s exact test for categorical variables and Mann-Whitney or Wilcoxon signed-rank tests for continuous variables. Box charts and line charts were used to illustrate the differences and trends between the two important time points of PIV. The association between blood biomarkers and the onset of irAEs was analyzed by logistic regression models. For data with binary data as the outcome variable, the cut-off value of the continuous independent variable is generally determined by ROC analysis. For dichotomous outcome variables with temporal dimensions, we use X-title to determine the cut-off value [31]. Cox regression models were used to determine the risk factors for PFS and OS. Moreover, PFS and OS were estimated using the Kaplan-Meier method and were compared using the log-rank test. Statistical significance was defined as a two-sided p-value of < 0.05. All statistical analyses were performed using X-tile versions 3.6.1, SPSS 20.0 (SPSS Inc., Chicago, IL) and GraphPad Prism version 9.2.0 (GraphPad Inc., San Diego, CA).

The clinical characteristics of 269 patients are outlined in Table 1. The median age of patients was 67 (41–87) years, with 232 (86.2%) patients being male. The most common histological type of NSCLC was adenocarcinoma (n = 164; 61.0%), followed by squamous carcinoma (n = 104; 38.6%) and a rare type of sarcoma (n = 1; 0.4%). Of all patients, 18 (6.7%) had an Eastern Cooperative Oncology Group performance status of 2. Overall, 215 (55.2%) patients had undergone testing for tumor PD-L1 expression; of these, 147 (68.4%) and 68 (31.6%) patients tested positive and negative for PD-L1, respectively. The median number of times patients underwent treatment at baseline was five, with 168 (62.5%) patients being treated with pembrolizumab and 101 (37.5%) with nivolumab. For blood biomarkers, the median PIV was 387 (17.8–9420), NLR was 3.37 (0.844–79.7), PLR was 164 (36.4–1490), SII was 682 (142.0–17300.0), PLR was 164 (36.4–1490), and dNLR was 2.07 (-1.34–31.4).

In total, 89 (33%) patients reported irAEs (Table 2), with the most common being hyperthyroidism/hypothyroidism (19.7%), skin-related events (17.4%), liver injury (15.1%), and enteritis/diarrhea (15.1%). Among organ-related toxicities, dermatological and endocrine toxicities were the most commonly reported irAEs, followed by hepatologic and digestive disorders. A total of 89 events of all-grade irAEs were documented, of which 12 were grade ≥ 3 in severity. The most common grade ≥ 3 irAEs were rash and adrenal insufficiency. The grade ≥ 4 irAEs included myocarditis and complex infections because of hypoimmunity. However, very few grade ≥ 4 events were observed; but these may be resulted in death. The majority of patients with grade ≥ 3 irAEs were treated with steroids, and for a few patients with higher grade irAEs, immunotherapy was discontinued permanently. All patients who received immunosuppressive therapies achieved alleviation of their irAEs.

Findings regarding the association of irAE onset with baseline clinical features and with blood parameters are listed in Table 3. None of the baseline clinical features affected the overall risk of irAEs. For example, an increased number of treatment cycles was not significantly associated with a higher probability of developing irAEs (odds ratio [OR]: 0.969; 95% confidence interval [CI]: 0.917–1.024; p = 0.262). Moreover, no association was noted between PD-L1 expression and irAE onset (OR: 0.787; 95% CI: 0.406–1.524; p = 0.477). However, the following blood markers showed a significant correlation with the occurrence of irAEs: L-PIV (OR: 0.112; 95% CI: 0.062–0.202; p < 0.001), L-SII (OR: 0.129; 95% CI: 0.067–0.250; p < 0.001), L-NLR (OR, 0.294; 95% CI, 0.171–0.504; p < 0.001), L-PLR (OR: 0.164; 95% CI: 0.092–0.293; p < 0.001), L-MLR (OR: 0.229; 95% CI: 0.129–0.407; p < 0.001), and L-dNLR (OR: 0.299; 95% CI: 0.165–0.541; p < 0.001). Finally, multivariate analysis confirmed that L-PIV (OR: 0.235; 95% CI: 0.117–0.472; p < 0.001; Table 3), L-SII (OR: 0.393; 95% CI: 0.175–0.883; p = 0.024), and L-PLR (OR: 0.483; 95% CI: 1.055–4.500; p = 0.035) were the independent predictors of irAEs.

In addition, determining collinearity using the variance inflation factor (PIV: 1.976; SII: 2.511; NLR: 2.777; PLR: 1.571; MLR: 1.691; dNLR: 2.391) revealed the absence of multicollinearity for blood biomarkers (Table 3).

Specific variables identified as by univariate analysis included the number of treatments with ICIs, PIV, SII, NLR, PLR, MLR, and dNLR (p < 0.05; Table 4). These variables were further analyzed using the multivariate model. Finally, only the number of treatments with ICIs (HR = 0.950, 95% Cl: 0.920–0.982, p = 0.002) and higher PIV (HR = 1.707, 95% Cl: 1.275–2.286, p < 0.001) were the independent prognostic factors for poor median PFS (Table 4).

A total of 196 tumor progression events were observed in 1 year of follow-up, with the median PFS being 7 months. The tumor progression events rate was 72.9%, with the median PFS being significantly longer in patients with low PIV (10 months, 95% CI: 8.6–11.4; Fig. 1A).

Univariate analysis demonstrated that stage, number of metastatic sites, PIV, SII, NLR, PLR, MLR, and dNLR were associated with OS (p < 0.05, Table 4). However, in multivariate analysis, only low PIV (HR = 2.414, 95% Cl: 1.509–3.862, p < 0.0001) and the number of metastatic sites (< 3, HR = 0.602, 95% Cl: 0.375–0.966, p = 0.035) were independently associated with longer survival outcomes (Table 4).

A total of 88 deaths occurred during the follow-up period, with the OS rate being 67.3%. The median OS for patients with low PIV was 29 months compared with 21 months in patients with high PIV (95% CI: 19.4–22.5, p < 0.0001; Fig. 1B).

A box plot plotted for comparing irAEs patients with PIV at T1 and T2 revealed a mean of 191.6 vs. 219.4 and an interquartile range (IQR) of 176.9 vs. 539.7. Moreover, the box plot suggested a significant difference in the association of PIV with irAEs between T1 and T2 (Fig. 2A, p = 0.006). For patients with no irAEs, the median of the box plot at T1 and T2 was 527.5 and 409.5, with the IQR being 620.8 and 727.9, respectively; however, these values did not significantly differ between the two time points (p = 0.056; Fig. 2B). In addition, PIV dynamics were compared between patients with irAEs and without irAEs using box plot analysis. The median obtained for patients with irAEs and without irAEs was 190.0 and 317.0, with an IQR of 447.6 and 550.3, respectively. This indicated a significant difference in PIV dynamics between the two groups (p = 0.001; Fig. 2C). The line chart (Fig. 2D) shows the changes in PIV with follow-up. The average change in PIV in patients with irAEs was 350 (Fig. 2D) and in those without irAEs was 678 (Fig. 2E). PIV dynamics were significantly lower in patients with irAEs than in those without irAEs.

The median PFS was longer for patients with a smaller PIV change (9 months; 95% CI: 7.9–10.0) than for patients with a greater PIV change (6 months; 95% CI: 4.8–7.1; log-rank test p = 0.0119; Fig. 1C). Likewise, median OS was significantly longer for patients with a smaller PIV change (28 months; 95% CI: 24.3–31.8) than for patients with a greater PIV change (21.6 months; 95% CI: 18.6–23.3; log-rank test p = 0.0021; Fig. 1D).