Comprehensive nomogram models for prediction of checkpoint inhibitors pneumonia

doi:10.21203/rs.3.rs-1668092/v1

PDF

Research Article

Comprehensive nomogram models for prediction of checkpoint inhibitors pneumonia

https://doi.org/10.21203/rs.3.rs-1668092/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted 24 May, 2022

You are reading this latest preprint version

Introduction: Checkpoint inhibitors pneumonia (CIP) is a common type of irAEs with poor clinical prognosis. Currently, there is a lack of effective biomarkers and predictive models to predict the occurrence of CIP.

Materials and Methods: The study retrospectively included 327 patients who received immunotherapy for the first time as a training cohort and 167 patients as a validation cohort. Patients were assessed for the occurrence of CIP, and were divided into any grade CIP cohort, grade ≥2 CIP cohort, and grade ≥3 CIP cohort. Multivariate logistic regression analyses were used to determine the independent risk factors. Based on that, predictive nomograms were constructed and validated to predict the risk of CIP.

Results: The overall incidence of CIP was 18.61%. Based on the results of multivariate analysis, we established nomogram models for predicting any grade CIP and grade ≥2 CIP, respectively. For the nomogram to predict any grade CIP, the C index of the model in the training cohort and validation cohort were 0.814 (95%CI=0.760-0.870) and 0.919 (95%CI=0.842-0.941), respectively. Similarly, as for the nomogram of grade 2 or higher CIP, the C index of the model in the training cohort and validation cohort were 0.879 (95%CI=0.81-0.92) and 0.936 (95%CI=0.812-0.982), respectively.

Conclusions: After internal verification and external verification, the predictive power of nomogram models is satisfactory and they are expected to be a convenient, visual, and personalized clinical tool for assessing the risk of developing CIP in patients receiving ICIs treatment.

Immunotherapy

Biomarkers

Nomogram model

Checkpoint inhibitors pneumonia

Since Ipilimumab was approved by the FDA in 2011 for the treatment of malignant melanoma, the field of tumor immunotherapy has enjoyed a great renaissance[1]. In recent years, immunotherapy, represented by anti-PD-1 /PD-L1 checkpoint inhibitors, has been widely used in a variety of tumors, achieving satisfactory efficacy and opening up a new era of tumor therapy, such as melanoma[2] and non-small cell lung cancer[3]. However, people's enthusiasm for immunotherapy is largely based on its long-term clinical efficacy, and the fact is that the clinical benefits of immunotherapy depend not only on its anti-tumor effectiveness, but also on its adverse events. Adverse events due to the unique antitumor mechanism of immune checkpoint inhibitors (ICIs) are called immune-related adverse events (irAEs). It is a nonspecific reaction produced by the immune system and can affect almost all tissues and organs[4], which is common in the skin, intestinal, endocrine organs, lungs and skeletal muscle tissues, while the heart, liver, kidney, nerve, and eye of irAEs are relatively rare[5–7].

Checkpoint inhibitor pneumonitis (CIP) is a type of irAEs with a poor prognosis. The incidence of CIP has been reported in previous clinical studies at approximately 3–5% [8, 9], but several real-world studies have shown that it can be as high as 10–20%[10], among which the fatality rate of severe CIP is high, up to 14–35%[4]. Mild CIP may have a good prognosis, however, severe CIP has a poor prognosis due to its single clinical treatment[4]. Although studies have indicated that mild to moderate irAEs are associated with better outcomes with ICIs therapy[11, 12], the severe irAEs may lead to the forced termination of ICIs, even death of the patient. And CIP has delayed radiological changes and lacks specific manifestations in the early stage, which makes it difficult to differentiate from other respiratory diseases and easily lead to missed diagnosis. Besides, the clinical treatment of CIP is difficult. Once CIP occurs, the disease deteriorates rapidly. Hormone intervention is the main treatment at present, but steroids may bring other adverse reactions and the risk of tumor progression to patients. Therefore, for patients who are susceptible to serious CIP, immunotherapy does not prolong survival, but rather endangers life. The quality of life of patients with severe CIP is significantly reduced, the risk of death is significantly increased, and the cost of health care is also significantly increased. Therefore, how to reduce the incidence of CIP, especially severe CIP, is the key issue to improve the efficacy and safety of immunotherapy in our clinical practice.

At present, there is still a lack of authoritative and unified judgment method for the screening and early warning of CIP high-risk groups. Previous studies have manifested that advanced age, smoking history, non-small cell lung cancer, especially lung squamous cell carcinoma, interstitial lung disease and emphysema at baseline are high risk factors for CIP[10, 13–16]. Salahaldin A. Tahir reported that the increase of CD74 autoantibody level was significantly correlated with the occurrence of CIP[17]. CD74 is an autoantibody active protein, mainly expressed on the cell membranes of immune cells including macrophages, which can stimulate the release of inflammatory mediators and participate in specific humoral immune response[18]. Su Chunxia et al. showed that the baseline peripheral blood eosinophils of patients in the CIP group were significantly higher than those in the non-CIP group[19]. Unfortunately, in general, there are still some limitations in the preliminary studies. The single dimension prediction has different emphasis, and its clinical prediction ability is limited, with poor specificity and low stability. There is a lack of effective biomarkers and comprehensive predictive models to predict the risk of CIP and screen out the susceptible population of CIP.

It can be seen from the above that screening the susceptible population of CIP and early prediction of its occurrence so as to avoid the development of severe CIP has become an urgent clinical problem to be solved. So the purpose of this study is to explore new biomarkers that can predict CIP, and to establish a novel nomogram model to predict the incidence of CIP, so as to reduce the risk of toxicity and optimize the benefits of patients.

1. Study population

We selected all malignant tumor patients who received ICIs (including Nivolumab, Pembrolizumab, Camrelizumab, Toripalimab, Atezolizumab and Sintilimab) for the first time from January 2019 to January 2021 in the First Affiliated Hospital of Xi'an Jiaotong University. The diagnosis of CIP was made by three professional clinical oncologists combined with two professional radiologists for double-blind review, which based on the guidelines for management of immunotherapy-related toxicities published by the national comprehensive cancer network in 2020[20], the clinical practice guidelines for the management of toxicity from toxicities published by the European society of clinical oncology[21] and the consensus of relevant domestic experts[22]. If their final diagnosis is inconsistent, a third experienced radiologist will be invited to re-evaluate to minimize the risk of misdiagnosis or missed diagnosis. CIP was classified according to common terminology criteria for adverse events version 4.0. In clinical practice, grade 1 CIP mostly has no obvious clinical manifestations, and can only be detected by imaging or other clinical examinations. Moreover, for grade 1 CIP, multiple clinical management guidelines and expert consensus recommend that ICIs treatment should not be suspended and close follow-up should be sufficient. However, grade 2 or higher CIP has great adverse effects on patients, which requires not only steroid therapy, but also suspension or even permanent cessation of ICIs treatment. Moreover, the clinical treatment method of CIP is single and the therapeutic effect is poor, especially for severe grade 3 or higher CIP patients, the risk of clinical mortality and disability is high[4]. Therefore, in order to prospectively screen or identify patients with grade ≥ 2 and grade ≥ 3 CIP, we established three cohorts based on the CTCAE grade: (1) any grade CIP cohorts (experimental group: patients with CIP, control group: patients without CIP); (2) Grade ≥ 2 CIP cohort (experimental group: patients with grade 2 or higher CIP, control group: patients without or with grade 1 CIP); (3) Grade ≥ 3 CIP cohort (experimental group: patients with grade 3 or higher CIP, control group: patients without or with grade 1–2 CIP). Inclusion criteria for patients were as follows: (1) Age ≥ 18 years; (2) Patients with histologically or cytologically proven malignant solid tumors can be treated with ICIs as assessed by professional oncologists, regardless of cancer type or stage; (3) The Eastern Oncology Collaboration Group (ECOG) physical status score was 0–2 when receiving immunotherapy; (4) No previous use of immunotherapy; (5) No previous exposure to immune-mediated therapy. Exclusion criteria were: (1) Previous target lesions received immune-related therapy or immune-mediated therapy; (2) A history of another primary malignancy; (3) Lack of baseline characteristic and radiologic images. The detailed flow chart is shown in Fig. 1.

2. Data collection

We collect all the baseline characteristics of patients, such as name, gender, age, diagnosis, and so on. At the same time, we also collect baseline of peripheral blood in patients with C reactive protein (CRP), white blood cells, neutrophil count (NEUT), absolute eosinophil count (AEC), absolute lymphocyte count (ALC), platelet count (PLT) and serum albumin levels. And the platelet/lymphocyte ratio (PLR), neutrophils/lymphocyte ratio (NLR), systemic immune-inflammation index (SII) (%) and prognostic nutritional index (PNI) (%) were calculated by following methods: PLR = PLT(×10⁹ cells/L)/ lymphocyte count(×10⁹ cells/L), NLR = NEUT(×10⁹ cells/L)/lymphocyte count(×10⁹ cells/L), SII = PLT(×10⁹ cells/L)× NEUT(×10⁹ cells/L)/ lymphocyte count(×10⁹ cell/L), PNI = serum albumin level(g/L) + 5× lymphocytes count (×10⁹ cells/L). Patients were assessed for interstitial lung disease (ILD) and emphysema at baseline before receiving ICIs.

3. Statistical analysis

Statistical analysis was performed using SPSS 22.0 statistical Software package (SPSS, Inc., Chicago, IL, USA) and GraphPad Prism 8.0 (GraphPad Software, La Jolla, CA, USA). Categorical variables are transformed into dichotomous variables, and continuous variables are expressed by concrete values. Age was the continuous variable. Then, 60 years old was taken as the cut-off value and divided into the elderly group (≥ 60 years old) and the non-elderly group (< 60 years old) to form the dichonomical variable. Receiver operating characteristic (ROC) curve and Youden index were used to determine the optimal cut-off value of laboratory indicators for continuity variables. The t-test, chi-square test or Mann-Whitney U test were used for comparison between groups, and Fisher's exact test was used when necessary. Multivariate logistic regression analysis was used to determine the independent risk factors directly related to the incidence of CIP. P < 0.05 (double-tailed) was considered statistically significant.

R Foundation for Statistical Computing (Vienna, Austria) version 3.6.3 was used to establish the Nomogram model, calibration curve, ROC curve, and the decision curve analysis (DCA). The Nomogram is used to establish a predictive model, and every independent risk factor corresponds to their respective scores, according to which the total scores of different patients can be obtained. The risk of CIP can be estimated according to the corresponding risk values of the total scores, which ranges from 0 to 1, indicating the probability of the incidence of an event. Harrell s concordance index (C-index) was used to evaluate the performance of prediction and discrimination. The C-index value should be range from 0.5 to 1.0, with 0.5 indicating random and 1.0 indicating a fully corrected discrimination[23]. The ROC curve shows the predictive power of each risk factor and the combined Nomogram model, and the area under curve (AUC) values are listed. Higher AUC presented a higher prediction power. Calibration curve is generated by rms package in R language, which reflects the relationship between the predicted incidence and the actual incidence. The abscissa is the predicted probability, and the ordinate is the actual probability of the patient (the actual incidence). DCA was used to evaluate the clinical value of the nomogram[24]. The cut-off value of the total score in the Nomogram model will be determined according to the ROC curve, and patients will be divided into high risk group and low risk group.

1. The demographic and clinicopathological features

This study retrospectively analyzed 547 patients who were first treated with ICIs for malignant solid tumors in the first affiliated hospital of Xi'an jiaotong university from January 1 2019 to January 1 2021. Patients were screened according to strict inclusion and exclusion criteria. Ultimately, 31 patients were excluded due to lack of baseline imaging features, 19 due to ECOG score > 2, and 12 due to lack of records of distant metastatic sites. In addition, 11 patients were excluded for lack of a record of hypertension or diabetes, and 7 patients were excluded for primary tumors that were not in a single site. Five patients were ruled out because they were diagnosed with hematological tumors. Finally, a total of 489 patients were included, covering 13 types of malignant tumors including lung cancer (82.21%), liver cancer (6.74%), gastroesophageal cancer (4.91%), colorectal cancer (2.25%) and renal cancer (1.63%). Following the principle of randomness, patients were divided into training cohort and validation cohort in a ratio of approximately 2:1. A total of 327 patients were eventually included as the training cohort and 162 patients as the validation cohort. The demographic and clinicopathological features of the training cohorts and validation cohort were shown in Table 1.

Among the patients finally analyzed, the median age was 61 years (24-86 years), and a total of 91 patients developed CIP, with an overall incidence of 18.61%. Most of the CIP patients presented with grade 1-2, of which, 32 patients (6.54%) presented with grade 1, 39 patients (7.98%) with grade 2, and 20 patients (4.09%) with grade 3 or higher (Figure S1A). The average time to onset of CIP was 94 days after taking immunotherapy, meaning that most CIP occurred around 3 months after initiation of immunotherapy. Moreover, more than 80% of CIP patients occurred within 4 cycles of immunotherapy, especially within the first or second cycle. However, grade 3 and higher CIP occurs more often after the third to fourth cycle of immunotherapy, which seems to occur later than lower grade CIP (Figure S1B-C).

2. Baseline characteristics of the training cohort

We analyzed the baseline characteristics of the training cohort in three cohorts (Table 2). In terms of patient characteristics, gender, age and smoking history had significant statistical differences in any grade CIP cohort, and male, elderly, and smokers were more likely to develop CIP (P=0.020, 0.006, 0.046, respectively). In tumor characteristics, compared with patients without CIP, patients with less than two metastatic sites and with bone metastases were more likely to develop CIP, and the difference was statistically significant (P=0.033 and 0.002, respectively). From the point of comorbidities, patients with ILD and emphysema at baseline had a higher risk of CIP, with statistically significant differences (P<0.001 for both).

In the grade ≥2 CIP cohort, male and older patients were more inclined to suffer from CIP, with statistically significant differences between the two groups (P=0.043 and 0.001, respectively). For comorbidities, patients with ILD and emphysema at baseline were more likely to develop grade 2 or higher CIP, with statistically significant differences (P< 0.001 for both). The remaining factors were not statistically significant in the grade ≥2 CIP cohort.

The presence of ILD and emphysema at baseline were statistically different in the grade ≥3 CIP cohort (P<0.001 for both), while no statistically significant differences were found for other factors.

3. Correlation between peripheral blood indicator and CIP

ROC curves were plotted for 11 peripheral blood indicators, including CRP, NEUT, ALC, AEC, white blood cell (WBC), PLT, NLR, PLR, procalcalonin (PCT), SII, and PNI. Then the cut-off values of laboratory indicators with AUC greater than 0.5 were obtained and the AUC of each indicator is shown in supplemental table 1. Based on this, the included patients were divided into high group and low group. The results showed that patients with any grade CIP had higher baseline CRP (cut-off value=16.25g/L), AEC (cut-off value=0.215×10⁹ cells/L), WBC (cut-off value=10.7×10⁹ cells/L), PLT (cut-off value=265.5×10⁹ cells/L), and SII index (cut-off value=1592.97), compared with patients without CIP. The differences were statistically significant (P=0.0002, 0.0041, 0.0173, 0.0061, 0.0364, respectively) (Figure S2A-E, respectively). In the grade ≥2 CIP cohort, the results suggested that patients with grade ≥2 CIP had higher baseline CRP (cut-off value=12.65g/L), AEC (cut-off value=0.22×10⁹ cells/L), WBC (cut-off value=5.765×10⁹ cells/L), PLT (cut-off value=176.5×10⁹ cells/L), SII index (cut-off value=411.14), with statistically significant differences (P=0.025, 0.001, 0.011, 0.021, 0.033, respectively) (Figure S3A-E, respectively). The same method was used to analyze the differences of laboratory indicators in the ≥3 grade CIP cohort, suggesting that patients with grade 3 and higher CIP had a high level of CRP (cut-off value=12.65g/L, P=0.025), AEC (cut-off value=0.22×10⁹ cells/L, P=0.001), NEUT (cut-off value=4.0×10⁹ cells/L, P=0.025), and WBC (cut-off value=6.19×10⁹ cells/L, P=0.023) (Figure S4A-D, respectively).

4. Exploration of independent risk factors for CIP in training cohort

To investigate independent risk factors associated with CIP, multivariate logistic regression analysis was performed for 25 factors, including patients’ baseline characteristics, tumor characteristics, treatment characteristics, comorbidity characteristics and peripheral blood indicators. In any grade CIP cohort, multivariate logistic regression analysis showed that several variables were independent risk factors for CIP, including ILD at baseline (P<0.001, OR=5.404, 95%CI=2.581-11.317), emphysema at baseline (P=0.006, OR=3.383, 95%CI=1.420-8.062), higher baseline CRP level (P=0.004, OR=4.967, 95%CI=1.650-14.953), and higher baseline AEC level (P=0.022, OR=2.553, 95%CI=1.146-5.687) (Table 3).

Multivariate logistic regression analysis was also performed in the grade ≥2 CIP cohort. The result revealed lung metastasis (P=0.012, OR=24.319, 95%CI=2.045-289.234), less than two sites of metastasis (P=0.031, OR=5.764, 95%CI=1.375-24.674), ILD at baseline (P=0.026, OR=8.949, 95%CI=1.300-61.614), emphysema at baseline (P=0.022, OR=9.252, 95%CI=1.373-62.329), higher baseline CRP level (P=0.045, OR=5.527, 95%CI=1.041-29.344), higher baseline AEC level (P=0.017, OR=10.442, 95%CI=1.509-72.258), and higher baseline SII level (P=0.028, OR=61.901, 95%CI=1.567-2444.729) were independent risk factors for grade 2 and higher CIP (Table 3). Similarly, multivariate logistic regression analysis was performed for the ≥3 CIP cohort, and no independent risk factors associated with CIP were found (Supplemental table 2).

5. Construction and validation of the nomogram model

Based on the independent risk factors, the predictive nomogram model was established derived from multivariate regression analysis. We used ILD and emphysema at baseline and baseline CRP and AEC levels as key factors to construct a nomogram model for predicting any grade CIP (Figure 2). The C index of the model in the training cohort and validation cohort were 0.814 (95%CI=0.760-0.870) and 0.919 (95%CI=0.842-0.941), respectively. And the calibration curves and ROC curves of the training cohort and validation cohort after internal verification were generated and illustrated in Figure 3A-D, respectively. The calibration curves suggested that the predicted and actual incidence rates are almost identical, which indicates that the nomogram performs well in predicting any grade CIP. Besides, the DCA curve showed that the nomogram model has better clinical predictive power than any single predictor, suggesting that the nomogram could serve as an effective diagnostic tool for CIP (Figure 3E-F, respectively). ROC curve was used to compare the value of combined nomogram model and single predictor in predicting the occurrence of CIP and the results indicate that the combined nomogram model has the highest AUC for both the training cohort and the validation cohort, which again verifies the predictive power of the nomogram model (Figure 3G-H, respectively).

Meanwhile, based on the independent risk factors for grade 2 and higher CIP, we construct another nomogram to predict the occurrence of grade 2 or higher CIP (Figure 4). After internal verification, the calibration curve of training cohort and validation cohort showed a small difference between the predicted incidence and the actual incidence, manifesting the nomogram was quite consistent with the actual observations (Figure 5A-B, respectively). The C index of training cohort and validation cohort were 0.879 (95%CI=0.81-0.92) and 0.936 (95%CI=0.812-0.982), respectively, which indicated the nomogram could perform well in effectively predicting the grade 2 or higher CIP (Figure 5C-D, respectively). DCA curve also manifested that the nomogram model had a stronger clinical prediction power than the single factor (Figure 5E-F, respectively). Moreover, the AUC of the nomogram model has the highest compared to every single indicator, which also supports the high predictive power for predicting grade 2 or higher CIP (Figure 5G-H, respectively).

6. The predictive power of the nomogram model scoring system

In order to further evaluate the predictive power of the nomogram prediction system, we divided the patients into high risk group and low risk group according to the cut-off value of the nomogram prediction model score. For any grade CIP cohort, the cut-off value of the total score of nomogram prediction model is 145 points, and patients will be divided into high risk group (total points ≥145 points) and low risk group (total points <145 points). After the cut-off value was applied to the training cohort, univariate analysis showed that there was a statistically significant difference in the occurrence of any grade CIP between the high risk group and the low risk group (P<0.001, OR=8.831, 95%CI=4.899-15.921), which was consistent with the results of the validation cohort (P<0.001, OR=6.457, 95%CI=2.339-17.824) (Table 4).

Similarly, in the grade ≥2 CIP cohort, patients were divided into high-risk group (total score ≥175) and low-risk group (total score < 175) with 175 points as the cut-off value. Univariate logistic regression analysis both manifested statistically significant differences in training cohort and validation cohort (training cohort: P<0.001, OR= 8.16, 95%CI=3.725-17.873; validation cohort: P<0.001, OR=35.100, 95%CI=4.447-277.065, respectively) (Table 4).

To the best of our knowledge, this is the first study of patients treated with ICIs to construct a visual and efficient nomogram model that can predict the occurrence of CIP, which can prospectively predict the risk of CIP according to the clinicopathological characteristics and the level of serum indexes of patients, so as to solve the clinical problems of screening the beneficiaries of ICIs treatment and identifying the high-risk groups of CIP. Our study manifested ILD and emphysema at baseline and higher CRP and AEC levels at baseline were closely associated with any grade of CIP. Besides, pulmonary metastases, less than two sites of metastases, ILD and emphysema at baseline, and higher levels of CRP, AEC, and SII at baseline were independent risk factors for grade ≥ 2 CIP. We then constructed a novel predictive nomogram model and patients were stratified according to the total score of the nomogram prediction model and divided into high-risk and low-risk groups and there was a significant statistical difference, which makes the nomogram model have a good value in clinical practice. if a patient who intends to receive ICIs therapy is classified as a high-risk group according to our nomogram model, especially high-risk group for grade 3 or higher CIP, clinicians should be vigilant about severe CIP and consider carefully whether to continue taking the ICIs. Repeated internal and external validations show that the nomogram models have a higher prediction performance than the single predictor, which indicates that the nomogram model constructed in our study is expected to be an efficient and convenient CIP prediction tool in clinical practice.

Several studies have confirmed that the incidence of CIP in the real world is much higher than the 3–5% reported in clinical trials[8, 9]. In our study, the incidence of CIP was about 18.61%, and the incidence of severe CIP was 4.09%, which was similar to other reports from the real world[10, 25]. But overall, the incidence of CIP in Asia appears to be higher than in Western populations. Although in many global clinical trials, the incidence of major irAEs in the Chinese population was consistent with that in the whole population, showing no significant ethnic difference between the East and the West[26–28], a number of real world studies also found that the risk of CIP was higher in the Asian population[29, 30]. A retrospective study based on the Japanese population also showed that Asian patients treated with ICIs monotherapy were more likely to develop CIP (5.7%-9.6%), and if combined with other drugs, the incidence of CIP would significantly increase[31]. Unfortunately, there is not enough data and mechanism studies to address this scientific question, which needs to be answered with continuous and in-depth global study.

Previous studies have indicated that patients with baseline ILD[13, 14, 32] and emphysema[32] have a higher incidence of CIP, and similar findings were found in our study. This may be related to the long-term inflammatory state of lung tissue caused by ILD and emphysema[33]. Such chronic pulmonary inflammation and irreversible lung parenchyma damage are risk factors for multi-drug pneumonia, including CIP[14]. In addition, the presence of lung metastases has also been shown to be an independent risk factor for grade ≥ 2 CIP, as it could result in poor basic lung condition, higher tumor burden, and stronger tumor-associated inflammation. Furthermore, this study is the first to show that patients with fewer than two distant metastases are more likely to develop CIP, and we hypotheized that this phenomenon may be due to the fact that the effective and lasting anti-tumor effect of immunotherapy depends on the relatively sound immune system and good physical condition of the patients[34]. To some extent, fewer metastatic sites can reflect patients' lower systemic tumor burden and relatively good physical status[15], which leads to better efficacy of ICIs and higher risk of CIP in patients[11, 12].

Also, in our nomogram models, a high level of baseline AEC and CRP are independent risk factors for CIP, and this was applicable for any grade CIP cohort and grade ≥ 2 CIP cohort. Moreover, more importantly, the optimal cut-off value of AEC and CRP is within the normal level, which gives us a great implication that clinicians should be vigilant against the occurrence of CIP in patients with AEC and CRP in the normal level, but exceeding the cut-off value in clinical practice. And previous retrospective studies have also supported that a higher baseline AEC level is closely related to the incidence of CIP[19]. The infiltration of eosinophils was found in the lung biopsies of patients with CIP[16] and the results of animal experiments also proved that eosinophils were involved in the response of pulmonary T cells[35]. The underlying mechanism is that eosinophils, as regulatory cells or effector cells, participate in a variety of immune responses, such as activating T cells and attracting tumor-specific CD8 + T cells by playing the role of antigen presentation, leading to inflammatory infiltration in the lungs[36–38]. CRP has been reported to be associated with many irAEs, such as immune-associated hypophysitis and colitis[39]. Elevated CRP reflects the presence of systemic inflammation in the host, including pulmonary inflammatory responses[40]. High level of baseline CRP was positively correlated with infiltration of CD8 + T cells and regulatory T cells[41], which play an important role in CIP[42]. In addition, it was observed in our study that higher SII was an independent risk factor for CIP. SII is a comprehensive parameter of lymphocytes, neutrophils and platelets, which is believed to objectively reflect the balance between inflammatory and immune responses in the body[43, 44] and a higher SII indicates a higher tumor burden and a stronger inflammatory response[45], which may be a potential reason to support this conclusion. However, some studies have pointed that higher SII is associated with worse OS in renal cancer patients treated with nivolumab[46], which seems to be in conflict with the previous conclusion that irAEs are positively correlated with good prognosis of ICIs therapy[11, 12], and the specific mechanism remains to be further studied.

Of course, there are many limitations in our study. First of all, this is a retrospective study, so it is subject to many interference factors, which may cause some bias. Secondly, the number of patients included in this study is still relatively insufficient, and larger patient cohorts and prospective studies are expected to verify the effectiveness of the models. Finally, in this retrospective study, many patients lack data on genetic characteristics, and the mechanism of CIP is still in early stage, which will lead to failure to fully consider the specific characteristics of the tumor itself and the characteristics of the tumor microenvironment when constructing the prediction model of CIP. With the further research on the CIP mechanism in the future, it is expected to build a more comprehensive and efficient CIP prediction model.

In this study, we established novel nomogram models to predict any grade CIP and grade 2 or higher CIP, respectively. Gratifyingly, the models have performed good clinical predictive ability after repeated internal and external verification, which are expected to be a convenient, visual, and personalized clinical tool for assessing the risk of developing CIP in patients receiving ICIs treatment.

AEC absolute eosinophil count;

ALC absolute lymphocyte count;

AUC the area under curve;

CIP checkpoint inhibitors pneumonia;

CRP C reactive protein;

DCA decision curve analysis;

ECOG The Eastern Oncology Collaboration Group;

ICIs immune checkpoint inhibitors;

ILD interstitial lung disease;

irAEs immune-related adverse events;

NEUT neutrophil count;

NLR neutrophils/lymphocyte ratio;

PCT procalcalonin;

PLR platelet/lymphocyte ratio;

PLT platelet count;

PNI prognostic nutritional index;

ROC receiver operating characteristic;

SII systemic immune-inflammation index;

WBC white blood cell

Acknowledgements

The authors are grateful to Dr. Kejun Nan of oncology hospital of Xi’an international medical center hospital for his valuable guidance on writing assistance and manuscript revision.

Authors' contributions

All authors contributed to the study conception and design. Xiaohui Jia, Ting Liang, Gang Niu and Hui Guo conceptualised the study and formulated the research protocol. Yonghao Du and Yanlin Li collected clinical data retrospectively. Yuan Shen designed the statistical analysis plan. Xiaohui Jia, Yajuan Zhang, Ziyang Mao and Longwen Xu analysed the data and developed the figures and tables. All authors had full access to all data, critically revised the paper, approved the final analysis.

Funding

This work was supported by Chinese Society of Clinical Oncology-Merck Cancer Research Fund (Y-MSD2020-0247) and C-Tong fund.

Data Availability

The datasets analysed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors have no relevant financial or non-financial interests to disclose.

Conflict of interest

The authors declare that they have no conflict of interest.

Consent for publication

Not applicable

Ethics approval and consent to participate

The study was performed in accordance with the Declaration of Helsinki. This retrospective study was approved by the Ethics Committee of the First Affiliated Hospital of Xi'an Jiaotong University (approval number: XJTU1AF2021LSK-001). And this study is stated that no sex-based or racial/ethnic-based differences were present.

Hoos A. (2016) Development of immuno-oncology drugs - from CTLA4 to PD1 to the next generations. Nature reviews Drug discovery 15(4):235-47.https://doi.org/10.1038/nrd.2015.35
Wolchok JD, Chiarion-Sileni V, Gonzalez R, et al. (2017) Overall Survival with Combined Nivolumab and Ipilimumab in Advanced Melanoma. The New England journal of medicine 377(14):1345-56.https://doi.org/10.1056/NEJMoa1709684
Paz-Ares L, Luft A, Vicente D, et al. (2018) Pembrolizumab plus Chemotherapy for Squamous Non-Small-Cell Lung Cancer. The New England journal of medicine 379(21):2040-51.https://doi.org/10.1056/NEJMoa1810865
Wang DY, Salem JE, Cohen JV, et al. (2018) Fatal Toxic Effects Associated With Immune Checkpoint Inhibitors: A Systematic Review and Meta-analysis. JAMA oncology 4(12):1721-8.https://doi.org/10.1001/jamaoncol.2018.3923
Postow MA, Sidlow R, Hellmann MD. (2018) Immune-Related Adverse Events Associated with Immune Checkpoint Blockade. The New England journal of medicine 378(2):158-68.https://doi.org/10.1056/NEJMra1703481
Hassel JC, Heinzerling L, Aberle J, et al. (2017) Combined immune checkpoint blockade (anti-PD-1/anti-CTLA-4): Evaluation and management of adverse drug reactions. Cancer Treat Rev 57:36-49.https://doi.org/10.1016/j.ctrv.2017.05.003
Brahmer JR, Lacchetti C, Schneider BJ, et al. (2018) Management of Immune-Related Adverse Events in Patients Treated With Immune Checkpoint Inhibitor Therapy: American Society of Clinical Oncology Clinical Practice Guideline. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 36(17):1714-68.https://doi.org/10.1200/jco.2017.77.6385
De Velasco G, Je Y, Bossé D, et al. (2017) Comprehensive Meta-analysis of Key Immune-Related Adverse Events from CTLA-4 and PD-1/PD-L1 Inhibitors in Cancer Patients. Cancer immunology research 5(4):312-8.https://doi.org/10.1158/2326-6066.Cir-16-0237
Khunger M, Rakshit S, Pasupuleti V, et al. (2017) Incidence of Pneumonitis With Use of Programmed Death 1 and Programmed Death-Ligand 1 Inhibitors in Non-Small Cell Lung Cancer: A Systematic Review and Meta-Analysis of Trials. Chest 152(2):271-81.https://doi.org/10.1016/j.chest.2017.04.177
Suresh K, Voong KR, Shankar B, et al. (2018) Pneumonitis in Non-Small Cell Lung Cancer Patients Receiving Immune Checkpoint Immunotherapy: Incidence and Risk Factors. Journal of thoracic oncology : official publication of the International Association for the Study of Lung Cancer 13(12):1930-9.https://doi.org/10.1016/j.jtho.2018.08.2035
Puzanov I, Diab A, Abdallah K, et al. (2017) Managing toxicities associated with immune checkpoint inhibitors: consensus recommendations from the Society for Immunotherapy of Cancer (SITC) Toxicity Management Working Group. J ImmunoTher Cancer 5(1):95.https://doi.org/10.1186/s40425-017-0300-z
Das S, Johnson DB. (2019) Immune-related adverse events and anti-tumor efficacy of immune checkpoint inhibitors. J ImmunoTher Cancer 7(1):306.https://doi.org/10.1186/s40425-019-0805-8
Sugano T, Seike M, Saito Y, et al. (2020) Immune checkpoint inhibitor-associated interstitial lung diseases correlate with better prognosis in patients with advanced non-small-cell lung cancer. Thoracic cancer 11(4):1052-60.https://doi.org/10.1111/1759-7714.13364
Shibaki R, Murakami S, Matsumoto Y, et al. (2020) Association of immune-related pneumonitis with the presence of preexisting interstitial lung disease in patients with non-small lung cancer receiving anti-programmed cell death 1 antibody. Cancer immunology, immunotherapy : CII 69(1):15-22.https://doi.org/10.1007/s00262-019-02431-8
Qiao M, Zhou F, Hou L, et al. (2021) Efficacy of immune-checkpoint inhibitors in advanced non-small cell lung cancer patients with different metastases. Annals of translational medicine 9(1):34.https://doi.org/10.21037/atm-20-1471
Naidoo J, Wang X, Woo KM, et al. (2017) Pneumonitis in Patients Treated With Anti-Programmed Death-1/Programmed Death Ligand 1 Therapy. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 35(7):709-17.https://doi.org/10.1200/jco.2016.68.2005
Tahir SA, Gao J, Miura Y, et al. (2019) Autoimmune antibodies correlate with immune checkpoint therapy-induced toxicities. Proceedings of the National Academy of Sciences of the United States of America 116(44):22246-51.https://doi.org/10.1073/pnas.1908079116
Starlets D, Gore Y, Binsky I, et al. (2006) Cell-surface CD74 initiates a signaling cascade leading to cell proliferation and survival. Blood 107(12):4807-16.https://doi.org/10.1182/blood-2005-11-4334
Chu X, Zhao J, Zhou J, et al. (2020) Association of baseline peripheral-blood eosinophil count with immune checkpoint inhibitor-related pneumonitis and clinical outcomes in patients with non-small cell lung cancer receiving immune checkpoint inhibitors. Lung cancer (Amsterdam, Netherlands) 150:76-82.https://doi.org/10.1016/j.lungcan.2020.08.015
Thompson JA, Schneider BJ, Brahmer J, et al. (2019) Management of Immunotherapy-Related Toxicities, Version 1.2019. Journal of the National Comprehensive Cancer Network : JNCCN 17(3):255-89.https://doi.org/10.6004/jnccn.2019.0013
Haanen JBAG, Carbonnel F, Robert C, et al. (2018) Management of toxicities from immunotherapy: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Annals of oncology : official journal of the European Society for Medical Oncology 29:iv264-iv6.https://doi.org/10.1093/annonc/mdy162
Lung Cancer group Chinese thoracic society. (2019) Expert consensus on the diagnosis and treatment of pneumonia associated with immune checkpoint inhibitors. Chin J Tuberc Respir Dis (11):820-1-2-3-4-5.https://doi.org/10.3760/cma.j.issn.1001⁃0939.2019.11.007
Harrell FE, Lee KL, Mark DB. (1996) Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Statistics in medicine 15(4):361-87.https://doi.org/10.1002/(sici)1097-0258(19960229)15:4<361::Aid-sim168>3.0.Co;2-4
Vickers AJ, Elkin EB. (2006) Decision curve analysis: a novel method for evaluating prediction models. Medical decision making : an international journal of the Society for Medical Decision Making 26(6):565-74.https://doi.org/10.1177/0272989x06295361
Nishino M, Giobbie-Hurder A, Hatabu H, Ramaiya NH, Hodi FS. (2016) Incidence of Programmed Cell Death 1 Inhibitor-Related Pneumonitis in Patients With Advanced Cancer: A Systematic Review and Meta-analysis. JAMA oncology 2(12):1607-16.https://doi.org/10.1001/jamaoncol.2016.2453
Garon EB, Hellmann MD, Rizvi NA, et al. (2019) Five-Year Overall Survival for Patients With Advanced Non‒Small-Cell Lung Cancer Treated With Pembrolizumab: Results From the Phase I KEYNOTE-001 Study. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 37(28):2518-27.https://doi.org/10.1200/jco.19.00934
Paz-Ares L, Vicente D, Tafreshi A, et al. (2020) A Randomized, Placebo-Controlled Trial of Pembrolizumab Plus Chemotherapy in Patients With Metastatic Squamous NSCLC: Protocol-Specified Final Analysis of KEYNOTE-407. Journal of thoracic oncology : official publication of the International Association for the Study of Lung Cancer 15(10):1657-69.https://doi.org/10.1016/j.jtho.2020.06.015
Mok TSK, Wu YL, Kudaba I, et al. (2019) Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial. Lancet (London, England) 393(10183):1819-30.https://doi.org/10.1016/s0140-6736(18)32409-7
Koyama N, Iwase O, Nakashima E, et al. (2018) High incidence and early onset of nivolumab-induced pneumonitis: four case reports and literature review. BMC Pulm Med 18(1):23.https://doi.org/10.1186/s12890-018-0592-x
Kato T, Masuda N, Nakanishi Y, et al. (2017) Nivolumab-induced interstitial lung disease analysis of two phase II studies patients with recurrent or advanced non-small-cell lung cancer. Lung Cancer 104:111-8.https://doi.org/10.1016/j.lungcan.2016.12.016
Hida T, Nishio M, Nogami N, et al. (2017) Efficacy and safety of nivolumab in Japanese patients with advanced or recurrent squamous non-small cell lung cancer. Cancer Sci 108(5):1000-6.https://doi.org/10.1111/cas.13225
Sun Y, Shao C, Li S, et al. (2020) Programmed cell death 1 (PD-1)/PD-ligand 1(PD-L1) inhibitors-related pneumonitis in patients with advanced non-small cell lung cancer. Asia Pac J Clin Oncol 16(6):299-304.https://doi.org/10.1111/ajco.13380
King TE. (2005) Clinical advances in the diagnosis and therapy of the interstitial lung diseases. Am J Respir Crit Care Med 172(3):268-79.https://doi.org/10.1164/rccm.200503-483OE
Blumenthal GM, Zhang L, Zhang H, et al. (2017) Milestone Analyses of Immune Checkpoint Inhibitors, Targeted Therapy, and Conventional Therapy in Metastatic Non-Small Cell Lung Cancer Trials: A Meta-analysis. JAMA oncology 3(8):e171029.https://doi.org/10.1001/jamaoncol.2017.1029
Jacobsen EA, Ochkur SI, Pero RS, et al. (2008) Allergic pulmonary inflammation in mice is dependent on eosinophil-induced recruitment of effector T cells. The Journal of experimental medicine 205(3):699-710.https://doi.org/10.1084/jem.20071840
Wen T, Rothenberg ME. (2016) The Regulatory Function of Eosinophils. Microbiol Spectrum 4(5).https://doi.org/10.1128/microbiolspec.MCHD-0020-2015
Simon SCS, Utikal J, Umansky V. (2019) Opposing roles of eosinophils in cancer. Cancer immunology, immunotherapy : CII 68(5):823-33.https://doi.org/10.1007/s00262-018-2255-4
Carretero R, Sektioglu IM, Garbi N, et al. (2015) Eosinophils orchestrate cancer rejection by normalizing tumor vessels and enhancing infiltration of CD8(+) T cells. Nature immunology 16(6):609-17.https://doi.org/10.1038/ni.3159
Khoja L, Day D, Wei-Wu Chen T, Siu LL, Hansen AR. (2017) Tumour- and class-specific patterns of immune-related adverse events of immune checkpoint inhibitors: a systematic review. Annals of oncology : official journal of the European Society for Medical Oncology 28(10):2377-85.https://doi.org/10.1093/annonc/mdx286
Abolhassani AR, Schuler G, Kirchberger MC, Heinzerling L. (2019) C-reactive protein as an early marker of immune-related adverse events. J Cancer Res Clin Oncol 145(10):2625-31.https://doi.org/10.1007/s00432-019-03002-1
Nakayama T, Saito K, Kumagai J, et al. (2018) Higher Serum C-reactive Protein Level Represents the Immunosuppressive Tumor Microenvironment in Patients With Clear Cell Renal Cell Carcinoma. Clin Genitourin Cancer 16(6):e1151-e8.https://doi.org/10.1016/j.clgc.2018.07.027
Suresh K, Naidoo J, Zhong Q, et al. (2019) The alveolar immune cell landscape is dysregulated in checkpoint inhibitor pneumonitis. The Journal of clinical investigation 129(10):4305-15.https://doi.org/10.1172/jci128654
Hu B, Yang XR, Xu Y, et al. (2014) Systemic immune-inflammation index predicts prognosis of patients after curative resection for hepatocellular carcinoma. Clinical cancer research : an official journal of the American Association for Cancer Research 20(23):6212-22.https://doi.org/10.1158/1078-0432.Ccr-14-0442
Wang K, Diao F, Ye Z, et al. (2017) Prognostic value of systemic immune-inflammation index in patients with gastric cancer. Chin J Cancer 36(1):75.https://doi.org/10.1186/s40880-017-0243-2
Naito T, Tanaka F, Ono A, et al. (2012) Prognostic impact of circulating tumor cells in patients with small cell lung cancer. Journal of thoracic oncology : official publication of the International Association for the Study of Lung Cancer 7(3):512-9.https://doi.org/10.1097/JTO.0b013e31823f125d
Villadolid J, Amin A. (2015) Immune checkpoint inhibitors in clinical practice: update on management of immune-related toxicities. Translational lung cancer research 4(5):560-75.https://doi.org/10.3978/j.issn.2218-6751.2015.06.06

Table 1 The demographic and clinicopathological features
	Training cohort（N=327）	Validation cohort（N=162）	*P value*
Patient Characteristic
Sex
Female	63（19.3%）	40（24.7%）	0.166
Male	264（80.7%）	122（75.3%）
Age
＜60	142（43.4%）	83（51.2%）	0.103
≥60	185（56.6%）	79（48.8%）
History of smoking
Never	138（42.2%）	82（50.6%）	0.078
Smoker	189（57.8%）	80（49.4%）
Tumor Characteristic
Cancer type
Lung cancer	280（85.6%）	122（75.9%）	0.055
Hepatocellular carcinoma	21（6.4%）	12（7.4%）
Stomach and esophagus cancer	11（3.4%）	13（8.0%）
Colorectal cancer	6（3.7%）	5（3.1%）
Others	9（2.8%）	9（5.6%）
NMS
≥2	150（45.9%）	63（38.9%）	0.143
＜2	177（54.1%）	99（61.1%）
Pulmonary metastasis
No	225（68.8%）	124（76.5%）	0.075
Yes	102（31.2%）	38（23.5%）
Lymphatic metastasis
No	161（49.2%）	80（49.4%）	0.976
Yes	166（50.8%）	82（50.6%）
Bone metastasis
No	226（69.1%）	111（68.5%）	0.894
Yes	101（30.9%）	51（31.5%）
Hepatic metastases
No	275（84.1%）	128（79.0%）	0.164
Yes	52（15.9%）	34（21.0%）
Brain metastases
No	275（84.1%）	135（83.3%）	0.829
Yes	52（15.9%）	27（16.7%）
Adrenal metastasis
No	297（90.8%）	152（93.8%）	0.254
Yes	30（9.2%）	10（6.2%）
Treatment Characteristic
Combined treatment
IO monotherapy	32（9.8%）	24（14.8%）	0.100
Combined IO	295（90.2%）	138（85.2%）
Specific combined treatment
IO monotherapy	32（9.8%）	24（14.8%）	0.139
IO+chemotherapy	231（70.6%）	109（67.3%）
IO+EGFR-TKI	16（4.9%）	15（9.3%）
IO+AVEGFR	21（6.4%）	3（1.9%）
IO+Chemo+AVEGFR	27（8.3%）	11（6.8%）
History of radiation therapy
No	272（83.2%）	125（77.2%）	0.109
Yes	55（16.8）	37（22.8%）
History of TKI drug therapy
No	288（88.1%）	138（85.2%）	0.370
Yes	39（11.9%）	24（14.8%）
History of antiangiogenic drug therapy
No	249（76.1%）	119（73.5%）	0.561
Yes	78（23.9%）	43（26.5%）
Characteristics of comorbidities
HBP
No	264（80.7%）	130（80.2%）	0.898
Yes	63（19.3%）	32（19.8%）
Diabetes
No	300（91.7%）	148（91.4%）	0.885
Yes	27（8.3%）	14（8.6%）
Baseline ILD
No	235（71.9%）	129（79.6%）	0.064
Yes	92（28.1%）	33（20.4%）
Baseline emphysema
No	257（78.6%）	135（83.3%）	0.216
Yes	70（21.4%）	27（16.7%）
NMS, number of metastatic sites; IO, immunotherapy; EGFR-TKI, epidermal growth factor receptor-tyrosine kinase inhibitor; AVEGFR, anti-vascular endothelial growth factor receptor; HBP, high blood pressure; ILD, interstitial lung disease

Table 2 Baseline characteristics of the training cohort
	Any grade CIP cohort			Grade ≥2 CIP cohort			Grade ≥3 CIP cohort
	No CIP（%）	CIP（%）	P value	No CIP or Grade 1 CIP（%）	Grade 2 or higher CIP（%）	P value	No CIP or Grade 1-2 CIP（%）	Grade 3 or higher CIP（%）	P value
All patients（n=327）	255（77.98）	72（22.02）		280（85.63）	47（14.37）		312（95.41）	15（4.59）
Patient Characteristic
Sex
Female	56（88.89）	7（11.11）		59（93.65）	4（6.35）		61（96.83）	2（3.17）
Male	199（75.38）	65（24.62）	0.020	221（83.71）	43（16.29）	0.043	261（95.26）	13（4.74）	0.551
Age/year,mean(range)	59.4（27-80）	63.1（29-82）		59.5(27-80)	64.9(29-82)		60.2(27-82)	61.8(29-76)
＜60	121（85.21）	21（14.79）		132（92.96）	10（7.04）		139（97.89）	3（2.11）
≥60	134（72.43）	51（27.57）	0.006	148（80.00）	37（20.00）	0.001	173（93.51）	12（6.49）	0.061
History of smoking
Never	115（83.33）	23（16.67）		122（88.41）	16（11.59）		134（97.10）	4（2.90）
Smoker	140（74.07）	49（25.93）	0.046	158（83.60）	31（16.40）	0.221	178（94.18）	11（5.82）	0.212
Tumor Characteristic
Cancer type
Lung cancer	213（75.80）	68（24.20）	0.193	237（84.64）	43（15.36）	0.477	265（94.64）	15（5.36）	0.620
Hepatocellular carcinoma	19（90.48）	2（9.52）		20（95.24）	1（4.76）		21（100.00）	0（0.00）
Stomach and esophagus cancer	10（90.91）	1（9.09）		10（90.91）	1（9.09）		11（100.00）	0（0.00）
Colorectal cancer	6（100.00）	0（0.00）		6（100.00）	0（0.00）		6（100.00）	0（0.00）
Others	8（88.89）	1（11.11）		7（77.78）	2（22.22）		9（100.00）	0（0.00）
NMS
≥2	109（72.67）	41（27.33）		128（85.33）	22（14.67）		169（95.48）	8（4.52）
＜2	146（82.49）	31（17.51）	0.033	152（85.88）	25（14.12）	0.889	143（95.33）	7（4.67）	0.950
Pulmonary metastasis
No	179（79.56）	46（20.44）		195（86.67）	30（13.33）		216（96.00）	9（4.00）
Yes	76（74.51）	26（25.49）	0.308	85（83.33）	17（16.67）	0.426	96（94.12）	6（5.88）	0.451
Lymphatic metastasis
No	132（81.99）	29（18.01）		140（86.96）	21（13.04）		154（95.65）	7（4.35）
Yes	123（74.10）	43（25.90）	0.085	140（84.34）	26（15.66）	0.500	158（95.18）	8（4.82）	0.839
Bone metastasis
No	187（82.74）	39（17.26）		198（87.61）	28（12.39）		216（95.58）	10（4.42）
Yes	68（67.33）	33（32.67）	0.002	82（81.19）	19（19.81）	0.126	96（95.05）	5（4.95）	0.834
Hepatic metastases
No	215（78.18）	60（21.82）		234（85.09）	41（14.91）		262（95.27）	13（4.73）
Yes	40（76.92）	12（23.08）	0.841	46（88.46）	6（11.54）	0.525	50（96.15）	2（3.85）	0.781
Brain metastases
No	213（77.45）	62（22.55）		233（84.73）	42（15.27）		262（95.27）	13（4.73）
Yes	42（80.77）	10（19.23）	0.597	47（90.38）	5（9.62）	0.286	59（96.72）	2（3.28）	0.781
Adrenal metastasis
No	235（79.12）	62（20.88）		256（86.20）	41（13.80）		284（95.62）	13（4.38）
Yes	20（66.67）	10（33.33）	0.117	24（80.00）	6（20.00）	0.357	28（93.33）	2（6.67）	0.568
Treatment Characteristic
ICIs agent
Pembrolizumab	110（74.83）	37（25.17）	0.200	123（83.67）	24（16.33）	0.087	138（93.88）	9（6.12）	0.522
Nivolumab	0（0.00）	1（100.00）		0（0.00）	1（100.00）		0（0.00）	1（100.00）
Camrelizumab	55（76.39）	17（23.61）		58（80.56）	14（19.44）		70（97.22）	2（2.78）
Toripalimab	20（80.00）	5（20.00）		22（88.00）	3（12.00）		24（96.00）	1（4.00）
Sintilimab	69（85.19）	12（14.81）		76（93.83）	5（6.17）		79（97.53）	2（2.47）
Atezolizumab	1（100.00）	0（0.00）		1（100.00）	0（0.00）		1（100.00）	0（0.00）
Combined treatment
IO monotherapy	26（81.25）	6（18.75）		27（84.37）	5（15.63）		31（96.88）	1（3.12）
Combined IO	229（77.63）	66（22.37）	0.639	253（85.76）	42（14.24）	0.832	281（95.25）	14（4.75）	0.677
Specific combined treatment
IO monotherapy	26（81.25）	6（18.75）	0.140	27（84.38）	5（15.62）	0.460	31（96.88）	1（3.12）	0.635
IO+chemotherapy	173（74.89）	58（25.11）		196（84.85）	35（15.15）		218（94.37）	13（5.63）
IO+EGFR-TKI	16（100.00）	0（0.00）		16（100.00）	0（0.00）		16（100.00）	0（0.00）
IO+AVEGFR	18（85.71）	3（14.29）		19（90.48）	2（9.52）		21（100.00）	0（0.00）
IO+Chemo+AVEGFR	22（81.48）	5（18.52）		22（81.48）	5（18.52）		26（96.30）	1（3.70）
History of radiation therapy
No	209（76.84）	63（23.16）		231（84.93）	41（15.07）		258（94.85）	14（5.15）
Yes	46（83.64）	9（16.36）	0.267	49（89.09）	6（10.91）	0.442	54（98.18）	1（1.82）	0.282
History of TKI drug therapy
No	224（77.78）	64（22.22）		245（85.07）	43（14.93）		273（94.79）	15（5.21）
Yes	31（79.49）	8（20.51）	0.809	35（89.74）	4（10.26）	0.435	39（100.00）	0（0.00）	0.145
History of antiangiogenic drug therapy
No	194（77.91）	55（22.09）		214（85.94）	35（14.06）		237（95.18）	12（4.82）
Yes	61（78.21）	17（21.79）	0.956	66（84.62）	12（15.38）	0.770	75（96.15）	3（3.85）	0.720
Characteristics of comorbidities
HBP
No	211（79.92）	53（20.08）		228（86.36）	36（13.64）		253（95.83）	11（4.17）
Yes	44（69.84）	19（30.16）	0.083	52（82.54）	11（17.46）	0.437	59（93.65）	4（6.35）	0.457
Diabetes
No	234（78.00）	66（22.00）		256（85.33）	44（14.67）		286（95.33）	14（4.67）
Yes	21（77.78）	6（22.22）	0.979	24（88.89）	3（11.11）	0.614	26（96.30）	1（3.70）	0.819
Baseline ILD
No	208（88.51）	27（11.49）		220（93.62）	15（6.38）		229（97.45）	6（2.55）
Yes	47（51.09）	45（48.91）	＜0.001	60（65.22）	32（34.78）	＜0.001	83（90.22）	9（9.78）	0.005
Baseline emphysema
No	219（85.21）	38（14.79）		237（92.22）	20（7.78）		250（97.28）	7（2.72）
Yes	36（51.43）	34（48.57）	＜0.001	43（61.43）	27（38.57）	＜0.001	62（88.57）	8（11.43）	0.002
CIP, checkpoint inhibitors pneumonia; NMS, number of metastatic sites; ICIs, immune checkpoint inhibitors; IO, immunotherapy; EGFR-TKI, epidermal growth factor receptor-tyrosine kinase inhibitor; AVEGFR, anti-vascular endothelial growth factor receptor; HBP, high blood pressure; ILD, interstitial lung disease

Table 3 Multivariate regression of any grade CIP cohort and grade ≥2 CIP cohort
	Any grade CIP cohort				Grade ≥2 CIP cohort
	P value	OR	95% CI		P value	OR	95% CI
	P value	OR	Upper limit	Lower limit	P value	OR	Upper limit	Lower limit
All patients（n=327）
Sex
Female	Ref.				Ref.
Male	0.512	0.677	0.211	2.176	0.440	0.275	0.010	7.297
Age
＜60	Ref.				Ref.
≥60	0.727	1.192	0.446	3.186	0.493	0.340	0.475	5.405
History of smoking
Never	Ref.				Ref.
Smoker	0.996	1.002	0.424	2.371	0.398	2.392	0.317	18.077
NMS
≥2	Ref.				Ref.
＜2	0.920	2.147	0.245	3.558	0.031	5.764	1.375	24.674
Pulmonary metastasis
No	Ref.				Ref.
Yes	0.538	1.360	0.511	3.622	0.012	24.319	2.045	289.234
Lymphatic metastasis
No	Ref.				Ref.
Yes	0.819	0.897	0.353	2.281	0.742	1.409	0.183	10.845
Bone metastasis
No	Ref.				Ref.
Yes	0.256	1.829	0.645	5.184	0.677	1.714	0.135	21.696
Hepatic metastases
No	Ref.				Ref.
Yes	0.996	0.997	0.331	3.007	0.321	0.310	0.031	3.135
Brain metastases
No	Ref.				Ref.
Yes	0.769	1.184	0.384	3.652	0.396	3.790	0.174	82.392
Adrenal metastasis
No	Ref.				Ref.
Yes	0.534	1.515	0.409	5.617	0.178	7.892	0.390	＞100
Combined treatment
IO monotherapy	Ref.				Ref.
Combined IO	0.213	2.216	0.633	7.761	0.569	1.483	0.382	5.762
Baseline ILD
No	Ref.				Ref.
Yes	＜0.001	5.404	2.581	11.317	0.026	8.949	1.300	61.614
Baseline emphysema
No	Ref.				Ref.
Yes	0.006	3.383	1.420	8.062	0.022	9.252	1.373	62.329
History of radiation therapy
No	Ref.				Ref.
Yes	0.486	0.699	0.256	1.910	0.211	0.110	0.003	3.496
History of TKI drug therapy
No	Ref.				Ref.
Yes	0.081	2.709	0.884	8.297	0.889	1.321	0.026	65.95
History of antiangiogenic drug therapy
No	Ref.				Ref.
Yes	0.657	0.821	0.343	1.962	0.523	2.543	0.145	44.653
HBP
No	Ref.				Ref.
Yes	0.174	1.922	0.749	4.932	0.256	2.836	0.470	17.096
Diabetes
No	Ref.				Ref.
Yes	0.685	0.754	0.193	2.947	0.245	0.114	0.003	4.434
Baseline CRP^*
0	Ref.				Ref.
1	0.004	4.967	1.650	14.953	0.045	5.527	1.041	29.344
Baseline NEUT^*
0	Ref.				Ref.
1	0.757	0.866	0.347	2.159	0.270	0.319	0.042	2.434
Baseline ALC^*
0	Ref.				Ref.
1	0.247	1.824	0.659	5.052	0.806	1.357	0.119	15.46
Baseline AEC^*
0	Ref.				Ref.
1	0.022	2.553	1.146	5.687	0.017	10.442	1.509	72.258
Baseline WBC^*
0	Ref.				Ref.
1	0.357	1.613	0.584	4.456	0.720	0.666	0.072	6.16
Baseline PLT^*
0	Ref.				Ref.
1	0.276	1.611	0.683	3.798	0.666	0.675	0.114	4.006
Baseline PLR^*
0	Ref.				Ref.
1	0.240	1.738	0.691	4.372	0.498	0.471	0.053	4.172
Baseline SII^*
0	Ref.				Ref.
1	0.811	0.875	0.292	2.618	0.028	61.901	1.567	＞100
* represents the cutoff value of different peripheral blood indicators for any grade CIP cohort and grade ≥2 CIP cohort. Lower than the cut-off value is defined as low level, otherwise it is considered as high level. For any grade cohort, the cut-off values of individual indicator are as follows: CRP=16.25g/L; NEUT=3.39×10⁹ cells/ L; ALC=2.145×10⁹cells/L; AEC=0.215×10⁹ cells/ L; WBC=10.7×10⁹ cells/ L; PLT=265.5×10⁹ cells/ L; PLR=191.96; SII=1592.97. For grade ≥2 CIP, the cut-off values of individual indicator are as follows: CRP=12.65g/L; NEUT=5.68×10⁹ cells/ L; ALC=0.87×10⁹cells/L; AEC=0.22×10⁹ cells/ L; WBC=5.765×10⁹ cells/ L; PLT=276.5×10⁹ cells/ L; PLR = 191.96; SII = 411.14.
Ref., reference; CIP, checkpoint inhibitors pneumonia; NMS, number of metastatic sites; IO, immunotherapy; ILD, interstitial pulmonary disease; EGFR-TKI, epidermal growth factor receptor-tyrosine kinase inhibitor; AVEGFR, anti-vascular endothelial growth factor receptor; HBP, high blood pressure; CRP, C reactive protein; NEUT, neutrophil count; ALC, absolute lymphocyte count; AEC, absolute eosinophil count; WBC, white blood cell; PLT, platelet count; PLR, the platelet/lymphocyte ratio; SII, systemic immune-inflammation index; CI, confidence interval.

Table 4

Univariate logistic regression analysis of total points of nomogram model in predicting any grade CIP and grade 2 or higher CIP in the training and validation cohorts

Any grade CIP cohort

Grade ≥2 CIP cohort

Group^*

Training Cohort

Validation Cohort

Training cohort

Validation cohort

P value

95%CI

P value

95%CI

P value

95%CI

P value

95%CI

Low risk

Ref.

High risk

＜0.001

8.831

4.899-15.921

＜0.001

6.457

2.339-17.824

＜0.001

8.16

3.725-17.873

＜0.001

35.100

4.447-277.065

*: For any grade CIP cohort: Low risk: total points＜145; High risk: total points ≥145; For grade ≥2 CIP cohort: Low risk: total points＜175; High risk: total points ≥175

Ref., reference; OR, odds ratio; CI, confidence interval

No competing interests reported.

PDF

Version 1

posted 24 May, 2022

You are reading this latest preprint version

Comprehensive nomogram models for prediction of checkpoint inhibitors pneumonia

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

1. Study population

2. Data collection

3. Statistical analysis

Results

1. The demographic and clinicopathological features

2. Baseline characteristics of the training cohort

3. Correlation between peripheral blood indicator and CIP

4. Exploration of independent risk factors for CIP in training cohort

5. Construction and validation of the nomogram model

6. The predictive power of the nomogram model scoring system

Discussion

Conclusion

Abbreviations

Declarations

References

Tables

Competing Interests

Supplementary Files

Status:

Version 1