A dynamic nomogram predicting symptomatic pneumonia in patients with lung cancer receiving thoracic radiation

Radiation pneumonitis (RP) is the most common side effect following radiotherapy for lung cancer. Currently, there is no effective treatment for RP. Symptomatic pneumonitis can impact a patient’s prognosis and reduce their quality of life. It is well-recognized that radiation dose parameters are the most important influencing factors for RP. The lung V20 is a classic and frequently described dosimetry parameter, and limiting lung V20 can significantly reduce the incidence of RP [30]. However, even when the whole lung V20 dose parameter is strictly restricted before radiotherapy, RP is still unavoidable. This suggests that multiple factors, in addition to dose parameter factors, contribute to RP.

Identifying high-risk patients for symptomatic pneumonitis before radiotherapy is crucial to prevent its occurrence. Patients with RP ≥ 2 represent a highly heterogeneous group; our study combined multiple clinical and dosimetry indicators to establish a predictive model. The results demonstrated that a combination of three dosimetry and three clinical factors could predict RP ≥ 2. More importantly, we developed a dynamic nomogram and risk classification system with good predictive performance, which is easy for physicians to understand and implement in practice.

Previous studies have estimated the incidence of symptomatic pneumonitis to be 15–40% among patients undergoing thoracic radiotherapy [14]. Our study’s incidence (17.5%) is consistent with these findings. However, the incidence of fatal pneumonitis in our study was 5% (8/160), significantly higher than the 2% reported in previous studies [8]. In our study, 13 patients had Grade 3 symptomatic pneumonitis, accounting for 46.43% (13/28) of all symptomatic pneumonitis cases. The high mortality rate in our study may be related to the increased number of patients diagnosed with ≥ Grade 3 pneumonitis, which is more likely to develop into fatal pneumonitis.

Older age has been shown to be a predictor of symptomatic pneumonitis [7, 31]. Previous studies have found that older patients (age ≥ 70) are more likely to develop symptomatic pneumonitis [31]. Similar results were confirmed in a prospective study with 7752 patients who received stereotactic body radiation therapy (SBRT) for stage IA to IB non-small-cell lung cancer (NSCLC) [10]. This is primarily due to the fact that older patients have more comorbidities that are a risk factor for RP than younger patients [7]. Although age ≥ 60 years was an independent risk factor in our study, a threshold value has not been determined due to additional risk factors such as smoking status and pulmonary function [3]. While age is an uncontrollable factor, it has minimal impact on the symptomatic pneumonitis score in the nomogram.

The relationship between smoking and radiation-induced lung injury remains controversial [32]. Tobacco smoking is the leading risk factor for developing lung cancer. In contrast, smoking has been found to have a protective effect on the risk of RP. Smokers are less likely to develop radiation pneumonitis, as it is thought that smoking-damaged lungs may be less sensitive to radiation damage [33, 7, 34]. However, smoking is a significantly poor prognostic factor for overall survival [35]. The smoking index, calculated as cigarettes smoked per day × years smoked, was an independent risk factor for symptomatic pneumonitis when the index was ≥ 400 in our study. The relationship between the smoking index and symptomatic pneumonitis has received insufficient attention in lung cancer patients; our results provide solid evidence, but the nomogram showed that the smoking index had the lowest impact among the six risk factors for symptomatic pneumonitis.

Conflicting results exist regarding the effects of concurrent, concomitant, or neoadjuvant chemotherapy on radiation toxicity. A dosimetric analysis of 25 patients with NSCLC revealed that approximately 30% of patients experienced at least a 20% reduction in tumor volume following induction chemotherapy, translating into a slight yet statistically significant reduction (5%) in the predicted risk of RP [36]. However, a retrospective analysis of 223 patients by Wang et al. [37] did not find a significant association in the risk of Grade 3 RP between patients with and without induction chemotherapy. In contrast, another retrospective study of 84 lung cancer patients receiving three-dimensional conformal radiotherapy (3D-CRT) showed that the risk of symptomatic pneumonitis was three-fold higher in the chemotherapy group than in the non-chemotherapy group (31.2%) [38]. Our findings showed that patients who did not receive chemotherapy were more likely to develop symptomatic pneumonia than those who did. We consider the reasons that lead our results to differ from those of previous studies as follows: Some patients in the symptomatic pneumonia group were more prone to symptomatic pneumonia due to older age and poor ECOG score, so we did not give chemotherapy to these patients. In addition, some patients with large lung tumors refuse chemotherapy due to their own wishes, resulting in a larger radiation treatment area, which in turn exposes more normal lung tissue to radiation. Therefore, we believe that the appropriate course of chemotherapy helps to shrink the tumor, and does not increase the occurrence of symptomatic pneumonia.

It should be noted that following the introduction of volumetric modulated arc therapy (VMAT) and SBRT into clinical practice, greater attention has been paid to the potential adverse effects of low-dose lung dose-volume histogram (DVH) parameters on RP ≥ 2. Several studies [12, 29, 39‐42] have reported a close relationship between whole lung V5, V10, and RP in lung cancer. Sheng L et al. [40] found that the dose threshold for V5 in patients with non-small cell lung cancer treated with induction chemotherapy was 44%. Another study discovered that ipsilateral lung V5 ≥ 55.65% might be associated with a higher risk of RP [21]. A previous clinical study reported that acute radiation pneumonitis was reduced from 29.2 to 5.7% in lung cancer patients receiving concurrent chemoradiotherapy when V10 ≤ 50 [41]. Moreover, Boonyawan et al. [43] identified a correlation between V10 > 30% and grade ≥ 2 RP in patients who had undergone surgery followed by PORT. Our study found that lung V5 was independent risk factors. We conducted a univariate receiver operating characteristic (ROC) analysis on lung V5, with the best cut-off value being 61.10% (AUC = 0.726, 95% CI: 0.626–0.827, sensitivity = 0.679, specificity = 0.720). The whole lung V5 > 61.1% in the symptomatic pneumonitis group was 67.86% (19/28) compared to 28.03% in the asymptomatic pneumonitis group (37/132). These data differences were statistically significant(P<0.05). Our whole lung V5 was mostly above the dose-optimal threshold, which may be associated with higher severity and mortality of symptomatic pneumonitis than previously reported. In conclusion, further research is needed to confirm that our findings are not accidental and to define the optimal cut-offs of whole lung V5 for more accurate prediction of symptomatic pneumonitis.

Mean lung dose (MLD) can be employed to predict the risk of symptomatic pneumonitis in lung cancer patients undergoing radical radiotherapy [44, 45]. A finding concur with these studies, indicating that the incidence of RP ≥ Grade 2 increased in cases where the MLD was ≥ 14.3 Gy [18]. A prior study [21] found that patients with a mean ipsilateral lung dose ≥ 11.86 Gy might be at a higher risk of RP ≥ Grade 2. Another study suggested that MLD of 10 Gy was significantly correlated with an increased risk of RP, while an MLD < 12 Gy was significantly correlated with a low risk of RP ≥ Grade 2 [17]. Lee et al. [46] used perfusion single-photon-emission computed tomography and fluorodeoxyglucose positron-emission tomography imaging to demonstrate that MLD (also functional MLD) was a significant predictor of grade ≥ 2 pneumonitis, with a cutoff value of 13.6 Gy (functional MLD: 13.2 Gy). Baker R et al. [42] found that patients receiving stereotactic body radiation therapy (SBRT) with MLD less than 600 Gy had a lower risk of RP ≥ Grade 2, and the same results were obtained in another study [45]. For patients receiving postoperative radiotherapy, it is recommended to control the whole lung MLD below 10.8 Gy [12]. A machine learning study based on 203 stage II-III lung cancer patients receiving chemoradiotherapy found that whole lung MLD was strongly associated with symptomatic pneumonitis, with a cutoff value of 13.4 Gy [39]. In our study, the best cut-off value of whole lung MLD was 948.15 Gy (AUC = 0.726, 95% CI: 0.622–0.799, sensitivity = 1.00, specificity = 0.424), indicating that the incidence of symptomatic pneumonitis was definite in cases where the whole lung MLD was > 948.15 Gy. Although the cutoff values for MLD differ across research institutions, our findings were within the range previously reported.

As is well-known, multifactorial combined prediction models are beneficial in enhancing RP prediction performance. Although several prediction models have been preliminarily established in prior studies, their implementation in the clinic has been limited due to their unsatisfactory ability to identify RP or the practical challenges they pose. In our study, however, a number of essential variables, including age, smoking index, whole lung V5 and whole lung MLD, were integrated into a nomogram to accurately personalize the risk assessment of RP ≥ 2 for each patient. The c-index of the nomogram in this study was 0.89, indicating that it outperformed previously reported models combining clinical dosimetric factors in predicting RP ≥ 2. The transformation into a dynamic nomogram makes it more convenient for clinicians to apply. More importantly, a new risk classification system was established that effectively stratified lung cancer patients into distinct RP risk groups. Our study demonstrates that the model exhibits excellent discrimination in predicting the occurrence of RP ≥ 2. As any physician can easily measure the included variables, the practicability of the model is enhanced. Despite the many advantages of our study, some limitations should be acknowledged. First, Due to the small sample size, we did not set up a validation group. Although there is no external validation, our concordance index and calibration plots confirmed the good predictive performance of our model. This study is a preliminary study, we will continue to collect more patients and expand the sample size to verify the accuracy of our nomogram. Second, as this study is retrospective and from a single institution, there may be a selection bias. In addition, our previous clinical data are incomplete, lacking crucial information such as inflammatory factors and specific chemotherapy courses, which may help improve the accuracy of symptomatic pneumonitis prediction. Lastly, only symptomatic pneumonitis was used as an endpoint, while local tumor control and overall survival were not considered. Reducing the risk of symptomatic pneumonitis may concurrently increase the risk of local tumor control failure and alter patient survival time. In the future, multi-objective models will be developed to assist physicians in determining the best course of treatment. This study only employed symptomatic pneumonitis as an endpoint and did not consider other treatment outcomes, such as local tumor control and overall survival rate. Reducing the risk of RP ≥ 2 may simultaneously increase the risk of tumor local control failure and decrease survival time; thus, our prospective study is ongoing.