Predictors of Post-chemoradiotherapy Pulmonary Complication in Locally Advanced Non–Small Cell Lung Cancer

1. Study population and treatment

The medical records of 317 patients who underwent definitive CCRT for LA-NSCLC (inoperable stage II and stage III) from January 2012 to August 2020 were reviewed retrospectively. No minimum radiation dose requirement was set. Patients with neoadjuvant chemotherapy were excluded. All patients underwent radiation therapy encompassing the known disease extent in the thoracic area identified by computed tomography (CT) scan, positron emission tomography (PET) scan, and endobronchial ultrasound-guided transbronchial needle aspiration. The gross tumor volume (GTV) was delineated for primary tumor and nodal metastasis identified by these examinations. The clinical target volume (CTV) was created by expanding the GTV toward surrounding normal tissue to encompass microscopic tumor extension. GTV-to-CTV expansion was 7 mm for primary GTV and 5 mm for nodal GTV. The planning target volume (PTV) was constructed by expanding the CTV by 7–10 mm. Radiation dose was prescribed to the PTV. Dose-fractionation regimen was decided by treating radiation oncologist, and disease involvement, baseline lung function, and lung radiation dose was mainly considered. Sixty Gy in 30 fractions were the most frequently prescribed. Three-dimensional conformal radiation therapy (3D-CRT) and volumetric modulated arc therapy (VMAT) were used for radiation therapy planning and delivery.

Chemotherapy was administered concurrently with radiation therapy. Frequently used chemotherapeutic regimens were carboplatin-paclitaxel and docetaxel-cisplatin. For carboplatin-paclitaxel regimen, carboplatin was administered to a targeted area under the plasma concentration-time curve of 2, and paclitaxel was administered at a dose of 50 mg/m². For docetaxel-cisplatin regimen, 20 mg/m² of docetaxel and 20 mg/m² of cisplatin were administered. The patient received these chemotherapeutic agents weekly. Consolidation immunotherapy with durvalumab became available in late 2018 and was administered to some patients.

The patients were follow-up with at 1 month after the completion of CCRT. Thereafter, follow-up interval was 3 months for 2 years. For the third to fifth years, follow-up interval was 6 months. Clinical examination and chest CT scan were performed for every follow-up visit. When radiation changes in chest CT were superimposed with suspected progressive lesion, short-term follow-up CT scan was performed to identify increase of nodular lesion or solid portion of suspected recurred site. If the progression was still uncertain after follow-up CT scan, PET scan was performed to identify hypermetabolism of suspected progressive lesion.

2. Post-chemoradiotherapy pulmonary complications

PCPC was defined as admission or emergency department visit for the treatment of pulmonary inflammatory complication, which was defined as pneumonia and pneumonitis, within 6 months from the initiation of CCRT. Hospital visits for other causes, such as supportive care or treatment of oesophagitis, were excluded. The events after confirmation of disease progression were censored. The adverse events of CCRT were assessed and graded in accordance with the Common Terminology Criteria for Adverse Events ver. 5.0. Distinguishing infectious pneumonia and radiation pneumonitis was performed by a board-certified radiation oncologist. Clinical features at the presentation such as location of consolidation in chest CT scan (in-field vs. out-field), onset of adverse event, signs of infection, and identification of infectious organism was considered first. If the category of pulmonary complication was uncertain, response for corresponding treatment and clinical course were taken into account. If the classification was still inconclusive, such pulmonary adverse event was classified as superimposed and counted as both pneumonia and pneumonitis.

Several covariates, including baseline patient characteristics, results of pulmonary function tests performed at disease diagnosis, dosimetric features of radiation therapy, and results of laboratory test performed at least once per 2 weeks during CCRT, were selected to evaluate their potential association with PCPC. Twenty-eight covariates were analyzed to investigate their association with PCPC. Sex, age at diagnosis, body mass index, smoking history (pack-year), histology (others vs. adenocarcinoma), and regimen of concurrent chemotherapy were chosen from patient characteristics. The absolute value and relative percentage of the predicted value of forced expiratory volume in 1 second (FEV₁), forced vital capacity (FVC), FEV₁/FVC, and diffusing capacity of carbon monoxide (DLCO) were selected from baseline pulmonary function tests. Radiation therapy technique, equivalent dose in 2 Gy fractions of the prescribed dose, bilateral lung mean dose, bilateral lung percentage volume that received more than 5 Gy (V5), bilateral lung percentage volume that received more than 20 Gy (V20), bilateral lung volume, mean heart dose, size of the PTV, and duration of radiation therapy were selected from radiation therapy-related features. The bilateral lung in the radiation plan was delineated by excluding the GTV. Anemia (grade ≥ 2), leucopenia (grade ≥ 2), neutropenia (grade ≥ 2), lymphocytopenia (grade ≥ 3), thrombocytopenia (grade ≥ 2), and hypoalbuminemia (grade ≥ 1, < 3.3 g/dL following the reference range of the institution) were selected from the laboratory tests. The grading of abnormal laboratory tests was evaluated using the nadir of each value during CCRT. The cutoff points of abnormal laboratory test results were selected considering the prevalence of each abnormality. Logistic regression analyses were performed to identify the predictors of PCPC. p-values were adjusted by controlling the false discovery rate [10]. Covariates with adjusted p-values lower than 0.1 were integrated into the multivariate logistic regression.

3. Prognostic endpoints and statistics

The prognostic endpoints in this study were the disease progression rate (DPR) and OS. A DPR event was defined as any disease progression. An OS event was defined as the death of a patient by any cause. These events were measured from the start of CCRT. A Kaplan–Meier estimator was applied to calculate the DPR and OS. A log-rank test was conducted to compare prognostic endpoints. OS events were subcategorized into ‘early death’ and ‘late death.’ Early death was defined as death that occurred within 6 months from CCRT initiation. Otherwise, events were considered late death. The crude rate of early death from all OS events was compared between patients with and without PCPC.

Pearson’s chi-square test with or without Yate’s continuity correction and Student’s t test were used to comparing categorical and continuous variables between two patient groups, respectively. Data with p < 0.05 were defined as statistically significant. All statistical analyses were conducted using R ver. 4.2.1 (The R Foundation for Statistical Computing, Vienna, Austria).

1. Patient characteristics

The median follow-up period was 30 months (range, 1 to 125 months). The patients were predominantly male (84.2%) with good performance (Eastern Cooperative Oncology Group scale 0–1, 98.7%) at initial diagnosis. Most patients (79.8%) were current or ex-smokers. Similar rates of adenocarcinoma (44.5%) and squamous cell carcinoma (46.1%) were reported. More than half (54.9%) of the patients had stage IIIA disease. As concurrent chemotherapy, carboplatin and paclitaxel were administered to 71.9% of the patients, and docetaxel and cisplatin were given to 24.9% of the patients. The percentage of the patients who received VMAT was slightly higher than that of the patients who underwent 3D-CRT (52.1% vs. 47.9%). Furthermore, 23 patients (7.3%) underwent consolidative durvalumab. The patient characteristics and details of the delivered treatment in terms of the occurrence of PCPC are summarised in Table 1. Patients with PCPC were older (median, 70 years vs. 65 years), and the number of ex-smokers was higher than that of current smokers (54.7% vs. 36.7%). The percentage of patients who had PCPC and were diagnosed with stage IIIA (62.3% vs. 53.4%) and IIIC disease (15.1% vs. 7.6%) was higher. In the lung function test, the patients who experienced PCPC had a lower absolute value of FVC, absolute value of DLCO, and relative rate of DLCO. A higher rate of utilization of VMAT was observed in patients with PCPC (66.0% vs. 49.2%). The patients with PCPC had larger PTV, higher bilateral lung doses (mean, V5, V20), and heart doses.

Table 1

Patient characteristics and treatment details

2. Adverse events

PCPC was reported in 53 patients (16.7%). Amongst them, 18 patients (34.0%) visited hospitals due to radiation pneumonitis, and 18 cases (34.0%) were due to infectious pneumonia. The 17 remaining cases (32.1%) had pulmonary inflammation, but they could not be easily categorized into radiation pneumonitis or infectious pneumonia. These pulmonary adverse events were considered superimposed events. Furthermore, 13 patients (4.1%) died from PCPC. The adverse event grading of symptomatic radiation pneumonitis and abnormal laboratory results is summarised in Table 2. Symptomatic (grade ≥ 2) pneumonitis was presented in 85 patients (26.8%), including 22 severe (grade ≥ 3) cases (6.9%). Infectious pneumonia was reported in 43 patients (13.6%), including 30 severe cases (9.5%). The most common grade 3–4 adverse events were lymphocytopenia (82.3%), leucopenia (19.9%), and neutropenia (12.4%). Anemia, thrombocytopenia, and hypoalbuminemia were also common, but no grade ≥ 3 events amongst these cases were reported.

Table 2

Toxicity grading of symptomatic radiation pneumonitis and abnormal laboratory results

3. Association of post-chemoradiotherapy pulmonary complications with prognostic endpoints

The prognostic effect of PCPC on DPR and OS was investigated (Fig. 1A and B). For DPR, the 1-, 2-, and 3-year rates were 51.9%, 70.7%, and 77.1% in patients without PCPC and 64.2%, 77.0%, and 87.7% in patients with PCPC, respectively. The DPR did not significantly differ between the two groups (p=0.087). For OS, the 1-, 2-, and 3-year rates were 85.5%, 67.0%, and 55.8% in patients without PCPC and 45.2%, 35.0%, and 24.4% in patients with PCPC, respectively. The OS differed significantly between the two groups (p < 0.001). Furthermore, the effect of symptomatic (grade ≥ 2) pneumonitis on DPR and OS was investigated. The log-rank test revealed that symptomatic pneumonitis was not significantly associated with DPR (p=0.906) and OS (p=0.198) (Fig. 1C and D).

Fig. 1

Actuarial rates of disease progression (A) and overall survival (B) in the presence of post-chemoradiotherapy pulmonary complications (PCPC), and disease progression (C) and overall survival (D) in the presence of symptomatic pneumonitis.

A total of 149 patients (56.4%) without PCPC died, whilst 42 patients (79.2%) with PCPC died. Furthermore, 12 early deaths (8.1%) were recorded in patients without PCPC, and 23 early deaths (54.8%) were documented in patients with PCPC. The rate of early deaths differed significantly between the two groups (p < 0.001).

4. Prediction of post-chemoradiotherapy pulmonary complications

Covariates potentially associated with PCPC were explored via univariate and multivariate logistic regression (Table 3). Univariate logistic regression showed that age at diagnosis, amount of smoking, FVC, DLCO, percentage predicted value of DLCO, radiation therapy technique, bilateral lung mean dose, bilateral lung V5, bilateral lung V20, mean heart dose, size of PTV, grade ≥ 2 anemia, grade ≥ 3 lymphocytopenia, and grade ≥ 1 hypoalbuminemia were significantly associated with PCPC. As two DLCO-related covariates and three lung dose-related covariates were selected, the percentage predicted value of DLCO, bilateral lung mean dose, and bilateral lung V20 were removed from the multivariate analysis based on p-value comparisons to avoid collinearity. Twelve covariates were incorporated into the multivariate logistic regression. DLCO (per mL/min/mmHg; odds ratio [OR], 0.855; 95% confidence interval [CI], 0.743 to 0.974; p=0.022), bilateral lung V5 (per 10%; OR, 1.872; 95% CI, 1.336 to 2.699; p < 0.001), and grade ≥ 1 hypoalbuminemia (OR, 5.670; 95% CI, 2.487 to 13.40; p < 0.001) remained as statistically significant factors in the multivariate analysis. A nomogram was constructed using these three factors (Fig. 2A). The receiver operating characteristic curve of the total points from the nomogram and occurrence of PCPC is illustrated in Fig. 2B. The area under the curve was 0.8545 (95% CI, 0.8020 to 0.9069). The calibration plot for the nomogram with bootstrapping of 100 repetitions is presented in Fig. 2C.

Table 3

Univariate and multivariate analyses of the occurrence of post-chemoradiotherapy pulmonary complications

Fig. 2

Prediction nomogram for post-chemoradiotherapy pulmonary complications. Nomogram to predict the probability of post-chemoradiotherapy pulmonary complications (A), receiver operating characteristic curve (B), and calibration plot with bootstrapping of 100 repetitions for the nomogram (C). DLCO, diffusing capacity of carbon monoxide.