Of the 204 patients reviewed, 121 patients matched the recruitment criteria while 83 patients were excluded from the analysis with reasons as follows: stage IV (n = 21), radiation dose < 50 Gy (n = 23), post-CCRT follow-up < 3 months (n = 32), histology other than squamous cell carcinoma (n = 5), and use of 3-dimensional conformal radiotherapy (n = 23). Table 1 summarized demographic and clinical characteristics of the 121 patients, including five women and 116 men. Upper or middle esophagus was involved by the tumor in 104 (86.0%) patients. Furthermore, six (5.0%) patients had a history of cardiovascular diseases (1 coronary artery disease, 3 congestive heart failure, 1 aortic valve infectious endocarditis after valve replacement, and 1 arrhythmia) at baseline. The six patients were considered medically fit for CCRT under the suggestion of our institutional multidisciplinary esophageal cancer team.
The utilized chemotherapy regimens and radiation doses were summarized in the Additional file 2: Table S1. The median radiation dose was 61.2 Gy (range, 50-66.6 Gy). The radiation doses were not different between patients with or without pre-existing cardiovascular diseases (p = 0.613). Fluoropyrimidine-based chemotherapy regimens were used in 113 (93.4%) patients. Most patients received either cisplatin (25 mg/m2) plus fluorouracil (1000 mg/m2) given intravenously every week or cisplatin (20 mg/m2 daily, on day 1–4) plus fluorouracil (800 mg/m2 daily, on day 1–4) given intravenously every 4 weeks. Other regimens were utilized at the discretion of physicians. Furthermore, during CCRT, enteral nutrition support was given via nasogastric, percutaneous endoscopic gastrostomy, and feeding jejunostomy tubes in eight (6.6%), 11 (9.1%), and 17 (14.0%) patients, respectively. Medications for emesis or pain as well as intravenous hydration were given as clinically indicated.
Clinical characteristics associated with overall survival
The median follow-up was 16.2 (range, 4.3-109.3) and 21.2 months (range, 5.8-104.6) for the whole cohort and the surviving patients, respectively. The median OS of the whole cohort was 18.4 months (Fig. 1a). In univariate analysis, body mass index, body surface area, Eastern Cooperative Oncology Group (ECOG) performance status, stage, chemotherapy regimens, and the volume of PTV prescribed to 50 Gy were associated with OS, but the pre-existing cardiovascular diseases was not correlated with OS (p = 0.87, Table 2). Moreover, ECOG performance status, stage, chemotherapy regimens, and the volume of PTV prescribed to 36 Gy and 50 Gy were independent prognostic factors of OS by multivariate analysis (Additional file 3-9: Table S2-8). The median OS was longer for patients with ECOG performance status 0-1 (19.0 vs. 6.8 months, p < 0.001; Fig. 1b), stage I-II (not reached vs. 17.7 months, p = 0.022; Fig. 1c), fluoropyrimidine-based chemotherapy (19.0 vs. 11.2 months, p = 0.0016), and volume of PTV prescribed to 50 Gy ≤ 640 ml (27.4 vs. 16.9 months, p = 0.008; Fig. 1d). On the other hand, the median volume of overlap between PTV and the heart was 25.1 ml. The median OS was not statistically different between patients with the overlapping volume > 25.1 ml and ≤ 25.1 ml (18.4 vs. 18.1 months, p = 0.192).
Dose-volume parameters associated with overall survival
Heart V5, V10, and V20 consistently served as independent prognostic factors of OS under consideration of individual pulmonary dose-volume parameters (Table 3 and Additional file 3-9: Table S2-8). The median heart V5, V10, and V20 were 94.3%, 86.4%, and 76.9%, respectively. A longer median OS was observed among patients with heart V5 ≤ 94.3% (24.7 vs. 16.3 months, p = 0.0025; Fig. 2a), heart V10 ≤ 86.4% (24.8 vs. 16.9 months, p = 0.0041; Fig. 2b), and heart V20 ≤ 76.9% (19.0 vs. 17.2 months, p = 0.047; Fig. 2c). In addition, mean lung dose was consistently shown to be a prognostic factor of OS in analytic models including different cardiac dose-volume parameters (Additional file 3-10: Table S2-9). Patients with mean lung dose ≤ 12.63 Gy had a superior median OS (24.8 vs. 17.5 months, p = 0.017; Fig. 2d)
Dose-volume parameters associated with symptomatic cardiac adverse events
There were 12 symptomatic CCRT-related cardiac adverse events, including ischemic heart disease in one, arrhythmia in three, and pericardial effusion in eight patients. The median interval from the start of IMRT to development of cardiac events was 9.8 months. The pre-existing cardiovascular diseases were not associated with the cumulative incidence of cardiac adverse events after CCRT (p = 0.459). Lower cumulative incidence of symptomatic cardiac adverse events was found among patients with heart V5 ≤ 94.3% (p = 0.017; Fig. 3a), heart V10 ≤ 86.4% (p = 0.02; Fig. 3b), and heart V20 ≤ 76.9% (p = 0.0057; Fig. 3c). Moreover, patients without symptomatic cardiac adverse events had a higher 3-year OS rate (33.8% vs. 0%, p = 0.03; Fig. 3d). There was a trend toward better survival at 2 years in patients without symptomatic cardiac complications (44.3% vs. 25.0%, p = 0.23).
Cancer-specific survival and cause of death
The causes of deaths were shown in Table 4. Fifty-eight (47.9%) patients died of esophageal cancer with or without other etiologies. The median esophageal cancer-specific survival was 24.7 months. As shown in the Additional file 11: Fig. S2, a longer median esophageal cancer-specific survival was observed among patients with heart V5 ≤ 94.3% (36.4 vs. 21.2 months, p = 0.042), heart V10 ≤ 86.4% (36.4 vs. 19.0 months, p = 0.032), and volume of PTV prescribed to 50 Gy ≤ 640 ml (37.2 vs. 19.0 months, p = 0.005). In addition, there was a trend toward better esophageal cancer-specific survival in cases with heart V20 ≤ 76.9% (p = 0.2). But the association between the tumor location and esophageal cancer-specific survival was not found (p = 0.67).