Characteristics of patients
A total of 1063 patients were included in this study, 272 (25.6%) patients were divided into adjuvant chemotherapy cohort and 791 (74.4%) divided into surgery alone cohort. The most commonly used regimens were pemetrexed plus carboplatin (n =91; 33.7%), pemetrexed plus cisplatin (n =52; 19.3%), pemetrexed plus nedaplatin (n =32; 11.9%), paclitaxel plus cisplatin (n =20; 7.4%), paclitaxel plus nedaplatin (n =11; 4.1%), paclitaxel plus carboplatin (n =16; 5.9%) and gemcitabine plus cisplatin (n =8; 3.0%). Other rarely used regimen included gemcitabine plus carboplatin or nedaplatin, vinorelbine plus cisplatin and paclitaxel plus lobaplatin. In addition, 30 patients were prescribed with only one single chemotherapeutic drug, including pemetrexed (n =26; 9.6%) or carboplatin (n =4;1.5%).
The detailed characteristics of patients before and after PSM are presented in Table 1. Compared to those in the surgery only cohort, patients in adjuvant chemotherapy cohort were younger (p < 0.001), had lower rates of adenocarcinoma (p = 0.002), had more rates of poor differentiation (p = 0.016) and visceral pleural invasion (p = 0.034). Beyond these, other baseline characteristics between the two cohorts were not significantly different. After PSM, 270 pairs of patients were matched in a 1:1 ratio in these two cohorts. The baseline clinicopathological characteristics were between the 2 cohorts were well-balanced.
Survival analyses before and after PSM
The median follow-up time for entire patients was 38.6 months. In this study, six variates were considered as risk factors, including poorly differentiated tumors, lymphovascular invasion (LVI), sublobectomy, size of tumor > 4 cm, visceral pleural invasion (VPI), and less than 6 lymph nodes. Based on their detailed risk factors, we calculated the CRS of patients from 0 to 5. Before PSM, Patients with higher CRS had a worse OS (P < 0.001; Fig. 1A) and DFS (P < 0.001; Fig. 1B). Then we divided all enrolled patients into two subgroups on basis of their CRS i.e., the score of 0-1group (low risk) vs the score of 2-5 group (high-risk). Compared to patients in the high-risk group, patients in the low-risk group had significantly longer OS (5-year OS rate 0-1 vs 2-5: 88.3% vs 77.4%, P < 0.001; Fig. 1C) and DFS (5-year DFS rate 0-1 vs 2-5: 71.2% vs 64.6%, P = 0.027; Fig. 1D). In addition, a significant improvement of OS was observed in the adjuvant chemotherapy cohort compared to those who in surgery only cohort (5-year OS rate 87.4% vs 80.5%, P = 0.031; Fig. 2A), but no significant difference in DFS was found between them (5-year DFS rate 71.0% vs 66.3%, P = 0.097; Fig. 2B) before PSM.
After PSM, patients in adjuvant chemotherapy cohort were found to better survival in both OS (5-year OS rate 87.3% vs 78.5%, P = 0.021; Fig. 2C) and DFS (5-year DFS rate 70.9% vs 64.6%, P = 0.029; Fig. 2D).
Table 2 shows the results of univariate and multivariate Cox proportional hazards regression for survival of the two cohorts. It showed that adjuvant chemotherapy was an independent prognostic for OS (HR=0.561, 95%CI 0.348-0.903, P = 0.017) and DFS (HR=0.688, 95%CI 0.492-0.961, P = 0.028). Patients who had lymphovascular invasion (HR=1.758,95%CI 1.114-2.775, P = 0.015) had a shorter DFS and patients who were older had a worse OS (HR=1.034, 95%CI 1.007-1.063, P = 0.014). Besides, female patients were also identified as an independent favorable predictor for both OS (HR=0.389, 95%CI 0.216-0.702, P = 0.002) and DFS (HR=0.625, 95%CI 0.435-0.897, P = 0.011).
Patients with higher CRS had a worse OS (P = 0.046; Fig. 3A) but no significant differences in DFS among these with different CRS were observed (P = 0.577; Fig. 3B). Similarly, compared to patients in the high-risk group, patients in the low-risk group had significantly longer OS (5-year OS rate 0-1 vs 2-5: 89.0% vs 79.6%, P = 0.036; Fig. 3C). whereas no significant DFS difference was found between these two subgroups (5-year DFS rate 0-1 vs 2-5: 68.8% vs 67.0%, P = 0.850; Fig. 3D).
We performed subgroup survival analyses based on the quantized CRS in patients after PSM (Fig. 4). In low-risk subgroup, patients in adjuvant chemotherapy cohort did not have a better OS and DFS (5-year OS rate 91.4% vs 86.5%, P = 0.231, Fig. 4A; 5-year DFS rate 74.2% vs 63.5%, P = 0.093, Fig. 4B). However, as for high-risk patients (risk scores ≥ 2), although patients in adjuvant chemotherapy cohort did not have better DFS, adjuvant chemotherapy was observed to significantly improve the patients’ OS (5-year OS rate 84.7% vs 73.2%, P = 0.038, Fig. 4C; 5-year DFS rate 68.8% vs 65.3%, P = 0.154, Fig. 4D).
For patients with a score of 2-5, adjuvant chemotherapy was an independent prognostic factor in multivariable analysis for OS (HR=0.535, 95%CI 0.325-0.880, P = 0.014) (Table 3). Also, patients who had lymphovascular invasion (HR=1.654, 95%CI 1.030-2.657, P = 0.037) had a worse DFS time and patients who were younger survived a longer time (HR=1.034, 95%CI 1.004-1.065, P = 0.027).
Exploratory analyses for patients with EGFR gene test
In order to identify explore the impact of EGFR mutations on adjuvant chemotherapy, we performed exploratory analyses in patients who accepted EGFR gene test. 164 of 270 patients without adjuvant chemotherapy and 199 of 270 patients with adjuvant chemotherapy accepted EGFR gene test and 141 patients (38.8%) had activating mutations in EGFR. 68 of 164 patients without adjuvant chemotherapy and 73 of 199 patients with adjuvant chemotherapy had activating mutations in EGFR. The mutation rate of two group did not have statistic difference (P=0.352).
Among patients with the wild-type EGFR, those who receiving adjuvant chemotherapy had better OS and DFS (5-year OS rate 85.1% vs 70.0%, P = 0.009, Fig. 5A; 5-year DFS rate 69.4% vs 58.6%, P = 0.035, Fig. 5B). However, among patients with the activating mutations in EGFR, those who received adjuvant chemotherapy had numerically but no statistically poor OS and DFS (5-year OS rate 91.4% vs 86.5%, P = 0.552, Fig. 5C; 5-year OS rate 84.9% vs 96.8%, P = 0.803, Fig. 5D).
For patients with wild-type EGFR, adjuvant chemotherapy was a positive independent prognostic factor in multivariable analysis for OS (HR=0.397, 95%CI 0.206-0.763, P = 0.006) and DFS (HR=0.551, 95%CI 0.334-0.911, P = 0.020) (Table 4). Also, female patients (HR=0.346, 95%CI 0.206-0.763, P = 0.019) had a lower recurrent risk.
Interaction analyses
As shown in Table 5, after adjusting for other factors, the interaction analysis showed an apparent interaction effect between adjuvant chemotherapy and activating mutations in EGFR on OS (HR (Adjuvant chemotherapy * Activating Mutations in EGFR) =4.491, 95%CI 1.028-19.616, P = 0.046) but not on DFS (HR (Adjuvant chemotherapy * Activating Mutations in EGFR) =2.045, 95%CI 0.843-4.959, P = 0.113). The positive impact of adjuvant chemotherapy (HR=0.381, 95%CI 0.199-0.729, P = 0.004) and activating mutations in EGFR (HR=0.272, 95%CI 0.081-0.917, P = 0.036) on OS were not independent. Adjuvant chemotherapy (HR=0.590, 95%CI 0.356-0.976, P = 0.040) instead of activating mutations in EGFR (HR=0.216, 95%CI 0.307-1.254, P = 0.184) showed a positive impact on DFS
After excluding 31 (11.5%) patients who had received single-drug chemotherapy, all chemotherapy regimens in this study were platinum-based chemotherapy and were divided into non-pemetrexed plus cisplatin chemotherapy (61 patients) and pemetrexed plus cisplatin chemotherapy (178 patients). As shown in Table 6, the interaction analyses showed an apparent interaction effect between pemetrexed plus cisplatin chemotherapy and non-squamous cell carcinoma on OS (HR (Adenocarcinoma * pemetrexed plus cisplatin) =0.090, 95%CI 0.010-0.838, P = 0.034) (HR (Others * pemetrexed plus cisplatin) =0.078, 95%CI 0.007-0.834, P = 0.035).