Patients’ characteristics
The clinical characteristics of the 300 patients (77 women, 223 men) with stage I–III GC who underwent curative gastrectomy are summarized in Table 1. The median age was 70 years (range: 36–95 years), and the median body mass index was 21.9 kg/m2 (range: 14.8–37.2 kg/m2). Fifty-five patients had diabetes mellitus, 44 patients had cardiac comorbidities, and 16 patients had pulmonary comorbidities. The surgical approaches were as follows: 187 patients underwent distal gastrectomy, 97 underwent total gastrectomy, and 16 underwent proximal gastrectomy. The median operative time was 277 min (range: 85–822 min), and the median estimated blood loss was 178 mL (range: 5–4031 mL). One hundred thirty-five patients, 74 patients, or 91 patients were diagnosed with pathological stage I, II, or III GC, respectively. The median preoperative NLR, PLR, PNI, CAR, and LCR were 2.3 (range: 0.4–18.7), 140.0 (range: 31.3–555.3), 50.1 (range: 30.2–68.7), 0.05 (range: 0.02–3.68), and 8501 (range: 134–51392), respectively.
Relationships between postoperative complications and the systemic inflammatory-based prognostic parameters in stage I–III GC patients
One hundred and one patients (33.7%) developed a postoperative complication of Clavien–Dindo grade II–IV. Thirty-five patients (11.7%) developed anastomotic complications, namely leak (14 patients, 4.7%), stenosis (15 patients, 5.0%), or hemorrhage (6 patients, 2.0%). Sixteen patients (5.3%) had pancreatic fistula, 14 patients (4.7%) had pneumonia, and 10 patients (3.0%) had ileus (Table 2).
To evaluate the relationships between each postoperative complication and preoperative NLR, PLR, PNI, CAR, or LCR, all systemic inflammatory-based prognostic parameters were assessed using ROC curves. For all postoperative complications, the AUCs for preoperative NLR, PLR, PNI, CAR, or LCR were 0.599, 0.527, 0.477, 0.545, or 0.471, respectively. For the anastomotic complications, the AUCs for preoperative NLR, PLR, PNI, CAR, or LCR were 0.446, 0.412, 0.534, 0.514, or 0.520, respectively. For pancreatic fistula, the AUCs for preoperative NLR, PLR, PNI, CAR, or LCR were 0.744, 0.696, 0.414, 0.485, or 0.504, respectively. Interestingly, the AUCs for preoperative NLR, PLR, PNI, CAR, or LCR for postoperative pneumonia were 0.817, 0.704, 0.252, 0.729, or 0.276, respectively, and postoperative pneumonia was the most sensitive parameter related to preoperative NLR, PLR, PNI, CAR, or LCR (Table 2). Therefore, postoperative pneumonia was identified as the postoperative complication closely related to the systemic inflammatory-based prognostic parameters. Using the ROC curves, the cutoff value for preoperative NLR, PLR, PNI, CAR, or LCR was set at 2.9 (sensitivity, 78.6%; specificity, 70.6%), 154.8 (sensitivity, 71.4%; specificity, 60.8%), 47.7 (sensitivity, 64.3%; specificity, 68.5%), 0.07 (sensitivity, 71.4%; specificity, 66.4%), or 5273 (sensitivity, 71.4%; specificity, 69.9%), respectively (Fig. 1).
Relationships between postoperative pneumonia and prognosis in stage I–III GC patients
The relationships between all postoperative complications and the prognosis of stage I–III GC patients were assessed, and we identified a significant difference between patients with and without any postoperative complication regarding OS (P=0.002) and RFS (P=0.013) (Fig. 2a and b). We also assessed the relationship between each postoperative complication and the prognosis of stage I–III GC patients. Regarding OS, postoperative pneumonia was the only complication with a significant difference; no significant difference was found for the other complications. Regarding RFS, postoperative pneumonia and pancreatic fistula showed a significant difference (Table 3). Considering each of these findings, postoperative pneumonia was the only complication showing a significant difference for OS (P<0.001) and RFS (P=0.009), and also affected the prognosis of stage I–III GC patients (Fig. 3a and b).
Identifying a useful predictor of postoperative pneumonia in stage I–III GC patients
To identify useful predictors for postoperative pneumonia related to poor prognosis in stage I–III GC patients, 27 clinical factors, including preoperative NLR, PLR, PNI, CAR, and LCR, were assessed using univariate analysis. Statistical significance in multiple comparisons in the univariate analysis was adjusted using a false discovery rate. Univariate analysis after false discovery rate correction showed that preoperative NLR (odds ratio (OR), 8.817; 95% confidence interval (CI), 2.399–32.409; P=0.001), preoperative LCR (OR, 0.172; 95% CI, 0.052–0.564; P=0.004), and total gastrectomy (OR, 3.284; 95% CI, 1.513–7.130; P=0.003) showed a significant difference. Moreover, multivariate analysis of the three significant factors identified using univariate analysis revealed that preoperative NLR (OR, 5.228; 95% CI, 1.269–21.541; P=0.022) and total gastrectomy (OR, 2.824; 95% CI, 1.217–6.556; P=0.016) were independent risk factors for postoperative pneumonia (Table 4).