Clinicopathological baseline characteristics across low and high ALI
Before PSM, the optimal cut-off ALI value for predicting CSS based on X-tile calculation was 43.0. Cases were divided into a low ALI cohort (<43.0) and a high ALI cohort (≥43.0). Nil major variations existed across sex, ASA score, tumor location, histology-based grade, surgical method, lymphadenectomy, intraoperative blood loss, and postoperative adjuvant chemotherapy across both cohorts (P > 0.05). Cases with low ALI had older age (> 65 years :24.9% vs 15.4%, P = 0.004), high tumor invasions (T3/T4a:64.2% vs 45.6%, P < 0.001), more lymph node metastases (N3:5.8% vs 3.0%; N2:23.9% vs 13.8%, P < 0.001), high TNM stage (III/IVA:53.6% vs 34.8%, P < 0.001), lower BMI (21.41±2.69 vs 22.98±2.97,P < 0.001), lower ALB (38.86 ±4.52 vs 42.31 ±3.80, P < 0.001) and lower NLR (2.70 ±0.70 vs 1.64 ±0.34, P < 0.001) (Table 1).
After PSM, on comparison to high ALI cohort, low ALI cohort had reduced BMI (21.41 ±2.69 vs 22.98 ±2.97, P < 0.001), reduced ALB (38.86 ±4.52 vs 42.31 ±3.80, P < 0.001), and reduced NLR (2.70 ±0.70 vs 1.64 ±0.34, P < 0.001). However, nil major variations existed across age, sex, ASA score, tumor location, histology-based grade, tumor invasions, lymph node metastasis, surgical method, lymphadenectomy, intraoperative blood loss, and postoperative adjuvant chemotherapy across both cohorts (P > 0.05) (Table 1).
Association across ALI and short-term postoperative endpoint
Within cohort before PSM, no variations existed within incidence of total complications (46.8% versus 44.9%, P=0.652) and mild complications (21.5% vs 28.2%, P=0.058) across both ALI cohorts. Major variation was observed within incidence of severe complications (25.3% vs 16.7%, P=0.01). After PSM, the incidence of total complications (46.8% vs 46.0%; P=0.861), mild complications (21.5% vs 27.7%, P=0.104), and severe complications (25.3% vs 18.3%, P=0.06) within both cohorts were not significantly different (Table 2).
Relationship across ALI and long-term prognosis
Within cohort before PSM, the 1-, 3-, and 5-year CSS of the low ALI cohort were 94.0%,79.1%, and 74.6%, accordingly, with 1-, 3-, and 5-year CSS of the high ALI cohort were 87.3%,59.4%, and 48.9%, accordingly, demonstrating major variations across both cohorts (P < 0.001) (Fig. 1A). Similarly, within PSM cohort, 1-, 3-, and 5-year CSS for low ALI cohort were 94.1%,77.0%, and 72.2%, accordingly, while 1-, 3-, and 5-year CSS for high ALI cohort were 88.4%, 59.4%, and 48.9%, accordingly, with major variations across both cohorts (P < 0.001) (Fig. 1B).
Table 3 shows univariate / multivariate analysis results before and after PSM. Before PSM, multivariate analysis demonstrated TNM stage (II, HR:3.371, P < 0.001; III/IVA, HR:5.853, P < 0.001), intraoperative blood loss (≥200 mL, HR:2.147, P < 0.001) and low ALI (HR:2.008, P=0.002) were independent risk factors for CSS. Post-matching, ASA score (III/IV, HR:1.531, P=0.027), TNM stage (II, HR:3.507, P < 0.001; III/IVA, HR:5.880, P < 0.001), intraoperative blood loss (≥200 mL, HR:2.162, P=0.001) and low ALI cohort (HR:2.113, P =0.001) were independent risk variables within adverse CSS.
Subgroup Analysis of ALI on Survival
Further subgroup analyses were performed within cohort before PSM. At stage I, nil major variations existed within 5-year CSS across both ALI cohorts (87.6% vs 89.2%, P=1.00), while at stages II and III, there was a major variation existed within 5-year CSS across both cohorts (II:75.9% vs 58.0%, P=0.006; III:60.0% vs 31.7%, P=0.002) (Fig. 2A–C). Similarly, within cohort after PSM, nil major variations existed within 5-year CSS across both ALI cohorts at stage I (86.3% vs. 89.2%, P=0.780). At stages II and III, there was still a major variation existing within CSS across both cohorts (II :75.3% vs 57.3%, P=0.006; III:61.7% vs 31.7%, P=0.001) (Fig. 2D–F).
Detailed subgroup analyses demonstrated that cases within high ALI cohort were at increased risk for adverse CSS in most of the clinicopathological features within cohort before PSM, except cases >65 years of age, stage I, and Ivor Lewis-MIE surgery (Fig. 3). Within cohort after PSM, cases within high ALI cohort had an increased risk of adverse CSS in most of the clinicopathological features, except for age >65 years, female gender, stage I, Lewis-MIE surgery, and intraoperative blood loss ≥200 mL (Fig. 4).
Comparison of ALI plus TNM staging model and TNM alone
To further verify the predictive value of ALI, we compared the predictive ability and clinical application value of model B (ALI+TNM stage) and model A (TNM stage alone). Within cohort before PAM, the t-AUC demonstrated that the AUC values of model B and model A were 0.692-0.712 and 0.677-0.670 during the follow-up period, accordingly. Within cohort after PSM, the t-AUC demonstrated that the AUC of model B and model A were 0.679-0.736 and 0.650-0.695, accordingly. Both before and after the PSM cohort, the predictive value of model B was better than that of model A (Fig. 5A-B).
Within cohort before PSM, the DCA demonstrated that model B had a greater clinical benefit rate than model A within range of 0.2 to 0.56. Within matched cohort after PSM, the DCA demonstrated that model B also had a greater clinical benefit rate than model A (0.16–0.52) (Fig. 5C-D).