Table 1 summarizes the distribution of the PLC1–3 results, which were categorized into eight transition patterns. Of the 439 patients, 398 remained negative at all three time points (consistently negative group), whereas 41 showed positivity at one or more points. Among these 41 patients, 20 were negative at PLC3 and 21 were positive at PLC3, either persistently positive (PLC1 + / PLC3 +) or converting from negative to positive (PLC1– / PLC3 +) (Fig. 1).

Transition of serial PLC among any-positive cases (n = 41). Patients with at least one positive PLC were stratified by the results immediately after thoracotomy (PLC1) and at chest closure (PLC3) to depict transition across surgery. Four patterns were observed: persistently positive (PLC1 + /PLC3 +), conversion-to-positive (PLC1–/PLC3 +), conversion-to-negative (PLC1 + /PLC3–), and PLC2-only positive (PLC1–/PLC3– with PLC2 +). The asterisk denotes cases that were positive only after lung resection (PLC2 +), while PLC1 and PLC3 were negative

Table 2 compares the clinicopathological characteristics between consistently negative patients and those with positive PLC. The median age was 70 years, 233 patients (53%) were male, and 199 (45%) were smokers. Smoking was significantly more frequent in the any-positive group (p = 0.006). Surgical procedures included lobectomy (n = 347), wedge resection (n = 44), segmentectomy (n = 41), and pneumonectomy (n = 7) with lymph node dissection in 314 patients (72%). No significant surgical differences were observed between the two groups. Histologically, adenocarcinoma accounted for 380 cases (86%), squamous cell carcinoma for 52 (12%), and others for 7 (2%). Pathological nodal involvement was found in 120 patients (27%), LVI in 144 patients (33%), and PL in 88 patients (20%). Only PL was significantly more frequent in the any-positive group (p < 0.001).

The median follow-up was 24.5 months (95% CI 22.8–26.8). Overall, 69 RFS primary events and 23 competing (non-lung cancer) deaths occurred during the follow-up period. The analysis of RFS demonstrated that patients with PLC3 + had significantly higher recurrence rates than consistently negative cases (p < 0.001), whereas patients with PLC3– showed outcomes similar to consistently negative patients (p = 1.000) (Fig. 2a). Among patients with ≥ 1 positive PLC, we stratified the outcomes by the PLC1/PLC3 combination into three groups: post-closure negative (PLC3 −), persistently positive (PLC1 + /PLC3 +), and conversion-to-positive (PLC1 − /PLC3 +). There was no significant difference between PLC3 − and PLC1 − /PLC3 + (p = 0.66), whereas the persistently positive group showed worse RFS than the conversion-to-positive group (PLC1 − /PLC3 + vs. PLC1 + /PLC3 + : p = 0.049) (Fig. 2b).

RFS according to PLC transition. a Kaplan–Meier RFS curves comparing consistently negative, closure-negative (PLC3–), and closure-positive (PLC3 +) groups. Pairwise log-rank p-values with Bonferroni adjustment were annotated (e.g., consistently negative vs. PLC3 + : p < 0.001; closure-negative vs. PLC3 + : p = 0.049; consistently negative vs. PLC–: p = 1.0). b RFS in any-positive PLC patients (n = 41) stratified by the PLC1/PLC3 combinations. Patients with at least one positive PLC were grouped by the result of PLC1 and PLC3 into three categories: post-closure negative (PLC3–), persistently positive (PLC1 + /PLC3 +), and conversion-to-positive (PLC1–/PLC3 +). Pairwise log-rank p-values with Bonferroni adjustment were annotated (e.g., PLC3 vs. PLC1–/PLC3 + : p = 0.66; PLC1–/PLC3 + vs. PLC1 + /PLC3 + : p = 0.049)

The cumulative incidence of pleural recurrence or malignant pleural effusion was significantly higher in PLC3-positive patients than in PLC3-negative patients (p = 0.003; Fig. 3a). Stratification by the total number of PLC-positive results (single vs. double vs. triple; Fig. 3b) showed that triple-positive patients had a significantly higher cumulative incidence of pleural recurrence than single-positive patients (p = 0.046), whereas the double- vs. triple-positive comparison was not significant (p = 0.378). Further analysis based on the PLC1/PLC3 combinations defined three groups: (1) PLC3-negative (regardless of PLC1), (2) persistently positive (PLC1 + /PLC3 +), and (3) conversion-to-positive (PLC1–/PLC3 +). The persistently positive group had a significantly higher incidence of pleural recurrence than the PLC3-negative group (p = 0.0005) and showed a non-significant trend toward higher incidence relative to the conversion-to-positive group (p = 0.059; Fig. 3c).

Cumulative incidence of pleural dissemination and malignant effusion. a Cumulative-incidence curves according to PLC3 status PLC3– vs. PLC3 + (p = 0.003). b Cumulative incidence according to the number of positive cytology results (Single, Double, Triple) Bonferroni-adjusted pairwise Gray’s tests: single vs. triple p = 0.046, double vs. triple p = 0.378. c Cumulative incidence by specific PLC1–PLC3 combinations. Bonferroni-adjusted pairwise Gray’s tests: PLC3 − vs. persistently positive (PLC1 + /PLC3 +) p = 0.0005; PLC3 vs. conversion-to-positive (PLC1 − /PLC3 +) p = 0.059

The clinicopathological correlates are shown in Table 3. PLC3-positive patients had significantly higher rates of pathological lymph node metastasis (p = 0.003), lymph node dissection (p = 0.004), and LVI (p = 0.012) in comparison to PLC3-negative patients.

Table 4 summarizes the Fine and Gray competing risks regression for RFS. In the univariate analysis, several factors were significantly associated with RFS. We then evaluated two scoring systems reflecting serial PLC results: (1) the number of positive PLC results (0–3) and (2) the combined PLC1/PLC3 pattern (PLC3-negative = 0, PLC1-negative/PLC3-positive = 1, PLC1-positive/PLC3-positive = 2). Both variables were strongly associated with RFS.

In the multivariate model, adjusted for non-lung cancer deaths, which included four covariates (PLC1 positivity, PLC3 positivity, LVI, and pathological nodal involvement), PLC3 positivity and LVI remained independent predictors of poor RFS (subdistribution hazard ratio [SHR] 3.056, 95% CI 1.663–5.616; p < 0.001 for PLC3; SHR 4.031, 95% CI 2.395–6.786; p < 0.001 for LVI), whereas PLC1 positivity (SHR 1.751, 95% CI 0.827–3.710; p = 0.14) and nodal status did not.

To further clarify the prognostic utility of scoring systems that incorporate serial PLC results (PLC1–3), we performed two additional multivariate analyses. First, a model that included the number of positive PLC results (0–3), maximum tumor size, LVI, and pathological nodal involvement demonstrated that the number of positive PLC results was an independent predictor of RFS (SHR 1.510, 95% CI, 1.133–2.014; p = 0.005; Supplementary Table 1). Second, when the combined PLC1/PLC3 pattern was analyzed together with maximum tumor size, LVI, and pathological nodal involvement, the PLC1/PLC3 score remained an independent predictor (SHR 1.64, 95% CI 1.12–2.41; p = 0.012; Supplementary Table 2).