Could less be enough: sublobar resection vs lobectomy for clinical stage IA non-small cell lung cancer patients with visceral pleural invasion or spread through air spaces

Abstract

Background:

While recent randomized controlled trials have demonstrated that sublobar resection (SLR) is non-inferior to lobectomy, the comparative efficacy of these procedures remains uncertain for early-stage non-small cell lung cancer (NSCLC; ≤3 cm) exhibiting invasive features postoperatively, such as visceral pleural invasion (VPI) or spread through air spaces (STAS).

Materials and Methods:

To identify eligible studies, a comprehensive search of PubMed, Embase, MEDLINE, the Cochrane Library, and Web of Science was conducted through 25 July 2024. Studies were screened according to predefined criteria in accordance with PRISMA guidelines. The primary endpoints were 5-year overall survival (OS) and recurrence-free survival (RFS). Hazard ratios (HR) and 95% confidence intervals (CI) were used to perform a meta-analysis.

Results:

The final analysis included 14 retrospective studies and 1 randomized controlled trial, encompassing a total of 8054 patients with NSCLC (tumors ≤3 cm) exhibiting VPI or STAS. The meta-analysis revealed that SLR was associated with impaired 5-year OS (HR: 1.25; 95% CI: 1.10–1.41) and slightly inferior RFS (HR: 1.25; 95% CI: 0.99–1.58) compared to lobectomy for pT2a (VPI) NSCLC patients with tumor ≤3 cm. Similarly, SLR was associated with significantly worse 5-year OS (HR: 2.58; 95% CI: 1.92–3.45) and 5-year RFS (HR: 2.42; 95% CI: 1.69–3.46) compared to lobectomy for stage IA NSCLC patients with STAS. Subgroup analysis revealed that statistically significant differences in 5-year OS (HR: 1.13; 95% CI: 0.92–1.38) and 5-year RFS (HR: 0.87; 95% CI: 0.56–1.36) were not observed between the SLR and lobectomy groups for pT2a (VPI) NSCLC patients with tumor ≤2 cm. Additionally, no statistically significant survival difference was observed between the segmentectomy and lobectomy groups for NSCLC patients (≤3 cm) with VPI (5-year OS: HR: 1.16; 95% CI: 0.89–1.52; 5-year RFS: HR: 1.07; 95% CI: 0.88–1.30) or STAS (5-year OS: HR: 3.88; 95% CI: 0.82–18.31; 5-year RFS: HR: 1.64; 95% CI: 0.70–3.80).

Conclusions:

For early-stage (≤3 cm) NSCLC with VPI or STAS, SLR was associated with worse survival outcomes compared to lobectomy. However, segmentectomy achieved survival outcomes comparable to those of lobectomy. For pT2a (VPI) NSCLC patients with tumor ≤2 cm, the differences in survival outcomes between SLR and lobectomy were not statistically significant.

HIGHLIGHTS

• Sublobar resection (SLR) demonstrates inferior 5-year overall survival and recurrence-free survival compared with lobectomy for clinical T1a-cN0M0 non-small cell lung cancer (NSCLC) with visceral pleural invasion (VPI) (upstage pT2aN0M0, IB) or spread through air spaces (STAS) (pT1N0M0, IA).

• SLR demonstrates comparable 5-year overall survival and recurrence-free survival compared with lobectomy for cT1a-bN0M0 NSCLC with VPI (pT2aN0M0, IB)

• For clinical Stage IA (T1a-cN0M0) NSCLC with VPI (pT2aN0M0, IB) or STAS (pT1N0M0, IA), segmentectomy achieves survival outcomes comparable to those of lobectomy.

Introduction

Non-small cell lung cancer (NSCLC) with visceral pleural invasion (VPI) exhibits increased aggressiveness and is associated with a higher risk of mediastinal lymph node metastasis and poorer prognosis.^[1–3] The eighth edition of the lung cancer staging system from the International Association for the Study of Lung Cancer reclassifies clinical T1 stage NSCLC with VPI as pathological T2^[4]. Similarly, spread through air spaces (STAS), a distinct pattern of lung cancer invasion^[5], has been observed in approximately 40% of patients with early-stage NSCLC^[6]. In 2015, the World Health Organization defined STAS as “micropapillary clusters, solid nests, or single cells extending beyond the tumor margin and infiltrating the alveolar spaces of the surrounding lung tissue.”^[7] Patients with STAS-positive tumors experience a worse prognosis and a higher risk of recurrence than those with STAS-negative tumors^[5,6,8]. While lobectomy has been the standard treatment for T1N0M0 NSCLC^[9], recent prospective randomized controlled trials demonstrate that sublobar resection (SLR), including segmentectomy, is not inferior to lobectomy in early-stage NSCLC^[10,11]. Notably, however, these trials did not consider the presence of VPI or STAS during randomization. Therefore, it remains unclear whether SLR or segmentectomy offers comparable survival benefits to lobectomy in early-stage (≤3 cm) NSCLC patients with VPI or STAS.

Evidence regarding the impact of VPI or STAS on surgical outcomes remains conflicting. Some studies indicate that for cT1N0M0 lung cancer patients with VPI (pT2a) or STAS (pT1), SLR is associated with worse recurrence-free survival (RFS) and overall survival (OS) compared to lobectomy,^[12–16] with significantly higher local and distant recurrence rates^[13,16]. Conversely, other studies suggest that SLR can achieve comparable benefits to lobectomy in these patients^[17,18]. However, existing comparative studies share a significant limitation: the failure to differentiate between segmentectomy and wedge resection (WR) within the broader category of SLR. This distinction is crucial because WR, a nonanatomical procedure, does not allow for hilar lymph node dissection, unlike segmentectomy^[19]. Therefore, it remains unclear whether the reduced survival benefit observed in some SLR studies is attributable to the inclusion of WR procedures. We hypothesize that in patients with NSCLC (tumor ≤3 cm) exhibiting VPI or STAS postoperatively, segmentectomy can achieve survival benefits comparable to those of lobectomy, and SLR could offer comparable oncological outcomes to lobectomy in patients with tumors ≤2 cm in size.

To investigate this hypothesis, we conducted a meta-analysis comparing the oncological outcomes of SLR (including segmentectomy) versus lobectomy in three cohorts of NSCLC patients: (1) those with tumors ≤3 cm and VPI (2) those with tumors ≤3 cm and STAS, and (3) those with tumors ≤2 cm and VPI.

Methods

Study design

This meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Supplementary Digital Content 1, http://links.lww.com/JS9/D936)^[20]. Furthermore, the methodological quality of this meta-analysis was evaluated using the AMSTAR 2 (A MeaSurement Tool to Assess Systematic Reviews) criteria (Supplementary Digital Content 2, http://links.lww.com/JS9/D937)^[21].

Search methods for identification of studies

Two investigators independently conducted a comprehensive literature search through 25 July 2024. The following databases were searched: PubMed, Embase, Cochrane Library, MEDLINE, and Web of Science. The search terms used were as follows: (visceral pleural invasion [Title/Abstract] OR spread through air spaces [Title/Abstract]) AND (segmentectomy [Title/Abstract] OR segment resection [Title/Abstract] OR wedge resection [Title/Abstract] OR sublobar resection [Title/Abstract] OR sublobectomy [Title/Abstract]) AND (lobectomy [Title/Abstract] OR lobar resection [Title/Abstract]). The detailed search strategy, including specific search terms for each database, is provided in Supplementary Digital Content, Table S1, http://links.lww.com/JS9/D933. Additionally, the reference lists of included studies were manually searched to identify any additional relevant articles.

Eligibility criteria

Studies were considered eligible for inclusion if they met the following criteria: (1) NSCLC (tumor ≤3 cm in size and no lymph node invasion) with VPI or STAS. (2) Compared SLR/segmentectomy and lobectomy with reported survival outcomes. (3) Survival outcomes included 5-year RFS and 5-year OS. (4) Confirmed VPI and STAS status through pathological examination.

Studies were excluded if they met any of the following criteria: (1) conference abstracts, case reports, reviews, or letters; (2) duplicates of other studies; (3) total quality assessment score <12 points; (4) overlapping pathological statuses of VPI and STAS; (5) inclusion of patients who received neoadjuvant therapy; (6) insufficient reporting of data on survival or other relevant outcomes.

Data collection and analysis

Data extracted from the included studies included: first author, publication year, study design, age, sex ratio, pulmonary function, comorbidity, smoking history, number of patients with VPI or STAS, number of patients who underwent SLR/segmentectomy/wedge resection (WR) and lobectomy, the number of patients receiving adjuvant therapy, tumor size, lung cancer stage (according to 8th edition lung cancer staging system^[4]), median follow-up time, and survival data. Two investigators independently performed data extraction. A third investigator resolved any discrepancies.

Hazard ratios (HRs) and 95% confidence intervals (CIs) were pooled using the inverse variance-weighted method in Review Manager version 5.4 (Cochrane Collaboration, UK). The threshold for statistical significance was set at P < 0.05. HRs and 95% CIs were either directly extracted from the included studies or estimated from published Kaplan–Meier curves^[22]. Heterogeneity among the included studies was assessed using the Higgins I² statistic and the χ² test. I² ≤ 50% or P ≤ 0.05 indicated low heterogeneity, in which case a fixed-effects model was applied. Otherwise, a random-effects model was used.

Publication bias and sensitivity analysis

Publication bias was assessed using a contour-enhanced funnel plot (CEFP), Egger’s test, and the trim-and-fill method in R 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria). Unlike traditional funnel plots, CEFPs delineate regions of statistical significance (P = 0.01, 0.05, 0.1), allowing for immediate visual identification of the significance levels of individual studies (represented as points). This facilitates the determination of whether funnel plot asymmetry arises from publication bias. When asymmetry is observed, a trim-and-fill analysis is conducted. Should the effect sizes of the filled studies fall within the nonsignificant region, publication bias is inferred as the likely source of asymmetry. Moreover, the trim-and-fill method serves as a robustness check for the pooled effect sizes of meta-analytic findings. A P-value ≤0.05 on Egger’s test was considered indicative of significant publication bias. Sensitivity analysis was performed using a leave-one-out approach to evaluate the robustness of the pooled effect sizes.

Quality assessment and subgroup analysis

Quality assessment of the included studies was conducted independently by two investigators using the Methodological Index for Non-Randomized Studies (MINORS), a scale designed for surgical intervention studies^[23]. MINORS comprises 12 items, each scored on a 0–2 point scale, with a maximum total score of 24 points. Studies scoring 0–8 points were considered low quality, 9–16 points moderate quality, and 17–24 points high quality. Studies scoring less than 12 points were excluded from our meta-analysis. The risk of bias assessment was conducted using ROBINS-I (Risk Of Bias In Non-randomized Studies—of Interventions), a validated tool for evaluating non-randomized interventional studies^[24]. Risk-of-bias assessment figures were generated using the online platform at www.riskofbias.info. The randomized control trail (RCT) was assessed by the risk of bias tool of Cochrane Collaboration. Subgroup analyses were performed for studies comparing segmentectomy versus lobectomy and for those specifically investigating pT2a (VPI) NSCLC with tumor ≤2 cm.

Results

Characteristics of included studies

Initially, 468 studies were retrieved. After removing 231 duplicate articles, 170 irrelevant studies, 6 nonclinical studies, 18 studies of data duplication, 27 studies without sufficient data, 1 study with a total MINORS score less than 12. Finally, 15 eligible studies^{[13,16,18,19,25–35]} comprising 8054 patients were identified after full-text review. The literature search and selection process is illustrated in Figure 1.

Figure 1.

The PRISMA literature search and selection process is illustrated in figure. PRISMA, preferred reporting items for systematic reviews and meta-analyses.

Table 1 summarizes the detailed characteristics of the 15 included studies, comprising 14 retrospective studies and one RCT. Among the retrospective studies, 10 were classified as high-quality and 4 as moderate-quality, based on an average MINORS score of 17 points (Table 2). Bias risk assessment indicated a moderate overall risk across the included studies (Supplementary Digital Content, Figure S1 http://links.lww.com/JS9/D923). Eight studies included patients with pT2aN0M0 (VPI) NSCLC, while seven included patients with pT1N0M0 NSCLC with STAS. Among the latter seven studies, three^[29,32,33] included patients who underwent SLR due to poor pulmonary function or severe comorbidities. Seven studies compared segmentectomy versus lobectomy for cT1 NSCLC patients, with four focusing on patients with VPI (pT2a) and three on patients with STAS (pT1). Five studies focused on NSCLC patients with tumor ≤2 cm and VPI. None of the patients in this study received neoadjuvant therapy. Of the selected cohort, 1789 patients underwent SLR, while 6265 underwent lobectomy. Subsequently, 98 (12.1%) of 813 patients in the SLR group and 904 (24.0%) of 3771 patients in the lobectomy group received adjuvant therapy. The mean age ranged from 64 to 71 years, and women accounted for 45.6% of the study population. Detailed baseline characteristics of age, pulmonary function, comorbidity, and smoking history are presented in Supplementary Digital Content, Table S2, http://links.lww.com/JS9/D934.

Table 1.

The characteristics of included studies

First author and publication year	Study design	Study population (VPI or STAS)	Age (average or median)	Sex ratio (female vs. male)	The number of VPI or STAS patients (Seg/SLR/WR vs. lob)	The number of patients receiving adjuvant therapy	Tumor size (median, range)	The 8th edition Lung Cancer TNM Stage	Median follow-up times (months)	Study endpoints
Schuchert 2012	RS without PSM	VPI	Seg: 70.0 ± 8.6	472 vs. 427	115 (39 vs. 76)	NR	seg:2.0 (1.5–2.8) cm;	pT1a-cN0M0	37	RFS
			Lob: 68.4 ± 10.1				lob:2.5 (1.8–3.0) cm
Mathey-Andrews 2024	RS with PSM	VPI	seg:70.0 (62–76)	1569 vs. 999	2227 (147 vs. 2080)	Seg:24 Lob:517	≤2 cm in size	cT1a-bN0M0 pT2aN0M0	NR	OS
			lob:71 (64–75)
Lula Lukadi 2023	RS with PSM	VPI	seg:67 (59–72)	155 vs. 98	253 (62 vs. 191)	Seg:8 Lob:37	≤3 cm in size	cT1a-cN0M0 pT2aN0M0	41	RFS, OS
			lob: 64 (58–71)
Kagimoto 2022	RS with PSM	VPI	seg:69 (66–75)	106 vs. 215	321 (80 vs. 241)	Seg:12 Lob:71	≤2 cm in size	cT1a-bN0M0 pT2aN0M0	32	RFS, OS
			lob:69 (62–75)
Kagimoto 2021	RS with PSM	STAS	seg:69 (64–75)	128 vs. 165	293 (107 vs. 186)	Seg:15 Lob:95	≤3 cm in size	cT1a-cN0M0	55	RFS, OS
			lob:68 (62–75)
Ikeda 2023	RS without PSM	STAS	69.4 (21–92)	260 vs. 295	148 (17 [Seg] vs. 112 vs. 19[WR])	25	≤3 cm in size	pT1N0M0	60.5	RFS, OS
Huang 2022	RS with PSM	VPI	70 (35–96)	1094 vs. 899	1497 (111 [Seg] vs. 1386 vs. 416 [WR])	80	≤3 cm in size	cT1a-cN0M0 pT2aN0M0	33	OS
Altorki 2024	RCT	VPI	66.8 (37.8–83.7)	61 vs. 52	113 (57 vs. 56)	None	≤2 cm in size	cT1a-bN0M0 pT2aN0M0	84	RFS, OS
Yu 2020	RS with PSM	VPI	NA	809 vs. 577	1386 (386 vs 1000)	SLR: 34 Lob:134	2 cm in size	cT1a-bN0M0 pT2aN0M0	58	OS
Shiono 2018	RS without PSM	STAS	SLR: 76 (68–80)	185 vs.329	104 (31 vs. 73)	SLR:5	≤3 cm in size	pT1a-cN0M0	48	RFS, OS
						Lob: 50
			Lob: 68 (60–74)
Yanagawa 2018	RS without PSM	STAS	NA	13 vs. 207	42 (8 vs. 34)	11	≤3 cm in size	pT1N0M0	55.2	RFS, OS
Ren 2019	RS without PSM	STAS	SLR: 65.9 ± 8.7	107 vs. 118	225 (43 vs. 182)	None	≤3 cm in size	pT1a-cN0M0	NR	RFS, OS
			Lob: 59.4 ± 8.5
Eguchi 2019	RS with PSM	STAS	SLR: 70 (64–76)	938 vs. 559	607 (218 vs. 389)	None	≤3 cm in size	pT1N0M0	NR	RFS
			Lob: 69 (61–75)
Choi 2020	RS without PSM	VPI	SLR: 67.5 ± 10.1	101 vs. 126	170 (21 vs. 149)	None	≤2 cm in size	cT1a-bN0M0 pT2aN0M0	46	RFS, OS
			Lob: 66.9 ± 10.8
Kadota 2019	RS without PSM	STAS	SLR: 71	355 vs. 380	137 (27 vs. 110)	11	≤3 cm in size	Stage I	65	RFS, OS
			Lob: 67

Lob, lobectomy; NA, not applicable; NR, not reported; PSM, propensity score matching; RS, retrospective study; Seg, segmentectomy; SLR, sublobar resection; STAS, spread through air spaces; TNM, tumor, node and metastasis; VPI, visceral pleural invasion.

Table 2.

Quality evaluation of included studies by methodological index for nonrandomized studies (MINORS)^a

Study (first- author)	A clearly stated aim	Inclusion of consecutive patients	Prospective collection of data	Endpoints appropriate to the aim of the study	Unbiased assessment of the study endpoint	Follow-up period appropriate to the aim of the study	Loss to follow up less than 5%	Prospective calculation of the study size	A gold standard control group	Contemporary groups	Baseline equivalence of groups	Adequate statistical analyses	Total scores
Schuchert	1	2	1	2	1	2	2	0	2	2	1	2	18
Mathey-Andrews	2	2	1	2	1	2	2	0	2	2	2	2	20
Lula Lukadi	2	2	1	2	0	1	2	0	2	2	2	2	18
Kagimoto	2	2	0	2	1	2	2	0	2	2	2	2	19
Kagimoto	2	2	0	2	1	1	2	0	2	2	2	2	18
Ikeda	2	2	1	2	0	2	2	0	2	2	0	1	16
Huang	2	2	0	2	0	2	2	0	2	2	2	2	18
Yu	2	2	1	2	0	2	2	0	2	2	2	1	18
Shiono	2	2	0	2	0	2	2	0	2	2	1	2	17
Yanagawa	2	2	0	2	0	2	2	0	2	2	1	2	17
Ren	2	2	1	2	0	1	1	0	2	2	0	2	15
Eguchi	2	2	2	2	0	0	0	0	2	2	2	2	16
Choi	2	2	0	2	1	2	0	0	2	2	1	2	16
Kadota	2	2	2	2	0	2	2	0	2	2	1	2	19

^aStudy scoring between 0 and 8 is considered as low quality, while scores of 9 to 16 indicate moderate quality, 17–24 indicate high quality.

SLR vs. lobectomy in NSCLC patients with tumor ≤3 cm and VPI

Nine studies with 1350 patients in the SLR group and 1681 patients in the lobectomy group reported 5-year OS outcomes, while five with 525 patients in the SLR group and 1145 patients in the lobectomy group reported 5-year RFS outcome (Supplementary Digital Content, Table S3, http://links.lww.com/JS9/D935). No significant heterogeneity was observed among studies reporting OS (I² = 0%, P = 0.65) or RFS (I² = 0%, P = 0.51). A Fixed-effects meta-analysis revealed that, compared to lobectomy, SLR was associated with impaired 5-year OS (HR: 1.25; 95% CI: 1.10–1.41; P = 0.0004) and slightly inferior 5-year RFS (HR: 1.25; 95% CI: 0.99–1.58; P = 0.06) in Stage IB (VPI) NSCLC patients with tumor ≤3 cm (Fig. 2). No publication bias was detected (P_OS = 0.582; P_RFS = 0.742).

Figure 2.

Forest plots of 5-year OS (A) and 5-year RFS (B) between the sublobar resection and lobectomy groups in Stage IB (VPI) non-small cell lung cancer with tumor ≤3 cm. OS, overall survival; RFS, recurrence-free survival; VPI, visceral pleural invasion.

SLR vs. lobectomy in NSCLC patients with tumor ≤3 cm and STAS

Seven studies with 252 patients in the SLR group and 809 patients in the lobectomy group reported 5-year OS outcomes, while ten with 522 patients in the SLR group and 1232 patients in the lobectomy group reported 5-year RFS outcomes (Supplementary Digital Content, Table S3, http://links.lww.com/JS9/D935). Heterogeneity among studies reporting OS was not significant (I² = 0%, P = 0.63), therefore, a fixed-effects model was employed for analysis. Conversely, significant heterogeneity was observed among studies reporting RFS (I² = 61%, P = 0.006), necessitating the use of a random-effects model. Meta-analysis revealed that, compared to lobectomy, SLR was associated with impaired 5-year OS (HR: 2.58; 95% CI: 1.92–3.45; P < 0.00001; Egger’s test P = 0.818) and 5-year RFS (HR = 2.42; 95% CI: 1.69–3.46; P < 0.00001; Egger’s test P = 0.581) in stage IA NSCLC patients with STAS (Fig. 3). No publication bias was detected (Egger’s test).

Figure 3.

Forest plots of 5-year OS (A) and 5-year RFS (B) between the sublobar resection and lobectomy groups in Stage IA non-small cell lung cancer patients with STAS. OS, overall survival; RFS, recurrence-free survival; STAS, spread through air spaces.

Subgroup analyses of SLR vs. lobectomy in NSCLC patients with tumor ≤2 cm and VPI

Within the VPI subgroup, five studies with 516 patients in the SLR group and 718 patients in the lobectomy group reported 5-year OS outcomes, while three with 158 patients in the SLR group and 360 patients in the lobectomy group reported 5-year RFS outcomes (Supplementary Digital Content, Table S3, http://links.lww.com/JS9/D935). No significant heterogeneity was observed among these studies reporting OS (I² = 0%, P = 0.40) or RFS (I² = 24%, P = 0.81). A fixed-effects meta-analysis revealed that patients who underwent SLR had comparable 5-year OS (HR: 1.13; 95% CI: 0.92–1.38; P = 0.24) and 5-year RFS (HR: 0.87; 95% CI: 0.56–1.36; P = 0.55) compared to those who underwent lobectomy in stage IB (VPI) NSCLC with tumor ≤2 cm (Supplementary Digital Content, Figure S2, http://links.lww.com/JS9/D924). Publication bias assessment using Egger’s test revealed no significant bias for either OS (P = 0.529) or RFS (P = 0.195).

Subgroup analyses of segmentectomy vs. lobectomy in NSCLC patients with tumor ≤3 cm and VPI

Within the segmentectomy subgroup, four studies with 391 patients in the SLR group and 2614 patients in the lobectomy group reported 5-year OS outcomes, while three with 200 patients in the SLR group and 1512 patients in the lobectomy group reported 5-year RFS outcomes (Supplementary Digital Content, Table S3, http://links.lww.com/JS9/D935). No significant heterogeneity was observed among these studies reporting OS (I² = 0%, P = 0.61) or RFS (I² = 52%, P = 0.13). A fixed-effects meta-analysis revealed that patients who underwent segmentectomy had comparable 5-year OS (HR: 1.16; 95% CI: 0.89–1.52; P = 0.28) and 5-year RFS (HR: 1.07; 95% CI: 0.88–1.30; P = 0.49) compared to those who underwent lobectomy in stage IB (VPI) NSCLC with tumor ≤3 cm (Supplementary Digital Content, Figure S3, http://links.lww.com/JS9/D925). No publication bias was detected for OS (P_egger = 0.948) or RFS (P_egger = 0.792).

Subgroup analyses of segmentectomy vs. lobectomy in NSCLC patients with tumor ≤3 cm and STAS

Within the segmentectomy subgroup, three studies with 164 patients in the SLR group and 439 patients in the lobectomy group reported the 5-year OS and RFS outcomes (Supplementary Digital Content, Table S3, http://links.lww.com/JS9/D935). Heterogeneity was observed among these studies reporting OS (I² = 79%, P = 0.008) and RFS (I² = 78%, P = 0.010). A random-effects meta-analysis indicated that patients who underwent segmentectomy had comparable 5-year OS (HR: 3.88; 95% CI: 0.82–18.31; P = 0.09) and 5-year RFS (HR = 1.64, 95% CI: 0.70–3.80, P = 0.25) compared to those who underwent lobectomy in stage IA NSCLC with STAS (Supplementary Digital Content, Figure S4, http://links.lww.com/JS9/D926). No publication bias was detected for OS (P_egger = 0.557) or RFS (P_egger = 0.461).

Publication bias and sensitivity analyses

The CEFP for OS revealed one unpublished study falling within the nonsignificant region (Fig. 4A), suggesting publication bias in studies comparing SLR to lobectomy for stage IB (VPI) NSCLC patients with tumor ≤3 cm. After adjustment using the trim-and-fill method, the trend of impaired 5-year OS with SLR compared to lobectomy remained unchanged (HR: 1.25; 95% CI: 1.10-1.41; P = 0.0004). The CEFP for RFS exhibited symmetry, indicating the absence of publication bias across the selected studies (Fig. 4B).

Figure 4.

The contour-enhanced funnel plots of sublobar resection vs. lobectomy for Stage IB (VPI) NSCLC patients with tumor ≤3 cm in terms of 5-year OS (A) and 5-year RFS (B). OS, overall survival; RFS, recurrence-free survival; VPI, visceral pleural invasion; NSCLC, non-small cell lung cancer.

The CEFPs for OS and RFS revealed two and three unpublished studies, respectively, falling within the nonsignificant regions (Fig. 5), suggesting publication bias in studies comparing SLR to lobectomy for stage IA NSCLC patients with STAS. After adjustment using the trim-and-fill method, the trends of impaired 5-year OS (HR: 2.39; 95% CI: 1.83–3.14; P < 0.0001) and 5-year RFS (HR: 1.75; 95% CI: 1.21–2.54; P = 0.003) with SLR compared to lobectomy remained unchanged.

Figure 5.

The contour-enhanced funnel plots of sublobar resection vs. lobectomy for Stage IA NSCLC patients with STAS in terms of 5-year OS (A) and 5-year RFS (B). OS, overall survival; RFS, recurrence-free survival; STAS, spread through air spaces; NSCLC, non-small cell lung cancer.

The CEFPs and adjustment results of OS and RFS in subgroup analysis are presented in Supplementary Digital Content (Figures S5, http://links.lww.com/JS9/D927, S6 http://links.lww.com/JS9/D928, S7 http://links.lww.com/JS9/D929).

Sensitivity analysis was conducted to assess the robustness of the pooled HRs and 95%CIs. The results (Supplementary Digital Content, Figures S8, http://links.lww.com/JS9/D930, S9 http://links.lww.com/JS9/D931, S10 http://links.lww.com/JS9/D932) indicated that the pooled HRs and 95% CIs were stable for 5-year OS and 5-year RFS comparing SLR to lobectomy in three cohorts of cT1 NSCLC patients: (1) those with VPI (pT2a) (2) those with STAS (pT1), and (3) those with tumors ≤2 cm and VPI. Similarly, the pooled results for 5-year OS and 5-year RFS comparing segmentectomy to lobectomy in NSCLC patients (≤3 cm) with VPI or STAS also remained stable (Supplementary Digital Content, Figure S11, http://links.lww.com/JS9/D938).

Discussion

Numerous retrospective studies^[36–40] and the JCOG0802 trial^[10] have demonstrated comparable survival outcomes between SLR (especially segmentectomy) and lobectomy for cT1N0M0 NSCLC, particularly for tumors ≤2 cm (cT1a-bN0M0)^[10]. However, the equivalence of these surgical approaches remains uncertain for cT1N0M0 NSCLC with adverse invasion features such as VPI or STAS identified postoperatively. Therefore, further investigation is warranted to clarify the role of SLR, particularly segmentectomy, in early-stage NSCLC patients with these high-risk features.

Our meta-analysis compared SLR, particularly segmentectomy, to lobectomy in NSCLC patients with tumor ≤3 cm exhibiting VPI or STAS. Unlike previous meta-analysis^[8], our meta-analysis considered tumor size, surgical approach, and VPI/STAS status. We separately analyzed outcomes for SLR in patients with pT2a (VPI) NSCLC with tumor ≤3 cm, pT1 (STAS) NSCLC with tumor ≤3 cm, and pT2a (VPI) NSCLC with tumor ≤2 cm. Furthermore, we compared survival outcomes between segmentectomy and lobectomy in those patients.

Our findings suggest that SLR (including WR and segmentectomy), when analyzed as a whole, is inferior to lobectomy in NSCLC (tumors ≤3 cm) patients exhibiting VPI or STAS postoperatively. However, the Higgins I² statistic and CEFP indicated potential heterogeneity among the included studies, possibly due to variations in tumor size. Notably, only 5 of the 15 included studies analyzed tumors ≤2 cm in size^{[13,18,26,28,35]}. This is significant because long-term surgical outcomes differ between lung cancer with tumors ≤2 cm and those between 2 and 3 cm^[41,42]. The randomized controlled trial CALGB 140503 demonstrated non-inferiority of SLR compared to lobectomy in T1a-bN0 NSCLC^[11]. Therefore, we further investigated the role of SLR in NSCLC patients with tumor ≤2 cm and VPI. Our results suggest that SLR provides comparable outcomes to lobectomy in this population. This is supported by recent secondary analyses of CALGB 140503, which also indicated comparable OS and RFS between SLR and lobectomy in patients with pT2a (VPI) NSCLC with tumor ≤2 cm^[35]. However, insufficient data exists to evaluate the role of SLR in NSCLC patients with tumor ≤2 cm and STAS, highlighting the need for further research in this specific population.

Upon analysis of heterogeneity sources among included studies, surgical approach also emerged as a potentially significant factor. Our study identified five studies that distinguished between segmentectomy and WR within the broader SLR category, reporting survival data comparing segmentectomy to lobectomy^{[16,19,26-28]}. WR, a more limited lung parenchyma resection, may not achieve adequate tumor margins, potentially elevating local recurrence risk^[43]. Furthermore, it lacks comprehensive lymph node assessment, potentially hindering accurate tumor staging and leading to missed opportunities for appropriate adjuvant therapy^[44], thus potentially compromising the OS benefits of SLR.

Consequently, we conducted separate analyses comparing segmentectomy and lobectomy for NSCLC (≤3 cm) with VPI and NSCLC (≤3 cm) with STAS. Results demonstrated comparable OS and RFS between segmentectomy and lobectomy in both NSCLC (≤3 cm) with VPI group and NSCLC (≤3 cm) with STAS group, although considerable heterogeneity was observed in the latter group. This heterogeneity may be attributable to publication bias given the limited number of included studies. Therefore, we employed a CEFP and the trim-and-fill method to assess for publication bias. The CEFP indicated asymmetry, and after applying the trim-and-fill method to simulate additional studies for a secondary meta-analysis, results continued to demonstrate that segmentectomy did not compromise OS or RFS compared to lobectomy in stage IA NSCLC with STAS (Supplementary Digital Content, Figure S7, http://links.lww.com/JS9/D929), further supporting the robustness of our findings.

Kagimoto et al^[19]. and Yanagawa et al^[31]. reported impaired OS in cT1N0 lung cancer patients with VPI or STAS, particularly among those who underwent SLR. However, our meta-analysis distinguishes between segmentectomy and WR within the SLR category, demonstrating for the first time that segmentectomy offers comparable survival outcomes to lobectomy in such population. This suggests that the inferior OS observed in some SLR studies might be attributed to the inclusion of WR, potentially due to its inadequacy in addressing hilar and intrapulmonary lymph nodes. Consequently, we posit that segmentectomy is a safe and effective treatment option for this patient population, and patients undergoing segmentectomy who are found to have VPI or STAS postoperatively may not require a secondary completion lobectomy. However, the necessity of adjuvant therapy following surgery in stage I lung cancer patients (tumor≤3 cm) with VPI or STAS remains uncertain^[31,45] and requires further investigation, including prospective randomized controlled trials.

The heterogeneity in adjuvant therapy administration across included studies represents another significant limitation. Among studies reporting adjuvant therapy data, 12.1% (98/813) of patients in the SLR group and 24.0% (904/3771) in the lobectomy group received adjuvant therapy. This disparity, along with three studies not specifying adjuvant therapy distribution between groups, may influence the survival outcomes and represents a potential source of bias. To address these limitations, future studies should focus on two key directions: First, conducting analyses that compare surgical approaches (SLR vs. lobectomy) in patients who did not receive adjuvant therapy, thereby isolating the effect of the surgical approach itself. Second, performing comprehensive evaluations of different treatment combinations, including SLR with adjuvant therapy versus lobectomy with adjuvant therapy, and comparing these to surgical treatment alone. Such studies would provide more precise guidance for treatment selection in early-stage NSCLC patients with VPI or STAS.

Another potential source of heterogeneity may be attributed to the diverse pathological types of NSCLC and adenocarcinoma subtypes. Studies^[46–49] have shown that the controversy regarding comparison of segmentectomy and WR in patients with different pathological types of NSCLC especially in lung squamous cell carcinoma and adenocarcinoma. In our study, squamous cell carcinoma constituted 20.6% of the SLR group, with some patients undergoing WR. Regarding adenocarcinoma, invasive adenocarcinoma carries a poorer prognosis compared to non-invasive adenocarcinoma^[50–52]. Notably, different histological subtypes of invasive adenocarcinoma are associated with varying prognoses^[53]. Studies^[54–56] indicate that for patients with papillary, micropapillary, and solid subtype stage IA adenocarcinoma, anatomical resections (lobectomy and segmentectomy) yield superior results compared to nonanatomical resection (WR). Our study included patients with invasive adenocarcinoma who underwent WR, which may partially explain the inferior outcomes observed in the SLR group compared to the lobectomy group in NSCLC patients (tumors ≤3 cm) with postoperative VPI or STAS.

Our meta-analysis has several limitations: First, 14 of 15 included studies were retrospective. Although seven studies employed propensity score matching, some confounding factors remained unadjusted. To mitigate this, we conducted sensitivity analyses, and heterogeneity tests further corroborated the robustness of our findings. Second, publication bias, which is inherent to retrospective studies, may have influenced the results. Although Egger’s test did not reveal significant publication bias among the included studies, the CEFPs indicated a certain degree of heterogeneity. Third, some HRs were derived from survival curves, which could have introduced bias in the outcomes. Fourth, differences in the cohorts receiving adjuvant therapy between the SLR and lobectomy groups in the included studies could also be a potential source of heterogeneity. Future studies should focus on comparing surgical approaches without adjuvant therapy interference, as well as evaluating the effectiveness of different combinations of surgical and adjuvant treatments for early-stage NSCLC patients with VPI or STAS. Fifth, including two pathological states (VPI and STAS) in the same study inevitably increases heterogeneity. Although there is currently no theoretical evidence supporting a pathological connection between VPI and STAS, studies^[15,57] have statistically shown that patients with VPI have higher rates of STAS. To address this issue, in our study, we have consistently analyzed VPI and STAS separately. Our research focuses on two distinct populations, analyzing them independently to understand the role of SLR in early-stage lung cancer patients with VPI or STAS. Lastly, none of the VPI-related studies reported on the grade of pleural invasion, which could have contributed to outcome bias.

Conclusion

For early-stage (≤3 cm) NSCLC with VPI or STAS, segmentectomy achieves comparable outcomes to lobectomy. Conversely, SLR demonstrates inferior 5-year OS and 5-year RFS. For stage IB NSCLC patients with tumor ≤2 cm and VPI, SLR achieves comparable outcomes to lobectomy. Our findings suggest that patients undergoing segmentectomy who are found to have VPI or STAS postoperatively do not require a secondary completion lobectomy.

Ethical approval

Not applicable.

Consent

Not applicable.

Sources of funding

This study was supported by the Key Research Project of Sichuan Province (No.2023YFS0199), Chengdu Science and Technology Project (No.2022-YF05-01833-SN), and Sichuan Cadre Health Care Research Project (GBKT23020).

Author’s contribution

Z.-Y.D.: conceptualization, methodology, software, formal analysis, writing – original draft preparation; C.S.: data curation, writing- original draft preparation; X.W.: formal analysis, visualization, investigation; F.-Q.W.: supervision, software, validation; Y.W.: writing- reviewing and editing, project administration; Z.D., C.S. and X.W. contributed equally to the manuscript.

Conflicts of interest disclosure

All the authors declare to have no conflicts of interest relevant to this study.

Research registration unique identifying number (UIN)

This meta-analysis is register at PROSPERO (registration ID: CRD42024568832).

Guarantor

Yun Wang.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Data availability statement

No other datasets were generated during and/or analyzed during the current study. All the information is available with the manuscript.

Assistance with the study

We are highly grateful to those who provided guidance during the writing of the manuscript.

Presentation

Not applicable.