Recently, multi-institutional randomized clinical trials (JCOG0802/WJOG4607L and CALGB140503) demonstrated the clinical value of segmentectomy in NSCLC. However, this benefit is limited to patients with small-sized peripheral NSCLC [3, 13]. Given the similar anatomical classification of the left upper division and lingula to that of the right upper and middle lobes, those who underwent segmentectomy (i.e. upper division segmentectomy and lingulectomy) in the left upper lobe may benefit in terms of the OS, even for tumors larger than 20 mm in diameter.

Several studies have reported similar oncological outcomes between left upper segmentectomy (upper division segmentectomy and lingulectomy) and left upper lobectomy. Among these studies, one of the largest was conducted by Zhou et al. who performed PSM, resulting in 273 pairs of patients undergoing thoracoscopic left upper division segmentectomy or left upper lobectomy for stage I NSCLC. The authors reported no compelling differences in clinical or oncological outcomes between the groups [9]. Furthermore, our institution also reported that the clinical and oncological outcomes following left upper division segmentectomy for clinical N0 NSCLC were not significantly inferior to those following lobectomy, even if the tumor was located close to the intersegmental plane as measured by three-dimensional computed tomography [10]. Although some studies have demonstrated similar oncological outcomes for left upper segmentectomy and lobectomy (5–9), our study cohort also included patients who underwent other single segmentectomy procedures, such as S1+2 segmentectomy. Since we assumed that all these segmentectomies for the left upper lobe were performed as a ‘split-lobe’ procedure, further studies are required to clarify whether or not the oncological clearance of this split-lobe concept would be applicable for all forms of segmentectomy in the left upper lobe. To our knowledge, this multi-institutional study has the largest cohort showing the non-inferiority of left upper segmentectomy to left upper lobectomy.

Corresponding to previous studies, this study revealed that segmentectomy exhibited an OS comparable to that of left upper lobectomy, irrespective of tumor size. However, the differences between the two groups in terms of pathological characteristics were concerning. Despite adjusting for the characteristics of segmentectomy and lobectomy, including tumor size, the rate of pN1/2 was significantly higher in the lobectomy group than in the segmentectomy group. One reason is that lobe-specific lymph node dissection may have been omitted during segmentectomy. For instance, although complete interlobar lymph node dissection is challenging, owing to variations in the divergence style of the lingular artery and vein during upper division segmentectomy, NSCLC located in the upper division tends to metastasize to the interlobar lymph nodes [14]. As presented in Table 2, mediastinal lymph node dissection during segmentectomy was not performed in some cases, which may have led to underestimation of the pathological stage in the segmentectomy group. Nevertheless, segmentectomy revealed an OS comparable to that of lobectomy. Indeed, a recent study also suggested that the survival was similar between lobectomy and segmentectomy in patients with clinical N0 and unsuspected pathological N1/N2 nodal metastases [15].

One of the most notable concerns after segmentectomy is local recurrence in the residual lobe [16]. Therefore, sufficient tumor margins should be ensured. Theoretically, as the subpleural lymphatic pathway can be blocked by the intersegmental septum, accurate intersegmental dissection during segmentectomy may allow resection of the tumor in the affected segment without the infiltration of cancer cells into the neighboring segment. As margins equal to the tumor diameter or >20 mm would be acceptable [17], and pleural lymphatic drainage might follow an intersegmental pathway [18], dissection into the neighboring segment while sacrificing the intersegmental vein is necessary if the tumor is close to the intersegmental plane. In addition, if the tumor located in the right upper or middle lobe is close to the minor fissure, most thoracic surgeons prefer to spare the middle or upper lobe and dissect into the neighboring lobe to secure the margin instead of performing bi-lobectomy. This may be applicable for large tumors (>20 mm) located in the left upper lobe during split-lobe segmentectomy if the tumor margin is sufficiently secured.

Early-stage NSCLC with a GGO component has been reported to be a uniform group of tumors that exhibit low-grade malignancy and have an excellent prognosis [19]. In contrast, solid predominant or pure solid tumors have more malignant potential, such as lymph node metastasis [20]. Our study revealed that left upper segmentectomy for radiological GGO or solid-dominant stage I NSCLC had long-term effects similar to those of left upper lobectomy, even when the entire tumor size exceeded 20 mm. Interestingly, segmentectomy had a significant advantage with regard to the RFS in patients with solid tumors ≤20 mm in size. One possible reason for this is that pathological upstaging might be infrequent in segmentectomy, probably because segmentectomy may be likely to be applied to cases with peripherally located tumors, which are less strongly associated with lymph node metastasis than inner-located tumor [21]. In contrast, segmentectomy showed a significantly worse RFS than lobectomy for solid tumors >20 mm, as segmentectomy might carry a high risk of recurrence due to insufficient hilar lymph node dissection. Nonetheless, segmentectomy exhibited an OS equivalent to that of lobectomy. One possible reason for this is that segmentectomy, which preserves more of the lung parenchyma than lobectomy, might have allowed more extensive treatment for relapse of primary lung cancer and second primary lung cancer than lobectomy, despite its higher recurrence rate. Thus, segmentectomy offers a prognosis similar to that of lobectomy, even for pure solid tumors.

Several limitations associated with the present study warrant mention. First, despite the propensity matching analysis, selection bias for surgical procedures and other biases may still exist. Indeed, thoracic surgeons performed segmentectomy as curative-intent resection for patients with predominantly GGO, low metabolic activity, slow-growth NSCLC, and compromised limited resection for patients unable to tolerate lobectomy. Second, this dataset, which was based on surgical cases in 2010, lacked information on minimally invasive surgical approaches, such as uniportal-, multiportal-, and robot-assisted thoracoscopic surgery. Third, our dataset lacked information regarding tumor location (i.e. tumor centrality and the specific segment location) and pathological margins. Finally, as the current study was retrospective, further multi-institutional prospective randomized trials are warranted in the future.

In conclusion, the current analyses suggest that the OS following segmentectomy for c-stage I NSCLC in the left upper lobe is not significantly inferior to that following lobectomy, irrespective of the tumor size and radiological tumor findings.