Prognostic impact of noninvasive areas in resected pathological stage IA lung adenocarcinoma

Abstract Main Problems In non‐small‐cell lung cancer, ground‐glass opacity on computed tomography imaging reflects pathological noninvasiveness and is a favorable prognostic factor. However, the significance of pathological noninvasive areas (NIAs) has not been fully revealed. In this study, we aimed to elucidate the prognostic impact of NIAs on lung adenocarcinoma. Methods We analyzed 402 patients with pathological stage (p‐Stage) IA lung adenocarcinoma who underwent surgery in 2013–2016 at two institutions and examined the association of the presence of NIAs with clinicopathological factors and prognosis. Furthermore, after using propensity‐score matching to adjust for clinicopathological factors, such as age, sex, smoking history, pathological invasive area size, pathological T factor (p‐T), p‐Stage, and histological subtype (lepidic predominant adenocarcinoma [LPA] or non‐LPA), the prognostic impact of NIAs was evaluated. Results Patients were divided into NIA‐present (N = 231) and NIA‐absent (N = 171) groups. Multivariable analysis showed that NIA‐present was strongly associated with earlier p‐T, earlier p‐Stage, LPA, and epidermal growth factor receptor mutation. Kaplan–Meier survival analysis showed that the NIA‐present group displayed a better prognosis than the NIA‐absent group in disease‐free survival (DFS) and overall survival (OS) (5‐year DFS 94.6% vs. 87.2%, 5‐year OS 97.2% vs. 91.1%). However, after adjusting for clinicopathological factors by propensity score matching, no significant differences in prognosis were identified between the NIA‐present and NIA‐absent groups (5‐year DFS 92.4% vs 89.6%, 5‐year OS 95.6% vs 94.3%). Conclusions Our current study suggests that the prognostic impact of the presence of NIAs on lung adenocarcinoma is due to differences in clinicopathological factors.


INTRODUCTION
The tumor-node-metastasis (TNM) classification of lung cancer was revised to the 8th edition in 2017. [1][2][3] One of the major changes in the 8th edition is that the T factor is assessed by the solid component size/invasive area size instead of the whole tumor size. [1][2][3] These changes are based on the findings of several studies showing that, especially on preoperative computed tomography (CT) findings, the solid component size reflects poor prognosis and pathological malignancy more accurately than the whole tumor size. [4][5][6][7] In non-small-cell lung cancer (NSCLC), the presence of ground-glass opacity (GGO) on CT findings has been reported to be a favorable prognostic factor for postoperative prognosis. [8][9][10][11][12][13][14] However, few reports have analyzed the association between the presence of pathological noninvasive areas (NIAs) and the postoperative prognosis of NSCLC. Because the presence of GGO is a favorable prognostic factor, the presence of NIAs might also be a positive factor. However, it is highly questionable whether NIAs, which do not appear to have metastatic or invasive potential, can influence the prognosis of NSCLC. In this study, to elucidate the prognostic impact of NIAs, we analyzed how their presence contributed to the prognosis of surgically resected lung adenocarcinoma using propensity-score matching.

Study patients
This retrospective research was reviewed and approved by our institutional review boards (Kyushu Cancer Center, IRB No. 2019-56; Kyushu University Hospital, IRB No. 2019-232). A total of 462 patients with pathological stage (p-Stage) IA NSCLC who underwent surgical resection between January 2013 and December 2016 at the Department of Thoracic Oncology, National Hospital Organization Kyushu Cancer Center and the Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University Hospital were enrolled. Of these, 409 patients were diagnosed as p-Stage IA lung adenocarcinoma. Seven patients who received preoperative treatments were excluded and 402 patients with p-Stage IA lung adenocarcinoma were analyzed in this research. As the routine postoperative follow-up, physical examination, blood tests, and chest radiographs were performed at 3-month intervals for the first 3 years followed by 6-month intervals, and CT scans were performed at 6-month intervals for the first 3 years followed by 1-year intervals. Patients who met the following criteria received uracil-tegafur as adjuvant chemotherapy: (i) pathological whole tumor size ≥2 cm, (ii) age < 76 years, (iii) Eastern Cooperative Oncology Group performance status of 0 or 1, and (iv) provided written informed consent.

Pathological diagnosis
The pathological diagnoses were performed at the Cancer Pathology Laboratory, National Hospital Organization Kyushu Cancer Center and Department of Anatomic Pathology, Graduate School of Medical Sciences, Kyushu University Hospital, based on the 8th edition of the TNM classification. Pathological whole tumor size and pathological invasive size were evaluated by at least two pathologists. Patients were classified into NIA-present and NIA-absent groups according to the presence of NIAs, and the relationship between clinicopathological factors and prognosis was analyzed. Representative histological findings of the NIA-present and NIA-absent groups are shown in Figure 1.

Statistical analysis
The Pearson's chi-square test and Student's t test were used to analyze the difference in clinicopathological factors between NIA-present and NIA-absent groups. The DFS was defined as the period from surgery to the date of the last follow-up, recurrence, or death, and OS was defined as the period from surgery to the date of the last follow-up or death. The Kaplan-Meier method with the log-rank test was used for plotting survival curves. Cox proportional hazards regression analysis was used to examine risk factors. In the multivariate analysis, only factors with p < 0.05 were extracted by the variable reduction method. JMP Pro 16.0 software (SAS Institute) was used for all statistical analyses.

Propensity-score matched analysis
A propensity-score matched analysis was performed to reduce the bias between NIA-present and NIA-absent groups. The propensity scores comprised the following variables: age, sex, smoking history, surgical procedure, pathological invasive size, p-T factor, p-Stage, histological predominant subtype, ly, and v. A propensity score difference of 0.05 was adopted as the maximum caliper width for matching NIA-present and NIA-absent groups. Finally, 100 matched patients from each group were enrolled in the analysis.

Clinicopathological factors
A total of 402 patients with p-Stage IA lung adenocarcinoma who underwent surgical resection were enrolled in this research. The clinicopathological factors of patients are shown in Supporting Information Table S1. The median age at surgery was 69 years (range 34-88 years). There were 173 men (43.0%) and 229 women (57.0%). In addition, 185 patients (46.0%) had a smoking history. The median radiological whole tumor size and solid component size were 19 mm (range 6-67 mm) and 12 mm (range 0-45 mm), respectively. The median C/T ratio was 0.67 (range 0.00-1.00), and GGO was present in 280 (69.7%) patients. In terms of surgical procedures, sublobar resection was performed on 91 patients (22.6%).

Clinicopathological factors according to the presence of NIAs
The numbers of patients in the NIA-present and NIA-absent groups were 231 (57.5%) and 171 (42.5%), respectively. The associations between the presence of NIAs and clinicopathological factors are demonstrated in Table 1. The Pearson's chi-square test and Student's t test showed that NIA-present was significantly associated with small radiological solid component size (p = 0.0042), smaller C/T ratio (p < 0.0001), GGO presence (p < 0.0001), larger pathological whole tumor size (p = 0.0055), smaller pathological invasive area size (p < 0.0001), earlier p-T factor (p < 0.0001), earlier p-Stage (p < 0.0001), LPA (p < 0.0001), v negativity (p = 0.0009), and EGFR mutation positivity (p = 0.0429).

Prognosis analysis according to the presence of NIAs
Prognostic analyses according to the presence of NIAs were performed using the Kaplan-Meier method. The DFS and OS in the NIA-present group were significantly better than those in the NIA-absent group (5-year DFS 94.6% vs. 87.2%, p = 0.0039, Figure 2a; 5-year OS 97.2% vs. 91.1%, p = 0.0106, Figure 2b). However, the multivariable analysis showed that NIA-present was not an independent prognostic factor for DFS and OS (Table 2).

Prognosis analysis according to the presence of NIAs after propensity-score matched analysis
We then performed propensity-score matched analysis to reduce the bias of clinicopathological factors between NIA-present and NIA-absent groups, and examined the prognostic impact of the presence of NIAs in more detail.
After adjusting for the bias of clinicopathological factors by propensity-score matched analysis, the Fisher's exact test and Student's t test demonstrated that most clinicopathological factors between the NIA-present and NIA-absent groups showed no significant difference (Table 3). Furthermore, Kaplan-Meier survival curves revealed that the prognostic differences in DFS and OS between the NIA-present and NIA-absent groups disappeared (5-year DFS 92.4% vs 89.6%, p = 0.7546, Figure 3a; 5-year OS 95.6% vs 94.3%, p = 0.9212, Figure 3b).

DISCUSSION
In the current research, we described the prognostic impact of NIAs in surgically resected p-Stage IA lung adenocarcinoma. The prognosis of the NIA-present group was significantly better than that of the NIA-absent group; however, after adjusting for the bias of clinicopathological factors by propensity-score matched analysis, the prognostic impact of NIAs disappeared (Figure 4).
In terms of the association between the presence of NIAs and clinicopathological factors, NIA-present was significantly associated with the presence of GGO, early F I G U R E 1 Representative histological images of NIA-present (a) and NIA-absent (b) tumors. NIA, noninvasive area stage, LPA, v negativity, and EGFR mutation positivity in our research. In particular, multivariable analysis showed that earlier p-T factor, LPA, and EGFR mutation positivity were strongly associated with NIA-present. These results were similar to the association between the presence of GGO and clinicopathological factors in previous reports that demonstrated the prognostic significance of GGO in NSCLC. [8][9][10][11][12][13][14][15] Therefore, our research appeared to be consistent with previous reports that the presence of radiological GGO reflected the presence of pathological NIAs.
Several reports have shown that the presence of radiological GGO is a favorable prognostic factor for NSCLC. [8][9][10][11][12][13][14] In those reports, the GGO-present group tended to have tumors with pathologically low-grade malignancy, lepidic   adenocarcinoma, and EGFR mutation, [8][9][10][11][12][13][14] and the GGO-absent group sometimes included histological types with poorer prognoses than adenocarcinoma, such as squamous cell carcinoma. 8,9 As a result, the presence of GGO has been reported as a favorable factor for NSCLC. Although many reports have analyzed the results without adjusting for background factors, some reports that adjusted for clinicopathological factors, such as histological features, using propensity-score matching have suggested that the presence of GGO may not directly affect prognosis. 16,17 In addition, a previous study demonstrated that the expression of program death ligand-1 protein on cancer cells was higher in GGO-absent tumors than in GGO-present tumors, and the tumor microenvironment may differ according to the presence of GGO. 18 On the basis of these reports, we think that the presence of GGO may be useful in predicting histological features, but GGO itself cannot directly influence the prognosis of NSCLC. Moreover, the results of our present study showed that the presence of pathological NIAs had no prognostic impact after adjusting for clinicopathological factors by propensity-score matched analysis. These results suggest that radiological GGO and pathological NIAs themselves might not directly impact the prognosis of lung adenocarcinoma.
As demonstrated in this study, NIA-present lung adenocarcinomas often contain lepidic growth components. Regarding the prognostic impact of the presence of lepidic components in lung adenocarcinoma, some reports showed that their presence is a favorable prognostic factor, 19,20 whereas other reports demonstrated that lepidic components are not an independent prognostic factor in multivariable analysis and do not affect the prognosis of lung adenocarcinoma. 21,22 Our results showing that NIAs do not influence the prognosis of lung adenocarcinoma may support reports that there are minimal effects of lepidic components on the prognosis of lung adenocarcinoma.
Some previous reports suggested that the TNM classification at the clinical stage should be reconsidered according to the presence of GGO. 9,10,23 However, considering that the prognostic impact of GGO and NIAs is strongly influenced by other histological features, and although the presence of radiological GGO is possibly useful to assess clinical stages where the histological features may not yet be fully known, the presence of pathological NIAs does not appear to be useful to assess p-Stage where the histological features may already be known. The results of our study might support the usefulness of the 8th edition of the TNM classification, which excludes NIAs in the assessment of T and suggests that clinicians should be careful about adding GGO and NIAs to the TNM classification.
This study had several limitations. Because this research was retrospective with a limited number of patients, a largescale prospective study in the future is desired. In addition, the data on pathological invasive area size were assessed only from 2013, and the observation period of this study was relatively short. Moreover, the insufficient number of patients made it difficult to examine more detailed histological subtypes of lung adenocarcinoma in this study.

CONCLUSION
Although the presence of NIAs was associated with a better prognosis of lung adenocarcinoma, the prognostic impact of NIAs disappeared after adjusting for the bias of clinicopathological factors by propensity-score matched analysis. Our current study suggests that the prognostic impact of NIAs on lung adenocarcinoma is due to differences in clinicopathological factors.