Prognostic Implications of the Predominant Histologic Subtype in Stage I Lung Adenocarcinoma: Lack of Association with Overall Survival Based on Multi-Aspect Survival Analyses


 We aimed to investigate the prognostic role of the predominant histologic subtype classification of lung adenocarcinoma in the era of the eighth-edition staging system. A total of 587 patients with stage I adenocarcinomas (lepidic, n=46; acinar/papillary, n=484; solid/micropapillary, n=57) were assessed. The prognostic value of this subtype classification for overall survival (OS) was investigated with three multivariable analyses adjusted for the eighth-edition clinical T category and radiologic nodule type: the adjusted log-rank test, multivariable Cox hazard regression, and multivariable mixture cure model to evaluate long-term and short-term prognostic associations separately. In the adjusted log-rank test for pairwise comparisons, no evidence was found for differences in OS among the subtype classifications (p>.05). The histologic subtype was not an independent prognostic factor in either the Cox regression analysis (hazard ratio [HR] of acinar/papillary subtypes: 3.03; 95% confidence intertval [CI]: 0.4-22.59; p=0.28; HR of solid/micropapillary subtypes: 2.78; 95% CI: 0.34-22.37; p=.34) or the mixture cure model (cure probability model; odds ratio: 1.06 [95% CI: 0.38-2.95; p=.91]; failure time distribution model; HR: 0.94 [95% CI: 0.36-2.44; p=.89]). The prognostic implication of the predominant histologic subtype classification is not clear for stage I lung adenocarcinoma in the eighth-edition staging system.


Introduction
Lung cancer is the leading cause of cancer mortality and second most common newly diagnosed type of cancer worldwide and in the United States. 1,2 The most common histologic type of lung cancer is adenocarcinoma, accounting for up to 60% of cases, and this single histologic type has a wide spectrum of clinical, radiologic, and pathologic characteristics. 3,4 According to the 2011 International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society (IASLC/ATS/ERS) classi cation, invasive lung adenocarcinoma is histologically categorized into lepidic, acinar, papillary, solid, and micropapillary subtypes based on the predominant growth pattern. 3 Many studies have reported the prognostic value of this classi cation system. [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] Speci cally, adenocarcinoma of the lepidic-predominant subtype showed the most favorable outcomes, followed by the acinar/papillary and solid/micropapillary subtypes. 4,5,7,[9][10][11][13][14][15][16][17]19,20,22,23 The value of this prognostic strati cation was proven in early-stage adenocarcinoma and in advanced disease. [4][5][6][7][8][9][10][11][13][14][15][16][17][18][19][20]22,23 The staging system by the American Joint Commission on Cancer (AJCC) is the cornerstone for predicting the prognosis of patients with lung cancer. 24 The eighth-edition staging system adopted the solid portion size on CT and the invasive component size on microscopy to determine the clinical and pathologic T category, respectively. 25,26 It is now well established that the solid portion size or invasive component size is a better prognostic predictor than the total tumor size. 25,26 In this context, the prognostic value of the predominant adenocarcinoma subtype has not yet been investigated in conjunction with the eighth-edition staging system. Therefore, the independent prognostic implication of histologic subtypes for early-stage adenocarcinoma in the era of the eighth-edition staging system remains unclear.
In this study, we aimed to validate the prognostic value of the predominant histologic subtype in patients with resected stage I adenocarcinoma using multi-aspect survival analyses, including the adjusted logrank test, Cox proportional hazard regression, and mixture cure model to distinguish short-term and longterm prognoses.
The adjusted log-rank test did not show longer survival for the lepidic subtype than for the acinar/papillary subtype (p=0.45) and solid/micropapillary subtype (p=0.63) (Supplementary

Discussion
To improve prognostic validity for non-small cell lung cancers, the eighth-edition staging system by the AJCC subclassi es the clinical and pathologic T categories in more detail based on the tumor extent; these categories are de ned according to measurements of the solid portion size on CT and pathologically invasive component, respectively. [24][25][26] In this context, we investigated the prognostic value of the predominant histologic subtype in conjunction with this new staging system in stage I adenocarcinomas. Although the predominant subtype classi cation is a prognostic factor for lung adenocarcinoma, this subtype was not found to be an independent predictor for OS in multi-aspect survival analyses, including the adjusted log-rank test, multivariable Cox hazard regression analysis, and the mixture cure model. The results were consistent in adenocarcinomas that presented as solid nodules on chest CT.
Murakami et al. 17 proposed that the histologic subtype was correlated with disease-speci c survival, Recently, Hattori et al. 20 applied the eighth-edition staging system in clinical stage I adenocarcinomas and demonstrated differences in OS according to the histologic subtype. Nevertheless, their study had limitations that only a univariable analysis was conducted and the micropapillary-predominant subtype was absent in the study population. 20 Other publications reporting the prognostic value of histologic subtypes differed from our study in that they included adenocarcinomas of stage II or higher. 4,5,7,[9][10][11][13][14][15][16]22,23 In this study, we performed multi-aspect survival analyses. In principle, Cox proportional hazard regression analysis is based on the assumption that the event will eventually occur and the survival curve will decrease to zero. 28 It is the conventional method for survival analysis in the lung cancer staging system. 26 In contrast, the mixture cure model assumes that the plateau of the survival curve in long-term follow-up re ects the cure state of patients. 29,30 Given the presence of long-term censored patients among those with stage I adenocarcinoma, the mixture cure model is applicable, [29][30][31] and investigators can separately analyze the predictors for long-term and short-term survival. 29,30 The consistent statistical results obtained from the multi-aspect survival analyses con rmed the robustness of our observations.
Long-term follow-up (at least 5 years) for the study population is another strength of our study. The median follow-up duration was 81.3 months, which enabled the investigation of a number of events in patients with stage I adenocarcinomas. Due to the homogeneous and su cient follow-up duration, milestone survival rates at 5 years could be directly described and compared according to the predominant histologic subtypes.
Some limitations of our study should be mentioned. First, we used the clinical T category as an input for the multivariable survival analyses, not the pathologic T category. However, a recent study proposed that the clinical T category may be a better prognostic indicator than the pathological T category. 32 Second, variants of invasive adenocarcinoma were not included (e.g., invasive mucinous, colloid, fetal, and enteric adenocarcinoma). 3 In addition, the latest reported histologic subtypes were not assessed. For example, although it is still not included in a formal classi cation, adenocarcinoma of the cribriform subtype, known as the high-grade acinar subtype, was reported to have poor outcomes similar to those of the solid/micropapillary-predominant subtype. [33][34][35][36] The absence of this concept, as part of the latest classi cation of histologic subtypes, may have led to an inaccurate histologic classi cation, as it can be arbitrarily classi ed as the acinar or solid subtype. 5 Further studies in a cohort with the latest pathological classi cation are warranted. Third, the novel IASLC grading system, which is based on a combination of the most predominant pattern and any high-grade histological pattern (20% or more), was not analyzed. 5 However, Sica's grading system based on two predominant patterns had a comparable discrimination performance to the novel IASLC system according to a validation study. 37 In conclusion, although the predominant histologic subtype classi cation is a prognostic factor for lung adenocarcinoma, no association of this classi cation with OS was observed in stage I adenocarcinoma according to the eighth-edition staging system. Multi-center studies with a large cohort and up-to-date pathologic information are warranted to further validate our results.

Methods
The Institutional Review Board of Seoul National University Hospital approved this study and waived the requirement for patients' informed consent. All experiments and methods were performed in accordance with relevant guidelines and regulations.

Study Population
This study was conducted at a single tertiary referral center in South Korea. Between January 2011 and December 2015, patients who underwent surgical resection for pathologic stage I primary lung adenocarcinoma according to the seventh-edition staging system were identi ed retrospectively through a dedicated search of electronic medical records. 38 The study population was then determined with the following exclusion criteria: 1) patients who underwent sublobar resection; 2) synchronous or metachronous lung cancer; 3) adenocarcinoma in situ and minimally invasive adenocarcinoma; 4) invasive mucinous adenocarcinoma; 5) pure ground-glass nodule on chest CT; 6) no available survival data; 7) clinical variables not obtainable; 8) adenocarcinoma histologic subtypes not described in pathology reports; and 9) a non-measurable primary tumor on CT. These exclusion criteria were determined to eliminate potential prognostic confounders. Consequently, 587 patients with pathologic stage I lung adenocarcinoma who underwent at least lobectomy were included in this study ( Supplementary Fig. 3).

Data Collection
The following data were obtained from patients' electronic medical records: demographic information (age, sex, history of malignancy other than lung cancer, and smoking status [never smoker, current, or exsmoker]), surgical record (operation date and operation modality), radiologic information on preoperative CT examinations (nodule type [part-solid nodule or solid nodule] and lobar location), and pathologic report (pathologic diagnosis, stage, and predominant histologic subtypes).
The pathologic diagnosis and stage were determined during routine clinical practice, and the specimens were not reviewed speci cally for this study. Therefore, the pathological stage was based on the seventh edition of the AJCC staging system for lung cancer, 38 as the participants underwent surgery before the release of the eighth edition of the staging system. 25,26 Thus, we measured the solid component size of each tumor on CT scans to determine the clinical T category according to the eighth edition of the staging system, 25,26 which was used as a covariate in the multivariable survival analyses. Indeed, a recent study reported that the solid portion size on CT demonstrated a better prognostic correlation than the invasive component size. 32 The longest diameter of the solid portion was measured on the lung window setting (window width, 1500 HU; level, -700 HU). All measurements were performed in the axial plane using an electronic caliper by one of the two trained radiology technicians (either M.L. with 10 years of research experience in chest CT or J.Y.J. with 3 years of research experience in chest CT) under the supervision of a board-certi ed thoracic radiologist (H.K. with 10 years of CT experience). Reviewers were blinded to the pathologic diagnosis, stage, and histologic subtypes, but the lobar location was provided.
In this study, the primary endpoint was overall survival (OS), which was measured from the date of surgery for the corresponding lung cancer to the date of death from any cause (i.e., all-cause mortality).
The censoring time was at March 18, 2021, and the survival status and date of death were obtained from the database of the Ministry of the Interior and Safety. 39 Pathologic Diagnosis In this study, the 2011 IASLC/ATS/ERS classi cation of lung adenocarcinoma was applied. 3 That is, nonmucinous invasive adenocarcinomas were categorized by their predominant architecture patterns (lepidic, acinar, papillary, micropapillary, or solid-predominant subtype). 3,4 The nal pathologic diagnosis was adjudicated by the board-certi ed attending pathologists of our hospital as part of their routine clinical practice. Then, according to Sica's three-tier grading system, 40 adenocarcinomas were categorized into three groups: 1) lepidic-predominant, 2) acinar/papillary-predominant, and 3) solid/micropapillarypredominant subtypes. [3][4][5]41 In 26% ( To investigate the prognostic value of the predominant histologic subtype for OS, we performed threelevel survival analyses: 1) pairwise comparisons among the three histological groups (lepidicpredominant, acinar/papillary-predominant, and solid/micropapillary-predominant subtypes) using unadjusted and adjusted log-rank tests; for the adjusted log-rank test, variables of age, sex, cancer history, smoking status, nodule type, and clinical T category (cT1mi/cT1a, cT1b, cT1, and cT2) were taken into account using inverse probability weighting; 43 2) multivariable Cox proportional hazard regression analysis with the histologic subtype and other input variables used for the adjusted log-rank test; and 3) a multivariable mixture cure model including a cure probability model for patients with long-term survival and a failure time distribution model for those who died. 29,30 For a study population in which a subset of patients are long-term survivors, the mixture cure model can be used to investigate the heterogeneity between patients with two distinct prognoses and to assess their predictors, respectively (a cure probability model for long-term survival and a failure time distribution model for short-term survival). 29,30 The relevance of this analysis depends on whether the survival curve has a plateau after long-term follow-up. 29,30 Since early-stage, resected lung adenocarcinomas show a very long-term censored survival time, the cure model was deemed applicable to our study. 31 For the mixture cure model, the same covariates were used as in the multivariable Cox regression, but the clinical T categories were collapsed into three groups (i.e., cT1mi/cT1a/cT1b, cT1c, and cT2) and the histologic subtype category was also collapsed into two groups (lepidic/acinar/papillary vs. solid/micropapillary) to estimate con dence intervals (CIs) for the hazard ratio (HR) and odds ratio (OR) robustly.
We then performed subgroup analyses to evaluate the prognostic association of the histologic subtype with OS solely in adenocarcinomas manifesting as solid tumors on CT. Because of the small number of adenocarcinomas with cT1mi/cT1a and the lepidic-predominant subtype, they were combined with cT1b and the acinar/papillary-predominant subtypes, respectively. In addition, sensitivity analyses were performed after excluding those with mixed subtype adenocarcinoma.
All statistical analyses were conducted using R software, Bar plot for the number of adenocarcinomas according to the predominant histologic subtype, clinical T category, and nodule type.