To improve prognostic validity for non-small cell lung cancers, the eighth-edition staging system by the AJCC subclassifies the clinical and pathologic T categories in more detail based on the tumor extent; these categories are defined according to measurements of the solid portion size on CT and pathologically invasive component, respectively.24–26 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 classification 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-specific survival, although the classification was not associated with the prognosis in a multivariable Cox regression analysis. Likewise, Ujiie et al.18 reported a difference in OS between low (lepidic), intermediate (acinar/papillary), and high (solid/micropapillary) grades. Song et al.8 and Hung et al.6 suggested that solid/micropapillary and lepidic-predominant subtypes were negative and positive prognostic factors for OS, respectively. However, those studies were performed on the basis of the seventh-edition staging system, and the solid portion size or the invasive component size was not analyzed.6,8,17,18 In addition, the nodule type (i.e., the presence of a ground-glass opacity component in tumors), which is one of the pivotal covariates for prognostication in adenocarcinomas,19,20,27 was not included in their analyses. 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.4,5,7,9–11,13−16,22,23
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 reflects 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,29–31 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 confirmed 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 sufficient 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 classification, 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.33–36 The absence of this concept, as part of the latest classification of histologic subtypes, may have led to an inaccurate histologic classification, as it can be arbitrarily classified as the acinar or solid subtype.5 Further studies in a cohort with the latest pathological classification 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 classification is a prognostic factor for lung adenocarcinoma, no association of this classification 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.