第一部分

基于IASLC组织学新分类的病理I期（pT1aN0M0~pT2aN0M0）肺腺癌CT影像学特征研究

目的：识别对病理I期肺腺癌患者IASLC新组织学分级具有独立预测价值的CT影像学特征，并探讨这些特征在不同性别、不同结节类型中的异同。

材料与方法：回顾性分析2014年1月至2017年6月于我院接受根治性切除术的379例病理I期肺腺癌患者的临床、影像及病理资料。复阅病理切片并根据2020年IASLC提出的浸润性肺腺癌三级分级系统进行组织学分级。采用多元有序logistic回归分析寻找组织学分级的独立预测因素，并采用ROC曲线评估其预测性能。分别基于性别和结节类型对纳入患者进行分层，观察不同亚组中组织学等级独立预测因素的异同。

结果：全肿瘤大小（OR=1.744，95%CI：1.225～2.479，P= 0.002）、CT值（OR=3.766，95%CI：1.768～8.020，P=0.001）和实性成分占比（consolidation tumor ratio，CTR）（OR=2.145，95%CI：1.534～2.995，P＜0.001）是组织学等级的独立预测因子。全肿瘤大小、CT值、CTR单独用于预测组织学1级vs. 2/3级和1/2级vs. 3级的曲线下面积（area under curve，AUC）分别为0.633和0.673、0.921和0.879、0.884和0.878，cutoff值依次分别为12mm和17mm、-420HU和-205HU、25%和75%。亚组分析的结果显示，对于女性，全肿瘤大小（OR=1.61，95%CI：1.02～2.53，P= 0.040）、CT值（OR=4.24，95%CI：1.57～11.44，P=0.004）和CTR（OR=2.03，95%CI：1.30～3.18，P=0.002）是组织学等级的独立预测因素；对于男性，全肿瘤大小（OR=2.37，95%CI：1.27～4.43，P=0.007）、CT值（OR=6.30，95%CI：1.61～24.71，P=0.008）、CTR（OR=1.84，95%CI：1.05～3.23，P=0.034）、胸膜牵拉（OR=5.41，95%CI：1.85～15.84，P=0.002）和从不吸烟（OR=0.34，95%CI：0.12～0.96，P=0.041）是组织学等级的独立预测因素。对于非实性结节，全肿瘤大小（OR=2.39，95%CI：1.06～5.38，P=0.036）和CT值（OR=1.012，95%CI：1.004～1.020，P=0.003）是组织学等级的独立预测因素；对于部分实性结节，CT值（OR=1.007，95%CI：1.002～1.012，P=0.007）和SD值（OR=1.011，95%CI：1.001～1.021，P=0.036）是组织学等级的独立预测因素；对于实性结节，只有全肿瘤大小（OR=2.89，95%CI：1.12～7.46，P=0.028）是组织学等级的独立预测因素。

结论：全肿瘤大小、平均CT值和CTR是病理I期肺腺癌IASLC新组织学分级系统的独立预测因子。对组织学等级具有独立预测价值的CT影像表现在不同性别的患者、不同类型的病变间存在差异。

第二部分

病理I期（pT1aN0M0~pT2aN0M0）肺腺癌术后复发危险因素的识别及预后模型的构建研究

目的：识别病理I期肺腺癌的独立预后危险因素并构建列线图模型。比较不同性别、不同结节类型间预后危险因素的异同。

材料与方法：对2014年1月至2017年6月于我院接受根治性切除术的379例病理I期肺腺癌患者进行术后随访。采用Kaplan-Meier法绘制无病生存曲线来验证临床T分期、IASLC组织学分级系统及与之关联的CT特征的预后分层表现。分别基于“术前+术后（即组织学分级）”和“仅术前”独立预测因子行两次多因素Cox回归分析，并构建预后列线图模型1和模型2。通过时间依赖性C指数、校准曲线和临床决策曲线来评估两个模型的预测表现。基于性别、结节类型将纳入对象分层，观察不同亚组中预后危险因素的差异。

结果：中位随访时长70.4个月，共60例患者在随访中复发。全肿瘤大小（cutoff值17mm）、CT值（cutoff值-205HU）和CTR（cutoff值25%和75%）可以实现对患者预后的有效分层。模型1中的预测因子包括：年龄（HR：1.05，95%CI：1.02～1.09，P=0.003）、毛刺征（HR：5.96，95%CI：2.30～15.43，P＜0.001）、临床T分期（HR：2.32，95%CI：1.53～3.52，P＜0.001）和组织学分级（HR：4.31，95%CI：2.28～8.14，P＜0.001）。模型2中的预测因子包括：年龄（HR：1.04，95%CI：1.01～1.08，P=0.015）、毛刺征（HR：4.55，95%CI：1.73～11.95，P=0.002）、临床T分期（HR：2.49，95%CI：1.58～3.93，P＜0.001）、肿瘤大小（HR：2.81，95%CI：1.16～6.77，P=0.022）和CTR（HR：2.49，95%CI：1.19～5.25，P=0.016）。模型1与模型2具有相似的预后区分度，模型1的时间依赖性C指数稳定在0.9左右，模型2接近0.9。基于性别分层的分析结果显示，在女性亚组中，肿瘤大小（HR：3.97，95%CI：1.02～15.41，P=0.046）、CTR（HR：3.38，95%CI：1.17～9.75，P=0.024）和毛刺征（HR：4.81，95%CI：1.33～17.42，P=0.017）是DFS的独立术前预测因子；而在男性亚组中，临床T分期（HR：3.80，95%CI：2.49～5.79，P＜0.001）和毛刺征（HR：5.57，95%CI：1.26～24.61，P=0.024）是DFS的独立术前预测因子。基于结节类型分层的结果显示，在部分实性结节亚组中，肿瘤大小（HR：1.14，95%CI：1.07～1.22，P＜0.001）、CTR（HR：2.40，95%CI：1.32～4.37，P=0.004）和毛刺征（HR：4.95，95%CI：1.27～19.34，P=0.021）是DFS的独立术前预测因子；而在实性结节亚组中，临床T分期（HR：3.27，95%CI：2.09～5.12，P＜0.001）和年龄（HR：1.09，95%CI：1.04～1.13，P＜0.001）是DFS的独立术前预测因子。

结论：通过整合与IASLC新组织学分级系统关联的传统CT影像特征（全肿瘤大小和CTR）和临床T分期而构建的术前预测模型与增加了组织学分级的术后预测模型相比，具有相近的DFS预测表现。在部分实性结节亚组和女性亚组中，全肿瘤大小和CTR是DFS的独立术前预测因子。在实性结节亚组中，临床T分期和年龄是DFS的独立术前预测因子。

PART I

CT imaging features of IASLC new histological grading system in patients with pathological stage I lung adenocarcinoma 

Objective To identify CT imaging features that have independent predictive value for IASLC new histological grading system in patients with pathological stage I lung adenocarcinoma, and to explore their similarities and differences in different genders and nodule types.

Materials and Methods We retrospectively analyzed the clinical, imaging and pathological data of 379 patients with pathological stage I lung adenocarcinoma who underwent radical resection in our hospital from January 2014 to June 2017. Pathological sections were reviewed and histologic grade was recorded according to the three-stage grading system for invasive lung adenocarcinoma proposed by IASLC in 2020. Multivariate ordered logistic regression analysis was used to identify independent predictors of histological grading system, and ROC curves were used to evaluate their predictive performance. Stratified analyses were also performed within the various subgroups according to sex and nodule type, and the similarities and differences of independent predictors for histological grade in different subgroups were observed.

Results Whole tumor size (OR=1.744, 95%CI: 1.225~2.479, P=0.002), CT value (OR=3.766, 95%CI: 1.768~8.020, P=0.001) and CTR (OR=2.145, 95%CI: 1.534~2.995, P<0.001) were independent predictors of histological grading system. Whole tumor size, CT value, and CTR alone were used to predict histologic grade 1 vs. grade 2/3 and grade 1/2 vs. grade 3 with AUC values of 0.633 and 0.673, 0.921 and 0.879, 0.884 and 0.878, respectively, and cutoff values of 12mm and 17mm, -420HU and -205HU, 25% and 75%, respectively. The results of subgroups analysis showed that in the female subgroup, whole tumor size (OR=1.61, 95%CI: 1.02~2.53, P=0.040), CT value (OR=4.24, 95%CI: 1.57~11.44, P=0.004) and CTR (OR=2.03, 95%CI: 1.30~3.18, P=0.002) were independent predictors of histological grade. In the male subgroup, whole tumor size (OR=2.37, 95%CI: 1.27~4.43, P=0.007), CT value (OR=6.30, 95%CI: 1.61~24.71, P=0.008), CTR (OR=1.84, 95%CI: 1.05~3.23, P=0.034), pleural traction (OR=5.41, 95%CI: 1.85~15.84, P=0.002) and never smoking (OR= 0.34, 95% CI: 0.12~0.96, P=0.041) were independent predictors of histological grade. For non-solid nodules, whole tumor size (OR=2.39, 95%CI: 1.06~5.38, P=0.036) and CT value (OR=1.012, 95%CI: 1.004~1.020, P=0.003) were independent predictors of histological grade. For part-solid nodules, CT value (OR=1.007, 95%CI: 1.002~1.012, P=0.007) and SD value (OR=1.011, 95%CI: 1.001~1.021, P=0.036) were independent predictors of histological grade. For solid nodules, only whole tumor size (OR=2.89, 95%CI: 1.12~7.46, P=0.028) was the independent predictor of histological grade.   

Conclusions Whole tumor size, mean CT value, and CTR were independent predictors of IASLC new histological grading system in patients with pathological stage I lung adenocarcinoma. CT imaging features with independent predictive value for histological grade were varied between different sexes and different nodule types.

PART II

Identification of prognostic factors and construction of nomogram models in pathological stage I lung adenocarcinoma

Objective To identify independent prognostic risk factors for pathological stage I lung adenocarcinoma and construct nomogram models. To explore the heterogeneity of prognostic factors between different sexes and different nodule types.

Materials and Methods 379 patients diagnosed with pathological stage I lung adenocarcinoma, who underwent radical resection in our hospital from January 2014 to June 2017, were enrolled and followed up. The disease-free survival curves were drawn by Kaplan-Meier method to verify the prognostic stratification of clinical T stage, IASLC new histological grading system and associated CT features. Twice multivariate Cox regression analyses were performed based on "preoperative + postoperative (i.e., histologic grade)" and "preoperative only" independent predictors, then prognostic nomogram model 1 and model 2 were constructed. The predictive performance of the two models was evaluated by time-dependent C-index, calibration curve, and clinical decision curve. In addition, we performed subgroup evaluation by stratifying analyses based on gender and nodule type to explore the differences of prognostic risk factors in different subgroups.

Result The median follow-up was 70.4 months, and a total of 60 patients relapsed during follow-up. Whole tumor size (cutoff of 17 mm), CT value (cutoff of -205HU) and CTR (cutoff of 25% and 75%) could effectively stratify patient prognosis. The predictors in model 1 included: age (HR: 1.05, 95% CI: 1.02~1.09, P=0.003), burrs sign (HR: 5.96, 95% CI: 2.30~15.43, P<0.001), clinical T stage (HR: 2.32, 95% CI: 1.53~3.52, P<0.001) and histological grade (HR: 4.31, 95% CI: 2.28~8.14, P<0.001). The predictors in model 2 included: age (HR: 1.04, 95% CI: 1.01~1.08, P=0.015), burrs sign (HR: 4.55, 95%CI: 1.73~11.95, P=0.002), clinical T stage (HR: 2.49, 95% CI: 1.58~3.93, P<0.001), whole tumor size (HR: 2.81, 95% CI: 1.16~6.77, P=0.022) and CTR (HR: 2.49, 95% CI: 1.19~5.25, P=0.016). Model 1 and Model 2 have similar prognostic discrimination, with the time-dependent C index of Model 1 stable at about 0.9 and Model 2 close to 0.9. The results based on gender stratification showed that in the female subgroup, tumor size (HR: 3.97, 95% CI: 1.02~15.41, P=0.046)，CTR (HR: 3.38, 95% CI: 1.17~9.75, P=0.024), and burrs sign (HR: 4.81, 95% CI: 1.33~17.42, P=0.017) were independent preoperative predictors of DFS. In the male subgroup, clinical T stage (HR: 3.80, 95% CI: 2.49~5.79, P<0.001) and burrs sign (HR: 5.57, 95% CI: 1.26~24.61, P=0.024) were independent preoperative predictors of DFS. The results stratified based on nodule types showed that in part-solid nodule subgroups, whole tumor size (HR: 1.14, 95% CI: 1.07~1.22, P<0.001), CTR (HR: 2.40, 95% CI: 1.32~4.37, P=0.004) and burrs sign (HR: 4.95, 95% CI: 1.27~19.34, P=0.021) were independent preoperative predictors of DFS. In pure solid nodule subgroup, clinical T stage (HR: 3.27, 95% CI: 2.09~5.12, P<0.001) and age (HR: 1.09, 95% CI: 1.04~1.13, P<0.001) were independent preoperative predictors of DFS.

Conclusions The preoperative prediction model constructed by integrating clinical T stage and the traditional CT imaging features (whole tumor size and CTR) associated with the IASLC new histological grading system has similar prediction performance for DFS, compared with the postoperative prediction model which added histological grading system into it. In part-solid nodule subgroup and female subgroup, whole tumor size and CTR were independent preoperative predictors of DFS. In the solid nodule subgroup, clinical T stage and age were independent preoperative predictors of DFS.