This retrospective study was approved by our institutional review board, which waived the need for informed consent. From October 2017 to December 2023, we identified 626 early-stage lung adenocarcinoma patients who underwent EGFR mutation and ALK rearrangement retesting after surgery at the Second Hospital of Dalian Medical University.

Inclusion criteria included pathologically confirmed lung adenocarcinoma by surgical resection, available EGFR and ALK mutation results, clinical data (age, sex, smoking history, hypertension, diabetes, tumor biomarkers), and CT scans performed within a month prior to surgery. Exclusion criteria included missing CT scans, preoperative treatments (Neoadjuvant chemotherapy, radiotherapy, targeted therapy and immunotherapy), incomplete data, severe comorbidities, or a history of lung cancer.

Histological specimens were formalin-fixed and stained with hematoxylin-eosin. Two experienced pathologists reviewed the samples and recorded the subtype and stage according to the 8th edition TNM system and IASLC/ATS/ERS classification. EGFR mutations (exons 18-21) and ALK mutations were analyzed using next-generation sequencing (NGS). The detection is based on the Illumina sequencing platform and employs the target region probe capture technology. It examines at least 9 genes that are highly relevant to personalized treatment of non-small cell lung cancer. These genes must include EGFR, ALK, ERBB2 (HER2), BRAF, KRAS, MET, RET, ROS1, and NTRK1/2/3.ALK positivity was further confirmed by FISH testing.

High-resolution CT scans (1 mm slice thickness) were reviewed by two chest subspecialists who were blind to the genetic classification. Tumor features assessed included size, location, margins (Figure 1), and density (pure or mixed ground-glass nodules, subsolid, or solid).

Statistical analyses were conducted using SPSS version 21.0. An independent-sample Student’s t-test compared continuous variables, and chi-square tests compared categorical variables. Multivariate regression analysis was performed after addressing multicollinearity. A prediction tool for EGFR and ALK mutations was developed using principal component analysis, and receiver operating characteristic (ROC) curves were generated to calculate the area under the curve (AUC). Furthermore, univariate and multivariate subgroup analyses were conducted for EGFR exon 19 and exon 21 mutation sites.

A total of 626 patients underwent surgery between October 2017 and December 2023, with EGFR mutation and ALK rearrangement retesting conducted post-surgery. The cohort consisted of 234 men (37.4%) and 392 women (62.6%), with a median age of 60.8 years (range 24–80). Of the patients, 452 (72.2%) were non-smokers, with 163 male and 11 female smokers. All cases were invasive lung adenocarcinomas. Among the 626 patients, 367 had EGFR mutations and 39 had EML4-ALK fusion mutations, EGFR mutations included 113 (30.8%) exon 19 deletions, 186 (50.7%) exon 21 mutations, and 68 cases (18.5%) of rare mutations (Figure 2).

The clinical characteristics of the patients are summarized in Table 1. EGFR-positive patients had significantly lower mean CYFRA21–1 levels (2.358 vs. 2.796 ng/ml; P = 0.003). A higher proportion of women (66%) and non-smokers (65%) were found in the EGFR-positive group compared to the EGFR-negative group (46% and 41%, respectively; P = 0.000 for both). There were few ALK-positive patients, and no significant differences in clinical features were observed for this group.

The pathological features of surgically resected cases are shown in Table 2. EGFR-positive cases were more frequently lepidic and acinar (69% vs. 51% and 68% vs. 50%, respectively; P < 0.05), and less frequently solid and mucinous (37% vs. 60% and 11% vs. 62%, respectively; P < 0.05). EGFR-positive patients were more likely to be moderately differentiated (P = 0.000) and less likely to be poorly differentiated (P = 0.000). No significant pathological differences were found between ALK-positive and wild-type (WT) patients.

The CT features of surgically resected cases are shown in Table 3. EGFR mutations were significantly associated with nodule location (P = 0.013), particularly in the upper lobe of the left lung (68% vs. 56%; P = 0.011) and middle lobe of the right lung (48% vs. 60%; P = 0.035). EGFR-positive tumors had smaller solid components (1.073 vs. 1.380 cm; P = 0.003), with a higher mutation rate in ground glass nodules with 25-50% solid components (P = 0.002) and a lower rate in those with 75-100% solid components (P = 0.000) or pure solid nodules (P = 0.001). CT features such as the spicule sign (P = 0.001) and air bronchogram (P = 0.034) were more common in EGFR mutations. For ALK mutations, the mutation rate was higher in the left lower lobe (11% vs. 5%; P = 0.031), with less frequent co-occurrence of pure ground glass nodules (2.7% vs. 7.3%; P = 0.035) and vessel convergence signs (1.8% vs. 7.2%; P = 0.030).

Table 4 shows clinical differences between EGFR exon 19 deletions and L858R mutations. Both mutations were more common in females and non-smokers compared to WT EGFR (P < 0.05). EGFR exon 19 deletions occurred more frequently in patients under 60 years of age (58.38 vs. 62.22 years; P = 0.002) and had lower mean CYFRA21–1 levels compared to WT EGFR (2.217 vs. 2.796 ng/ml; P = 0.012). Additionally, EGFR exon 19 deletions were more common in younger patients compared to L858R mutations.

Table 5 shows the pathological features of EGFR 19 deletion and L858R mutations. Both mutations were more commonly associated with moderate differentiation (P = 0.000) and high frequencies of lepidic and acinar subtypes, while being rare in mucinous subtypes. EGFR 21 mutations were also rare in solid pathological subtypes (16% vs. 41%; P = 0.002).

Table 6 shows the CT features of EGFR 19 deletion and L858R mutations. Compared to WT EGFR, EGFR 21 mutations were more frequent in the upper lobe of the left lung (P = 0.038) and rare in the middle lobe of the right lung. EGFR L858R mutations had smaller solid components (1.030 vs. 1.380 cm; P = 0.007), with higher mutation rates in ground glass nodules with 25-50% solid components (P = 0.001). Both EGFR 19 and L858R mutations were rare in ground glass nodules with 75-100% solid components or pure solid nodules. EGFR 19 mutations were more likely to occur in ground glass nodules with 50-75% solid components. The CT characteristics that significantly differed between EGFR mutations and WT EGFR were the spicule sign (P = 0.011, P = 0.003). EGFR 21 mutations were more likely to show a bronchial cutoff sign (P = 0.034).

Multivariate analysis (Table 7) identified several independent predictors of EGFR mutations: gender, moderate differentiation, lepidic, acinar, and mucinous subtypes, middle lobe of the right lung, spicule sign, and air bronchogram (P < 0.05). For EGFR exon 19 mutations, independent predictors included age, degree of differentiation, lepidic, acinar subtypes, 75-100% solid component ratio, and spicule sign (P < 0.05). For EGFR exon 21 mutations, gender, moderate differentiation, lepidic, acinar subtypes, and spicule sign were significant (P < 0.05). Pure GGO and vessel convergence sign were independent predictors of ALK gene fusion (P < 0.05).

To validate the logistic regression model for predicting EGFR, EGFR19, EGFR21, and ALK mutation status, ROC curves were generated using the regression equations. The ROC for EGFR showed an AUC of 0.782 (P = 0.000, 95% CI: 0.745-0.819), for EGFR19 an AUC of 0.809 (P = 0.000, 95% CI: 0.764-0.855), for EGFR21 an AUC of 0.791 (P = 0.000, 95% CI: 0.750-0.832), and for ALK an AUC of 0.675 (P = 0.000, 95% CI: 0.595-0.755) (Figure 3).

The study reaffirmed that EGFR mutations are more prevalent in non-smoking women, which aligns with prior studies by Paez et al. (28) and Lynch et al. (29). The PIONEER study (6) also demonstrated a high EGFR mutation rate (51.4%) in Asian women and non-smokers with lung adenocarcinoma. In this study, smoking history did not emerge as an independent predictor of EGFR mutations in multivariate analysis, suggesting that smoking may not play a direct role in EGFR mutation status in this cohort, likely due to sample bias from the single-center study.

The results of this study showed that there was no correlation between the mean age or the 60-year age stratification and EGFR mutation status, which is consistent with previous literature (30, 31). Although some studies have suggested that diabetes may influence tumor biology and EGFR signaling pathways, thus participating in tumor development (32–34), no association between diabetes and EGFR mutation rate was found in this study. Hypertension and coronary heart disease also did not show a significant correlation with EGFR mutation.

Regarding tumor markers, although some studies have reported that elevated CEA levels are associated with a higher EGFR mutation rate (35–37), its independence as a risk factor remains controversial. NSE levels are not significantly correlated with EGFR mutation (38), but it may serve as a predictor of EGFR-TKI treatment response and a prognostic indicator for patients with EGFR mutations (39). In this study, the mean CYFRA21–1 level in the mutation group was lower than in the WT group (2.358 vs. 2.796 ng/ml, P = 0.003), but multivariate regression analysis did not identify it as an independent predictor of EGFR mutation. This is consistent with the retrospective study by Takeuchi et al. (40).

In terms of EGFR mutation subtypes, this study found that EGFR exon 19 and 21 mutations were more common in women and patients without a smoking history, which is in line with the conclusions of Cao et al. (41). However, after multivariate analysis, smoking was not an independent risk factor. Female gender was an independent predictor for EGFR exon 21 mutation, and EGFR exon 19 mutations were more common in patients under 60 years of age compared to those with wild-type EGFR, which was an independent predictor after multivariate analysis. The CYFRA21–1 level was lower in patients with EGFR exon 19 mutations but was not an independent predictor after regression analysis. Patients with EGFR exon 19 mutations were younger than those with exon 21 mutations (58.38 vs 62.22 years, P = 0.002).

Compared to EGFR gene mutations, ALK rearrangement is more likely to occur in younger, non-smoking or light-smoking lung adenocarcinoma patients (42). In this study, it was also found that the average age of patients with ALK mutations was younger, and the mutation rate was significantly higher in those under 60 years old. The mutation rate in patients without a smoking history was higher than in those with a smoking history. However, due to the small number of ALK fusion cases, no statistical difference was found. The research on the proportion of male and female ALK fusion mutation-positive patients is still controversial (43, 44). In this study, the ALK mutation rate was slightly higher in female patients, but no significant statistical difference was observed. There were no significant differences in tumor markers, hypertension, diabetes, or coronary heart disease.

Regarding tumor differentiation, this study found that EGFR mutations were more common in moderately differentiated adenocarcinomas and rare in poorly differentiated tumors, which slightly differs from previous research suggesting a lower mutation rate in poorly differentiated cases (45). In general, better tumor differentiation correlates with a better prognosis, supporting that EGFR mutation detection aligns with tumor differentiation in determining lung cancer prognosis.

Histopathological subtypes of lung adenocarcinoma show varying sensitivities to molecular-targeted drugs. Studies have indicated that certain subtypes, such as papillary and lepidic, have higher EGFR mutation rates, although the exact relationship remains unclear (46, 47). In this study, the mutation rate was higher in lepidic and acinar dominant subtypes, consistent with previous reports (21, 48–50). Previous studies (51) also showed that lepidic as the main subtype manifested as GGO on CT.The more ground glass components of GGO, the higher the EGFR mutation rate, which was consistent with the above results.Conversely, solid and mucinous subtypes had a lower EGFR mutation rate, with mucinous subtypes showing rare mutations, aligning with Inamura et al.’s findings (52).

Further analysis of EGFR mutations at exon 19/21 sites showed higher mutation rates in moderately differentiated adenocarcinomas, particularly in lepidic and acinar subtypes, and low mutation rates in mucinous subtypes. Solid subtypes also had a low incidence of EGFR21 mutations, though multivariate analysis did not identify this as an independent predictor.

Regarding ALK mutations, Wang et al. reported an association with solid dominant subtypes, which was also observed in this study, though the difference was not statistically significant (53). Several studies suggest ALK mutations are linked to lymph node metastasis, indicating poor prognosis (54–56). However, this study found no significant correlation between ALK mutations and lymph node metastasis, likely due to the inclusion of early-stage tumors and no significant differences in tumor differentiation or histological subtypes.

Regarding lesion location, previous studies have shown no correlation between lesion location and EGFR mutation status (23, 57). However, some studies suggest that EGFR mutations are more common in the upper lung lobes, possibly due to a complex interaction between genetic and environmental factors (58). In this study, univariate analysis found a higher EGFR mutation rate in tumors located in the upper lobe of the left lung and a lower rate in the middle lobe of the right lung. Multivariate regression analysis identified the middle lobe of the right lung as an independent risk factor for EGFR mutations, indicating the need for further research to determine whether lobe distribution in peripheral small lung adenocarcinomas can predict EGFR mutations.

In terms of CT features, early lung adenocarcinomas often present as pure ground glass opacity (pGGO), partial solid nodules, or solid nodules. Previous studies have suggested a relationship between EGFR mutations and solid components of GGO (59, 60), although findings are inconsistent. Most studies report that EGFR mutations correlate with a higher proportion of GGO (61, 62), suggesting that GGO could predict EGFR mutations. However, other studies (63, 64) found that EGFR mutations were more common in lesions with more than 50% solid components. This study showed that EGFR mutation patients had smaller maximum solid component diameters, with a higher mutation rate in tumors with 25-50% solid components and a lower rate in tumors with 75-100% solid components. In pure solid nodules, the mutation rate was also low, supporting a positive correlation between EGFR mutations and GGO components. The presence of solid components in GGO may indicate increased tumor aggression and pathological grade (60).

CT features like the air bronchogram, spicule sign, and tumor size have been linked to EGFR mutations. Some studies suggest a positive correlation between air bronchogram and EGFR mutation (58, 65), while others report a negative correlation (66). In this study, EGFR mutations were positively correlated with the spicule sign and negatively correlated with the air bronchogram, with no significant correlation found with other features. Multivariate analysis identified spicule sign and air bronchogram as independent predictors of EGFR mutations.

Further analysis of EGFR mutation subtypes, specifically exon 19/21 mutations, found that EGFR21 mutations were more common in the upper lobe of the left lung and rare in the middle lobe of the right lung, consistent with Tengeng et al.’s study (58). However, multivariate regression showed no correlation between lesion location and EGFR mutation. Solid components of GGO in EGFR21 patients were smaller than in EGFR19 patients, and EGFR21 mutations were more frequent in patients with 25-50% solid components, consistent with Lee et al.’s findings (18). The mutation rate increased with the proportion of GGO. Both EGFR19 and EGFR21 mutations had lower mutation rates in pure solid nodules or GGO with 75-100% solid content and higher rates with spicule signs. There was no significant difference in the air bronchogram between the two mutation types.

In terms of ALK mutations, previous studies have shown that ALK gene fusion is associated with larger tumor volumes and higher solid component proportions (55). This study found the same trend, but no statistical significance due to the small number of ALK fusion cases. ALK mutations were more common in the left inferior lobe, but this was not an independent predictor. Studies have shown that ALK mutations are associated with larger tumor diameters and higher solid component proportions, which was confirmed in this study, although no significant statistical differences were observed due to the small sample size. Pure GGO nodules had a very low ALK mutation rate (P<0.05), with mutation rates increasing as the solid component proportion increased.

Other studies (67–71) have shown that ALK-positive tumors are less likely to exhibit cavity signs and air bronchograms, and are associated with larger tumor size and a higher proportion of solid components. In this study, ALK mutations were less common in patients with vessel convergence signs, with no statistical differences found in other imaging features. This may be related to the fact that the lung cancer patients in this study were early-stage and had fewer lymph node or distant metastases.

Multivariate analysis identified several independent predictors for EGFR mutations: female gender, moderate differentiation, lepidic and acinar subtypes, mucinous subtype, middle lobe of the right lung, spicule sign, and air bronchogram. Subgroup analysis revealed that age under 60, degree of differentiation, lepidic and acinar subtypes, 75-100% GGO solid component, and spicule sign were independent predictors for EGFR19 mutation. For EGFR21 mutation, female gender, moderate differentiation, lepidic and acinar subtypes, and spicule sign were independent predictors. Pure GGO and vessel convergence sign were independent predictors for ALK mutation.

An ROC curve analysis showed the diagnostic value of clinical, pathological, and HRCT models for predicting EGFR, EGFR19, EGFR21, and ALK mutations. The AUC values were 0.782 for EGFR, 0.809 for EGFR19, 0.791 for EGFR21, and 0.675 for ALK, indicating that the combined radiomic and clinical feature models can effectively predict gene mutation status.

Limitations: (1) The study was a single-center, retrospective analysis with 100% Asian patients, which may introduce selection bias. (2) The low ALK fusion rate resulted in fewer positive cases, and differences from previous studies may be due to the early-stage tumors. Further research with a broader patient population and extended follow-up is needed to assess the relationship between gene mutations, clinicopathological features, and prognosis. (3) Only early-stage lung adenocarcinomas were included, so it remains unclear whether the findings apply to advanced stages.