Epidermal growth factor receptor mutations in lung adenocarcinoma: associations between dual-energy spectral CT measurements and histologic results

Abstract

Objective

To analyze the relationship between dual-energy spectral CT and epidermal growth factor receptor (EGFR) mutation status in patients with lung adenocarcinoma.

Methods

The quantitative parameters of spectral CT were analyzed in 208 patients with lung adenocarcinoma. The quantitative parameters including CT_40keV and CT_70keV values, effective atomic number (Zeff), iodine concentration (IC), water concentration (WC), and the slope of the spectral curve (λ _HU) were calculated. Statistical analysis was used to determine the clinical characteristics and quantitative parameters for the diagnosis of EGFR-mutation status. The ROC curves were used to calculate diagnostic efficiency.

Results

Sex (p = 0.027) and smoking history (p = 0.019) differed significantly according to the EGFR-mutation status. Spectral CT quantitative parameters (CT_40keV and CT_70keV values, λ _HU, Zeff and IC) differed significantly between the EGFR mutant and the EGFR wild-type groups (p < 0.05) during the arterial phase (AP) and venous phase (VP). However, WC was not statistically different between the two groups (p > 0.05). ROC curve analysis revealed the combination of the significantly different quantitative parameters provided the best diagnostic performance for determining the EGFR-mutation status (AUC: 76.0%) in the AP, while the AUC during the VP was 75.6%.

Conclusion

The quantitative parameters of dual-energy spectral CT have potential value for identifying the EGFR-mutation status.

Introduction

The morbidity and mortality of lung cancer remain the highest among all malignancies worldwide, accounting for an incidence of 2.2 million cases and 1.9 million deaths in 2017, respectively (Global Burden of Disease Cancer et al. 2018). Non-small cell lung cancer (NSCLC) accounts for more than 80% of cases of lung cancer, and adenocarcinoma is the most common histopathological type (Hsu et al. 2018). In the past decade, great progress has been made in research on the pathological analysis and molecular biology of lung cancer, especially NSCLC (Chen et al. 2014). NSCLC is defined as a heterogeneous condition (Kerr 2009; Travis 2009). Recently, epidermal growth factor receptor (EGFR) was a key molecule in investigation of lung adenocarcinoma (Liu et al. 2016b).

Patients with EGFR mutations treated with EGFR tyrosine kinase inhibitors (TKI) exhibited significantly longer survival than those treated with standard platinum-based chemotherapy (An et al. 2016; Jeon et al. 2015; Kim et al. 2013; Maemondo et al. 2010; Rosell et al. 2012). EGFR TKI therapy has a significant initial effect in approximately 70–80% of patients with EGFR mutations (Choi et al. 2014). Therefore, it is important to determine the EGFR-mutation status of patients with lung adenocarcinoma before starting TKI therapy. However, before starting targeted therapy, due to the small biopsy samples, sampling errors or inoperability, it is not always possible to obtain sufficient sample tissue to determine the EGFR-mutation status (Choi et al. 2014). In this situation, non-smoking, female, East Asian ethnicity, and histological subtype of adenocarcinoma have been tentatively identified as potential preconditions for responsiveness to TKI (Dearden et al. 2013; Kris et al. 2003; Liu et al. 2016b; Rosell et al. 2009; Shepherd et al. 2005). Some studies have used of computed tomography (CT) radiomics features to study its association with EGFR mutations. A majority of these studies found that certain characteristics CT radiomics can predict the EGFR-mutation status (Hong et al. 2016; Jia et al. 2019; Kim et al. 2016; Liu et al. 2016a). Moreover, some researchers, who tried to use CT features to predict EGFR mutations, found that some CT features, such as ground-glass opacity, air bronchogram, bubble-like lucencies, vascular convergence, and pleural retraction, may be associated with EGFR mutations (Choi et al. 2014, 2015; Lee et al. 2013; Liu et al. 2016b; Suh et al. 2018). However, these CT features are dependent on the observer’s subjective judgment and cannot be used for quantitative evaluation.

Dual-energy spectral CT uses a single bulb for instantaneous dual kVp (80 kVp and 140 kVp) fast-switching alternate scans, which enables the acquisition of a total of 101 single-energy images between 40–140 keV, and provides quantitative information with multiple base material pairs (such as iodine/water, iodine/calcium, iodine/fat, etc.) (Matsuda et al. 2012). A characteristic performance spectral curve, reflecting different types of tissues and lesions of the human body, can be obtained based on the single-energy data, and the slope of the lesion can be calculated using the energy spectral curve. The adaptive statistical iterative reconstruction technology feature of spectral CT uses an iterative algorithm to acquire images with the best resolution, and can effectively suppress noise and achieve high-definition imaging at low-dose conditions (Koc et al. 2014). The most significant advantage of spectral CT over conventional CT is that spectral CT provides a variety of quantitative analyses tools and multi-parameter imaging-based comprehensive diagnostic modes, with broad prospects for clinical applications. Spectral CT has obvious advantages for identifying benign and malignant lung tumors, distinguishing between the histological subtypes, and detecting lymph-node metastasis in lung cancer (Chae et al. 2008; González-Pérez et al. 2016; Ohana et al. 2014). Therefore, the purpose of this study was to retrospectively analyze the value of the quantitative parameters of spectral CT for identifying the EGFR-mutation status.

Materials and methods

Patient enrollment

This retrospective study was approved by the Institutional Review Board of our institution, and the need for informed consent was waived. We included patients who had undergone histological biopsy or surgical resection and were pathologically confirmed as lung adenocarcinoma, and these patients underwent preoperative chest CT examination at our hospital between March 2017 and October 2019. The inclusion criteria for our study were as follows: (1) patients aged 18 years or older; (2) lung adenocarcinoma confirmed by pathological examination after surgical resection; (3) those who underwent EGFR-mutation testing; (4) patients who underwent dual-phase enhanced chest CT under the gemstone spectral imaging (GSI) mode 2 weeks before surgery; (5) no chemotherapy, radiotherapy, or immunotherapy before the CT scan; and (6) no history of other malignancies. Imaging data with poor quality, respiratory artifacts, and other conditions affecting observation were excluded.

According to the above-mentioned criteria, a total of 208 patients with lung adenocarcinoma [mean age, 57.35 years ± 10.19 (range 27–85 years)] were included, of which 111 were men [mean age, 58.89 years ± 10.05 (range 27–85 years)] and 97 were women [mean age, 55.59 years ± 10.12 (range 29–78 years)]. A significant difference was observed between the mean ages of men and women (P = 0.019, two-sample t test). Flowchart of inclusion and exclusion criteria is shown in Fig. 1

Fig. 1.

Flowchart of inclusion and exclusion criteria. EGFR: epidermal growth factor receptor

EGFR-mutation detection

EGFR mutations were detected in patients who underwent surgical excision, followed by pathological confirmation of the adenocarcinoma. Molecular analysis of the mutation state of EGFR exons 18–21 was performed with the polymerase chain reaction (PCR)–-based amplification refractory mutation system (ARMS) using the EGFR-mutation detection kit (Beijing SinoMD Gene Detection Technology Co., Ltd, Beijing, China).

CT scanning protocol

The Discovery CT750 HD (GE Healthcare, Waukesha, WI, USA) scanner was used to perform CT for all the patients. The scanning parameters were as follows: tube voltage, 120 kVp; tube current, 375 mA; and reconstruction thickness and reconstruction interval, 1.25 mm each. A non-ionic iodine contrast agent (Ultravist 300, Bayer Pharma, Berlin, Germany) was injected through the cubital vein using a high-pressure syringe (XD8000, Ulrich, Germany) at a dose of 1.3–1.5 mL per kilogram of body weight at a rate of 3.0 mL/s. The auto-tracking method was used to obtain scans in the arterial phase (AP) and venous phase (VP) of the GSI mode, 30 s and 60 s after contrast injection, respectively.

Quantitative image analysis

Two thoracic radiologists, with 7 and 5 years of experience, analyzed the CT images independently and retrospectively at the image processing workstation (AW4.6, GE Healthcare). Two radiologists were blinded to the clinical information and pathological results. Consensus was achieved through discussion in cases of disagreement.

Round or oval lesions were placed in areas where the lesions were uniformly strengthened; attempts were made to avoid areas such as calcification, blood vessels, necrosis, and cystic areas that were visible to the naked eye, and could affect the measurement results. The area of interest was placed in the largest possible lung tumor area except for necrotic or cystic areas. To ensure the consistency of the results, three consecutive images were measured, and the average value was calculated. For all measurements, the copy and paste function was used to maintain consistency of the size, shape, and location of the area of interest as far as possible during the AP and VP. The GSI software automatically generated single-energy CT values, iodine concentration (IC), water concentration (WC), and effective atomic number (Zeff) of the lesion at 40 and 70 keV and calculated the slope of the spectral curve (λ). The CT values corresponding to the two energy levels (40 keV and 70 keV, respectively) were calculated by dividing the energy difference using the following equation:

$${\lambda }_{\text{70keV}}=\left({\text{CT}}_{\text{40keV}}-{\text{CT}}_{\text{70keV}}\right)/\left(70-40\right).$$

Statistical analysis

Statistical analysis was performed using SPSS Statistics for Windows (version 22.0, IBM Corporation, Armonk, NY, USA). The Wilcoxon rank sum test and the Chi-square test were used to compare the continuous and categorical variables between the two groups, respectively. Continuous variables were expressed as mean ± standard deviation, and categorical variables were expressed as percentages. The two-sample t test was used to compare the quantitative parameters of EGFR-mutation and wild-type EGFR. Two-sided P < 0.05 indicated a significant difference. A receiver-operating characteristic curve (ROC) was generated to evaluate the diagnostic efficiency for values with statistically significant differences. The cutoff, sensitivity, and specificity of the highest value of Youden’s index (YI) were calculated.

Results

Clinical characteristics of the patients

The clinical characteristics of the patients are presented in Table 1. No significant differences were observed in the age (p = 0.654) and histopathological stage (p = 0.856) between the EGFR mutation and EGFR wild-type groups. Sex and smoking history differed significantly between the two groups (p = 0.027 and p = 0.019, respectively). EGFR mutations were more common in women and never-smokers, while the wild-type EGFR was more common in men and smokers. The long-axis and short-axis diameters of EGFR mutation and wild-type patients were 4.54 ± 1.85 vs. 5.22 ± 2.47 (P = 0.026) and 3.38 ± 2.34 vs. 3.65 ± 1.74 (P = 0.841), respectively.

Table 1.

Relationship between clinical characteristics and EGFR-mutation status

Variable	All patients (n = 208)	EGFR+ (n = 128)	EGFR− (n = 80)	P
Age (years)
Mean ± SD	57.35 ± 10.19	57.86 ± 9.88	57.03 ± 10.41	0.654
Median (range)	58.5 (27–85)	59 (36–78)	58 (27–85)
Sex
Male	102 (49.0%)	55 (43.0%)	47 (58.8%)	0.027
Female	106 (51.0%)	73 (57.0%)	33 (41.2%)
Smoking history*
Smoker	88 (42.3%)	46 (35.9%)	42 (52.5%)	0.019
No smoker	120 (57.7%)	82 (64.1%)	38 (47.5%)
Pathological stage
I or II	131 (63.0%)	80 (62.5%)	51 (63.7%)	0.856
III or IV	77 (37%)	48 (37.5%)	29 (36.3%)
Long-axis diameter	4.80 ± 2.13	4.54 ± 1.85	5.22 ± 2.47	0.026
Short-axis diameter	3.60 ± 2.83	3.38 ± 2.34	3.65 ± 1.74	0.841

EGFR epidermal growth factor receptor, EGFR⁺ EGFR mutation, EGFR⁻ EGFR wild type, SD standard deviation

*Smoking history is defined as follows: smokers, former and current smokers; no smoker, never smoked

Quantitative image analysis

Figures 2 and 3 shows two sets of representative images (slice thickness: 1.25 mm) acquired by single-spectral CT (70 keV) for patients with EGFR mutations and wild-type EGFR, respectively. Statistically significant differences were observed between CT_40keV, CT_70keV, λ_70keV, IC, and Zeff (p < 0.05) during the AP and VP in patients with EGFR mutations and wild-type EGFR; however, no significant differences were observed in the WC between the two groups during the AP and VP (p > 0.05). CT_40keV, CT_70keV, λ_70keV, Zeff, and IC were higher in the EGFR-mutation group than that in the EGFR wild-type group during the AP and VP. Comparisons between the quantitative parameters of single-spectral CT in the EGFR-mutation group and the EGFR wild-type group are presented in Table 2 and Fig. 4.

Fig. 2.

Spectral computed tomography (CT) images and pathological section of a 66-year-old man with EGFR mutation. Lung window (a); arterial phase (b~j); venous phase (k~o, q~t); mediastinal window (f, k). b, l 70 keV single-energy pseudocolor image (CT number of 70 keV was 62.35 HU and 49.65 HU, respectively); (c, m) water concentration pseudocolor image (WC was 1032.71 mg/cm³ and 1026.12 mg/cm³, respectively); (d, n) iodine concentration pseudocolor image (IC was 17.20 100 ug/cm³ and 12.64 100 ug/cm³, respectively); (e, o) effective atomic number pseudocolor image (Zeff was 8.37 and 8.26, respectively); (g, q) spectral curve (λ_70keV was 2.50 and 2.16, respectively); (h, r) water (iodine) concentration scatter plot; (i, s) iodine (water) concentration scatter plot; (j, k) effective atomic number histogram; (p) hematoxylin and eosin staining (original magnification × 200)

Fig. 3.

Spectral computed tomography (CT) images and pathological section of a 70-year-old man with EGFR wild-type. Lung window (a); arterial phase (b~j); venous phase (k~o, q~t); mediastinal window (f, k). (b, l) 70kev single-energy pseudocolor image (CT number of 70kev was 75.39 HU and 46.02 HU, respectively); (c, m) water concentration pseudocolor image (WC was 1043.70 mg/cm³ and 1021.19 mg/cm³, respectively); (d, n) iodine concentration pseudocolor image (IC was 12.85 100 ug/cm³ and 13.94 100 ug/cm³, respectively); (e, o) effective atomic number pseudocolor image (Zeff was 8.36 and 8.44, respectively); (g, q) spectral curve (λ_70kev was 2.23 and 2.41, respectively); (h, r) water (iodine) concentration scatter plot; (i, s) iodine (water) concentration scatter plot; (j, k) effective atomic number histogram; (p) hematoxylin and eosin staining (original magnification × 400)

Table 2.

Comparison of parameter between EGFR⁺ and EGFR⁻ in the arterial and venous phases

Parameter	EGFR+ (n = 128)	EGFR− (n = 80)	t	P
AP
CT40keV (HU)	172.85 ± 9.31	164.70 ± 9.04	6.21	< 0.001
CT70keV (HU)	71.33 ± 3.91	68.15 ± 3.93	5.68	< 0.001
λ70keV	3.38 ± 0.29	3.22 ± 0.28	4.06	< 0.001
Zeff	8.62 ± 0.07	8.59 ± 0.07	2.80	0.02
IC (100ug/cm3)	17.11 ± 1.20	16.66 ± 1.39	2.93	< 0.01
WC (mg/cm3)	1029.81 ± 5.97	1030.24 ± 5.19	0.52	0.51
VP
CT40keV (HU)	139.76 ± 6.35	134.58 ± 5.32	6.33	< 0.001
CT70keV (HU)	65.82 ± 5.52	61.60 ± 4.22	6.22	< 0.001
λ70keV	2.46 ± 0.06	2.43 ± 0.07	3.29	< 0.01
Zeff	8.41 ± 0.06	8.39 ± 0.06	3.00	0.01
IC (100 ug/cm3)	13.81 ± 0.84	13.33 ± 0.95	3.77	< 0.01
WC (mg/cm3)	1029.81 ± 5.97	1029.34 ± 6.37	0.31	0.72

AP arterial phase, EGFR epidermal growth factor receptor, EGFR⁺ EGFR mutation, EGFR⁻ EGFR wild-type, HU Hounsfield, IC iodine concentration, VP venous phase, WC water concentration, Zeff effective atomic number

Fig. 4.

Scatter plots and histograms of the dual-energy spectral attenuation parameters for EGFR⁺ and EGFR⁻. The horizontal lines represent error bars. a, b Scatter plot histograms of (a) CT40keV value and (b) CT70keV value in the AP and VP for EGFR⁺ and EGFR⁻. c, d Scatter plots histograms of (C) λ _HU and (d) effective atomic number (Zeff) in the AP and VP for EGFR⁺ and EGFR⁻. e, f Scatter plots histograms of (e) iodine concentration (IC) and (f) water concentration (WC) in the AP and VP for EGFR⁺ and EGFR⁻. AP arterial phase, EGFR epidermal growth factor receptor, EGFR⁺ EGFR mutation, EGFR⁻ EGFR wild type, VP venous phase. *P < 0.05; **P < 0.01; Ns. No significance

ROC curve analysis

ROC curve analysis was used to determine the optimal sensitivity and specificity of the threshold parameters required to distinguish between EGFR mutations and wild-type EGFR (Table 3). The sensitivity and specificity for distinguishing between patients with EGFR mutations and those with wild-type EGFR were 60.2% and 75.0%, respectively, when the threshold λ_70keV was 3.365 during the AP. The sensitivity and specificity were 54.7% and 75.0%, respectively, when the threshold λ_70keV was 2.465 during the VP.

Table 3.

Performance of the different parameters in distinguishing between EGFR⁺ and EGFR⁻⁻ using ROC analysis

Parameter	AUC (%)	YI	Threshold	Sensitivity (%)	Specificity (%)
AP
CT40 keV (HU)	72.8	0.395	167.375	69.5	70.0
CT70 keV (HU)	71.5	0.325	69.305	68.8	63.8
λ70keV	66.4	0.352	3.365	60.2	75.0
Zeff	59.9	0.166	8.605	57.8	58.8
IC (100ug/cm3)	60.7	0.192	17.595	55.5	63.8
CT40 keV +70 keV + λ70 keV + Zeff + IC	76.0	0.380	0.621	68.0	70.0
VP
CT40 keV (HU)	72.1	0.327	135.460	72.7	60.0
CT70 keV (HU)	71.4	0.350	64.080	62.5	72.5
λ70keV	66.4	0.297	2.465	54.7	75.0
Zeff	60.2	0.173	8.435	42.2	83.8
IC (100ug/cm3)	64.2	0.230	12.880	86.7	46.3
CT40 keV +70 keV + λ70 keV + Zeff + IC	75.6	0.350	0.582	68.8	66.3

AP arterial phase, AUC area under cure, EGFR epidermal growth factor receptor, EGFR⁺ EGFR mutation, EGFR⁻ EGFR wild type, IC iodine concentration, VP venous phase, YI Youden index, Zeff effective atomic number

Only the area under the curve (AUC) of CT_40+70 keV + λ_{70 keV} + Zeff + IC (0.76) was greater than 0.75 during AP for distinguishing between patients with EGFR mutations and wild-type EGFR, while the sensitivity and specificity were 68.0% and 70.0%, respectively (Table 3; Fig. 5a). During the VP, only CT_40+70 keV + λ_70 keV + Zeff + IC (0.756) were greater than 0.75, while the sensitivity and specificity were 68.8% and 66.3%, respectively (Table 3; Fig. 5b).

Fig. 5.

Receiver-operating characteristic (ROC) curves for all parameters. ROC curves for the computed tomography (CT) value at 40 keV and 70 keV, λ _HU, iodine concentration (IC), effective atomic number (Zeff) and combination of all that parameters (for distinguishing between EGFR⁺ and EGFR⁻) in the arterial phase (AP) (a), venous phase (VP) (b). EGFR: epidermal growth factor receptor, EGFR⁺: EGFR mutation, EGFR⁻: EGFR wild type

Discussion

In this study, we demonstrate that quantitative parameters of dual-energy spectral CT have potential value for identifying the EGFR-mutation status.

We used ARMS technology to detect EGFR mutations in patients with lung cancer. The incidence of EGFR mutations in Asian populations with lung adenocarcinoma is approximately 40–50% (Shi et al. 2014). In this study, the incidence of EGFR mutations was 61.5% (128/208), which was higher than that reported by previous studies. This may be attributed to our small sample size. We found that EGFR mutations were more common in non-smokers and women, which was consistent with previous reports (Choi et al. 2015, 2014; Lee et al. 2013; Liu et al. 2016b; Sekine et al. 2013; Song et al. 2013; Suh et al. 2018; Sun et al. 2012; Usui et al. 2011; Zhang et al. 2012).

Our results showed that the quantitative parameters of spectral CT (CT_40keV value, CT_70keV value, λ_70keV, Zeff, and IC) were significantly correlated with EGFR mutations, which is consistent with previous reports (Li et al. 2019a, 2019b). The quantitative parameters of the EGFR-mutation group were significantly higher than those of the EGFR wild-type group during the AP and VP, while WC did not differ significant between the two groups. CT values, λ _HU, and IC are quantitative parameters that reflect the blood supply of lung tumors, while Zeff reflects the effective atomic number of inorganic substances within the tumor. Our results may be related to the increased blood supply to lung adenocarcinoma caused by EGFR mutations. Iodine is the main component of the CT contrast agent. The iodine concentration measured by spectral CT reflects the degree of enhancement of the region of interest, which may have the potential to reflect the relative vascular distribution in the region of interest (Knöss et al. 2011). EGFR plays an important role in tumorigenesis. Overexpression or abnormal expression of EGFR in tumor cells is associated with tumor cell proliferation, apoptosis, tumor invasion, metastasis, and angiogenesis (Bruns et al. 2000). EGFR can cause the expression of vascular growth factors (such as VEGF and bFGF) to be up-regulated, which indirectly stimulates the formation of new blood vessels in tumors (Klapper et al. 2000). The increased blood supply to lung adenocarcinomas in patients with EGFR mutations may be related to the stimulation of angiogenesis by EGFR, and this increase may be reflected in the quantitative parameters of spectral CT.

ROC curve analysis showed that the AUC of the combination of all the quantitative parameters was larger than that obtained for each individual statistically significantly CT parameter (for the comparison between the EGFR-mutation and EGFR wilt-type groups) (Table 3; Fig. 5). The AUC values in the AP and VP were 76.0% and 75.6%, respectively. Moreover, CT_{40 keV} also showed large AUC values (72.8% and 72.1%, respectively). However, the thresholds evaluated in this study were based on this cohorts, and their accuracy needs to be further verified in a larger sample size. These findings suggest the potential value of the quantitative parameters of spectral CT for identifying the EGFR-mutation status.

This study has several limitations. First, the study population was relatively small, this is related to the low incidence of EGFR mutations. Therefore, larger cohorts are needed for verifying our results in future studies. Second, our study was limited to lung adenocarcinoma and did not include other histological subtypes, since most of the EGFR mutations occur in adenocarcinomas.

Conclusion

Our results indicate a correlation between the quantitative parameters of spectral CT and EGFR mutation in lung adenocarcinoma. This study provides more objective and detailed information from a new perspective, thereby contributing to the treatment options and prognosis of lung cancer.

Author contributions

GJZ and YTC contributed equally to this study. All authors had access to the study data. All authors read and approved the final manuscript.

Funding

This study has received funding from Talent Innovation and Entrepreneurship Project of Lanzhou (2016-RC-58). Open Fund project of Key Laboratory of Medical Imaging of Gansu Province (GSYX202010).

Compliance with ethical standards

Conflict of interest

The authors declares that there is no conflict of interest.

Ethical approval

The Ethics Committee of the Second Hospital of Lanzhou University (Lanzhou, China) approved the use of patient materials in this study.