The predictive value of 18F-FDG PET/CT in an EGFR-mutated lung adenocarcinoma population.

Background: A non-invasive, simple, and convenient method to evaluate the presence of epidermal growth factor receptor (EGFR) mutations is important for initial treatment decisions in lung adenocarcinoma.
Methods: We retrospectively reviewed 297 untreated primary lung adenocarcinoma patients with exact EGFR status. Based on their EGFR status, the patients were divided into a mutant-type group (138 patients) and wild-type group (159 patients). General patient characteristics and possible factors reflecting the status of EGFR were also evaluated.
Results: Of the 297 lung adenocarcinoma patients analyzed for EGFR status who underwent positron emission tomography (PET)/computed tomography (CT) between January 2013 and December 2017, mutations in the EGFR gene were detected in 138 patients (46.5%). EGFR mutations were more frequently associated with women, never smokers, and low 18F-fluoro-2-deoxy-glucose (18F-FDG) PET/CT maximal standard uptake value of the primary tumor (pSUVmax). Multivariate analysis indicated that women [odds ratio (OR) =2.853; 95% confidence interval (CI): 1.451-5.611; P=0.002], never smokers (OR =2.414; 95% CI: 1.217-4.789; P=0.012), tumor size <3.5 cm (OR, 2.170; 95% CI: 1.205-3.908; P=0.010), and pSUVmax <8.2 (OR =1.904; 95% CI: 1.098-3.302; P=0.022) were effective predictors of EGFR mutation. In addition, the area under the curve (AUC) of pSUVmax and tumor size was 0.623 and 0.600, respectively. Combined with clinical characteristics, including sex and smoking status, the AUC of the 4 predictors was 0.770. Conclusions: These indicators could be helpful for enhancing predictive accuracy of EGFR mutations in lung adenocarcinoma patients, especially in those for whom EGFR detection is unavailable. 2022 Translational Cancer Research. All rights reserved.

Introduction

The introduction of novel therapies, especially the use of tyrosine kinase inhibitors (TKIs), has improved patient outcomes and enabled a significant paradigm shift in the therapeutic agent management of non-small cell lung cancer (NSCLC) (1-5). Numerous randomised clinical trials have shown that mutations in epidermal growth factor receptor (EGFR) are frequently present in lung adenocarcinoma. While progression-free survival (PFS) is longer when using TKIs than when using chemotherapy in patients harbouring EGFR mutations (6-8), the effectiveness of TKIs is affected by the presence of EGFR mutations (9,10). Therefore, molecular profiling of EGFR or anaplastic lymphoma kinase (ALK) has been recommended as standard care for patients with advanced-stage adenocarcinoma regardless of race, sex, smoking history, or other clinical factors (1). However, obtaining sufficiently high-quality tumor tissues for EGFR testing is challenging in many situations due to a deficiency of biopsy samples and the physical condition of patients. Therefore, a non-invasive, simple, and convenient method to evaluate the presence of EGFR mutations is necessary when managing treatment strategies, especially in making initial treatment decisions.

Positron emission tomography (PET) using 18F-fluoro-2-deoxyglucose (18F-FDG) is a widely applied, non-invasive, functional imaging modality based on differential glucose metabolism and is important for diagnosis, staging, assessment of therapeutic efficacy, and radiotherapy planning in lung cancer (11). High rates of 18F-FDG uptake illustrate both active glucose metabolism and cancer proliferation of malignant cells (12,13). The EGFR signalling pathway regulates glucose metabolism and inhibits apoptosis in EGFR-mutated lung adenocarcinoma cells, and TKIs act to decrease glucose consumption (14,15). Studies have shown that maximal standard uptake values (SUVmax) are associated with EGFR mutational status in both cellular and animal models (16,17). Clinical studies have demonstrated an association between EGFR mutations and 18F-FDG uptake, but the results are conflicting (18-25). We retrospectively analyzed patients with lung adenocarcinoma to investigate whether 18F-FDG PET could be an accurate indicator for predicting EGFR mutations in lung adenocarcinomas. We present the following article in accordance with the STARD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-22-1726/rc).

Methods

The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by committee of Shanxi Province Cancer Hospital (No. 201835). Individual consent for this retrospective analysis was waived.

Study design and inclusion criteria

We retrospectively examined all untreated primary lung adenocarcinoma patients who had undergone EGFR detection and PET/CT simultaneously between January 2013 and December 2017. In total, 432 cases without diseases which influence the standard uptake value (such as diabetes, acute inflammation, and metabolic disease, etc.) were identified, and 297 individuals were ultimately selected. Of these, 135 cases were excluded for one of the following reasons: (I) 73 patients had a history of another malignancy or mixed pathological types; (II) 54 cases had incomplete data; (III) the period between PET/CT and EGFR detection exceeded 1 month in 8 individuals.

The parameters of 18F-FDG PET/CT and clinical information, including age, sex, smoking history, immunohistochemistry (IHC) analysis, EGFR status, and tumor/node/metastasis (TNM) stage were obtained from the hospital’s medical records. Patients who never smoked or had smoked no more than 100 cigarettes in total were defined as never smokers. The patients were divided into a mutant group (n=138) and a wild-type group (n=159) based on EGFR mutational status. TNM staging was based on the International Association for the Study of Lung Cancer (IASLC) 8th TNM Lung Cancer Staging System and clinical testing results. All procedures were performed by at least 2 experienced clinicians according to international guidelines.

18F-FDG PET/CT scanning and analysis

A Discovery STE PET/CT system (GE Healthcare, Waukesha, WI, USA) was used to perform PET/CT. The PET/CT center of Shanxi Cancer Hospital provided 18F-FDG with a radiochemical purity of >95%. Patients fasted for >6 h before examination. The fasting blood glucose concentration was checked and confirmed to be <11 mmol/L before examination. Image scanning was performed 60 min after intravenous injection of 18F-FDG (0.12–0.15 mCi/kg). PET was performed in the 3-dimensional mode using 3.75 mm per slice. CT was performed using the following parameters: 120 kV, 200 mA, 0.8 s/lap, and 22.5 mm/s bed speed. The images were immediately obtained from the top of the skull to the upper femur (6–8 bed positions and 3 min per bed position). The PET images were reconstructed and attenuation-corrected by CT images. The images were then observed on an Xeleris Workstation (GE Healthcare) for assessment. Two experienced nuclear medicine radiologists reviewed all PET/CT data independently and reached a consensus on the results. The reviewers were blinded to the EGFR mutational status.

EGFR mutation analysis

EGFR mutations were detected according to the amplification refractory mutation system (ARMS). Four paraffin-embedded wax slices with a continuous 7-µm thickness were taken to extract DNA using the QIAamp DNA FFPE Kit (Qiagen NV, Venlo, The Netherlands) according to the manufacturer’s instructions. Polymerase chain reaction was performed to detect EGFR mutations using the EGFR 29 Mutation Detection Kit (Amoy Diagnostics, Xiamen, People’s Republic of China). All steps were completed according to the ADx-EGFR detection instructions. The results of EGFR mutation analyses were interpreted by at least 2 experienced pathologists.

IHC analysis

Tissue samples were acquired through surgical resection or biopsy of the primary tumor, primary or distant lymph node, or local or distant metastases. The samples were fixed in 10% neutral formalin and embedded in paraffin wax. Slices 3-µm thick were taken, one of which was used for hematoxylin and eosin (HE staining), and the rest were used for IHC staining. The process was performed using the fully automated Ventana IHC assay. Positive and negative controls were included in each experiment. The results of pathological diagnoses and IHC tests were interpreted by at least 2 experienced pathologists as being positive for a brown or tan nuclear appearance. Positive results for thyroid transcription factor-1 (TTF-1) were defined as >5% of tumor cells with nuclear staining and positive expression of napsin A. Cytokeratin 7 (CK7) positivity was defined as >50% of tumor cells with nuclear staining. The percentage of nuclei staining positive for Ki67 in tumor cells was denoted with a Ki67 score.

Statistical analysis

All data were analyzed using SPSS for Windows (version 22.0). Clinical characteristics, including age, sex, smoking history, IHC results, and staging, were compared according to the EGFR mutational status using the chi-squared test and the t-test. A significant difference was defined as a two-sided P value <0.05. Receiver operating characteristic (ROC) curves were plotted to acquire the cut-off value of maximal standard uptake value of primary tumor (pSUVmax) for predicting EGFR mutations. Logistic regression analysis was used to identify independent predictors of EGFR mutation.

All variables with P<0.05 in the univariate analysis were further analyzed by multivariate regression analysis. Indicators with P<0.05 in the multivariate analysis were considered to be independent predictors, and the odds ratios (OR) and 95% confidence intervals (CI) of the independent predictors were obtained. In addition, ROC curves and area under the curves (AUCs) were plotted for the combined predictors of EGFR mutational status.

Results

Patients and tumor characteristics

Of the 297 lung adenocarcinoma patients analyzed for EGFR status who underwent PET/CT between January 2013 and December 2017, 138 (46.5%) were identified as having mutant EGFR, and 159 participants were wild type. The basic characteristics of the 2 groups are summarized in . The median age of the 297 patients was 59.27 years (range, 30–88 years), and there were 131 women (44.1%) and 166 men (55.9%). In total, 168 (76.1%) were never smokers. In addition, 66 (22.2%), 31 (10.4%), 59 (19.9), and 141 (47.5%) patients had stage I, II, III, and IV disease, respectively. The mutant group contained 69 cases (50.0%) with an exon 19 mutation, 62 patients (44.9%) with an exon 21 mutation, and 7 cases (5.1%) with rare mutations (). The median pSUVmax was 9.52 (range, 0.67–39.20).

Table 1

Associations between the 2 groups

Variables	EGFR		Total (N=297)	P value
	Mutant (N=138)	Wild-type (N=159)
Age (years), mean ± SD [range]	58.95±9.87 [31–83]	59.54±10.52 [30–88]	59.27±10.21 [30–88]	0.619a
Gender, n (%)				<0.001b
Male	48 (34.8)	118 (74.2)	166 (55.9)
Female	90 (65.2)	41 (25.8)	131 (44.1)
Smoking status, n (%)				<0.001b
Never smoker	105 (76.1)	63 (39.6)	168 (76.1)
Ever smoker	33 (23.9)	96 (60.4)	129 (43.4)
pSUVmax, mean ± SD	8.24±4.51	10.64±5.77	9.53±5.35	<0.001a
nSUVmax, mean ± SD	5.64±3.77	7.25±4.53	6.58±4.29	0.009a
mSUVmax, mean ± SD	7.20±3.86	7.32±4.72	7.27±4.37	0.870a
Tumor size, mean ± SD	2.89±1.46	3.54±1.90	3.24±1.74	0.001a
Nodal metastasis, n (%)				0.141b
Negative	51 (37.0)	46 (28.9)	97 (32.7)
Positive	113 (71.7)	87 (63.0)	200 (67.3)
Distant metastasis, n (%)				0.293b
Negative	77 (55.8)	79 (49.7)	156 (52.5)
Positive	61 (44.2)	80 (50.3)	141 (47.5)
Stage, n (%)				0.138b
I	38 (27.5)	28 (17.6)	66 (22.2)
II	16 (11.6)	15 (9.4)	31 (10.4)
III	23 (16.7)	36 (22.6)	59 (19.9)
IV	61 (44.2)	80 (50.3)	141 (47.5)
TTF-1, n (%)				0.001b
Negative	2 (4.0)	24 (24.5)	26 (17.6)
Positive	48 (96.0)	74 (75.5)	122 (82.4)
Napsin A, n (%)				<0.001b
Negative	2 (4.2)	33 (37.5)	35 (25.7)
Positive	46 (95.8)	55 (82.5)	101 (74.3)
CK7 (n, %)				0.548b
Negative	0 (0.0)	3 (4.9)	3 (3.3)
Positive	30 (100.0)	58 (95.1)	88 (96.7)
Ki67 score, mean	35.95±19.07	49.08±23.96	44.60±23.19	0.003a
EGFR mutations, n (%)
Exon 18	3 (2.2)
Exon 19	69 (50.0)
Exon 20	4 (2.9)
Exon 21	62 (44.9)

a, t-test calculation; b, Chi-Square calculation. EGFR, epidermal growth factor receptor; SD, standard deviation; pSUVmax, maximal standard uptake value of primary tumor; nSUVmax, maximal standard uptake value of lymph node; mSUVmax, maximal standard uptake value of distant metastasis; TTF-1, thyroid transcription factor 1.

Association between EGFR mutations and clinical characteristics

To analyze the association between EGFR mutations and clinical information, the characteristics of the 2 groups were compared. EGFR mutations were detected more frequently in women (65.2% vs. 34.8%, P>0.001), never smokers (76.1% vs. 23.9%, P<0.001), and those with positive expression of TTF-1 (96.0% vs. 4.0%, P=0.001) and napsin A (95.8% vs. 4.2%, P<0.001). In addition, the Ki67 scores were lower in the EGFR-mutant group (35.95±19.07) than in the wild-type group (49.08±23.96, P=0.003; ). The means of the pSUVmax (8.24±4.51 vs. 10.64±5.77; P<0.001) and lymph node SUVmax (5.64±3.77 vs. 7.25±4.53; P=0.009) of PET/CT were lower, respectively, in the mutant group compared to the wild-type group. The mean tumor size of the primary mass (2.89±1.46 vs. 3.54±1.90; P=0.001) was smaller in the mutant group than that in the wild-type group. There were no statistical differences in the parameters of PET/CT among the EGFR-mutation subtypes.

A ROC curve was plotted to calculate the cut-off point (8.20) of the pSUVmax with 64.8% sensitivity, 55.8% specificity, a 62.8% positive predictive value, a 57.9% negative predictive value, and 60.6% accuracy; the AUC was 0.623 (95% CI: 0.560–0.686; P<0.001). The cut-off point for tumor size was 3.5 cm with 46.5% sensitivity, 71.2% specificity, 71.2% positive predictive value, 56.0% negative predictive value, and 61.3% accuracy; the AUC was 0.600 (95% CI: 0.536–0.665; P=0.003). Thus, even though EGFR mutations were detected more frequently in individuals with a pSUVmax <8.2 (P<0.001) and tumor size <3.5 cm, the 2 indicators had a marginal predictive value for EGFR mutations.

Prediction of EGFR mutations

To confirm the predictive factors of EGFR mutations, univariate and multivariate logistic regression analyses were used. Univariate regression analysis indicated that sex, smoking status, tumor size, pSUVmax, and stage were correlated with EGFR mutations. Multivariate regression indicated that sex, smoking status, tumor size, and pSUVmax remained independent factors for predicting EGFR mutation. Women (OR =2.853; 95% CI: 1.451–5.611; P=0.002), never smokers (OR =2.414; 95% CI: 1.217–4.789; P=0.012), tumor size <3.5 cm (OR =2.170; 95% CI: 1.205–3.908; P=0.010), and pSUVmax <8.2 (OR =1.904; 95% CI: 1.098–3.302; P=0.022) were effective predictors of EGFR mutation (). In addition, the AUCs of the pSUVmax and tumor size were 0.623 and 0.600, respectively. When using pSUVmax combined with tumor size to calculate the ROC curve, the AUC was 0.656. Combined with clinical characteristics, including sex and smoking status, the AUC of the 4 predictors increased to 0.770 ().

Table 2

Univariate and multivariate analysis of various predictive indicators for EGFR status

Characteristics	Univariate analysis			Multivariate analysis
	OR (95% CI)	P value		OR (95% CI)	P value
Age		0.230
<60 years	Reference
≥60 years	1.323 (0.838–2.090)
Sex		<0.001			0.002
Male	Reference			Reference
Female	5.396 (3.276–8.888)			2.853 (1.451–5.611)
Smoking status		<0.001			0.012
Never	4.848 (2.929–8.026)			2.414 (1.217–4.789)
Ever	Reference			Reference
Tumor size		<0.001			0.010
<3.5 cm	3.314 (1.881–5.222)			2.170 (1.205–3.908)
≥3.5 cm	Reference			Reference
pSUVmax		<0.001			0.022
<8.2	2.322 (1.454–3.706)			1.904 (1.098–3.302)
≥8.2	Reference			Reference
Nodal metastasis		0.142
Negative	Reference
Positive	0.694 (0.427–1.130)
Distant metastasis		0.293
Negative	Reference
Positive	0.782 (0.495–1.236)
Stage		0.046
I	Reference
II	0.786 (0.334–1.852)	0.582
III	0.471 (0.230–0.963)	0.039
IV	0.562 (0.311–1.015)	0.056

EGFR, epidermal growth factor receptor; OR, odds ratio; CI, confidence interval; pSUVmax, maximal standard uptake value of primary tumor.

Figure 1

ROC curves of EGFR mutation. pSUVmax gained 0.623 (continuous line) of AUC, and the value increased to 0.770 when combined with all predictive factors (gender, smoking status, tumor size, and pSUVmax, dotted line). P value <0.05. AUC, area under curve; ROC, receiver operating characteristic; EGFR, epidermal growth factor receptor; pSUVmax, maximal standard uptake value of primary tumor.

Discussion

TKIs have played an important role in improving the outcomes of a particular group of patients and enabled a significant paradigm shift in the therapeutic agent management of NSCLC (1-5). EGFR mutations are mostly found in lung adenocarcinoma, which is the predominant subtype of lung cancer (24). Nevertheless, due to the deficiency of biopsy samples and the physical condition of patients, obtaining sufficient high-quality tumor tissues for EGFR testing remains challenging in many situations. Therefore, a noninvasive, simple, and convenient method to detect the presence of EGFR mutations is needed for managing treatment strategies, especially when making initial treatment decisions. In this study, we showed that lung adenocarcinoma patients harboring EGFR mutations had lower pSUVmax and smaller tumor size than wild-type cases according to 18F-FDG PET/CT. As a noninvasive method, pSUVmax and tumor size based on PET/CT together with clinical characteristics such as gender and smoking status could enhance EGFR status discriminability in those patients for whom EGFR detection is not feasible.

Previous studies have revealed a relationship between EGFR mutations and 18F-FDG uptake, but the results have been conflicting (18-29). The majority of studies demonstrated that a lower pSUVmax was an indicator for predicting EGFR mutations in either lung adenocarcinomas or NSCLC, regardless of tumor stage (18-25). The present study also showed a lower pSUVmax was an independent indicator in lung adenocarcinoma. Our results were consistent with the findings of multiple studies (18-25) but contradicted with 2 studies (26,27). Notably, one of the 2 studies that showed a higher pSUVmax was a predictor of EGFR mutations involved 132 advanced adenocarcinoma (ADC) cases regardless of tumor stage (26), while the other study included 77 stage III and IV ADC patients (27). Further, retrospective analyses conducted by Lee et al. (28) and Caicedo et al. (29) found there was no relationship between pSUVmax and EGFR mutations, which also differed from our results. Further studies of ADC patients in stage IV demonstrated that there was no difference in pSUVmax between EGFR-mutant patients and wild-type patients (21,28).

Several possible factors may have contributed to these contradictory results. First, these studies involved either NSCLC or lung adenocarcinomas with different tumor stages. Thus, the histological type and tumor stage of the included cases could have influenced the results. Second, of the 2 studies showing results that were contrary to ours (26,27), one included 132 ADC cases regardless of tumor stage and also required pretreatment serum indicator and CT assessments (26). The other study, which involved 77 cases of stage III and IV ADC patients, reported an EGFR-mutation rate of 64% (27). The pretreatment requirement, small sample size, and high EGFR mutation rate of both studies may reflect the bias in patient selection. Moreover, although pSUVmax was significantly related to EGFR mutation in univariate regression analysis, no relationship was found in multivariate regression analysis in the study conducted by Lee et al. (28). In our study, despite the fact that pSUVmax showed predictive value in both univariate and multivariate regression analysis, the AUC was only 0.623, which was consistent with the results of Lv et al. (AUC =0.557) (21). Hence, the marginal discriminating ability of pSUVmax could have led to discrepancy among results.

The results of tumor size as a predictive indicator for EGFR mutations were also conflicting and there are several possible reasons for this (21,28,30). The studies performed by Lee et al. (28) and Guan et al. (30) showed a smaller tumor size could predict EGFR mutations in both univariate and multivariate regression analysis, which was consistent with our study. The results from Lv et al. (21) revealed no relationship between tumor size and EGFR status in a separate analysis of an ADC-only group. However, tumor size was smaller in EGFR-mutant individuals than that in wild-type patients when ADC and non-ADC cases were analyzed as a whole in Lv’s study. In our study, the rate of EGFR mutations was 46.5%, and a smaller tumor size was weakly associated with EGFR mutations (AUC =0.600). Further, due to the limited availability of testing technology, the detection rate of EGFR mutation was only 9.6% in China about a decade ago (31). One large-sample survey reported the EGFR-mutation rate was 42.5% in a noninterventional, real-world study of IIIB/IV NSCLC patients in northern China (32). However, patients with early-stage disease and a sound economic background were more likely to undergo genetic detection. Therefore, the modest discriminating power of tumor size and genetic-testing patient selection bias may have affected the results.

Although detecting EGFR mutations from tissue is the gold standard for therapeutic agent management using EGFR-TKIs for first-line treatment, sufficient tissue for EGFR-mutation detection is not available in many situations. Previous randomized head-to-head data (33,34) have demonstrated the potential benefit of EGFR-TKIs in patients who did not undergo EGFR-status detection and were selected based on clinical characteristics only. More and more diagnoses of NSCLC rely on cytologic specimens or small biopsies which result in insufficient quality and quantity of tumor cells for EGFR detection. A retrospective study including 11 Asian Pacific countries showed 71.4% (53.8% in China) of samples used for detecting EGFR mutations were cytology samples and/or small biopsies (35). Overall failure rates for EGFR testing using cytology samples and/or small biopsies have been reported to be 5–30% (36,37). Therefore, EGFR-mutation detection from tissue rather than cytology samples and/or small biopsies is highly recommended before first-line treatment with EGFR-TKIs. In addition to the limitations of clinical resources and testing technology mentioned above, the subjective willingness of patients also restrict invasive examination and obtainment of sufficient high-quality tumor tissues for EGFR testing.

While EGFR-mutation testing from tissue is important, noninvasive methods to predict EGFR mutations is also necessary when tissue detection is not available in order to identify patients likely to benefit from EGFR-TKIs. This study aimed to investigate the predictive value of pSUVmax for EGFR mutations in those patients for whom EGFR testing was not feasible. The results showed that a lower pSUVmax and a smaller tumor size could be helpful for distinguishing EGFR mutations from wild type. Although both indicators had marginal predictive value for EGFR mutations, they could be combined with other clinical characteristics to enhance patient stratification in cases where EGFR detection is not feasible. Reasonable application of clinical factors and parameters of PET/CT could be used to manage therapeutic strategies in lung adenocarcinomas without available EGFR detection.

Our study had several limitations. Above all, the retrospective nature of the study may have led to sample availability bias and patient selection bias. The nature of a single-center study can also induce bias in the results. Thus, larger samples and multicenter prospective studies should be performed to validate the results. In addition, maximal standard uptake value of lymph node (nSUVmax) and maximal standard uptake value of distant metastasis (mSUVmax) were not included in this study for patients without lymph node and distant metastasis. Hence, individuals with a similar background (stage of disease, race, gender, smoking status, etc.) should be compared to verify the results. Finally, the study did not detect all mutations, and several drive mutations may exist in a patient concurrently. Each driver mutation may influence the glucose metabolism process through different signaling pathway activations. Therefore, more comprehensive genetic detection is needed to confirm the results.

Conclusions

This study aimed to reveal whether pSUVmax could be a noninvasive indicator to predict EGFR mutations in untreated primary lung adenocarcinomas. The results identified that pSUVmax <8.2 was an independent predictor of EGFR mutations, and the AUC was 0.629. When using all the predictors, including female, never smoker, tumor size <3.5 cm, and pSUVmax <8.2, the AUC increased to 0.770. Other biomarkers could be integrated with the parameters of PET/CT to predict EGFR status and efficiency of TKI treatment in the future in cases where genetic detection is not feasible.

The article’s supplementary files as