Usefulness of cumulative smoking dose for identifying the EGFR mutation and patients with non‐small‐cell lung cancer for gefitinib treatment

Abstract

We examined the diagnostic accuracy of the cumulative smoking dose for identifying the epidermal growth factor receptor (EGFR) exon 19 deletion and L858R mutation among Japanese patients with non‐small‐cell lung cancer (NSCLC). EGFR mutations in exon 19 and exon 21 were determined in 1001 NSCLC patients. A receiver–operating characteristic (ROC) curve methodology was applied to estimate the diagnostic accuracy. EGFR mutations were detected in 314 patients (31.4%). A cumulative smoking dose of less than 13 pack‐years (PY) was the optimal cut‐off point for predicting a positive EGFR mutation status, producing a balance between the sensitivity (73.5%) and the specificity (77%). The area under the ROC curve was 0.77, indicating that the smoking dose had a moderate diagnostic accuracy. The median survival time or the median progression‐free survival time of patients who had smoked less than 13 pack‐years (PY) were 18.6 and 6.3 months, respectively, while those of patients with equal to or more than 13 PY were 9.6 and 2.4 months, respectively. The overall survival (OS) and progression‐free survival (PFS) rates were significantly different between the two groups (OS: hazard ratio [HR] = 0.64, 95% confidence interval [CI] = 0.51–0.80, P = 0.0001) (PFS: HR = 0.58, 95% CI = 0.47–0.71, P < 0.0001). Our study indicated that the smoking dose predicted EGFR mutations with a moderate diagnostic accuracy. Thus, patients who have smoked less than 13 PY might be candidates for gefitinib treatment when EGFR mutation status cannot be determined. (Cancer Sci 2009; 100: 1931–1934)

Epidermal growth factor receptor (EGFR) is a receptor tyrosine kinase that is highly expressed in cancer cells.^{( 1 )} Mutations in EGFR have been reported in non‐small‐cell lung cancers (NSCLC).^{( 2} , ³ , ^{4 )} EGFR mutations are frequently located in exons 18 to 21 of the EGFR tyrosine kinase domain and play an oncogenic role, especially in adenocarcinoma.^{( 5 )} Gefitinib and erlotinib are reversible EGFR tyrosine inhibitors (EGFR‐TKI) and are used for the treatment of NSCLC patients.^{( 6 )} Previous studies have focused on identifying factors that are useful indicators for selecting candidates for EGFR‐TKI treatment. An adenocarcinoma histology, a never‐smoking status, and the female sex have been shown to be associated with sensitivity to gefitinib. Since 2004, EGFR mutations, especially exon 19 deletions and the L858R mutation, have been demonstrated to be associated with sensitivity to EGFR‐TKIs and are considered to predict a favorable clinical outcome for NSCLC patients treated with EGFR‐TKIs.^{( 7} , ^{8 )} Based on accumulating data, an examination of the EGFR mutation status prior to the start of EGFR‐TKI treatment is now encouraged. However, there are some situations when a mutation analysis is not feasible, such as, for instance, when the available clinical samples are inappropriate for determining the EGFR genotype. Previous studies have shown that smoking status, adenocarcinoma histology, and East Asian ethnicity were significantly related to EGFR mutations.^{( 4 )} In addition, we and others have reported that the presence of EGFR mutations is inversely correlated with the cumulative smoking dose, suggesting that the degree of smoking predicts the prevalence of EGFR mutations.^{( 9} , ¹⁰ , ^{11 )} Pham et al. reported that a smoking history could be a predictor of EGFR mutation based on a study using a receiver–operating characteristic (ROC) curve methodology, suggesting that the total smoking dose could assist clinicians in assessing the likelihood of EGFR mutations.^{( 11 )} Because the frequency of EGFR mutation differs according to ethnicity, the relationship between EGFR mutation and smoking dose is an issue of interest in East Asian patients, including Japanese patients, in countries where gefitinib and erlotinib have been approved for the treatment of NSCLC.

In the present study, we determined the diagnostic accuracy of the cumulative smoking dose for identifying the EGFR mutation status of Japanese NSCLC patients. We also showed the survival of patients according to the smoking dose, as determined using an ROC curve methodology.

Materials and Methods

Patients’ characteristics.  We collected the data of 1001 NSCLC patients undergoing surgical procedures between 2000 to 2008 that contained clinical records and EGFR mutation status from Aichi Cancer Center Hospital (410 patients), Nagoya, Japan, and Okayama University Hospital (591 patients), Okayama, Japan. Total patients consisted of 365 (36.5%) females and 636 (63.5%) males, 375 (37.5%) never‐smokers and 626 (62.5%) ever‐smokers (median, 43.7 pack‐years [PY]; range, 0.15–324 PY), 779 (77.8%) adenocarcinomas and 222 (22.2%) non‐adenocarcinomas. Smoking categories were defined as follows: never‐smokers were those with lifetime exposure of 100 cigarettes or less, ever‐smokers were those with lifetime exposure of more than 100 cigarettes. Patients’ characteristics are shown in Table 1. We previously reported the survival of 408 gefitinib‐treated patients with clinical records and EGFR mutation status collected from four institutions.^{( 12 )} Briefly, 408 NSCLC patients consisting of 362 adenocarcinomas and 46 non‐adenocarcinomas were obtained form from the National Cancer Center Hospital (207 patients), Aichi Cancer Center Hospital (103 patients), and Okayama University Hospital with NHO Yamaguchi–Ube Medical Center (98 patients). All patients had advanced or recurrent NSCLC and initiated gefitinib treatment (250 mg/day) between November 2000 and August 2006 in each institution.^{( 12 )} We used this cohort to investigate the clinical outcome of patients treated with gefitinib.

Table 1.

Patients’ characteristics and relevance of EGFR mutation

Variables	Number	EGFR mutation (%)	P‐values
Sex
Female	365	198 (54.2)	<0.0001
Male	636	116 (18.2)
Smoking history
Never	375	223 (59.5)	<0.0001
Ever	626	91 (14.5)
Histology
Ad	779	308 (39.5)	<0.0001
Non‐Ad	222	6 (2.7)

Ad, adenocarcinoma; EGFR, epidermal growth factor receptor.

Tumor response was assessed based on World Health Organization criteria (National Cancer Center Hospital and Okayama University Hospital with Sanyo National Hospital)^{( 8} , ¹³ , ^{14 )} and image analysis and serum carcinoembryonic antigen level as reported (Aichi Cancer Center Hospital).^{( 7 )} The overall survival (OS) and progression‐free survival (PFS) were calculated from the start of gefitinib treatment until the date of death or the last follow‐up for OS and until confirmed disease progression or death for PFS. The Kaplan–Meier method was applied to estimate OS and PFS. This study was permitted by the Institutional Review Board at each institution and informed consents were obtained from each patient.

Detection of EGFR mutations in primary tumors.  The DNA‐based analysis using direct‐sequencing or PCR‐based length polymorphisms (exon 19) or RFLP (exon 21) assays were performed to detect EGFR mutation in samples from Okayama University.^{( 15 )} The RNA‐based analysis using one‐step reverse transcription–polymerase chain reaction for EGFR mutation detection was carried out at Aichi Cancer Center.^{( 9 )}

Receiver–operating characteristic (ROC) curve analysis for prediction of the EGFR mutation.  We used ROC analysis to determine the cut‐off point for the smoking dose at which optimal sensitivity and specificity were achieved, maximizing accuracy. The best cut‐off point for balancing the sensitivity and specificity of a test was assumed to be the point on the curve closest to the (0, 1) point.^{( 16 )} The diagnostic accuracy of smoking dose for predicting the incidence of EGFR mutations was summarized as the area under the curve (AUC). The AUC greater than 0.9 has high accuracy, while 0.7–0.9 indicates moderate accuracy; 0.5–0.7, low accuracy; and 0.5 a toss‐up.^{( 17 )}

Statistical analyses.  The differences of significance among categorized groups were compared using the χ²‐test. Differences in OS and PFS among groups were assessed by the log‐rank test. Statistical data were analyzed with StatView 5.0 for Windows (SAS Institute, Cary, NC, USA). All statistical tests were two‐sided and P < 0.05 were defined as being statistically significant.

Results

Frequency of EGFR mutation and clinicopathological factors.  EGFR mutations were present in 314 of the 1001 patients and were comprised of 164 mutations in exon 19 and 153 mutations in exon 21. Three patients had mutations in both exons 19 and 21. The relationships between the EGFR mutation status and clinicopathological factors are shown in Table 1. EGFR mutations were significantly more frequent among females (P < 0.0001), patients with an adenocarcinoma histology (P < 0.0001), and never‐smokers (P < 0.0001). Regarding smoking status, EGFR mutations were identified in 223 (59.5%) never‐smokers and 91 (14.5%) ever‐smokers (P < 0.0001). According to the cumulative smoking dose, EGFR mutation was present in 12 (41.2%) patients with a PY of >0 and 10, 12 (28.6%) patients with a PY >10 and 20, 28 (15.5%) patients with a PY >20 and 40, 20 (10.7%) patients with a PY >40 and 60, and 19 (10.1%) patients with a PY >60 (Table 2). Overall, the incidence of EGFR mutations decreased as the number of PY increased.

Table 2.

Cumulative smoking dose and EGFR mutation

Smoking dose (pack‐years [PY])	Number	EGFR mutation (%)	P‐values*
0 (Never‐smokers)	375	223 (59.5)
0 < PY ≤ 10	29	12 (41.2)	0.070
10 < PY ≤ 20	42	12 (28.6)	0.0002
20 < PY ≤ 40	181	28 (15.5)	<0.0001
40 < PY ≤ 60	187	20 (10.7)	<0.0001
60 < PY	187	19 (10.1)	<0.0001

^* The difference was examined between pack‐year categories and never‐smokers. EGFR, epidermal growth factor receptor.

Diagnostic accuracy of smoking dose in identifying EGFR mutation status.  The smoking dose, described in terms of PY, predicted the prevalence of EGFR mutations among all the patients and among only the adenocarcinoma patients (areas under the ROC curves were 0.77 and 0.73, respectively) (Fig. 1a). According to the ROC curve for all the patients, a 12.8 PY dose level was the best cut‐off of a positive EGFR mutation status, with a 73.5% sensitivity and a 77% specificity for prediction. For convenience, we choose a dose level of 13 PY as the best cut‐off value for a positive EGFR mutation status. Among only the adenocarcinoma patients, the ROC curve indicated that a cut‐off of 11.3 PY was the best predictor for a positive EGFR mutation status (67.3% sensitivity and 77.6% specificity) (Fig. 1b). The ROC curves for specific mutation types, i.e. in exon 19 or exon 21, were also similar (data not shown). Though the frequency of EGFR mutation was higher among patients with adenocarcinomas than among all the patients with NSCLCs, the optimal cut‐off value was similar and the sensitivity and the specificity were not superior to those obtained for all the NSCLC patients.

Figure 1.

Receiver–operating characteristic (ROC) curve of the association between epidermal growth factor receptor (EGFR) mutation and cumulative smoking dose. The optimal cut‐off for the test is the point closest to the upper‐left corner of the graph, which corresponds to a mutation in EGFR. AUC, area under the ROC curve (a) ROC for total 1001 cases. (b) ROC for 779 adenocarcinoma cases.

Response and survival of patients treated with gefitinib stratified according to smoking dose.  Analyses for tumor response and survival were performed in 408 patients who were treated with gefitinib.^{( 12 )} A total of 211 patients had smoked less than 13 PY and 197 patients had smoked 13 PY or more. A total of 109 (51.7%) patients with less than 13 PY smoking history showed tumor response and 47 (23.9%) patients with equal or more than 13 PY showed response (P = 0.0001). The median survival time (MST) or the median progression‐free survival time (MPFS) of patients who smoked less than 13 PY were 18.6 and 6.3 months, respectively, while those of patients with equal to or more than 13 PY were 9.6 and 2.4 months, respectively. Significant differences in OS and PFS were observed between the two groups (OS: hazard ratio [HR] = 0.64, 95% CI = 0.51–0.80, P = 0.0001) (PFS: HR = 0.58, 95% CI = 0.47–0.71, P < 0.0001) (Fig. 2a,b). Furthermore, tumor response rate, MST, and MPFS according to smoking dose are shown in Table 3. When the patients were stratified according to EGFR mutation status, significant differences in OS and PFS were observed between the patients with EGFR mutations and the wild‐type patients, as expected (OS: HR = 0.43, 95% CI = 0.33–0.54, P < 0.0001) (PFS: HR = 0.30, 95% CI = 0.24–0.37, P < 0.0001) (Fig. 3a,b).

Figure 2.

Kaplan–Meier plot of survival times stratified by the cumulative smoking dose (13 pack‐years [PY]). (a) Overall survival of patients treated with gefitinib. (b) Progression‐free survival of patients treated gefitinib. MPFS, median progression‐free survival time; MST, median survival time. P‐values were calculated using the log‐rank test.

Table 3.

Cumulative smoking dose and clinical outcomes

Smoking dose (pack‐years [PY])	Number	Response rate (%)	P‐values*	MST (months)	P‐values*	MPFS (months)	P‐values*
0 (Never‐smokers)	178	94 (52.8%)		18.3		6.3
0 < PY < 13	33	15 (45.5%)	0.78	18.7	0.71	5.1	0.61
13 < PY < 40	92	26 (28.3%)	0.019	9.6	0.0024	2.6	0.0015
40 < PY	105	21 (20.0%)	0.0004	9.5	0.0015	2.2	< 0.0001

^* The difference was examined between pack‐year categories and never‐smokers.

MPFS, median progression‐free survival time; MST, median survival time.

Figure 3.

Kaplan–Meier plot of survival times stratified byepidermal growth factor receptor (EGFR) mutation status. (a) Overall survival of patients treated with gefitinib. (b) Progression‐free survival of patients treated gefitinib. MPFS, median progression‐free survival time; MST, median survival time; Mut, EGFR mutation; Wt, EGFR wild‐type. P‐values were calculated using the log‐rank test.

Discussion

Previous studies have indicated that an EGFR mutation, but not smoking status or sex, was a predictor of a better clinical outcome among patients treated with gefitinib.^{( 12 )} However, understanding the ability or limitation of clinical factors as a predictor of patients’ prognosis when treated with EGFR‐TKI would be useful if the EGFR mutation status were not available. In this study, we determined the utility of the cumulative smoking dose for identifying the EGFR mutation status in a Japanese cohort. Our results showed that a cumulative smoking dose of less than 13 total PY yielded the highest discriminative ability with a 73.5% sensitivity and 77% specificity for predicting the presence of EGFR mutations. The AUC was 0.77, indicating that the cumulative smoking dose had a moderate diagnostic accuracy for predicting EGFR mutation status.^{( 18 )}

For the survival analysis, we used a previously reported cohort that had demonstrated the impact of EGFR mutation.^{( 12 )} Both the OS and PFS were significantly longer among patients who had smoked less than 13 PY, compared with patients who had a smoking history of 13 PY or more. Although the OS and PFS were longer in patients with a positive EGFR mutation status than in patients with a smoking history of less than 13 PY, our results indicated that the cumulative smoking dose was a predictor of survival among patients treated with gefitinib when EGFR mutation status was unknown. It should be noted that never‐ or light‐smoking status itself could be a favorable prognostic factor of NSCLC. On this point, Hotta et al. reported that the effect of smoking dose on survival was more significant in patients with gefitinib treatment than those without treatment, indicating that smoking status was a predictive factor among patients for gefitinib treatment, rather than a prognostic factor.^{( 19 )} In addition, the tumor response rate was also better in patients with never‐ or light‐smoking history than in patients with heavy smoking history. Taken together, patient selection based on smoking dose may be useful when EGFR mutation status is not available and might be a determiner for EGFR‐TKI treatment.

Using an ROC curve methodology, Pham et al. reported that a smoking history of less than 15 PY had an 82% sensitivity and a 70% specificity for predicting the presence of EGFR mutations, with an AUC of 0.78.^{( 11 )} While there are some differences, their ROC results are similar to our data. The fact that the ROC curve and the optimal cut‐off were very similar between American and Japanese patients is interesting, since the frequency of EGFR mutation differs between these two groups. Matsuo et al. suggested that smoking itself might not cause EGFR mutation.^{( 20 )} The reason why EGFR mutation seemed to be low in smokers is that lung cancers in smokers have more chance of having other molecular alterations such as K‐ras mutation, LKB1 alteration, or DNA methylation.^{( 4} , ²¹ , ^{22 )} The effect of smoking on lung cancer in patients without an EGFR mutation might be similar between American and Japanese patients, although further investigation of this possibility is necessary. While Pham et al. reported that smoke‐free years were an effective predictor of EGFR mutation status, this type of data was not available in our cohort.

In conclusion, cumulative smoking dose predicted EGFR mutation status with a moderate diagnostic accuracy. NSCLC patients who have smoked less than 13 PY might be candidates for gefitinib treatment.

Disclosure Statement

Authors have no conflict of interest to disclose.