Can smoking duration alone replace pack-years to predict the risk of smoking-related oncogenic mutations in non-small cell lung cancer? A cross-sectional study in Japan.

OBJECTIVE: To investigate whether smoking duration alone can replace pack-years to predict the risk of oncogenic mutations in non-small cell lung cancer (NSCLC).
DESIGN: A cross-sectional study using the baseline dataset from the Japan Molecular Epidemiology for Lung Cancer Study.
SETTING: Forty-three medical institutions nationwide in Japan. PARTICIPANTS: From July 2012 to December 2013, 957 patients with newly diagnosed stage I-IIIB NSCLC who underwent surgery were enrolled, and molecular analyses were performed on 876 samples (from 441 ever-smokers and 435 never-smokers). MAIN OUTCOMES MEASURED: We calculated the area under the receiver operating characteristic curve (AUC) values using logistic regression to compare between the predictive values of smoking duration and pack-years for mutational frequencies in the v-Ki-ras2 Kirsten rat sarcoma (KRAS), tumour suppressor p53 (TP53), and epidermal growth factor receptor (EGFR) genes and for cytosine-to-adenine base substitution (C>A).
RESULTS: For predicting KRAS mutations, the AUC values for smoking duration and pack-years were 0.746 (95% CI 0.682 to 0.800) and 0.759 (95% CI 0.700 to 0.810), respectively (p=0.058). For predicting KRAS mutations in smokers, the AUC values for smoking duration and pack-years were 0.772 (95% CI 0.697 to 0.833) and 0.787 (95% CI 0.714 to 0.845), respectively (p=0.036). There were no significant differences between the AUC values for smoking duration and pack-years in terms of predicting TP53 and EGFR mutations and C>A. Pack-years was a significantly better predictor of KRAS mutations than smoking duration.
CONCLUSION: Smoking duration was not significantly different from pack-years in predicting the likelihood of smoking-related gene mutations. Given the recall bias in obtaining smoking information, smoking duration alone should be considered for further investigation as a simpler alternative to pack-years. © Author(s) (or their employer(s)) 2020. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

This study is the first to show the comparison between two indices (smoking duration vs pack-years) in predicting the risk of smoking-related oncogenic mutations in non-small cell lung cancer (NSCLC).

It focuses on the mutations in the v-Ki-ras2 Kirsten rat sarcoma (KRAS), tumour suppressor p53 (TP53) and epidermal growth factor receptor (EGFR) genes and cytosine-to-adenine base substitution (C>A), which were associated with pack-years based on the results of the Japan Molecular Epidemiology (JME) study, a prospective multicentre molecular epidemiology study.

A limitation of this study was that the JME study data were obtained from targeted sequencing and not from whole genome or whole exome sequencing.

Introduction

Lung cancer is the leading cause of cancer-related morbidity and mortality worldwide.1 Low-dose CT has been shown to be effective as a screening test for lung cancer, but optimal eligibility for screening remains undetermined.2 According to epidemiological studies, cancer development primarily occurs due to environmental factors.3 Smoking is the most assessed cause of cancer and contributes to lung cancer development. There is convincing evidence that tobacco smoking strongly increases the risk of lung cancer, with a relative risk of approximately 4.4 in men and 2.8 in women for current smokers compared with never-smokers.4

V-Ki-ras2 Kirsten rat sarcoma (KRAS) and epidermal growth factor receptor (EGFR) mutations are well documented in the pathogenesis of lung adenocarcinoma according to smoking status.5 KRAS mutations show no sex predilection but are more frequent in Caucasians than in Asians, and most patients with these mutations are former or current cigarette smokers.6 7 Unlike KRAS mutations, it has been reported that EGFR mutations are more frequently found in women, Asians, and never-smokers.8 9 Information regarding pack-years has been widely used to assess the risks of lung cancer and chronic obstructive pulmonary disease (COPD).10 11 Pack-years of smoking is calculated by multiplying the number of packs of cigarettes smoked per day by the number of years the person has smoked. The total number of base substitution mutations is positively correlated with pack-years smoked for all cancer types; based on these correlation rates, it is estimated that the approximate number of mutations accumulated in a normal cell of each tissue due to smoking a pack of cigarettes per day for a year is 150, particularly in the lungs.12 A recent study showed that smoking duration alone reportedly provides stronger risk estimates of COPD than the composite index of pack-years.13

However, it remains unknown whether smoking duration can also replace pack-years for predicting smoking-related oncogenic mutations in non-small cell lung cancer (NSCLC). We previously reported a prospective, multicentr, molecular epidemiology study, the Japan Molecular Epidemiology (JME) study, which included comprehensive smoking information based on a detailed questionnaire and the mutational profiles of 72 genes using next-generation sequencing.14 The prevalence of KRAS, tumour suppressor p53 (TP53) and EGFR mutations in this study were associated with smoking dose. In addition, cytosine-to-adenine base substitutions (C>A) were reported as the most significant smoking-related base substitution pattern.

Using the baseline dataset of this prospective cohort study, we investigated whether smoking duration alone could be an alternative index to pack-years in predicting the risk of oncogenic mutations in NSCLC.

Patients and methods

Study design and patient population

The eligibility criteria and questionnaire were previously described in the JME protocol (see online supplemental file 1).14 Patients with newly diagnosed stage I–IIIB NSCLC who underwent surgery were considered eligible. Patients with a history of chemotherapy and/or radiotherapy and patients with a history of malignancies, other than adequately treated basal cell or squamous cell skin cancer or in situ cervical cancer, were excluded. The participants were required to complete a questionnaire before surgery that was modelled after the one designed for SWOG study S042415 to assess the following parameters in detail: smoking history, occupational exposures, reproductive and hormonal risk factors, weight loss, family history of cancer, medication history and current lifestyle (diet and exercise). The S0424 was originally designed to address the association between sex and lung cancer carcinogenesis by using a detailed questionnaire and tissue specimens from smoker and never-smoker men and women with newly diagnosed stage I–III NSCLC.15

Formalin-fixed paraffin-embedded surgical tissues were sent to a central laboratory for genomic analysis and immunohistochemical staining. DNA was extracted from the samples, and quality-control assessments were performed as described previously.16 Multiplexed, targeted deep sequencing was performed on MiSeq (Illumina, San Diego, California, USA) using a TruSeq Amplicon Cancer Panel and an additional custom panel (Illumina) to evaluate the tumours. Somatic mutations in 72 cancer-associated genes and copy numbers of five cancer-associated genes were selected based on previous reports17–19 and were evaluated to cover the range of critical mutations.

Statistical analyses

The JME study followed and extended the concept of S0424 by using the similar approach with the same questionnaire that would allow for direct comparison of the data. The sample size of the JME study was adjusted with reference to the sample size of S0424.

The correlations between smoking status (ever-smoker or never-smoker) and demographic factors, such as age, sex, histology and pathological stage, were examined using the χ2 test. A logistic regression model was used for multivariate analysis. To evaluate the predictive values of pack-years and smoking duration for the above-mentioned mutations or C>A, we compared the area under the receiver operating characteristic curve (AUC) calculated using logistic regression, considering sex, age, stage and histology as covariates.

Patient and public involvement

This research was performed without patient involvement. Patients were not invited to comment on the study design and were not consulted for determining patient-relevant outcomes or interpreting the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy.

Results

Clinical characteristics

From July 2012 to December 2013, 957 patients were recruited from 43 institutions of the National Hospital Organization, and information regarding environmental factors was obtained through questionnaires. For molecular analyses, 876 samples were successfully tested for gene mutations. Overall, 622 cases involved at least one mutation, and a total of 860 mutations were detected. Clinicopathological characteristics according to smoking status are shown in table 1.

Table 1

Clinicopathological characteristics according to smoking status

Characteristic	Ever smoker		Never smoker		P value
	Number of patients	%	Number of patients	%
Age at surgery, years
Median	69		71
Range	41–89		23–92
Sex					<0.001
Male	366	83	53	12.2
Female	75	17	382	87.8
Smoking duration					<0.001
Never smoker	0	0	435	100
0<years<20	28	6.3	0	0
20≤years<40	126	28.6	0	0
40≤years	246	55.8	0	0
Pack-years					<0.001
Never smoker	0	0	435	100
0<pack-years<30	102	23.1	0	0
30≤pack-years<60	184	41.7	0	0
60≤pack-years	109	24.7	0	0
Histology					<0.001
Adenocarcinoma	265	60.1	415	95.4
Squamous cell carcinoma	135	30.6	7	1.6
Others	41	9.3	13	3
Pathological stage					<0.001
I	280	63.5	338	77.7
II	81	18.4	50	11.5
III	65	14.7	39	9
IV	15	3.4	8	1.8
Total	441	50.3	435	49.7

There were 441 ever-smokers and 435 never-smokers in the JME study. The median smoking duration and pack-years were 41 (1–65) years and 43 (1–189) in ever-smokers, respectively. The frequencies of mutations in KRAS, TP53 and EGFR were 9.4%, 26.8% and 42.5%, respectively, and C>A was observed in 12.7% of cases. The distributions of smoking duration and pack-years are shown in figure 1. In ever-smokers, the most frequent mutations were in TP53 (38.3%), EGFR (20.2%) and KRAS (13.2%), and C>A was observed in 21.1% of cases, whereas in never-smokers, EGFR (65.1%), TP53 (15.2%) and KRAS (5.5%) harboured the most frequent mutations, and C>A was observed in only 4.1% of cases.

Figure 1

Scatter diagram showing the distributions of smoking duration (longitudinal axis) and pack-years (horizontal axis). The frequency of smoking <20 cigarettes per day is higher than that of smoking ≥20 cigarettes per day.

Mutational frequencies associated with smoking duration or pack-years

We divided all cases into four groups according to smoking duration (never, light (0KRAS mutations in the never, light, middle and heavy smoking duration groups were 4.1%, 7.1%, 11.1% and 14.2%, respectively, while those in the never, light, middle and heavy pack-year groups were 4.1%, 8.8%, 13.0% and 14.7%, respectively (figure 2A). The frequencies of TP53 mutations in the never, light, middle and heavy smoking duration groups were 15.1%, 25.0%, 38.9% and 41.5%, respectively; those in the never, light, middle and heavy pack-year groups were 15.1%, 32.4%, 39.1% and 46.8%, respectively (figure 2B). The frequencies of EGFR mutations in the never, light, middle and heavy smoking duration groups were 60.9%, 53.6%, 24.6% and 13.0%, respectively; those in the never, light, middle and heavy pack-year groups were 60.9%, 32.4%, 19.0% and 9.2%, respectively (figure 2C). The frequencies of C>A in the never, light, middle and heavy smoking duration groups were 4.1%, 10.7%, 23.0% and 22.3%, respectively; those in the never, light, middle and heavy pack-year groups were 4.1%, 15.7%, 19.6% and 17.4%, respectively (figure 2D).

Figure 2

The frequency of (A) KRAS, (B) TP53 and (C) EGFR mutations and (D) C>A according to smoking duration and pack-years. As the smoking dose increased, the frequencies of KRAS and TP53 mutations increased in the smoking duration and pack-year groups, but the frequency of EGFR mutations decreased with the increase in smoking dose. As the smoking dose increased, the frequency of C>A tended to increase in the smoking duration and pack-year groups.

We examined the associations between the frequency of mutations or C>A and the smoking duration or pack-year groups using logistic regression. The frequency of KRAS and TP53 mutations or C>A in the smoking duration or pack-year groups increased significantly with an increase in smoking exposure (all groups: p<0.001). In contrast, the frequency of EGFR mutations in the smoking duration or pack-year groups decreased significantly with an increase in smoking exposure (all groups: p<0.001).

The ORs calculated using logistic regression are shown in table 2. Although the ORs for smoking duration were slightly higher than those for pack-years, no significant differences were observed.

Table 2

ORs of smoking duration and pack-years for predicting KRAS, TP53 and EGFR mutations and C>A

Mutations or C>A	Smoking index	OR (95% CI)	P value
KRAS	Duration	1.03 (1.01 to 1.04)	1.07×10−3
	Pack-years	1.01 (1.01 to 1.02)	1.14×10−3
TP53	Duration	1.02 (1.01 to 1.03)	<0.001
	Pack-years	1.01 (1.00 to 1.02)	7.31×10−3
EGFR	Duration	0.968 (0.957 to 0.978)	<0.001
	Pack-years	0.978 (0.969 to 0.987)	<0.001
C>A	Duration	1.04 (1.02 to 1.05)	<0.001
	Pack-years	1.01 (1.00 to 1.02)	6.25×10−3

Comparison of mutational frequencies between smoking duration and pack-years

To compare between the predictive values of smoking duration and pack-years for mutational frequencies, we calculated the AUC values using logistic regression (table 3). For KRAS mutations in the overall population, the AUC values for smoking duration and pack-years were 0.746 and 0.759, respectively (p=0.058), whereas for KRAS mutations in cases involving smokers, the AUC values for smoking duration and pack-years were 0.772 and 0.787, respectively (p=0.036). There were no significant differences in the AUC values of smoking duration and pack-years for TP53 and EGFR mutations and C>A. However, pack-years was a significantly better predictor of KRAS mutations than smoking duration in ever-smokers.

Table 3

The AUC values of smoking duration and pack-years for KRAS, TP53 and EGFR mutations and C>A in all cases and in smokers

Mutations or C>A	Cases	Smoking index	AUC (95% CI)	P value
KRAS	All	Duration	0.746 (0.682 to 0.800)	0.058
		Pack-years	0.759 (0.700 to 0.810)
	Smokers	Duration	0.772 (0.697 to 0.833)	0.036
		Pack-years	0.787 (0.714 to 0.845)
TP53	All	Duration	0.700 (0.658 to 0.739)	0.894
		Pack-years	0.700 (0.658 to 0.738)
	Smokers	Duration	0.627 (0.571 to 0.681)	0.774
		Pack-years	0.629 (0.573 to 0.682)
EGFR	All	Duration	0.801 (0.770 to 0.829)	0.911
		Pack-years	0.801 (0.770 to 0.828)
	Smokers	Duration	0.850 (0.803 to 0.888)	0.454
		Pack-years	0.844 (0.795 to 0.882)
C>A	All	Duration	0.746 (0.693 to 0.792)	0.472
		Pack-years	0.736 (0.687 to 0.780)
	Smokers	Duration	0.660 (0.593 to 0.721)	0.129
		Pack-years	0.644 (0.576 to 0.707)

AUC, area under the receiver operating characteristic curve.

Discussion

This study showed that there were no significant differences in AUC values between smoking duration and pack-years for TP53 and EGFR mutations and for C>A, but pack-years was a significantly better predictor of KRAS mutations than smoking duration in ever-smokers. Dogan et al reported that pack-years of smoking has a significant predictive value for KRAS and EGFR mutations in lung adenocarcinomas,20 but they did not compare pack-years with smoking duration with respect to prediction of KRAS and EGFR mutations. To the best of our knowledge, our study is the first to compare these two indices (smoking duration vs pack-years) in predicting the risk of smoking-related oncogenic mutations in NSCLC. For COPD, the strength of the association between smoking duration and COPD was greater than that between pack-years and forced expiratory volume in 1 s (FEV1)/forced vital capacity, emphysema, gas trapping, FEV1, 6 min walking distance and St George’s Respiratory Questionnaire.12

The relative contributions of smoking duration and cigarettes smoked per day to lung cancer incidence have been examined but not in terms of the incidence of driver gene alterations.21 22 Smoking duration was more strongly associated with lung cancer development than cigarettes smoked per day,23 but no comparisons were made in terms of pack-years. It is accepted that a longer duration of smoking is associated with increasing accumulation of genetic and epigenetic changes. A long smoking history is almost always self-reported, and it is conceivable that the lower predictive value of smoking intensity is due to the fact that smoking duration may be recalled and reported with greater accuracy than average daily intensity over a lifetime of smoking history.23 We believe that the duration of smoking is more easily and accurately recalled than the average number of cigarettes smoked per day, which tends to fluctuate over time. It is also harder to accurately quantify the number of cigarettes smoked per day, and the measurements are correlated poorly with the biochemical assessments of smoking exposure.24

A retrospective analysis of individuals referred to centralised lung cancer screening programmes25–27 serving a 5-hospital health services system in Seattle, Washington between October 2014 and January 2016 has been reported. The study assessed the eligibility of individuals referred for lung cancer screening and compared the information extracted from electronic medical records (EMRs) with the information derived from a shared decision-making conversation to determine the eligibility for lung cancer screening. They found a 96.2% discordance in pack-year smoking history between EMRs and shared decision-making conversations. The EMRs under-reported pack-years of smoking for 85.2% of participants. If the identification of eligible individuals relied solely on the accuracy of pack-year smoking history recorded in EMRs, 53.6% of participants would have failed to meet the 30-pack-year threshold for screening. Over-reliance on EMRs for the identification of individuals at risk may lead to missed opportunities for appropriate lung cancer screening.28

Although figure 2 shows similar graphs for smoking duration and pack-years for KRAS, TP53 and EGFR mutations and C>A, pack-years was a significantly better predictor of KRAS mutations than smoking duration in ever-smokers; this was not the case for TP53 and EGFR mutations and C>A. Figure 1 shows an unbalanced distribution of smoking duration and pack-years; therefore, it is possible that the exact smoking dose has not been reflected. It has been reported that a subtype of KRAS mutations was associated with smoking dose.29 In Dogan et al’s study, the observed KRAS mutation subtypes were G12C (39.4%), G12V (20.7%), G12D (17.0%) and G12A (10.7%).20 Never-smokers were significantly more likely to harbour transition mutations (G>A), rather than the transversion mutations known to be smoking-related (G>T or G>C), than ever-smokers.30 G12C, a transversion mutation, was the most frequent mutation among ever-smokers, and G12D, a transition mutation, was the most frequent mutation among never-smokers.20 30 In our study, the observed KRAS mutations were G12C (26.0%), G12V (24.7%), G12D (19.2%) and G12A (15.1%). The frequency of KRAS G12C mutations in our study was lower than that reported by Dogan et al. This may reflect the difference in the proportion of ever-smokers (50.3% in our study, 72.6% in their study), since KRAS G12C was found to be more strongly associated with smoking-associated signature four in lung adenocarcinoma than other KRAS mutations.29

To explain why pack-years was superior to smoking duration in predicting the frequencies of KRAS mutations, we divided KRAS subtypes (G12A, G12C, G12D, G12V) into four groups according to smoking duration and pack-years. The frequencies of KRAS subtypes in each group are shown in online supplemental figure 1. We examined the association between the frequencies of KRAS subtypes and smoking duration or pack-year groups using logistic regression. The frequencies of KRAS G12C in both groups and that of G12V in the pack-year group increased significantly with an increase in smoking dose (G12C (duration): p<0.001, G12C (pack-years): p=1.03×10−3, G12V (pack-years): p=0.017). There were no significant increases in the frequencies of G12A (duration, pack-years), G12D (duration, pack-years) or G12V (duration) with an increase in smoking dose. Based on the results of this subset analysis, it can be reasonable to conclude that pack-years was superior to smoking duration in predicting the frequencies of KRAS mutations.

The main limitation of our study was that the JME study data were obtained by targeted sequencing and not from whole genome or whole exome sequencing. We focused on cancer development and chose 72 oncogenic driver genes as the targets for mutational analysis, and C>A was examined in limited lesions. Various targeted sequencing panels have recently been developed to efficiently determine tumour mutation burden, and a strong correlation has been observed between targeted sequencing and whole genome sequencing in some studies.31 32 Another limitation of our study was recall bias, which cannot be ruled out when obtaining smoking information, even in a prospective cohort study. Additionally, data were extracted only from Japanese patients. The frequencies of gene mutations are known to differ according to ethnicities.33 Therefore, to confirm our results, further studies using data from a large cohort involving people of different ethnicities with detailed smoking information are needed.

Conclusion

Smoking duration was not significantly different from pack-years in predicting the likelihood of smoking-related gene mutations. Given the recall bias in obtaining smoking information, smoking duration alone should be considered for further investigation as a simpler alternative to pack-years.