Changes in survival of patients with non–small cell lung cancer in Japan: An interrupted time series study

Abstract

The epidermal growth factor receptor tyrosine kinase inhibitors (EGFR‐TKIs) gefitinib and erlotinib were approved for metastatic or relapsed non–small cell lung cancer (NSCLC) in Japan in 2002 and 2007, respectively. EGFR mutation testing was also approved in 2007. Although clinical trials showed efficacy in NSCLC patients harboring activating EGFR mutations, these effects have rarely been reported in real‐world practice. We evaluated changes in survival in NSCLC patients following introduction of these agents and EGFR mutation testing by extracting patients diagnosed with NSCLC from 1993 through 2011 from six prefectural population‐based cancer registries in Japan. Relative survival (RS) was calculated by sex, histological subtype, and cancer stage. We conducted interrupted time series analysis to assess survival changes following introduction of EGFR‐TKIs and EGFR mutation testing. 120,068 patients with NSCLC were analyzed. One‐ and three‐year RS gradually increased in overall NSCLC for men and women. For adenocarcinoma, among men, slopes of 1‐ and 3‐year RS increased steeply in patients diagnosed from 2007 through 2011; among women, significant level increases were seen in 1‐year RS in patients diagnosed in 2002 (4.55% [95% confidence interval: 1.76‐7.33]) and 2007 (3.40% [1.27‐5.52]). These significant level increases were particularly obvious in women with adenocarcinoma at an advanced stage. Our results suggest that recent improvements in survival in men and women with adenocarcinoma are due at least partly to introduction of EGFR‐TKIs into real‐world practice, and to prescription based on appropriate patient selection following introduction of EGFR mutation testing into real‐world practice in Japan.

This population‐based study provided information on the effects of introducing EGFR‐TKIs and EGFR mutation testing in real‐world practice in a country with different proportions of NSCLC subtypes from Europe and the United States. A significant level change in 1‐year relative survival was detected among women with metastatic adenocarcinoma, for whom EGFR‐TKIs were most likely to be prescribed. Improving survival in women with adenocarcinoma would be at least partly due to the introduction of EGFR‐TKIs into real‐world practice and to prescription based on appropriate patient selection following introduction of EGFR mutation testing into real‐world practice in Japan.

Abbreviations

CT

computed tomography

EGFR‐TKI

epidermal growth factor receptor tyrosine kinase inhibitor

FDG‐PET

f‐18‐fluorodeoxyglucose positron emission tomography

NSCLC

non–small cell lung cancer

RS

relative survival

1. INTRODUCTION

The question of whether innovative cancer drugs with highly positive results in clinical trials improve survival outcomes at the population level is not always clear. In particular, while patient entry in clinical trials is restricted to exclude certain subjects, such as those with significant comorbidity, the range of patients treated in general clinical practice is typically substantially broader. ¹ , ² Validating the effect of new drugs using population‐based data will minimize the influence of selection bias and allow the assessment of outcomes in patients in general clinical practice. To date, however, very little research focusing on the effect of novel treatments at the population level has been conducted in Japan.

The development of molecularly targeted drugs for non–small cell lung cancer (NSCLC) with driver mutations has dramatically changed the treatment algorithm of NSCLC. One prototype of these drugs is gefitinib, an epidermal growth factor receptor tyrosine kinase inhibitor (EGFR‐TKI), which has demonstrated substantial tumor shrinkage in patients with adenocarcinoma histology, patients with East‐Asian ethnicity, or never smokers. ³ , ⁴ , ⁵ Later findings demonstrated that the actual determinant of gefitinib sensitivity was activation mutations of the EGFR gene, which are frequent in patients with the clinical characteristics described above. ³ , ⁴ , ⁵ , ⁶ , ⁷

In 2002, gefitinib was approved in Japan for the first time in the world for unresectable or relapsed NSCLC. The identification of EGFR mutations as predictive of TKI response followed in 2004, and EGFR mutation testing was approved in Japan in 2007. A second EGFR‐TKI, erlotinib, was approved the same year. Since 2009, several phase III trials have demonstrated that the benefit of gefitinib is limited to those patients with NSCLC harboring EGFR mutations. ⁸ , ⁹ , ¹⁰ Although these studies all showed far longer progression‐free survival in patients treated with gefitinib than in those treated with platinum‐doublet chemotherapy, this advantage did not translate into an overall survival benefit, probably because of high cross‐over to gefitinib after disease progression in the chemotherapy group. For erlotinib, while several phase III trials have reported significant improvements in progression‐free survival compared with platinum‐based chemotherapy, these clinical trials were limited to patients with EGFR mutation–positive disease. ¹¹ , ¹² Thus, these studies were unable to demonstrate a general clinical effect of EGFR‐TKIs on survival trends in lung cancer.

Nevertheless, hospital‐based studies in Japan and several population‐based studies in the United States and Europe have reported improvements in overall survival. ¹³ , ¹⁴ , ¹⁵ For example, Howlader et al. ¹⁴ showed that population‐level mortality from NSCLC in the United States fell sharply from 2013 to 2016 and that this was attributable to a reduction in incidence and the use of targeted therapies. Given the high prevalence of EGFR mutation in Japanese (~40% of NSCLC) compared with Caucasian populations (~15%), a greater improvement of survival outcomes in Japan than in the United States might be anticipated. ¹⁶ , ¹⁷ Indeed, Takano et al. ¹⁸ compared overall survival between patients with advanced lung adenocarcinoma who began first‐line systemic therapy before and after gefitinib approval in Japan and found that overall survival was significantly longer among EGFR‐mutant patients treated after approval than in those treated before approval and in those without EGFR mutations. However, this study was conducted at a single institution with a small number of patients (~400).

We hypothesized that the introduction of EGFR‐TKIs contributed to this improved survival in NSCLC patients, particularly those with nonresectable adenocarcinoma. In addition, the approval of EGFR mutation testing also likely improved survival because of the more likely appropriate prescription of EGFR‐TKIs to these patients. Hence, in our present study, we assessed time trends in patient outcomes in “real‐world” Japanese using population‐based cancer registry data. If an improvement in survival outcome was detected, we then aimed to determine which patient subset benefitted most in terms of both patient attributes and time of diagnosis in order to assess the contribution of the clinical introduction of EGFR‐TKIs and EGFR mutation testing to improved survival. Although several other molecular targets and corresponding inhibitors have been approved for clinical use over the last 15 years or more, including those targeting ALK, ROS1, BRAF, MET, and RET, we did not include them due to their relative infrequency compared with EGFR mutations. ¹⁶

2. MATERIALS AND METHODS

2.1. Data source

All data of patients diagnosed with NSCLC were obtained from population‐based cancer registries of six prefectures in Japan, namely Miyagi, Yamagata, Niigata, Fukui, Osaka, and Nagasaki. All were members of the Monitoring of Cancer Incidence in Japan (MCIJ) project in 2015, and their combined population accounts for 13.2% of the total population of Japan. ¹⁹ , ²⁰ They include patients with cancer who were diagnosed from 1993 through 2015. The percentage of patients for whom information on cancer incidence could only be collected from death certificates (DCO) during this period was between 6.0% and 20.4%, giving them better accuracy than other prefectural cancer registries in Japan at that time. The registries included information on individual patients, tumors, and treatment. Morphologies of lung cancer were recorded according to the morphology codes of the International Classification of Disease for Oncology, Third Edition (ICD‐O‐3). ²¹ NSCLC morphology was classified according to Cancer Incidence in Five Continents, Volume XI into the following groups: adenocarcinoma (8140, 8211, 8230, 8231, 8250‐8260, 8323, 8480‐8490, 8550‐8552, 8570‐8574, 8576); squamous cell carcinoma (8050‐8078, 8083, 8084) and others, including large cell carcinoma and part of other specified carcinomas (8010‐8012, 8014‐8031, 8035, 8046, 8310, 8560). ²² Cancers were staged in four groups: (I) localized, cancer confined to the original organ; (II) lymph node metastasis, cancer spread to regional lymph nodes; (III) adjacent metastasis, cancer spread to immediately adjacent tissues; and (IV) distant, cancer metastasized to distant organs. Cancer stage was then modified into three groups by creating a regional stage which combined lymph node involvement and adjacent metastasis, on the basis that the original groupings both involved metastasis near the primary cancer lesion. In five of the six prefectures, patients were followed up after diagnosis and routinely linked to the prefectural death certificate database and residence records to confirm their vital status; for Niigata prefecture, only the prefectural death certificate database was used for linkage. The cutoff time for follow‐up was set as December 31, 2016.

2.2. Subject selection and inclusion criteria

Patients diagnosed with primary NSCLC from January 1, 1993 to December 31, 2011 were identified from the six prefectural cancer registries. Survival status was followed for up to 5 years after diagnosis. For survival analysis, patients who met any of the following criteria were excluded: (1) DCO; (2) age less than 15 or more than 100 years; and (3) presence of multiple cancers, among which NSCLC was diagnosed as the second or later cancer.

2.3. Outcome

The primary outcome of this study was survival at 1 and 3 years after cancer diagnosis. Observation was terminated on the date of death.

2.4. Statistical analysis

Non–small cell lung cancer patients diagnosed after 2002 were more likely to receive EGFR‐targeted drugs than those diagnosed before 2002, as the first and second EGFR‐TKI, gefitinib and erlotinib, were approved for metastatic NSCLC in 2002 and 2007, respectively. Moreover, NSCLC patients diagnosed after 2007 were more likely to be prescribed EGFR‐TKIs appropriately due to the approval of EGFR mutation testing in 2007. We therefore defined three diagnostic periods, period 1 (1993‐2001), period 2 (2002‐2006), and period 3 (2007‐2011), according to whether EGFR‐TKIs were included as possible treatments and EGFR mutation testing was available as part of the treatment process for NSCLC. Patient characteristics by sex were described by histological subtype in each of the three diagnostic periods. Baseline characteristics were compared using the chi‐square test. Prognosis was assessed using relative survival (RS) analysis due to its ability to show survival probability in the absence of causes of death other than cancer. Background mortality of NSCLC patients was calculated using the complete national population life tables by birth year, age, and sex. ²³ , ²⁴ , ²⁵ RS was estimated using the maximum likelihood method of Esteve et al. ²⁴ , ²⁶ One‐ and three‐year RS of NSCLC patients according to year of diagnosis were calculated by sex, histological subtype, and cancer stage because these factors affect the prognosis of these patients.

The effect of the introduction of EGFR‐TKIs and EGFR mutation testing on clinical outcomes in NSCLC patients was determined by estimating the change in RS in three diagnostic periods using the interrupted time series and segmented regression models. ²⁷ , ²⁸ The model is a strong quasi‐experimental method that provides statistical evidence of the causal effect of the introduction of EGFR‐TKIs and EGFR mutation testing because it assesses changes in study outcomes following interventions after adjusting for pre‐existing levels and trends. ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ We considered the years in which EGFR‐TKIs and EGFR mutation testing were approved to be the time points at which they were introduced into general practice. We used segmented linear regression models to estimate the effects of introduction on change in the level and slope of survival in NSCLC patients. The formula for this segmented linear regression model is:

$$\begin{array}{c}{\text{RS}}_t={\beta }_0+{\beta }_1*{\text{time}}_t+{\beta }_2*{\text{intervention}}_{1t}+{\beta }_3*{{\mathrm{time}}_{{\mathrm{after}}_{\text{intervention}}}}_{1t}+{\beta }_4*{\text{intervention}}_{2t}+{\beta }_5*{{\mathrm{time}}_{{\mathrm{after}}_{\text{intervention}}}}_{2t}+{ϵ}_t\hfill \end{array}$$

where ${\beta }_0$ denotes the intercept, which means the pre‐existing level at time point $t=0$; ${\beta }_1$ is the baseline slope; ${\beta }_2$ is the level change in RS following approval of the first EGFR‐TKI (gefitinib); ${\beta }_3$ indicates the slope change in RS following the approval of gefitinib; ${\beta }_1+{\beta }_3$ is the slope after the approval of gefitinib; ${\beta }_4$ is the level change following the approval of EGFR‐TKIs and EGFR mutation testing in the period; ${\beta }_5$ indicates the difference between the first‐ and second‐intervention slopes of RS; and ${\beta }_1+{\beta }_3+{\beta }_5$ is the slope after the approval of EGFR‐TKIs and EGFR mutation testing (Figure 1). ²⁸ , ²⁹ , ³⁰ We interpreted this to mean that statistically significant level and slope changes reflect the effect of the introduction of EGFR‐TKIs (in 2002) and of EGFR‐TKIs and EGFR mutation testing (in 2007) according to the time point at which these changes were introduced. We controlled for significant autocorrelation terms in the initial regression models.

FIGURE 1.

Estimated parameters from interrupted and segmented linear regression models in this study. Level and slope changes represent the sustained change immediately after the interruption and the difference in trends before and after the interruption, respectively

Two‐sided P values <0.05 were considered statistically significant. Statistical analysis and graph description were conducted using STATA version 16 (Stata Corp) and RStudio version 1.3.959.

3. RESULTS

We identified 120,068 NSCLC patients as eligible for RS analysis and observed 93,241 deaths in six prefectural registries from 1993 through 2011. Differences in patient characteristics among the three periods are shown by sex and histological subtype in Table 1. Approximately 30% of overall NSCLC patients were women. The proportion of elderly patients increased over time for both sexes and with each histological subtype. By cancer stage, the proportions of localized and metastatic cancer increased remarkably among patients with adenocarcinoma from period 1 to period 3, while that of regional cancer decreased. Adenocarcinoma occurred more frequently than squamous cell carcinoma over time for both sexes.

TABLE 1.

Characteristics of non–small cell lung cancer (NSCLC) patients diagnosed from 1993 through 2011

	Overall NSCLC							Adenocarcinoma							Squamous cell carcinoma
	Period 1 a		Period 2 b		Period 3 c		P value	Period 1 a		Period 2 b		Period 3 c		p Value	Period 1 a		Period 2 b		Period 3 c		P value
	N	%	N	%	N	%		N	%	N	%	N	%		N	%	N	%	N	%
Men
Number of patients	32,744		23,191		27,878			15,890		12,168		14,905			13,679		8587		9651
Age at diagnosis
15‐39	261	0.8	156	0.7	194	0.7	< 0.001	189	1.2	118	1.0	135	0.9	< 0.001	21	0.2	18	0.2	28	0.3	<0.001
40‐74	23,808	72.7	15,552	67.1	17,580	63.1		12,055	75.9	8693	71.4	10,068	67.6		9466	69.2	5367	62.5	5588	57.9
75‐99	8675	26.5	7483	32.3	10,104	36.2		3646	23.0	3357	27.6	4702	31.6		4192	30.7	3202	37.3	4035	41.8
Cancer stage
Localized	7167	21.9	5499	23.7	7406	26.6	< 0.001	3501	22.0	3162	26.0	4543	30.5	< 0.001	3236	23.7	1988	23.2	2373	24.6	<0.001
Regional	11,090	33.9	7458	32.2	8234	29.5		4828	30.4	3443	28.3	3689	24.8		5311	38.8	3317	38.6	3627	37.6
Distant	9165	28.0	7226	31.2	9890	35.5		5186	32.6	4158	34.2	5551	37.2		2816	20.6	2090	24.3	2838	29.4
Unknown	5322	16.3	3008	13.0	2348	8.4		2375	15.0	1405	11.6	1122	7.5		2316	16.9	1192	13.9	813	8.4
Women
Number of patients	13,045		10,128		13,082			10,393		8315		10,723			1741		1147		1410
Age at diagnosis
15‐39	162	1.2	115	1.1	120	0.9	< 0.001	139	1.3	99	1.2	106	1.0	< 0.001	6	0.3	2	0.2	7	0.5	0.055
40‐74	9069	69.5	6648	65.6	8076	61.7		7537	72.5	5638	67.8	6875	64.1		1016	58.4	670	58.4	760	53.9
75‐99	3814	29.2	3365	33.2	4886	37.4		2717	26.1	2578	31.0	3742	34.9		719	41.3	475	41.4	643	45.6
Cancer stage
Localized	3825	29.3	3511	34.7	4998	38.2	< 0.001	3389	32.6	3208	38.6	4564	42.6		340	19.5	226	19.7	309	21.9	<0.001
Regional	3503	26.9	2414	23.8	2874	22.0		2662	25.6	1855	22.3	2165	20.2		621	35.7	402	35.1	493	35.0
Distant	3710	28.4	2965	29.3	4169	31.9		2898	27.9	2358	28.4	3285	30.6		456	26.2	339	29.6	442	31.4
Unknown	2007	15.4	1238	12.2	1041	8.0		1444	13.9	894	10.8	709	6.6		324	18.6	180	15.7	166	11.8

^a Period 1 included patients diagnosed with NSCLC from 1993 through 2001.

^b Period 2 included patients diagnosed with NSCLC from 2002 through 2006.

^c Period 3 included patients diagnosed with NSCLC from 2007 through 2011.

3.1. Changes in 1‐ and 3‐year RS in overall NSCLC patients

Changes in 1‐ and 3‐year RS in overall NSCLC patients diagnosed after the introduction of EGFR‐TKIs and EGFR mutation testing from segmented regression models are shown in Figure 2 and Table S1. One‐ and three‐year RS improved in overall NSCLC patients for both sexes. Among women, there was a significant level increase in 1‐year RS in patients diagnosed in 2002, the year of introduction of the first EGFR‐TKI, gefitinib (3.49% [95% confidence interval: 1.33‐5.65]). A significant slope change in 1‐year RS was also observed in women diagnosed in period 2 (−1.13% per year [−1.81 to −0.46]). Significant changes in 1‐year RS were observed in both level and slope in women diagnosed in period 3 (2.66% [0.75‐4.57] and 0.57% per year [0.08‐1.06], respectively). We observed similar patterns for 3‐year RS, albeit there was no level change in women diagnosed in 2002.

FIGURE 2.

Trends in 1‐ and 3‐year relative survival (RS) rates in overall non–small cell lung cancer patients by sex. Error bars indicate the 95% CIs of the point estimates of 1‐ and 3‐year RS for each diagnosis year. Asterisks indicate that the slope of the regression line is significant. *p < 0.05, ** p < 0.001. Only significant level and slope changes are depicted

Among men, there was a significant level increase in 1‐year RS in patients diagnosed in 2002 (2.47% [0.02‐4.92]). A significant slope change in 1‐year RS was observed in patients diagnosed in period 3 (0.83% per year [0.45‐1.21]), and their 1‐year RS improved more steeply than that in those diagnosed in period 2. Three‐year RS improved significantly across the three periods. Unlike the case with 1‐year RS, there was no level change in men diagnosed in 2002; nevertheless, a significant slope change in period 3 was observed (1.15% per year [0.16‐2.15]).

3.2. Changes in 1‐ and 3‐year RS in NSCLC patients with adenocarcinoma

Changes in 1‐ and 3‐year RS of NSCLC are shown by histological subtype in Figure 3 and Table S2. For adenocarcinoma, among women, significant level increases in 1‐year RS were seen in patients diagnosed in 2002 (4.55% [1.76‐7.33]) and 2007 (3.40% [1.27‐5.52]). A significant slope change of 1‐year RS was seen in women diagnosed in period 2 (−1.04% per year [−1.70 to −0.38]). Slope in this period was not significant, whereas it improved in those diagnosed in period 1. A similar pattern was seen with 3‐year RS, except the level change in women diagnosed in 2002 was not significant, while a significant slope change was observed in those diagnosed in period 3 (0.67% per year [0.08‐1.25]).

FIGURE 3.

Trends in 1‐ and 3‐year relative survival (RS) rates in non–small cell lung cancer patients diagnosed with adenocarcinoma and squamous cell carcinoma by sex. Error bars indicate the 95% CIs of the point estimates of 1‐ and 3‐year RS for each diagnosis year. Asterisks indicate that the slope of the regression line is significant. *p < 0.05, **p < 0.001. Only significant level and slope changes are depicted

Among men, no level change was seen in 1‐year RS, while a significant slope change was seen in those diagnosed in period 3 (0.98% per year [0.36‐1.60]). The pattern in 3‐year RS was similar, except for a slope change in those diagnosed in period 2 (−0.91% per year [−1.45 to −0.37]).

3.3. Changes in 1‐ and 3‐year RS in NSCLC patients with squamous cell carcinoma

For squamous cell carcinoma, women showed no level or slope changes in 1‐ or 3‐year RS. Among women, 1‐year RS improved only in patients diagnosed in period 1. Three‐year RS also improved in women diagnosed in periods 1 and 3. Among men, there was no level change in 1‐year RS, but a slope change was seen in patients diagnosed after 2007 (0.76% per year [0.05‐1.46]). One‐year RS did not improve in men diagnosed in periods 1 and 2 but improved in those in period 3 (0.79% per year [0.44‐1.14]). For 3‐year RS, significant slope change was seen in men diagnosed in period 2 (0.61% per year [0.22‐1.01]). Three‐year RS improved in men diagnosed in period 2, which improved further in those diagnosed in period 3.

3.4. Changes in 1‐ and 3‐year RS in NSCLC patients with adenocarcinoma by cancer stage

Figure 4 and Table S3 show changes in 1‐ and 3‐year RS in NSCLC patients with adenocarcinoma by cancer stage. Among women with metastatic cancer, level in 1‐year RS increased significantly in patients diagnosed in 2002 (14.37%, [10.34‐18.39]). A significant slope change was observed in women diagnosed with metastatic cancer in period 2 (−1.69% per year [−3.0 to −0.31]), while the slopes for both periods 1 and 2 were not significant. In addition, a significant level increase in 1‐year RS was observed in women diagnosed with metastatic cancer in 2007 (8.49% [3.52‐13.45]), and survival improved in those women diagnosed in period 3 due to a significant slope change (2.49% per year [1.22‐3.77]). Among women with localized or regional cancer, there was no change in level or slope in 1‐year RS. For 3‐year RS, the slope of 3‐year RS was not significant in women diagnosed with metastatic cancer in period 2 due to a significant slope change (−1.24% per year [−2.47 to −0.02]). Level in 3‐year RS increased significantly in women diagnosed with metastatic cancer in 2007 (5.60% [1.11‐10.09]), and survival improved in those diagnosed in period 3. We observed similar patterns for women diagnosed with regional cancer, except the level change in those diagnosed in 2002 was significant (6.23% [1.28‐11.17]), while a slope change was not significant in those diagnosed in period 3. Among women with localized cancer, 3‐year RS improved across three periods, although no changes were observed in level or slope.

FIGURE 4.

Trends in 1‐ and 3‐year relative survival (RS) rates in non–small cell lung cancer patients diagnosed with adenocarcinoma by sex and cancer stage. Error bars indicate the 95% CIs of the point estimates of 1‐ and 3‐year RS for each diagnosis year. Asterisks indicate that the slope of the regression line is significant. *p < 0.05, **p < 0.001. Only significant level and slope changes are depicted

Among men with any cancer stage, 1‐ and 3‐year RS tended to improve over time. For 1‐year RS, no level or slope changes were observed in men with metastatic cancer, unlike the case in women. A significant slope change in 1‐year RS was observed in men diagnosed with regional cancer in period 3 (1.21% per year [0.56‐1.86]). For 3‐year RS, slope increases were steeper in men diagnosed with regional or metastatic cancer in period 3.

4. DISCUSSION

In this study, we described long‐term trends in the survival of NSCLC patients in the context of changing treatment patterns using population‐based data in Japan. In patients with adenocarcinoma, among women, level changes in survival rate significantly increased in the years when EGFR‐TKIs (in 2002) and EGFR mutation testing (in 2007) were introduced; specifically, 1‐ and 3‐year RS increased in those diagnosed in 2002, and 3‐year RS in those diagnosed in 2007. It is noteworthy that the level changes in 1‐ and 3‐year RS were significant among women with metastatic adenocarcinoma, for whom EGFR‐TKIs were most likely prescribed. Given the very high incidence of EGFR mutations in Asian women with adenocarcinomas of the lung, ⁴ , ⁵ these significant level changes in 1‐ and 3‐year RS may be attributable to the introduction of EGFR‐TKIs and prescription based on appropriate patient selection due to the introduction of EGFR mutation testing, and suggest their effect in real‐world practice in Japan.

The slope in 1‐year RS in men diagnosed with non–small cell lung adenocarcinoma from 2002 through 2006 and in 1‐ and 3‐year RS in women diagnosed from 2002 through 2006—in particular in women diagnosed with metastatic non–small cell lung adenocarcinoma from 2002 through 2006—significantly decreased compared with the slope in those diagnosed from 1993 through 2001. These differences may indicate the possibility of inappropriate pharmacotherapy based on inconclusive patient selection. This therapy might have included the use of gefitinib as an alternative to anticancer drugs in NSCLC patients without EGFR mutations, prior to clarification of the significance of these mutations. Mok et al. ¹⁰ reported that the effect of gefitinib was inferior to that of carboplatin‐paclitaxel among these patients; we therefore speculate that inappropriate pharmacotherapy might have abrogated the improvement in survival.

In contrast, the slope in 1‐year and 3‐year RS among men diagnosed with non–small cell lung adenocarcinoma from 2007 through 2011 and in 3‐year RS among women diagnosed with it from 2007 through 2011 increased compared with the slopes in men and women diagnosed from 2002 through 2006. By cancer stage, these slope changes appeared evident in advanced cancers. These changes might be attributable to the increase in appropriate prescribing of EGFR‐TKIs to patients with EGFR mutation–positive NSCLC due to the approval of EGFR mutation testing in 2007, rather than to the effect of approval of erlotinib, given that the efficacy of erlotinib does not substantially differ from that of gefitinib. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹²

With regard to NSCLC patients with squamous cell carcinoma, level changes were not seen in these patients, and slope changes were seen only in men. The slope change in 1‐year RS among these men diagnosed from 2007 through 2011 might also be due to the introduction of EGFR mutation testing, which led to prescribing based on appropriate patient selection and consequent infrequent use of EGFR inhibitors in NSCLC patients with squamous cell carcinoma.

In our study, the survival of NSCLC patients improved over the study period. This might have been due to progress in the accuracy of staging, at least in part. The use of computed tomography (CT) and f‐18‐fluorodeoxyglucose positron emission tomography (FDG‐PET) may have facilitated the detection of cancers at an early stage. Low‐dose CT was reported to help detect lung cancer at an early stage, albeit at the risk of overdiagnosis. ³² , ³³ , ³⁴ , ³⁵ Also, combined evaluation with CT and FDG‐PET was demonstrated to minimize unnecessary invasive procedures for benign lesions and to improve the accuracy of staging. ³² , ³⁶ Our results showed an increase in the proportion of NSCLC patients who were diagnosed at an early stage. In addition, we also reported improvements in survival in NSCLC patients diagnosed at any cancer stage. Accordingly, the improved survival in NSCLC patients might have been attributable in part to any or all of overdiagnosis in the early stage due to improvements in diagnostic imaging and progression in other treatments, such as radiotherapy or palliative care. ³⁷ , ³⁸

Although mortality in lung cancer patients has decreased consistently in men and intermittently in women, ³⁹ the degree of this decrease does not appear to be large, despite the dramatic improvement in short‐term survival in NSCLC patients with adenocarcinoma due to the introduction of EGFR‐TKIs and EGFR mutation testing. This suggests that the improvement in short‐term survival is not reflected in a decrease in mortality.

A descriptive study evaluated the population‐level effect of the introduction of EGFR‐TKIs in lung cancer patients in Taiwan, who have a high proportion of EGFR mutation–positive cases. ²⁹ One‐ and two‐year RS among lung cancer patients improved from 1994 through 2014, but survival could not be calculated by cancer stage or histological subtype because of a lack of individual data. We reported the effect of introducing EGFR‐TKIs and EGFR mutation testing on survival trends in NSCLC patients by histologic subtype and cancer stage using high‐quality population‐based cancer registry data, and were thus able to expand the evidence provided in this previous study. ²⁹

Our study has a number of strengths. First, it was conducted using population‐based data, and was therefore able to provide evidence of the effect of introduction of EGFR‐TKIs and EGFR mutation testing in the general population. ⁴⁰ Second, the quasi‐experimental design method—an interrupted time series analysis to assess the effect of introducing EGFR‐TKIs and EGFR mutation testing in NSCLC—provided strong evidence for a causal effect of the introduction of EGFR‐TKIs and EGFR mutation testing on survival among NSCLC patients, particularly the level and slope changes in survival beginning with the introduction of EGFR‐TKIs and EGFR mutation testing. Accordingly, our results more clearly indicate the effect of introducing EGFR‐TKIs than previous studies using joinpoint analysis, which can identify the slope change in outcome trends but not the level change. ¹⁴ , ⁴¹

Our study also has several limitations. First, population‐based cancer registries in Japan do not collect information on patients and clinical treatments, such as comorbidities, performance status, smoking habit, gene mutation, and types of drugs, and we were unable to consider the presence or absence of mutations or evaluate specific drugs for their role in improved survival. Furthermore, we were unable to consider other factors that may have affected patient survival. However, this limitation is unlikely to have biased our results because the lack of information was consistent across observation periods. Second, a degree of selection bias may have been present because we only analyzed patients who had been diagnosed with NSCLC and registered in the cancer registry system. Our data were developed prior to the 2016 enactment of the Cancer Registry Promotion Act, which mandated registration of all the cancer patients nationwide in Japan. ⁴² Our patient data are therefore more likely to have been sourced from designated cancer hospitals, especially in the earlier periods. Considering the substantially better survival in designated cancer hospitals, our finding of improved NSCLC survival likely underestimates the improving trend, and our results are likely robust despite this limitation. Finally, we could not directly assess the effect of introducing EGFR‐TKIs and EGFR mutation testing and their timing on survival because we could not calculate yearly survival but rather only survival by year of diagnosis. The effect of introducing the EGFR‐TKIs and EGFR mutation testing—analyzed using the change of level and slope in survival—may be underestimated because NSCLC patients diagnosed before 2001 might have also received treatment with these drugs.

In conclusion, using representative population‐based cancer registry data in Japan, we observed a gradual improvement in the survival of patients with NSCLC between 1993 and 2011. Furthermore, we indicated the level increase in survival in women diagnosed with advanced adenocarcinoma in 2002 or 2007, the years of introduction of EGFR‐TKIs and EGFR mutation testing for NSCLC, and the slope increase in survival in both sexes diagnosed with it after 2007. We suggest that the improved survival in these specific populations was due to the introduction of two factors into general clinical practice, namely EGFR‐TKIs for treatment and EGFR mutation testing for rational prescribing based on appropriate patient selection. Since 2012, new EGFR‐TKIs for the treatment for NSCLC have been regularly introduced, along with other molecularly targeted drugs that correspond to ALK, ROS1, BRAF, MET, and RET mutations. ¹⁶ Ongoing observation of survival trends in NSCLC patients to assess the effect of introducing these novel drugs into general clinical practice is required.

FUNDING INFORMATION

This work was supported by the Ministry of Health, Labour and Welfare, Japan by a Grant‐in‐Aid for Research for Promotion of Cancer Control Programs; Grant numbers: H29‐political‐general‐015, H29‐political‐general‐016, and 20EA1026 and JSPS KAKENHI Grant numbers: JP20K23247 and JP21H03210.

DISCLOSURE

Dr. Mitsudomi reports receiving honoraria for lectures from AstraZeneca, Eli‐Lilly, MSD, and Ono Pharmaceutical and grants from Boehringer Ingelheim and Bridge Biopharma. Dr. Mitsudomi and Dr. Matsuo are current Editorial Board Members of Cancer Science. The remaining authors declare no conflict of interest.

ETHICAL APPROVAL

Approval of the research protocol by the institutional review board of the National Cancer Center, Japan (2019–202).

Supporting information

Table S1. Change in relative survival rates following approval of EGFR‐TKIs and EGFR mutation testing in overall non–small cell lung cancer patients by sex

Table S2. Change in relative survival rates following approval of EGFR‐TKIs and EGFR mutation testing in non–small cell lung cancer patients diagnosed with adenocarcinoma and squamous cell carcinoma by sex

Table S3. Change in relative survival rates following approval of EGFR‐TKIs and EGFR mutation testing in non–small cell lung cancer patients with adenocarcinoma by sex and cancer stage

Taniyama Y, Oze I, Koyanagi YN, et al. Changes in survival of patients with non–small cell lung cancer in Japan: An interrupted time series study. Cancer Sci. 2023;114:1154‐1164. doi: 10.1111/cas.15646

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.