Comparative effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors: Analysis of real-world data in a tertiary hospital in Taiwan.

INTRODUCTION: Comparison of the effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors (TKIs) would improve patients' clinical benefits and save costs. Using real-world data, this study attempted to directly compare the effectiveness and cost-effectiveness of first-line afatinib, erlotinib, and gefitinib.
METHODS: During May 2011-December 2017, all patients with non-small cell lung cancer (NSCLC) visiting a tertiary center were invited to fill out the EuroQol five-dimension (EQ-5D) questionnaires and World Health Organization Quality of Life, brief version (WHOQOL-BREF), and received follow-ups for survival and direct medical costs. A total of 379 patients with EGFR mutation-positive advanced NSCLC under first-line TKIs were enrolled for analysis. After propensity score matching for the patients receiving afatinib (n = 48), erlotinib (n = 48), and gefitinib (n = 96), we conducted the study from the payers' perspective with a lifelong time horizon.
RESULTS: Patients receiving afatinib had the worst lifetime psychometric scores, whereas the differences in quality-adjusted life expectancy (QALE) were modest. Considering 3 treatments together, afatinib was dominated by erlotinib. Erlotinib had an incremental cost-effectiveness of US$17,960/life year and US$12,782/QALY compared with gefitinib. Acceptability curves showed that erlotinib had 58.6% and 78.9% probabilities of being cost-effective given a threshold of 1 Taiwanese per capita GDP per life year and QALY, respectively.
CONCLUSION: Erlotinib appeared to be cost-effective. Lifetime psychometric scores may provide additional information for effectiveness evaluation.

Introduction

In Asian countries such as Japan and Taiwan, more than half of non-small cell lung cancer (NSCLC) patients tested for epidermal growth factor receptor (EGFR) mutations have shown positive results [1]. In addition to new generation osimertinib [2], three EGFR-tyrosine kinase inhibitors (TKIs)–afatinib, erlotinib, and gefitinib—are commonly used as first-line therapies for advanced NSCLC. Although a randomized trial showed afatinib is superior to first-generation TKIs in progression-free survival [3], a significant difference in overall survival has not been revealed [4]. Based on our clinical observation, the quality of life (QoL) and costs among patients receiving different EGFR-TKIs may differ. To improve patients’ clinical benefits and save costs, the comparative effectiveness and cost-effectiveness of these drugs warrant further exploration.

Previous studies comparing the effectiveness and cost-effectiveness of first-line erlotinib versus gefitinib, erlotinib versus afatinib, and afatinib versus gefitinib, usually used model analyses and trial data [5-7]. Constructing the model analyses requires several assumptions, and utility values of QoL are often borrowed from other investigations. Although trial data are generally cleaner, their restrictive inclusion and exclusion criteria and limited length of follow-up period may limit the application in daily practice. A cost-effectiveness study using real-world approach would be useful in assisting healthcare resources allocation. Moreover, most previous analyses used chemotherapy as the reference group for indirect treatment comparisons [5, 6].

From May 2011 to December 2017, we prospectively invited all lung cancer patients visiting a tertiary center to provide their survival, QoL, and costs data for analysis. By integrating the long-term survival with utility values of QoL and costs, we developed a method to estimate the quality-adjusted life expectancy (QALE) and lifetime costs. Because psychometric scores are more sensitive than the utility values [8] and may provide additional information for effectiveness evaluation, lifetime psychometric scores were also estimated. Using the new method and real-world data of a tertiary hospital in Taiwan, this study attempted to directly compare the effectiveness and cost-effectiveness of three first-line EGFR-TKIs.

Methods

This study was approved by the Institutional Review Board of National Cheng Kung University Hospital (NCKUH) before commencement (A-ER-107-107). All participants provided written informed consent. We performed the study from payers’ perspective, and the time horizon was lifelong.

From May 2011 to December 2017, we invited all lung cancer patients who visited the outpatient departments of NCKUH to fill out QoL questionnaires, and receive follow-ups for survival and healthcare expenditures. Throughout 2017, we also recruited patients from the thoracic ward. There were 729 patients with EGFR mutation-positive advanced NSCLC under first-line TKIs during the study period. After excluding patients without informed consent and cases with missing values on EuroQol five-dimension (EQ-5D) questionnaires, all subjects were included for analysis. More specifically, the QIAamp DNA Mini Kit (Qiagen, Valencia, CA, USA) was used to analyze EGFR mutations of effusion cytology and tissue samples. We excluded patients with tumor stages I, II, and IIIA at the initiation of EGFR-TKIs, leaving only subjects with recurrent or newly-diagnosed advanced NSCLC in the analysis. Afatinib, erlotinib, and gefitinib [9] were defined as the standard first-line therapies because osimertinib [2] had not yet become a standard therapy during the study period.

Propensity score matching

We created a system to abstract age, sex, performance and recurrence statuses at the initiation of therapy from electrical medical records. Because all these data are required to be approved for receiving the first-line EGFR-TKIs in our hospital, the information collected were relatively complete, leaving few patients with missing performance statuses. In addition, we reviewed the reports of brain magnetic resonance imaging and computed tomography with contrast to define brain metastasis. That is, subjects who did not receive brain images or show any radiographic evidence at the initiation of therapy were categorized as negative for brain metastasis.

To account for observed covariates among three different EGFR-TKIs, we used propensity score matching via greedy algorithm [10]. That is, the first treated unit was selected to find its closest control based on the difference of their propensity scores using logistic regression. The procedure was repeated for all the treated units. We first matched patients receiving afatinib versus erlotinib one-to-one, followed by one-to-two matching for gefitinib versus erlotinib. Previous literatures have found performance status, recurrence status, metastasis, and mutation subtypes to be prognostic factors of survival [11] and QoL outcomes [8, 12] among patients receiving first-line EGFR-TKIs. Therefore, we computed propensity scores by age, sex, and these clinical characteristics upon initiation of treatment. The balances between afatinib, erlotinib, and gefitinib were tested using standardized differences, an absolute value less than 0.1 suggests each two groups are well balanced.

Effectiveness

Each matched patient underwent follow-ups from the initiation of EGFR-TKI until September 2018 to verify the survival status. By using a semiparametric method explained in detail in our previous article [13], we extrapolated the survival to lifetime to estimate the life expectancy of patients receiving one of three first-line treatments. The extrapolation method has been shown to be effective via computer simulations [14], mathematical proof [15] and corroboration by examples of lung cancer cohorts [13, 16, 17]. The iSQoL statistical package (www.stat.sinica.edu.tw/isqol/) was used to perform the computations.

A thoracic oncologist independently reviewed every chest computed tomography, bone scan, positron emission tomography, and brain image to determine if there is any disease progression. Disease progression was defined according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 [18]. Switching of therapy due to adverse events without an image progression was not considered as disease progression.

Quality of life

The EQ-5D and World Health Organization Quality-of-Life—Brief (WHOQOL-BREF) questionnaires were used to estimate the QoL utility values and psychometric scores, respectively. We invited patients to fill out the questionnaires each time they visited our hospital to capture dynamic changes in their QoL along the follow-up course (i.e., QoL at different stages of the disease). To minimize collinearity, repeated measurements were performed more than 2 weeks apart. Using the EQ-5D scoring function from Taiwan [19], we transformed the health state parameters into a utility value ranging from 0 to 1, where 0 represents death and 1 indicates full health. To present the utility value for each group, we constructed linear mixed models to consider random effects from subjects because EQ-5Ds were repeatedly measured. The intercept represents the mean utility value. Each facet in the WHOQOL-BREF was scored from 1 to 5, where a higher score indicated a better QoL [20]. By multiplying the average of the scores of all facets in the same domain by four, a domain score was also calculated, ranging from 4 to 20. High correlation coefficients between Rasch scores and the crude domain scores have been documented [21], which supports the WHOQOL-BREF as a sound instrument to measure QoL for cancer patients.

The time after treatment for each QoL measurement was defined as the period between the initiation of treatment and that of the interview. We used kernel-smoothing (i.e., a moving average of the nearby 10%) to estimate the mean QoL function after initiation of treatment [22]. The QoL scores beyond the follow-up period were assumed to be the same as the average of the last 10% near the end of follow-up. We multiplied lifetime survival function by the mean QoL functions of EQ-5D and WHOQOL to obtain quality-adjusted survival curves, with the sum of the areas under the curves being the QALE and lifetime psychometric scores, respectively. We applied a 3% annual discount when the QALE was employed in the estimation of the incremental cost-effectiveness ratio (ICER).

Medical costs

We used the reimbursement database at NCKUH to obtain spending details from the initiation of treatment to December 2017. These data included all expenditures reimbursed by National Health Insurance (NHI) plus out-of-pocket money paid to the hospital, of which direct medical costs along time course of the disease could be obtained. Specifically, the total monthly healthcare expenditures were divided by the effective sample sizes, i.e., the number of patients who survived in that month, to obtain the average monthly healthcare expenditures per case. Similar to the estimation of QALE, costs beyond the follow-up period were assumed to be the same as the average of the last 10% near the end of follow-up. These values were subsequently multiplied by the corresponding monthly survival probabilities and summed to obtain the lifetime costs per case. All payments in different calendar years were adjusted based on the related consumer price indices and made them equivalent to 2017 dollars. To discount costs in future years, an annual discount rate of 3% was applied.

We did not collect transportation costs, payments to caregivers, or home adaptations due to illness or human capital loss in this analysis. The costs of EGFR-TKIs were stratified according to the order codes established by the NHI. The costs of chemotherapies included different-line regimens and administration fees.

Probabilistic sensitivity analysis

We assumed a normal distribution for life expectancy and QALE [23], and a gamma distribution for costs, with means and standard deviations set to base-case values. To determine the most cost-effective option using net life years or QALY gained, a Monte Carlo simulation with 1,000 iterations was conducted to construct acceptability curves of the three different EGFR-TKIs. We adopted the criterion suggested by WHO-CHOICE (World Health Organization-CHOosing Interventions that are Cost Effective), and applied one gross domestic product (GDP) per capita as the threshold for cost effectiveness. Cost-effectiveness scatter plots were also developed. SAS 9.4 and Amua 0.2.7 were used to perform the analyses.

Results

From May 2011 to December 2017, a total of 729 patients in NCKUH received afatinib, erlotinib, or gefitinib as first-line therapies for EGFR mutation-positive advanced NSCLC. Among them, 346 cases did not sign the informed consent and 4 cases with missing values on EQ-5D questionnaires, leaving 379 subjects (Fig A in S1 Appendix). After 1:1:2 (afatinib: erlotinib: gefitinib) propensity-score matching the patients, 192 patients were used for analysis. Table 1 shows the 192 propensity score-matched patients stratified by treatment as well as those excluded after matching. The daily prices per person were US$48.2, US$37.0, and US$36.3 for 40mg afatinib, 150mg erlotinib, and 250mg gefitinib, respectively. In general, propensity-score matched patients had higher proportions of men, brain metastasis, and common mutations compared with those without matching. After propensity score matching, most of the characteristics in the three groups were balanced. The progression-free survival and overall survival under three first-line treatments were also similar (Fig B in S1 Appendix). Patients receiving erlotinib had a higher mean utility value compared with the afatinib and gefitinib groups.

Table 1

Clinical characteristics of the 192 propensity score-matched patients and those excluded after matching.

	Propensity score-matched patients n = 192					Excluded patients after matching
	Afatinib (A)	Erlotinib (E)	Gefitinib (G)	Standardized differences
	n = 48	n = 48	n = 96	A vs. E	G vs. E	n = 187
Daily price per person, US$	48.2	37.0	36.3
Age, n (%)
< 67 years	32 (66.7)	31 (64.6)	60 (62.5)	0.04	-0.04	120 (64.2)
≥ 67 years	16 (33.3)	17 (35.4)	36 (37.5)	-0.04	0.04	67 (35.8)
Male, n (%)	21 (43.8)	21 (43.8)	39 (40.6)	0	-0.06	63 (33.7)
Mutation subtype, n (%)
Exon 19 deletions	24 (50.0)	23 (47.9)	46 (47.9)	0.04	0	70 (37.4)
L858R substitution	22 (45.8)	24 (50.0)	49 (51.0)	-0.08	0.02	91 (48.7)
Other mutations	2 (4.2)	1 (2.1)	1 (1.0)	0.12	-0.08	26 (13.9)
Performance status, n (%)
ECOG: 0–1	45 (93.8)	45 (93.8)	85 (88.5)	0	-0.18	168 (89.8)
ECOG: 2–4	3 (6.3)	3 (6.3)	11 (11.5)	0	0.18	18 (9.6)
Missing	0	0	0			1 (0.5)
Disease by recurrence, n (%)
Recurrent	9 (18.8)	8 (16.7)	12 (12.5)	0.06	-0.11	45 (24.1)
Newly-diagnosed	39 (81.3)	40 (83.3)	84 (87.5)	-0.06	0.11	142 (75.9)
Brain metastasis, n (%)	20 (41.7)	25 (52.1)	52 (54.2)	-0.21	0.04	14 (7.5)
PFS, median (IQR) months	12.3 (7.1–22.2)	12.7 (6.4–22.0)	11.5 (8.2–24.3)			12.1 (6.7–23.8)
Number of QoLs, n	200	194	491			951
Utility value, β0 (SE)a	0.80 (0.02)	0.85 (0.02)	0.81 (0.02)			0.82 (0.01)

aIntercept of linear mixed model considering random effects from subjects. ECOG: Eastern Cooperative Oncology Group; PFS: progression-free survival

Base case scenarios

Fig 1 depicts the cost- and quality-adjusted survival curves along time courses for different treatments. In this figure, the survival probability was multiplied by the costs and QoL at each time point t (see Fig C in S1 Appendix for the mean QoL curve using moving averages of the nearby 10% values), the sums of the shaded areas under the curves represent the lifetime costs and QALE, respectively. As expected, costs dropped after initiation of therapies but increased in final months due to end-of-life care [24]. Lifetime psychometric scores in 2 domains and 4 facets are depicted in Fig D in S1 Appendix.

Fig 1

Lifetime costs and QALE of patients receiving different first-line treatments.

The survival curves (dashed lines), costs and QoL functions (dotted lines); cost- and quality-adjusted survival curves (solid lines) are shown, with the shaded areas representing the lifetime costs and QALE, respectively. 1 US dollar = 29.848 Taiwanese dollars. QALE: quality-adjusted life expectancy; QoL: quality of life.

Costs, effectiveness, and ICER of the 192 propensity score-matched patients are summarized in Table 2. Patients receiving afatinib incurred the highest costs in both the progression-free and lifetime periods. Lifetime psychometric scores were lower in the afatinib group, including those in the physical and psychological domains, as well as facet scores of pain, sleep, bodily appearance, and negative feelings. However, the differences in QALE appeared to be modest. The erlotinib group dominated the afatinib group and had an incremental cost-effectiveness of US$17,960/life year and US$12,782/QALY when compared with the gefitinib group.

Table 2

Costs, effectiveness, and ICER of the 192 propensity score-matched patients.

	Erlotinib	Afatinib	Gefitinib
	n = 48	n = 48	n = 96
Costs, US$
Costs in progression-free period	31,734 (3,210)	36,001 (2,874)	31,873 (2,177)
EGFR-TKIs	19,122 (1,647)	19,970 (1,519)	16,546 (1,161)
Other than EGFR-TKIs	12,612 (1,576)	16,040 (2,052)	15,328 (1,251)
Lifetime costs	59,005 (3,390)	64,465 (3,856)	55,227 (2,249)
EGFR-TKIs	34,693 (2,004)	34,094 (2,565)	26,740 (1,246)
Other than EGFR-TKIs	24,360 (1,867)	30,397 (1,775)	28,490 (1,564)
Chemotherapies	4,805 (792)	11,980 (1,212)	8,683 (844)
Lifetime psychometric score, score year
Physical	43.9 (2.3)	37.1 (1.7)	37.2 (1.3)
Pain	12.7 (0.6)	11.1 (0.6)	11.3 (0.4)
Sleep	10.2 (0.5)	8.1 (0.3)	8.8 (0.3)
Psychological	41.7 (2.0)	35.2 (1.7)	36.8 (1.2)
Bodily appearance	10.6 (0.6)	8.9 (0.5)	9.5 (0.3)
Negative feelings	11.2 (0.6)	9.7 (0.5)	10.7 (0.4)
Effectiveness
Life expectancy, life year	3.06 (0.14)	2.94 (0.13)	2.84 (0.10)
QALE, QALY	2.53 (0.12)	2.21 (0.11)	2.20 (0.08)
ICER
ΔCost /ΔLife expectancy	17,960 (6,766)	dominated	—
ΔCost /ΔQALE	12,782 (25,001)	dominated	—

Data presented as mean (standard error) after 100 bootstrap samplings. EGFR: epidermal growth factor receptor; ICER: incremental cost-effectiveness ratio; QALE: quality-adjusted life expectancy; QALY: quality-adjusted life year; TKI: tyrosine kinase inhibitor

Sensitivity analysis

Fig 2 shows the cost-effectiveness scatter plots. A Monte Carlo simulation with 1,000 iterations was conducted to construct the acceptability curves (Fig 3A), which show erlotinib had a probability of 58.6% being cost-effective at a cost-effectiveness threshold of US$24,408 (1 GDP per capita of Taiwan in 2017) / life year. If the willingness-to-pay threshold was set at US$24,408 / QALY (Fig 3B), the probability became 78.9%.

Fig 2

Cost-effectiveness scatter plots using (A) life expectancy and (B) QALE.

Individual dots represent results after 1,000 iterations. QALE: quality-adjusted life expectancy; QALY: quality-adjusted life year.

Fig 3

Acceptability curves of cost-effectiveness thresholds using (A) US$/life year, and (B) US$/QALY.

The dash line represents a threshold of US$24,408 (1 GDP per capita of Taiwan in 2017) / life year or QALY. GDP: gross domestic product; QALY: quality-adjusted life year.

We also examined the overall results of the 379 patients before propensity score matching (Table 3).

Table 3

Costs, effectiveness, and ICER of the 379 patients before propensity score matching.

	Afatinib	Erlotinib	Gefitinib
	n = 71	n = 57	n = 251
Costs, US$
Costs in progression-free period	44,785 (3,681)	30,967 (2,813)	29,668 (1,177)
EGFR-TKIs	24,992 (2,015)	18,234 (1,368)	16,316 (757)
Other than EGFR-TKIs	19,781 (1,850)	12,736 (1,737)	13,354 (723)
Lifetime costs	78,612 (6,046)	62,057 (3,197)	52,812 (1,343)
EGFR-TKIs	41,823 (3,097)	35,737 (2,255)	26,185 (895)
Other than EGFR-TKIs	36,859 (2,852)	26,338 (2,151)	26,628 (1,012)
Chemotherapies	11,481 (1,101)	5,106 (797)	10,040 (588)
Lifetime psychometric score, score year
Physical	44.5 (1.4)	44.9 (1.9)	37.2 (0.9)
Pain	13.1 (0.5)	13.0 (0.6)	11.1 (0.3)
Sleep	10.1 (0.4)	10.5 (0.5)	8.7 (0.2)
Psychological	42.6 (1.5)	42.8 (1.7)	36.4 (0.8)
Bodily appearance	10.8 (0.5)	10.9 (0.5)	9.4 (0.2)
Negative feelings	11.6 (0.4)	11.5 (0.5)	10.2 (0.2)
Effectiveness
Life expectancy, life year	3.48 (0.12)	3.16 (0.14)	2.79 (0.05)
QALE, QALY	2.61 (0.10)	2.50 (0.12)	2.17 (0.05)
ICER
ΔCost /ΔLife expectancy	47,765 (26,544)	24,769 (2,028)	—
ΔCost /ΔQALE	156,385 (55,500)	31,506 (15,760)	—

Data presented as mean (standard error) after 100 bootstrap samplings. EGFR: epidermal growth factor receptor; ICER: incremental cost-effectiveness ratio; QALE: quality-adjusted life expectancy; QALY: quality-adjusted life year; TKI: tyrosine kinase inhibitor

Discussion

Most studies analyzed the cost-effectiveness of EGFR testing [25-29] or EGFR-TKI versus chemotherapy as first-line treatment [30-34]. Direct comparisons between different first-line EGFR-TKIs, however, have been performed less frequently. Using OPTIMAL and IPASS trials, Lee et al. found that the cost per QALY gained for erlotinib versus gefitinib was US$62,419; however, their approach applied an indirect treatment comparison [5]. Similarly, from the experiences of EURTAC and LUX-Lung 3 trials, Ting et al. calculated an ICER value of $61,809/QALY for erlotinib versus afatinib via an indirect approach [6]. Recently, Chouaid et al. directly compared the cost-effectiveness of afatinib versus gefitinib using LUX-Lung 7 data [7]. Nevertheless, all these studies applied model construction to compare the cost per life year or cost per QALY of 2 EGFR-TKIs [5-7]. In contrast, our study, based on real-world data, directly assessed the lifetime survival, QoL, and medical costs of 3 different treatments along the follow-up courses to estimate the life expectancy, QALE, lifetime psychometric scores, and lifetime costs (Fig 1 and Fig D in S1 Appendix). Our analysis requires fewer assumptions, and the effectiveness and cost-effectiveness estimates produce figures much closer to reality. This study was limited to advanced NSCLC patients with EGFR mutations, and all patients concomitantly using other first-line therapies were excluded. Although instrumental variables can control unobserved covariates [35], such a variable may not easily be found in all study settings. Meta-analysis pooling randomized trials data can avoid confounding [36], but the inclusion and exclusion criteria may limit its application in every day practice. Thus, our method matching each group with propensity scores based on real-world data could still be a viable alternative in providing useful information for allocation of limited resources. We thus tentatively concluded that erlotinib appeared to be cost-effective (Table 2).

Similar to the costs in the progression-free period, more than half of the lifetime costs were attributable to EGFR-TKIs because erlotinib and gefitinib could still be used as the subsequent treatment. That is, the costs of EGFR-TKIs were a major determinant of cost-effectiveness. During the study period of 2011–2017, generic drugs of gefitinib were not yet available in Taiwan. If the price of afatinib had been reduced 25% to match that of gefitinib (Table 1), it would not have been strongly dominated. Interestingly, the costs of chemotherapies were different among the three groups. However, whether these cost differences are related to the uneven use of subsequent osimertinib remains unknown. Since the out-of-pocket costs of subsequent osimertinib were not recorded in our database, we reviewed the medical records of each patient. Although there was no uneven frequency distribution of the drug use (Table A in S1 Appendix [37]: 18.8%, 18.8%, and 16.7% in the afatinib, erlotinib, and gefitinib groups, respectively), the frequencies of positive serum/tissue T790M mutation differed. In contrast to the erlotinib group (18.8%), the afatinib (12.5%) and gefitinib (14.6%) groups had fewer patients with T790M mutation after first-line therapies. Patients without T790M mutation had a shorter progression-free survival using osimertinib [37]; namely, the cumulative use and additional costs of osimertinib would be less. However, after adjusting the additional costs, erlotinib still remained cost-effective.

In accordance with our previous report [12], lower lifetime psychometric scores measured with WHOQOL-BREF for patients receiving afatinib were observed. Lower lifetime scores for pain, bodily appearance, and negative feelings in the afatinib group might result from more severe paronychia, folliculitis, and diarrhea related to afatinib. Nevertheless, more subsequent chemotherapy treatments with increased adverse events in the afatinib group might also contribute to the results. The differences of QALE among three groups were modest, a lower QALE in the afatinib group was not observed.

Several limitations must be acknowledged in this study. First, the work was done in a tertiary hospital in Taiwan, generalizing the results must be cautious. Besides, the spending details on medical services were obtained from the NCKUH database. Because patients might incur expenses outside the hospital, the costs we calculated constitute a conservative estimate. However, we compared the NHI-reimbursed costs in our hospital with the total charges recorded in the NHI database and found that the former accounted for more than 80% of the latter [16], indicating a small bias at most. Second, patients in this study were generally younger and had a better performance status. Thus, the progression-free survival and overall survival would be longer than those excluded, which might lead to an overestimate of effectiveness. Nevertheless, since patients who live longer incur more costs and probabilistic sensitivity analyses accounting for uncertainties showed consistent results, we believe that the estimates would not be overly biased. Third, unobserved prognostic factors including smoking were not considered into propensity score matching because of the lack of data. However, most of EGFR mutation-positive NSCLC patients in Asia (including Taiwan) are never smokers [1], the results would not be biased too much. Fourth, because it is difficult to measure QoL scores close to death, we hypothesized the values beyond the follow-up period to be the same as that near the end of follow-up, which might lead to an over-estimation of QALE. Moreover, numbers of subjects in the afatinib and erlotinib groups were smaller than those receiving gefitinib. Consequently, mean QoL scores and costs after 2 years of follow-up were more easily influenced by outliers. However, as the survival rates after 2 or more years would be low, the bias would not be too big to affect the inference.

This real-world analysis directly compared the effectiveness and cost-effectiveness of three first-line EGFR-TKIs. Erlotinib appeared to be cost-effective from payer’s perspective. Lifetime psychometric scores may provide additional information for effectiveness evaluation.

Supplemental figures and tables to accompany the primary results.

(DOC)

Click here for additional data file.

Anonymized data set for the study findings.

(XLSX)

Click here for additional data file.

(PDF)

Click here for additional data file.

2 Jan 2020

PONE-D-19-29926

Comparative effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors: Analysis of real-world data

PLOS ONE

Dear Dr. Yang,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

In addition to addressing 2 reviewers' comments please address the following comments:

1. Please add Cost-effectiveness Planes (Figures) using your ICERs

2. Please add ICER results in your abstract

3. Please also interprete ICERs in your results section.

We would appreciate receiving your revised manuscript by Feb 16 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Khurshid Alam, Ph. D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at http://www.plosone.org/attachments/PLOSOne_formatting_sample_main_body.pdf and http://www.plosone.org/attachments/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3.Please include your tables as part of your main manuscript and remove the individual files. Please note that supplementary tables (should remain/ be uploaded) as separate "supporting information" files

4. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Cost-effectiveness of drugs using real-world data is truly the area of interest for many researchers/physicians and payers.

However, when using real-world data, the followings must be addressed in the paper to ensure reliability and transparency ,and not to mislead the readers.

Therefore, I suggest to make 3 major revisions and 5 minor revisions for this great work by the team in Taiwan.

Major revisions

・Lack of generalizability must be addressed in Discussion and you need to change the Title to honestly indicate that this work is done in a single tertiary hospital in Taiwan.

・Methods has to be fully revised as there is lack of information. Especially, what information other than QOL and survival are collected prospectively for this research purpose. If you are utilizing the information from EMR which was collected as a part of routine care, it is fine but you need to mention it and how those additional information are abstracted from EMR. It might be helpful to just attach the study protocol approved by IRB before commencement as a supplemental document if you wish to avoid re-writing Method entirely.

・Information is missing for how you handled missing data in QOL or data abstracted from other source which needs to be implemented in either body or supplemental document. Moreover, information about how you summarized QOL data in Figure 1 while every patient have at different timing of visit in real-world setting need to be included.

Minor revisons

・In section of Effectiveness, it is written that the authors determined disease progression based on RECIST v1.1. I wounder if this is true because disease progression is not strictly assessed based on RECIST criteria but also by other evidence such as patient symptoms in real-world setting. Some may decide the switching of therapy due to adverse event too. If the authors wish to mention that the authors reevaluated disease progression by independent review, there needs to be clarification.

・Please make a change to supplemental figure 1 in order to specify how you identify 379 patients from 729 patient overall. you need to clarify the reason of exclusion as it is not clear by just mentioning the detailed information of QOL and cost was provided by 379 patients. (what do you mean by the detailed information?)

・Table 1 has to be explained in detail why these clinical characteristics are chosen for propensity score matching. Smoking status is well known prognostic factor in NSCLC. If information was not collected, you have to mention in Discussion that not all known prognostic factors are adjusted in this research. Moreover, clarification is needed for why these certain comorbidities are selected even though there are several other comorbidities which may affect QOL of NSCLC patient treated with EGFR-TKI such as skin rash.

・It is better to mention each drug costs in Taiwan as Table 2 only show the total healthcare expenditure as you mention afatinib is too expensive in Discussion. Moreover, availability of generic drugs for gefitinib or erlotinib in Taiwan can be included in Discussion.

・Figure 1 is key result for this research, therefore, need to be explained in details. Especially, there seems sudden increase in cost and QOL for afatinib and erlotinib. Also, there may be a over prediction of QoL when reaching to survival probability close to 0.

Reviewer #2: PONE-D-19-29926: Comparative effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors: Analysis of real-world data

General comments

The research was innovative on methodology in two approaches. First, it used a semi-parametric approach to extrapolate the overall survival (OS) over lifetime. Second, the study patients were matched individually on propensity scores.

The authors claimed the study was based on the real-world dataset of Taiwanese patients that compared treatment effectiveness directly across three TKIs. Regarding the first approach, the OS relied on a statistical model of data projection as the study patients were not followed exhaustively through lifetime. The study justified this approach adequately by citing previous studies for validity of the employed model.

Regarding the second approach, comparability across patients which was based on the propensity-score matching (PSM) across the three TKI groups have not been adequately addressed. To make the three groups comparable, the PSM handles selection bias through the observed covariates. This was justified through balances in the selected 7 covariates shown in Table 1.

The PSM, however, could not control for the unobserved prognostic factors which can lead to omitted-variable bias. An instrumental-variable (IV) technique is a viable alternative to randomization and probably is better than PSM. First, the study should provide adequate argument on how well PSM performed as compared to meta-analysis (direct and indirect treatment comparisons) based on RCTs and pseudo-randomization technique like IV.

The second issue for the second approach is on the study-excluded and included patients. There were 729 patients in total that received the first-line TKIs. Then 350 patients were excluded due to data availability. The remaining 379 patients (52% of 729) were subject to the PSM, where only another 50% (192/379 patients) were successfully matched. The study should shed some lights on the excluded 537 patients in total (74% of 729) that were eligible but did not contribute the analytic dataset.

Other comments about terminology used and reporting/presentation styles are as follows.

Specific comments

Abstract

Introduction:

1. Page 3: “… would improve patients’ values.” The term “patients’ values” is ambiguous; the values on which exact aspects should be pointed out: quality of life, clinical benefits, values for money, or else? This should be more specific.

Main text

Introduction:

2. Page 5: “Randomized trials have shown similar efficacy of these three agents [3, 4].” Reference #3 which was not relevant should not be cited in the present study since it was conducted in patients mixed between 1st line and 2nd line treatments. Reference #4 showed a statistically significant difference in the efficacy measured by PFS, in which afatinib was better than gefitinib (despite non-significant difference in OS because the study was not mature yet).

3. Page 5: “To improve patients’ values, the comparative effectiveness …” The term “patients’ values” again is ambiguous. How the patients’ values could be improved by exploration of the comparative effectiveness and cost-effectiveness of TKIs was unclear?

4.Page 5: “When the cost-effectiveness is considered from the payers’ perspective, real-world data seem

more credible.” Credibility on the baseline survivals or credibility on the comparative effectiveness and cost-effectiveness? The real-world data based on local population tend to be useful for generating the baseline outcomes for the comparator which is least effective such as placebo or chemotherapy. When comparing across innovative drugs (TKIs), the data should be based on the “randomized, controlled” studies, the level-1 evidence. Please provide the references on the evidence that real-world data are more credible from the payer’s perspective.

Methods:

5. Page 6: “This study was approved by the International Review Board of …” This is typo? It should say the Institutional Review Board of …

6. Page 6: “Among 1,828 enrolled cases, 379 patients with EGFR mutation-positive advanced NSCLC under first-line TKIs were abstracted for analysis.”

There were 350 patients not included in the study (Supplementary Table 1, the rightmost column) and 379 patients had detailed information of QoL and costs (shown in Supplementary Figure 1), of which 192 patients were successfully matched. The total of 729 (379+350) patients were relevant to the study and should be mentioned as the enrolled cases instead of all invited lung cancer cases (1,828 patients), which some of them did not receive the EGFR-TKIs.

Otherwise, Supplementary Figure 1 should show the patient flow diagram beginning with a total of 1,828 patients invited, followed by those 1,099 patients who did not receive TKIs were excluded, then 350 patients without information were excluded and 379 remained in the dataset.

The study should elaborate on (1) what were the reasons they were excluded? (2) How did the excluded patients look like and were they much different in characteristics from those included in the study?

To answer (2), the excluded 537 patients in total (in Table 1) should be divided further into two columns separately between 350 patients first excluded (as in Supplementary Table 1) and 187 patients without PSM.

Propensity score matching:

7. Page 6” “To minimize selection bias and ensure better comparability among three different EGFR-TKIs, we used propensity score matching via a greedy algorithm.”

The term “selection bias” should be replaced with more specific terms that the propensity score matching (PSM) aims for and can handle. For example, the PSM is only able accounted for the observed covariates that have been used for determining the probability of obtaining a treatment. However, it is not able to handle selection bias due to the “omitted variables”.

The term “greedy algorithm” should be elaborated technically. For example, how is this method different from conventional approaches in calculating the propensity scores, such as probit or logistic regressions? Does the algorithm perform well for the not so large sample size (379 patients in this study).

QoL and QALE-QALYs

8. Page 7: As the research used both EQ-5D and WHOQOL-BREF for QoL measures, it should be point out which specific measure was used for adjusting the survival years. What are the reasons to analyze the lifetime QoL scores if they were not used for estimating QALE or QALYs.

Otherwise, the study should propose in the beginning the lifetime QoL scores as one of the objectives, say the patients’ values.

Medical costs

9. Page 8: Data on the medical costs were collected until December 2017 for calculating monthly expenditures per case. However, the censor date for survival was in September 2018. Were the medical costs occurring in patients survived during December 2017 to September 2018 not accounted for?

10. Since costs of the TKIs per se played an important role, especially for afatinib in the lifetime costs as shown in the study results. The data on the unit price of each TKI along with the descriptive statistics of utilities per TKI should be shown as the baseline parameters.

Probabilistic sensitivity analysis (PSA)

11. Page 9: “We assumed a gamma distribution for life expectancy, QALE, and costs, …” As healthcare costs are known for very skewed distribution, assuming a gamma distribution of cost data is well accepted. However, a gamma distribution for life expectancy, QALE is not familiar with. References for justifying the distribution of LE and QALE would be helpful.

12. Page 9: “A Monte Carlo micro-simulation with 1,000 iterations was conducted to construct acceptability curves…” A Monte Carlo micro-simulation is too general to be mentioned for the PSA. There were three TKI options to be compared at the same time on cost-effectiveness measures. Specific approaches, either net monetary benefit or net health benefit used for determining the most cost-effective option among the three should be mentioned instead.

Results:

13. The study should present descriptive statistics in the beginning of the Eesults on the unit price of each TKI along with the utilities per TKI as the baseline parameters.

Base-case analysis

14. Table 2: The results on CER should be omitted as the “average ratio” has no implications for policy decision and can create confusions if they were contradicted to the incremental analysis.

Instead, the results on ICER should be calculated hierarchically either by an increasing order in the total costs or by an increasing or in the total effectiveness (LYs, QALE, QALYs).

Based on an increasing order of QALYs, the first ICER should be calculated for afatinib (vs. gefitinib). As erlotinib was more effective and less costly than afatinib, there was no need to calculate the ICER for erlotinib (vs. afatinib) because erlotinib was cost-saving in lay language or technically economic dominance. In this case, it should not say afatinib was least cost-effective. Instead, it should say afatinib was not cost effective since it was just simply dominated by erlotinib and should be dropped totally for the cost-effectiveness comparison.

Then, the ICER for erlotinib (vs. gefitinib) could be calculated and compared with Taiwan’s GDP per capita. In this case, the ICER of erlotinib (vs. gefitinib) was below Taiwanese GDP per capita, hence, the study should conclude that erlotinib was cost effective.

Sensitivity analysis

15. Page 11: Since the PSA compared all three TKIs mutually at the same time, Supplementary Figure 4 should not do a pairwise-plot. Instead, it should plot the scatters for all three TKIs in the same graph by using total cost and total QALYs of each TKI rather than using the incremental values by specifying any TKI as the comparator. In addition, there is no need to draw the CE threshold of Taiwan at this stage.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: comments on PONE-D-19-29926.docx

Click here for additional data file.

30 Jan 2020

Please kindly refer to the Response to Reviewers.

Submitted filename: Response to Reviewers.doc

Click here for additional data file.

19 Feb 2020

PONE-D-19-29926R1

Comparative effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors: Analysis of real-world data in a tertiary hospital in Taiwan

PLOS ONE

Dear Dr. Yang,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Apr 04 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Khurshid Alam, Ph. D.

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: As for the my previous comment 5, you included the detailed explanation regarding the reason why 350 patients were excluded from this research in the manuscript and supplemental figure 1.

If the reason for exclusion of 346 out of 350 patients is no informed consent obtain for this research, then it is not appropriate to use their data and summarize in Table 1 from the ethical point of view.

Please delete all relevant data from patients without signed informed consent in this manuscript.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

20 Feb 2020

Please kindly refer to the Response to Reviewers. Thank you!

Submitted filename: Response to Reviewers_1.doc

Click here for additional data file.

24 Mar 2020

Comparative effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors: Analysis of real-world data in a tertiary hospital in Taiwan

PONE-D-19-29926R2

Dear Dr. Yang,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Khurshid Alam, Ph. D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I believe all the comments have been addressed. I have no objection to accept this manuscript for PLOS ONE journal.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

25 Mar 2020

PONE-D-19-29926R2

Comparative effectiveness and cost-effectiveness of three first-line EGFR-tyrosine kinase inhibitors: Analysis of real-world data in a tertiary hospital in Taiwan

Dear Dr. Yang:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Khurshid Alam

Academic Editor

PLOS ONE