Advanced Lung Cancer Patients’ Use of EGFR Tyrosine Kinase Inhibitors and Overall Survival: Real-World Evidence from Quebec, Canada

EGFR tyrosine kinase inhibitors (EGFR-TKIs) are breakthrough palliative treatments for advanced lung cancer patients with tumors harboring mutations in the EGFR gene. Using healthcare administrative data, three cohorts were created to describe the use of three EGFR-TKIs that are publicly funded in Quebec for specific indications (i.e., 1st-line gefitinib, 1st-line afatinib, and post-EGFR-TKI osimertinib). The main objective was to compare overall survival (OS) among patients receiving these treatments to those in previous experimental and real-world studies. The patients who received EGFR-TKIs for indications of interest between 1 April 2001, and 31 March 2019 (or 31 March 2020, for post-EGFR-TKI osimertinib) were included to estimate the Kaplan-Meier-based median OS for each cohort. An extensive literature search was conducted to include comparable studies. For the gefitinib 1st-line (n = 457), the afatinib 1st-line (n = 80), and the post-EGFR-TKI osimertinib (n = 119) cohorts, we found a median OS (in months) of 18.9 (95%CI: 16.3–21.9), 26.6 (95%CI: 13.7-NE) and 19.9 (95%CI: 17.4-NE), respectively. Out of the 20 studies that we retained from the literature review and where comparisons were feasible, 17 (85%) had similar OS results, which further confirms the value of these breakthrough therapies in real-world clinical practice.


Introduction
With about half of all cases being diagnosed at stage IV [1], lung cancer is the deadliest cancer in Canada, including Quebec [2]. Patients diagnosed with lung cancer in Canada between 2015 and 2017 had a net survival rate of 22% [2]. Only 24% of all cases in Quebec diagnosed between 2014 to 2016 survived 5 years or more from any cause of death [3].
Patients with advanced cancer (i.e., locally advanced, or metastatic) with limited curative treatment options are typically given palliative treatment with chemotherapy, immunotherapy, or targeted therapy, the latter two being breakthrough therapies introduced in the past decade. Since 2011 in Quebec, oral EGFR tyrosine kinase inhibitors (EGFR-TKI) have been a standard targeted treatment for patients with tumors harboring EGFR-TKI sensitizing mutations, which include resistance mutations [4,5]. In Ontario, up to 20% of advanced patients with non-squamous, non-small-cell lung cancer have the most common EGFR-TKI sensitizing mutations (exon 19 deletions and exon 21-L858R) [6]. About 50-65% of patients progressing on first-and second-generation EGFR-TKIs develop an EGFR-TKI resistance mutation (T790M) that can be treated with third-generation EGFR-TKIs [7,8].

Study Design and Data
Our study included patients who had public coverage for prescriptions filled at community pharmacies in Quebec and were treated with EGFR-TKIs for indications included in Quebec's drug formulary (Table 1): gefitinib or afatinib for 1st-line palliative treatment, or osimertinib as a second EGFR-TKI treatment following 1st-line palliative treatment with an EGFR-TKI. Since the public coverage of EGFR-TKIs requires a treating physician to confirm their patient's status as advanced cancer with an activating EGFR mutation, payments of these drugs by the public insurer allowed us to infer this status for patients in our study. We did not investigate EGFR-TKIs that were (1) not in Quebec's drug formulary (e.g., dacomitinib), (2) in Quebec's drug formulary for indications included after our recruitment period (e.g., osimertinib for 1st-line treatment), and (3) in Quebec's drug formulary without EGFR mutation status as an indication and reimbursement criterion (e.g., erlotinib for 2nd-or 3rd-line treatments).
Inpatient and outpatient care administrative data (i.e., hospitalization [MED-ECHO], physician billing [SMOD], drugs dispensed in community pharmacies [SMED], health insurance registry [FIPA]) that is managed by the Régie de l'assurance maladie du Québec (RAMQ) and the Ministère de la Santé et des Services sociaux (MSSS) were used to create three EGFR-TKI cohorts: (1) 1st-line gefitinib, (2) 1st-line afatinib, and (3) post-EGFR-TKI osimertinib. For all cohorts, we selected patients who had their first lung cancer diagnostic code in MED-ECHO or SMOD between 1 April 2001, and 31 March 2019, (lung cancer index date) [3] with at least one gefitinib, afatinib, or osimertinib prescription in the same period. We also included in the osimertinib cohort patients with at least one osimertinib prescription before 31 March 2020, after receiving gefitinib, afatinib, or erlotinib between 1 April 2001, and 31 March 2019. International nonproprietary names and drug identification numbers were used to identify EGFR-TKI prescriptions in SMED (Appendix A). All patients were followed until death (captured in FIPA), or 31 March 2020, whichever came first. Table 1. Indications for gefitinib, afatinib, and osimertinib included in the drug formulary of Quebec's universal public drug coverage program [15].

EGFR-TKI Line of Palliative Treatment Indication and Coverage Criteria Date Added to Formulary
Gefitinib 1st line 1st-line treatment of patients with locally advanced or metastatic NSCLC, with an activating EGFR mutation, and with an ECOG performance status of 0 to 2.

November 2011
Afatinib 1st line In monotherapy for the 1st-line treatment of patients with metastatic NSCLC, with an activating EGFR mutation, and with an ECOG performance status of 0 or 1.

May 2016
Osimertinib 2nd line or more Treatment of locally advanced, unresectable, or metastatic NSCLC, with the T790M EGFR mutation in patients: • Whose disease progressed during or following treatment with an EGFR-TKI • Whose ECOG performance status is 0 or 1.
November 2018 1st line 1 1st-line treatment of people with locally advanced, unresectable, or metastatic NSCLC, with an activating EGFR mutation, and with an ECOG performance status of 0 to 1.

December 2018
Abbreviations: NSCLC: non-small-cell lung cancer; ECOG: Eastern Cooperative Oncology Group. 1 EGFR-TKI for this indication was not studied.
We then excluded from each cohort patients that did not receive an EGFR-TKI for the line of treatment of interest. Since we had limited information on the line of palliative treatment and chemotherapy history, we developed an algorithm to identify the line of therapy associated with patients' 1st EGFR-TKI (Appendix B). From each cohort, we first excluded patients who were covered by a public drug insurance plan in FIPA for less than 90% of the period from 3 months before the 1st EGFR-TKI treatment to the end of follow-up (targeted treatment observation period). From the gefitinib and afatinib cohorts, we also excluded patients whose 1st EGFR-TKI was not gefitinib or afatinib, and for the osimertinib cohort, we excluded those whose 2nd EGFR-TKI was not osimertinib after receiving gefitinib, afatinib, or erlotinib. All patients identified by the algorithm as receiving their 1st EGFR-TKI other than 1st-line palliative treatment were then excluded to form the final cohorts.
This study was carried out at INESSS as part of a Health System Impact Fellowship (SQ) of the Canadian Institutes of Health Research. It was part of an initiative to promote the use of real-world data for health technology assessment in line with INESSS's work on real-world evidence in its 2019-2022 Three-Year Business Plan which aims to "Assess the value of interventions in real care settings using medico-administrative data" [16]. INESSS is not responsible for the content of this publication, however, the results reported here were part of a larger INESSS project to study EGFR-TKIs in Quebec [17]. Access to de-identified data for this study was made possible through a tripartite agreement between the MSSS, the RAMQ, and INESSS [18]. Clinical experts were consulted to select relevant intervention codes and develop the algorithm used to identify line of treatment. A scientific advisory committee and INESSS' Comité de l'évolution des pratiques en oncologie (CEPO), a panel of experts in cancer (hemato-and radio-oncologists, surgeons and pharmacists), were consulted throughout the study [17].

Outcomes
For each cohort, we reported the annual number of new users and other estimates from the distributions of patient characteristics, total days of the EGFR-TKI supplied, and OS. The date of the 1st EGFR-TKI prescription fill was used to count the number of new users by fiscal year (April to March) and calculate patients' age. The age and sex distributions in each cohort were compared to distributions of a previous population-level lung cancer cohort created by INESSS with Quebec's health administrative databases [3]. For the post-EGFR-TKI osimertinib cohort, we reported the EGFR-TKI that was also received as 1st-line treatment. The total number of days' supply of EGFR-TKI per patient was obtained by adding the treatment duration (in days) of each EGFR-TKI prescription reimbursed by the RAMQ.
Patients' overall survival time was calculated as months elapsed between patients' first EGFR-TKI prescription of interest and the date of death from any cause. Administrative censoring occurred on 31 March 2020. Patients' follow-up in the 1st-line gefitinib and afatinib cohorts was also censored at the time of receiving osimertinib, if applicable; osimertinib increases OS relative to standard chemotherapy when given as a subsequent line after another EGFR-TKI [19,20]. Censoring at the time of receiving osimertinib was applied to compare our survival results to studies in which patients did not receive osimertinib as a subsequent therapy, including studies submitted to INESSS for evaluation of gefitinib and afatinib.

Statistical Analyses
Frequencies and proportions, means and standard deviations, or medians and ranges (minimum to maximum), were reported for distributions of age, sex, the 1st-line EGFR-TKI drug, and new users. Boxplots were used to display the distribution of total days' supply. Median follow-up time was estimated with the reverse Kaplan-Meier method in the prodlim package in R [21], and the survival package was used to plot Kaplan-Meier survival curves and estimate median overall survival times [22].

Comparison of Overall Survival
Our study could not include a comparator group of patients with EGFR mutated tumors who were not treated with EGFR-TKIs since we did not have information on mutation status, and more importantly, in the real-world clinical setting, it would be unethical to withhold EGFR-TKI treatment in the presence of an activating EGFR mutation. Therefore, we relied on the indirect comparison of the median OS in our cohorts to that of cohorts with similar EGFR-TKI use in previous studies. In May 2021, we conducted an extensive literature review to include all published experimental trials and real-world studies that reported OS estimates related to the three EGFR-TKIs for the indications of interest (Appendix C). The studies submitted to INESSS for evaluation of these treatments that reported estimates of median OS were automatically included in our review [20,[23][24][25][26][27][28].
The similarity between our results and those of other studies (point estimates and 95% confidence intervals) was assessed with the degree of overlap that took into consideration the absolute difference between point estimates and the width of the confidence intervals, but without any formal statistical tests.

Results
Among 552 patients who received gefitinib, 457 (82.8%) were included in the 1st-line gefitinib cohort. Out of 117 patients receiving afatinib, 80 (68.4%) were included in the 1st-line afatinib cohort. Similarly, of 178 patients who received osimertinib, 119 (66.9%) were included in the post-EGFR-TKI osimertinib cohort ( Figure 1). New users of 1st-line palliative treatments with gefitinib or afatinib started to accrue in the same year as their listing in Quebec's drug formulary in November 2011 and May 2016, respectively. In contrast, 32 patients in the osimertinib cohort received the drug in the 2 years before it was listed on the formulary in November 2018, when the highest numbers of users were observed ( Figure 2 and Table 1). line afatinib cohort. Similarly, of 178 patients who received osimertinib, 119 (66.9%) were included in the post-EGFR-TKI osimertinib cohort ( Figure 1). New users of 1st-line palliative treatments with gefitinib or afatinib started to accrue in the same year as their listing in Quebec's drug formulary in November 2011 and May 2016, respectively. In contrast, 32 patients in the osimertinib cohort received the drug in the 2 years before it was listed on the formulary in November 2018, when the highest numbers of users were observed ( Figure 2 and Table 1).  Compared to lung cancer patients in Quebec, there was a larger proportion of women in all three cohorts (60-71% vs. 49%) ( Table 2 and Appendix D). The median age for the gefitinib cohort was the same as the overall cohort (71 years), whereas the median age for the afatinib cohort was lower (68 years) and that for the osimertinib cohort was higher (72 years). There was a much smaller portion of patients that were 80 years and over in the afatinib cohort in comparison to the overall cohort (7.5% vs 20.6%).   Compared to lung cancer patients in Quebec, there was a larger propor in all three cohorts (60-71% vs. 49%) ( Table 2 and Appendix D). The med gefitinib cohort was the same as the overall cohort (71 years), whereas the the afatinib cohort was lower (68 years) and that for the osimertinib cohort years). There was a much smaller portion of patients that were 80 years a afatinib cohort in comparison to the overall cohort (7.5% vs 20.6%). Compared to lung cancer patients in Quebec, there was a larger proportion of women in all three cohorts (60-71% vs. 49%) ( Table 2 and Appendix D). The median age for the gefitinib cohort was the same as the overall cohort (71 years), whereas the median age for the afatinib cohort was lower (68 years) and that for the osimertinib cohort was higher (72 years). There was a much smaller portion of patients that were 80 years and over in the afatinib cohort in comparison to the overall cohort (7.5% vs 20.6%). The median total days' supply was 300 days (9.9 months) for gefitinib and osimertinib, and 274.5 days (9.0 months) for afatinib ( Figure 3). Approximately 5% of patients had a total days' supply of gefitinib that was greater than 3 years. Only 1 patient received a supply of afatinib for more than 3 years, while no one received a supply of osimertinib for this length of time.  The median total days' supply was 300 days (9.9 months) for gefitinib and osimertinib, and 274.5 days (9.0 months) for afatinib ( Figure 3). Approximately 5% of patients had a total days' supply of gefitinib that was greater than 3 years. Only 1 patient received a supply of afatinib for more than 3 years, while no one received a supply of osimertinib for this length of time.   With a median follow-up of 29.6 months in the gefitinib cohort, we observed 295 (65%) deaths and a median OS of 18.9 months (95%CI: 16.3-21.9) (Figure 4). The afatinib cohort had a median follow-up of 17.6 months, 36 (45%) deaths and a median OS of 26.6 months (95%CI: 13.7-NE). With a median of 17.3 months of follow-up after receiving osimertinib in the post-EGFR-TKI osimertinib cohort, we observed 44 (38%) deaths and median OS was 19.9 months (95%CI: 17.4-NE). The upper bounds of the 95%CI for median OS in the afatinib and osimertinib groups were non-evaluable (NE) due to the immaturity of the data (i.e., less than 50% of patients in the cohort had the outcome).
Curr. Oncol. 2022, 29, FOR PEER REVIEW 7 OS in the afatinib and osimertinib groups were non-evaluable (NE) due to the immaturity of the data (i.e., less than 50% of patients in the cohort had the outcome).  From 2565 articles identified, we selected 10 RCTs (9 articles) and 17 real-world studies (18 articles) for indirect comparison with our OS results (Appendix C). We retained 7 RCTs [26,27,[29][30][31][32][33] and 10 real-world studies [34][35][36][37][38][39][40][41][42][43] on gefitinib as 1st-line treatment ( Figure 5). Among these studies, 5 RCTs and 6 real-world studies had median OS estimates similar to ours, 1 RCT [30] and 1 real-world study [40] had results superior to ours. We were limited in making comparisons with the estimates from other studies, which did not overlap our results and lacked confidence intervals. For afatinib as 1st-line treatment we retained 2 RCTs [28], and 1 study combining these RCTs [28], and 3 realword studies [34,44,45]. All studies had similar results to ours, except for 1 real-world study for which we were limited in making a comparison due to a lack of overlap of results and the non-evaluable upper bound for our estimate's 95%CI. We retained only 1 RCT [20] and 5 real-world studies [46][47][48][49][50] on osimertinib as a post-EGFR-TKI treatment. One real-world study [49] had a median inferior to ours and two others [47,50] had results similar to ours. Due to a lack of overlap of results and the non-evaluable upper bound of our estimate's 95%CI, our comparisons with the other 2 real-world studies [46,48] were limited. Similarly, we were limited in the comparison with the single RCT, which had a longer median OS. confidence interval is non-evaluable. The mustache around each estimate represents the 95% confidence interval. The orange or green square with a blue outline represents a median that is considered similar to Quebec's median.

Discussion
In our study, the annual number of new users of gefitinib always exceeded the number of new patients using the second-generation EGFR-TKI afatinib as a 1st-line treatment. However, a decline in the use of gefitinib relative to afatinib was noted in the From 2565 articles identified, we selected 10 RCTs (9 articles) and 17 real-world studies (18 articles) for indirect comparison with our OS results (Appendix C). We retained 7 RCTs [26,27,[29][30][31][32][33] and 10 real-world studies [34][35][36][37][38][39][40][41][42][43] on gefitinib as 1st-line treatment ( Figure 5). Among these studies, 5 RCTs and 6 real-world studies had median OS estimates similar to ours, 1 RCT [30] and 1 real-world study [40] had results superior to ours. We were limited in making comparisons with the estimates from other studies, which did not overlap our results and lacked confidence intervals. For afatinib as 1st-line treatment we retained 2 RCTs [28], and 1 study combining these RCTs [28], and 3 real-word studies [34,44,45]. All studies had similar results to ours, except for 1 real-world study for which we were limited in making a comparison due to a lack of overlap of results and the non-evaluable upper bound for our estimate's 95%CI. We retained only 1 RCT [20] and 5 real-world studies [46][47][48][49][50] on osimertinib as a post-EGFR-TKI treatment. One real-world study [49] had a median inferior to ours and two others [47,50] had results similar to ours. Due to a lack of overlap of results and the non-evaluable upper bound of our estimate's 95%CI, our comparisons with the other 2 real-world studies [46,48] were limited. Similarly, we were limited in the comparison with the single RCT, which had a longer median OS.
Curr. Oncol. 2022, 29, FOR PEER REVIEW 7 OS in the afatinib and osimertinib groups were non-evaluable (NE) due to the immaturity of the data (i.e., less than 50% of patients in the cohort had the outcome). From 2565 articles identified, we selected 10 RCTs (9 articles) and 17 real-world studies (18 articles) for indirect comparison with our OS results (Appendix C). We retained 7 RCTs [26,27,[29][30][31][32][33] and 10 real-world studies [34][35][36][37][38][39][40][41][42][43] on gefitinib as 1st-line treatment ( Figure 5). Among these studies, 5 RCTs and 6 real-world studies had median OS estimates similar to ours, 1 RCT [30] and 1 real-world study [40] had results superior to ours. We were limited in making comparisons with the estimates from other studies, which did not overlap our results and lacked confidence intervals. For afatinib as 1st-line treatment we retained 2 RCTs [28], and 1 study combining these RCTs [28], and 3 realword studies [34,44,45]. All studies had similar results to ours, except for 1 real-world study for which we were limited in making a comparison due to a lack of overlap of results and the non-evaluable upper bound for our estimate's 95%CI. We retained only 1 RCT [20] and 5 real-world studies [46][47][48][49][50] on osimertinib as a post-EGFR-TKI treatment. One real-world study [49] had a median inferior to ours and two others [47,50] had results similar to ours. Due to a lack of overlap of results and the non-evaluable upper bound of our estimate's 95%CI, our comparisons with the other 2 real-world studies [46,48] were limited. Similarly, we were limited in the comparison with the single RCT, which had a longer median OS. confidence interval is non-evaluable. The mustache around each estimate represents the 95% confidence interval. The orange or green square with a blue outline represents a median that is considered similar to Quebec's median.

Discussion
In our study, the annual number of new users of gefitinib always exceeded the number of new patients using the second-generation EGFR-TKI afatinib as a 1st-line treatment. However, a decline in the use of gefitinib relative to afatinib was noted in the of 95% confidence interval is non-evaluable. The mustache around each estimate represents the 95% confidence interval. The orange or green square with a blue outline represents a median that is considered similar to Quebec's median.

Discussion
In our study, the annual number of new users of gefitinib always exceeded the number of new patients using the second-generation EGFR-TKI afatinib as a 1st-line treatment. However, a decline in the use of gefitinib relative to afatinib was noted in the years following the introduction of afatinib in 2016. In contrast with other EGFR-TKIs, the use of osimertinib began before it was listed in Quebec's drug formulary, which most likely occurred through the "exceptional patient program" that provides coverage for drugs that are not listed in Quebec's drug formulary under exceptional circumstances. These trends indicate physicians' proactiveness in integrating newer generations of EGFR-TKIs into clinical practice.
The high proportion of women (60-71%) we observed in all three of our cohorts is concordant with a previous study reporting that 57% of EGFR mutations in lung cancer tumours are found in women in Canada [6]. Our results are also concordant with the studies we selected from our literature review and one recent single-center study in Quebec that reported 71% of its 1st-line EGFR-TKI users being women [51]. Knowledge of a higher EGFR mutation rate in women could also drive a higher testing rate in women, as seen previously in Canada [52], and further contribute to the higher rates of EGFR-TKI use in women.
In general, all our cohorts included patients above the age of 80 years. Older patients may be less likely to receive cytotoxic chemotherapy due to a lack of clinical data on tolerability, but also due to clinical experience with the elderly who show more complications. In contrast, EGFR-TKIs have better tolerability than chemotherapy [19,32,33,[53][54][55]. Furthermore, in RCTs on EGFR-TKIs, patients older than 80 years (maximum of 89 years) have been included [19,29,31,56,57], and studies on patients of 75 years or older, have reported median OS estimates similar to those of the overall population (19-35.2 months) [58][59][60][61][62][63]. However, we noted that patients in the afatinib cohort, compared to the gefitinib cohort, had a lower proportion of patients of 80 years and older (7.5% vs. 17.5%). Physicians may be hesitant in prescribing afatinib for the elderly due to its higher toxicity profile in comparison to gefitinib [56]. The trend of fewer older patients using afatinib in comparison to gefitinib has also been observed in other real-world studies [34,44,64].
When daily treatment is continuous until disease progression, estimates of total days' supply should resemble time to treatment discontinuation, which is a proxy for PFS [65]. Indeed, the total days' supply of EGFR-TKI in our cohorts were close to the PFS reported in the RCTs we selected from our literature review, and in one recent single-center study in Quebec. The studies reported a PFS of 8.4-11.9 months (median 10.4) for gefitinib as 1st-line treatment [27,29,[31][32][33]51,53,54], and a PFS of 10.1 months for osimertinib as a post-EGFR-TKI treatment [20], while we found medians of total days' supply of 9.9 months for both treatments. Two RCTs reported a PFS of 11.0-11.1 months for afatinib as a 1st-line treatment [28], while we found a total days' supply of 9 months. Total days' supply and PFS may not be equivalent when daily treatment is paused or discontinued for reasons other than progression, such as adverse events. Therefore, the small difference between both parameters that we observed for afatinib as a 1st-line treatment may have been led by its higher toxicity profile.
When indirect comparisons with our OS estimates were possible, we found 3 RCTs submitted to INESSS for evaluation of gefitinib and afatinib (IPASS, LUX-Lung 3, and LUX-Lung 6) to have results similar to ours. Among the 12 other RCT and real-world studies on gefitinib that we were able to indirectly compare with the current study, 10 studies had similar results. The 2 real-world studies on afatinib that we were able to compare with our study also had similar results. Similarly, among the 3 real-world studies on osimertinib that we were able to compare with, 2 had similar results. There were 7 out of 27 studies that reported longer median OS estimates than those in the current study, but these comparisons were limited because of a lack of overlap of results due to unreported or non-evaluable confidence intervals. Despite the uncertainty in these comparisons, it is likely that the OS gap between the RCT of osimertinib (AURA3) and the current study is real. Contrary to AURA3, patients in our post-EGFR-TKI osimertinib cohort may have received chemotherapy in between their 1st-line treatment with an EGFR-TKI and osimertinib as a second EGFR-TKI, which may have reduced the length of survival at the time of receiving osimertinib.
Based on the indirect comparisons, we conclude that the real-world estimates of OS in Quebec related to the use of the three EGFR-TKIs for specific indications are not different from most published studies that we selected from our literature review. This is further supported by one recent population-level study in Ontario that reported results similar to ours: a median OS of 21.6 months (95%CI: 19.3-23.3) for 1st-line gefitinib, and a median OS of 31.0 months (95%CI: 23.4-42.1) for 1st-line afatinib [64].
The limitations of this study include that we could not directly verify whether patients had a non-small cell type histology and an activating EGFR mutation. However, public coverage of EGFR-TKIs is conditional on these characteristics, making it unlikely for patients receiving public coverage of these medicines not to have them. The indirect comparisons in this study are subject to confounding by multiple factors such as age, smoking status, the type of EGFR mutation, and palliative treatments following EGFR-TKIs. For example, we did not include patients that had received EGFR-TKIs through private drug insurance, which likely rendered our cohorts slightly older than the ones in the RCTs submitted to INESSS for evaluations. However, we estimate the proportion of patients with private drug insurance among all EGFR-TKI users to be low since about 70% of lung cancer patients in Quebec are diagnosed at age 65 or older, when 90% of Quebec residents are registered for public drug coverage [3,66,67]. Finally, the survival estimates for the afatinib and osimertinib cohorts are less precise due to data immaturity, and the indirect comparisons of results from these cohorts with other studies imply a greater level of uncertainty.
Our study also has strengths. To our knowledge, this is the first population-level study on the use of EGFR-TKIs for specific indications in a Canadian province where public drug insurance offers full coverage of oral cancer drugs [66]. The involvement of clinical experts and a scientific committee helped us generate real-world evidence on these breakthrough therapies with analyses of Quebec's health administrative data that are grounded in clinical and scientific expertise. Our holistic analytical approach through an extensive literature review also produced more meaningful results.

Conclusions
RCTs have shown that EGFR-TKIs are breakthrough therapies for advanced lung cancer patients. EGFR-TKIs have been included in Quebec's drug formulary for specific indications, mainly based on the efficacy observed in surrogate endpoints of OS. Our study further confirms the value of these treatments in real-world clinical care, with OS rates in Quebec similar to those reported in most real-world studies and RCTs, including most RCTs used in the evaluation of these treatments for public drug coverage. Future studies should re-evaluate EGFR-TKIs with mature and richer data and, given their potential real-world value, further investigate inequities in access to these treatments.

Informed Consent Statement:
This study was carried out as part of a larger evaluation project at INESSS and used de-identified secondary data made available from the tripartite agreement between the MSSS, the RAMQ, and the INESSS; informed consent was not required. Ethics review was requested before submitting to a peer-review journal, and granted by McGill University's Faculty of Medicine and Health Sciences Institutional Review Board on 15 September 2022 (study number: A09-E38-22B).

Data Availability Statement:
Restrictions apply to the availability of these data. Data are available to INESSS through RAMQ servers due to the tripartite agreement.
Acknowledgments: Nicole Bouchard, Alexis Bujold, and Catherine Labbé served as external collaborators for this project. They provided guidance, notably for developing an algorithm using health administrative databases to identify the line of palliative treatment associated with patients' 1st EGFR-TKI treatment.

Appendix B. Algorithm to Identify the Line of Therapy Linked to Patients' 1st EGFR-TKI Treatment
We first excluded patients who had a public drug insurance plan for less than 90% of the time covering 3 months before their 1st EGFR-TKI treatment to the time of death or 31 March 2020, whichever came first (i.e., targeted treatment observation period). This minimized the possibility of missing information on targeted treatments that patients may have received through private coverage. We then established the chronological sequence of all publicly reimbursed targeted therapies received between 1 April 2001, and 31 March 2020, including anaplastic lymphoma tyrosine kinase inhibitors (ALK-TKI). A new line of treatment was considered every time a patient switched targeted therapy drugs, with the possibility of an EGFR-TKI drug being assigned multiple lines.
History of cancer-related treatments (i.e., lung surgery, chemotherapy supervision, and radiotherapy) received between patients' diagnosis date [3] and their 1st EGFR-TKI was retrieved from intervention codes in hospitalization and physician billing databases: MED-ECHO and SMOD. Patients were first grouped into 6 treatment code scenarios, and their 1st EGFR-TKI was further classified as 1st-line palliative treatment or 2nd-line or more of palliative treatment based on the previous use of chemotherapy and its context (curative versus palliative). We considered neoadjuvant chemotherapy, adjuvant chemotherapy, and chemoradiation as curative intent treatments, and, in consultation with medical experts, rules were created to identify these chemotherapies. Patients' 1st EGFR-TKIs were assigned 1st-line treatment only in the absence of palliative chemotherapy. Table A2 and Figure A1 further describe the steps to identify the line of treatment of patients' 1st EGFR-TKI treatment. Table A1 (Appendix A) and Table A3 (Appendix B) provide international non-proprietary names and drug identification numbers used to identify targeted treatments in SMED and Table A4 provides intervention codes used to identify cancer-related treatments in SMOD and MED-ECHO.
More details on this algorithm can be found in a previous report [17]. Table A2. Treatment Scenarios and Rules to Identify Chemotherapy Context.

No chemotherapy codes
When no chemotherapy code is found prior the 1st EGFR-TKI, the latter is considered as 1st-line palliative treatment.

All chemotherapy codes found before patient's last lung resection surgery
Since lung resection is considered a curative treatment (except in some cases), chemotherapy that is received prior to this surgery is also considered a curative treatment and it usually corresponds to neoadjuvant chemotherapy. The EGFR-TKI treatment that follow this chemotherapy is considered 1st-line palliative treatment.

Chemotherapy codes found after patient's last lung resection surgery and no radiotherapy codes
Chemotherapy that comes after lung resection can correspond to adjuvant chemotherapy, which is given with curative intent if certain temporal conditions are met. Firstly, the chemotherapy should be given within 6 weeks of surgery. In practice, there may be a longer delay, therefore, we selected a more realistic margin of 12 weeks (≤84 days) that will allow the identification of most patients that received adjuvant chemotherapy (experts' opinion). Secondly, since this type of chemotherapy usually involves 4 treatment cycles (4 months) [68], it should occur within a limited timeframe within patient's treatment trajectory. We selected a margin of 6 months after the 1st chemotherapy code to account for potential treatment delays. When the 1st temporal condition is met (chemotherapy ≤ 12 weeks or 84 days after surgery), the chemotherapy is considered as adjuvant (curative intent; scenario 3B). If no additional chemotherapy code is found after 6 months, the patient is considered to receive no palliative chemotherapy and their 1st EGFR-TKI is considered as 1st-line palliative treatment (scenario 3B1). If ≥1 chemotherapy code is found after the 6 months period (scenario 3B2), the patient is moved into scenario 6 (see scenario 6). When the 1st chemotherapy occurs >12 weeks after surgery (scenario 3A), the patient is moved into scenario 6 (see scenario 6).

Chemotherapy and radiotherapy codes, and no lung resection surgery codes
The physician billing code 8519 corresponds to treatment verification of an irradiated site (under treatment), and, according to experts that were consulted, it is systematically billed on a weekly basis throughout the treatment. Therefore, this code allows the estimation of the duration of radiotherapy. In the absence of a lung surgery, a combination of chemotherapy and radiation therapy can be administered with curative intent (chemoradiation) or in palliative intent (chemotherapy + palliative radiotherapy). The difference between these two types of treatments is that curative chemoradiation involves a higher dose of radiotherapy (generally 60-66 Gy for 6-7 weeks according to standard fractionation) [4]. Identifying ≥4 weeks of radiotherapy (≥4 × code 8519) can be used as a marker of curative treatment, whereas the presence of ≤3 weeks of radiotherapy corresponds to a palliative treatment (experts' opinion). If the radiotherapy is considered curative (scenario 4A), the algorithm verifies if all chemotherapies occur within 6 months. If no chemotherapy is found beyond 6 months of the 1st chemotherapy, the patient is considered to receive no palliative chemotherapy and their 1st EGFR-TKI is considered as 1st-line palliative treatment (scenario 4A1). If ≥1 chemotherapy code is found after the 6 months of the 1st chemotherapy (scenario 4A2), the patient is moved into scenario 6 (see scenario 6).

Chemotherapy and radiotherapy codes found after patient's last lung resection surgery
When chemotherapy and radiotherapy codes are observed after lung surgery, they could be related to (1) adjuvant/post-operative chemotherapy and radiotherapy treatments with curative intent, (2) chemotherapy and radiotherapy treatments with palliative intent, or (3) salvage chemoradiotherapy with curative intent. If the 1st chemotherapy code occurs in ≤12 weeks of the most recent lung resection (scenario 5B), it is considered as a curative treatment, irrespective of radiotherapy codes that may be present. If no chemotherapy is observed beyond 6 months after the 1st chemotherapy, the patient is considered to receive no palliative chemotherapy and their 1st EGFR-TKI is considered as 1st-line palliative treatment (scenario 5B1). If ≥1 chemotherapy code is found beyond 6 months of the 1st chemotherapy (scenario 5B2), the patient is assessed for receiving salvage chemoradiotherapy (curative) with the same steps as in scenario 4 (scenario 5B2A = 4A and scenario 5B2B = 4B). If the 1st chemotherapy code occurs in >12 weeks of the most recent lung resection (scenario 5A), the patient is assessed for receiving salvage chemoradiotherapy (curative) with the same steps as in scenario 4 (scenario 5A1 = 4A and scenario 5A2 = 4B).

Chemotherapy codes only OR other scenarios that mention "proceed to 6"
When chemotherapy is the only treatment observed, it always administered with palliative intent. Patients with this chemotherapy are considered to receive their 1st EGFR-TKI as 2nd-line or more of palliative treatment. An exception to this rule is when only 1 chemotherapy code is found which is within 30 days prior to the 1st EGFR-TKI, the latter is considered as 1st-line palliative treatment (scenario 6A). According to the experts we consulted, it is possible in clinic that some patients received a chemotherapy as an immediate treatment due to long delays in receiving mutation test results and patients pressing health state. In this case, the experts recommended that patients' 1st EGFR-TKI be considered as a 1st-line palliative treatment

C. Canadian classification of diagnostic, therapeutic, and surgical procedures (CCP) codes 4639
Exérèse lésion cage paroi thoracique/Exc./destruction de lesion de la cage thoracique 4669 Réparation paroi thoracique 1 Even though palliative immunotherapy is not indicated for treating lung cancer patients whose tumors harbor an EGFR mutation, the billing code for intravenous infusion of gamma globulins (152) was included in the chemotherapy category to account for situations where it may have been prescribed prior to an EGFR-TKI treatment. 2 All radiotherapy codes were used to group patients into 1 of 6 initial treatment code scenarios, however, only the code 8519 (in bold) was used for subsequent steps of the algorithm.

Appendix C. Literature Review on Experimental and Real-World Studies Reporting Survival Data for EGFR-TKIs for Indications of Interest
A preliminary literature search of the "snowball" type was carried out through the MEDLINE (PubMed) database. The goal of this preliminary search was to explore the literature to develop the search strategy and inclusion and exclusion criteria of our final literature review. PubMed's "similar articles" function allowed us to consult randomized controlled trials (i.e., experimental trials) on EGFR-TKIs similar to those presented in the summary tables of the algorithm for care of lung cancer patients [69].
The final systematic search of the literature was conducted to include all experimental trials and real-world studies reporting overall survival data related to the three EGFR-TKIs for indications of interest. In May 2021, we searched for articles in French or English that were published since 2005 in the following bibliographic databases: MEDLINE (Ovid), Embase (Ovid) et EBM Reviews: Cochrane Database of Systematic Reviews (EBSCO). Tables A5 and A6 provide information on the PICO-based (i.e., population, intervention, comparator, outcome) inclusion/exclusion criteria and search strategy, respectively. The studies that were submitted to INESSS for evaluation of the EGFR-TKI for indications of interest and reported overall survival results were included in our review, irrespective of inclusion/exclusion criteria. [20,[23][24][25][26][27][28] The results from the article selection process are presented in a flow diagram in Figure A2.
Two reviewers (SQ and GB) were each responsible for a part of the abstract and full-text screening. The references of retained articles were also searched. Data extraction was divided between the two reviewers and particular attention was given to subsequent treatments (i.e., osimertinib) in studies on 1st-line treatment with gefitinib or afatinib. Table A7 provides the data abstracted from each study, in addition to the median overall survival results that are presented in Figure 5 of the main text.

Study Design
Experimental trials, real-world studies, systematic reviews 1 , meta-analyses 1 Narrative reviews 1 Systematic reviews and meta-analyses were only included to search for additional studies that may have been missed by the literature review. 2 Osimertinib is known to increase overall survival relative to standard chemotherapy when given as a subsequent EGFR-TKI [19,20].
ncol. 2022, 29, FOR PEER REVIEW 23 Figure A2. Flow Diagram of the Article Selection Process. a 1 article [28] included 2 randomized controlled trials (i.e., 2 studies) on afatinib as 1st-line treatment. b 1 article [34], which included 1 realworld study, was counted twice since it was used in the indirect comparisons of EGFR-TKIs for 2 indications: 1st-line gefitinib treatment and 1st-line afatinib treatment. Figure A2. Flow Diagram of the Article Selection Process. a 1 article [28] included 2 randomized controlled trials (i.e., 2 studies) on afatinib as 1st-line treatment. b 1 article [34], which included 1 real-world study, was counted twice since it was used in the indirect comparisons of EGFR-TKIs for 2 indications: 1st-line gefitinib treatment and 1st-line afatinib treatment.   4 Patients' dissemination areas (DA: smallest Canada-wide geographic unit with a standard population size of 400-700) were obtained from the health insurance registry (FIPA) and linked to a DA-level material vulnerability index that was developed at INESSS with the 2016 Census data (DA-level unemployment ratio, median income, and low education). The index was adapted from a previous method and quintiles were created based on the total RAMQ population (Q1: least deprived and Q5: most deprived [70]. 5 The Population Grouping Methodology was applied to obtain the total number of comorbidities for each patient [71,72]. Patients' hospitalization data was screened for diagnostic codes for 226 possible health conditions (excluding lung cancer) in the three years prior to patients' 1st EGFR-TKI.