Real-world management of patients with epidermal growth factor receptor (EGFR) mutation-positive non–small-cell lung cancer in the USA

Background Randomized phase III trials have established the efficacy of epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors as first-line treatment for EGFR mutation-positive advanced non–small-cell lung cancer (EGFR Mut+ NSCLC). This retrospective cohort study examined the management patterns and outcomes of patients with EGFR Mut+ NSCLC in a real-world setting. Materials and methods Data were extracted from the US Flatiron Electronic Health Record-derived database. Adult patients with stage IIIB/IV EGFR Mut+ NSCLC (exon 19 deletion or exon 21 L858R mutation) who had received first-line systemic therapy between 2011 and 2016 were included. Demographic and clinical characteristics were analyzed. Outcomes evaluated were time to next treatment (a surrogate for progression-free survival) and overall survival. Results Of the 22,258 patients with advanced NSCLC in the database, 961 met the inclusion criteria. Median age was 69.0 years (range: 61–78) and the majority were female (68.0%), with stage IV (93.9%), non-squamous cell carcinoma (97.4%). EGFR tyrosine kinase inhibitors were the most widely prescribed first-line therapy (72.8%). The likelihood of receiving an EGFR tyrosine kinase inhibitor or chemotherapy was unaffected by the type of medical insurance patients had. Patients treated with an EGFR tyrosine kinase inhibitor had significantly longer time to next treatment than those given other first-line systemic therapies (p < 0.0001). There were no significant differences in overall survival according to treatment type. Conclusion Results from this large US cohort study reflect those obtained in randomized trials of patients with advanced EGFR Mut+ NSCLC and demonstrate their transferability into a real-world setting.


Introduction
Non-small-cell lung cancer (NSCLC) comprises up to 80% of all newly diagnosed lung cancer cases, and more than half of all patients (57%) are diagnosed with metastatic disease [1]. Despite advances in treatment, 5-year survival rates for advanced NSCLC remain low, at around 15-20% [1][2][3]. Approximately 19% of Western patients (Europe, North America, or Australia) with NSCLC have tumors that harbor a mutation in the epidermal growth factor receptor gene (EGFR Mut+) while the corresponding prevalence in Asian patients is around 48% [4]. The introduction of tyrosine kinase inhibitors (TKIs) directed against EGFR, such as erlotinib, afatinib, and gefitinib, has improved the outlook for this subgroup of patients [5] and these drugs now represent standard first-line treatment for EGFR Mut+ NSCLC [6].
Despite the wealth of published clinical trial data, information concerning the use of EGFR TKIs in a real-world setting is limited. Real-world data complement results from randomized clinical trials and generate evidence on the effectiveness and use of medical products in daily practice [17], often providing a better understanding of patient response, disease patterns, safety data, and economic outcomes in patients of all ages [18]. This study was designed to evaluate the management and clinical outcomes of patients with EGFR Mut+ NSCLC by analyzing real-world data from a large US healthcare database.

Data source
This study used data extracted from the Flatiron Health Cloud-based Oncology Electronic Health Record (EHR)-derived database. The longitudinal database represents one of the most comprehensive sources of real-world evidence in oncology. At the data cut-off date of August 31, 2017, the Flatiron Health database included data from more than 265 cancer clinics (~800 sites of care) representing more than 2 million active US cancer patients available for analysis. This database captures data from both structured and unstructured EHR fields using a technology-enabled abstraction process and multi-pronged quality assurance approaches. Data are refreshed monthly to provide near real-time research-ready datasets [19].
Approval of the study protocol was obtained from Copernicus Group Independent Review Board (CGIRB) prior to study conduct, and included a waiver of informed consent. Data provided to third parties were de-identified and provisions were in place to prevent re-identification in order to protect patients' confidentiality.

Patients and methods
This retrospective study examined treatment patterns, demographic and disease characteristics, and clinical outcomes of patients with locally advanced unresectable or metastatic, EGFR Mut+ NSCLC. The patient cohort for this study was selected from the Flatiron advanced NSCLC population, which included patients with a diagnosis of lung cancer, �2 clinical visits on or after January 1, 2011, and evidence of stage IIIB, IV, or recurrent metastatic NSCLC on or after January 1, 2011.
The selection criteria for this study included: age �18 years at diagnosis, histologically confirmed stage IIIB/IV NSCLC (American Joint Committee on Cancer [AJCC], 7th edition), documented EGFR Mut+ disease (exon 19 deletion or exon 21 L858R mutation), and receipt of systemic therapy within 90 days of diagnosis and between January 1, 2011-June 30, 2016. Patients with evidence of data discrepancies, such as diagnosis date before birth year, date of treatment initiation before diagnosis date, mortality date before last visit date, were excluded. The start of treatment of interest in the first-line setting was defined as the index date to evaluate clinical outcomes for the treatment subgroup. Patients were retrospectively followed up until June 30, 2017, to allow at least 1-year follow up for each patient.

Study endpoints
Demographic variables evaluated were age, gender, race, and smoking status. Disease clinical characteristics included tumor histology, stage at diagnosis, absence or presence of brain metastases, Eastern Cooperative Oncology Group performance status, and EGFR mutation information. The proportions and sequence of patients receiving chemotherapy and targeted agents, with a focus on TKIs, were examined.
Time to next treatment (TTNT) was used as a surrogate for PFS and was defined as the time from the start of first-line treatment for advanced EGFR Mut+ NSCLC until initiation of second-line treatment, discontinuation of first-line treatment, the data cut-off, or death. OS was defined as the time from initiation of first-line treatment until death from any cause. Due to the de-identification processing of the analysis dataset, all death dates were approximated to the 15th of the month. Patients were censored at the last visit date or at the data cut-off date, whichever occurred first, if they had no death information or if they were still on treatment at the end of the study period.
The predefined variable of line of therapy in the Flatiron database was used to define the firstline setting. However, since this variable was derived from administrative or prescription dates, and name and dosage of regimens, rather than date of disease progression, line of therapy described in the study should be interpreted as a surrogate to line of therapy in a clinical trial.

Statistical analyses
Baseline patient and clinical characteristics were described at the start of the first-line treatment in the overall study cohort and by treatment subgroups. Categorical variables were described as frequency and percentages. For continuous variables, means, standard deviations, medians, range and percentiles (25th and 75th) were reported when appropriate. Comparisons between treatment subgroups were made using chi-square tests for categorical variables and Wilcoxon rank-sum test for continuous variables (two-sided test, p-value threshold � 0.05). No formal hypotheses were tested and p-values were reported purely for descriptive purposes. All analyses were performed in the overall study cohort and by treatment subgroups (e.g., TKIs, erlotinib, afatinib, chemotherapy, or other non-TKI target therapy in the first-line). A Kaplan-Meier analysis was conducted to estimate median TTNT and OS with 95% CIs; HRs and 95% CIs with p-values were calculated using a Cox-proportional hazards model after adjusting for patient age, gender, race, smoking status, and histology. All statistical analyses were conducted using SAS software (SAS Institute, Cary, NC) or R.

Patients
In total, 22,258 patients aged �18 years with stage IIIB/IV NSCLC were identified from the Flatiron advanced NSCLC cohort (Fig 1). Of these, 1,564 patients had an activating EGFR mutation in their tumors, and 1,293 patients received first-line systemic therapy for NSCLC between January 1, 2011-June 30, 2016. After excluding patients with evidence of data discrepancies, a total of 961 patients were included in the main analysis. Since only 13 patients were treated with gefitinib, these patients were not included in subgroup analyses.

Baseline demographic and clinical characteristics
Patients were older than the median age observed in clinical trials of EGFR Mut+ NSCLC (~60 years) [7,13,16] with a median age of 69 years (range: 61-78) ( Table 1). The majority of

Insurance
The likelihood of receiving an EGFR TKI was generally unaffected by the type of medical insurance patients had ( Table 1). The most common insurance type was a commercial health plan, which was held by 36.9% (n = 258) of patients who received EGFR TKIs, 40.8% (n = 99) of patients who received chemotherapy alone and by 41.2% (n = 35) of patients on other non-TKI targeted treatments ( Table 1). Patterns of insurance were similar among patients receiving an EGFR TKI and patients receiving other systemic therapy.

Discussion
This retrospective study evaluated real-world treatment patterns of patients with advanced EGFR Mut+ NSCLC in oncology practices across the USA, and examined demographic and disease characteristics among patients receiving EGFR TKIs vs other systemic anti-cancer therapies. With the exception of age (clinical trial patients are approximately 9 years younger), the baseline characteristics of the patients eligible for this analysis were similar to those observed in randomized, clinical trials of EGFR Mut+ NSCLC [7,13,16].
The rate of EGFR mutation testing in our analysis was low in 2011, at just 30.5%. This contrasts with rates of up to 65% reported in Japan during the same year, although testing rates in China were only 18% [20]. Following a steady increase in testing rates between 2011 and 2015, we observed a rise in EGFR mutation testing during 2016 to approximately 78%. Apart from the potential bias introduced by data completeness during the earlier years, this increase probably reflects an enhanced awareness of the need for EGFR mutation testing in patients with advanced NSCLC before initiation of first-line therapy, as recommended in clinical practice guidelines published at that time [21,22], or to a wider availability of a validated test. Notably, the results of a retrospective chart review in patients with advanced non-squamous NSCLC initiating therapy between 2011 and 2013, with follow-up until 2016, revealed EGFR mutation testing rates of more than 97% in Taiwan [23]. Compared with our study population, fewer of the Taiwanese patients were aged >75 years (22% vs 32%, respectively), with a history of smoking (33% vs 44%, respectively), and a greater proportion of them received first-line therapy for advanced NSCLC (100% vs 83%, respectively) [23]. Reports from the Western literature during the same time-period have suggested testing rates in the region of 70%, congruent with the results of our analysis [24]. These results suggest that the introduction of the broad use of molecular testing took more than 5 years to achieve levels of almost 80%, assuming that in about 5-10% of patients tumor testing cannot be performed for various reasons (e.g., no tumor tissue available, no biopsy possible, or need for immediate treatment intervention without a delay).
Treatment patterns were as expected, based on the mutation status of the patients' tumors, with single-agent EGFR TKIs, predominantly erlotinib, having been the most widely prescribed first-line therapy. The type of medical insurance that patients had did not influence the likelihood of them receiving an EGFR TKI. There has been a lack of evidence supporting the efficacy of combining EGFR TKIs with chemotherapy, nevertheless, erlotinib and afatinib were used in combination with chemotherapy in a few patients (n = 32; 3.3% of the entire cohort). We speculate that a TKI was intercalated with chemotherapy rather than given concomitantly.
Approximately 27% of patients with EGFR Mut+ NSCLC were treated with a platinumcontaining chemotherapy doublet either with or without bevacizumab. Single-agent erlotinib was the second-line therapy of choice in the majority of patients who continued with treatment, presumably given to those patients who did not receive a TKI as first-line treatment.
Patients who were given an EGFR TKI as first-line therapy had significantly longer TTNT, a surrogate for PFS, than those receiving other systemic treatments. Median TTNT was 13.1 months (95% CI 12.1-14.3) with erlotinib and 12.1 months (95% CI 9.7-14.9) with afatinib. Erlotinib and afatinib demonstrated comparable efficacy in this regard (both p < 0.0001 vs other systemic therapies). The slightly longer duration of TTNT in this study compared with PFS in clinical trials [7][8][9][13][14][15] most likely reflects the time that would typically elapse between clinical or radiologic progression and the start of the next course of treatment in Real-world management of patients with EGFR mutation-positive NSCLC in the USA eligible patients. The comparable values seen for TTNT in this real-world setting and PFS observed in a clinical trial setting are important and reassuring. The patients in our study had a higher median age than patients enrolled in clinical trials, confirming that increased age is not associated with reduced efficacy in the real-world setting. As observed in randomized studies [7,25], no significant difference in OS was noted among patients receiving EGFR TKIs vs those receiving chemotherapy as their initial therapy. This may be because almost half of the patients who received chemotherapy initially, received an EGFR TKI on progression (49.7%, n = 84/169). However, 31.9% of patients (n = 54) who received chemotherapy upfront were never exposed to an EGFR TKI. While the reason for this may vary, this finding requires further investigation. The median OS of 23.2 months (95% CI 21.2-24.9) achieved with erlotinib and 20.7 months (95% CI 16.2-35.1) achieved with afatinib were in-line with the survival times reported in phase III studies [7,11,16]. Of note, a much higher number of patients received treatment with erlotinib than with afatinib throughout the study (n = 593 vs n = 87, respectively).
The results from this real-world study support the efficacy of the EGFR TKIs established in clinical trials. However, the population studied in the USA may not be representative of a global population, and additional analyses may be needed to evaluate real-world treatment patterns in other countries. In a retrospective chart review examining real-world practice patterns in 175 Japanese patients treated for stage IIB/IV NSCLC between 2011 and 2013, EGFR TKIs were the most commonly prescribed therapies for EGFR Mut+ NSCLC across all treatment lines [26], similar to our findings. Of note, median OS from the start of first-line therapy was 9.9 months (95% CI 7.6-11.7) for all patients compared with 17.9 months (95% CI 9.9-24.4) for those with EGFR/ALK Mut+ tumors. The authors note that due to the retrospective nature of the study, these results are not reflective of the current clinical landscape, given the availability of newer therapies for NSCLC, namely third-generation EGFR TKIs [26,27].
Limitations of this analysis are reflective of the data source and collection. Flatiron data are generated from real-world clinical practice, and are subject to miscoding and errors. The data are mainly drawn from community oncology centers and therefore may not be representative of treatment patterns at academic medical centers. Since information about patients prior to the clinical oncology practice's adoption of the EHR may not be available, particularly in the structured data, the extent to which historical data are entered into the EHR is a practice-specific decision and varies widely. An important strength of the present study over previous realworld data studies [28,29] was the availability of information on mortality in patients with EGFR Mut+ NSCLC from which OS data could be estimated. Moreover, the sensitivity and specificity of mortality in the advanced NSCLC dataset is 91% and 96%, respectively, which supports its application to evaluate clinical outcomes in this setting [30]. Finally, the study enabled a description of longitudinal treatment patterns with a minimum of 12 months of follow-up for every patient. Notwithstanding the limitations, our results represent the largest available real-world data in this setting. Concurrence between these data and clinical trial evidence is reassuring, with respect to the accuracy and validity of our findings on one hand and applicability of the clinical data in real-life practice on the other.

Conclusions
These real-world data reflect the results of randomized clinical trials of EGFR Mut+ NSCLC, with the exception of patients in this analysis being generally older, by approximately 9 years. Testing rates for EGFR mutations steadily increased over time, reaching approximately 78% in 2016, although there is room for improvement. The majority of patients with EGFR Mut+ disease received EGFR TKIs as first-line therapy, which were predominantly administered as monotherapy. Patients who received first-line EGFR TKIs had significantly longer TTNT than those who received other non-EGFR TKI systemic therapies. No significant differences in OS were observed by treatment type.
Supporting information S1 Table. TTNT and OS outcomes by EGFR mutation status. (DOCX)