Concurrent Metformin Use and Survival Among Finnish Non-Small Cell Lung Cancer Patients Treated With EGFR TKIs

Purpose Many studies have shown correlation between metformin use and lower incidence and improved outcomes of lung cancer. We investigated the potential association between metformin use and survival among Finnish non-small cell lung cancer (NSCLC) patients treated with EGFR TKIs. Methods Nationwide registries were utilized to identify all the patients with use of EGFR TKIs in NSCLC between 2011–2016, and 1242 patients were included in further analyses. Data related to cancer, survival, and drug purchases were combined with personal identity codes. Concurrent use of diabetes medications was dened as their purchase +/-120 days from the rst purchase of EGFR TKI. The impact of diabetes medication use was investigated separately in the whole and in the EGFR mutant type cohort (n = 481).


Introduction
Epidermal growth factor receptor (EGFR) targeting with tyrosine kinase inhibitors (TKI) was initially studied in non-small cell lung cancer (NSCLC) without molecular selection (Shepherd et  ). The most common side effects of EGFR TKIs include acneiform rash and diarrhea but in general EGFR TKIs are well tolerated and treatment discontinuations due to toxicity are rare (Ding et al. 2017). A mutation-speci c EGFR TKI, osimertinib, has quite recently introduced further improvements in progression free survival (PFS) compared to both rst-generation EGFR TKIs and chemotherapy, with a favorable side effect pro le (Mok et al. 2017; Soria et al. 2018). Even though most of the patients with EGFR mutated NSCLC respond to EGFR TKIs, eventually an acquired resistance develops, and, in addition, a small proportion of cases are primary resistant (Rotow and Bivona 2017).
Metformin is commonly used in the treatment of type 2 diabetes (T2D). Insulin resistance, hyperinsulinemia and hyperglycemia are characteristic metabolic dysfunctions in T2D, and they have also been linked to cancer progression (Shlomai et al. 2016). Metformin increases insulin sensitivity and decreases glucose and insulin levels in the blood (Shlomai et al. 2016), and may enhance immunological response against cancer (Levy and Doyen 2018). Metformin also modulates signaling pathways including the activation of adenosine monophosphate activated protein kinase (AMPK) pathway and inhibition of insulin-like growth factor-1 receptor (IGF-1R) pathway, resulting in reduced tumor cell proliferation and increased apoptosis (Levy and Doyen 2018; Deng et al. 2019). These signaling pathways are also involved in EGFR TKI resistance mechanisms which provides a rationale for combining metformin with EGFR TKIs in NSCLC (Deng et al. 2019

Collection of patient material
Collection of the primary patient material has been described in detail previously (Alanen et al. 2020;Alanen et al. 2021). Brie y, all the patients with special reimbursement for EGFR TKIs (erlotinib, ge tinib or afatinib) between 2011-2016 were identi ed from the Special Reimbursement Register of Social Insurance Institution (SII) of Finland (n = 1541). In Finland, reimbursement for erlotinib in advanced NSCLC is based on the presence of activating EGFR mutation in the tumor or progression during prior chemotherapy and both are registered under the same code and cannot be separated. On the contrary, reimbursements for ge tinib and afatinib are entitled only in advanced NSCLC with activating EGFR mutations. The information on the drug purchases for EGFR TKIs, metformin, and other diabetes medications (purchase dates and the number and strength of purchased units) was collected from the Prescription database of SII, cancer related data from the Finnish Cancer Registry, and data on dates of death from Statistics Finland. The data were combined using personal identity codes and pseudonymized before analyses. The patients with at least one EGFR TKI purchase in the Prescription database and data available at the Finnish Cancer Registry were included in the study (n = 1271). The patients with afatinib as their 1st EGFR TKI were excluded from the study due to a small sample size (n = 29) resulting in a nal cohort of 1242 patients, 1072 with erlotinib and 170 with ge tinib. The follow-up for drug purchases and deaths ended on December 31st, 2017. Collection of the data was performed according to the national legislation, the ethical standards of the Ethical Board of Oulu University Hospital (study no.43/2017) and under a permit from the Social Insurance Institution of Finland (study no.48/522/2017), the Finnish Institute of Health and Welfare (study no. THL/1391/5.05.00/2017), and Statistics Finland (study no.TK-53-1277-17). Informed consent was not required as this is a retrospective register-based study.

Study cohorts and endpoints
Concurrent use of metformin or other diabetes medication was de ned as a purchase of diabetes medication within 120 days before or after the rst purchase of EGFR TKI. A 120-day time period was selected since the number of reimbursed drugs is limited to a three-month supply per purchase and this would re ect the concurrent use of the medication. As metformin may have different effects in EGFR mutant and EGFR wild-type (wt) lung cancers, we created an EGFR mutant type cohort which included all the patients with ge tinib (based on the reimbursement criteria requiring EGFR mutant NSCLC) and patients with > 180 days of erlotinib use as patients with EGFR mutant NSCLC more often obtain longlasting responses than patients with EGFR wt lung cancer. The impact of the concurrent metformin and other diabetes medication use was investigated separately in the whole patient cohort (n = 1242) and in the EGFR mutant type cohort (n = 481).
The purchases for medication were collected from 1st January 2010 until 31st December 2017. OS was calculated from the date of rst EGFR TKI purchase to death or end of the follow-up (December 31st, 2017), and death of any cause was counted as an event. The length of EGFR TKI treatment was calculated from the rst EGFR TKI purchase to the last and the number of days according to number of tablets in the last purchase were added to the treatment length. Treatment was considered continuous based on dates of purchases and the number of tablets allowing a maximum of 10 days gap between the purchases. Discontinuation of treatment before the end of the follow-up was counted as an event. EGFR TKI dose reduction was de ned as EGFR TKI purchase of lower dose compared to the rst prescribed dose within 200 days from the beginning of the treatment. EGFR TKI treatment break was de ned as a break of more than 30 days within the rst 200 days of treatment and calculated from the EGFR TKI purchase dates and number of tablets.

Statistical analyses
Statistical analyses were performed with IBM SPSS Statistics v. 25.0.0 (IBM Corporation, Armonk, NY, USA). Chi-square test was used for analyzing the differences between the groups. Kaplan-Meier method using log-rank test was utilized for survival and EGFR TKI treatment length analyses and for plotting the curves. Univariate and multivariate (tumor histology; concomitant metformin with EGFR TKI/erlotinib/ge tinib; other diabetes medication with EGFR TKI) analyses were performed with the Cox's model. The results are presented with 95% con dence level. P-values < 0.05 were regarded as statistically signi cant.

Patient characteristics
Of the 1242 patients, 54% (n = 670) were male and 46% (n = 572) were female, 65% (n = 812) had adenocarcinoma while the rest had other histology or de ned as unknown (n = 430). The rst purchased EGFR TKI was erlotinib in 86% (n = 1072) and ge tinib in 14% (n = 170) of the patients. Concurrent metformin use (-120 to + 120 days from the rst EGFR TKI purchase) was found in 10% (n = 124) and the use of other diabetes medication without metformin in 5 % (n = 60) of the patients. The median follow-up time was 288 days (range 0-2606) and 1073 (86%) patients had died by the end of the follow-up. (Table   1) EGFR, epidermal growth factor receptor; NSCLC, non-small cell lung cancer; TKI, tyrosine kinase inhibitor Erlotinib use for > 180 days was found in 25% (n = 311) of the patients. The EGFR mutant type cohort (n = 481) was formed from all the patients with ge tinib (n = 170) and the patients with erlotinib with a use of the TKI for > 180 days. Of these patients, 45% (n = 216) were male and 76% (n = 367) had adenocarcinoma. Concurrent metformin use was found in 9% (n = 41) and other diabetes medication without metformin in 4% (n = 21) of the patients. The median follow-up time was 568 days (range 1-2606) and 360 (75%) patients had died by the end of the follow-up. (Table 1) Concurrent metformin use and survival In the whole patient cohort (n = 1242), the concurrent metformin use with EGFR TKI did not signi cantly correlate with the survival of the patients (Fig. 1A, Table 2). With the ge tinib users (n = 170), there was a non-signi cant trend for inferior survival among those with concurrent metformin use whereas no difference was found among the patients with erlotinib (n = 1072) ( Table 2). The survival of the patients with other diabetes medication than metformin was similar compared to patients without diabetes medication (Fig. 1A, Table 2). In the EGFR mutant type cohort (n = 481), the use of metformin associated with a narrowly non-signi cant trend for inferior survival (Fig. 1A, Table 2). The use of other diabetes medication than metformin did not correlate with survival (Fig. 1A, Table 2).   among patients with and without metformin, respectively (p = 0.02). The use of metformin associated signi cantly also with a shorter ge tinib treatment duration (HR 1.89, 95% CI 1.08-3.32) and there was a non-signi cant trend for inferior EGFR TKI treatment duration among the erlotinib users (Table 3). Concurrent metformin use remained as a signi cant risk factor for shorter EGFR TKI treatment duration (HR 1.23, 95% CI 1.02-1.49) in the Cox multivariate analysis including also tumor histology (Table 3). In the EGFR mutant type cohort, the median length of EGFR TKI treatment was 336 days (range . EGFR TKI treatment duration was shorter among patients with metformin compared to the others (HR 1.59, 95% CI 1.12-2.26) (Fig. 1B, Table 3) with median treatment durations of 269 days (30-679) and 348 days , respectively (p = 0.009). The use of other diabetes medication than metformin did not associate with EGFR TKI treatment durations in the whole patient cohort or in the EGFR mutant type cohort (Fig. 1B, Table 3).

Discussion
To our knowledge, this is the rst study investigating the role of metformin in Nordic lung cancer patient population treated with EGFR TKIs. Concurrent metformin use with EGFR TKIs showed a non-signi cant trend for inferior survival in the EGFR mutant type cohort. In addition, metformin use associated signi cantly with a shorter EGFR TKI treatment duration. The use of other diabetes medication than metformin did not associate with survival of the patients or EGFR TKI treatment duration.
Preclinical studies have suggested that many of the anti-tumoral effects of metformin may require high doses but especially in combination with other cancer treatments metformin could provide e cacy and tolerability in conventional doses (Pollak 2014; Levy and Doyen 2018). Diarrhea is a common side effect of both metformin and EGFR TKIs, and it is possible that especially high doses of metformin may increase the side effects leading to dose reductions or interruptions of EGFR TKI treatment and subsequently impairing the e cacy of the treatment. In our study, metformin use associated with a shorter EGFR TKI treatment duration which likely re ects shorter PFS but may also indicate higher incidence of side effects leading to treatment discontinuation. Nevertheless, no signi cant correlations between metformin use and EGFR TKI dose reductions or treatment breaks were found but it should be In the present study, information of metformin dose was unavailable but based on the previous data, metformin dose may be important in the treatment combination with EGFR TKIs.
The effects of metformin in NSCLC may also be in uenced by ethnicity of the patients. In a recent metaanalysis (Xiao et al. 2020), lung cancer incidence was reduced in T2D patients treated with metformin compared to those without only in studies including Asian patients, but not among European patients.
Also, survival of lung cancer patients with T2D and metformin medication was better compared to those without metformin, and even though this association was found in both Asian and non-Asian populations, the possible protective effect of metformin seemed to be greater among the Asian patients patients with other diabetes medication than metformin, survival and duration of EGFR TKI treatment were similar compared to those without any diabetes medications. It is possible that T2D medication used re ects some characteristics of the patient populations, e.g., it has been suggested that metabolic dysfunction may be more di cult among T2D patients whose medication includes metformin whereas metformin monotherapy may associate with a better general health status (Fatehi Hassanabad and MacQueen 2021). It is also important to notice that the metabolic state is different among T2D patients and patients without T2D, and also the effects of metformin in cancer treatment may be different in these patient populations.
The strengths of our study include a large nationwide patient material which was collected using reliable nationwide registries and combining the data with personal identity codes. The study has also some limitations. Several possible confounding factors for which the data was unavailable, e.g., comorbidities, smoking status, and metformin doses used, may have in uenced on the associations found in this study.
In addition, we had no information on possible hospital treatment periods of the patients and medication received during hospitalization. Also, even though the patient material is quite large, the number of patients in some of the subgroup analyses remained rather small. Furthermore, the results cannot be extrapolated to non-diabetic patients as the effects of metformin may be different in patients without diabetes.
To conclude, in this material of Finnish NSCLC patients treated with EGFR TKIs, metformin use associated with a shorter EGFR TKI treatment duration and also a non-signi cant trend for inferior survival was found. Differences in the patient materials, e.g., ethnicity, differences in mutation pro les,  Figure 1 Kaplan-Meier analysis for overall survival and EGFR TKI treatment duration according to purchase of diabetes medication. A. Overall survival in the whole patient cohort and EGFR mutant type cohort. B.
EGFR TKI treatment duration in the whole patient cohort and EGFR mutant type cohort. Blue= no diabetes medication purchases, red= other diabetes medication purchases than metformin, green= metformin purchases -120 to +120 days from the 1st EGFR TKI purchase. Crosses indicate censored events.