Advertisement

Pharmacovigilance Analysis of Cardiac Toxicities Associated With Targeted Therapies for Metastatic NSCLC

  • Sarah Waliany
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California
    Search for articles by this author
  • Han Zhu
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, California

    Stanford Cardiovascular Institute, Stanford University School of Medicine, Stanford, California
    Search for articles by this author
  • Heather Wakelee
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Oncology, Stanford University School of Medicine, Stanford, California

    Stanford Cancer Institute, Stanford, California
    Search for articles by this author
  • Sukhmani K. Padda
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Oncology, Stanford University School of Medicine, Stanford, California

    Stanford Cancer Institute, Stanford, California
    Search for articles by this author
  • Millie Das
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Oncology, Stanford University School of Medicine, Stanford, California

    Stanford Cancer Institute, Stanford, California

    Department of Medicine, VA Palo Alto Health Care System, Palo Alto, California
    Search for articles by this author
  • Kavitha Ramchandran
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Oncology, Stanford University School of Medicine, Stanford, California

    Stanford Cancer Institute, Stanford, California
    Search for articles by this author
  • Nathaniel J. Myall
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Oncology, Stanford University School of Medicine, Stanford, California

    Stanford Cancer Institute, Stanford, California
    Search for articles by this author
  • Thomas Chen
    Affiliations
    Division of Oncology, Stanford Cancer Center South Bay, San Jose, California
    Search for articles by this author
  • Ronald M. Witteles
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, California
    Search for articles by this author
  • Joel W. Neal
    Correspondence
    Corresponding author. Address for correspondence: Joel W. Neal, MD, PhD, Division of Oncology, Stanford University School of Medicine, 875 Blake Wilbur Drive, Palo Alto, CA 94304, USA.
    Affiliations
    Department of Medicine, Stanford University School of Medicine, Stanford, California

    Division of Oncology, Stanford University School of Medicine, Stanford, California

    Stanford Cancer Institute, Stanford, California
    Search for articles by this author
Open ArchivePublished:August 18, 2021DOI:https://doi.org/10.1016/j.jtho.2021.07.030

      Abstract

      Introduction

      Targeted therapies have transformed treatment of driver-mutated metastatic NSCLC. We compared cardiovascular adverse events between and within targeted therapy classes.

      Methods

      We used WHO pharmacovigilance database VigiBase to compare odds of heart failure, conduction disease, QT prolongation, supraventricular tachycardia (SVT), and ventricular arrhythmias between inhibitors of EGFR (erlotinib, gefitinib, afatinib, osimertinib), BRAF (dabrafenib), MEK (trametinib), and ALK and ROS1 (alectinib, brigatinib, ceritinib, crizotinib, lorlatinib).

      Results

      Of 98,765 adverse reactions reported with NSCLC targeted therapies, 1783 (1.8%) were arrhythmias and 1146 (1.2%) were heart failure. ALK and ROS1 inhibitors were associated with increased odds of conduction disease (reporting OR [ROR] = 12.95, 99% confidence interval [CI]: 10.14–16.55) and QT prolongation (ROR = 5.16, 99% CI: 3.92–6.81) relative to BRAF and EGFR inhibitors. Among ALK and ROS1 inhibitors, crizotinib had highest odds of conduction disease (ROR = 1.75, 99% CI: 1.30–2.36) and QT prolongation (ROR = 1.91, 99% CI: 1.22–3.00). Dabrafenib (ROR = 2.24, 99% CI: 1.86–2.70) and trametinib (ROR = 2.44, 99% CI: 2.03–2.92) had higher odds of heart failure than other targeted therapies. Osimertinib was strongly associated with QT prolongation (ROR = 6.13, 99% CI: 4.43–8.48), heart failure (ROR = 3.64, 99% CI: 2.94–4.50), and SVT (ROR = 1.90, 99% CI: 1.26–2.86) relative to other targeted therapies.

      Conclusions

      ALK and ROS1 inhibitors are associated with higher odds of conduction disease and QT prolongation than other targeted therapies. Osimertinib is strongly associated with QT prolongation, SVT, and heart failure relative to other EGFR inhibitors and targeted therapies. Monitoring for heart failure and arrhythmias should be considered with NSCLC targeted therapies, especially osimertinib.

      Keywords

      Introduction

      Molecularly targeted therapies have transformed the treatment of metastatic NSCLC. Advances in molecular profiling have identified multiple targetable mutations or gene rearrangements in lung adenocarcinoma including in EGFR, ALK, ROS1, and BRAF.
      • Skoulidis F.
      • Heymach J.V.
      Co-occurring genomic alterations in non-small-cell lung cancer biology and therapy.
      In lung adenocarcinoma, although there is variation on the basis of clinical characteristics such as smoking, sex, and race, EGFR mutations are found in up to 38% of tumors,
      • Zhang Y.L.
      • Yuan J.Q.
      • Wang K.F.
      • et al.
      The prevalence of EGFR mutation in patients with non-small cell lung cancer: a systematic review and meta-analysis.
      ALK rearrangements in up to 3% to 5%,
      • Kim H.
      • Chung J.H.
      Overview of clinicopathologic features of ALK-rearranged lung adenocarcinoma and current diagnostic testing for ALK rearrangement.
      ,
      • Bergethon K.
      • Shaw A.T.
      • Ou S.H.
      • et al.
      ROS1 rearrangements define a unique molecular class of lung cancers.
      BRAF mutations in up to 5%,
      • Marchetti A.
      • Felicioni L.
      • Malatesta S.
      • et al.
      Clinical features and outcome of patients with non-small-cell lung cancer harboring BRAF mutations.
      and ROS1 rearrangements in up to 1% to 2%.
      • Bergethon K.
      • Shaw A.T.
      • Ou S.H.
      • et al.
      ROS1 rearrangements define a unique molecular class of lung cancers.
      Although cytotoxic chemotherapy was once the predominant treatment for metastatic NSCLC, targeted agents are now the preferred first-line therapy for those with actionable genetic aberrations.
      Although targeted therapies are generally well tolerated in NSCLC, cardiovascular adverse events have been described, including arrhythmias and heart failure. Understanding these cardiovascular effects is particularly critical given the high prevalence of cardiovascular diseases in patients with NSCLC. At baseline, approximately 14% to 22% of patients with stage I to IV NSCLC have heart failure and 26% to 31% have arrhythmias.
      • Kravchenko J.
      • Berry M.
      • Arbeev K.
      • Lyerly H.K.
      • Yashin A.
      • Akushevich I.
      Cardiovascular comorbidities and survival of lung cancer patients: Medicare data based analysis.
      In addition, for patients with metastatic NSCLC initially diagnosed with having stage I to IIIA diseases, previous radiation increases long-term risk of clinically relevant arrhythmia and heart failure.
      • Wang K.
      • Eblan M.J.
      • Deal A.M.
      • et al.
      Cardiac toxicity after radiotherapy for stage III non-small-cell lung cancer: pooled analysis of dose-escalation trials delivering 70 to 90 Gy.
      Although previous studies have compared cardiovascular adverse drug reactions (ADRs) between EGFR inhibitors,
      • Anand K.
      • Ensor J.
      • Trachtenberg B.
      • Bernicker E.H.
      Osimertinib-induced cardiotoxicity: a retrospective review of the FDA Adverse Events Reporting System (FAERS).
      to our knowledge, no previous studies have comprehensively described and compared cardiovascular ADRs for all typically used targeted therapies in NSCLC. Therefore, we evaluated the association between NSCLC targeted agents and arrhythmia and heart failure using a pharmacovigilance database.

      Materials and Methods

      Study Design

      This retrospective pharmacovigilance study used VigiBase, the WHO database of deidentified ADRs from more than 130 countries.
      • Lindquist M.
      VigiBase, the WHO global ICSR database system: basic facts.
      Managed by the Uppsala Monitoring Centre in Sweden, VigiBase contains more than 21 million ADRs received during the postmarketing period from health care professionals, patients, and pharmaceutical companies.

      Pharmacovigilance Cohort

      We extracted all cases of arrhythmia (including conduction disease, QT prolongation, supraventricular tachycardia [SVT], ventricular arrhythmias) and heart failure available between database inception in November 14, 1967, and July 10, 2020. ADRs were coded according to WHO Adverse Reaction Terminology and Medical Dictionary for Regulatory Activities and categorized as conduction diseases, SVTs, ventricular arrhythmias, long QT syndrome, or heart failure (Supplementary Table 1).
      We restricted this analysis to small-molecule tyrosine kinase inhibitors (TKIs) used in metastatic NSCLC. Selected drug categories were inhibitors of EGFR (erlotinib, gefitinib, afatinib, osimertinib), BRAF (dabrafenib), MEK (trametinib), and ALK (alectinib, brigatinib, ceritinib, crizotinib, lorlatinib); the latter three ALK inhibitors are also inhibitors of ROS1. Drugs with less than 150 ADRs were excluded: larotrectinib (n = 114), dacomitinib (n = 100), selpercatinib (n = 49), entrectinib (n = 42), pralsetinib (n = 40), and capmatinib (n = 20).
      For each ADR, we retrieved administrative information (country of origin, report date, reporter qualification), patient characteristics (sex, age), drug characteristics, and reaction details (time to onset and seriousness). Severe adverse events were those considered life threatening or resulting in death, significant disability, or hospitalization.

      Statistical Analysis

      Characteristics of ADR cases were described using means (with SDs) or medians (with interquartile ranges) for quantitative variables and numbers and proportions for categorical variables. Two-sided t tests compared quantitative variables; chi-square tests compared categorical variables.
      Disproportionality analyses evaluated if arrhythmias or heart failure were differentially reported with NSCLC targeted therapies compared with the full database. In addition, known as case–noncase analysis, disproportionality analysis compares the proportion of a selected ADR for a single drug or drug class with the proportion of that same ADR reported for a control drug group.
      • Norén G.N.
      • Hopstadius J.
      • Bate A.
      Shrinkage observed-to-expected ratios for robust and transparent large-scale pattern discovery.
      Disproportionality was assessed using information component (IC) and reporting OR (ROR). ROR was used to compare the rate of each ADR from each of 11 investigated drugs to the ADR rate from (1) the same drug class (e.g., crizotinib versus other ALK inhibitors), (2) the other 10 NSCLC targeted therapies investigated (e.g., crizotinib versus other 10 NSCLC targeted therapies), and (3) all drugs in VigiBase (not limited to antineoplastic drugs). Because dabrafenib and trametinib are usually co-administered, cardiotoxicities of these agents were not compared with each other. To reduce type 1 error from multiple comparisons, two-sided α level of 0.01 was used to determine statistical significance. ROR was statistically significant if the lower end of the 99% confidence interval (CI) was greater than 1.
      Calculated using Bayesian Confidence Propagation Neural Network, IC is a logarithm of the ratio of the observed rate of drug ADRs to the expected rate assuming the null hypothesis (Supplementary Fig. 1). IC is a measure of the strength of a drug-ADR association. IC025 is the lower end of the 95% credibility interval for IC, with positive IC025 value (>0) indicating a statistically significant drug-ADR association. IC has been validated for comparing ADRs from a specific drug or drug class against the full VigiBase database; IC cannot compare reporting between individual drugs.
      • Norén G.N.
      • Hopstadius J.
      • Bate A.
      Shrinkage observed-to-expected ratios for robust and transparent large-scale pattern discovery.
      • Norén G.N.
      • Bate A.
      • Orre R.
      • Edwards I.R.
      Extending the methods used to screen the WHO drug safety database towards analysis of complex associations and improved accuracy for rare events.
      • Bate A.
      • Lindquist M.
      • Edwards I.R.
      • et al.
      A Bayesian neural network method for adverse drug reaction signal generation.
      As a result, the association between NSCLC targeted therapies and cardiovascular ADRs was assessed with both IC and ROR when using the full database as the comparator and only by ROR when comparing individual drugs or drug classes to each other.

      Results

      Of 98,765 ADRs reported with metastatic NSCLC targeted therapies, there were 61,383 with EGFR inhibitors, 15,540 with ALK inhibitors, and 21,842 with BRAF and MEK inhibitors. Among those ADRs, 1783 (1.8%) represented arrhythmias and 1146 (1.2%) represented heart failure. The total number of ADRs from all drugs in VigiBase was 21,740,021.

      Conduction Diseases

      We evaluated the association between each NSCLC targeted therapy and conduction diseases (bradycardia, sinus node dysfunction, atrioventricular block, bundle branch block, and others). ALK inhibitors (n = 376) were associated with increased odds of conduction disease relative to both EGFR (n = 89) and BRAF (n = 70) inhibitors combined (ROR = 12.95, 99% CI: 10.14–16.55) and the full VigiBase database (ROR = 4.48, 99% CI: 3.92–5.13, IC025 = 1.98) (Fig. 1). Crizotinib was associated with higher odds of conduction disease (Fig. 2) relative to both other ALK inhibitors (ROR = 1.75, 99% CI: 1.30–2.36) and other NSCLC targeted therapies collectively (ROR = 10.11, 99% CI: 8.08–12.66), with median time to onset of 53 days (Supplementary Table 2). Alectinib had a trend toward higher odds relative to other ALK inhibitors (ROR = 1.38, 99% CI: 0.99–1.93) and significantly increased odds relative to other NSCLC targeted therapies (ROR = 6.77, 99% CI: 4.92–9.33), with median time to onset of 53 days (Supplementary Table 3). Lorlatinib (ROR = 0.21, 99% CI: 0.07–0.69) had lower odds of conduction disease relative to other ALK inhibitors.
      Figure thumbnail gr1
      Figure 1Assessment of odds of arrhythmias and heart failure associated with inhibitors of ALK, BRAF, MEK, or EGFR relative to other targeted therapies collectively or relative to full VigiBase database. ALK inhibitors have 13 times higher odds of conduction disease and five times higher odds of long QT syndrome relative to other NSCLC targeted therapies collectively. BRAF and MEK inhibitors had 2.9 times higher odds of heart failure relative to other targeted therapies. EGFR inhibitors collectively had lower odds of conduction disease, QT prolongation, and heart failure relative to other targeted therapies. ADR, adverse drug reaction; CI, confidence interval; IC, information component; ROR, reporting OR.
      Figure thumbnail gr2
      Figure 2Assessment of odds of conduction disease associated with NSCLC targeted therapies. Crizotinib had significantly higher odds of conduction disease relative to other ALK inhibitors, to other targeted therapies collectively, and to the full database. Ceritinib and lorlatinib had lower odds of conduction disease relative to other ALK inhibitors. Gefitinib had two times higher odds relative to other EGFR inhibitors but lower odds relative to other targeted therapies collectively. ADR, adverse drug reaction; CI, confidence interval; IC, information component; ROR, reporting OR.
      The EGFR inhibitor gefitinib was associated with higher odds of conduction disease relative to other EGFR inhibitors (ROR = 2.17, 99% CI: 1.14–4.14) but decreased odds relative to other NSCLC targeted therapies (ROR = 0.49, 99% CI: 0.27–0.86).

      Long QT Syndrome

      EGFR inhibitors had increased odds of QT prolongation relative to the full database (ROR = 1.39, 99% CI: 1.08–1.80, IC025 = 0.18) but decreased odds relative to other targeted therapies (ROR = 0.26, 99% CI: 0.19–0.35). Nevertheless, osimertinib had significantly increased odds of QT prolongation relative to other EGFR inhibitors (ROR = 49.19, 99% CI: 25.88–93.47), other NSCLC targeted therapies (ROR = 6.13, 99% CI: 4.43–8.48), and the full VigiBase database (ROR = 14.34, 99% CI: 10.79–19.04, IC025 = 3.38), with median time to onset of 29.3 days (Fig. 3AC and Supplementary Table 4).
      Figure thumbnail gr3
      Figure 3Assessment of odds of long QT syndrome associated with inhibitors of (A) EGFR, (B) ALK, and (C) BRAF or MEK. (A) Of EGFR inhibitors, osimertinib was associated with 49 times higher odds of long QT syndrome relative to other EGFR inhibitors, six times higher odds relative to other targeted therapies collectively, and 14 times higher odds relative to the full database. (B) Crizotinib had 1.9 times higher odds of QT prolongation relative to other ALK inhibitors, 5.5 times higher odds relative to other targeted therapies, and 11 times higher odds relative to the full database. Ceritinib also had higher odds relative to other targeted therapies and the full database but not relative to other ALK inhibitors. (C) Dabrafenib and trametinib had higher odds of QT prolongation relative to the full database but not relative to other targeted therapies. ADR, adverse drug reaction; CI, confidence interval; IC, information component; ROR, reporting OR.
      ALK inhibitors had increased odds of QT prolongation (n = 171) relative to EGFR (n = 104) and BRAF (n = 75) inhibitors combined (ROR = 5.16, 99% CI: 3.92–6.81) and to the full VigiBase database (ROR = 9.18, 99% CI: 7.53–11.20, IC025 = 2.92) (Fig. 1). Crizotinib had increased odds of QT prolongation (Fig. 3B) relative to other ALK inhibitors (ROR = 1.91, 99% CI: 1.22–3.00), with median time to onset of 30 days and with 92.7% being severe cases (Supplementary Table 2). Ceritinib had increased odds of QT prolongation relative to other NSCLC targeted therapies (ROR = 3.43, 99% CI: 2.02–5.81) and the full database (ROR = 9.48, 99% CI: 5.70–15.75, IC025 = 2.41) but not compared with other ALK inhibitors, with median time to onset of 45 days (Supplementary Table 5).

      Supraventricular Tachycardia

      Osimertinib had increased odds of SVT (n = 44) relative to other EGFR inhibitors (ROR = 2.12, 99% CI: 1.39–3.24), other NSCLC targeted therapies (ROR = 1.90, 99% CI: 1.26–2.86), and the full VigiBase database (ROR = 1.71, 99% CI: 1.16–2.52, IC025 = 0.30) (Fig. 4), with median time to onset of 21 days and with 90.7% being severe (Supplementary Table 4). Dabrafenib had increased odds of SVT relative to other NSCLC targeted therapies (ROR = 1.43, 99% CI: 1.03–1.99), with median time to onset of 61.5 days (Supplementary Table 6).
      Figure thumbnail gr4
      Figure 4Assessment of odds of supraventricular tachycardia associated with NSCLC targeted therapies. The third-generation EGFR inhibitor osimertinib had approximately two times higher odds of supraventricular tachycardia relative to other EGFR inhibitors, other targeted therapies, and the full database. ADR, adverse drug reaction; CI, confidence interval; IC, information component; ROR, reporting OR.

      Ventricular Arrhythmias

      None of the targeted therapies were associated with increased odds of ventricular arrhythmias relative to other same-class drugs, NSCLC targeted therapies collectively, or the full VigiBase database (Fig. 5).
      Figure thumbnail gr5
      Figure 5Assessment of odds of ventricular arrhythmia associated with NSCLC targeted therapies. None of the ALK, BRAF, MEK, or EGFR inhibitors were associated with increased odds of ventricular arrhythmia relative to agents within the same treatment class, other targeted therapies collectively, or the full database. ADR, adverse drug reaction; CI, confidence interval; IC, information component; ROR, reporting OR.

      Heart Failure

      Osimertinib had increased odds of heart failure (Fig. 6) relative to other EGFR inhibitors (ROR = 6.75, 99% CI: 5.29–8.62), to other NSCLC targeted therapies (ROR = 3.64, 99% CI: 2.94–4.50), and to the full database (ROR = 5.37, 99% CI: 4.42–6.54, IC025 = 2.14), with median time to onset of 85 days.
      Figure thumbnail gr6
      Figure 6Assessment of odds of heart failure associated with NSCLC targeted therapies. Osimertinib was associated with 6.8 times higher odds of heart failure relative to other EGFR inhibitors, 3.6 times higher odds relative to other targeted therapies, and 5.4 times higher odds relative to the full database. Dabrafenib and trametinib were also associated with two to three times higher odds of heart failure relative to other targeted therapies and the full database. ADR, adverse drug reaction; CI, confidence interval; IC, information component; ROR, reporting OR.
      Dabrafenib and trametinib were associated with increased odds of heart failure relative to other targeted therapies (dabrafenib: ROR = 2.24, 99% CI: 1.86–2.70; trametinib: ROR = 2.44, 99% CI: 2.03–2.92) and to the full database (dabrafenib: ROR = 3.29, 99% CI: 2.79–3.89, IC025 = 1.50; trametinib: ROR = 3.50, 99% CI: 2.99–4.11, IC025 = 1.60), with median time to onset of 116 days for dabrafenib (Supplementary Table 6) and 87 days for trametinib (Supplementary Table 7).
      No ALK inhibitors had increased odds of heart failure relative to other same-class drugs or other NSCLC targeted therapies combined.

      Discussion

      To our knowledge, this study provides the first comparison of cardiovascular ADRs between and within classes of most often used NSCLC targeted therapies. This differs from previously published research that primarily compares cardiotoxicities between EGFR inhibitors alone.
      • Anand K.
      • Ensor J.
      • Trachtenberg B.
      • Bernicker E.H.
      Osimertinib-induced cardiotoxicity: a retrospective review of the FDA Adverse Events Reporting System (FAERS).
      In our analysis, ALK and ROS1 inhibitors, especially crizotinib, had strong associations with conduction disease (including bradycardia, atrioventricular block, bundle branch block, and intraventricular conduction delay) and QT prolongation. In contrast, BRAF and MEK inhibitors had stronger associations with heart failure; the third-generation EGFR TKI osimertinib had strong associations with QT prolongation, SVT, and heart failure.
      The odds of conduction disease and QT prolongation was 13 and five times higher, respectively, with ALK inhibitors relative to other NSCLC targeted therapies. These findings seemed to be driven by ALK and ROS1 inhibitors crizotinib and ceritinib and ALK inhibitor alectinib for conduction disease and by crizotinib for QT prolongation. In contrast, among ALK and ROS1 inhibitors, lorlatinib was associated with lower odds of conduction disease. To the best of our knowledge, these are novel findings with clinical implications. There was overlap between conduction disease and QT prolongation among ALK inhibitors, with 38.5% of ceritinib- and 18.4% of crizotinib-associated QT prolongation cases having conduction disease and 7.9% of alectinib-associated conduction disease cases having QT prolongation. Understanding the cardiovascular toxicities of alectinib and crizotinib is important given the prevalent usage of alectinib for ALK-rearranged NSCLC and of crizotinib for ROS1-rearranged NSCLC (in addition to NSCLC with MET abnormalities).
      • Vuong H.G.
      • Nguyen T.Q.
      • Nguyen H.C.
      • Nguyen P.T.
      • Ho A.T.N.
      • Hassell L.
      Efficacy and safety of crizotinib in the treatment of advanced non-small-cell lung cancer with ROS1 rearrangement or MET alteration: a systematic review and meta-analysis.
      ,
      • Landi L.
      • Chiari R.
      • Tiseo M.
      • et al.
      Crizotinib in MET-deregulated or ROS1-rearranged pretreated non–small cell lung cancer (METROS): a phase II, prospective, multicenter, two-arms trial.
      Our findings of an association between both crizotinib and alectinib and conduction disease are concordant with clinical trial data. Phase 3 PROFILE 1007 and 1014 trials detected 10% incidence of sinus bradycardia and 2% of QT prolongation from crizotinib.
      • Shaw A.T.
      • Kim D.W.
      • Nakagawa K.
      • et al.
      Crizotinib versus chemotherapy in advanced ALK-positive lung cancer.
      ,
      • Solomon B.J.
      • Mok T.
      • Kim D.W.
      • et al.
      First-line crizotinib versus chemotherapy in ALK-positive lung cancer.
      PROFILE 1005 and 1007 detected a mean decrease in heart rate of 25 beats per minute with crizotinib.
      • Ou S.H.
      • Tang Y.
      • Polli A.
      • Wilner K.D.
      • Schnell P.
      Factors associated with sinus bradycardia during crizotinib treatment: a retrospective analysis of two large-scale multinational trials (PROFILE 1005 and 1007).
      Given these risks, clinicians should be aware that patients initiating crizotinib should avoid concomitant use of beta blockers or non-dihydropyridine calcium channel blockers.

      Crizotinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2017.

      Similar findings have been found with alectinib; in phase 2 NP28761 and NP28673 studies, 4% of patients on alectinib developed sinus bradycardia (with mean decrease in heart rate of 11–13 beats per minute) and 2% developed other conduction diseases.
      • Morcos P.N.
      • Bogman K.
      • Hubeaux S.
      • et al.
      Effect of alectinib on cardiac electrophysiology: results from intensive electrocardiogram monitoring from the pivotal phase II NP28761 and NP28673 studies.
      Mouse models have revealed that crizotinib inhibits hyperpolarization-activated cyclic nucleotide-gated channel 4 in cardiomyocytes, potentially affecting the conduction system.
      • Zhang Z.
      • Huang T.Q.
      • Nepliouev I.
      • et al.
      Crizotinib inhibits hyperpolarization-activated cyclic nucleotide-gated channel 4 activity.
      Crizotinib may also contribute to bradycardia by decreasing testosterone levels.
      • Weickhardt A.J.
      • Rothman M.S.
      • Salian-Mehta S.
      • et al.
      Rapid-onset hypogonadism secondary to crizotinib use in men with metastatic nonsmall cell lung cancer.
      Further research is needed to confirm our findings and understand the mechanisms underlying differences in the odds of arrhythmias between different therapies.
      Compared with ALK, ROS1 and EGFR targeted therapies, BRAF and MEK inhibitors were associated with two to three times increased odds of heart failure. Although more common in melanoma, BRAF mutations are found in up to 4.9% of lung adenocarcinomas and 0.3% of lung squamous cell carcinomas.
      • Marchetti A.
      • Felicioni L.
      • Malatesta S.
      • et al.
      Clinical features and outcome of patients with non-small-cell lung cancer harboring BRAF mutations.
      In a phase 2 trial in patients with BRAF V600E-mutant NSCLC, 8% of patients on dabrafenib and trametinib developed heart failure with reduced ejection fraction.
      • Planchard D.
      • Smit E.F.
      • Groen H.J.M.
      • et al.
      Dabrafenib plus trametinib in patients with previously untreated BRAFV600E-mutant metastatic non-small-cell lung cancer: an open-label, phase 2 trial.
      Consequently, our data support the recommendations of the Food and Drug Administration to monitor left ventricular ejection fraction (LVEF) at baseline, 1 month, and every 2 to 3 months thereafter during dabrafenib or trametinib use.

      Trametinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2014.

      ,

      Dabrafenib [package insert]. Silver Spring, MD: Food and Drug Administration; 2018.

      Collectively, EGFR inhibitors were associated with lower odds of QT prolongation and heart failure relative to other NSCLC targeted therapies. Nevertheless, the third-generation EGFR inhibitor osimertinib had a strong signal for cardiotoxicity relative to other EGFR inhibitors and other targeted therapies. Of 4095 ADRs for osimertinib, 179 (3.7%) were heart failure cases and 143 (2.9%) were arrhythmias. Given that osimertinib has become the first-line preferred agent in EGFR-mutant NSCLC and was recently approved in the early stage adjuvant setting, understanding cardiotoxicity risk with osimertinib is clinically important.
      • Zhang Y.L.
      • Yuan J.Q.
      • Wang K.F.
      • et al.
      The prevalence of EGFR mutation in patients with non-small cell lung cancer: a systematic review and meta-analysis.
      ,
      • Wu Y.L.
      • Tsuboi M.
      • He J.
      • et al.
      Osimertinib in resected EGFR-mutated non-small-cell lung cancer.
      Osimertinib had an especially strong association with QT prolongation, with 49 times higher odds relative to other EGFR inhibitors and with approximately 4% of cases being fatal. The strong association between osimertinib and QT prolongation highlights importance of monitoring corrected QT interval in patients with previous history of QT prolongation or with concurrent use of other QT-prolonging medications.

      Osimertinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2018.

      Our practice is to obtain electrocardiograms at baseline, 2 weeks, and every 2 to 3 months to 6 months if no other QT-prolonging medications are newly prescribed. Osimertinib also had a strong association with SVT and heart failure. Most cases of osimertinib-associated cardiotoxicity were severe, with 25.0% of heart failure cases being fatal. This may reflect reporting bias as severe cases are more likely to be reported. There was also notable overlap among cardiotoxicities: among cases of osimertinib-associated QT prolongation, 14.6% also developed heart failure and 6.1% had ventricular arrhythmias. Similarly, 30.2% of osimertinib-associated SVT cases had heart failure.
      Clinical trials of osimertinib in NSCLC also found high rates of cardiac toxicities. The FLAURA trial comparing osimertinib to gefitinib and erlotinib in EGFR-mutated NSCLC found a higher rate of QT prolongation from osimertinib (10%) than gefitinib and erlotinib (5%)
      • Soria J.C.
      • Ohe Y.
      • Vansteenkiste J.
      • et al.
      Osimertinib in untreated EGFR-mutated advanced non-small-cell lung cancer.
      ,
      • Ramalingam S.S.
      • Vansteenkiste J.
      • Planchard D.
      • et al.
      Overall survival with osimertinib in untreated, EGFR-mutated advanced NSCLC.
      and higher rate of LVEF decrease from osimertinib (5%) than gefitinib and erlotinib (2%).
      • Ramalingam S.S.
      • Vansteenkiste J.
      • Planchard D.
      • et al.
      Overall survival with osimertinib in untreated, EGFR-mutated advanced NSCLC.
      Another study found that osimertinib had 2.2 times higher odds of heart failure, 2.1 times higher odds of atrial fibrillation, and 6.6 times higher odds of QT prolongation relative to other EGFR inhibitors for NSCLC.
      • Anand K.
      • Ensor J.
      • Trachtenberg B.
      • Bernicker E.H.
      Osimertinib-induced cardiotoxicity: a retrospective review of the FDA Adverse Events Reporting System (FAERS).
      Our study is the first to compare osimertinib to all NSCLC targeted therapies collectively, revealing a strong signal for QT prolongation, SVT, and heart failure.
      Given overlap between osimertinib-associated arrhythmias and heart failure with earlier time to onset for arrhythmias, it is important to promptly recognize arrhythmias to prevent tachycardia-induced cardiomyopathies. Although the Food and Drug Administration recommends regular LVEF assessments during dabrafenib and trametinib, similar guidelines do not exist for osimertinib, with current recommendations including LVEF evaluation at baseline and during treatment in those with cardiac risk factors.

      Osimertinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2018.

      Given predominant use of osimertinib in patients with lower incidence of cardiac risk factors (nonsmokers and women), the remarkable association between osimertinib and heart failure highlights a potential role for routine LVEF monitoring during osimertinib use, including in those without cardiac risk factors. Our current practice is to evaluate LVEF at baseline, 2 to 3 months, and 6 months in all patients regardless of risk factors. Although the mechanism underlying cardiotoxicity from osimertinib is not fully understood, osimertinib has been found through in vitro models to have greater activity against HER-2 compared with erlotinib and afatinib.
      • Cross D.A.
      • Ashton S.E.
      • Ghiorghiu S.
      • et al.
      AZD9291, an irreversible EGFR TKI, overcomes T790M-mediated resistance to EGFR inhibitors in lung cancer.
      Given known cardiotoxicity of HER-2–inhibiting monoclonal antibody trastuzumab, it is plausible that HER-2 inhibition underlies cardiotoxicity from osimertinib. Further research is needed to uncover mechanisms underlying osimertinib-associated cardiotoxicity.
      Regarding limitations, the pharmacovigilance database, VigiBase, allows signal detection to identify drug-ADR associations. Cardiac ADRs in VigiBase are reported by health professionals and nonhealth professionals to national pharmacovigilance centers which are reviewed and sent to VigiBase by more than 130 member countries of WHO Programme for International Drug Monitoring. Because not all suspected ADRs are reported to pharmacovigilance centers, our analysis is limited to reported cases. Analyzing VigiBase may be affected by reporting bias as individuals are more likely to report severe ADRs.
      In addition, owing to this lack of exposure data, we are unable to determine the incidence of cardiovascular ADRs using VigiBase. Because the reporting odds ratios in this study are calculated using the total number of ADRs for each drug as a surrogate for the total number of patients who received the therapies, this limitation has the potential to affect the magnitude of the reporting odds ratios in our study depending on the overall incidence of ADRs from each drug. For example, because EGFR mutations are more prevalent among patients with NSCLC than ALK rearrangements and BRAF mutations, if ADRs (including cardiac and noncardiac events) had a higher incidence from ALK inhibitors than EGFR inhibitors, then the calculated odds ratios of cardiac ADRs from ALK inhibitors relative to EGFR inhibitors would be skewed. Nevertheless, the landmark clinical trials on NSCLC targeted therapies have found similar incidence of adverse events from most often used ALK inhibitors, EGFR inhibitors, and BRAF and MEK inhibitors. For example, PROFILE 1005 trial detected a 96% incidence of any adverse event from crizotinib (38% incidence of grade 3–4 events)
      • Blackhall F.
      • Ross Camidge D.
      • Shaw A.T.
      • et al.
      Final results of the large-scale multinational trial PROFILE 1005: efficacy and safety of crizotinib in previously treated patients with advanced/metastatic ALK-positive non-small-cell lung cancer.
      ; FLAURA detected a 98% incidence of any adverse event from osimertinib (32% incidence of grade 3–4 events)
      • Soria J.C.
      • Ohe Y.
      • Vansteenkiste J.
      • et al.
      Osimertinib in untreated EGFR-mutated advanced non-small-cell lung cancer.
      ; the phase 3 EURTAC trial detected a 98% incidence of any adverse event from erlotinib (45% incidence of grade 3–4 events); and a phase 2 trial detected a 100% incidence of any adverse event from dabrafenib and trametinib combination therapy (69% incidence of grade 3–4 events).
      • Planchard D.
      • Smit E.F.
      • Groen H.J.M.
      • et al.
      Dabrafenib plus trametinib in patients with previously untreated BRAFV600E-mutant metastatic non-small-cell lung cancer: an open-label, phase 2 trial.
      Similar adverse event incidences have been reported with afatinib,
      • Wu Y.L.
      • Zhou C.
      • Hu C.P.
      • et al.
      Afatinib versus cisplatin plus gemcitabine for first-line treatment of Asian patients with advanced non-small-cell lung cancer harbouring EGFR mutations (LUX-Lung 6): an open-label, randomised phase 3 trial.
      lorlatinib,
      • Shaw A.T.
      • Bauer T.M.
      • de Marinis F.
      • et al.
      First-line lorlatinib or crizotinib in advanced ALK-positive lung cancer.
      and alectinib.
      • Peters S.
      • Camidge D.R.
      • Shaw A.T.
      • et al.
      Alectinib versus crizotinib in untreated ALK-positive non–small-cell lung cancer.
      The similar incidence of adverse events across the targeted therapies studied in our analyses makes it unlikely for our calculated odds ratios to be significantly skewed. In addition, the proportions of cardiac adverse events from each drug category are consistent with the findings from our disproportionality analyses. For example, among all ADRs reported with ALK inhibitors, 2.42% were conduction diseases; in contrast, conduction disease made up only 0.32% of all ADRs from BRAF inhibitors and 0.14% of all ADRs from EGFR inhibitors. This is consistent with our findings of a higher odds of conduction disease from ALK inhibitors relative to other NSCLC targeted therapies. Among all adverse events reported with BRAF inhibitors, 2.34% were reports of heart failure, whereas heart failure reports made up only 0.91% of ADRs from ALK inhibitors and 0.80% of ADRs from EGFR inhibitors. These proportions are also consistent with our findings of a higher odds of heart failure from BRAF inhibitors relative to other NSCLC targeted therapies.
      Furthermore, for drugs with previously characterized unique toxicities (e.g., QT prolongation with osimertinib), it is possible that patients receiving those drugs in the real world or clinical trial setting received more frequent monitoring for these toxicities compared with other drugs in VigiBase, leading to detection of more cases. Another limitation is lack of data on smoking status in VigiBase. Nevertheless, EGFR mutations, ALK rearrangements, and ROS1 rearrangements in lung cancer are more prevalent in never smokers than in former or current smokers, with three to six times higher odds of EGFR mutations, two to four times higher odds of ALK rearrangements, and three times higher odds of ROS1 rearrangements in never smokers relative to ever smokers.
      • Chapman A.M.
      • Sun K.Y.
      • Ruestow P.
      • Cowan D.M.
      • Madl A.K.
      Lung cancer mutation profile of EGFR, ALK, and KRAS: meta-analysis and comparison of never and ever smokers.
      ,
      • Zhu Q.
      • Zhan P.
      • Zhang X.
      • Lv T.
      • Song Y.
      Clinicopathologic characteristics of patients with ROS1 fusion gene in non-small cell lung cancer: a meta-analysis.
      This suggests that findings on cardiovascular events associated with EGFR, ALK, and ROS1 inhibitors are less likely driven by smoking and more by the therapies.
      Because of incomplete data in VigiBase on concurrent medications received by patients, we were unable to analyze for interactions between NSCLC targeted therapies and other drugs with cardiovascular effects, such as antidepressants, antiemetics, antiarrhythmic agents, and antiplatelet agents. Given that patients with NSCLC may be concurrently taking agents associated with QT prolongation (such as selective serotonin reuptake inhibitors or antiemetics, such as ondansetron), clinicians should be cognizant of the potential risks of concurrent use of NSCLC targeted therapies and other QT-prolonging medications. In addition, a promising development in the treatment of metastatic NSCLC is the use of combination therapy for patients with oncogenic driver mutations; for example, clinical trials are investigating the combination of EGFR TKIs and agents targeting vascular endothelial growth factor or the vascular endothelial growth factor receptor, such as bevacizumab or ramucirumab, respectively.
      • Jiang T.
      • Zhang Y.
      • Li X.
      • et al.
      EGFR-TKIs plus bevacizumab demonstrated survival benefit than EGFR-TKIs alone in patients with EGFR-mutant NSCLC and multiple brain metastases.
      • Yu H.A.
      • Paz-Ares L.G.
      • Yang J.C.
      • et al.
      Phase I study of the efficacy and safety of ramucirumab in combination with osimertinib in advanced T790M-positive EGFR-mutant non–small cell lung cancer.
      • Nakagawa K.
      • Garon E.B.
      • Seto T.
      • et al.
      Ramucirumab plus erlotinib in patients with untreated, EGFR-mutated, advanced non-small-cell lung cancer (RELAY): a randomised, double-blind, placebo-controlled, phase 3 trial.
      Because of risks of hypertension and thrombosis from antiangiogenesis agents, studies are needed to evaluate whether combining these drugs with EGFR TKIs is associated with higher risk of cardiovascular disease relative to EGFR TKI monotherapy.
      In conclusion, our pharmacovigilance database analysis reveals that several targeted therapies used in NSCLC are strongly associated with cardiac adverse events, such as crizotinib and alectinib with conduction disease and osimertinib with QT prolongation and heart failure. There is a need for an understanding of the mechanisms underlying these toxicities and for additional studies to establish standardized guidelines for monitoring, particularly for osimertinib, crizotinib, and alectinib.

      CRediT Authorship Contribution Statement

      Sarah Waliany: Conceptualization, Methodology, Statistical analysis, Writing—original draft preparation.
      Han Zhu: Conceptualization, Writing—review and editing.
      Heather Wakelee, Sukhmani K. Padda, Millie Das, Kavitha Ramchandran, Nathaniel J. Myall, Thomas Chen: Writing—review and editing.
      Ronald M. Witteles, Joel W. Neal: Conceptualization, Supervision, Writing—review and editing.

      Supplementary Data

      References

        • Skoulidis F.
        • Heymach J.V.
        Co-occurring genomic alterations in non-small-cell lung cancer biology and therapy.
        Nat Rev Cancer. 2019; 19: 495-509
        • Zhang Y.L.
        • Yuan J.Q.
        • Wang K.F.
        • et al.
        The prevalence of EGFR mutation in patients with non-small cell lung cancer: a systematic review and meta-analysis.
        Oncotarget. 2016; 7: 78985-78993
        • Kim H.
        • Chung J.H.
        Overview of clinicopathologic features of ALK-rearranged lung adenocarcinoma and current diagnostic testing for ALK rearrangement.
        Transl Lung Cancer Res. 2015; 4: 149-155
        • Bergethon K.
        • Shaw A.T.
        • Ou S.H.
        • et al.
        ROS1 rearrangements define a unique molecular class of lung cancers.
        J Clin Oncol. 2012; 30: 863-870
        • Marchetti A.
        • Felicioni L.
        • Malatesta S.
        • et al.
        Clinical features and outcome of patients with non-small-cell lung cancer harboring BRAF mutations.
        J Clin Oncol. 2011; 29: 3574-3579
        • Kravchenko J.
        • Berry M.
        • Arbeev K.
        • Lyerly H.K.
        • Yashin A.
        • Akushevich I.
        Cardiovascular comorbidities and survival of lung cancer patients: Medicare data based analysis.
        Lung Cancer. 2015; 88: 85-93
        • Wang K.
        • Eblan M.J.
        • Deal A.M.
        • et al.
        Cardiac toxicity after radiotherapy for stage III non-small-cell lung cancer: pooled analysis of dose-escalation trials delivering 70 to 90 Gy.
        J Clin Oncol. 2017; 35: 1387-1394
        • Anand K.
        • Ensor J.
        • Trachtenberg B.
        • Bernicker E.H.
        Osimertinib-induced cardiotoxicity: a retrospective review of the FDA Adverse Events Reporting System (FAERS).
        JACC CardioOncol. 2019; 1: 172-178
        • Lindquist M.
        VigiBase, the WHO global ICSR database system: basic facts.
        Drug Inf J. 2008; 42: 409-419
        • Norén G.N.
        • Hopstadius J.
        • Bate A.
        Shrinkage observed-to-expected ratios for robust and transparent large-scale pattern discovery.
        Stat Methods Med Res. 2013; 22: 57-69
        • Norén G.N.
        • Bate A.
        • Orre R.
        • Edwards I.R.
        Extending the methods used to screen the WHO drug safety database towards analysis of complex associations and improved accuracy for rare events.
        Stat Med. 2006; 25: 3740-3757
        • Bate A.
        • Lindquist M.
        • Edwards I.R.
        • et al.
        A Bayesian neural network method for adverse drug reaction signal generation.
        Eur J Clin Pharmacol. 1998; 54: 315-321
        • Vuong H.G.
        • Nguyen T.Q.
        • Nguyen H.C.
        • Nguyen P.T.
        • Ho A.T.N.
        • Hassell L.
        Efficacy and safety of crizotinib in the treatment of advanced non-small-cell lung cancer with ROS1 rearrangement or MET alteration: a systematic review and meta-analysis.
        Target Oncol. 2020; 15: 589-598
        • Landi L.
        • Chiari R.
        • Tiseo M.
        • et al.
        Crizotinib in MET-deregulated or ROS1-rearranged pretreated non–small cell lung cancer (METROS): a phase II, prospective, multicenter, two-arms trial.
        Clin Cancer Res. 2019; 25: 7312-7319
        • Shaw A.T.
        • Kim D.W.
        • Nakagawa K.
        • et al.
        Crizotinib versus chemotherapy in advanced ALK-positive lung cancer.
        N Engl J Med. 2013; 368: 2385-2394
        • Solomon B.J.
        • Mok T.
        • Kim D.W.
        • et al.
        First-line crizotinib versus chemotherapy in ALK-positive lung cancer.
        N Engl J Med. 2014; 371: 2167-2177
        • Ou S.H.
        • Tang Y.
        • Polli A.
        • Wilner K.D.
        • Schnell P.
        Factors associated with sinus bradycardia during crizotinib treatment: a retrospective analysis of two large-scale multinational trials (PROFILE 1005 and 1007).
        Cancer Med. 2016; 5: 617-622
      1. Crizotinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2017.

        • Morcos P.N.
        • Bogman K.
        • Hubeaux S.
        • et al.
        Effect of alectinib on cardiac electrophysiology: results from intensive electrocardiogram monitoring from the pivotal phase II NP28761 and NP28673 studies.
        Cancer Chemother Pharmacol. 2017; 79: 559-568
        • Zhang Z.
        • Huang T.Q.
        • Nepliouev I.
        • et al.
        Crizotinib inhibits hyperpolarization-activated cyclic nucleotide-gated channel 4 activity.
        Cardiooncology. 2017; 3: 1
        • Weickhardt A.J.
        • Rothman M.S.
        • Salian-Mehta S.
        • et al.
        Rapid-onset hypogonadism secondary to crizotinib use in men with metastatic nonsmall cell lung cancer.
        Cancer. 2012; 118: 5302-5309
        • Planchard D.
        • Smit E.F.
        • Groen H.J.M.
        • et al.
        Dabrafenib plus trametinib in patients with previously untreated BRAFV600E-mutant metastatic non-small-cell lung cancer: an open-label, phase 2 trial.
        Lancet Oncol. 2017; 18: 1307-1316
      2. Trametinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2014.

      3. Dabrafenib [package insert]. Silver Spring, MD: Food and Drug Administration; 2018.

        • Wu Y.L.
        • Tsuboi M.
        • He J.
        • et al.
        Osimertinib in resected EGFR-mutated non-small-cell lung cancer.
        N Engl J Med. 2020; 383: 1711-1723
      4. Osimertinib [package insert]. Silver Spring, MD: Food and Drug Administration; 2018.

        • Soria J.C.
        • Ohe Y.
        • Vansteenkiste J.
        • et al.
        Osimertinib in untreated EGFR-mutated advanced non-small-cell lung cancer.
        N Engl J Med. 2018; 378: 113-125
        • Ramalingam S.S.
        • Vansteenkiste J.
        • Planchard D.
        • et al.
        Overall survival with osimertinib in untreated, EGFR-mutated advanced NSCLC.
        N Engl J Med. 2020; 382: 41-50
        • Cross D.A.
        • Ashton S.E.
        • Ghiorghiu S.
        • et al.
        AZD9291, an irreversible EGFR TKI, overcomes T790M-mediated resistance to EGFR inhibitors in lung cancer.
        Cancer Discov. 2014; 4: 1046-1061
        • Blackhall F.
        • Ross Camidge D.
        • Shaw A.T.
        • et al.
        Final results of the large-scale multinational trial PROFILE 1005: efficacy and safety of crizotinib in previously treated patients with advanced/metastatic ALK-positive non-small-cell lung cancer.
        ESMO Open. 2017; 2e000219
        • Wu Y.L.
        • Zhou C.
        • Hu C.P.
        • et al.
        Afatinib versus cisplatin plus gemcitabine for first-line treatment of Asian patients with advanced non-small-cell lung cancer harbouring EGFR mutations (LUX-Lung 6): an open-label, randomised phase 3 trial.
        Lancet Oncol. 2014; 15: 213-222
        • Shaw A.T.
        • Bauer T.M.
        • de Marinis F.
        • et al.
        First-line lorlatinib or crizotinib in advanced ALK-positive lung cancer.
        N Engl J Med. 2020; 383: 2018-2029
        • Peters S.
        • Camidge D.R.
        • Shaw A.T.
        • et al.
        Alectinib versus crizotinib in untreated ALK-positive non–small-cell lung cancer.
        N Engl J Med. 2017; 377: 829-838
        • Chapman A.M.
        • Sun K.Y.
        • Ruestow P.
        • Cowan D.M.
        • Madl A.K.
        Lung cancer mutation profile of EGFR, ALK, and KRAS: meta-analysis and comparison of never and ever smokers.
        Lung Cancer. 2016; 102: 122-134
        • Zhu Q.
        • Zhan P.
        • Zhang X.
        • Lv T.
        • Song Y.
        Clinicopathologic characteristics of patients with ROS1 fusion gene in non-small cell lung cancer: a meta-analysis.
        Transl Lung Cancer Res. 2015; 4: 300-309
        • Jiang T.
        • Zhang Y.
        • Li X.
        • et al.
        EGFR-TKIs plus bevacizumab demonstrated survival benefit than EGFR-TKIs alone in patients with EGFR-mutant NSCLC and multiple brain metastases.
        Eur J Cancer. 2019; 121: 98-108
        • Yu H.A.
        • Paz-Ares L.G.
        • Yang J.C.
        • et al.
        Phase I study of the efficacy and safety of ramucirumab in combination with osimertinib in advanced T790M-positive EGFR-mutant non–small cell lung cancer.
        Clin Cancer Res. 2021; 27: 992-1002
        • Nakagawa K.
        • Garon E.B.
        • Seto T.
        • et al.
        Ramucirumab plus erlotinib in patients with untreated, EGFR-mutated, advanced non-small-cell lung cancer (RELAY): a randomised, double-blind, placebo-controlled, phase 3 trial.
        Lancet Oncol. 2019; 20: 1655-1669