long-term electrocardiographic safety monitoring in ... · long-term electrocardiographic safety...

Cardiac Safety Research Consortium

Long-term electrocardiographic safetymonitoring in clinical drug development:A report from the Cardiac SafetyResearch Consortium
Jonathan P. Piccini, MD, MHS, a Richard L. Clark, BS, b Peter R. Kowey, MD, c Suneet Mittal, MD, d
Preston Dunnmon, MD, e Norman Stockbridge, MD, e James A. Reiffel, MD, f Mintu P. Turakhia, MD, MAS, g

Paul D. Ziegler, MS, b Robert B. Kleiman, MD, h Fraz Ismat, MD, i and Philip Sager, MD j Durham, NC; Mounds View,MN; Philadelphia, PA; Ridgewood, Princeton, NJ; New York, NY; and Stanford, CA

This white paper, prepared by members of the Cardiac Safety Research Consortium (CSRC), discusses important issuesregarding scientific and clinical aspects of long-term electrocardiographic safety monitoring during clinical drug development.To promote multistakeholder discussion of this topic, a Cardiac Safety Research Consortium–sponsored Think Tank was held on2 December 2015 at the American College of Cardiology's Heart House in Washington, DC. The goal of the Think Tank wasto explore how and under what circumstances new and evolving ambulatory monitoring technologies could be used to improveand streamline drug development. This paper provides a detailed summary of discussions at the Think Tank: it does notrepresent regulatory guidance. (Am Heart J 2017;187:156-169.)

A Think Tank sponsored by the Cardiac Safety ResearchConsortium (www.cardiac-safety.org) was convened atthe American College of Cardiology's Heart House inWashington, DC, on December 2, 2015, to discusslong-term electrocardiographic (ECG) safety monitoringduring clinical drug development. Based on the principlesof the United States Food and Drug Administration'sCritical Path Initiative,1 the CSRC was created in 2006 tofacilitate collaborations among academicians, industryprofessionals, and regulators to develop consensusapproaches addressing cardiovascular safety issues thatcan arise in the development and therapeutic use ofmedical products.2

From the aDuke University Medical Center & Duke Clinical Research Institute, Durham, NC,bMedtronic Diagnostics and Monitoring Research, Mounds View, MN, cLankenau HeartInstitute and Jefferson Medical College, Philadelphia, PA, dThe Valley Health System,Ridgewood, NJ, eDivision of Cardiovascular and Renal Products, US Food and DrugAdministration, fDivision of Cardiology, Columbia University, Medical Center, New York,NY, gVA Palo Alto Health Care System and Stanford University, Palo Alto, CA, heResearchTechnology, Inc. (ERT), Philadelphia, PA, iBristol-Myers Squibb, Princeton, NJ, andjStanford University, Stanford, CA.DisclaimerThe opinions and conclusions expressed in this article are solely the views of theauthors and do not necessarily represent views of the United States Food and DrugAdministration or other author affiliations.Submitted January 24, 2017; accepted January 24, 2017.Reprint requests: Jonathan P. Piccini, MD, MHS, Duke University Medical Center & DukeClinical Research Institute, Durham, NC.E-mail: [email protected]© 2017 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.ahj.2017.01.012

A key issue in drug development is the potentialfor drugs to have deleterious electrophysiologiceffects. Although drug-induced changes in ECG inter-vals (eg, the PR, QRS, and QT intervals) are readilyevaluated via careful ECG collection and analysis, theelucidation of actual drug-induced arrhythmias such asclinically significant bradyarrhythmias, atrial fibrilla-tion (AF), or ventricular tachycardia (VT) is difficult tocapture and assess using traditional ambulatory mon-itoring techniques, including Holter monitoring. Thesetraditional techniques have limited sensitivity due tothe limited temporal sampling for ECG analysis.Recently, approaches to collect longer periods ofECG data using patch, implanted, real-time telemetry,or mobile technologies have created the potential formore comprehensive safety and, in some cases,efficacy assessments during pharmacologic develop-ment programs. However, these technologies are notcost neutral, generate a tremendous amount of data,and increase the risk of detecting “ambient” or back-ground arrhythmia of little or no clinical significance.This CSRC Think Tank convened to explore how andunder what circumstances these new and evolvingambulatory monitoring technologies could be used toimprove and streamline drug development. This articlesummarizes the Think Tank findings by a broad rangeof experts, now further extended by the CSRC writinggroup. The CSRC views expressed herein do notrepresent regulatory policy.

Table II. Characteristics of ambulatory ECG monitors

ECG type

Typicalduration of

monitoring (d)

Monitors forasymptomaticarrhythmias

Provides real-timefeedback of ECGbeing acquired

irst-generationAECG monitorstandard Holter 1-2 Yes Noatient-activatedevent recorder

30 No⁎ No†

oop recorder 30 No Noecond-generationAECG monitorsatch-based Holter 14 Yes Nombulatory telemetry 30 Yes Yeshird-generationAECG monitorsatch-basedtelemetry

30 Yes Yes

martphone-basedapplications

Indefinite No⁎ Yes

sertable cardiacmonitors

Several years Yes No‡

Patient can screen for asymptomatic arrhythmias by acquiring an ECG randomly.Patient can transmit acquired data via a telephone line for further review; if thisappens soon after the ECG is acquired, it provides near–real-time information to thelinician.ECG data acquired during the day are downloaded that evening and then availabler review the next day.

Table I. ECG parameters for consideration

• Heart rate• Ectopic beats• Repolarization morphologyo T wave alternans

• Beat-to-beat variabilityo Heart rate variability

• Heart rate dependent repolarization abnormalitieso Corrected QT interval

• Autonomic nervous systemo Heart rate turbulence

Adapted from Farkas and Nattel.3

Piccini et al 157American Heart JournalVolume 187, Number 0

Historic development, current state, andlimitations of cardiac monitoringECGmonitoring plays a critical role in the assessment of

safety (and in some cases efficacy) of a new cardiovas-cular drug. Table I outlines the ECG parameters ofconcern in cardiovascular safety.3 In particular, there isthe need to detect (1) a change in a specific ECGparameter (eg, interval measurements such as heart rate,PR interval, QRS duration, and QT interval as well as ECGmorphologic changes), (2) development of a newarrhythmia (eg, atrial or ventricular ectopy, AF, non-sustained VT, torsade de pointes [TdP]), or (3) suppres-sion of an existing arrhythmia. There are few tools thatare currently available for extended ECG monitoring inclinical practice. Each has its inherent strengths andlimitations.4,5 Moreover, the background rates for dys-rhythmias are poorly known, so decisions about whatarrhythmias might be related to drug treatment are poorlyinformed. Thus, there is no accepted criterion standardfor extended ECG monitoring to assess cardiac safety (orefficacy of a cardiovascular drug) during the drugdevelopment process. To determine the optimal methodfor ECG monitoring in the assessment of drug safety,the following 3 critical questions must be kept in mind(Table II): (1) how long do we need to monitor (informedby pharmacokinetics and patient population), (2) do weneed to monitor for asymptomatic arrhythmias, and (3) isreal-time ECG feedback necessary? This section reviewsthe strengths and limitation of the various forms of ECGmonitoring available today (Figure 1).

Twelve-lead ECGHistorically, the 10-second 12-lead ECG has served as

the criterion standard for assessment of drug-inducedECG effects. Although well validated, the 12-lead ECGalso has important limitations. First, from a practicalstandpoint, it is burdensome for a patient or investiga-tional site to record ECGs multiple times over the courseof a day or multiple times per week. Second, the standard10-second 12-lead ECG is not well suited to picking upchanges that may be occurring only intermittently, or

A

F

SP

LS

PAT

P

S

In

⁎†hc‡fo

asymptomatically, or during physiologic conditions suchas exercise or sleep. In this regard, the evolving field ofnovel ambulatory ECG monitoring technology providespromise for extended continuous monitoring that pro-vides enhanced disclosure with enhanced arrhythmiccharacterization, that is, presence, frequency, duration,and burden.

First-generation (“legacy”) ECG monitoring toolsHistorically, 24-48 hours of Holter monitoring has been

used when a single ECG is insufficient to make a clinicaldiagnosis or too brief to assess a safety or efficacy issueduring drug development. Holter monitors are well suitedto determine the average heart rate and heart rate range,quantify atrial and ventricular ectopy counts, determinewhether more persistent forms of AF are present, andcorrelate rate and rhythm findings with activity and withsymptoms if present (via the diary or event marker).Information regarding the shortest and longest durationof AF, burden of AF, the heart rate during AF, and patternof initiation and termination of AF (and other arrhyth-mias) can be determined during the monitored period.3

However, it is generally impractical to wear a Holtermonitor for more than 24-48 hours because of therequirements to change batteries and flashcards, skinirritation from the skin electrodes, and the inconvenienceof wearing a device connected to the skin electrodes bymultiple cables. A 1- or 2-day monitoring period maysimply be too short to detect infrequent but importantatrial or ventricular arrhythmias or episodes of

Figure 1

Spectrum of ambulatory external ECG monitoring modalities. As one moves from left to right, the duration of monitoring increases, which in turnincreases the diagnostic yield. ILR, implantable loop recorder; LINQ, Medtronic LINQ ILR (Minneapolis, MN). Adapted from Mittal et al.5

158 Piccini et alAmerican Heart Journal

Month 2017

conduction block. Longer-term monitoring is possibleusing patient-activated event recorders and nonmemoryloop recorders; however, both require an intervention bythe patients before an ECG recording is stored, andtherefore, these devices are better suited for assessmentof symptomatic arrhythmias. When assessing safety andefficacy of a drug in development, both longer-term ECGmonitoring and the ability to capture information aboutboth symptomatic as well as asymptomatic arrhythmiasare required. Limitations of these first-generation toolsinclude lack of real-time feedback on recorded ECGevents, challenges to patient acceptance and adherence,and the logistical burden for operationalizing withinclinical research (ie, consistent equipment availability,flash cards, data transfers, and central adjudication).

Second-generation ECG monitoring toolsSecond-generation ECG monitoring technologies gen-

erally have the ability to monitor for both symptomaticand asymptomatic arrhythmias for an extended durationof time. These include patch-based extended Holtermonitors, autotriggered memory loop monitors, andlead-based mobile cardiac telemetry systems.5,6 Themajor difference among the 3 systems is the ability toobtain real-time feedback on the ECG. In the case ofpatch-based Holters, no information is available until theECG data are acquired, the monitoring period iscompleted, the patch is returned to the manufacturerfor analysis, and a report is created for physician review.

Lead-based mobile cardiac telemetry systems overcomethis limitation, although the requirement for a 24/7monitoring center which must be staffed at all timesmarkedly increases the costs. Patch-based extendedHolter systems with a median 11 days of monitoringhave been shown to detect more events than a 24-hourHolter monitor.7

Third-generation ECG monitoring toolsFurther refinements have led to the development of

third-generation systems. These include smartphone-basedsystems where an electrode-embedded module is attachedto a patient's smartphone. The module detects electricalimpulses from the user's fingertips and transmits signals tothe smartphone when contact is maintained with theelectrodes. The recording can be stored as a PDF file thatcan then be directly e-mailed from the phone to theclinician. The advantages to this type of technologyinclude the use of inexpensive hardware and the captureof real-time and high-fidelity ECG data that are immediatelyavailable to the patient (albeit with many implications thatrequire careful consideration in research). In addition,ECG data can be intermittently acquired over an indefiniteperiod of time. A limitation of this type of system is that theECG data are only collected when the patient makescontact with the electrodes, that is, not continuousmonitoring to characterize arrhythmic burden and asymp-tomatic events, and it is really best suited for intermittentmonitoring as controlled by the user/patient.


The other type of third-generation system uses patch-based real-time telemetry. These systems combine thepatient convenience of patch-based systems with theability to acquire all ECG data and the capability ofallowing near–real-time review. Additional benefitsinclude the following: early notification of any arrhyth-mia of concern (specifically useful when the studycompound has a potential arrhythmic signal), the abilityto remotely monitor patient compliance with ECGmonitoring, and reduction in “dead time” inherent towaiting for data to enable clinical or study decisionmaking, that is, potential to reduce study timelines orreduce patient or site burden.Again, an importantconsideration is whether the cost of maintaining a 24/7 ECG monitoring center may preclude the use of suchsystems during drug development.

Insertable cardiac monitorsImplantable subcutaneous ECG monitors can collect

ECG data over a period of several years. As an example,the most recent iteration of the device (Reveal LINQ™,Medtronic Inc) is small enough to allow simpler insertion,records cardiac information automatically in response todetected arrhythmias (or on demand based on patientactivation), and uses cardiac telemetry to transmit data tothe physician. These type of devices are particularly wellsuited to defining the occurrence (including time to firstoccurrence) and burden of episodes of AF or otherinfrequent arrhythmias.7 Although the costs and invasivenature of this technology are limitations for clinicaltrial application, it is the only device available thatallows continuous ECG monitoring for up to 3 yearsand that can detect arrhythmia burden over time(including AF) without the need for active interventionby the patient.8

Electrocardiographic monitoring technology is evolv-ing and provides novel approaches for integration in drugdevelopment. There are clinical benefits and limitationsto each technology. In the context of clinical research,the technology must be validated and supported witha scalable compliant global operational process. Thesmartphone-based and patch-based technologies holdpromise (Figure 2).

Cardiac monitoring and big dataBig data have revolutionized healthcare, including

cardiovascular medicine.9 Although there is a growth ofbig data in general, this is particularly true in the arena ofarrhythmia science and heart rhythm medicine. Forexample, the availability of continuous monitoring datafrom cardiac implanted electronic devices (CIEDs) innationwide remote monitoring databases has allowedclinicians, industry, and other stakeholders to assessdevice and lead performance in a much more timelyfashion with much greater power to detect significant

differences in device and hardware function.10 Theseadvantages also extend to improved and more detaileddiagnosis in individuals with suspicion for arrhythmia.For example, in one study of 120,000 patients monitoredwith patch-based Holter technology and median analyz-able time of 9 days, up to 20% had evidence of ventriculararrhythmia (18% nonsustained and 1.4% sustained).11

Thus, analysis of big datasets in continuous monitoringhas shown that sensitivity is clearly increased. This alsoraises the question as to what is “ambient” backgroundarrhythmia in different populations (eg, young, old,structural heart disease).Big data sources have also facilitated comparative

effectiveness analyses that compare emerging technol-ogies with more established therapies.12 Continuousmonitoring in CIEDs has also allowed for more“in-depth” data, even without very “big data.” Forexample, continuous monitoring in only 1,195 CIEDpatients has demonstrated that patterns of AF andtemporal persistence are highly variable acrossphysician-based clinical classifications of AF type. Inthis study, authors observed that fewer than 50% ofpatients diagnosed with paroxysmal AF truly hadparoxysmal AF that terminated within 7 days as definedby continuous device monitoring.13

Given these contributions in the clinical and arrhythmiascience realm, how can “big data” in heart rhythmdiagnostics positively impact drug safety assessment?There are many possibilities, including continuous QTmonitoring in larger populations and improved predic-tive modeling of arrhythmic risk by harnessing dataacross multiple populations with different drug expo-sures (and thus different target channel profiles).Continuous monitoring will enable even more compre-hensive analyses of emerging QT analysis strategies,including dynamic beat-to-beat QT assessment and ECGrestitution (the QT-TQ relationship).14 “Big data” mayalso be helpful in improving the characterization ofambient arrhythmia in studied populations and thus helpdiscern proarrhythmia (signal) from background arrhyth-mia (noise). The availability of these large data sets willalso benefit from new data analytic methodologies,including machine learning and “deep learning.”That said, there are many limitations to the use of big

data. First, big data often lack depth. Although a givendata source may contain many observations, thenumber of characteristics at the individual level isusually less than traditional data formats (ie, administra-tive claims versus registry data). Loss of critical variablesthat influence prognosis may predispose “big data”analyses to greater levels of residual confounding. Inaddition, big data analyses are complex and requireautomation and verification to ensure validity. Finally,given the real-time nature of many big data analyses,methods to ensure frequent updating of core data areimportant.15

Figure 2

Promising ambulatory external ECG monitoring techniques to assess the QT/QTc interval. (Top) Novel smartphone-based application to assess theheart rate and produce all 6 limb leads of the ECG. In this manner, lead II (a commonly used lead to assess the QT/QTc interval) can be obtainedand analyzed as many times as clinically appropriate. The acquired ECG data can be transmitted to the clinician using e-mail. (Image reproducedwith permission of AliveCor) (Bottom) A patch monitor is applied in the left infraclavicular space in an orientation designed to mirror lead II of thesurface 12-lead ECG. The acquired ECG data are downloaded to a receiver carried by the patient, which transmits the information using cellularsignal to a central monitoring station. In this manner, real-time information about the heart rate and arrhythmias is available. (Image reproducedwith permission from Medtronic.)


Month 2017

When should cardiac monitoring beused in drug development?BradyarrhythmiaThe assessment of any potential ECG and dysrhythmic

effects is an important part of the development process forall new pharmaceutical agents. Of chief concern is thedevelopment of new tachyarrhythmias, including bothsupraventricular proarrhythmia (eg, AF) and ventricularproarrhythmia (eg, TdP). However, significant bradyarrhyth-mias can also occur and can pose significant risk in somepatients.The type, duration, and timing of monitoring are

important variables in ECG assessment. Selecting thetype and timing of monitoring must be based, in part, onthe pharmacokinetics of both the parent compound andany active metabolites. Drug-induced ECG changes aretypically stable by the time steady state is reached,whereas the development of arrhythmias may bedependent not only upon attainment of steady state butalso upon a change of concomitant medications, changeof electrolyte or metabolic status, comorbid disorders,and/or changes in autonomic tone. For example, if a drugprolongs the QT interval modestly with an increased riskfor TdP, such TdP may not become manifest until thepatient adds another drug that further prolongs the QT

interval, such as erythromycin; the patient developshypokalemia or hypomagnesemia, such as with gastro-enteritis; or the patient becomes bradycardic, such asduring a vasovagal episode.The type and duration of monitoring are also

dependent upon issues of practicality. For a drug witha short half-life (eg, 4-6 hours) (parent, active metabo-lite) and no drug or disease interactions, an appropriatemonitoring period regarding potential bradycardia/conduction changes will most likely be the first24-48 hours. Serial ECGs, Holter monitoring, orin-hospital telemetry would all suffice. In the case of adrug with a prolonged and uncertain half-life (eg,amiodarone), many drug interactions, or active metab-olites, the timing and duration of bradyarrhythmias aredifferent than those for short half-life drugs, as would beits monitoring requirements. Finally, drugs that caninduce hypothyroidism or other secondary causes ofbradyarrhythmia with long-term therapy (such asamiodarone and lithium) require prolonged recordingtechniques (such as mobile cardiac outpatient teleme-try [MCOT], 30-day loop recorders, daily transtelepho-nic tracings over time, or insertable cardiac monitors)and/or repeat periods of monitoring depending uponthe addition/development or removal/resolution ofinteracting agents or disorders.

Figure 3

A textual diagram of the reported cases of Harvoni-associated bradyarrhythmias.


Case study in drug-induced bradyarrhythmia: HarvoniPostmarketing cases of serious and life-threatening

symptomatic bradycardia, as well as 1 fatal cardiac arrestand several cases requiring pacemaker insertion, havebeen reported16-22 when either Harvoni (a fixed-dosecombinationof ledipasvir/sofosbuvir) or Sovaldi (sofosbuvir)was given to patients taking amiodarone (Figure 3).Bradycardia has been observed within the first few hoursor days of initiation, as well as up to 2 weeks aftercoadministration. Accordingly, cautions have been addedto the package inserts of these agents. More recently, 3more cases of bradyarrhythmias associatedwith sofosbuvir-based regimens were reported—this time, only 1 inassociation with amiodarone.23

The mechanism of action for Harvoni- or Sovaldi-associated bradycardia is uncertain. All patients hadunderlying cardiac disease, concomitant β-blocker therapy,and/or advanced liver disease. It is possible that the newdrug altered the If current, already prone to dysfunctiondue to the underlying disease or concomitant medica-tion, and/or impaired conduction out of the sinus node(SN). Several characteristics of these cases suggest acausal association: (1) the short time to symptom onsetfrom starting either Harvoni or Sovaldi combined with adirect-acting antiviral, (2) resolution of symptoms upondrug discontinuation, and (3) recurrence of symptomsupon rechallenge. Understanding the mechanism un-derlying the Harvoni-amiodarone story might helpprevent a similar outcome with other agents in the

future. Without this understanding, it is difficult to knowhow one might extrapolate this experience to futuredrug development or which patients might needprophylactic monitoring.

SN, atrioventricular node, His-PurkinjetissueSinus node function, atrioventricular node (AVN)

function, and His-Purkinje (H-P) conduction can all beimpaired by medications and result in bradyarrhythmia.Furthermore, SN and AVN function can also be affectedby changes in autonomic balance due to drug effects. Theextent to which their function becomes abnormal is inlarge part dependent upon any baseline dysfunctionpresent, the dose of the drug given, its direct and indirectactions, any interacting coadministered agents, andsometimes the time at which it is examined. With respectto the latter, a drug that is parasympathomimetic mightonly produce substantial SN or AVN dysfunction at timeswhen physiologic vagotonia is also high, such as duringsleep or gastrointestinal distress, when the effects can besynergistic.Importantly, ECG monitoring is the major noninvasive

method to assess the effect of a drug on these specializedcardiac tissues. However, we need to recognize that theinterpretation of monitor recordings requires knowledgeof the normal range of function seen in these tissuesacross (1) the diurnal cycle, (2) activity levels during


Month 2017

monitoring, and (3) the patient's age. Consider the AVnode as an example:Atrioventricular Wenckebach and even periods of

complete heart block at the AVN can be normal duringsleep in young or physiologically trained healthy individ-uals. This has been demonstrated by monitoring studiesin military recruits and in medical students, amongothers.24 However, AV Wenckebach and completeheart block in older individuals, especially during wakefulhours, are pathologic. Hence, the interpretation offindings must be made with an understanding of whatobservations represent physiology within the normalrange in the population being examined and whichrepresent pathophysiology that requires further study ortreatment. This challenge becomes even more importantas more sensitive monitoring technologies are introducedinto research and clinical practice.Medications that are parasympathomimetic or are

direct depressants of the AVN may produce high-degreeAV block—but do so most commonly if there isunderlying dysfunction present prior to their administra-tion (such as noted by a prolonged PR interval) or if thereis disease-mediated AVN injury concomitant with theiradministration (such as with an acute inferior infarction).Accordingly, if a drug with potential adverse effects onthe AVN is given to a patient with either baselinedysfunction or a concomitant potentially interactingcondition, the probability of developing a bradycardicresponse should be greater. It should also increase fromdrug initiation to attainment of steady state. Hence, ECGmonitoring would best be used from initiation untilsteady state is reached. A similar precaution should betaken if a patient with manifest H-P dysfunction, such asbundle-branch block, is given a drug with sodiumchannel–blocking properties.Understanding a new drug's pharmacology and phar-

macokinetics should help determine what risks may beencountered and what type and duration of monitoringshould be used during its development. When a scenariosuch as the Harvoni story is encountered, the mostimportant consideration to address is the mechanism,which will play a major role in determining whatmonitoring should be used going forward.

Representative examplesThere are well-known and underappreciated medica-

tions that can affect the SN and AVN. Among thewell-known agents are β-blockers, nondihydropyridinecalcium channel blockers, sympatholytic agents (includ-ing centrally acting ones such as clonidine), amiodarone,digitalis (infrequently at the SN), and ivabradine (at theSN). Less well-known but also important are lithium(more at the SN), cimetidine, ticagrelor, and others. Somedrugs rarely affect SN or AVN function in healthy heartsbut can have adverse effects if underlying dysfunction

(recognized or not) is present, including flecainide andpropafenone. Examples of drugs that can affect H-Pfunction include sodium channel blockers, including classIA and IC antiarrhythmics, tricyclic antidepressants, andintravenous amiodarone. In the Cardiac Arrhythmia PilotStudy,25 new conduction abnormalities were seen as oftenwith imipramine as with encainide and flecainide.

Steady state should affectmonitoring durationSteady state is generally thought to be attained after 5

half-lives. However, many drugs have a wide range ofhalf-lives, depending upon the patient's age, hepatic andrenal function, genetic factors, and concomitant therapies.Figure 4 uses flecainide as an example. In this figure, thesignificant variability in half-life can be seen.When monitoring a patient after dosing until steady

state is attained (parent compound and any activemetabolite), to assess ECG effects, safety, and possiblyefficacy before increasing the dose or changingtreatment—especially in the outpatient setting—it isbest to use the maximal, not the mean, half-life whenestimating the optimal duration of ECG monitoring.For example, monitoring flecainide for 24 hours may

suffice if its half-life were short (eg, 3-5 hours) butshould be almost 6 days if the half-life were 29 hours.Thus, monitoring might require a Holter (24 hours) or1 week of a patch monitor, autotriggered loop record-ing, MCOT, or daily ECG tracings. This same consider-ation would be appropriate for agents underdevelopment once their pharmacokinetic characteris-tics are known.

How should we monitor forbradyarrhythmias indrug development?Considering the example of Harvoni, where the

observation of significant bradyarrhythmias only becameapparent long after its approval and clinical use, and onlyunder the specific circumstance of amiodarone coadmin-istration, can regulators be better prepared to detect suchissues prior to a new drug's approval? The challenge is notsimply what type of monitoring to use but to understandthe mechanism of such interactions so that perhaps theymay be more predictable and investigations mostappropriately designed and timed during the drugdevelopment stage. Short of that, recognizing whichagents are known to affect SN, AVN, and H-P tissues(channels, receptors, currents, autonomic interactions),and by what mechanism, plus extrapolation to newagents with similar mechanistic characteristics coulddetermine what monitoring approach might best be usedduring phase I-III trials (Table III). The Harvoni case

Figure 4

Variation in pharmacokinetics observed with flecainide.


reminds us that we do not know all we need to knowduring drug development. Moving forward, when asignificant pharmacologicallyinduced bradyarrhythmiaoccurs with a mechanism that is uncertain, regulatorsshould consider requiring that additional mechanisticstudies be initiated as a requirement for continuingmarketing of the drug.

TachyarrhythmiaLong-term ECG monitoring is often considered during

drug development to document tachycardias—bothsupraventricular arrhythmias (primarily AF) as well asVTs (primarily nonsustained VT). This is most commonlyperformed using Holter monitors with recording dura-tions of 24-48 hours but may also involve use of long-termECG monitoring with event recorders, MCOT, or evenimplantable monitoring devices. The decision to performlong-term ECG monitoring during a drug developmentprogram is generally related to concerns that a therapymay either promote the initiation of arrhythmias directlyor increase the prevalence of arrhythmias in a particularlysusceptible population with preexistent cardiac disease,where patients are already at substantial risk ofarrhythmias.Drugs that do not have effects on cardiac ion channels

or autonomic tone are unlikely to initiate arrhythmias. Inthe absence of any mechanism that might promotearrhythmias, long-term ECG monitoring would generallynot be warranted during a drug development program.When long-term ECG monitoring is performed during

drug development, any detected tachyarrhythmias are farmore likely to represent the background prevalence ofarrhythmia in healthy individuals than drug-relatedeffects. It is often difficult or impossible to determine ifan arrhythmia detected in an individual subject in aclinical trial is related to a drug-mediated effect or is amanifestation of underlying cardiac disease or the lowprevalence of tachyarrhythmias in healthy individuals.Thus, when long-term monitoring is performed and asmall number of tachyarrhythmias are detected, it may beimpossible to tell if a small imbalance in the eventfrequency between treatment groups represents a truedrug-related effect or not. In light of ambient arrhythmiadetected in healthy individuals during long-term ECGmonitoring, it is prudent to consider long-term ECGmonitoring only when a drug has characteristics whichmight promote arrhythmias or when the treatmentpopulation is at particularly high risk. It can be helpfulto have a relatively long period of monitoring prior todrug administration to better define an individual'sbackground arrhythmia frequency.

Supraventricular tachyarrhythmiasFew drugs currently available are unequivocally known

to produce supraventricular tachyarrhythmias in healthyindividuals. Many drugs, however, do produce increasesin sinus heart rate. Other than adrenergic agonists, mostdrugs that increase chronotropy increase the heart rateby only a few beats per minute. It may be very difficult todetect such an effect with standard vital sign assessmentsor with data recorded from standard ECGs, but long-term

Table III. Proposed monitoring for arrhythmias during drug development

Monitoring technology Potential circumstances for use

ECG • Daily monitoring from dosing until achievement of steady state• Documentation with symptoms

Holter monitor (24-48 h) • Daily monitoring from dosing until achievement of steady state• Documentation with frequent but intermittent symptoms(b1-2 d between episodes)

14-d patch ambulatory monitor • Drugs with longer-half lives30-d ambulatory monitor with memory (loop monitor) • Documentation of intermittent symptoms less frequent

(N48 h between episodes)MCOT • High risk of severe arrhythmia

• Detection of arrhythmias that require a rapid response• If remotely monitoring patient compliance with monitoring is desired• If reduced timeline to obtaining data provides valueo Reduced study timelineso Reduction in site or patient burden

Implantable loop monitor • Drugs with very long half-lives• Rare and infrequent arrhythmias• If defining time to reoccurrence or burden of arrhythmicepisodes is desired


Month 2017

monitoring with Holter recorders is very useful fordocumenting small drug-induced changes in heart rate.Holter recordings allow assessment of heart rate over afull 24-hour period or during shorter periods; mostHolters can report mean heart rate on an hourly basisor even in smaller increments. This may be useful forrelating heart rate changes to drug dosing and to drugpharmacokinetics.However, in patients who may have preexisting cardiac

disease or pulmonary disease, some drugs are thought toinduce AF (and atrial flutter). These drugs includeadrenergic agonists (dobutamine), adenosine, digoxin,and anticholinergics. There has been concern thatinhaled β-agonists and anticholinergics may induce AF,particularly in patients with severe chronic obstructivepulmonary disorder, many of whom may also havepreexistent cardiovascular disease and very high risk ofAF. Similar concerns hold for patients with severe heartfailure and left ventricular dysfunction, who are at highrisk for both ventricular and supraventricular arrhyth-mias. It is therefore common during drug development oftherapies targeted at populations with underlying leftventricular dysfunction or severe chronic obstructivepulmonary disorder to include some assessment of therisk of induction of AF. This may include Holtermonitoring (24-48 hours). However, Holter monitoringmay be inadequate to detect a small increase in theincidence of AF. More extended monitoring for a periodof up to a month is more likely to document asymptom-atic episodes of self-terminating or paroxysmal AF. Amajor concern, however, is that long-term monitoring in

susceptible populations is more likely to detect AFunrelated to drug therapy. It is therefore critical tocollect adequate-duration pretreatment ECG monitoringdata and, whenever possible, to provide for a placebo orat least standard-of-care control group. Careful attentionto the duration of pretreatment monitoring and to therandomization ratio is also warranted to avoid biaseddetection of episodes of AF.Adrenergic agonists may also promote the develop-

ment of reentrant supraventricular tachycardias (SVTs)(atrial tachycardias, AV nodal reentry, or AV reentry) insusceptible individuals. Such patients may already havehad episodes of SVT in the past, but adrenergicstimulation may promote SVT in patients with dual AVnode pathways or an accessory AV pathway who havenever had documented SVT previously. Drugs whichhave indirect effects on autonomic tone may alsopromote AF or SVT indirectly; for instance, a potentvasodilator may lead to a large sympathetic responsewhich might provoke supraventricular tachyarrhythmiasin a susceptible patient.

Ventricular tachyarrhythmiasDrugs may also cause ventricular proarrhythmia, which

may also be detected during drug development vialong-term ECG monitoring. The ventricular arrhythmiathat provokes the greatest concern during drug develop-ment is drug-induced TdP.26 Fourteen drugs have beenremoved from the market worldwide because ofunanticipated drug-induced TdP27 generally due todrug-induced block of the hERG-encoded IKr potassium


channel. Long-term ECG monitoring is not, however,likely to be useful for detection of drug-induced TdPduring clinical drug development programs because of itsextreme rarity even with relatively potent torsadogenicdrugs. Although class I and class III antiarrhythmics mayproduce TdP in 1%-3% of patients, most of the drugswithdrawn because of TdP produced documentedarrhythmias far less frequently. It has been estimatedthat terfenadine produced TdP in approximately 1 in50,000 prescriptions. It is therefore impractical to detectdrugs which are at risk for drug-induced TdP withlong-term ECG monitoring, and instead, the use ofsurrogate markers (as described in the ICH E14 guidance)is recommended.Drugs that block the cardiac sodium channel NaV1.5

have also been associated with VTs (tricyclic antidepres-sants, flecainide in patients with heart failure and reducedsystolic function). Few such drugs are in common use,and the incidence of VT in patients receiving sodiumchannel blockers (other than antiarrhythmic agents) isunknown. During the development of a new drug withknown sodium channel blockade in which the channelkinetics (slow onset and offset kinetics are moreconcerning than faster kinetics) and/or effect on theQRS interval raises significant concern, it may beappropriate to perform long-term ECG monitoring toallay concerns about drug-induced VT. For drugs thoughtto have a significant risk of inducing VT, initial monitoringin an inpatient setting with continuous telemetry wouldbe most appropriate. For drugs thought to have a possiblelow risk of inducing VT, outpatient dosing, possibly whilepatients wear an MCOT device, capable of detectingsustained ventricular arrhythmias and immediately noti-fying a full-time monitoring center who are able to notifyemergency services may be appropriate. For such drugs,monitoring with a Holter recorder or standard eventrecorder would not be sufficient because of the lack ofreal-time arrhythmia detection and activation of emer-gency responders.Nonpharmacologic therapies for heart failure, such as

stem cell therapy, have also raised concerns aboutincreasing ventricular proarrhythmia. Early trials ofskeletal muscle–derived stem cells seemed to show anincrease in ventricular tachyarrhythmias; current trials ofcardiac stem cells now generally include long-term ECGmonitoring to assess for the risk of promoting VT.

Dose-finding studiesFirst-in-human (FIH) phase I ascending-dose studies and

phase I/II dose-finding studies in patients are trials inwhich long-term ECG monitoring may also be used.These trials generally involve administration of new drugs(or dosages of established drugs which have never beentested) to either healthy individuals or individuals withthe condition of clinical concern. In such cases, thehuman tolerability and adverse effect profiles are not

known. In these studies, inpatient telemetry may be usedto detect potential arrhythmic adverse events. However,long-term ECG monitoring may also be of value. A newagent might produce bradyarrhythmias (SN slowing orAV block) or tachyarrhythmias (SVT or VT); long-termECG monitoring during trials testing increasing drugdosages might detect a subtle, but clinically relevant,signal during dose escalation.In reality, most ascending-dose phase I trials in healthy

volunteers or phase I/II patient trials generally use verysmall cohort sizes, and detection of new brady- ortachyarrhythmias during such trials would be veryunlikely. For drugs that have not shown any signals ofconcern during preclinical testing (no ion channel blockin patch clamp studies; no arrhythmias or ECG intervalchanges during telemetered animal studies), long-termECG monitoring during ascending-dose trials is generallynot warranted. As discussed previously, it is far morelikely that an arrhythmia detected in such a small trialrepresents underlying cardiac pathology or simply theoccasional occurrence of AV block or tachyarrhythmiasin healthy populations.However, for drugs that do have a preclinical signal

suggesting a potential for arrhythmogenesis, a knownclass effect, or for use in a very high risk population,long-term ECGmonitoring may be warranted during earlyascending-dose trials. Common wisdom would suggestthat the incidence of arrhythmias should rise with dosageand exposure, although there are clearly drugs that havedifferent ion channel effects at different exposures andfor which the risk of arrhythmia may not be directly doseproportional. It is therefore recommended that if long--term ECG monitoring is being considered, one shouldperform the same extent of monitoring at all planneddosages. It is also recommended that careful consider-ation be given to the duration of pretreatment monitoring(to document the baseline occurrence of arrhythmias)and to the use of a placebo or standard-of-care controlgroup. Again, the presence of an adequately sizedpretreatment assessment period or control group isimportant to discriminate proarrhythmia versus ambientarrhythmia.

Pragmatic useLong-term ECG monitoring has an important role in

clinical drug development, but a number of consider-ations suggest that its indiscriminate use is likely to resultin more confusion than clear benefit. Foremost amongthese is our relatively poor understanding of the truebackground occurrence of arrhythmias in healthy volun-teers as well as in patient populations. If AV block andnonsustained SVT and VT never occurred in healthyindividuals, it would be a simple matter to use long-termECG monitoring in clinical trials and, if any arrhythmiaswere detected, to conclude that they were drug induced.However, AV block, short episodes of SVT, and even


Month 2017

episodes of nonsustained VT (eg, benign outflow tractVT) are often documented in healthy individuals with nostructural heart disease and on no medications. Further-more, the incidence of these arrhythmias increases withage and in the presence of cardiovascular and pulmonarydisease and is more likely to be observed with longerversus shorter monitoring. Discrimination of morphology(monomorphic inferiorly directed axis versus polymor-phic) also provides important information. High-qualitymultilead ECG monitoring can be very helpful in thisregard.It is therefore important that long-term ECG monitoring

be used appropriately during drug development pro-grams. For drugs with a clean preclinical slate, long-termECG monitoring is far more likely to detect unrelatedarrhythmias than drug-related effects. Furthermore, thelonger the monitoring duration, the higher the likelihoodof detecting “baseline” arrhythmias. One should there-fore use long-term ECG monitoring in circumstanceswhere it is likely to be helpful. For drugs with a knownclass effect, such as S1P1 modulators, which are knownto produce first-dose bradycardia and AV block, long-termECG monitoring is critical for detecting the risk ofarrhythmic adverse events and for determining howclinicians should best monitor for and treat sucharrhythmias. Careful attention should be paid to trialdesign. One should ensure an adequate duration ofpretreatment monitoring to detect patients who mayhave preexisting arrhythmias and to document thebaseline incidence of arrhythmias in the trial population.Careful consideration should be given to having anadequate-sized placebo or standard-of-care treatmentgroup. In cases where a placebo or other control groupis not feasible, one might consider a longer duration ofpretreatment monitoring or potentially long-term ECGmonitoring after the end of drug treatment.Consideration should also be given to the type of

long-term ECG monitoring to be used. Standard 24-hourHolter recordings may be sufficient to detect (or exclude)a large increase in arrhythmias but may be inadequate todetect a small increase in either bradycardic or tachycar-dic events. In such cases, either a larger sample size or useof longer-term ECG monitoring with one of the newtechnologies such as MCOT for 2-4 weeks, or even aninsertable cardiac monitor, might be required.

Interpretation of monitoringdata and limitationsDiscerning pathologic versus “ambient” arrhythmia: thepitfalls of small datasetsBy the time a new drug reaches the stage of large phase

III trials, sponsors and investigators typically understandthe channel-blocking profile of the drug being tested andhopefully have some insight into the organ-specificmetabolic toxicities of a compound and/or its metabolites

that were demonstrated during nonclinical testing.Pharmacokinetic profiles of the parent molecule and itsmetabolites are usually well understood. In the circum-stance of a known membrane-active channel-blockingagent for which the overall proarrhythmic profile iseither unclear or potentially different for different subjectsubstrates (eg, underlying coronary artery disease or not,important myocardial hypertrophy or not, poor leftventricular function or not), phase III trials can bedesigned and powered appropriately to delineate benefit/risk profiles in subpopulations that may be vulnerable todrug-induced proarrhythmia or drug-induced failures ofimpulse generation and conduction. Cardiac rhythmsafety can be assessed based on placebo-adjustedoccurrences of these types of events.In contrast, FIH studies and phase I single- and

multiple-dose exploration studies are typically conductedat a time when much about the human safety of the drugis unknown, and are generally performed in small subjectcohorts. These small cohort sizes make individual patientassessments mandatory because of the potentially highinter-subject variability in the occurrence of arrhythmias,with each subject serving as his or her own control forassessments of the on-drug occurrence or worsening ofarrhythmic events (typically ventricular but occasionallyatrial as well). Accordingly, the primary source ofvariability that limits the interpretation of arrhythmiasin early development is the intra subject variability ofthese events that can be affected by diurnal variation,autonomic tone, dietary factors, and environmentalfactors. In this circumstance, if an individual's ambientarrhythmia baseline is either poorly defined (or notdefined at all), the documentation of new potentiallyimportant arrhythmias during telemetry in the phase Isetting can be profoundly problematic; cause substan-tial program delays; and, in the worstcase scenario,cause the development of a promising, potentiallylife-saving therapy to be wrongly abandoned. Thefollowing example of real-life drug X illustrates thispoint.Case example. Drug X is a first-in-class therapy for a

serious and often fatal disease. During preclinical animaltesting, elevated cardiac troponin-I (cTnI) was seen inseveral animals of 2 species, and 1 animal experienced a4-beat run of nonsustained ventricular tachycardia(NSVT) following exposure to this drug. There were nocardiac histopathological changes, and there were nosafety issues during FIH and phase I single ascending dosetesting. However, during the human multiple ascendingdose testing, 2 drug-related, treatment-emergent adverseevents occurred. First, several subjects developed tran-sient low-grade cTnI elevation. Second, an episode ofNSVT occurred in an apparently healthy, middle-agedsubject (Figure 5).This 14-beat, N3 second episode of monomorphic

NSVT (rate N 200 beats per minute) caused much

Figure 5

Nonsustained VT complicating monitoring in drugdevelopment with evidence of low-grade troponin elevation in an animal model.


concern because it had a left superior axis that is nottypical for either outflow tract VT or fascicular VT(common “normal heart” VTs), and this subject had nohistory of ischemic cardiovascular disease, and his QT/QTc was not prolonged. Because of the small number ofsubjects that had been exposed to this drug, togetherwith the troponin elevations and NSVT that had beenseen in animal testing, and because there was no predrugbaseline monitoring to document NSVT as a baselinearrhythmia for this subject, this development programwas placed on full regulatory hold in the interest of trialparticipant safety. However, there was ongoing concernboth within the drug-specific Review Division and theDivision of Cardiovascular and Renal Products that thefew findings in the animals may not have beendrug-related and that the ECG could have representedthis subject's normal ambient ventricular arrhythmiaprofile.Given the potential importance of this first-in-class

therapy for a fatal disease, the sponsor pushed forwardwith a comprehensive strategy to either confirm orexclude the possibility that drug-induced cardiotoxicitycaused this arrhythmia. This strategy included thefollowing:

• Longer-term preclinical studies were repeated withstringent handling, acclimation, and animal selec-tion criteria, and no evidence of cardiovasculartoxicity was seen.

• In vitro and human ECG data did not suggest that thisdrug or its metabolites affected the QTc interval.

• The individual who experienced NSVT underwentextensive post-trial cardiac evaluations to furtherdefine baseline cardiovascular status includingmagnetic resonance imaging, echocardiography,stress-perfusion exercise testing, and thyroid func-tion testing. The results of all of these were normal.Thirteen-day Holter monitoring off any drugsdemonstrated a run of monomorphic NSVT onmonitoring day 11 as well as frequent prematureventricular contractions (PVCs) on monitoring day13. After reviewing this information, the sponsor'sexternal cardiovascular expert review panel deter-mined that the probability was low that the NSVTthat this subject experienced in the clinical trial wasdrug-related.

Based on these findings, clinical testing was allowed toresume with the incorporation of baseline andon-treatment 13-day Holter monitoring, cardiac telemetryduring drug initiation, ECGs, cTnI assessments, sequentialechocardiograms, and stringent exclusion criteria basedon screening Holter monitoring. Importantly, this processof post-trial subject evaluation and preclinical studyrepetition caused an approximately 1-year delay in thisprogram. Had the involved subject not been willing totake part in poststudy cardiac testing, a study restart maynot have been possible at all.


Month 2017

How much monitoring is necessary? The questionthat frequently arises from sponsors is, “How muchpost-exposure monitoring is necessary?” Given whatoccurred in the example of drug X, we would encouragesponsors and investigators to think about this issue morebroadly, to include the question, “How much pre-expo-sure monitoring is necessary?” These are related ques-tions, and as is often the case, the answers depend onwhere a new therapy falls on the continuum ofcardiotoxic risk and, separately, on the vulnerability ofthe patient population in which it is intended to be used.Thinking about postexposure monitoring first, it has

been shown previously that “…increasing the duration ofrecording from 24 to 36 hours increased the probabilityof detecting maximal and more complex grade PVCs by25% and 50%, respectively.”28,29 We also note that thesubject of concern in the above example of drug X didnot demonstrate NSVT in posttrial follow-up testing untilday 11 of long-term continuous monitoring off drug.Therefore, we see several possible lines of reasoning forthe intensity of postexposure monitoring, some examplesof which would include the following:

• A limited monitoring approach for a drug for whichthere is long experience with the class beingnoncardiotoxic and forwhich therewere no concernsfor cardiotoxicity raised in preclinical studies.

• A limited or, depending on the target patientpopulation, intermediate monitoring approach fora first-in-class therapy which has no importantchannel-blocking activity (reported as IC50 valuesfor the panel of channels assessed), no metabolicblocking properties (either for nucleic acid medi-ated grow-repair-reproduction or for intracellularenergetics pathways), and no evidence of cardio-toxicity in preclinical testing.

• A more extensive monitoring approach for drugsthat have (or belong to a class that has) knowncardiotoxic effects, either in channel blockingstudies or in preclinical studies, or are known tobe metabolic blockers of some variety.

The issue, then, that the drug X example brings up iswhat to do if, from limited monitoring, a worrisomefinding emerges. As in that case, monitoring had to beintensified to define whether the observed NSVT was adrug toxicity or an ambient arrhythmia. In the absence ofbaseline testing, a year-long program delay occurred. Thisargues for some degree of baseline monitoring in allsubjects undergoing postexposure monitoring. In thedrug X example, the NSVT did not recur until monitoringday 11 off drugs. New technologies make longer-durationmonitoring feasible. The degree to which sponsorsincorporate comprehensive baseline monitoring intodevelopment programs will likely be influenced bynewness of the class, experience with other class

members, and preclinical testing results.Some mayconsider 24 to 48 hours of preexposure baselinemonitoring as part of phase I drug exposures, eitherprior to or at the time of phase I unit admission. In theend, this is a balance between the cost of obtainingbaseline monitoring information versus the risk of asubstantial program delay in the absence of thatinformation. These decisions will undoubtedly be madeon a case-by-case basis.

Summary of recommendationsIn summary, evolving technologies for cardiac rhythm

monitoring will enable improved collection of heartrhythm function during drug development. Similarly,these technologies will also present new challengesalongside their benefits. Based upon the input frommultiple stakeholders in academic medicine, clinicalpractitioners, industry, and regulatory bodies, the follow-ing consensus recommendations have been put forward.

1. The selection of heart rhythm monitoring technol-ogies during drug development should be princi-pally based upon the half-life of the drug andmetabolites, the proposed mechanism of action ofany potential proarrhythmia, and the underlyingrisk of the studied population (Table III).

2. A single bipolar ECG recording represents theminimum standard for cardiac rhythm monitoring.The device should be medical grade with appro-priate regulatory agency approval. Devices such assmart watches that use non-ECG electrode modal-ities, including but not limited to photoplethysmo-graphy and pulse rate accelerometry, should not beused until there is sufficient evidence, regulatoryapproval, and clinical consensus.

3. The intensity of heart rhythm monitoring should beguided by the level of anticipated risk. The highestintensity of monitoring should be reserved for drugswith cardiotoxic effects demonstrated in nonclinicalinvestigation or other analyses (Table III).

4. Cardiac rhythmmonitoring during drug developmentis of greatest value when pretreatment monitoring isincorporated to help discern background arrhythmiaversus drug-induced proarrhythmia.

5. Signals for potential arrhythmia need to be inter-preted in light of the underlying treated population(eg, incidence of AF in younger versus elderlypatients).

DisclosuresJonathan P. Piccini receives grants for clinical research

from ARCA biopharma, Boston Scientific, Gilead, Johnson& Johnson, and St Jude Medical Spectranetics and servesas a consultant to Amgen, GSK, Janssen Scientific,Medtronic, and Spectranetics.


Paul D. Ziegler is an employee of Medtronic and reportsstock ownership.Richard L. Clark is an employee of Medtronic and

reports stock ownership.Peter R. Kowey reports serving as a consultant to

Medtronic and to Cardionet.Suneet Mittal serves as a consultant to Boston Scientific,

Medtronic, and St Jude Medical.Preston Dunnmon reports no disclosures.Norman Stockbridge reports no disclosures.Fraz Ismat is an employee of BMS.Mintu P. Turakhia serves as a consultant to Medtronic,

St Jude, Janssen, iRhythm, and AliveCor and receivesresearch support from Medtronic and Janssen.Robert B. Kleiman is an employee of ERT, providing

ECG Core Laboratory services to and cardiac safetyconsulting services to pharma.Philip Sager consults with Medtronic, Inc, and Alivecor.

References

1. FDA. Innovation or stagnation: challenge and opportunity on thecritical path to new medical products. http://www.fda.gov/ScienceResearch/SpecialTopics/CriticalPathInitiative/CriticalPathOpportunitiesReports/ucm077262.htm 2004. [AccessedDecember 4, 2015].

2. Finkle J, Bloomfield D, Uhl K, et al. New precompetitive paradigms:focus on cardiac safety. Am Heart J 2009;157(5):825-6.

3. Farkas AS, Nattel S. Minimizing repolarization-related proarrhythmic riskin drug development and clinical practice. Drugs 2010;70(5):573-603.

4. ICH harmonized tripartite guideline E14. The clinical evaluation ofQT/QTc interval prolongation and proarrhythmic potential fornon-antiarrhythmic drugs. http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Efficacy/E14/E14_Guideline.pdf 2016. [Accessed February, 2016].

5. Mittal S, Movsowitz C, Steinberg JS. Ambulatory external electro-cardiographic monitoring: focus on atrial fibrillation. J Am CollCardiol 2011;58(17):1741-9.

6. Fung E, Jarvelin MR, Doshi RN, et al. Electrocardiographic patchdevices and contemporary wireless cardiac monitoring. Front Physiol2015;6:149.

7. Barrett PM, Komatireddy R, Haaser S, et al. Comparison of 24-hourHolter monitoring with 14-day novel adhesive patch electrocardio-graphic monitoring. Am J Med 2014;127(1):95.e11-97.

8. Mittal SRJ, Sarkar S, Koehler J, et al. Real-world performance of anenhanced atrial fibrillation detection algorithm in an insertablecardiac monitor. Heart Rhythm 2016.

9. Antman EM, Benjamin EJ, Harrington RA, et al. Acquisition, analysis,and sharing of data in 2015 and beyond: a survey of the landscape: aconference report from the American Heart Association Data Summit2015. J Am Heart Assoc 2016;13:1624-30.

10. Sung RK, Massie BM, Varosy PD, et al. Long-term electrical survivalanalysis of Riata and Riata ST silicone leads: National VeteransAffairs experience. Heart Rhythm 2012;9(12):1954-61.

11. Solomon MD, Yang J, Sung SH, et al. Incidence and timing of potentiallyhigh-risk arrhythmias detected through long term continuous ambulatoryelectrocardiographic monitoring. BMC Cardiovasc Disord 2016;16:35.

12. Turakhia MP, Cao M, Fischer A, et al. Reduced mortality associated withquadripolar compared to bipolar left ventricular leads in cardiacresynchronization therapy. JACC Clin Electrophysiol 2016;2(4):426-33.

13. Charitos EI, Purerfellner H, Glotzer TV, et al. Clinical classifications ofatrial fibrillation poorly reflect its temporal persistence: insights from1,195 patients continuously monitored with implantable devices. JAm Coll Cardiol 2014;63(25 Pt A):2840-8.

14. Fossa AA, Zhou M. Assessing QT prolongation and electrocardiog-raphy restitution using a beat-to-beat method. Cardiol J 2010;17(3):230-43.

15. Schneeweiss S. Learning from big health care data. N Engl J Med2014;370(23):2161-3.

16. FDA. Hepatitis C treatments containing sofosbuvir in combination withanother direct acting antiviral drug: drug safety communication—serious slowing of heart rate when used with antiarrhythmic drugamiodarone. http://www.fda.gov/safety/MedWatch/safetyinformation/safety/alertsforhumanmedicalproducts/ucm439662.htm. [Accessed September 9, 2016].

17. EMA. EMA recommends avoidance of certain hepatitis C medicines andamiodarone together. http://www.ema.europa.eu/ema/index.jsp?curl=pages/news_and_events/news/2015/04/news_detail_002313.jsp&mid=WC0b01ac058004d5c1. [Accessed March 2, 2017].

18. Renet SCM, Antonini T, et al. Extreme bradycardia after first dose ofsofosbuvir and declatasvir in patients receiving amiodarone: 2 casesincluding a rechallenge. Gastroenterology 2015;149:1378-80.

19. Back DJ, Burger DM. Interactions between amiodarone andsofosbuvir-based treatment for hepatitis C viral infection: potentialmechanisms and lessons to be learned. Gastroenterology 2015;149:1315-7.

20. Fazel YLB, Golabi P, Younossi Z. Safety analysis of sofosbuvir andledipasvir for treating hepatitis C. Expert Opin Drug Saf 2015;14:1317-76.

21. Sofosbuvir + amiodarone: bradycardia and conduction disturban-cesPrescrire Int 2015;24(166):294-5.

22. Ledipasvir + sofosbuvir (Harvoni). A therapeutic advance in genotype1 hepatitis C virus infection, despite uncertaintiesPrescrire Int2015;24(166):285-9.

23. Fontaine H, Lazarus A, Pol S, et al. Bradyarrhythmias associated withsofosbuvir treatment. N Engl J Med 2015;373(19):1886-8.

24. Bigger JT, Reiffel JA. Ambulatory electrocardiography. In: Platia E,ed. Non-phmaracologic management of cardiac arrhythmias.Lippincott Co.; 1986.

25. Cardiac Arrhythmia Pilot Study I. Effects of encainide, flecainide,imipramine and moricizine on ventricular arrhythmias during the yearafter acute myocardial infarction: the CAPS. Am J Cardiol1988;61(8):501-9.

26. Piccini JP, Pritchett EL, Davison BA, et al. Randomized, double-blind,placebo-controlled study to evaluate the safety and efficacy of a singleoral dose of vanoxerine for the conversion of subjects with recent onsetatrial fibrillation or flutter to normal sinus rhythm: RESTORE SR. HeartRhythm 2016;13(9):1777-83.

27. Vicente J, Stockbridge N, Strauss DG. Evolving regulatory paradigmfor proarrhythmic risk assessment for new drugs. J Electrocardiol2016;49(6):837-42.

28. Min SS, Turner JR, Nada A, et al. Evaluation of ventricular arrhythmiasin early clinical pharmacology trials and potential consequences for laterdevelopment. Am Heart J 2010;159(5):716-29.

29. Kennedy HL, Chandra V, Sayther KL, et al. Effectiveness of increasinghours of continuous ambulatory electrocardiography in detectingmaximal ventricular ectopy. Continuous 48 hour study of patientswith coronary heart disease and normal subjects. Am J Cardiol1978;42(6):925-30.

long-term electrocardiographic safety monitoring in ... · long-term electrocardiographic safety...

Documents