Remote Cardiac Monitoring For Clinical Trials - FNIH

1m ago
1.65 MB
27 Pages

Remote Cardiac Monitoring for Clinical TrialsCase Study Working Group:Remote Digital Monitoring WorkshopFebruary 18 – 19, 2020Elena IzmailovaDan BloomfieldJason HomsyQi LiuWilliam WoodVadim ZipunnikovJohn Wagner

Overview Problem statement Statement of needStudies used for cardiac monitoring caseExperimental design and key findingsContext of useRelationship to the existing biomarker evidentiary criteria frameworkBenefit and risk assessmentsState of evidenceStatistical considerationsRemote cardiac monitoring in clinical trialsQ&A panel Key topics to be presentedfor each Case Study2

Disclaimer for the Cardiac Monitoring Case Study Both studies described in publications used as a starting point forthis working group were designed and initiated in Q1 and Q2 of2016, prior to the mobile technologies CTTI recommendations and2018 Biomarker Qualification Evidentiary Framework FDAGuidance being available publicly3

Evidentiary Criteria Framework Certain elements of safety profile (HR, RR, bodytemp) can be built comprehensively in early stageclinical trials using wearable devices to collectdense continuous data both during the clinicalpharmacology unit (CPU) confinement and afterthe discharge from the CPU Collect the data in the real-world settings Early signal detection can inform doseadjustments or discontinuation of drugcandidates with safety liabilities earlier in thedrug development processNeedstatement Benefits of continuous monitoring using wearable devices Earlier detection of a potential safety signal Dose adjustment Early discontinuation of drug candidates with an unfavorable safety profile A reliable pharmacodynamic assessment if related to drug MOARisks False negative - missing a potential safety signal False positive - time consuming data review and reporting COU may be different than intended use and indications of use stipulated under 510(k) clearance Missing dataRisk mitigation Validation should be performed according to the COU pertinent to a specific clinical trial Establishing analytical validity and statistical methods for continuous ambulatory monitoringCOUVital signs, such as HR and RR,evaluated in normal healthyvolunteers for safety monitoringin Phase I clinical trials but maybe applied to any stage of drugdevelopmentBenefitRiskEvidentiaryCriteria Biological rationale is well established for conventionalsafety monitoring Device measurement characteristics, reportable range andreference interval need to be established for continuousremote ambulatory monitoring Retrospective data analysis and ad hoc if needed Confirmation with independent datasets is needed Novel data analytics and statistical approaches are needed4

Problem StatementClinical trials in normal healthy volunteers (NHV)The goal of early-stage clinical trials is to establish a pharmacokinetic, pharmacodynamic and safetyprofile of an investigational drug Early stage clinical trials in multiple therapeutic areas, excluding Oncology, are conducted in NHVo The PK, PD and safety data are collected while study subjects are confined to the clinical pharmacology units(CPU) and after the discharge from the CPU during the follow-up visitso The duration of the confinement varies from one to several weeks depending on the study design,investigational compound properties and anticipated/emerging safety profile Safety data collection is done at predefined time points and includes vital signs (e.g. ECG andlaboratory safety tests) The CPU confinement for extended periods of time is inconvenient for study subjects and may notprovide the data reflective of normal day-to-day person’s activity Little or no safety informationo Other than subject’s memory recall, is available after subject's discharge from the CPU and in-between thefollow-up visits making difficult to interpret potential safety findings5

Statement of Need How is this needed in drug development?o Certain elements of safety profile (HR, RR, body temp) can be built comprehensivelyin early stage clinical trials using wearable devices to collect dense continuous databoth during the CPU confinement and after the discharge from the CPU Collect the clinical trial subject data in the real-world setting, a.k.a. “in the wild”o Early signal detection can inform dose adjustments or discontinuation of drug candidateswith safety liabilities earlier in the drug development process Why take the path of digital measure vs. current modalities?o Data collected at predefined time points – not clear what happens in-between – a signal can bemissed Holter continuous monitoring is available for limited duration, e.g. 24-48 ho No clear how early safety profile is impacted by activities of daily living, e.g. physical exercise,once the study subjects leave the unit6

Flow of data from biology to decisionEach step needs evidence!RawdataBiology(Biomarker orCOA)Raw sensor dataData transfer can occur at any stepRawprocessingConvert toreadoutIn DeviceProcesseddatae.g. Heart rateor Blood AnalysisToAnalysisPackaging; luationFDAQualificationFromDeviceTool use* Stakeholder input should be addressed as appropriate7

Studies Used for Cardiac Monitoring CaseNPopulation1 lead ECG patchlike devicesWrist wornactigraphyDurationStudy #1Study #265NHVNHVHealthPatch by Vital Connect HR, RR, skin temperature Step count ( via an accelerometer)BodyGuardian by Preventice HR, RR Activity counts (via an accelerometer)Actiwatch Spectrum Pro by Philips ( activity counts, physical activity, sleep)10 days confinement period2 confinement periods separated by at homeperiod510(k) clearanceYesYesECG raw *AlgorithmIn-study datareview*Study#1 1111/cts.12602Study#2 1111/cts.12673* Not deployed in the study8

Experimental Design and Key FindingsStudy DesignOperational Execution Exploratory endpoint Data "safe harbor“ - not used for anyclinical decision-making Comparison to conventional safetymeasures using time matching datapoints Assessment of wearable devices asan exploratory objective in aninterventional study is feasibleConventional measurements, e.g."gold standard", need to beconsidered carefullyInclusion of appropriate controls isessential Data AnalysisAll analyses were performed aftercompleting the data collection inall subjectsAnalytical evaluation was carriedout by:Separate optional informedconsent formDevices were administered andmanaged by the site personnelThe sites were trained to assigna device to a specific studysubject and train subjects ondevice management, e.g.battery recharging Study subjects expect to becompensated for additional studyprocedures such as wearabledevicesThe sites emphasized theimportance of having hands ontraining prior to deploying deviceswith the study subjectsThe satisfaction of studyparticipants with variable deviceswas high Ambulatory ECG data can be noisy,data review can be timeconsuming Appropriate data filtering isessential Novel statistical approaches arerequired to facilitate the review ofcontinuous data compared to theanalysis of conventional datacollected at predefined time points Comparison to the correspondingconventional measures for HR and RR Conformity of randomly selected HRvalues (low, medium, high) to the resultsof ECG tracer manual review Face validity of vital sign andactigraphy data was confirmed byexamining aggregate diurnalvariation patterns9

Context of Use (COU)Definition: A statement that fully and clearly describes theway the medical product development tool is to be used andthe medical product development-related purpose of the use Does the COU of the device fulfill the need?o Example 1: Retrospective multimodal analysis of early safety signals is needed Vital sign (HR and RR) analysis suitable Dense continuous data Physical activity (PA) needed as a part of metadata to interpret the resultsCardiac rhythm analysis: more feasible than for real-time, but still requires a lot of manual curationo Example 2: Real-time ad hoc analysis of early safety signals is needed Vital signs (HR and RR) Rhythm analysis (arrhythmias): analytical validation is key Important considerations:o Analytical validation (what the device is meant to measure) and human factor testingo Changes in patient population may require reassessment, e.g. NHV vs. disease population10

Relationship to the existing biomarker evidentiary criteriaframework Existing Measureo Single-lead ECG for remote monitoringo Conventional measurements: 12-lead ECG (resting and supine) Holter monitoring (ambulatory) Manual RR Oral temperature Is the relationship of the remote measure to the clinical outcome known?o Reference ranges – normal/ abnormal are established and apply to both conventionaland remote measurementso Not all features of the conventional ECG are collected as part of safety monitoring inclinical trials (e.g. QT or PR interval prolongation) are available from a single lead ECG11

Alignment with biomarker evidentiary criteria framework What fits?o Variables HR RR What doesn’t fit?o Conventional value reference ranges and reference intervalo Interpretation of continuous data under ambulatory conditionso Physical activity by means of actigraphy does not have conventional counterparts forpurposes of safety monitoringo Skin temperature is highly variable, can be impacted by a number of factors difficult orimpossible to control (ambient temperature, clothing, body movements) and difficult tointerpretStudy subjects were asked to complete the technology satisfactionquestionnaire. The response indicated high acceptance of technology12

Benefit Assessment By qualifying vital sign measurements using single-lead ECG and wrist wornactigraphy devices:ooooEarlier detection of a potential safety signalDose adjustmentEarly discontinuation of drug candidates with unfavorable safety profileA reliable pharmacodynamic assessment if related to drug MOA When in the drug development lifecycle is the measure intended to be used?o Predominantly in Phase I clinical trials, but can be deployed at any stage as needed Is the benefit of the measure to the individual or society?o More effective drug developmento Easier participation in clinical trialso More feedback from the study subjects how they are doing13

Risk Assessment Device performance:o What is the potential consequence or harm if the measure’s performance isnot aligned with expectations based on the COU? False negative - missing a potential safety signal leading to a misuse of certaintherapeutics False positive - time consuming data review and reporting Potential signal needs to be verified by a trained professionalo Data losses due to subjects not wearing devices when unsupervised, connectivityissue etc.o Analytical validity – comparison with the raw/ source data to establish accuracybeyond the comparison to conventional measurements Reference ranges and interval are needed for ambulatory conditions14

Risk Assessment Regulatory:o COU may be different than intended use and indications of use stipulatedunder 510(k) clearanceo Validation should be performed according to the COU pertinent to a specificclinical trial Data analysis:o Correlation and limits of agreement with conventional measurements Limitations: conventional measurements are done at predefined time orcontinuous monitoring for limited time, e.g. Holter monitoring for 24-48 h – fullscale comparison is not feasibleo Data filtering – what is noise and what is a real signal? Striking the right balance – more analytical work is required using larger data sets15

Risk Assessment – mitigation strategy Device performance and regulatory:o Establishing device performance characteristics according to the COU prior to collecting data Access to the raw data is a must – required for both analytical validation and a potential safety signalconfirmation during clinical study results reviewo Establishing a minimal threshold of subjects’ adherence to contributing the data Prior human factor testing may be requiredo Establishing reference ranges for ambulatory conditions for variables of interest Physical activity data is essential for result interpretation Data analysis:o Establish a process for data review and reportingo Retrospective analysis at a predefined timeo Novel analytical approaches are required for data analysis Filtering out noiseDefine acceptable false-negative and false-positive rate16

State of EvidenceEquivalence with conventional measurementsCorrelation and limits ofagreement Comparison to conventionaltechnologies - correlation andlimits of agreement– conventional 5 minutes restingand supine protocolDevice to device datacomparison is appropriate, acomparison to the manual datacollection method is om/doi/full/10.1111/cts.1267317

Face validity of the data – aggregate levelAggregate data shows clear and consistent diurnal variation full/10.1111/cts.1260218

Approach to analytical validationReasonableness andphysiological validity of thedata needs to be evaluated atthe individual and trial Gap countLongest s.1267319

Approach to analytical validationNon-physiological or any otherdata representing a potentialsafety signal requires a follow-upGap countTotal gaptime(hours)58001 0003270.658001 0007581.558001 0011161.558001 001241728.358001 0015118.358001 doi/full/10.1111/cts.1260220