Thecollectionofspontaneouscasereportstoevaluatepotentialadverseevents(AEs)inducedbydrugsorvaccinesisparamountforpharmacovigilance,allowingthetimelyidentificationofsafetyissuesandenablingappropriateactionincludingchangesinprescribinginformation.
Weassessedpatient-generatedsafetydatafromonlinehealthcareforainthreeareasofpotentialimportancetopharmacovigilanceactivities:
1)changesintherelativefrequencyofAEdiscussionsinsupportoftraditionalsignaldetection
2)utilityoflongitudinalAErecordsobtainedbylinkingpostsfromthesameindividual
3)abilitytoactivelyengagepatientstoobtainmoreinformationaboutAEs.
Foreachofthethreeexperimentscorrespondingtothestudyobjectives,differentdatasetsandmethodologywereused,asdescribedbelow.
In-scopeusecases(drugandAEpairs)werefirstidentifiedasfollows:
Identificationofproduct:toidentifycontemporaryexamples,weconsideredimmunemodulatorsthathavebeenapprovedintheUnitedStates(US)since2010.
Inclusioncriteria:tobepotentiallyincludedinthisassessment,eachAEhadtobemedicallyserious,hadnottohaveappearedintheinitialUSPIatapproval,hadtobeidentifiedpost-approvalwithatleasta3-yearsgapbetweenapprovalandtheeventbeingaddedtotheUSPI,andhadtoreflectamedicalconceptthatcouldbearticulatedbythepatient/caregiverinsocialmedia.
Fromthelist,fivepotentialusecaseswereidentified.Subsequentfeasibilityassessmentsdeterminedminimaldiscussionsforthreeofthepotentialusecaseswhichnarrowedthelisttothefollowingtwousecases:denosumabandmultiplevertebralfracturesfollowingdiscontinuationandpembrolizumabandimmune-mediatedskinadversereactions(whichincludesStevens–Johnsonsyndrome,toxicepidermalnecrolysis,exfoliativedermatitis,andbullouspemphigoid).
FIGURE1.Dataprocessing.CSV,CommaSeparatedValue;ML,machinelearning;AI,artificialintelligence;NLP,naturallanguageprocessing;ADR,adversedrugreactions.
Duplicateposts(samecontent,poster,andpostdate/time)wereremoved,andaspamremovalalgorithmwasruntolookforkeywordssuchas“onlinestore”orpressreleases(toexcludeanynon-patientgenerateddata).Medicallyrelevantposts(withatleastonementionoftheproductbygenericname,brandnameordrugclass)wereidentifiedbyusingkeywordfilters,andallpostsweremergedintoaunifiedstructuresuitableforanalysis(includingmeta-datasuchasauniquepostidentificationnumber,date,source,contentofpostandtaggedAEs).
Allpostsmentioningthedrugsofinterestwerecollected.Countsforthefollowingwerecomputed:1)totalnumberofpostsmentioningthedrugofinterestand2)thetotalnumberofpoststhatmentionthedrugofinteresttogetherwithanyoftheAEsofinterest(AEOI)and3)theuniquenumberofpoststhatmentionadrugandaspecificAEOIwerecollected.Fromthesefigures,wecomputedthefrequencyofeachAEOIbydividingthenumberofeventsbythetotalnumberofpoststhatmentionedthedrugofinterest.
Foreachgroupofin-scopelinkedpoststhefollowingdataelementsweremanuallyreviewedandannotatedbyahealthcareprofessional:medicalhistory,diseaseburden,useofnon-medicaltreatments,laboratoryresults,treatmenthistory,concomitantmedications,andmentionofAEs.IfthepostmentionedanAE,otherdataelementswerealsoannotated:timetoonset,outcome,treatmentofAE,andstatementofcausality.Thepercentagesofgroupedpostscontainingeachdataelementwerecalculated.
Thefollowingstepswereusedtoidentifyin-scopepostsforfollow-up:
1.UsingInspiredata,weidentifiedallpoststhatmentioneddenosumab(Xgeva,Prolia)upto20April2021.
2.Postswerede-identifiedbyInspirebeforesendingthemtothestudysponsoranduniqueidentifierswereassignedforeachpersonmakingapost.
4.AnexperiencedsafetyscientistmanuallyreviewedthefulllistofpotentialAEs,bydescendingAEprobabilityscoreandwithapostdateof1January2020orlater,toconfirmatrueAEandtemporalrelationshiptodenosumab.FalseAEs,orthosenottemporallyrelatedtodenosumab,wereremovedfromthelist.ThirtyAEswereselectedbasedonthefollowing,startingwiththelistoftrue/temporalAEs:
i.non-serious/non-severeAEswereremoved;
ii.ifnofollow-upinformationwasdeemedimportantnoressentialforassessmentbythereviewer,thentheAEwasremoved;
iii.oftheremainingpostsarandomsamplewaschosen(however,alimitoftwosimilarAEswasenforced).Ifthesamplewastoosmall,inclusioncriteriawerewidened;
iv.onlyoneAEperuniqueidentifierwaschosen.
5.BasedonthesampleofAEschosen,eachpostwasreviewedindependentlybytwoexperiencedsafetyscientistsandrequisitedataelementsrequiredforproperassessment(definedascontainingtheminimuminformationtoassessapotentialcausalassociation,whichmayvarybyAE)wasevaluatedandmissingdataelementsdocumented.Resultswerecomparedbetweenreviewersandconcordanceofrelevantmissingdataelementswasachieved.Thefollow-upattemptsoughtonlythemissingdataelements.
6.Requestforfollow-upwassentforthefirst15AEsonthelist:
a.Ifapersondeclinedtosubmitfollow-up,whichwasdocumented,adifferentAEwaschosen.
b.Forthoseindicatingawillingnesstosubmitfollow-upinformation:
i.after1month,iffollow-upwasnotreceived,additionalAEswerechosenforfollow-upuntil15completedrequestswereobtained;
ii.iffollow-upwascompleted,theindividualreceivedanhonorariumof$100.00fromInspire.
7.Intheevent15follow-upswerenotreceivedasoutlinedabove,thenpostsfortwootherdrugs(teriparatide[Forteo]oradalimumab[Humira])werecollectedbyInspireusingthesameinclusionarydates(1January2020orlater)andsenttothestudysponsor,andstepsabovewererepeated.
8.Afterfollow-upwasreceived,thereportswerereassessedbyeachsafetyscientists,andthenewassessmentswererecorded,aswellasanycommentsabouttherelevancyoftheindividualdataelementsprovided,whetherthefollow-upinformationreceivednowallowedforproperassessment,andtheneedforfurtherinformationtoproperlyassesstheAE.Concordanceacrossthetwoassessorswasevaluated.
Oncetheprocessabovewasfinalized,thestudysponsorandInspiredraftedaninformedconsentformthatwastobesharedwitheachparticipantpriortoinitiatingtheengagement.TheWIRBCopernicusGroup,Inc.(WCG)IRB’sInstitutionalReviewBoardAffairsDepartmentreviewedtheinformedconsentandprocessoutlinedaboveundertheCommonRuleandapplicableguidanceanddetermineditexemptunder45CodeofFederalRegulation§46.104(d)(2),becausetheresearchonlyincludedinteractionsinvolvingeducationaltests,surveyprocedures,interviewprocedures,orobservationsofpublicbehaviorandtherewereadequateprovisionstoprotecttheprivacyofparticipantsandtomaintaintheconfidentialityofdata.
Wecalculatedthepercentageoffollow-uprequestssentandreceived,timefromrequesttoreceiptoffollow-up,thepercentageofcaseswithatleastone(important)pieceofinformationobtaineduponfollow-up(meaningfulfollow-up),caseswhichcouldbeproperlyassessedafterfollow-up,andthepercentageofcaseswhereallfollow-upinformationrequestedwasobtained.
TABLE1.NumberandpercentageofpostsmentioningthedrugsofinterestandatleastoneadverseeventmappedtoaMedDRApreferredterm,bysocialmediaforumandoverall.
TABLE2.Adverseevents(byMedDRApreferredterm)reportedwithfrequency≥2%inpostsmentioningthedrugsofinterest,betweenJanuary2010andDecember2021,bysocialmediaforumandoverall.
ThefollowingMedDRAPTswereusedasAEOIcervicalvertebralfracture,lumbarvertebralfracture,thoracicvertebralfracture,andspinalfracturefordenosumabandtoxicepidermalnecrolysis,Stevens-Johnsonsyndrome,erythemamultiforme,dermatitisexfoliative,andpemphigoidforpembrolizumab.
TABLE3.Adverseeventsofinterest(byMedDRApreferredterm)identifiedinpostsmentioningdenosumab.
TABLE4.Adverseeventsofinterest(byMedDRApreferredterm)identifiedinpostsmentioningpembrolizumab.
FIGURE2.Adverseeventsofspecificinterestfordenosumab(A)andpembrolizumab(B)detectedinFAERS.FAERS,FoodandDrugAdministrationAdverseEventReportingSystem;EBGM,empiricalBayesgeometricmean.
FIGURE3.Percentageofsocialmediapoststhatmentiondenosumabandspinalfractureasanadverseevent(A)andEBGMscoresfordenosumabandspinalfracturesfromFAERSdata(B).FAERS,FoodandDrugAdministrationAdverseEventReportingSystem;Q,quarter;EBGM,empiricalBayesgeometricmean.
FIGURE4.Percentageofsocialmediapoststhatmentiondenosumabandthoracicvertebralfractureasanadverseevent(A)andEBGMscoresfordenosumabandthoracicvertebralfracturefromFAERSdata(B)FAERS,FoodandDrugAdministrationAdverseEventReportingSystem;Q,quarter;EBGM,empiricalBayesgeometricmean.
FIGURE5.PercentageofsocialmediapoststhatmentionpembrolizumabandStevens-Johnsonsyndromeasanadverseevent(A)andEBGMscoresforpembrolizumabandStevens-JohnsonsyndromefromFAERSdata(B)FAERS,FoodandDrugAdministrationAdverseEventReportingSystem;Q,quarter;EBGM,empiricalBayesgeometricmean.
Atotalof1,672postswereevaluated,whichaveragedto104postsperauthor(median34,range11–534).Relevantmedicalinsightsmentionedbytheauthorsinthe16postswiththehighestcomplexityscoresincludedmedicalhistory(16/16,100%),diseaseburden(15/16,94%),useofnon-medicaltreatments(15/16,94%),laboratoryresults(14/16,88%),treatmenthistory(13/16,81%),AEs(13/16,81%),andconcomitantmedications(12/16,75%).
AmongthepoststhatcontainedAEinformation,additionaldetailswereprovidedsuchas:outcomeoftheevent(13/16,81%),timetoonset(11/16,69%),treatmentoftheevent(10/16,63%),mentionofcausality(10/16,63%),doseofthemedications(9/16,56%),andmentionofrechallenge/de-challenge(3/16,19%).
FIGURE6.Relevantmedicalandadverseeventinsightsfromspontaneouslyreportedadverseeventsinsocialmedia,postedbyasingleindividual.Note:Postsarelistedverbatimbutmentionsofdrugnameshavebeenredacted.
AmongasubsetofInspireusers,39requestsweresenttofollow-uponpostsmentioninganAEinrelationtooneofthedrugsofinterest.Eight(8/39,21%)oftheserequestswereansweredwithaconsentforfurtherfollow-up,forwhichsix(6/8,75%)patientsprovidedtheinformationrequested.Themediantimeforreceiptoffollow-upfromthesepatientswas2.5days(range0–27days)fromthedateoftheinitialcontactseekingconsentand0.5days(range0–4days)fromthedatethefollow-upquestionsweresent.
Dependingontheinitialinformationreported,therequestforfollow-upcontainedbetweenfourand11questions.Mostofthequestionsfocusedonobtainingfurtherinformationregardingthediagnosis,duration,andriskfactorsfortheAE,historyofsimilareventspriortotreatment,timetoonset,concomitantmedicationandrelevantmedicalhistory.Allquestionsaskedduringthefollow-upwereansweredbytheinitialposters(37/37,100%).Afterreviewingthefollow-up,theAEwasassessablefor4/6(67%)andnon-assessablefor2/6(33%)ofthereports.Allpatients(6/6,100%)werewillingtosharetheirHCP’scontactinformationforadditionalfollow-up.
FIGURE7.Examplesofanassessable(A)andnon-assessable(B)caseafterfollow-upinformationwasrequestedandprovidedbythepatient.Note:Thedrugofinterestwasdenosumab.Postsarelistedverbatimbutmentionstotradenamesofdrugsorsensitivepersonalinformationhavebeenredacted.
Allauthorsparticipatedinthedesignoftheexperiments,analysis,andinterpretationofthedata.Allauthorsagreetobeaccountableforthecontentofthework.
GlaxoSmithKlineBiologicalsSAcoveredallcostsassociatedwiththeconductofthestudyandthedevelopmentofthemanuscript.
ThestudywasfundedbyGlaxoSmithKlineBiologicalsSA.Thefunderwasinvolvedinstudydesign,datacollectionandanalysis,decisiontopublishandpreparationofthemanuscript.GP,VK,JP,LS,andABareemployeesoftheGSKgroupofcompaniesandholdsharesintheGSKgroupofcompanies.
Theremainingauthordeclaresthattheresearchwasconductedintheabsenceofanycommercialorfinancialrelationshipsthatcouldbeconstruedasapotentialconflictofinterest.
Allclaimsexpressedinthisarticlearesolelythoseoftheauthorsanddonotnecessarilyrepresentthoseoftheiraffiliatedorganizations,orthoseofthepublisher,theeditorsandthereviewers.Anyproductthatmaybeevaluatedinthisarticle,orclaimthatmaybemadebyitsmanufacturer,isnotguaranteedorendorsedbythepublisher.
TheauthorsthanktheModisplatform,onbehalfofGSK:PetronelaM.PetrarprovidedwritingsupportandCarlosMarinPcoordinatedmanuscriptdevelopment.
Anderson,J.T.,Bouchacourt,L.M.,Sussman,K.L.,Bright,L.F.,andWilcox,G.B.(2022).TelehealthAdoptionduringtheCOVID-19Pandemic:ASocialMediaTextualandNetworkAnalysis.Digit.Health8,20552076221090041.doi:10.1177/20552076221090041
Arnaud,M.,Bégaud,B.,Thurin,N.,Moore,N.,Pariente,A.,andSalvo,F.(2017).MethodsforSafetySignalDetectioninHealthcareDatabases:aLiteratureReview.ExpertOpin.DrugSaf.16(6),721–732.doi:10.1080/14740338.2017.1325463
Bate,A.,andStegmann,J.U.(2021).SafetyofMedicinesandVaccines-BuildingNextGenerationCapability.TrendsPharmacol.Sci.42(12),1051–1063.doi:10.1016/j.tips.2021.09.007
Bulcock,A.,Hassan,L.,Giles,S.,Sanders,C.,Nenadic,G.,Campbell,S.,etal.(2021).PublicPerspectivesofUsingSocialMediaDatatoImproveAdverseDrugReactionReporting:AMixed-MethodsStudy.DrugSaf.44(5),553–564.doi:10.1007/s40264-021-01042-6
Chen,X.,Faviez,C.,Schuck,S.,Lillo-Le-Lout,A.,Texier,N.,Dahamna,B.,etal.(2018).MiningPatients'NarrativesinSocialMediaforPharmacovigilance:AdverseEffectsandMisuseofMethylphenidate.Front.Pharmacol.9,541.doi:10.3389/fphar.2018.00541
Correia,R.B.,Wood,I.B.,Bollen,J.,andRocha,L.M.(2020).MiningSocialMediaDataforBiomedicalSignalsandHealth-RelatedBehavior.Annu.Rev.Biomed.DataSci.3(1),433–458.doi:10.1146/annurev-biodatasci-030320-040844
DuMouchel,W.,andPregibon,D.(2001).EmpiricalBayesScreeningforMulti-ItemAssociationsinProceedingsoftheseventhACMSIGKDDinternationalconferenceonKnowledgediscoveryanddatamining.SanFrancisco,California,.doi:10.1145/502512.502526
Freifeld,C.C.,Brownstein,J.S.,Menone,C.M.,Bao,W.,Filice,R.,Kass-Hout,T.,etal.(2014).DigitalDrugSafetySurveillance:MonitoringPharmaceuticalProductsinTwitter.DrugSaf.37(5),343–350.doi:10.1007/s40264-014-0155-x
Gartland,A.,Bate,A.,Painter,J.L.,Casperson,T.A.,andPowell,G.E.(2021).DevelopingCrowdsourcedTrainingDataSetsforPharmacovigilanceIntelligentAutomation.DrugSaf.44(3),373–382.doi:10.1007/s40264-020-01028-w
Gendreau,J.,Ramsubeik,S.,andPitesky,M.(2022).WebCrawlingofSocialMediaandRelatedWebPlatformstoAnalyzeBackyardPoultryOwnersResponsestothe2018–2020NewcastleDisease(ND)OutbreakinSouthernCalifornia.TransboundingEmerg.Dis.[Epubaheadofprint].doi:10.1111/tbed.14454
Golder,S.,Smith,K.,O'Connor,K.,Gross,R.,Hennessy,S.,andGonzalez-Hernandez,G.(2021).AComparativeViewofReportedAdverseEffectsofStatinsinSocialMedia,RegulatoryData,DrugInformationDatabasesandSystematicReviews.DrugSaf.44(2),167–179.doi:10.1007/s40264-020-00998-1
Inácio,P.,Cavaco,A.,andAiraksinen,M.(2017).TheValueofPatientReportingtothePharmacovigilanceSystem:aSystematicReview.Br.J.Clin.Pharmacol.83(2),227–246.doi:10.1111/bcp.13098
Inch,J.,Watson,M.C.,andAnakwe-Umeh,S.(2012).PatientversusHealthcareProfessionalSpontaneousAdverseDrugReactionReporting:aSystematicReview.DrugSaf.35(10),807–818.doi:10.1007/bf03261977
Jokinen,J.D.,Walley,R.J.,Colopy,M.W.,Hilzinger,T.S.,andVerdru,P.(2019).PoolingDifferentSafetyDataSources:ImpactofCombiningSolicitedandSpontaneousReportsonSignalDetectioninPharmacovigilance.DrugSaf.42(10),1191–1198.doi:10.1007/s40264-019-00843-0
Kara,V.,Powell,G.,Merico,E.,Kaur,N.,andBate,A.(2021).20thISoPAnnualMeeting“IntegratedPharmacovigilanceforSaferPatients”8-10November2021Muscat,Oman(HybridMeeting).DrugSaf.44(12),1391–1470.doi:10.1007/s40264-021-01129-0
Lee,J.Y.,Lee,Y.S.,Kim,D.H.,Lee,H.S.,Yang,B.R.,andKim,M.G.(2021).TheUseofSocialMediainDetectingDrugSafety-RelatedNewBlackBoxWarnings,LabelingChanges,orWithdrawals:ScopingReview.JMIRPublicHealthSurveill7(6),e30137.doi:10.2196/30137
Li,Y.,JimenoYepes,A.,andXiao,C.(2020).CombiningSocialMediaandFDAAdverseEventReportingSystemtoDetectAdverseDrugReactions.DrugSaf.43(9),893–903.doi:10.1007/s40264-020-00943-2
Makady,A.,Kalf,R.R.J.,Ryll,B.,Spurrier,G.,deBoer,A.,Hillege,H.,etal.(2018).SocialMediaasaToolforAssessingPatientPerspectivesonQualityofLifeinMetastaticMelanoma:aFeasibilityStudy.HealthQual.LifeOutcomes16(1),222.doi:10.1186/s12955-018-1047-z
Nguyen,T.,Larsen,M.E.,O'Dea,B.,Phung,D.,Venkatesh,S.,andChristensen,H.(2017).EstimationofthePrevalenceofAdverseDrugReactionsfromSocialMedia.Int.J.Med.Inf.102,130–137.doi:10.1016/j.ijmedinf.2017.03.013
Platt,R.,Brown,J.S.,Robb,M.,McClellan,M.,Ball,R.,Nguyen,M.D.,etal.(2018).TheFDASentinelInitiative-anEvolvingNationalResource.N.Engl.J.Med.379(22),2091–2093.doi:10.1056/NEJMp1809643
Rakhsha,M.,Keyvanpour,M.R.,andVahabShojaedini,S.(2021).“DetectingAdverseDrugReactionsfromSocialMediaBasedonMultichannelConvolutionalNeuralNetworksModifiedbySupportVectorMachine”,inProceedingofthe7thInternationalConferenceonWebResearchICWR,TEHRAN,Iran,.doi:10.1109/icwr51868.2021.9443128
Sarker,A.,Ginn,R.,Nikfarjam,A.,O'Connor,K.,Smith,K.,Jayaraman,S.,etal.(2015).UtilizingSocialMediaDataforPharmacovigilance:AReview.J.Biomed.Inf.54,202–212.doi:10.1016/j.jbi.2015.02.004
Schachterle,S.E.,Tresnan,D.,Hauben,M.,Murugesan,S.,Sobel,R.E.,Bate,A.,etal.(2019).Abstractsofthe35thInternationalConferenceonPharmacoepidemiology&TherapeuticRiskManagement,PennsylvaniaConventionCenter,Philadelphia,PA,USA,August24‐28,2019.PharmacoepidemiolDrugSaf.28,980Socialmediadataanalysiswithpopulation-basedmethods:Apilotinvestigationofmildhypersensitivityamongusersofsubcutaneouslyadministeredmonoclonalantibodies,5-586.doi:10.1002/pds.4864
Stergiopoulos,S.,Fehrle,M.,Caubel,P.,Tan,L.,andJebson,L.(2019).AdverseDrugReactionCaseSafetyPracticesinLargeBiopharmaceuticalOrganizationsfrom2007to2017:AnIndustrySurvey.Pharm.Med.33(6),499–510.doi:10.1007/s40290-019-00307-x
Thomas,M.,Curry,A.,Painter,J.L.,Akhtar,A.,Schifano,L.,andPowell,G.E.(2017).“T-26:CaseStudy:ComputingComplexityScorestoIdentifyPatientsofInterestfromInspire.ComForumsforSafetyandBeyond”,inDrugInformationAssociationAnnualMeeting,Chicago,IL,.
Trigger,S.,andColeman,B.(2019).SocialMediaMentionsofElectronicNicotineDeliverySystems(ENDS)Battery-RelatedOverheating,Fires,andExplosions:FindingsfromaPilotStudy.Int.J.Environ.Res.PublicHealth16(8),1308.doi:10.3390/ijerph16081308
vanStekelenborg,J.,Ellenius,J.,Maskell,S.,Bergvall,T.,Caster,O.,Dasgupta,N.,etal.(2019).RecommendationsfortheUseofSocialMediainPharmacovigilance:LessonsfromIMIWEB-RADR.DrugSaf.42(12),1393–1407.doi:10.1007/s40264-019-00858-7
Wang,R.,Liu,W.,andGao,S.(2016).HashtagsandInformationViralityinNetworkedSocialMovement.OnlineInf.Rev.40(7),850–866.doi:10.1108/OIR-12-2015-0378