characterisation of databases (db(bs) used for signal ...€¦ · bhlthbsp2 ph ti l it ti la ic:...
TRANSCRIPT
Characterisation of databases (DBCharacterisation of databases (DB(Results of a survey of IMI PROTECTResults of a survey of IMI PROTECTy
Antoni FZ Wisniewski1 Kristina Juhlin2 Mona Vestergaard Laursen3 Miguel MAntoni FZ Wisniewski , Kristina Juhlin , Mona Vestergaard Laursen , Miguel M 1AstraZeneca, Alderley Park, UK; 2The Uppsala Monitoring Centre, Uppsala, Sweden; 3The Danish Health and Mediciny pp g pp
Health Care Pharmaceuticals Berlin Germany; 6GlaxoSmithKline London UK; 7Medicines and Healthcare products RHealth Care Pharmaceuticals, Berlin, Germany; 6GlaxoSmithKline, London, UK; 7Medicines and Healthcare products R
⎯ ♦ ⎯ DISCLOSURE STATEMENT: SEE BOTTOM OF POSTER ⎯ ♦ ⎯ ACKNOWLEDGEMENT: ALEX ASIIM
B k dBackgroundg-
The work was conducted as part of the PROTECT consortium Surveye o as co ducted as pa t o t e O C co so t u(Pharmacoepidemiological Research on Outcomes of
SurveyS ti(Pharmacoepidemiological Research on Outcomes of
Th i b E C Ti i i )Survey question
Therapeutics by a European ConsorTium; www.imi-protect.eu) survey was dividp y p p )which is a public-private partnership coordinated by the
survey was dividwhich is a public private partnership coordinated by the European Medicines Agency (EMA)European Medicines Agency (EMA). i. general inforg
signal and dWork-package 3 includes assessment of signal detection
signal and dii t i l dWork-package 3 includes assessment of signal detection
th d li d t d t b f t d dii. counts includ
methods applied to databases of spontaneous adverse drug iii. data elemenreactions. PROTECT Partners surveyed include 3 national
iii. data elemeniv database coreactions. PROTECT Partners surveyed include 3 national
regulatory agencies (AEMPS DKMA MHRA) 3 pharmaceuticaliv. database co
regulatory agencies (AEMPS, DKMA, MHRA), 3 pharmaceutical companies (AZ, BSP, GSK), the EMA and the Uppsala Data collection ap ( ) ppMonitoring Centre (UMC)
Data collection aThe survey wasMonitoring Centre (UMC) The survey was communicationThe software allThe software all
d i i t tadministrator wacompleted the s
Objectivescompleted the s
Objectivesj-
T d ib diff t t t d t b ithResponses were
To describe different spontaneous report databases with p
created within Eregard to size and content as context for future assessment of
created within Et d b lregard to size and content as context for future assessment of
the relative performance of SD and duplicate detectionpresented below
the relative performance of SD and duplicate detection and may not refmethods in these databases
and may not ref
T bl 1 PROTECT P t th t ti i t d i th D t b SRe
Table 1 – PROTECT Partners that participated in the Database SurveyRePartner Abbrevn. Affiliation Setting Database
nameD i h H lth & DMKA1 N ti l C t t N ti l /Danish Health & Medicines Agency
DMKA1 National Competent Authority
National n/a Table 2 – Signal algorithms & definitions of statistics of dispMedicines Agency AuthorityUppsala Monitoring UMC International Drug International VigiBase
g greporting employed across survey participants’ databases Uppsala Monitoring
CentreUMC International Drug
Monitoring International VigiBase
Partner Algorithm How is a statistic of disproportionatorganisation
E M di i EMA E I t ti l E d i il
g p pdefined?IC025 0European Medicines
AgencyEMA European
Competent AuthorityInternational Eudravigilance UMC IC IC025 >0
Agency Competent AuthorityMedicines and MHRA National Competent National Sentinel EMA PRR N >2 and lower end of 95%CI on PRR Medicines and Healthcare products
MHRA National Competent Authority (UK)
National SentinelMHRA EBGM EBGM ≥2.5, EB05 ≥1.8, N ≥3Regulatory Agency
y ( )
Agencia Espaniola de AEMPS National Competent National FEDRA
MHRA EBGM EBGM ≥2.5, EB05 ≥1.8, N ≥3
AEMPS IC ROR ROR L li it 95% CI 1 d NAgencia Espaniola de Medicamentos y
AEMPS National Competent Authority (Spain)
National FEDRA AEMPS IC, ROR ROR: Lower limit 95% CI >1 and No. oIC: Lower limit 95% CI >0 and No. of ICy
Productos SanitariosAuthority (Spain)
B H lth BSP2 Ph ti l I t ti l A
IC: Lower limit 95% CI 0 and No. of IC
BSP PRR PRR ≥2 Chi² ≥4 N ≥3Bayer Healthcare Pharmaceuticals
BSP2 Pharmaceutical company
International Argus BSP PRR PRR ≥2, Chi ≥4, N ≥3
AZ EBGM EB05 ≥1 8 and/or +ve trend flag A trePharmaceuticals companyAstraZeneca AZ Pharmaceutical International Sapphire
AZ EBGM EB05 ≥1.8, and/or +ve trend flag. A treeither of the following are true: AstraZeneca AZ Pharmaceutical
companyInternational Sapphire
• An EB05 based on current data is d e pair 52 weeks agoGlaxoSmithKline GSK Pharmaceutical International Oceans d-e pair 52 weeks ago
• An 50% increase in EBGM score wcompany data are compared with the EBGM
k1 Since completing the survey The Danish Medicines Agency and Danish National Board of Health weeks agoGSK EBGM EB05 >2 for non-serious unlisted adve
1. Since completing the survey, The Danish Medicines Agency and Danish National Board of Health merged under the name Danish Health and Medicines Authority. Throughout this poster, the former GSK EBGM EB05 >2 for non-serious unlisted adve
event whose reporting rate has increasg y g p ,
abbreviation is retained.2 Si l ti th B S h i Ph AG (BSP) d B Ph compared to 6 months previously2. Since completing the survey, Bayer Schering Pharma AG (BSP) was renamed as Bayer Pharma
AG Throughout this poster the former abbreviation is retainedAG. Throughout this poster, the former abbreviation is retained.
Figure 3 SPONTANEOUS repoFigure 3 – SPONTANEOUS repo
Figure 2 –Number of SPONTANEOUS reports broken down by Seriousness UMC
GSKUMC
AZ*
EMA
BSPBSP
GSKAEMPS
GSKAEMPS
MHRA*BSP
MHRA*
MHRA DKMAMHRA
0% 10% 20% 30% 40%0% 10% 20% 30% 40%
P fAZ
Percentage of spontaneous rep*counts of AE report for AZ and MHRA are not bacounts of AE report for AZ and MHRA are not ba
Physician Pharmacist Other health professioAEMPS Physician Pharmacist Other health professio
DKMATable 3 – Presence of demographic data in partner d
DKMA Millions of Spontaneous ReportsTable 3 Presence of demographic data in partner d
R i t D t A /D B G0 1 2 3 4 5 6 Receipt Date Age/DoB Gen0 1 2 3 4 5 6
DKMA AEMPS AZ MHRA BSP GSK EMA UMCDKMA (unk) (10
DKMA AEMPS AZ MHRA BSP GSK EMA UMCTotal Serious 18,737 38,758 112,577 381,317 157,000 361,606 1,564,981 1,058,090
UMC (77%) (9Total Serious 18,737 38,758 112,577 381,317 157,000 361,606 1,564,981 1,058,090
SPONTANEOUS50 398 104 951 359 045 251 431 490 000 1 061 856 396 127 752 842 ( ) (
EMA (unk) (uNon-serious50,398 104,951 359,045 251,431 490,000 1,061,856 396,127 752,842
( ) (MHRA (80%) (9
SPONTANEOUSUnknown Seriousness
0 0 5,395 0 0 5,411 0 3,580,534 (80%) (9AEMPS (96%) (9
Unknown Seriousness
(96%) (9BSP (74%) (9 (74%) (9AZ (73%) (9 (73%) (9GSK (79%) (8 (79%) (8
1 - not recorded; value present in ≤ 5% of reports1. - not recorded; value present in ≤ 5% of reports
Fi T SOC E dFi T SOC E dFigure 5 - Top 5 SOCs – Expressed asFigure 5 - Top 5 SOCs – Expressed as
Summary & Conclusioy
• Data from the 8 respondents were obtained betw• Data from the 8 respondents were obtained betwAugust 2011 Database size varies greatly (rangAugust 2011. Database size varies greatly (rangspontaneous reports)spontaneous reports)
• Databases are comparable in terms of: main cou• Databases are comparable in terms of: main coupredominant body systems to which events are cpredominant body systems to which events are ccompleteness of certain data elements such as gcompleteness of certain data elements such as gcasecase
• Pharmaceutical company databases have highePharmaceutical company databases have higheserious adverse events and consumer reports thserious adverse events and consumer reports th
• There is little comparability in SD algorithms empThere is little comparability in SD algorithms emp(e g flags for targeted designated or important m(e.g. flags for targeted, designated or important mPredominant drugs and drug-event pairs vary anPredominant drugs and drug event pairs vary anparochial issues Annual report numbers continuparochial issues. Annual report numbers continucompany, EMA and UMC databases. The patterncompany, EMA and UMC databases. The patternin pharmaceutical company databases, has chanin pharmaceutical company databases, has chan
• The heterogeneity of spontaneous databasesg y pimportant consideration when assessing the p gdetection algorithms in future studies to be cagPROTECT WP 3
Disclosure: The PROTECT project has received support from the Innovp j ppundertaking (www.imi.europa.eu) under Grant Agreement n° 115004, resfi i l t ib ti f th E U i ' S th F k Pfinancial contribution from the European Union's Seventh Framework Procompanies' in kind contribution Authors A Wisniewski K Manlik and V Pcompanies in kind contribution. Authors A.Wisniewski, K Manlik and V PFederation of Pharmaceutical Industries and Association) member comptheir part in the research were carried by the respective company as in-kschemescheme.
Bs) used for signal detection (SD):Bs) used for signal detection (SD):) g ( )T work package (WP) 3 participantsT work package (WP) 3 participantsp g ( ) p p
Macia4 Katrin Manlik5 Vlasta K Pinkston6 Suzie Seabroke7 and Jim Slattery8Macia , Katrin Manlik , Vlasta K Pinkston , Suzie Seabroke and Jim Slatterynes Authority, København, Denmark; 4Agencia Española de Medicamentos y Productos Sanitarios, Madrid, Spain; 5Bayer y g p y p y
Regulatory Agency London UK and 8European Medicines Agency London UKRegulatory Agency, London, UK and 8European Medicines Agency, London, UK
MWE, previously of AstraZeneca1,for help with electronic survey design, administration and data management ⎯ ♦ ⎯p y p y g g
MethodsMethods
d l d ll b ti l t WP3 ti i t d i 2Q2010 Thns were developed collaboratively amongst WP3 participants during 2Q2010. The ded into five domains as follows:ded into five domains as follows:
rmation including types of therapeutic agent, coding dictionaries, use of metadata and g yp p g , g ,uplicate detection algorithmsuplicate detection algorithmsdi i t t t f i i th ti t d tding seriousness, reporter type, country of origin, therapeutic agents and eventsts including presence of demographic data and drug detailsts including presence of demographic data and drug detailsverage in terms of predominant drugs and events v vaccine specific informationverage in terms of predominant drugs and events v. vaccine specific information
and compilation of responsesand compilation of responsesadministered electronically using commercially available software (2ask amundisadministered electronically using commercially available software (2ask, amundiss Gmbh, Felix-Wankel Strasse 4, D-78467 Konstanz, Deutschland; http://www.2ask.net). , , , ; p )ows for on line completion of surveys with password protected access The surveyows for on-line completion of surveys with password protected access. The survey
WP3 ti i t d h d l bilit t d l d d t P ti i tas a WP3 participant and had sole ability to download survey data. Participants survey between September 2010 and August 2011survey between September 2010 and August 2011.
e downloaded by the survey administrator into Microsoft Excel format. Tables were y yExcel for the purpose of summarising the responses and producing figures The resultsExcel for the purpose of summarising the responses and producing figures. The results
t th t t d f th d t b t th ti th i d tw represent the status and use of the databases at the time the survey was carried out lect the status and use of the databases as it is today.lect the status and use of the databases as it is today.
Fi 1 S hi ill t ti th l f t f th 8 d t b th t ti i t d i thesults
Figure 1 – Summary graphic illustrating the general features of the 8 databases that participated in the surveyesults
proportionate
te reporting p g
>1
f ICSR ≥3of ICSR ≥3CSR ≥3CSR ≥3
end flag is +ve ifend flag is +ve if
>EB95 for the
when current M score 26
erse events; anyerse events; any sed significantly
orts broken down by Reporter TypeFigure 4 - Top 5 Countries of origin
percentage of case reports used for signal detectionFigure 4 - Top 5 Countries of origin
percentage of case reports used for signal detectionorts broken down by Reporter Type percentage of case reports used for signal detectionpercentage of case reports used for signal detection
50% 60% 70% 80% 90% 100%50% 60% 70% 80% 90% 100%
f h iports for each reporter occupationased on unique reports for each category of reporterased on unique reports for each category of reporter
onal Lawyer Consumer Unknown Otheronal Lawyer Consumer Unknown Other
Table 4 Number of drugs with >=10 reports in the databasedatabases (% of reports with a value recorded1) and time to onset data
Table 4 – Number of drugs with >=10 reports in the database databases (% of reports with a value recorded ) and time to onset data
d Eth i it C t f S bj t ID Ti t
N b f d ith > 10 tnder Ethnicity Country of case
Subject ID Time to onset2 Number of drugs with >=10 reports
case onset 00%) (100%) (unk) MHRA 1,814 94%) (11%) (100%) (>0%) (54%)
,AEMPS 1 955) ( ) ( ) ( ) ( )
unk) (unk) AEMPS 1,955 GSK 2 146) ( )
97%) (100%) (57%) (5%) GSK 2,146 AZ 2 931
9 %) ( 00%) (5 %) (5%)99%) (100%) (59%) AZ 2,931
199%) (100%) (59%) 97%) (58%) (100%) (83%) (37%) EMA1 3,200 97%) (58%) (100%) (83%) (37%) 92%) (26%) (100%) (41%) (37%)
,UMC 5 71392%) (26%) (100%) (41%) (37%)
86%) (10%) (96%) (59%) (32%)UMC 5,713
1 (Estimate based only on reports including IMEs)86%) (10%) (96%) (59%) (32%) 2 First dose to event first occurrence
1. (Estimate, based only on reports including IMEs) 2. First dose to event first occurrence
f f h SOCf f h SOCs percentage of events for the top 5 SOCss percentage of events for the top 5 SOCs
Figure 6 Histogram of the number or reports (ALL REPORTS) by year 2000 2009Figure 6 – Histogram of the number or reports (ALL REPORTS) by year 2000 - 2009
ons
ween September 2010 andween September 2010 and e 69 000 to 5 391 000e 69,000 to 5,391,000
untries of origin of reports;untries of origin of reports; coded; availability andcoded; availability and gender age and country ofgender, age and country of
er proportions of non-er proportions of nonhan non-pharma databaseshan non pharma databases
ployed or use of meta-dataployed or use of meta data medical events)medical events). nd appear to reflect historicnd appear to reflect historic ue to rise in pharmaceuticalue to rise in pharmaceutical n of reporters, particularlyn of reporters, particularly nged over timenged over time
s is likely to be an yperformance of signal p garried out as part of p
vative Medicine Initiative Joint sources of which are composed of
(FP7/2007 2013) d EFPIAogramme (FP7/2007-2013) and EFPIA Pinkston belong to EFPIA (EuropeanPinkston belong to EFPIA (European anies in the IMI JU and costs related to
kind contribution under the IMI JU