Vol. 25, No. 1, (2004) Early Detection of Important Safety Information Recent Methods for Signal Detection 1 2, Hiroyuki Watanabe

Size: px

Start display at page:

Download "Vol. 25, No. 1, (2004) Early Detection of Important Safety Information Recent Methods for Signal Detection 1 2, Hiroyuki Watanabe"

ゆゆこがうん
4 years ago
Views:

1 Vol. 25, No. 1, (2004) Early Detection of Important Safety Information Recent Methods for Signal Detection 1 2, Hiroyuki Watanabe 1, Yasuyuki Matsushita 2, 3, Atsushi Watanabe 4, Toshiro Maeda 5, Kazuhiko Nukui 6, Yoshimasa Ogawa 7, Jungo Sawa 8 and Hiroshi Maeda PMS Data Management and Biostatistics, Clinical Development Institute, BANYU PHARMACEUTICAL CO., LTD. 2 Clinical Pharmacology and Biostatistics Department, New Drug Development Division, SANKYO CO., LTD. 3 Management Sciences Department, Faculty of Engineering, Tokyo University of Science 4 Medical and Drug Information, ELI LILLY JAPAN K.K. 5 Post-Marketing Data Management Development, Pharmacovigilance Unit, Pharmacovigilance & Quality Assurance Division, Mitsubishi Pharma Corporation. 6 Drug Safety & Post Marketing Surveillance Department, Sumitomo Pharmaceuticals Co., Ltd. 7 Post Marketing Surveillance Department, JANSSEN PHARMACEUTICAL K.K. 8 Biometrics Department, R & D Institute, Schering-Plough K.K. 9 Biostatistics and Planning, Development Division, FUJISAWA PHARAMACEUTICAL CO., LTD. hiroyuki watanabe@merck.com Received March Revised April Accepted April 2004.

2 38 It is very important to provide safety information of new drugs to physicians and patients as soon as possible after the early postmarketing period. For that purpose, it is important to appropriately collect and analyze the spontaneous reports accumulated in databases of companies and regulatory agencies. This paper reviews the analytical methods to assess spontaneous reports. Bate et al. (1998) presented Bayesian Confidence Propagation Neural Network (BCPNN) Method used by Uppsala Monitoring Centre (UMC) of the World Health Organization (WHO). DuMouchel (1999) presented Gamma-Poisson Shrinker (GPS) Program of U. S. Food and Drug Administration (FDA), and Evans et al. (2001) presented Proportional Reporting Ratios (PRR) of the Medicines Control Agency (MCA). Furthermore, DuMouchel and Pregibon (2001) extended the GPS Program, proposing the Multi-Item Gamma Poisson Shrinker (MGPS) Program, which then became the standard method for the FDA. This report also reviews the practical problems (e.g. database, duplication cases, code of Medical Dictionary for Regulatory Activities (MedDRA)) encountered in Japan. Key words: Post-Marketing Surveillance; Pharmacovigilance Method; Signal Detection; Spontaneous Reports 1. ( Post-Marketing Surveillance; PMS Good Post-Marketing Surveillance Practice; GPMSP Spontaneous Reports Adverse Drug Reaction; ADR 6

3 39 3 Bate et al World Health Organization WHO Uppsala Monitoring Centre UMC Bayesian Confidence Propagation Neural Network BCPNN Method DuMouchel 1999 Food and Drug Administration FDA Gamma- Poisson Shrinker GPS Program Evans et al Medicines Control Agency MCA Medicines and Healthcare products Regulatory Agency MHRA Proportional Reporting Ratios PRR Du- Mouchel and Pregibon 2001 GPS Program Multi-Item Gamma Poisson Shrinker MGPS Program FDA Signal Detection, 2001 DuMouchel GPS Program DuMouchel, MCA FDA WHO Signal Generation Symposium International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use ICH E2E Pharmacovigilance Planning Step2 ICH Routine pharmacovigilance practices Pharamacovigilance Methods Systematic methods

4 40 2. Signal WHO Edvards and Aronson, 2000 Signal Detection, 2001 MHRA FDA WHO n 11 n 12 n 1+ n 21 n 22 n 2+ n +1 n +2 n P 11 P 12 P 1+ P 21 P 22 P 2+ P +1 P +2 P ++ (= 1) n ij n ij n 11 3 k n 11k ADR Adverse Event; AE 2 2

5 2.1.1 Proportional Reporting Ratios (PRR 41 PRR MHRA 95% Evans et al., 2001; van Puijenbroek, 2002 n 11,n 21 n 1+,n 2+ P RR = n11/n1+ = P11/P1+, SE(logP RR) = n 21/n 2+ P 21/P 2+ r 95%CI = e log(p RR)± n n 1 1+ n 1 21 n 2+ s 1 n 11 1 n n 21 1 n 2+ 3 (1) P RR = 2 (2) χ 2 n++( n11n22 n12n21 n++/2)2 = = 4 p < 0.05 (3) n 11 = 3 n 1+n 2+n +1n +2 van Puijenbroek et al P RR 1.96SE > Bayesian Confidence Propagation Neural Network BCPNN Method BCPNN WHO Bayes Bate et al., 1998; Lindquist et al., 2000; Bate et al., 2002 IC ij information component log 2(P ij/p i+p +j) P i+,p +j P ij IC ij 2 IC 11 IC 1 P i+ (i = 1,2) P +j (j = 1,2) Beta distribution P ij((i,j) = (1,1),(1,2),(2,1),(2,2)) Dirichlet distribution P i+ α 1 α 2 α 1 + α 2 = α P +j β 1 β 2 β 1 + β 2 = β P ij γ 11,γ 12,γ 21,γ 22 IC 11 P 1+,P +1 P 11 P γ 11 γ α 1 = β 1 = 1, α = β = 2, γ 11 = γ 12 = γ 21 = γ 22 =

6 42 1 1/2 Haldane-Anscombe 1/2 correction Breslow, Greenland, 2000 IC 11 Orre et al., 2000 (n 11 + γ 11)(n ++ + α)(n ++ + β) E(IC 11) = log 2 (n ++ + γ)(n 1+ + α 1)(n +1 + β 1) V (IC 11) = 1 log 2 2 n ++ n 11 + γ γ 11 (n 11 + γ 11)(1 + n ++ + γ) n++ n+1 + α α1 n++ n+1 + β β1 + + (n 1+ + α 1)(1 + n ++ + α) (n +1 + β 1)(1 + n ++ + β) (n ++ + α)(n ++ + β) γ = γ 11, γ11 = 1, α1 = β1 = 1, α = β = 2 (n 1+ + α 1)(n 1+ + β 1) 95% 0 E(IC 11 2SE 11) = E(IC 11 2 p V (IC 11)) > Gamma-Poisson Shrinker GPS Program GPS FDA DuMouchel, 1999 MGPS Szarfman et al., 2002 GPS BCPNN IC GPS 1 2 Gamma distribution RR ij DuMouchel 1999 RR ij Relative Risk SMR; Standardized Mortality Ratio Breslow and Day, 1987 RR ij = n ij/e ij E ij FDA 5 1 s X E ij = s ni+sn+js/n++s RR ij Empirical Bayes 2

7 43 µ ij n ij λ ij = µ ij/e ij λ ij 2 λ λ π(λ; α 1,β 1,α 2,β 2,P ) = P g(λ; α 1,β 1) + (1 P )g(λ; α 2,β 2) P α 1/β 1 + (1 P )α 2/β 2 P (1 P )(α 1/β 1 α 2/β 2) 2 + P α 1/β (1 P )α 2/β 2 2 g(λ; α,β) = β α λ α 1 e βλ /Γ(α), Γ(α) = R 0 e t t α 1 dt L(θ) = Π ij{p f(n ij; α 1,β 1,E ij) + (1 P )f(n ij; α 2,β 2,E ij)} f(n; α,β,e) = (1 + β/e) n (1 + E/β) α Γ(α + n)/γ(α)n! λ f( ) Q n = P f(n; α 1,β 1,E)/[P f(n,α 1,β 1,E) + (1 P )f(λ; α 2,β 2,E)] E[λ N = n] = Q n(α 1 + n)/(β 1 + E) + (1 Q n)(α 2 + n)/(β 2 + E) E[log(λ) N = n] = Q n[ψ(α 1 + n) log(β 1 + E)] + (1 Q n)[ψ(α 2 + n) (β 2 + E)] Ψ(x) log[γ(x)] digamma Ψ(x) = Γ (x)/γ(x) L(θ) 5 θ = (α 1,β 1,α 2,β 2,P ) DuMouchel θ θ = (α 1 = 0.2, β 1 = 0.1, α 2 = 2, β 2 = 4, P = 1/3) λ Empirical Bayes Geometric Mean; EBGM EBGM ij = exp{e[log(λ ij) n ij]} EBGM ij E ij RR ij E ij shrink λ 5% EB05 95% EB05 signal score FDA EB05 = 2 Szarfman et al., 2002 Dumouchel and Pregibon 2001 EBGM, EB05 EXCESS EXCESS EXCESS ij = E ij(eb05 ij 1)

8 Multi-Item Gamma-Poisson Shrinker MGPS Program MGPS FDA GPS Program DuMouchel and Pregibon, 2001; Szarfman et al., 2002 GPS, MGPS Empirical Bayes Screening EBS Hauben and Zhou, 2003; Kubota et al., E A,B s A,B E P As,P Bs,P Es s n s A,B,E E0 E0 ij = P s nspispjs (i, j = A, B, E, i j) E0ABE = P s nspaspbspes m ABE A,B,E 0 1 a,b 1,b 2,b 3,c 12,c 13,c 23 logm ABE = a + b A 1 + b B 2 + b E 3 + c AB 12 + c AE 13 + c BE 23, 2003 (EBGM ij E ij) A = 1, B = 1, E = 1 E[m 111] E2 ABE EXCESS2 ABE = EBGM ABE E0 ABE E2 ABE EBGM ij, EBGM ABE GPS ROR Yule s Q Poisson Chi square Yates van Puijenbroek et al., ) Reporting Odds Ratio ROR ROR 95% s ROR = (n11/n21) (n = n11n22 1, SE(logROR) = /n 22) n 12n 21 n 11 n 12 n 21 r 95%CI = e ln(ror)± n n n 1 21 n 22 95% 1 2) Yule s Q n 22 Yule s Q 95%

9 Q = 45 s n11n22 n12n21, SE Q = 1 1 n 11n 22 + n 12n 21 2 (1 Q2 ) n 11 n 21 n 12 n 22 95%CI = Q ± 1.96SE Q 95% 0 3) Poisson Poisson p p = 1 a 1 X k=0 e µ µ k k! a µ p ) Chi square Yates Chi square Yates 2 2 χ 2 = n++( n11n22 n12n21 n++/2)2 n 1+n 2+n +1n +2 p 0.05 χ ) 1 β 1 exp(β 1) logit(y) = β 0 + β 1x 1 x y 1 0 β 0,β 1 β 1 95% 0 logit(y) = β 0 + β 1x 1 + β 2x 2 + β 3x 3 + β 4x 4 x x 2,x 3,x 4 y 1 0 β 0,β 1,β 2,β 3,β 4

10 46 β 1 6) 2 β 12 exp(β 12) logit(y) = β 0 + β 1x 1 + β 2x 2 + β 12x 1x 2 x x y 1 0 β 0,β 1,β 2,β 12 β 12 95% 0 3 NSAIDs van Puijenbroek et al., 2000; Egberts et al., n 11 E 11 xxxxx Bate et al WHO captopril % BCPNN GPS Szarfman et al CDER Monitoring Adverse Reports Tracking System; MART GPS MART MART 1 4 GPS

11 47 1. IC captopril IC IC 95% European Journal of Clinical Pharmacology 1998; 54: MART GPS MART MART GPS GPS MART Drug Safety 2002; 25(6):

12 48 MGPS 3 A 4 EB05 = 1.5,2,4,8 Receiver Operating Characteristic; ROC Szarfman et al., ROC A MGPS true positive TP, y 1 false positive FP, x Drug Safety 2002; 25(6): GPS GPS FDA ftp://ftp.research.att.com/dist/gps DuMouchel, 1999 EBGM 10 ACE EBGM = 5

13 49 ACE 3 N Rank EBGM 3. ACE Rank N E RR EBGM EBlog2 Drug enalapril benazepril hydrochloride lisinopril ramipril quinapril hydrochloride fosinopril sodium captopril lovastatin 4 CK 7 Coding Symbols for Thesaurus of Adverse Reaction Terms COSTART COSTART Medical Dictionary for Regulatory Activities MedDRA Terminology MedDRA System Organ Class; SOC High Level Group Term; HLGT High Level Term; HLT Preferred Term; PT Lowest Level Term; LLT 5 Szarfman et al MGPS MedDRA PT FDA PT 4. Rank N E RR EBGM EBlog2 Event CK

14 Kubota et al BCPNN, GPS, PRR MHRA PRR P RR 1.96SE > 1 ROR κ κ PRR MHRA BCPNN GPS 2 7.5% BCPNN, PRR MHRA 2 Van Puijenbroek et al PRR, ROR BCPNN 4 Gould 2003 DuMouchel 1999 BCPNN GPS ,000 BCPNN PRR 25,000, FDA Szarfman et al., WHO Bate et al., , PRR, BCPNN 1/100 1/50,000 Meyboom et al., /250,000 5 WHO 2002 Drug Safety Bate et al., 2002; Brown, 2002; Coulter, 2002; Egberts et al., 2002; Kubota, 2002; Meyboom et al., 2002; Nelson et al., 2002; Peachey, 2002; Purcell and Barty, 2002; Shakir and Layton, 2002; Szarfman et al., 2002 Gold Standard ICH E2B-M Standard Generalized Markup Language SGML

15 51,

16 ICH E2A PMS AE ADR AE ADR AE ADR

17 ) 2) 3) 4) ICH E2B-M2 B.4.k ) 2) 1)

18 54 WHO Drug 2) 7% PMS Japanese Adverse Reaction Terminology J-ART ICH E2B-M2 MedDRA MedDRA 3 1) MedDRA 2) 5 Special Search Categories; SSC 3) J-ART MedDRA MedDRA MedDRA MedDRA 5 MedDRA PT HLT MedDRA World Health Organization Adverse Reaction

19 55 5. PT HLT V5.1J PRIMARY SOC SSC SSC PT HLT (981) 490(357) 6. 2 PT 69 PT PT NOS NOS NOS Terminology WHO-ART Brown, 2002 MedDRA PT SOC 5 PT HLT 2 PT 6 SMQ Standardised MedDRA Queries J-ART MedDRA J-ART J-ART MedDRA MedDRA J-ART Mapping

20 56 7. J-ART MedDRA V5.1 mapping mapping PT LLT (Y) Y N (Y) Y N N J-ART LLT mapping J-ART 2 5,048 MedDRA mapping J-ART 62 MedDRA V5.0 V LLT PT N Y J-ART 62 MedDRA mapping 1,711 J-ART MedDRA mapping PT LLT primary SOC MedDRA LLT mapping LLT N 7 J-ART MedDRA V5.1 mapping N J-ART mapping N Y 1 V Y N PT V Y 1) 2) MedDRA SOC primary PT PT PT SOC

21 PRR PRR 5. PRR, BCPNN, GPS, MGPS 1 1

22 58 1) 2) 3) PMS PMS, PMS, 2003 Clark, 2001; Hauben and Zhou, DM DM DM Bate, A., Lindquist, M., Edwards, I. R., Olsson, S., Orre, R., Lansner, A. and De Freitas, R. M. (1998). A Bayesian neural network method for adverse drug reaction signal generation. European Journal of Clinical Pharmacology 54, Bate, A., Lindquist, M., Edwards, I. R., Orre, R. (2002). A data mining approach for signal detection and analysis. Drug Safety 25, Breslow, N. E. (1981). Odds ratio estimators when the data are sparse. Biometrika 68, Breslow, N. E. and Day N. E. (1987). Statistical Methods in Cancer Research, Vol. II-Design and Analysis of Cohort Studies. Oxford University Press. Brown, E. G. (2002). Effects of coding dictionary on signal generation: A consideration of use of MedDRA compared with WHO-ART. Drug Safety 25, Clark, J. A., Klincewicz, S. L. and Stang, P. E. (2001). Spontaneous adverse event signaling methods: classification and use with health care treatment products. Epidemiologic Reviews 23,

23 59 Coulter, D. M. (2002). Signal generation in the New Zealand Intensive Medicines Monitoring Programme: a combined clinical and statistical approach. Drug Safety 25, DuMouchel, W. (1999). Bayesian data mining in large frequency tables, with an application to the FDA spontaneous reporting system. The American Statistician 53, DuMouchel, W. and Pregibon, D. (2001). Empirical bayes screening for multi-item associations. Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining, Edvards, I. R. and Aronson, J. K. (2000). Adverse drug reactions: definitions, diagnosis and management. THE LANCET 356, Egberts, A. C. G., Meyboom, R. H. B. and van Puijenbroek, E.P. (2002). Use of measures of disproportionality in pharmacovigilance: three Dutch examples. Drug Safety 25, Evans, S. W., Waller, P. C. and Davis, S. (2001). Use of proportional reporting ratios (PRRs) for signal generation from spontaneous adverse drug reaction reports. Pharmacoepidemiology and Drug Safety 10, Gould, A. L. (2003). Practical pharmacovigilance analysis strategies. Pharmacoepidemiology and Drug Safety 12, Greenland, S. (2000). Small-sample bias and corrections for conditional maximum-likelihood odds-ratio estimators. Biostatistics Hauben M. and Zhou X. (2003). Quantitative methods in pharmacovigilance. Drug Safety 26, (2003). ( 15 )., 411. (2001). MCA, FDA, WHO. 6, Kubota, K. (2002). Prescription-Event-Monitoring in Japan (J-PEM). Drug Safety 25, Kubota, K., Koide, D., Hirai, T. (2004). Comparison of data mining methodologies using Japanese spontaneous reports. Pharmacoepidemiology and Drug Safety, in press. Lindquist, M., Stahl, M., Bate, A., Edwards, I. R. and Meyboom, R. H. B. (2000). A retrospective evaluation of a data mining approach to aid finding new adverse drug reaction signals in the WHO international database. Drug Safety 23, Meyboom, R. H. B., Egberts, A. C. G., Edwards, I. R., Hekster, Y. A., de Koning, F. H. and Gribnau, F. W. (1997). Principles of signal detection in pharmacovigilance. Drug Safety 16, Meyboom, R. H. B., Lindquist, M., Egberts, A. C. G. and Edwards, I. R. (2002). Signal selection and follow-up in pharmacovigilance. Drug Safety 25,

24 60 Nelson, R. C., Palsulich, B. and Gogolak, V. (2002). Good pharmacovigilance practices: technology enabled. Drug Safety 25, (2001) Orre, R., Lansner, A., Bate, A. and Lindquist, M. (2000). Bayesian neural networks with confidence estimations applied to data mining. Computational Statistics & Data Analysis 34, Peachey, J. (2002). From pharmacovigilance to pharmacoperformance. Drug Safety 25, PMS (2002). PMS PMS. 7(Suppl.), S22-S28. PMS (2003). PMS 8, Purcell, P. and Barty, S. (2002). Statistical techniques for signal generation. Drug Safety 25, Shakir, S. A. W. and Layton, D. (2002). Causal Association in pharmacovigilance and pharmacoepidemilogy: thoughts on the application of the Austin Bradford-Hill criteria. Drug Safety 25, Szarfman, A., Machado, S. G. and O Neill, R. T. (2002). Use of screening algorithms and computer systems to efficiently signal higher-than-expected combinations of drugs and events in the US FDA s spontaneous reports database. Drug Safety 25, Van Puijenbroek, E. P., Egberts, A. C. G., Heerdink, E. R. and Luefkens, H. G. M. (2000). Detecting drug-drug interactions using a database for spontaneous adverse drug reactions: an example with diuretics and non-steroidal anti-inflammatory drugs. European Journal of Clinical Pharmacology 56, Van Puijenbroek, E. P., Bate, A., Leufkens, H. G. M., Lindquist, M., Orre, R. and Egberts, A. C. G. (2002). A comparison of measures of disproportionality for signal detection in spontaneous reporting systems for adverse drug reactions. Pharmacoepidemiology and Drug Safety 11, 3-10., (2003) Signal Detection,,,,, ( ) (2003)...

Vol. 26, No. 2, (2005) Rule of Three Statistical Inference for the Occurrence Probability of Rare Events Rule of Three and Related Topics Manabu

Vol. 26, No. 2, (2005) Rule of Three Statistical Inference for the Occurrence Probability of Rare Events Rule of Three and Related Topics Manabu Vol. 26, No. 2, 53 63 (25) Rule of Three Statistical Inference for the Occurrence Probability of Rare Events Rule of Three and Related Topics Manabu Iwasaki and Kiyotaka Yoshida Department of Computer