Database size and power to detect safety signals in pharmacovigilance


×
Rating : Rate It:
 
Post a comment
    Post Comment on Twitter
Comments:  
1 Favorites
ansarhameed,   favourited this   3 Years ago.
First Prev [1] Next Last



  Notes
 
 
1 : Database size and power to detect Safety Signals in Pharmacovigilance Isaac W. Hammond, PhD, MD, MPH Trevor G. Gibbs, MD Harry A. Seifert, MD, MSCE Donna S. Rich, MT (ASCP)
2 : Introduction Signal detection is intended to identify an association between a drug and an adverse drug reaction which is considered important, previously unknown or not fully characterized. Pharmaceutical companies and most regulatory agencies collect ADRs directly from healthcare professionals and consumers for independent evaluation to protect public health. Pharmaceutical companies submit only reports of ADRs that are serious, unexpected and suspected to regulatory agencies. The United States Food and Drug Administration (US FDA) collects and evaluates adverse drug reactions on drugs approved in the United States using a database called the Adverse Event Reporting System (AERS). The US FDA uses the Bayesian Gamma Poisson Shrinker (BGPS) disproportionality method of signal detection1-4. Similarly the World Health Organization (WHO) collects and evaluates adverse drug reactions on drugs approved internationally using the WHO database for ADRs called Vigibase. The WHO Uppsala Monitoring Centre uses the Bayesian Confidence Propagation Neural Network (BCPNN) method of signal detection. GlaxoSmithKline uses BGPS method of signal detection
3 : Methods Evaluation of databases Three large global databases included in this study: GSK’s global safety database (OCEANS) FDA’s safety database – (Adverse Events Reporting System (AERS)), and WHO’s global safety database – (Vigibase). Sample Data Ten drugs marketed by GSK were randomly selected using a computerized random number generator and the three databases were queried to determine the number of adverse drug reactions in each database for these 10 drugs. GSK does not have access to Vigibase, so we received cooperation from the World health Organizations’ Uppsala Monitoring Center for their data. Using the Bayesian Gamma Poisson Shrinker disproportionality method of signal detection, we searched GSK’s OCEANS and FDA’s AERS databases for adverse drug reactions that were unlabeled at the time of identification for each drug; such events would represent a safety signal.
4 : Power calculation Odds ratios and power were calculated using the maximum likelihood theory where: odds ratio (OR) = {p2/1 + p2)}/{p1/1 + p1)}, and Power = Cln(sqrt(Vn) * Ln(OR) – 1.96) Where Cln(x) = probability (standard normal distribution
5 : Results Data cut off date was September 2005 FDA’s AERS database contained 6.2 million records WHO’s Vigibase database contained 7.2 million records GSK’s OCEANS database contained 2 million records. However, for each of the 10 drugs OCEANS contained more records than AERS and vigibase (except for 2 drugs – topotecan and timentin in vigibase)
6 : Results
7 : Results
8 : Results
9 : Results
10 : Results
11 : Results
12 : Results
13 : Results
14 : Results
15 : Results
16 : Results
17 : Discussion Early safety signal detection, i.e., when the reporting rates are low, is essential for the protection of public health, because it allows for intervention using risk management and risk minimization plans. This could potentially prevent unnecessary exposure of other subjects. When reporting rates are high, there is no difference in the power to detect signals regardless of the size of the drug specific database. Power is mathematically expressed as 1- ß (the complement of ß [beta]). Beta is a designation for type II error, which is the probability of accepting a false null hypothesis. The probability of committing a Type II error (ß) decreases as the sample size (N) increases, and thus power increases with increasing sample size.
18 : Discussion contd. GSK has the smallest overall database size of the three, but in general has the largest number of reports for each drug. This is to be expected, because drug manufacturers are required by law to provide their addresses and telephone numbers with the drug for the purpose of receiving questions and complaints about their drugs. Additionally, drug manufacturers are also required by law to submit certain reports of adverse drug reactions to regulators (i.e., events that are serious and unexpected). However, not all reports received by manufacturers meet the regulatory definition for submission to regulatory authorities. Therefore it is reasonable to expect that a drug manufacturer would have more drug specific ADR reports than regulatory authorities. From Table 1, it can be seen that the database pair that resulted in the highest power was the pair with the highest total drug specific sample size (GSK/WHO) (Figures 1, 2, 3, 5, 7, 8 and 9). Similarly, where the GSK and FDA database pairs resulted in the highest total, they resulted in the highest power (Figures 4, 6 and 10). It appears that the overall database size may not be the determinant of the power to detect safety signals. Rather the number of reports for a given drug may be the determinant of the power to detect safety signals for that given drug.
19 : Conclusion We concluded from this study that: the overall size of a database is not a determinant of its statistical power to detect safety signals, but rather the size of the drug-specific data is a determinant of the power to detect safety signals. Also, we believe consideration should be given to the use of multiple large global databases, and that reliance on a single database could reduce statistical power and diversity. Further studies are needed to expand the focus of this study, for example, to investigate drugs from other pharmaceutical companies in order to confirm our findings. This would make the conclusions more general.
20 : Acknowledgements The authors would like to express their sincere gratitude to Anders Viklund, Erica Walette, and Mats Persson of the World Health Organization’s Uppsala Monitoring Center in Sweden for their assistance and patience in providing us with information from Vigibase.
21 : References Szarfman A, Machado SG, O’Neil RT. 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 (2002) 25(6): 381–392. Hauben M, Reich L, Chung S. Postmarketing surveillance of potential fatal reactions to oncology drugs: potential utility of two signal-detection algorithms. Eur. J. Clin. Pharmacol. (2004) 60: 747–750. Bayesian data mining in large frequency tables, with an application to the FDA spontaneous reporting system. The American Statistician (1999) 53(3): 177–190. Almenoff JS, DuMouchel W, Kindman LA, Yang X, Fram D. Disproportionality analysis using Bayes data mining: a tool for the evaluation of drug interactions in the post-market setting. Pharmacoepidemiology & Drug Safety (2003) 12(6):517–521. Bate A, Lindquist M, Edwards IR, Orre R. A data mining approach for signal detection and analysis. Drug Safety (2002) 25(6): 393–397. Bate A, Lindquist M, Orre R, Edwards IR, Meyboom RHB. Data-mining analyses of pharmacovigilance signals in relation to relevant comparison drugs. Eur. J. Clin. Pharmacol. (2002) 58: 483–490. Lindquist M, Stahl M, Bate A, Edwards IR, Meyboom RHB. A retrospective evaluation of a data mining approach to aid finding new adverse drug reaction signals in the WHO international database. Drug Safety (2000) 23(6): 533–542.
22 : References contd Bate A, Lindquist M, Edwards IR, Olsson S, Orre R. Lasner A, De Freitas RM. A Bayesian neural network method for adverse drug reaction signal generation. Eur. J. Clin. Pharmacol. (1998) 54: 315–321. Whitehead John. Measurement of treatment difference. In: The Design and Analysis of Sequential Clinical Trials. Revised 2nd Edition; John Wiley, New York, (2000): 29-68. Hammond IW, Rich D. Consumers usurp spontaneous adverse event reporting in the United States. Pharmacoepidemiology & Drug Safety (2005) 14(suppl. 2): 017, S8. Kubota K, Koide D, Hirai T. Comparison of data mining methodologies using Japanese spontaneous reports. Pharmacoepi. & Drug Safety (2004) 13: 387–394. WWW References 201. http://www.fda.gov/cder/dpe/annrep96/index.htm Stepper H. Annual adverse drug experience report: 1996 FDA, 202. http://www.fda.gov/cder/dpe/annrep95/index.htm Knapp DE, Robinson JI, Britt AL. Annual adverse drug experience report: 1995 FDA

 

Add as Friend By : Isaac W. Hammond, PhD, MD, MPH
Added On : 6 Years ago.
PowerPoint Presentation on Database size and power to detect safety signals in pharmacovigilance or    more
Views 2129 | Favourite 1 | Total Upload :1

Embed Code:

Flag as inappropriate


Related  Most Viewed



Free Powerpoint Templates



 



Medical PowerPoint Templates | Powerpoint Templates | Tags | Contact | About Us | Privacy | FAQ | Blog

© Slideworld