Presentation is loading. Please wait.

Presentation is loading. Please wait.

Issues in the Practical Application of Data Mining Techniques to Pharmacovigilance A. Lawrence Gould Merck Research Laboratories May 18, 2005.

Similar presentations


Presentation on theme: "Issues in the Practical Application of Data Mining Techniques to Pharmacovigilance A. Lawrence Gould Merck Research Laboratories May 18, 2005."— Presentation transcript:

1 Issues in the Practical Application of Data Mining Techniques to Pharmacovigilance A. Lawrence Gould Merck Research Laboratories May 18, 2005

2 18 May 20051 Spontaneous AE Reports Clinical trial safety information is incomplete ° Few patients -- rare events likely to be missed ° Not necessarily ‘real world’ Need info from post-marketing surveillance & spontaneous reports : Pharmacovigilance Carried out by skilled clinicians & epidemiologists Long history of research on issue, e.g. ° Finney (1974, 1982) Royall (1971) ° Inman (1970)Napke (1970)

3 18 May 20052 Signal Generation: The Traditional Method Single suspicious case or cluster Potential Signals Identify Potential Signals Statistical Output Consult Programmer Consult Marketing Patient Exposure Integrate Information Refined Signal(s) Background Incidence Consult Literature Consult Database Comparative Data Consultation Action

4 18 May 20053 Some Limitations of Traditional Approach Incomplete reports of events, not reactions How to compute effect magnitude Many events reported, many drugs reported Bias & noise in system Difficult to estimate incidence because no. of pats at risk, pat-yrs of exposure seldom reliable Inappropriate to consider incidence using only spontaneous reports

5 18 May 20054 The Pharmacovigilance Process Detect Signals Traditional Methods Data Mining Generate Hypotheses Refute/Verify Type A (Mechanism-based) Type B (Idiosyncratic) Insight from Outliers Estimate Incidence Public Health Impact, Benefit/Risk Act Inform Change Label Restrict use/ withdraw

6 18 May 20055 Major Uses of Data Mining Identify subtle associations that might exist in large databases Early identification of potential toxicities Identify complex relationships not apparent by simple summarization Screening tool to identify potential associations to undergo clinical/epidemiological followup

7 18 May 20056 More to Pharmacovigilance than Data Mining Data mining a refinement to discover subtleties Still need initial case review respond to reports involving severe, potential life- threatening events eg., Stevens-Johnson syndrome, agranulocytosis, anaphylactic shock Clinical/biological/epidemiological verification of apparent associations is essential Need to think about most effective use of data mining in routine pharmacovigilance practice

8 18 May 20057 Statistical Methodology (1) Not the key issue Most use variations of 2-way table statistics No. ReportsTarget AEOther AETotal Target DrugabnTD Other DrugcdnOD TotalnTAnOAn Some possibilities Reporting Ratio: E(a) = nTD  nTA/n Proportional Reporting Ratio: E(a) = nTD  c/nOD Odds Ratio: E(a) = b  c/d Basic idea: Flag when R = a/E(a) is “large”

9 18 May 20058 Statistical Methodology (2) Estimate variability in various ways, e.g., usual chi- square statistic, Bayesian & Empirical Bayesian models) Similar results for all methods if more than a few drug/event combinations reported (e.g., 10) No non-clinical “gold standard” → can’t assess diagnostic utility of any method in usual sense OR > PRR > RR when a > E(a), doesn’t mean OR identifies real associations better than RR RR probably most stable

10 18 May 20059 Spontaneous Report Database Limitations Significant under reporting (esp. OTC) -- depending on seriousness or novelty of event, newness of drug, intensity of monitoriing Different regulatory reporting requirements Reflects only reporting practice, not incidence Synonyms for drugs & events → sensitivity loss Much duplication of reports Exposure rate unknown For any given report, there is no certainty that a suspected drug caused the reaction

11 18 May 200510 A Major Limitation (Often Ignored) Accumulated reports cannot be used to calculate incidence or to estimate drug risk. Comparisons between drugs cannot be made from these data Unfortunately, this still is done – disclaimers do not balance the effect of the misrepresentation Easy to show differences with data mining techniques, but impossible to make valid inferences about causality and may mislead

12 18 May 200511 Implementation Issues Portfolio bias in company databases can lead to inaccurate estimates of relative reporting rates Does public health benefit justify cost of following up signals detected by routine data mining methods? Variation in tools and databases among regulators could lead to significant cost without public health benefit Do frequency-based signal detection methods useful to regulators have business value in industry settings? Need examples of situations where computerized approach failed to identify important issues and where signals were “created” by publicity or reporting artifacts

13 18 May 200512 Mining is Easy, Refining Low-grade Ore is Hard What is data mining activity intended to accomplish -- what are the clinical/epidemiological/regulatory questions that need to be answered Need to address the impact of various factors, e.g., evolution of apparent association over time, association with key demographic factors such as age, sex, disease classification

14 18 May 200513 More Issues Composition of database may be important, important associations of a new drug could be cloaked by events associated with an old drug with similar mechanism of action Individual company databases tend to have comprehensive information about company products, but not general spectrum of drugs/ vaccines Databases contain reports mentioning drugs, not demonstrations of causality

15 18 May 200514 Discussion Most apparent associations represent known problems Some reflect disease or patient population ~ 25% may represent signals about previously unknown associations Statistical involvement in implementation & interpretation is important The actual false positive rate is unknown as are the legal and resource implications

16 18 May 200515 What Next? PhRMA/FDA working group is considering ways to address issues – white paper will be published May be worthwhile to construct & maintain a cleaned-up canonical database from AERS to provide a common resource for checking data mining findings based on individual company proprietary databases


Download ppt "Issues in the Practical Application of Data Mining Techniques to Pharmacovigilance A. Lawrence Gould Merck Research Laboratories May 18, 2005."

Similar presentations


Ads by Google