Monitoring New Drugs for Safety in

. . Monitoring New Drugs for Safety in Insurance Claims Alexander M. Walker Sr VP Epidemiology I 3 Magnifi An Company

Monitoring New Drugs The Need A routine, comprehensive system § “All” drugs, “all” outcomes § Capable of generating signals – Verified elsewhere – Tested internally § Capable of testing signals from other sources § Serving all stakeholders – – – Copyright © 2005 i 3 Patients Doctors Payers Managed care Regulators Manufacturers 1

Monitoring New Drugs An i 3 Response to the Need • An active drug safety surveillance program • Using the full data assets of United Health Group concerning 11 million individuals • Pushing closer and closer to “real time” surveillance Copyright © 2005 i 3 2

Monitoring New Drugs Technical Number of drugs § ~20 NMEs introduced each year § Bring into follow when there is a critical number of users Data source § ~11 million individuals, open formulary § Growing constantly: Oxford, MAMSI, Americhoice, . . Data quality § These are insurance claims data, no better no worse § Significant in-house experience in sorting, cleaning, extracting medical data from these Reflects real-life drug use Copyright © 2005 i 3 3

Monitoring New Drugs Three Filters for Effective Claims. Based Surveillance • Treatment-emergent diagnoses § Events associated with diagnoses not seen prior to drug initiations § Filters out: continuation of disease • Comparator groups § Stat methods now exist for identifying comparison groups with similar distributions on demographics, diagnoses, drug use, and health services. Focus on drug-comparator differences. § Filters out: “confounding by indication, ” concomitant illnesses and their consequences • Data mining § Look for population and subgroups, as well as outcome combinations that may not be apparent in the crude tables § Filters out: deception by the typical result Copyright © 2005 i 3 4

Monitoring New Drugs Trade-offs in Filtering • Advantage § Systematic removal of confounding, background noise, and the dominance of the whole reduces false positives § Differences that emerge have greater scientific and rhetorical power • Disadvantage § Loss of ability to detect AE’s that are caught in the filter Copyright © 2005 i 3 5

Monitoring New Drugs Treatment-emergent diagnoses • Identify only § Hospitalizations, MD visits, other services § Occurring after dispensing § Not sharing 1 st 3 digits of ICD with any service in the 6 months preceding 1 st dispensing • Removes § Progression of disease § Concomitant illnesses Copyright © 2005 i 3 6

Monitoring New Drugs Comparator drug Problem: False positives and noise § Not all new Dx’s are drug AEs – “natural history” / background incidence § Claims far from perfect Response: Similar drug, similar people for comparison § Choose a standard drug (usually the most prescribed for same indication) § Choose users of that drug who resemble the users of Registry Drug with respect to medical history, drugs, doctors, procedures, health services utilization Statistically balanced for all elements available from claims Copyright © 2005 i 3 7

Monitoring New Drugs Drug Launch Time Treated Group Follow-Up Control Group Follow-Up Outcome Follow-up conducted in all sets of matched cohorts P Treated Group Follow-Up Control Group Follow-Up P Dynamic comparator matching Copyright © 2005 i 3 8

Monitoring New Drugs Data mining Good comparisons, good data still not enough – Drug/comparator tables can still obscure important relations – So many outcomes to look at Evaluate clinically meaningful subsets Search for hidden multi-way associations Copyright © 2005 i 3 9

Monitoring New Drugs The program Quarterly reports on all New Molecular Entities § Treatment patterns § New Diagnoses Emerging During Treatment § Data-mining § Interactive query § Data feed Annual print and web-based summaries Copyright © 2005 i 3 10

Monitoring New Drugs Timeline Beta-version of four drugs available end of August – – – NME Cialis Cymbalta Spiriva Ketek Comparator Viagra Effexor Atrovent Biaxin All NMEs with >1000 users since 2004 by end of 2005 Copyright © 2005 i 3 11