Selecting all cases in cohort – i.e. all subjects with adverse event
Randomly selecting a sample of pre-determined size of subjects from the cohort
Reduced version of cohort, with all cases added
Study design
Large Simple trial (LST)
Large randomized trials made simple by reducing data collection to minimum needed to test specific hypotheses
Randomization of treatment assignment
Controls of confounding by known or unknown risk factors
Sufficient statistical power small risks of common events + large risks of rare events
Study design
Efficiency of LSTs could be enhanced using health delivery systems databases
Relevant outcomes (e.g., hospitalizations for gastrointestinal bleeding) could be captured electronically which would eliminate need to contact patients for follow-up
Still necessary to identify eligible subjects, obtain consent and randomize treatment
Could be a very effective hybrid method
A typical scenario: identifying risks INCIDENCE OF COMMONLY OCCURRING EVENT FROM CLINICAL TRIALS Background Epidemiology Risk management strategy INCIDENCE OF EVENT IN GENERAL POPULATION IDENTIFICATION OF RISK FACTORS POTENTIAL SIGNALS OF RARE EVENTS Spontaneous Reports Observational Studies Other Activities 1/1,000 1/500 1/100 1/10,000 1/5,000 1/1,000 1/100,000 1/50,000 1/10,000 HYPOTHESIS TESTING 1/1,000,000 1/500,000 1/100,000, Clinical trial Data Insufficiently powered Evidence base Self-report bias Best study design….
Safety challenges in pre-approval phase
Challenges of safety signal identification during pre-approval phase:
pre-approval studies cannot usually be statistically powered to identify readily low frequency safety events of concern
samples often narrowly defined – therefore may not be truly representative of final treatment population
short time frames: mid to long-term effects cannot be thoroughly evaluated
Post-marketing pharmacovigilance
Limitations of post-marketing spontaneous reporting systems:
patterns of spontaneous reporting vary by country, increase during early post-launch period and are affected by publicity -> s pontaneous reporting rates biased
limited population level denominator information
for common events (e.g., MI) difficult to separate actual signals from background noise
Product life-cycle and safety studies Approval Product life-cycle FIM Ph I Ph II Ph III Ph IV
Toxicology (e.g. genotoxicity assays)
In silico analysis/ structural alerts
Margin of safety & no observed adverse effect level defined
Ph I & II: dose ranging, efficacy & toxicity
Phase III: demonstration of efficacy and safety signal monitoring
Ph IV: spontaneous reporting systems
Postmarketing pharmacovigilance Clinical Exposure in humans (Potential Denominator) Preclinical Sufficiently powered Easy study replication Large database studies No self-reporting bias
Example 1
The YASMIN (ethinylestradiol with drospirenone) study
Case Study - Yasmin and the Safety of DRSP
Background
Yasmin is an OC that contains drospirenone (DRSP)
DRSP has antimineralocorticoid activity and can raise serum potassium
Phase IV commitment
To obtain safety data related to:
Complications related to hyperkalemia (elevated serum potassium)
Physician prescribing and patient monitoring
Breakthrough pregnancy and birth outcomes
In 25,000 users of Yasmin and 50,000 users of other OCs
Wider business implications
Sponsor planned to introduce DRSP into menopausal therapy
Needed strong basis for later safety claims
Cohort Formation INGENIX Research Database Women (10-59 yrs) n = 959,482 Propensity score matching of 12 initiator cohorts (n = 22,429) New dispensing of Yasmin (n = 31,149) New dispensing of other OCs (n = 360,505) ≥ 6 months continuous enrollment (n = 22,887) ≥ 6 months continuous enrollment (n = 227,596) Propensity score matching of 12 initiator cohorts (n = 44,858) 1:2 Ratio
Data Collection
• Create matched cohorts of Yasmin and
other OC users
Collect comorbidities, demographics,
healthcare utilization, exposure to oral
contraceptives
• Identify women with hepatic, renal or
adrenal insufficiency
• Identify potassium-sparing meds
New OC or New type of OC OC Rx 6 Months Before Dispensing Prospective Data Collection
Identify clinical outcomes based on
claims data (prescription, diagnoses,
procedures and service)
Hospitalizations for new conditions,
clinical events potentially related to
hyperkalemia, electrolyte abnormality,
inappropriate prescribing, potassium
monitoring, pregnancy/fetal malformations
Yasmin Launch Q3 Time Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 2001 2002 2003 2004 Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow-Up Follow-Up Yasmin Control P Follow Follow Yasmin Control P 2005 Q1 Q2 Q3 Q4 Claims-Based Outcomes Chart-Based Outcomes Apr. Report Oct. Report Apr. Report Oct. Report Apr. Report Oct. Report Apr. Report Final Report
Blinded Chart Reviews to Confirm Outcomes Claims Based Abstraction Medical Chart Abstraction Screen Claims Data Medical Review
Rate Ratios Some outcomes may be continuations of pre-existing conditions * No rate ratio calculated as no case in Yasmin Cohort ** Composite hyperkalemia outcome comprised of chart-confirmed cases of arrhythmia, syncope, electrolyte disturbance, hyperkalemia, and myocardial infarction. Syncope Arrhythmia Hyperkalemia Other Electrolyte Disturbance Dialysis* Myocardial Infarction* Hospitalization with Hyper/Hypokalemia* Death Composite Hyperkalemia** 0 0 0 0 Incidence Rate Ratio –Yasmin versus Other OC (95% CI) 5.0 1.0 2.0 4.0 3.0 0.5 0.25 0.33 0.2 0.1
Does Yasmin lead to more VTE than other OCs?
NO. We added a new outcome which we could evaluate both retrospectively and prospectively for the full cohort.
Why do physicians prescribe to women with apparent contraindications of renal, hepatic, or adrenal disease?
MAY IGNORE NOMINAL CONTRAINDICATION FOR MILD DISEASE. We directly interviewed physicians with apparently contraindicated prescribing patterns.
Why do physicians not perform potassium monitoring as recommended in the label?
DISAGREE WITH LABEL. We surveyed physicians with patients who by label should have been monitored, but were not.
Flexible Data Collection Facilitates Risk Management
Example 2
The CRESTOR (rosuvastatin) studies
Brief background
Rosuvastatin
Hepato-selective and relatively hydrophilic hydroxy-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitor (statin)
Approved by FDA on 12 August, 2003
Management of dyslipidemia
Potential safety issues
Added concerns from withdrawal of cerivastatin in 2001 due to excess cases of fatal rhabdomyolysis
Comorbidities and concomitant medications may increase risk of myotoxicity, depending on properties of given statin
Evolution of Lipid Management ATP Guidelines* ATP I (1988) ATP II (1993) ATP III (2001) Diet; low-dose, non-statin monotherapy High-dose statin, combination therapy Low- to moderate-dose statin monotherapy Increasing aggressiveness of cholesterol-lowering therapy * The National Cholesterol Education Program Adult Treatment Panel (ATP)
The CRESTOR study program
An international collaborative pharmacoepidemiology
safety program involving many groups…
AstraZeneca R&D
Centro Espa ñol de Investigación Farmacoepidemiológica (CEIFE), Spain
PHARMO Institute, Netherlands
i3 Drug Safety, US
Drug Safety Research Unit (DSRU), UK
Pharmacoepidemiology program design
Program consisted of nine studies grouped into three components:
Patient characteristics studies (x 4 databases)
Safety evaluation studies (x 4 databases)
Prescription-event monitoring study (DSRU)
All studies performed according to GPP guidelines and reviewed and approved by ethics committees
Pharmacoepidemiology program overview
Table: Study parameters
Numbers and incidence rates of study outcomes
Results from McAfee et al. (2006) (Ingenix)
Safety evaluation studies
Pooled person-time and outcomes
Summary
Contributing areas and reasons to perform pharmacoepidemiology studies
Study design
Database studies
Two practical examples:
YASMIN (ethinylestradiol with drospirenone)
CRESTOR (rosuvastatin)
Dr David Neasham [email_address] +44 (0)1628 408442 Questions & answers Company logo here Practical use of Pharmacoepidemiology in clinical drug development
0 comments
Post a comment