Transcript Analysis of Safety Data

Analysis of Safety Data
Is More Enough? Marc Andersen
Senior Statistician
Outline / Conclusion
Analysis - For what purpose
• Different usages
Approaches and issues
• Best practice to be developed
Is more enough?
• It's needed
– Consolidation of methods
Some slides will be skipped, but present in handout
*: examples will be shown separately
Safety Analysis - For what
•
•
•
•
•
•
•
Annual Report (AR)
Investigator’s Brochure (IB)
Integrated Safety Summary (ISS, eCTD)
For planning of next study (protocol rationale)
Data Monitoring Commitees (DMC, DSMB)
Assessing spontaneous reports
Publications
Annual Report (AR) and
Investigators Brochure (IB)
Used for different purposes:
• IND's:
– 21 CFR 312.33
– http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/cfrsea
rch.cfm
• IRB's:
•
– ICH E6 section 8.3.19 (note wording: annual or interim)
– http://www.ich.org
Investigator’s Brochure (IB)
– ICH E6 section 7
– http://www.ich.org
Integrated Safety Summary (ISS)
•
M4E: The CTD - Efficacy. U.S. Department of Health and Human Services, Food
and Drug Administration, Center for Drug Evaluation and Research (CDER),
Center for Biologics Evaluation and Research (CBER), August 2001.
–
•
Guideline for the Format and Content of the Clinical and Statistical Sections of an
Application. Center for Drug Evaluation and Research, Food and Drug
Administration, Department of Health and Human Services, July 1988.
–
•
http://www.fda.gov/cder/guidance/statnda.pdf
Conducting a Clinical Safety Review of a New Product Application and Preparing a
Report on the Review. U.S. Department of Health and Human Services Food and
Drug Administration Center for Drug Evaluation and Research (CDER)
–
•
http://www.fda.gov/cder/guidance/4539E.pdf
http://www.fda.gov/cder/guidance/3580fnl.pdf
Common Terminology Criteria for Adverse Events v3.0 (CTCAE), Cancer Therapy
Evaluation Program, Common Terminology Criteria for Adverse Events, Version
3.0, DCTD, NCI, NIH, DHHS.
– http://ctep.cancer.gov/forms/CTCAEv3.pdf
May claim that ISS corresponds to eCTD section 2.7.4
Tips
• Get the timeslines
– More difficult to identify that imagined (!)
• There is no shortcut
– by study and integrated output are both needed
• Keep the perspective
– analysis on demand; aim for just in time
• Safety is looking for the unknown :-)
– so they will probably ask for more
What is the purpose of analysis
•
•
•
•
Exploratory – is it safe …
Signal detection – is there something …
Confirmatory – there is nothing …
Regulatory request – somebody else had …
Absense of Evidence is not Evidence of Absence
Alderson, BMJ editorial, 2004,
http://bmj.bmjjournals.com/cgi/reprint/328/7438/476
Altman, Bland, BMJ Letter, 2004,
http://bmj.bmjjournals.com/cgi/content/full/328/7446 /1016-b,
Altman, Bland, BMJ Statistics notes 1995,
http://bmj.bmjjournals.com/cgi/reprint/311/7003/485
Attributed to Carl Sagan http://www.quotationspage.com/quote/37901.html
Regulatory approach
Signal detection
Different Approaches
•
Proportional Reporting Ratio (PRR)
•
Reporting Odds Ratio
–
Pharmacoepidemiology and Drug Safety Volume 13, Issue 8:
–
http://www3.interscience.wiley.com/cgi-bin/jissue/109593697
–
Ronald D. Mann: Assessments of disproportionality
–
Kenneth J. Rothman, Stephan Lanes, Susan T. Sacks: The reporting odds ratio and its advantages over the proportional
–
•
reporting ratio
Patrick Waller, Eugène van Puijenbroek, Antoine Egberts, Stephen Evans: The reporting odds ratio versus the proportional
reporting ratio: deuce
Bayesian Confidence Propagation Neural Network (BCPNN)
–
Lindquist M, Ståhl 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;23:533-42
Lareb is the centre for knowledge about adverse drug reactions in the Netherlands. The government has instructed
us to register and analyse the adverse drug reactions from drugs and vaccines:
•
E.P. van Puijenbroek: Quantitative Signal Detection in Pharmacovigilance (2001)
–
http://www.lareb.nl/documents/thesis_evp_1059.pdf
•
Hauben M, Madigan D, Gerrits CM, Walsh L, van Puijenbroek EP: The role of data mining in
pharmacovigilance (2005)
–
http://www.lareb.nl/documents/eods2005_2057.pdf
Google search: adverse drug reactions data mining
Before anything: Integration ...
• Why: What is the purpose?
– Several study versus one study
– Subgroup analysis
• Integrated data may be more heterogenous
• Alternative
– extract results from study reports and pool (using
excel)
• Claim: integration is always nescessary!
To integrate without trouble
Consider implications for integration
• At protocol stage,
• At database design stage (e.g. lab
identification in database)
• View it as meta analysis problem
• Which treatment groups can be pooled:
– Context depending (e.g. by active, placebo; by
dose, placebo)
Integration - ideal versus reality
Ideal: SAS set statement combines the study datasets:
data lab;
set lab001 lab002 lab003 ...;
run;
• Be pragmatic *
•
– no formats, all formatted values to characters
Make reporting programs general
– simple to make a summary for any subpopulation, when
safety asks
• Tip: Data definition tables by study *
– definition and automation in SAS
http://www.lexjansen.com/pharmasug/2002/proceed/fdacomp/fda
04.pdf
When integrating
Get an overview early
• Study overview in calendar time *
Overview of study characteristics *
• Study characteristics
• Timing of treatment administration by study
• Dose in single dose / multiple dose studies
• Study Populations
• Study Periods
Coding / Standardisation
• AE coding: MEDRA
–
–
–
–
http://www.meddramsso.com
hierachy SOC, HLGT, HLT, PT and LLT
special searches categories (SSC)
Jürgen Kübler et all: "Adverse Event Analysis and MedDRA:
Business as Usual or Challenge?" Drug Information Journal,
Vol. 39, pp. 63-72, 2005
http://www.diahome.org/NR/rdonlyres/ECDCE5F2-B5D5-4313B64E-8026AAB684F6/0/DIJ39_1_63.pdf
• Note: WHO-ART: Adverse Reaction Terminology
http://www.umc-products.com/DynPage.aspx?id=4918&mn=1107
• CM coding: WHO Drug Dictionary
http://www.umc-products.com/DynPage.aspx?id=2829&mn=1107
• Issue: version of dictionaries
Structure: Domains
FDA: Regulatory Submissions in Electronic Format; New Drug
Applications (Issued 1/1999, Posted 1/27/1999) (IT3)
http://www.fda.gov/cder/guidance/2353fnl.pdf (not recent, but still
relevant)
– For single study datasets:
– Demographics, Inclusion criteria, Exclusion criteria, Concomitant
medication, Medical history, Drug exposure, Disposition, Efficacy
results, Human pharmacology and bioavailability/bioequivalence
data, Microbiology data, Adverse Events, Lab - chemistry, Lab hematology, Lab - urinalysis, ECG, Vital signs, Physical examination
CDISC, CDISC Submissions Data Model Version 2.0,
http://www.cdisc.org/publications/index.html
– Domains for DEMO, AE, CONMEDS, DISPOSIT, ECG, EXPOSURE,
LAB, PE, VITALS
Question: when is domain = dataset?
Integration - AE
Has the same conventions been used in all studies
• Symptoms not recorded as AE same across studies?
• Does out of range lab parameters lead to AE
registration
• Relevant of subpopulations may be domain specific
– One measurement may only be present in a few studies
– Only pregnancy test for ...
• When AE intensity changes, how is that captured: new
event or ...
• Is intensity registered using the same approach?
– mild/moderate/severe vs CTCAE Grade 1-5
AE data are standardised
• Typical items
–
–
–
–
–
–
–
–
Investigator term
Start (date or date and time)
Stop (duration)
Intensity
Severity
Ongoing (ongoing as of when)
Action taken
Relatedness
AE and Serious adverse events
SAE reconciliation
• Clinical database must be consistent with
Safety database
– Safety has the final word
Tip: usefull to list CIOMs numbers with AE’s *
Integration - lab
Lab parameters
• obtain documentation for each study:
– parameter names,
– lab analysis method,
– units (conversion !),
– reference ranges
Tips for analyses
• Do the simple stuff
– as it was a study, just with more treatment arms
• QC approach: number of records by study
• Agree on intervals for grouping continous
variables:
– exposure
– Age
• Identify the sub population
Summaries patient level
• This is covered for eCTD
– 2.7.4.1.2 Overall Extent of Exposure
– 2.7.4.1.3 Demographic and Other Characteristics of
Study Population
•
•
•
•
Treatments
Follow-up status / Vital status
Exposure
Discontinuation (treatment, study)
Presenting AE analysis
• What is the question?
– Number of patients with event relative to number of patients
treated (frequency and proportion)?
AE’s where proportion is > 5%, 1% etc? In which group?
Point Prevalence *
Incidence
Duration
–
–
–
–
• Tables or Graphs
– (usefull with row x columns layout) *
• Which AE’s occurs at the same time?
Question: Cross over studies?
AE tables lessons
• Ask!
– Identify groups of AE's to tabulate
• AE's in tables distinguish between N and E
• Full MEDRA hierachy is relevant *
• There will be lot of columns or two *
Relationships: continous variable
and AE time to event
• Table: grouped value before versus any AE (+/•
•
) by time interval
Graphical - AE start or measurement date on xaxis, cont. var on x axis
Time to event analysis
– Proportional hazards model (time dependent
predictor straightforward)
Issue: multiple events per patient; first event
only?
Challenges / further analyses
• Which level for analysis of AE's:
– SOC, HLGT, HLT, PT, LLT, SSC, selected AE's?
• Model occurrence of AE's
• Pharmacodynamic analysis (PD) modelling for
all lab parameters
– Simple PD model may work
Questions