Roche Template

Causality Assessment in Practice
Pharmaceutical Industry Perspective
Lachlan MacGregor
Senior Safety Scientist
F. Hoffmann-La Roche Ltd., Basel

Disclaimer:
The opinions expressed in this presentation are those of the author, and do not reflect those of F. Hoffmann-La Roche
Ltd.

Overview of presentation
• Causality and case processing
• Guidance from regulators and CIOMS
• Naranjo algorithm at Roche
• Language to express uncertainty
• Reaching definitive conclusions about drug-event pairs
• Conclusions
• Final thoughts

Causality and case processing
• Spontaneous case reports
‟ Reporter’s assessment of causality is often not explicit
‟ Roche standard AE form:
„ Reporter selects `related’, `not related’ `unknown’ or `not provided’ for each drug-event pair.
„ Interpretation of `related’ is left to the reporter
‟ Spontaneously reported events are considered related, for the purpose of case processing and regulatory
reporting, unless reporter says `not related’
• Study case reports
‟ Investigator codes events as related or not related to the study drug
‟ Some instruction for the investigator are provided in the study protocol
‟ SUSAR: `Reaction’ implies `related’

Instructions for investigators
Typical study protocol provides information like this:
Investigators should use their knowledge of the patient, the circumstances surrounding the event, and an evaluation of any potential alternative causes to
determine whether or not an adverse event is considered to be related to the study drug, indicating "yes" or "no" accordingly.
The following guidance should be taken into consideration:
• Temporal relationship of event onset to the initiation of study drug
• Course of the event, considering especially the effects of dose reduction, discontinuation of study drug, or re-introduction of study drug (where
applicable)
• Known association of the event with the study drug or with similar treatments
• Known association of the event with the disease under study
• Presence of risk factors in the patient or use of concomitant medications known to increase the occurrence of the event
• Presence of non treatment-related factors that are known to be associated with the occurrence of the event

Standard SAE Report Form
Causality information

Case processing Study cases

Serious unexpected events in randomized trials
What if the patient is taking placebo
• General principles:
‟ When assigning causality (e.g. for SUSARs), the investigator should assume that the patient is on the study
treatment
‟ Aim to preserve blinding where possible. Unblind in emergencies only
‟ Submit SUSAR report regardless of actual treatment
„ Investigators and sponsor receive blinded SUSAR reports
„ If unblinded+placebo, no SUSAR report is submitted to authorities
‟ DSMBs may request unblinded data, while investigator and patient remain blinded

Adding `events’ to the reporter’s original list
• `Events’ may be added by the company to the reporter’s original list, where there are additional symptoms, signs, other
medical conditions etc. described in
…the original fax/email
…the free text section of the report form
…the `past history’ or `concurrent illness’ section,
unless clearly consistent with the reported diagnosis
• Interpretation of these events is sometimes difficult
‟ May not have been of concern to the reporter
‟ No `causality’ information provided by the reporter

Reporter’s causality assessment
Example: One drug, 12 month period 2010-2011
• Events from study cases:
568/2155 (26%) `not related’ (2 re-assigned `related’ by the company)
944/2155 (44%) `related’
643/2155 (30%) no reporter causality assessment
• Events from spontaneous cases:
199/2681 (8%) `not related’
1191/2681 (44%) `related’
1291/2681 (48%) no reporter causality assessment
Note: All spontaneous events are considered `related’ by the company for reporting purposes

Methods for the assessment of individual cases
Guidance
CIOMS VI (2005):
• Survey of pharmaceutical companies:
‟ 12/21 companies use `introspection’, 2 home-grown algorithms, 3 published methods, 4 non- specific methods
• Recommendations:
‟ Binary yes/no causality (i.e. related/not related) for study investigators
„ Grades of causality (e.g. `possible’, `probable’, `definite’) offer little practical advantage. Only `related’ versus
`unrelated’ is needed for regulatory reporting requirements
„ Poor inter-rater agreement using terms such as `possible’ or `probable’
‟ Investigator should complete a checklist of potential causes
‟ Events should be considered related if there is “a reasonable possibility of a causal relationship” rather than if “a causal
relationship cannot be ruled out”

When should an event be considered `related’
Guidance
FDA (2010)
• Investigators/sponsors were too cautious in their interpretation of “reasonable possibility”. They submitted reports for events that
were likely to be manifestations of disease, common, probably unrelated events, or study endpoints.
• FDA received too many SUSAR reports for which there was no evidence of `relatedness’.
• Document provides guidance and examples to illustrate “evidence to suggest a causal relationship between the drug and the event”.
EMA guideline (2012)
• For regulatory reporting purposes, all spontaneous reports are considered suspected adverse reactions unless the reporters specifically
state they believe the events to be unrelated (page 5)

Naranjo algorithm at Roche
• Pre-2001
‟ Home-made algorithm with 6 categories
• 2001-2009
‟ Used to `triage’ cases for immediate physician review
‟ Events from study cases with score ≥5 `upgraded’ to `related’ for regulatory reporting
• 2009
‟ Naranjo withdrawn
‟ Physicians review all cases
‟ Events classified `unrelated’ or `related’ for regulatory reporting

Naranjo algorithm at Roche
• Survey of 55’000 cases scored over a 2 year period 2005-2007:
‟ Study cases: 43 causality `upgraded’ because score 5
‟ Spontaneous cases: 12 causality `upgraded’ because score 5
• We now assume `related’ for all spontaneously reported events unless reporter explicitly states `not related’
• Perceived limitations
‟ Little added value
‟ Time consuming (strictly, requires a review of previous cases)
‟ Information needed to assess probability of a causal relationship depends on the drug-event pair in question (i.e.
not the same for all DEPs)

Expressing causality in words
`Problem’ expressions
• “A causal association cannot be ruled-out”
‟ A causal association can never be ruled-out by an individual case.
• “It is possible that the event was caused by Drug X…”
‟ Possible has many shades of meaning.
• “Insufficient information for an adequate assessment”
‟ Often true, but for certain drug-event pairs, no additional information will help assess causality. ICSR is not
necessarily helpful for all drug-event pairs.
• “The case was confounded by…”
‟ Risk factors/alternative explanations are not necessarily confounders

Assessment of potential new adverse drug reactions
• Hypothesis (may be a case report, publication etc.) +/- review of other cases
• `Drug Safety Report’ (includes detailed review all relevant information)
• Presentation to the Drug Safety Committee
-> Decision concerning risk management (including label update etc.)
-> Request for additional information etc.

Combining evidence from different sources
Examples
• Mechanism of drug action
• Effects of similar molecules
• In vitro data
• Animal studies
• Pharmacokinetics
• Clinical trials
• Observational studies, claims databases etc.
• Individual case reports

Presenting information to the Drug Safety Committee
Bradford Hill criteria
Austin Bradford Hill. The Environment and Disease: Association or Causation Proc Royal Soc Med 1965; 58, 295-300

Single compelling case reports
• In some instances, one or two case reports can provide compelling evidence of causality, sufficient to change the
product label, risk management etc.
• Examples
‟ Progressive multifocal leukoencephalopathy
‟ Stevens Johnson syndrome (single case)
‟ Fatal anaphylaxis (single case)
• See also: Anecdotes that provide definitive evidence. Aronson JK, Hauben M. BMJ 2006;333;1267-1269

Conclusions
• Structured causality assessment methods for individual cases have been used to facilitate case processing rather than to
reach definitive conclusions
‟ Naranjo algorithm proved to be generally unhelpful
‟ Events are considered `related’ or `not related’ for the purpose of case processing and regulatory reporting
• Definitive decisions about causality (leading to RMP/label changes etc.) usually incorporate evidence from various
sources
‟ Structured approaches e.g. Bradford Hill criteria
‟ No single `once size fits all’ approach
• Individual case reports sometimes provide compelling evidence of causality

Final thoughts…
• Communicating risk (probabilities) to the patient and prescriber
‟ Drug attributable risk, effect of time, etc.
• Listening to the reporter
‟ “Why do you think the drug caused the event”

References
• Aronson JK, Hauben M. Anecdotes that provide definitive evidence. BMJ 2006; 333; 1267-9
• Bradford Hill A. The Environment and disease: association or causation Proc Royal Soc Med 1965; 58, 295-300
• Naranjo CA, Busto U, Sellers EM, Sandor P et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther 1981; 30 (2): 239‟45
• Taofikat B. Agbabiaka TB, Savovi J, Ernst E. Methods for causality assessment of adverse drug reactions. A systematic review. Drug Safety 2008; 31 (1): 21-37
• Guidelines for preparing core clinical safety information on drugs. 2 nd Ed. Report of CIOMS Working Groups III and IV. CIOMS Geneva 1999
• Management of safety information from clinical trials. Report of CIOMS Working Group VI. CIOMS Geneva 2005
• EMA: Guideline on good pharmacovigilance practices (GVP); Module IV ‟ Management and reporting of adverse reactions to medicinal products, 20 February 2012;
EMA/873138/2011
• FDA: Guidance for Industry and Investigators: Safety Reporting Requirements for INDs and BA/BE Studies. Sep. 2010 Drug Safety
• FDA: Federal Register/Vol75, No 188/Wednesday, September 29, 2010/Rules and Regulations p59945

Back-up slides, additional references etc.

Guidance
• Guidance for Industry and Investigators: Safety Reporting Requirements for INDs and BA/BE Studies. Sep. 2010 Drug Safety
‟ FDA found that investigators/sponsors were too cautious in their interpretation of “reasonable possibility”. They submitted reports for events
that were likely to be manifestations of disease, common, probably unrelated events, or study endpoints.
‟ FDA received too many SUSAR reports for which there was no evidence of `relatedness’.
• Federal Register/Vol75, No 188/Wednesday, September 29, 2010/Rules and Regulations p59945
‟ Examples to illustrate `evidence to suggest a causal relationship between the drug and the adverse event’ for reporting purposes
„ A single occurrence of an event that is uncommon and known to be strongly associated with drug exposure (e.g. angioedema,
hepatic injury, SJS)
„ One or more occurrences of an event that is not commonly associated with drug exposure, but is otherwise uncommon in the
population exposed to the drug (e.g. tendon rupture)
„ An aggregate analysis of specific events observed in trials (such as known consequences of the underlying disease or condition under
investigation or other events that commonly occur in the study population independent of drug therapy), that indicates those events
occur more frequently in the drug treatment group than in a concurrent or historic control group

Guidance
• CIOMS VI
‟ A CRF is never likely to contain all the fields needed to evaluate causality for all possible types of adverse event (p82)
‟ Pre-study training of investigators is very important (p82)
‟ Causality assessment for individual SAEs is more relevant for determining regulatory reporting status than for clinical analysis (p120)
• CIOMS III/V
‟ Recommends against expressions in the product label such as “The following have been reported, but a causal relationship has not been
established”, but recognizes some possible advantages to such an approach (p25)
‟ Good quality information is essential (p30)
‟ The threshold (strength of evidence required) may differ according to the nature and seriousness of the event (p30)
‟ No quantitative criteria (such as risk difference >5%) p56
‟ Investigators should always be strongly encouraged to express their opinion on what the cause of an event might be (p56)

Guidance
• ICH E2A
• ICH E2B
‟ The term `adverse reaction’ implies at least a reasonable possibility of a causal relationship between a suspected product and an adverse
event
‟ Spontaneously reported events should be considered `related’ for the purpose of regulatory reporting

Capturing the reporter’s assessment of causality
Proposal #1
Note: Probability categories derived from: Rockey DC, Seeff LB, Rochon J, Freston J et al. for the US Drug-Induced Liver Injury Network. Causality assessment in druginduced
liver injury using a structured expert opinion process: Comparison to the roussel-uclaf causality assessment method. Hepatology 2010; 51: 2117-26