A CASE FOR BINARY CAUSALITY IN CLINICAL TRIALS

October 29 2021 Valerie Jurasek

Related, Not Related, Likely, Possibly Related, Unlikely Related, Certain, Definitely, Unassessable, Unknown, Probably Related…if you recognize these terms, then there is a reasonable possibility you are familiar with what the Drug Safety and Pharmacovigilance world refers to as the causality assessment.

For decades, Sponsors have grappled with clearly defining the reporting options. Could ‘unknown’ mean ‘possibly’ or ‘unassessable?’ Does anyone ever select ‘certain’ when they are not certain of the treatment received in a blinded study or are not certain whether concomitant medications could be the cause of the event? A recent trend in Drug Safety is to simplify the options…is there a reasonable possibility the event is related to the suspect product, Yes or No? This approach is referred to as “binary causality” and is the current recommendation of the CIOMS VI working group.

Causality assessment in clinical trials is a critical component in determining reportability of adverse events, detecting trends and identifying safety issues. The causality assessment seeks to determine the relationship between the suspect product (i.e., drug, device, vaccine, biologic), and an adverse event. Determining whether an event was caused by a product depends on many factors, and may result in varying levels of certainty. Investigators and Sponsors must consider the time between study drug administration and onset of the event, preclinical data, the half-life of the product, underlying conditions, concomitant medications, medical history and other factors to make a causality determination.

The goal of adopting binary causality is to increase consistency across all products and Investigators. The idea is not to over-simplify the assessment of the event’s relationship to the product, but simplify the options from which the Investigators choose when reporting causality. For regulatory reporting purposes, the only two options available are Related or Not Related.  Whether the event was reported by the Investigator as Related, Probably Related or Possibly Related, it does not make a difference in regulatory reporting; these are all reported as Related. The binary causality recommendation introduces consistency between the original safety event report and the eventual end reporting to regulatory authorities by stripping out unnecessary subcategories.

“The goal of adopting binary causality is to increase consistency across all products and Investigators.”

Share This Article

A Case for Binary Causality:

  • Increase repeatability in reporting from Investigators
    • Different Investigators might assess the same situation as Possibly Related or Unlikely Related. Assessments are not always repeatable.
    • According to a 2005 report from the CIOMS VI working group there is no consensus among stakeholders in the industry regarding the definitions of the common terms used today, such as probably vs possibly vs likely.
  • Ensure assessments by safety case processors are consistent
    • If Unlikely related is reported by the Investigator, one case processor may interpret that as a reasonable possibility, and another may interpret that as not a reasonable possibility. This results in subjective variability among case processors that is less based on facts and more on individual feelings. 
  • Increase consistency across studies
    • One study might provide Investigators with the options: Related, Unrelated and Possible. Another may provide the options of Definitely Related, Possibly Related, Unlikely Related, and Not Related. There can be any combination of causality options for the Investigator to select depending on the Sponsor and the study. This issue becomes more complex when Sponsors don’t work with the same safety vendors in each study across a program, resulting in variance within the Sponsor safety database. The Sponsor ends up needing to create mapping algorithms between studies to create a unified reporting structure after the fact.
  • Encourage Investigators to provide a very thorough assessment
    • When given the option of Possibly Related, an Investigator might choose that option if an alternative etiology is not apparent and causality cannot be ruled out. When given the options for binary causality, the assessment is more likely to answer the question “is there a reasonable possibility” and not “can the relationship be ruled out?” Per the FDA, a “reasonable possibility” means there is evidence to suggest a causal relationship between the drug and the adverse event. Binary causality helps Investigators answer the right question (Naidu, R.P., 2013).

A common practice in clinical trial drug safety reporting has been to apply a level of how likely it is that the product caused the adverse event, resulting in inconsistent, endless options that are not always easy to interpret and carry different meanings among different Investigators and Sponsors. The causality determination already depends on the amount of data available and somewhat subjective opinion of a trained medical professional making the assessment, so taking the ambiguity out of available options for causality seems like a smart move.

With a portfolio containing many clinical trials using a variable causality assessment model, it may seem cumbersome to re-configure and re-train in order to facilitate a new binary causality approach. As a full service clinical research organization (CRO), PROMETRIKA enables consistent and clear communication among data management, pharmacovigilance and clinical monitoring which would ensure a smooth transition to this approach in the electronic data capture (EDC) system, SAE reporting forms and instructions, and site training.

Is your company ready to take the plunge into a binary causality approach – Yes or No?

References

CIOMS (2005). Management of safety information from clinical trials. Report of CIOMS Working Group VI.

Naidu, R. P. (2013). Causality assessment: A brief insight into practices in pharmaceutical industry. Perspectives in Clinical Research, 4(4), 233–236. https://doi.org/10.4103/2229-3485.120173