Charles M. Beasley, Jr and Roy Tamura: What We Know and Do Not Know by Conventional Statistical Standards About Whether a Drug Does or Does Not Cause a Specific Side Effect (Adverse Drug Reaction)Overview

(Full Text)

Charles M. Beasley’s response to Hector Warnes’ response to Charles Beasley’s reply to his comment

        We strongly agree with Hector Warnes’ admonition that whether an individual patient experiences good therapeutic benefit and/or one or more adverse drug reactions (ADRs) to a medication is highly dependent on the combination of multiple intrinsic factors (e.g., genetically determined processes of drug metabolism) and multiple extrinsic factors (e.g., concomitant medications taken by the patient).  Not every patient who takes an SSRI antidepressant to treat a well-diagnosed episode of Major Depressive Disorder will experience remission and not every patient will experience nausea (probably the most common ADR associated with the SSRI class and the ADR with the most robust data supporting the hypothesis that the adverse event (AE) is an ADR).  Likewise, not every patient treated with a second-generation antipsychotic will experience substantial improvement in psychotic symptoms and not every patient will experience weight gain.  We did not discuss this critical point in our original commentary (Beasley and Tamura 2019a,b) and we appreciate Hector emphasizing this point. 

        The dependency of experiencing an ADR on the combination of a large number of factors can give rise to a seemingly paradoxical situation for some drug and AE combinations where the AE might be an ADR.  It is a logical possibility that some ADRs are so rare that on a population basis, they cannot be demonstrated to be ADRs.  However, an individual can be affected by a unique, complex set of these intrinsic and extrinsic factors such that an AE of interest is an ADR for an individual patient.  In an extreme case, there might even be robust evidence on a population basis that the AE of interest is not an ADR for the drug of interest.  What follows below is a hypothetical example of such a phenomenon.

        Let us say 100 million persons are treated with a drug from the time it is first approved until newer agents replace it and its original developer and generic manufacturers no longer manufacture the drug.  During the life of that drug, one patient experiences an AE, which for that individual is an ADR.  Out of that 100 million treated, 1,000 experience the AE, but for the 999 other patients, the drug has no contribution to the cause of the AE.  In such a case, the totality of data would argue against the AE being an ADR (total incidence:  1,000/100,000,000 = 0.001%) if the background incidence of the AE is considerably greater than 0.001%.  Although the data would not support the hypothesis that the AE was an ADR, these data would not rise to the level of being robust evidence that the AE was not an ADR depending on the estimated background incidence of the AE in general population that might take the drug.

        We can carry this hypothetical example even further concerning the result of a paradox.  Let us say that for the one patient who experienced the AE and for whom it was an ADR, the ADR nature of the AE was confirmed by an “N-of-One” experiment.  Let us further say that the AE is QT prolongation leading to frank TdP.  A positive “N-of-One” experiment would establish the AE as an ADR for the individual patient with high-quality evidence.  One well-confirmed case is enough to establish the AE as an ADR associated with the drug.  However, for this one patient, the occurrence of the ADR was dependent on the patient’s highly unusual milieu of intrinsic and extrinsic influences.  For this patient, while the drug was necessary to cause the ADR, the drug alone was not sufficient to cause the ADR.

        Continuing with QT prolongation (that might lead to Torsade de Pointes) as the AE of interest, a high-quality TQT study has demonstrated that the drug does not increase QT but results in a very slight decrease in QT length in this hypothetical scenario.  Here then we would have an AE proven to be an ADR in one specific patient, but evidence of equal quality proving that the ADR is highly unlikely to occur in the general population  This situation might or might not apply to some drugs.  We would never know. 

        Beasley is not aware of such extreme examples as that above, where robust evidence supported both the occurrence of an ADR in an individual or a very small number of individuals and equivalently robust evidence from alternative experimental designs argued against the occurrence of the AE as an ADR in a population.  However, Beasley is aware of instances where there was some credible evidence of AE causation by a drug in one individual but at least comparable evidence of lack of causation of the AE by the drug on a population basis.  This potential has profound but somewhat philosophical implications for product labeling, AE reporting, and product liability litigation matters.

References:

Beasley CM, Tamura R.  Full Text (Charles M. Beasley, Jr., and Roy Tamura: What We Know and Do Not Know by Conventional Statistical Standards About Whether a Drug Does or Does Not Cause a Specific Side Effect [Adverse Drug Reaction]). inhn.org.ebooks.  November 21, 2019 (2019a).

Beasley CM. A Post-Script (Charles Beasley and Roy Tamura: What We Know and Do Not Know by Conventional Statistical Standards About Whether a Drug Does or Does Not Cause a Specific Side Effect [Adverse Drug Reaction]). inhn.org.ebooks.  October 24, 2019 (2019b).

June 11, 2020