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)

Charles M. Beasley, Jr’s reply to Daniel Kanofsky’s comments

 

       We appreciate Dr. Kanofsky’s comment (Kanofsky 2020) on our work (Beasley and Tamura 2019) and are gratified to see that it has stimulated a call for data that should be easily available and would be of some aid to physicians in making important risk-benefit decisions regarding patient care.

       We consider schizophrenia to be a critical illness for which clozapine can be a life-saving treatment.  However, as pointed out by Dr. Kanofsky, highly infrequent (approaching only one case in 1,000 treated patients) to rare (≤ one case in 1,000 treated patients) and very rare (< one case in 10,000 treated patients) cases of acute renal failure and inflammatory disorders (colitis, pancreatitis, pericarditis, myocarditis, among others) have been reported during treatment with clozapine.  Given the infrequent to very rare occurrences (observations and reports) of these disorders in temporal association with clozapine treatment, it cannot be said with conventional statistical standards that such cases are adverse drug reactions (ADRs) to clozapine or simply coincidental cases not due to or influenced by clozapine and therefore simply adverse events (AEs) observed during clozapine treatment.  To complicate clinical care further, as pointed out again by Dr. Kanofsky, the occurrence of these disorders might be facilitated by clozapine, but clozapine treatment by itself does not directly cause the disorders.  Concomitant treatment with clozapine and other drugs might be required to induce what still might be an infrequent to very rare ADR in patients treated with the relevant combinations.

       Readers will have their own opinions about whether these cases are ADRs to clozapine, ADRs to clozapine plus another drug or AEs not related to clozapine or clozapine in combination with an additional, single medication or medications.

       Consistent with the medical dictum of “first, do no harm,” discontinuing clozapine in the face of one of these reactions and not reintroducing clozapine as a treatment is good medical practice.  However, there is the uncertainty about whether these serious and potentially fatal events are ADRs to clozapine (or a clozapine combination) or AEs with no relationship to clozapine other than temporal co-occurrence of treatment with clozapine and the observation of the medical disorder. Given the gravity of schizophrenia, especially in some specific patients with the disorder, careful consideration of the individual patient’s risk-benefit ratio in receiving subsequent treatment with clozapine favors re-treatment with clozapine and some patients have received such treatment.

       Such patients constitute “rechallenge cases.”  We have discussed the use of the rechallenge paradigm and the more formal “N-of-One” experimental design (Beasley 2020) in response to Shorter (2019) that had not appeared in INHN at the time Dr. Kanofsky provided his comment (Kanofsky 2020).  As we pointed out, there are limitations to how the outcome of a single rechallenge can be interpreted.  However, such data, even with interpretive limitations and potential for supporting incorrect beliefs about the safety of a rechallenge, are likely more useful than the absence of such data when making critical treatment decisions.

       At the very least, with a sufficient number of rechallenge cases with their outcomes reported for a specific medical disorder (disorder recurred or did not recur on rechallenge), the binomial probability of the ratio of non-recurrence to recurrence could be computed.  The following example illustrates the potential utility of only a relatively small number of cases.

       Assume that the probability non-recurrence of myocarditis on rechallenge with clozapine is 0.50 (the outcome is binary [recurrence or non-recurrence], and non-recurrence is a completely random event with a probability similar to that of getting a “heads” on a single coin toss). With 20 reported rechallenge cases, non-recurrence is observed in 14 cases.  Then, the probability of this outcome is 0.037 (https://www.thecalculator.co/math/Binomial-Calculator-741.html).  Believing that the observed AE is not an ADR, it would be reasonable to assign a higher probability to non-recurrence with each rechallenge case.  Assigning a probability of 0.80 to non-recurrence with the same 14-to-6 non-recurrence to recurrence ratio, the probability of that outcome is 0.11.  With the belief that the outcome is an ADR and assigning a probability of 0.20 to non-recurrence, the probability of observing the same 14-to-6 non-recurrence to recurrence ratio is 2.0x10-6.  More sophisticated analytical models could likely be employed that would consider potential confounding factors and mediating variables if the data were available for the cases.  Without a comparative group that would include non-treatment episodes in “N-of-One” designs, the assumptions about the probability of observing non-recurrence in an individual rechallenge in a series of such cases are quite an important consideration in the application of numerical methods to the assessment and then the interpretation of such case series data.

       While such information is far from rigorously definitive based on what we have previously described as definitive by conventional standards and the caveats regarding open rechallenge compared to formal “N-of-One” experiments, we believe such data could be immensely helpful in real-world clinical treatment of schizophrenia with what might well be the most powerful treatment option on a population basis currently available.  Notably, myocarditis is not a symptom where a variety of factors might influence its reporting.  While its diagnosis might be missed or it might be diagnosed when not present, these possibilities are much less likely than an inaccurate Positive and Negative Syndrome Scale (PANSS) Total score (Kay, Fiszbein and Opler 1987) due to a poor rating interview without good patient-rater rapport.   The utility of such single rechallenge case report data is highly dependent on the extent and quality of the available data.  However, our example demonstrates that even a small number of cases can provide highly useful guidance if there is a substantial preponderance of either recurrences or non-recurrences.

       Dr. Kanofsky has pointed out that all patients treated with clozapine in the US are registered in a Risk Evaluation and Mitigation Strategy (REMS) related database.  Physicians who treat with clozapine and do rechallenge patients with the medication following the occurrence of serious medical disorders should be strongly encouraged to report these results in the medical literature.  However, it should be kept in mind that such work would be another “unfunded activity” in the life of many very busy, and in the view of some, under-compensated physicians, perhaps under the pressure of “performance quotas.”  With clozapine available generically, there is no single manufacturer of the medication highly vested in extending the knowledge of both its positive and negative clinical potentials.

 

References:

Beasley CM, Tamura R.  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.

Beasley CM.  Reply to Edward Shorter’s Comments on 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. August 15, 2019.

Beasley CM. Charles Beasley’s additional reply to Edward Shorter’s comments on his “introductory comments” on 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. May 21, 2020.

Kay SR, Fiszbein A, Opler LA. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr Bull. 1987; 13(2): 261-76.

Shorter E.  Comments on 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.  May 9, 2019.

 

June 4, 2020