You are here: Home / Courses / Jose de Leon: Training psychiatrists to think like pharmacologists / 23. Valproate Case 2: Safety / Michael Goldberg’s response to Jose de Leon’s reply
Wednesday, 28.10.2020

Jose de Leon: Training psychiatrists to think like pharmacologists
Valproate Case 2 Safety

Michael Goldberg’s response to Jose de Leon’s reply


            Dr Warnes comments on appropriate evaluation of adverse events are quite logical and expand on my comments regarding methods to assess causality and evaluate event mechanisms.  I agree with Dr. de Leon's assessment of them.

            Much of Dr de Leon’s commentary centers on the oft cited challenges to using industry data as a source of valid information--obviously our opinions differ on the reliability of such data, which is a point to mention as physicians consider the experiences of their patients and the science of randomized, controlled trials.  I would reference, also, Charles Beasley’s correspondence regarding Blackwell's commentaries on the nature of information provided by the pharmaceutical industry.  In aggregate, in most cases, an ethical Company has the advantage of having controlled and verified data on hundreds to thousands of patients. I suspect these vastly exceed reports from an individual physician (which are often resistant to follow up inquiries or verified by chart review). Though there are event databases (WHO, FDA, etc) which, while not controlled, may provide useful information if appropriate disproportionality analyses are applied; such reports often overlap with Company reports.  Further, patients experiencing an event in a trial are under control of the investigator who can further evaluate and/or respond to Company suggestions regarding ancillary evaluations to better understand mechanisms and relatedness.

            Dr de Leon's comments completely overlook the differentiation of adverse events from adverse reactions and their respective reporting in prescribing information and elsewhere.  In the past, at least, there have been differences between reporting requirements with some venues requiring report of events (as in the tabulations in the valproate US label) vs reactions (more prominent in EU SPCs).  This is a critical learning point for any audience of an analysis of drug safety reports.

             My biggest concern about Dr de Leon’s presntation is the use in an educational presentation of data from controlled clinical trials without presenting the data from the control group, or citing information on indication(s) studied, concomitant treatments, patient characteristics, dose, or duration of treatment, which, with other data allow for interpretation of these rates.  Additionally, presentation of data from small studies of unapproved indications is flawed in its anecdotal essence. Frankly, these approaches represent biased data presentation of the highest order and make any references to deficiencies of industry data pale in comparison.  Data presentations in an educational setting should be completely presented, balanced with limitations of the data emphasized, as apparent in a table in the US prescribing information showing events for both drug and control treatments.     

            Regarding alternative presentations of patient data by Companies, in many venues, an approved patient package insert would serve this purpose of providing “highlights” without short-changing the full presentation of data for the prescriber as agreed in the approval process. Warnings and precautions are required in advertising, at least in the US.  A highlight as the only presentation would necessarily be incomplete and leave the Company open to criticism of the method of selection of "highlights" and the lack of complete disclosure, which a prescriber should have available; based on his or her patient population.  The Company can do no less than complete disclosure subject to regulatory review of trial data, although.  It is acknowledged that, in agreement with reviewing authorities, data for those events reported by a meaningful percentage of patients (e.g., 2, 5 percent) including those with counts greater than in the control group often represent a cut point for presentation in labeling to infer events of significance.                  

            Regarding peer reviewed publications, my experience is that Companies are required to review these promptly and to adjust labeling based on the quality of the published data in individual and aggregated publication.  Published reports are carefully analyzed, including visits to authors, as permitted for chart review and assessment of the data. Necessarily, events such as toxic epidermal necrolysis and acute hepatic failure are, by consensus, deserving of inclusion if there is even a single well-documented report.               

            What to include in a label based on post marketing reports is also subject to careful review for a cluster of similar events, or single events of reportable significance (above).  Reviews should take all information into account, with careful analysis of the nature of events and their number.  As an example of a suspect event from my past:  I have worked on review of data from a highly prescribed drug ("X"), available in oral and intravenous form.  Over a period of years, the Company regularly received reports of an excess of thrombocytopenia associated with administration of X.  Analysis of these reports was that they were all from a single country, despite the world-wide use of the drug.  In this particular country, the drug was used extensively in critically ill patients receiving a large number of supportive treatments and medications in an intensive care situation where spontaneous thrombocytopenia would not be unexpected, independent of X. NO reports of X-associated thrombocytopenia were ever reported from other countries.  Multiple twice-a year reviews of these reports by 3 physicians and their management, and label review teams all concluded that inclusion of thrombocytopenia in the label of this product would be misleading and not representative of the clinical profile of X.  Even if a population based metabolic or other susceptibility factor were at play for this event, it defies credulity that absolutely no such events were reported from other venues in which residents from the country of interest would be likely to receive X by chance.      

            I recommend to Dr de Leon review of Cobert's Manual of Drug Safety and Pharmacovigilance (Second Edition (2012); Barton Cobert, MD; Jones and Bartlett Learning; Sudbury, MA) which summarizes well industry standards and methods for management of drug safety data.


March 7, 2019