Michael R. Goldberg’s Commentary

Jose de Leon: Training Psychiatrists to Think Like Pharmacologists

23.Valproate Case 2: Safety


            The purpose of this commentary is to provide a perspective on the Valproate Case 2: Safety presentation (posted on 2-12-16) by Dr. Jose de Leon.  This review is based on my experience as a safety physician and clinical pharmacologist in the US pharmaceutical industry.  Generally, I will not focus on the clinical profile of valproate, except in passing.  Rather, my focus is on the data presentation and key concepts in the evaluation of the safety profile of a pharmaceutical product (AbbVie Inc., 2017; Sanofi, 2016).  From a pharmaceutical industry medical perspective, the Package Circular and Summary of Product Characteristics (SPC) are the most appropriate source documents for information on a product as they are prepared by a drug’s sponsor from registration trials, updated based on post-marketing reports, including the medical literature and approved by regulatory agencies. These are the legal data sources to be used by a practicing physician to inform a patient on risks of treatment, when prescribing a product. 

            My perspective is that communication of product information between physicians (especially in an educational setting) should be consistent with product labeling.  In general, the experience of an individual physician is unlikely to be greater or more accurate than the prescribing information, except when an unusual finding is observed.  The case described by Dr. de Leon may be an example of such a finding, although careful review of product labeling does contain information consistent with this report.

            Preliminarily, it is unclear what “Case 2” in the title references.  On my first read of the presentation, I thought there were two cases of eosinophilic pleural effusion.  On further review, a second pleural effusion case was not described.

            The educational objectives which are presented are not discussed to a major degree, except in the circumstance of the unusual pleural reaction.  Since treatment of the primary indications for valproate often involves multiple drugs, a more extensive discussion of polypharmacy and the challenge of interpretation of safety observations would be an important lead-in.

            Although the labels of drugs often consider adverse “reactions," the events tabulated in the slide set are crude adverse event rates from clinical trials.  The distinction between an adverse event and an adverse reaction is important.  Thus, an adverse event is described in the United Sates Food and Drug Administration (US FDA) regulations (CFR 21, 2016) as “any untoward medical occurrence associated with the use of a drug in humans, whether or not considered drug related.”  In contrast, an adverse reaction (EMA) is defined as: “A response to a medicinal product which is noxious and unintended. Response in this context means that a causal relationship between a medicinal product and an adverse event is at least a reasonable possibility.”  These differences, while subtle, depend on physician interpretation of the event and consideration of the difference between treatment-associated, such as event rates in clinical trials, for the most part, and treatment-related, which requires an assessment of relatedness.  In the conduct of blinded clinical trials, investigators are required to report events, whether or not considered treatment related.  For each event, both the investigator and sponsor make an assessment as to whether the event is definitely not, possibly, probably, or definitely related to treatment, without knowing the treatment assignment.  For post-marketing reports, a relationship is implied in the act of reporting, as indicated in the label.

            For post-marketing reports, it is the responsibility of the sponsor to assess events, individually, if serious, and/or in aggregate, for the strength of an association with the treatment, to be followed by appropriate updates of the prescribing information.  In the presentation, Dr. de Leon is considering the reports of eosinophilic pleural effusion as related to valproate (possibly).  However, it should be noted that the US 2017 prescribing information does not include pleural effusion as a post-marketing reaction, and the EU Summary of Product Characteristics (SPC) does (Sanofi, 2016).  Further, the US prescribing (AbbVie, 2017) information describes the association of an eosinophilic syndrome with the use of valproate as summarized below:

Drug Reaction with Eosinophilia and Systemic Symptoms (DRESS)/

Multiorgan Hypersensitivity Reactions

Drug Reaction with Eosinophilia and Systemic Symptoms (DRESS), also known as Multiorgan Hypersensitivity Reaction, has been reported in patients taking valproate. DRESS may be fatal or life threatening.  DRESS typically, although not exclusively, presents with fever, rash, and/or lymphadenopathy, in association with other organ system involvement, such as hepatitis, nephritis, hematological abnormalities, myocarditis, or myositis sometimes resembling an acute viral infection. Eosinophilia is often present. Because this disorder is variable in its expression, other organ systems not noted here may be involved. It is important to note that early manifestations of hypersensitivity, such as fever or lymphadenopathy, may be present even though rash is not evident. If such signs or symptoms are present, the patient should be evaluated immediately. Valproate should be discontinued and not be resumed if an alternative etiology for the signs or symptoms cannot be established.

            The purpose of defining the safety profile of a drug product is to provide information to the practicing physician regarding potential risks of that product.  The primary source of such information should be the package circular and not a source of unclear relationship to the approved information (Lexi-Comp Inc., 2012).  In the circular, the data from controlled clinical trials are presented with placebo or a control treatment as reference to provide perspective on the experience of patients in the trial.  Event rates for valproate in the US circular are reported by indication and not generally, as in the presentation.  In the presentation, adverse event reporting rates are reported without the control group rates; these rates are greater than could be attributed to valproate, alone (unless the control rate were to be zero).  Reporting rates across indications are combined without considering that the dose could differ for different indications and provide different profiles of the drug.  The potential for treatment-by-indication interaction must also be considered when aggregating rates across indications.

            For example, the cited study in children and adolescents regarding suicidality appears to be from an uncontrolled study.  In none of the other data cited does the rate of suicidality rise to the level observed in that experience.  Furthermore, examination of the labeling statement may be an example of class labeling for antiepileptic drugs and may not necessarily reference valproate, specifically.  Such statements may be included by regulatory agencies, even when an actual event is not necessarily reported for a product.  A similar situation applies to the incorporation of suicidality as a risk for all antidepressants in children and adolescents, independent of mechanism and specific observations with any drug.  Indeed, in depression, it has been considered that the risk of suicide increases as a patient’s energy level increases to the point of taking some action at a time before antidepressant activity has manifested, and not necessarily as mechanism-based suicidal ideation.

            In discussing the efficacy of valproate, the core indications are summarized.  However, Section 2.2 of the presentation sites a book chapter covering the use of the drug in “adults with intellectual disabilities.”  This is not an indication described in the product labeling.  For a general discussion, as intended, I would not separately present such an indication except in the section on unlabeled indications.

            The presentation of the safety of valproate is challenging.  Critically, all the presentations of event rates for common events are presented only in relation to the reporting of events in those trial patients receiving valproate.  This is an incomplete presentation and cannot be interpreted as describing the safety profile of the drug.  Rather, these rates require the fair balance of the comparison with a control, especially placebo, with the difference between valproate and placebo as an index of attributable risk for an event during treatment for a particular indication.  These rates may reflect identification from the source reference, or selection of only the valproate rates from the prescribing information.

            Certain rare events such as osteoporosis, Parkinsonism and others may represent isolated post-marketing reports to an agency or the manufacturer, or published case reports.  A perspective on the relationship of these to treatment with valproate would be useful.

            Several events are labeled in the prescribing information as Warnings and Precautions, and discussed as potentially lethal events.  I leave the assessment of these as potential risks to those who prescribe the drug.  It is fair to emphasize these in the presentation; however, there are Warnings and Precautions which are not discussed.  These include such as eosinophilic syndrome (which could be related to the pleural effusion case report), interaction with carbapenem antibiotics, hypothermia and use in patients with mitochondrial disorders.  A reference to the prescribing information for additional safety information would have been useful and would indicate that those potential harms presented were judged to be the most significant by the presenter.  A “black box warning” for life threatening adverse reactions is included in the US circular for hepatotoxicity (generally and for patients with mitochondrial disease), fetal risk and pancreatitis.


            Dr. de Leon describes a case of eosinophilic pleural effusion in a patient with schizophrenia and receiving multiple medications, with the last added treatment being valproate, off-label, for “remaining symptoms.”  The patient develops pulmonary symptoms and is found to have an effusion.  After discontinuation of valproate, the effusion resolves.  A lengthy discussion follows which leads to the conclusion that the effusion is related to treatment with valproate.  Note that the above cited DRESS associated with valproate is not referenced, but should be.

            In considering the case, Dr. de Leon completed a literature search and identified 18 case reports of pleural effusion with valproate, 123 reports of drug-induced pleural effusion and 30 review articles.  It is unclear how many were personally reviewed by Dr. de Leon beyond one case report and one review.  The review article was chosen subjectively (It “appears to be a good review article” because it mentions valproate and was published in in a pulmonary journal [which may or may not indicate whether the authors <s>to</s> have special qualifications].)  A second “literature review” is included in a two-page case report of eosinophilic effusion in a patient receiving valproic acid; its quality cannot be determined. 

            A comparative tabulation of the reported cases by clinical profile, or individually would have been useful and would be necessary to include eosinophilic pleural effusion in the prescribing information.  Inclusion could be either as an individual post-marketing event, in the discussion of DRESS, or both.  In my industry experience, a detailed summarization of the published reports would have been appropriate and could have resulted in a specific label update.

            A complicated discussion of the hypothesis that the effusion was related to a metabolic drug interaction with clozapine follows.  No conclusions can be drawn from this discussion except that 1) a drug interaction is a possible mechanism of the effusion and 2) Dr. de Leon has expertise on the metabolism of clozapine.  The case represents a complex situation of polypharmacy such that clozapine, or interaction with it, may not be the only alternative cause.

            The final section of the presentation addresses the assessment of causality for a potential adverse drug reaction.  Assessment of causality for an individual adverse drug event is always difficult.  Indeed, FDA only goes as far as “suggesting” a causal relationship in its Pharmacovigilance Guidance.  My personal experience is that the only criteria that can be met for a single case are temporal-present during or in proximity to treatment, positive dechallenge and positive rechallenge.  However, such observations are at risk of being coincidental.  It is only in aggregate and applying the features listed below that an association can be suggested.  It is the strength of this association which allows for label updates, not necessarily with demonstration of an absolute cause-effect relationship.

            In assessing case reports, FDA (U.S. Department of Health and Human Services Food and Drug Administration, 2005) recommends that sponsors look for features that may suggest a causal relationship between the use of a product and the adverse event, including:

1.     Occurrence of the adverse event in the expected time (e.g., type 1 allergic reactions occurring within days of therapy, cancers developing after years of therapy);

2.     Absence of symptoms related to the event prior to exposure;

3.     Evidence of positive dechallenge or positive rechallenge;

4.     Consistency of the event with the established pharmacological/toxicological effects of the product, or for vaccines, consistency with established infectious or immunologic mechanisms of injury;

5.     Consistency of the event with the known effects of other products in the class;

6.     Existence of other supporting evidence from preclinical studies, clinical trials, and/or pharmacoepidemiologic studies; and

7.     Absence of alternative explanations for the event (e.g., no concomitant medications that could contribute to the event; no co- or pre-morbid medical conditions).

            With respect to causality determinations, if there were criteria which left no doubt about causality, agencies would be applying these consistently and universally.  Thus, consider the Naranjo et al. algorithm which has been applied by some since 1981.  While aspects of the algorithm correspond to the suggestive features indicated above, such a scoring system has not been captured in regulation.  In Dr. de Leon’s discussion, he has pulled only those criteria from the algorithm that relate to his case; it would have been useful to present the other criteria.  The important criteria of positive rechallenge and dose relationship, which could be applied to the multiple reports, in aggregate, as further suggestion of a causal relationship, are not discussed.  Various methods have been used to mine safety data for potentially causal associations and applied to a range of data bases to determine the strength of association between certain adverse events and drug treatment (Lehman et al., 2007; Strom et al., 2013).

            In conclusion, Dr. de Leon’s presentation is ambitious in its scope, addressing both core information on valproate and an interesting case report.  From this pharmaceutical physician’s perspective, there are short-comings in both the presentation of core safety information and interpretation of the case report considering the core safety information.  These touch on methods for use of controlled trials to assemble the core safety information, challenges associated with evaluation of a drug for multiple indications at multiple doses, the understanding of pharmacokinetic and pharmacodynamic drug-drug interactions, assessment of drug-disease interactions, and the use of post-marketing safety data.  Also raised are questions related to the most appropriate sources of drug information, causality assessments for a particular adverse event and use of a product for unlabeled indications. 



AbbVie, Inc.  Depakote Full Prescribing Information.  March, 2017 [http://www.rxabbvie.com/pdf/depakote.pdf , accessed 21-June, 2017].

Code of Federal Regulations][Title 21, Volume 5][Revised as of April 1, 2016][CITE: 21CFR314.80] [https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch.cfm?fr=314.80 , accessed 21-June-2017].

Drug Information Handbook, 2012-2013:  A comprehensive resource for all clinicians and healthcare professionals.  Lexi-Comp Inc. (COR), 2012.

Lehman HP, Chen J, Gould AL, Kassekert R, Beninger PR, Carney R, Goldberg M, Goss MA, Kidos K, Sharrar RG, Shields K, Sweet A, Wiholm BE, Honig PK. An evaluation of computer-aided disproportionality analysis for post-marketing signal detection. Clin Pharmacol Ther. 2007; 82:173-80. Epub 2007 May 16.

Naranjo CA , Busto U, Sellers EM, Sandor P, Ruiz I, Roberts EA, Janecek E, Domecq C, Greenblatt  DJ. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther 1981; 30: 239-45.

Sanofi, Inc. Depakote 250mg Tablets Summary of Product Characteristics Updated 09Dec2016. [http://www.medicines.org.uk/emc/medicine/25929, accessed 21-June, 2017].

Strom BL, Kimmel SE, Hennessy S.  Textbook of Pharmacoepidemiology, 2nd edition.  Oxford: Wiley-Blackwell; 2013.

U.S. Department of Health and Human Services Food and Drug Administration, Guidance for Industry:  Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment, 2005.


July 27, 2017