Commentary: Musculoskeletal adverse events in dogs receiving bedinvetmab (Librela)

Anthony Simon^*Beatriz MonteiroOliver KneslAdam Werts

• Zoetis, Parsippany, United States

The article "Musculoskeletal adverse events in dogs receiving bedinvetmab (Librela)" by Farrell et al. 1 reports an increased reporting rate of musculoskeletal adverse events (AE) in dogs treated with bedinvetmab. It is divided into three parts: a descriptive disproportionality analysis, a review of purported data entry errors and a case series of 19 clinical cases.Bedinvetmab is a medication approved for the alleviation/treatment/control of OA-related pain in dogs. Zoetis welcomes researchers' interest in the safety of bedinvetmab and thanks the authors for raising these concerns. Zoetis does not exclude the possibility of any AE. Adverse events reported in the study are taken seriously and Zoetis continues to assess these and other reported AEs.The present commentary highlights concerns related to the case series (section 1) and methodology of the disproportionality analysis (section 2), addresses allegations made regarding data entry errors (section 3) and corrects two other identified errors (section 4). We would like to thank Farrell at al. for publishing the case series and reporting them to Zoetis or local regulatory agencies.Farrell et al. describe their study as a "case-control study and case series analysis" aimed at investigating a potential association between bedinvetmab administration and rapidly progressive osteoarthritis (RPOA). The publication discusses 19 cases of suspected joint-related AEs in dogs treated with bedinvetmab. These cases were selected by nine clinicians based on "evidence to support a causal relationship" and assessed by an adjudication panel using a subjective three-tiered system ('very suspicious'; 'suspicious'; 'insufficient evidence'). The study, however, lacks a true case-control design and scientific rigor in case selection and evaluation. Including cases from dogs with other joint diseases and chronic injuries, and blinding reviewers to treatment status, would have helped to reduce biases. Additionally, there is no control for baseline OA severity or comorbidities, and no discussion of the history of non-steroidal antiinflammatory drug (NSAID) use (14/19 cases). Long-term NSAID use in humans is associated with accelerated OA progression 2 and RPOA. 3 The cases presented by Farrell et al., though severe, are heterogeneous, lack biological similarity, and do not consistently resemble human RPOA. Human RPOA is distinct from fast-progressing OA. It is characterized by significant radiographic joint space narrowing or subchondral bone collapse in a short period of time (6 months to 1 year) 4 and features atrophic (non-bone forming) destructive arthropathy. 11,12 RPOA lacks established veterinary diagnostic criteria or pathophysiologic descriptors and has not been previously reported in dogs. In humans, OA progression varies with individuals showing 'stable, 'slow' or 'fast' progression. 5 Limited data suggest similar variability in dogs, influenced by various factors; these studies did not involve bedinvetmab. 6,7 Target animal safety studies for bedinvetmab were undertaken whereby laboratory animals received 7 monthly doses at 1X, 3X, and 10X the recommended dose. Joint risk assessments included pre-study and end-of-study radiographs of major joints, and extensive histopathological evaluation of bones and joints beyond the scope of traditional preclinical toxicology assessments. No joint risk was identified. The latter was conducted according to Good Laboratory Practices, 8,9 which are the highest regulatory laboratory safety standards globally, including auditing of the data. Nevertheless, regulatory agencies (European Medicines Agency (EMA); U.S. Food and Drug Administration (FDA)) required bedinvetmab labels to mention human RPOA as a precautionary statement, while acknowledging that this condition had not been reported in dogs. 10 Based on the available evidence, the cases do not support the existence of a canine equivalent of human RPOA, or a specific clinical syndrome associated with bedinvetmab administration. Disproportionality analysis is a well-defined methodology 13 used for signal detection in large databases (e.g. Marketing Authorization Holder (MAH) databases or regulatory agencies such as EudraVigilance Veterinary (EVV)) 14 to help identify AEs with a higher-than-expected reporting frequency. 15 Disproportionality analysis most commonly involves calculation of Proportional Reporting Ratio (PRR) or Reporting Odds Ratio (ROR). Both these approaches are frequentist statistics that are inverses of each other and calculated using a 2x2 contingency table. The PRR compares the proportion (frequency) of reports for a specific AE (or clinical sign at the Veterinary Dictionary for Drug Related Affairs (VeDDRA) Preferred Term level) associated with the drug of interest to the proportion of reports for the same AE associated with all other drugs from a reference database. A PRR greater than 1 suggests that the AE is reported more frequently for the drug of interest compared to other drugs (or has higher odds in the case of ROR). 16 Thus, disproportionality analysis should be reported as PRR or ROR including their respective 95% confidence interval. Other disproportionality analysis methods include information component and empirical Bayes geometric mean. None of these approaches appear to have been performed in the study by Farrell et al. The authors refer to a 'descriptive disproportionality analysis', but such methodology does not appear to have been previously reported in the literature. In the abstract, disproportionality analysis seems to be referred to as 'case-control study' which does not accurately reflect the nature of the study.Farrell et al. extracted a small subset of data from the EVV database of suspected Adverse Drug Reaction Reports (EV-ADR), which is the public facing database based on EVV, and appear to have compared the number of musculoskeletal AE reports identified for bedinvetmab and other selected products. For such comparison to be valid, the datasets should be directly comparable. However, this was not the case. There are significant limitations of the datasets selected which make the analysis incomplete and invalid. As part of the transition to new European Veterinary Medicines Regulation EU 2019/06, Zoetis took the decision to report up to three years of historical non-serious data to EVV. This is the reason for the peak in reports between January and June 2022 in Figure 3. The dataset is complete for bedinvetmab, but not for any other product. Zoetis (then Pfizer Animal Health) reported all European and third country (non-European) AE reports to EVV from September 2008 onwards. Before this date, there were over 23,900 Rimadyl ® (carprofen) AE reports that were not reported to EVV but are available in another open-access database (OpenFDA). 17 Nevertheless, a major limitation remains that data from non-US sources (including European AEs) are not available in any open-access database prior to 2004 and is of variable completeness after that. At least 3,500 'non-serious' AE reports related to carprofen in the United States between 2004 and 2019 are available in OpenFDA but not in EVV. As a result, while the dataset for bedinvetmab is complete, the datasets for other products omit a substantial number of 'non-serious' cases. The MAHs have an obligation to submit global AE reports to EVV for all products authorised in Europe (serious cases prior to 28 January 2022 and all cases since that date). However, if a product was authorised earlier in countries outside of Europe, the respective AE reports would not be submitted to EVV. For example, Galliprant ® (grapriprant) was authorised in 2016 in the US and in 2018 in Europe. These cases would not have been reported to EVV. Visual inspection of the data should be considered as multiple sources may report the same incident. 18,19 Detailed review of full case information is required to identify duplicates; however, this is not possible with data extracted from EV-ADR (see below). This is due to EMA's EVV Access Policy 20 and data protection laws. For example, a very simple case in EV-ADR may contain <20 datapoints whereas a similar case in EVV would contain >60 data points including detailed case narrative. An adequate disproportionality analysis would require full case details including the case narrative to allow for interpretation of the case. 21 It is not possible to conduct an accurate disproportionality analysis using only the data available in EV-ADR. Published literature indicates that it is not appropriate to use disproportionality analysis to compare different products because they are missing incidence denominators, are subject to severe reporting bias and are not adjusted for confounding. 22 In conclusion, and based on the scientific literature on the topic, the methodology reported by Farrell et al. is inconsistent with disproportionality analysis. Referring to the methodology described in the study as disproportionality analysis does not align with established definitions of disproportionality analysis and may lead to misinterpretation of the findings. In addition, the extracted data are complete for bedinvetmab but significantly incomplete for every other product listed. It is therefore not appropriate to make any comparison or analysis between products based on the extracted data. Farrell et al. define 'translation error' as 'a clinically important discrepancy between the adverse event report (AER) submitted by the attending veterinarian and the report filed by Zoetis'. Upon review, only one error was identified (case #10 -see Table 1 Farrell et al. stated "Furthermore, despite being invited to provide annotated images to clarify this discrepancy, Zoetis declined to do so". However, the cited reference 28 actually states "A summary and characterization of these incidental histopathological background findings in Beagle dogs has since been presented at the 2024 Society of Toxicologic Pathology Symposium (Cole et al., 2024). 29 A manuscript detailing these findings is in preparation". 28 Zoetis is actively working to ensure this information is made available in the scientific record.Farrell et al. stated "Notably, the joint safety claims outlined in Librela's datasheet are based on radiographic assessment of five healthy beagles who received the recommended dose". In contrast, the cited reference 30 describes a radiographic assessment involving 24 dogs (eight in each group being administered 1X, 3X and 10X the recommended dose). 30 It is also important to note that in the cited reference no joint risk was observed even at overdose conditions (3X and 10X), which provides additional context to the safety profile. Farrell et al. conclude a strong suspicion of causality based on a 'descriptive disproportionality analysis' and expert opinion reviewing cases without comparator controls. The methodology described does not align with established pharmacovigilance practices, and the conclusion exceeds the evidentiary strength of the data. The study design cannot prove/support the causation asserted by the authors. Reports of a series of cases can be useful to create awareness and inform future research, but do not allow for meaningful scientific conclusions to be drawn.Similar concerns to those identified above have been raised in the human literature about the rapid increase in studies based on open-access pharmacovigilance databases inappropriately using disproportionality methods to compare the safety profiles of different drugs. In human medicine, guidelines address the issue of poorly reported methodology and results in published articles of disproportionality analysis. 31 Furthermore, a meta-epidemiological study highlights a common issue in disproportionality studies referred to as 'spin'. Spin is characterized by overinterpretation of results including inappropriate interpretations and extrapolations or misleading reporting. 32 This has been described as 'pharmacovigilance syndrome'. 33 The authors of this letter support increasing transparency and greater understanding of veterinary pharmacovigilance, but it is also important that publications related to veterinary pharmacovigilance follow sound scientific principles and current scientific knowledge. The publication by Farrell et al. also illustrates the need for providers of open-access veterinary pharmacovigilance databases to provide detailed explanations of potential limitations of the data available in the database and what uses of the data may be appropriate. This may help ensure that all consumers of such data are guided to avoid similar misunderstandings in the future.Zoetis does not exclude the possibility of any AE and continues to investigate AE reports received, especially those with imaging provided, to try to identify any causal relationship between musculoskeletal AEs and bedinvetmab administration. Veterinary professionals are encouraged to report any cases they have concerns with (for bedinvetmab or any other veterinary medicine) by following national reporting requirements. Effective post-marketing pharmacovigilance relies on the collaboration of all stakeholdersanimal owners, veterinary professionals, MAHs, and regulatory agencies-for the benefit of the users of veterinary medicines and the veterinary patients themselves.