Conceiving, conducting, reporting, interpreting, and publishing disproportionality analyses: A call to action

Abstract

Pharmacovigilance is becoming a clinically oriented discipline dealing with the detection, assessment, understanding and prevention of adverse effects or any other possible drug-related issues, thus proactively supporting safe prescribing in the real world. Therefore, thanks to increasing awareness among clinicians and patients, the reporting of adverse events after drug exposure to a pharmacovigilance system is becoming pivotal not only for regulatory purposes, but also for research activities, namely, the analysis of large spontaneous reporting systems (SRSs). These post-marketing studies usually rely on the disproportionality concept, through the so-called case/non-case approach. The detection of a “statistically significant” disproportionality signal should not be intended as an adverse drug reaction (ADR) or a proven drug-event association, but rather as an alert requiring additional characterization (case-by-case assessment), including biological plausibility and analytical pharmaco-epidemiological investigations, when needed.1 Indeed, positive prognostic values of disproportionality analyses to detect true ADR are estimated between 0.2 and 0.6 according to method and threshold; thus, a large amount of false-positive signals is expected.2 In the recent past, the public availability of SRSs collecting anonymous data, not requiring approval by ethics committee or a pre-planned protocol, together with the relative simplicity in data analyses, generated a plethora of publications, sometimes redundant, especially in the cardiovascular and oncological areas.1 The latest meta-epidemiological investigations found major concerns on transparency and reporting of disproportionality analyses, which were extremely heterogenous on key methodological aspects such as selection of comparator3; this undermines the credibility and reproducibility of the results,4 which are also frequently over-interpreted, notably in the abstract.5 Ten years ago, the British Journal of Clinical Pharmacology fuelled the debate on the strengths and limitations of disproportionality analyses, including the benefits of publishing relevant results in medical journals.6, 7 A key overlooked aspect is represented by the rationale and actual added value of a disproportionality analysis for the scientific community. The availability of already published pharmaco-epidemiological and/or spontaneous reporting data should be carefully taken into account before planning a disproportionality analysis to avoid research waste and redundancies. We believe that conception of a study based on SRSs shall rely on (a) early detection of rare, unexpected, late-onset or long-lasting ADRs (including those from drug interactions), which cannot be fully appreciated in pivotal trials; (b) characterization of adverse events of special interest (following an imbalance from pre-approval clinical trials), especially in terms of clinical features such as time to onset and reversibility; (c) exploration of the underlying pharmacological bases, including potential correlation with pharmacokinetic (e.g., lipophilicity) and pharmacodynamic features (e.g., receptor affinity/occupancy); (d) description/demonstration of methodological issues (i.e., potential biases and confounders to be accounted for in study interpretation). The recent study by Chen et al., described the safety profile of alpelisib and captured our attention.8 The study stemmed from the importance of monitoring the post-marketing safety of a novel medication, a regulatory requisite. In other words, there is no strong clinical rationale. The authors applied a traditional disproportionality analysis, although the comparator was not reported, and generated 129 disproportionality signals. Of these, several signals, including important medical events, can be ascribable to the underlying disease and its severity (e.g., metastasis), rather to be actually drug induced (the so-called confounding by indication and channelling bias, respectively). Calculation of disproportionality by therapeutic area would be advisable to mitigate the former (i.e., by comparing alpelisib vs. other anticancer drugs or agents used for breast cancer). The remaining list of adverse events found by Chen et al. is in agreement with safety data from pivotal SOLAR-1 trial. Although we support the publication of negative findings (i.e., no novel safety signals), more emphasis should have been put on further clinical characterization of major known safety issues (e.g., hyperglycaemia, rash, and diarrhoea), to uncover new information on severity/seriousness, discontinuation/dechallenge, co-reported drugs/events, and latency; rash and hyperglycaemia occurred early (median onset 13–15 days), whereas diarrhoea was delayed (median onset 139 days) in SOLAR-1.9 These clinical features, together with conventional or ad hoc developed prioritization criteria (e.g., the strength of disproportionality), should be used systematically to select adverse events of clinical interest and targeting proactive monitoring strategies in the real world. Therefore, in the era of Open Science, we believe there is an urgent need to harmonize the publication of a disproportionality analysis through common criteria, ultimately resulting in shared recommendations for conceiving, conducting, reporting, and interpreting its results. These guidelines, accounting for quality assessment, will complement Good Signal Detection Practices,10 and will hopefully assist: (a) Researchers, to standardize, at least partially, disproportionality analysis and increase reproducibility, (b) Editors and Reviewers, in prioritizing the publication of a given study on SRSs; (c) Clinicians, to effectively apply results from spontaneous reporting studies in clinical practice, thus supporting a real proactive benefit–risk evaluation. We welcome contribution from interested stakeholders to pursue this call towards an effective concerted proposal. ER reports personal fees from Novartis, outside the content of this letter. FS and CK have no competing interests to declare. ER conceive the idea and wrote the first draft of the letter. FS and CK make substantial contribution to the content of the letter and approved the final version.

Topics & Concepts