Table 1 Filtering rules used to select candidate positive test cases based on labeling revisions.

Event listed in earlier label

Rationale: The association is not new

Example: The event myalgia associated with the drug Arimidex posted at MedWatch on May 2013 as ‘added.....myalgia’ in the Adverse Reaction section, was mentioned in a labeling revision dated April 2009. Similarly, the event anaphylaxis associated with the drug Azulfidine EN posted on July 2013 under the Adverse Reactions section, was mentioned in a labeling revision dated December 2012

Event part of a contraindication

Rationale: Associations where the event is qualified by a co-existing contraindicated situation (e.g., presence of a comorbid condition, physiological state, demographic risk factor, coadministered drug) require non-standard analyses approaches to properly identify or test because the co-existing contraindicated situation will also need to be identified and incorporated into the analysis

Example: The association Depakene (Valproate)-hepatotoxicity posted on July 2013 as an update to the Warnings and Precautions section: ‘Valproate is contraindicated in patients known to have mitochondrial disorders caused by POLG mutations… Valproate-induced acute liver failure and liver-related deaths have been reported in patients with hereditary neurometabolic syndromes caused by mutations in the gene for mitochondrial DNA polymerase γ (POLG)…’

Event is too general

Rationale: Challenging to accurately identify the cases (data records) needed for analysis

Example: The event hypersensitivity posted on February 2013 in association with Aromasin (exemestane) is a broad term that refers to wide range of adverse reactions such Anaphylaxis, Thrombocytopenia, Lupus Nephritis, Montoux and more

Events like drug abuse, misuse, dependence, and withdrawal

Rationale: Challenging to accurately identify the cases (data records) needed for analysis, e.g., drug abuse or withdrawal are not necessarily coded (billed for or mentioned) in health records and may need to be inferred via other mechanisms that are prone to error

Example: Withdrawal effects associated with the drug Ambien (Zolpidem) posted on April 2013

Multi-ingredient drugs where one of the ingredients is already known to cause the event

Rationale: The association is unlikely new given that the ingredient already known to be associated with the event is most likely the culprit

Example: The association between Symbyax (olanzapine+fluoxetine) and the event QT Prolongation posted on July 2013. Fluoxetine (Prozac) is already known to be associated with the event QT Prolongation

The event is inferred from other drugs with the same/similar ingredients to the drug in question

Rationale: The association is not new, and possibly not observed or reported with the specific drug in question

Example: The association between Xyzal (levocetirizine) and myoclonus posted on November 2013; ‘Besides these events reported under treatment with XYZAL, other potentially severe adverse events have been reported from the post-marketing experience with cetirizine. Since levocetirizine is the principal pharmacologically active component of cetirizine, one should take into account the fact that the following adverse events could also potentially occur under treatment with XYZAL: orofacial dyskinesia, severe hypotension, cholestasis, glomerulonephritis, still birth, tic, myoclonus, and extrapyramidal symptoms’

Events that seem to have been identified through or supported by pre-marketing clinical trials

Rationale: The association was observed in pre-marketing clinical trials and therefore unlikely newly observed

Example: The association between Potiga (ezogabine) and retinal abnormalities posted September 2013; ‘The retinal abnormalities observed with POTIGA have been reported in patients who were originally enrolled in clinical trials with POTIGA’

A different preparation or time release mechanism (IR versus ER) of the drug already known to be associated with the event

Rationale: Cannot definitively determine that the association is new

Example: The association between Kapvay (Clonidine) extended-release and hallucinations posted on February 2013. A labeling revision from September 2010 states the following: ‘The following less frequent adverse reactions have also been reported in patients receiving immediate-release clonidine, but in many cases patients were receiving concomitant medication and a causal relationship has not been established… hallucinations’

Associations included in existing reference standards

Rationale: The association is most likely not new. There is also a need to separate associations used in retrospective benchmarks from those designed for prospective benchmarks since the two types of benchmarks should complement each other

Example: The association between Clozaril (clozapine) and liver injury posted on March 2013 is part of the OMOP reference standard (positive test case)

The event is polymorphic to another event included in an earlier label

Rationale: The association is unlikely new

Example: Torsades de pointes posted on July 2013 in association with Agrylin (anagrelide) is a type of tachycardia included in earlier label dated November 2011