MethodsUsing the nationwide Danish registries, we identified AF patients initiating dabigatran, rivaroxaban, or apixaban between 2011 and 2015 (n=50,627). Applying a symmetry analysis design, we screened for AEs of NOAC, as reflected by new drug treatments, incident diagnoses, or procedures. For signals with the lowest number needed for one additional patient to be harmed (NNTH), we evaluated whether they likely represented genuine AEs or other types of associations. Signals assessed as potential AEs were grouped into five categories for analysis of effect modification according to patient and drug characteristics.
ResultsOf the identified signals, 61 were classified as potential AEs. Most signals could be categorized according to the following types of AEs: bleedings, non-bleeding gastrointestinal symptoms, mental disease, urinary tract disorders, and musculoskeletal symptoms. Older age and first-ever use of anticoagulants was associated with strengthening of all 'NOAC-adverse effect'-associations. Conversely, use of low-dose NOAC and apixaban led to attenuation of most associations.
ConclusionThrough a symmetry analysis based hypothesis-free screening of large scale health care databases, we were able to confirm well-established AEs to NOAC therapy in clinical practice as well as potential AEs that deserve further investigation.3
Key Points• A hypothesis-free screening of automated health care databases generated signals of potential adverse effects to NOAC therapy. The vast majority of signals could be categorized as bleedings, non-bleeding gastrointestinal adverse effects, mental disease, urinary tract disorders, or musculoskeletal symptoms.• The risk of experiencing adverse effects seems to vary according to patient and drug characteristics such as age and NOAC dose.• The present study has generated several specific signals of potential adverse effects to NOAC therapy. These signals should be evaluated in future studies. 5
MethodsUsing a symmetry analysis design [7], we analyzed the occurrence of events (drug initiation, new diagnosis, and procedures) in a symmetrical time-window before and after initiation of a NOAC for AF. Any non-symmetrical distribution of the occurrence of these events before and after NOAC initiation might reflect effects of NOAC treatment.
Data sourcesVirtually all medical care in Denmark is furnished by the national health authorities, allowing population-based register-linkage studies covering all inhabitants of Denmark [12]. We used data from three nationwide health registries. Data on drug use was retrieved from the National Prescription Registry [13] as filled prescriptions. From the National Patient Register [14] we obtained data on registered in-and outpatient diagnoses (classified according to International Classification of Diseases, 10 th Edition (ICD-10)) and procedures from non-psychiatric hospitals. Finally, the Civil Registration System[15] was used to keep track of study subjects with respect to deaths and migrations. Data were linked by using the set-up provided by Dan...