Phase 2: Illustrative analysis
In our sample analysis, we identified 132 patients who fulfilled the
inclusion/exclusion criteria within the defined study period (Figure 4).
Most patients were warfarin users (n=101, 76.5%), followed by
rivaroxaban users (n=26,19.7%). There were only 5 patients on apixaban
(3.8%) in the period studied (from January 2013 to December 2016). The
median age of patients across the different OACs were similar, ranging
from 70 to 71 years old. Racial and gender distributions in the cohort
were similar to the ratios of the local population in Singapore. The
descriptive characteristics of the cohort were detailed in Table 2.
The cohort was analysed for the type of AF diagnosis (Table 3). Majority
of patients were assigned the code for “atrial arrhythmia” (92.4% of
all patients), followed by “Atrial Fibrillation” (11.4%). There were
eight patients in the warfarin arm diagnosed with both “Atrial
Fibrillation” and “Atrial Arrhythmia”, and four diagnosed with both
“Atrial Fibrillation and Flutter” and “Atrial Arrhythmia”. Forty
patients (30.3%) experienced a haemorrhagic event and 14 (10.6%)
experienced a thromboembolic event following OAC exposure over the
duration of follow up in our study. Only five patients (3.8%) had a
documented exposure of concurrent bleed-inducing medications. All five
were on warfarin for anticoagulation.
Overall, the incidence of bleeding was highest among warfarin users at
36.6% (37 out of 101 patients), followed by rivaroxaban at 11.5% (3
out of 26 patients). There were no haemorrhagic events among apixaban
users, however there were only five patients in the cohort exposed to
apixaban. Thromboembolic events were less common, with eight (7.9%)
from the warfarin arm, five (19.2%) from rivaroxaban, and one (20.0%)
from apixaban. Majority (59.0%) of patients did not experience any
haemorrhagic nor thromboembolic events within the observation period and
were ‘event-free’.
To visualize relative adverse event rates while accounting for
differences in utilization, we extended the previous work by Hripcsak et
al and created a novel method of visualizing efficacy, safety and
utilization in one chart.12 Figure 5 shows a 100%,
horizontally stacked, bar chart combining relative utilization
information (vertical thickness) with the proportions experiencing
events of interest (bleeding and thromboembolic events, horizontal
proportion). The central grey region represents the proportion not
experiencing any haemorrhagic or thromboembolic events at the end of
follow up. As patient-time of exposure (and therefore incidence rates)
could not be incorporated into the figure, a fixed-timepoint analysis
was conducted only including patients who had completed 6 months of
continued OAC exposure, to equalize exposure times (Figure 6).