Keywords hierarchical composite outcome - atrial fibrillation - anticoagulants - ABC pathway
- thrombosis
Introduction
Atrial fibrillation (AF) is the most common cardiac arrhythmia and remains a significant
global health burden. The Global Burden of Diseases database estimates that AF affects
over 37.5 million people worldwide in 2017, with a rising prevalence and incidence.[1 ] AF is associated with an increased risk of stroke, systemic embolism, and mortality.[2 ] To improve patients' outcome, current guidelines recommend the use of a holistic
or integrated care approach based on the AF Better Care (ABC) pathway, which includes
the pillars of AF care, as follows: “Anticoagulation to avoid stroke,” “Better symptom
control,” and “Cardiovascular risk and comorbidity management.”[3 ]
[4 ] Adherence with the ABC pathway has been associated with improved clinical outcomes.[5 ]
The standard choice of oral anticoagulants (OACs) has been vitamin K antagonists (VKAs),
including warfarin; however, warfarin possesses several limitations, such as narrow
therapeutic range, frequent international normalized ratio (INR) monitoring, and drug–drug/drug–food
interaction.[6 ] From recent trials and meta-analyses, non-VKA oral anticoagulants (NOACs), such
as dabigatran, rivaroxaban, apixaban, and edoxaban, have emerged as an alternative
therapeutic option for stroke prevention with noninferior efficacy, better safety
profile, and more convenient use.[7 ]
[8 ]
In cardiovascular trials, a primary composite endpoint is often used to evaluate the
efficacy of randomized treatment, which often consists of two or more types of clinical
events (i.e., cardiovascular death, stroke, hospitalization). This conventional composite
endpoint analysis focuses on time-to-first event, neglecting severity or clinical
importance of events. The effect size of time-to-first event analysis was mainly contributed
by less severe events that occurred earlier than more severe events, including death.[9 ] Pocock et al proposed the “Win Ratio” as a new approach to analyze composite outcome
in the way that account for both clinical priorities and timing of the events.[10 ] In recent years, the Win Ratio analysis has gained attention in many cardiovascular
trials and post-hoc analysis of trials,[11 ]
[12 ]
[13 ]
[14 ] especially from Western population; however, its application in disease registries
and Asian cohorts remains scarce.
While randomized controlled trials (RCTs) have demonstrated the efficacy and safety
of NOACs,[15 ]
[16 ]
[17 ]
[18 ] real-world data from large registries are crucial to providing valuable insights
into how anticoagulants perform in routine clinical practice, especially in clinically
complex AF patients.[19 ]
[20 ] We therefore hypothesize that the Win Ratio method could be a novel tool in analyzing
disease registry outcomes and providing a more comprehensive understanding of the
treatment benefits when used in routine clinical practice.
In this study, we applied a Win Ratio analysis in the context of the COhort of antithrombotic
use and Optimal INR Level in patients with nonvalvular Atrial Fibrillation in Thailand
(COOL-AF) registry, a multicenter nationwide prospective cohort of patients with AF.[21 ] The objectives of our study are, firstly, to demonstrate the analysis and interpretation
of Win Ratio in a disease registry in relation to clinical outcomes; secondly, to
access the comparative efficacy of NOACs and warfarin in Asian patients with AF; and
thirdly, we applied a Win Ratio analysis to AF patients who were ABC pathway adherence
compared with those who were nonadherent.
Methods
Study Population
We utilized the data from the COOL-AF registry,[21 ] a nationwide prospective study that recruited nonvalvular AF (NVAF) patients aged
18 years or older from 27 hospitals in Thailand. Patients with rheumatic or severe
valve disease, prosthetic valve or valve repair, AF from transient reversible cause,
bleeding disorders, such as thrombocytopenia or myeloproliferative disorders, etc.,
ischemic stroke within 3 months, pregnancy, current participation in a clinical trial,
life expectancy less than 3 years, or inability to attend follow-up were excluded
from the registry. The protocol for the COOL-AF registry was approved by the Institutional
Review Board of Central Research Ethics Committee (COA: CREC 003/2014). All patients
provided written informed consent. The patients in the COOL-AF registry, who were
treated with OACs at baseline, were included in the present analysis.
Study Protocol and Data Collection
The details of the COOL-AF registry protocol were previously published.[21 ] Clinical data were collected by investigators from the medical record and patient
interview, which were recorded in the study case record form, entered in the web-based
system, and verified via central data management. All participating hospitals underwent
site monitoring and followed the good clinical practice. The data were collected at
baseline, and at 6, 12, 18, 24, 30, and 36 months. The collected data included demographic
data, vital signs, time after AF diagnosis, AF symptoms, type of AF, medical history,
and medications.
Outcomes
The clinical outcomes were all-cause death, intracranial hemorrhage (ICH), ischemic
stroke/transient ischemic attack (TIA)/systemic embolism (SSE), non-ICH major bleeding,
and myocardial infarction (MI) or heart failure (HF). ICH was defined as bleeding
within the cranium, such as intracerebral bleeding, subdural bleeding, and subarachnoid
bleeding, but did not include microbleeds or hemorrhagic transformation.[22 ] Ischemic stroke was defined as an acute onset of a focal neurological deficit that
lasted longer than 24 hours, whereas TIA had a similar definition but lasting less
than 24 hours. Systemic embolism was defined as sudden loss of end-organ perfusion
supported by both clinical and objective evidence. Major bleeding was defined using
the International Society of Thrombosis and Hemostasis criteria.[23 ] The definition of MI was derived from the Fourth Universal Definition of Myocardial
Infarction.[24 ] HF event was defined as either a hospitalization or an urgent or unscheduled visit
to clinic, office, or emergency department due to symptoms and objective evidence
of new or worsening HF. The definition also requires initiation or intensification
of treatment specifically for HF.[25 ] All clinical events in the COOL-AF registry were adjudicated by the clinical event
committee.
For hierarchical composite outcome analysis, we determined the order of the outcome
based on clinical severity, prioritizing death, systemic embolism, and major bleeding
events over MI or HF events. The hierarchical order was (1) all-cause death, (2) ICH,
(3) SSE, (4) non-ICH major bleeding, and (5) MI or HF.
To compare adherence to the ABC pathway with nonadherence, we applied the unmatched
Win Ratio to analyze the hierarchical composite outcome in the order mentioned above.
Since the entirety of our patient cohort received OAC therapy, we conducted the Win
Ratio analyses to compare (1) adherence to ABC pathway versus nonadherence, (2) adherence
to component B versus nonadherence, and (3) adherence to component C versus nonadherence.
The definition of ABC pathway adherence was established based on the original definition.[26 ] Adherence to component A was achieved if patient received appropriate OAC strategy
to their stroke risk at baseline. Component A adherence was met when male patients
with CHA2 DS2 -VASc score ≥1 or ≥2 in females who received an OAC, and male patients with CHA2 DS2 -VASc score = 0 or ≤1 in females who did not receive an OAC. Component B adherence
was fulfilled if patients had a European Heart Rhythm Association score ≤2. For component
C, adherence was considered if patients had appropriate comorbidity management. This
included (1) hypertension management with angiotensin-converting enzyme inhibitors
(ACEi)/angiotensin (II) receptor blockers (ARB), calcium channel blockers, diuretics,
and β-blockers (BBs), (2) coronary artery disease management with ACEi/ARB, BB, and
statins, (3) ischemic stroke/TIA management with statins, (4) HF management with ACEi/ARB
and BB, and (5) diabetes management with oral antidiabetics or insulin. Adherence
to the ABC pathway was achieved when the patient met all three components.
Statistical Analysis
In this study, we conducted an analysis of all patients from the COOL-AF registry
who were treated with NOACs or warfarin at baseline. The baseline characteristics
of patients were presented as means ± standard deviation (SD), and frequency for continuous
and categorical variable, respectively. The incidence rates of each clinical outcome
were presented as rate per 100 person-years with a Poisson 95% confidence interval
(95% CI) and a two-sided p -value.
Win Ratio is an emerging concept for analyzing composite outcome in clinical trials,
which focuses on the severity of the outcome and timing of occurrence instead of solely
relying on time-to-first event in conventional analysis. There are two approaches
for analyzing hierarchical composite outcome using Win Ratio—unmatched-pair and matched-pair
approaches.[10 ]
[11 ]
In the unmatched approach, all patients from the NOAC group were paired with all patients
from the warfarin group. In each pair, the NOAC group was counted as a winner when
a patient in the warfarin group developed all-cause death before the NOAC group. In
contrast, the NOAC group was considered a loser if a patient in this group developed
all-cause death first. If neither of these conditions were met, we assessed the next
outcome in the prespecified hierarchical order until we determined the winner or loser.
If the winner or loser could not be determined, the pair was a tie. The Win Ratio
was calculated by dividing the number of win pairs by the number of lose pairs. The
value of win ratio greater than 1.00 indicated that the treatment group (NOACs) had
better outcomes compared with the control group (warfarin). Win Ratio, 95% CI, two-sided
p -value, and numbers of wins, losses, and ties were calculated with WINS package in
R.[27 ]
In the matched-pair approach, we aim to compare the outcomes of the similar-risk patients
between two groups. Our study employed propensity score matching by nearest neighbor
method with logistic regression from the MatchIt package in R.[28 ] Each NOAC patient was matched with four warfarin patients based on all variables,
including age, sex, body mass index, duration after AF diagnosis, AF symptoms, type
of AF, history of congestive HF, history of revascularization, implanted cardiac device,
peripheral arterial disease, carotid occlusive disease, ischemic stroke or TIA, hypertension,
dyslipidemia, diabetes mellitus, smoking, renal replacement therapy, dementia, history
of bleeding, and antiplatelet use. Win ratio, 95% CI, and two-sided p -value were calculated as described by Pocock et al.[10 ]
Because our study utilized data from a prospective registry, it is possible that there
are some differences in the baseline characteristics between the NOACs and the warfarin
group. Therefore, we created a propensity-matched population to minimize the differences
and conducted a conventional win ratio analysis as a secondary analysis. The propensity-matched
population was generated based on the same method as described in the matched-pair
approach. Then we conducted an analysis, in which all patients from the NOAC group
were paired with all patients from the warfarin group, to determine the Win Ratio,
95% CI, two-sided p -value, and numbers of wins, losses, and ties of the propensity-matched population.
For sensitivity analysis, we analyzed hierarchical composite outcome in the order
of (1) ICH, (2) SSE, (3) non-ICH major bleeding, and (4) MI or HF, using the unmatched
win ratio approach.
As demonstrated by Oakes and Finkelstein and colleagues, Win Ratio and win proportion
could vary over the follow-up time.[29 ]
[30 ] Specifically, the beneficial effect of the treatment that contributes to wins may
be driven by different outcomes at each time point in the study period. The win proportion
of a group was defined as the number of wins for that group divided by the total number
of pairs.[31 ] We used WIN package to calculate Win Ratio and win proportion for each group over
the follow-up time to investigate the changes in the contribution to wins by each
outcome. All analyses were done using R version 4.2.3 (www.r-project.org ), IBM SPSS Statistics version 28.0 (IBM Corp., Armonk, New York, United States),
and MedCalc Statistical Software version 20 (MedCalc Software Ltd., Ostend, Belgium).
Results
Study Population
Of the total cohort of 3,461 patients from the COOL-AF registry, follow-up data were
unavailable for 50 patients, and 837 patients did not receive OACs. Therefore, a total
of 2,568 patients (mean age: 68.8 ± 10.7 years; 43.4% female) were included in this
analysis. The flow diagram of the study population is illustrated in [Fig. 1 ].
Fig. 1 Flow diagram of study population and hierarchical of composite endpoints. ABC, Atrial
fibrillation Better Care; HF, heart failure; ICH, intracranial hemorrhage; MI, myocardial
infarction; OAC, oral anticoagulants; SSE, ischemic stroke or transient ischemic attack
or systemic embolism.
Baseline Characteristics
The number of patients receiving warfarin and NOACs was 2,340 (91.1%) and 228 (8.9%),
respectively. The median follow-up time was 35.9 months, with an interquartile range
of 34.7 to 36.0. Among the study population, 1,302 (50.7%) patients adhered to the
ABC pathway, with adherence to components B and C observed in 1,942 (75.6%) and 1,787
(69.6%) patients, respectively.
[Table 1 ] shows the baseline characteristics of the study population. Permanent AF was the
majority of the warfarin group (52.7%), while paroxysmal AF was prominent in the NOAC
group (50.0%). The NOAC group had more patients with a cardiovascular implantable
electronic device. Patients in the warfarin group had more of the following characteristics:
history of HF, history of ischemic stroke/TIA, hypertension, chronic kidney disease,
anemia, and receiving antiplatelet. There were no statistically significant differences
in other baseline factors between the two groups.
Table 1
Baseline characteristics of the study population
Characteristics
All
(N = 2,568)
Warfarin
(n = 2,340)
NOACs
(n = 228)
p -Value
Age (y)
68.8 ± 10.7
68.8 ± 10.7
68.5 ± 10.6
0.701
Female sex
1,115 (43.4%)
1,017 (43.5%)
98 (43.0%)
0.889
Body mass index (kg/m2 )
25.2 ± 4.8
25.2 ± 4.8
25.3 ± 4.4
0.748
Time after diagnosis of AF (y)
3.5 ± 4.4
3.5 ± 4.4
3.6 ± 4.7
0.755
Atrial fibrillation
Paroxysmal
778 (30.3%)
664 (28.4%)
114 (50.0%)
<0.001
Persistent
482 (18.8%)
443 (18.9%)
39 (17.1%)
Permanent
1,308 (50.9%)
1,233 (52.7%)
75 (32.9%)
Symptomatic AF
1,974 (76.9%)
1,797 (76.8%)
177 (77.6%)
0.775
History of heart failure
702 (27.3%)
660 (28.2%)
42 (18.4%)
0.002
History of coronary revascularization
416 (16.2%)
378 (16.2%)
38 (16.7%)
0.841
History of PAD
32 (1.2%)
31 (1.3%)
1 (0.4%)
0.250
History of ischemic stroke/TIA
538 (21.0%)
502 (21.5%)
36 (15.8%)
0.045
History of bleeding
273 (10.6%)
252 (10.8%)
21 (9.2%)
0.466
Diabetes mellitus
690 (26.9%)
637 (27.2%)
53 (23.2%)
0.196
Hypertension
1,862 (72.5%)
1,710 (73.1%)
152 (66.7%)
0.039
Smoking
473 (18.4%)
434 (18.5%)
39 (17.1%)
0.592
Dyslipidemia
1,507 (58.7%)
1,373 (58.7%)
134 (58.8%)
0.977
Renal replacement therapy
22 (0.9%)
21 (0.9%)
1 (0.4%)
0.715
CHA2 DS2 -VASc score
Low risk
102 (4.0%)
81 (3.5%)
21 (9.2%)
<0.001
Intermediate risk
348 (13.5%)
310 (13.2%)
38 (16.7%)
High risk
2,118 (82.5%)
1,949 (83.3%)
169 (74.1%)
HAS-BLED score
0
336 (13.1%)
286 (12.2%)
50 (21.9%)
<0.001
1–2
1,818 (70.8%)
1,653 (70.7%)
165 (72.4%)
≥3
414 (16.1%)
401 (17.1%)
13 (5.7%)
Dementia
25 (1.0%)
21 (0.9%)
4 (1.8%)
0.273
CIED
270 (10.5%)
230 (9.8%)
40 (17.5%)
<0.001
Antiplatelet
309 (12.0%)
294 (12.6%)
15 (6.6%)
0.008
Abbreviations: AF, atrial fibrillation; CIED, cardiac implantable electronic device;
CKD, chronic kidney disease; PAD, peripheral arterial disease; TIA, transient ischemic
attack.
For the propensity-matched population, 228 patients from the NOAC group were matched
to 912 patients from the warfarin group, resulting in a total of 1,140 patients. The
mean age (±SD) of the patients was 67.7 ± 11.1 years, and 43.0% were female. There
were no statistically significant differences of baseline variables between NOACs
and warfarin groups. The baseline characteristics are displayed in [Supplementary Material ] ([Supplementary Table S1 ] [available in the online version]).
Incidence Rate of Outcomes
The warfarin group exhibited greater event rates (events per 100 person-years) than
the NOAC group consistently across all outcomes; however, statistically significant
differences were observed only in all-cause death and MI or HF outcome. Specifically,
patients who received warfarin had an all-cause death rate of 4.51 (95% CI: 3.99–5.07),
while the rate in NOAC group was 2.86 (95% CI: 1.69–4.52; p = 0.025). The event rates for all other outcomes are summarized in [Supplementary Material ] ([Supplementary Table S2 ] [available in the online version]).
Principal Composite Outcome
Two approaches of principal hierarchical outcome analysis were conducted as described
in the Methods section. In the unmatched approach, all patients in the NOAC group
were paired with all patients in the warfarin group, resulting in 228 × 2,340 = 533,520
matched pairs. The NOAC group had a total of 108,571 wins (20.3%), while experiencing
66,109 losses (12.4%) and 358,840 ties (67.3%). The Win Ratio was 1.64 (95% CI: 1.22–2.20;
p < 0.001), with all-cause death as a major contributor to wins (52.9%). There were
more wins than losses for the NOAC group in every outcome except for ICH events, which
yielded 4,714 wins and 7,314 losses. The comprehensive breakdown of the hierarchical
analysis is illustrated in [Fig. 2 ].
Fig. 2 Unmatched Win Ratio for the hierarchical composite outcome analysis of all-cause
death, ICH, SSE, non-ICH major bleeding, and MI or HF in NOACs versus warfarin group.
HF, heart failure; ICH, intracranial hemorrhage; MI, myocardial infarction; NOACs,
non-vitamin K antagonist oral anticoagulants; SSE, ischemic stroke or transient ischemic
attack or systemic embolism.
In matched pairs, each patient in the NOAC group was matched to four patients in the
warfarin group as previously described, yielding 912 pairs. The total number of wins
for the NOAC group was 156 (17.1%), the total number of losses was 117 (12.8%), and
the number of ties was 639 (70.1%). The Win Ratio for matched pairs was 1.33 (95%
CI: 1.17–1.52; p = 0.017) with all-cause death accounted for a half of the total wins. Similar to
the unmatched approach, the NOAC group encountered a greater number of wins than losses
in all outcomes, with the exception of ICH events, which had 11 wins and 16 losses.
The details of the hierarchical analysis are shown in [Fig. 3 ].
Fig. 3 Matched pairs Win Ratio for the hierarchical composite outcome analysis of all-cause
death, ICH, SSE, non-ICH major bleeding, and MI or HF in NOACs versus warfarin group.
HF, heart failure; ICH, intracranial hemorrhage; MI, myocardial infarction; NOACs,
non-vitamin K antagonist oral anticoagulants; SSE, ischemic stroke or transient ischemic
attack or systemic embolism.
Conventional Win Ratio analysis was also conducted on a propensity-matched population,
resulting in 207,936 pairs from 228 patients in the NOAC group and 912 patients in
the warfarin group. The Win Ratio for this analysis was 1.42 (95% CI: 1.02–1.99; p = 0.039), which is consistent with both previous analyses. The comprehensive details
of this analysis are displayed in Supplementary Material ([Supplementary Fig. S1 ] [available in the online version]).
Win Ratio and Win Proportion Changes over Follow-Up
The unmatched Win Ratio showed an initial increase, reaching its peak at around 200
days of follow-up time, which was then followed by a decrease and plateau for the
remainder of the study period ([Fig. 4A ]). The NOAC group had a higher total win proportion than the warfarin group, with
all-cause death and MI or HF outcomes being the main contributors to the determination
of wins and losses in both groups during the entire follow-up time. Interestingly,
all outcomes contributed to the win proportion of the NOAC group since the beginning
of the study. In contrast, only all-cause death and MI or HF outcomes contributed
to the win proportion of the warfarin group in the early stage, as depicted in [Figs. 4(B, C) ].
Fig. 4 Changes over the course of follow-up of: the unmatched Win Ratio with 95% confidence
interval in NOACs versus warfarin group (A ) and ABC pathway adherence versus nonadherence group (D ), win proportion of the NOAC group (B ), win proportion of the warfarin group (C ), win proportion of the ABC pathway adherence group (E ), and win proportion of the ABC pathway nonadherence group (F ). ABC, Atrial fibrillation Better Care; HF, heart failure; ICH, intracranial hemorrhage;
MI, myocardial infarction; SSE, ischemic stroke or transient ischemic attack or systemic
embolism; TIA, transient ischemic attack.
Sensitivity Analysis
The unmatched Win Ratio for the hierarchical outcome analysis in the order of ICH,
SSE, non-ICH major bleeding, and MI or HF event was 1.84 (95% CI: 1.32–2.57; p < 0.001), which was consistent with the primary analyses. The NOAC group achieved
a total of 82,367 wins (15.4%), 44,716 losses (8.4%), and 406,437 ties (76.2%). The
NOAC group achieved more wins than losses in all outcomes, including ICH events with
12,339 wins and 10,067 losses (see [Supplementary Material ]: [Supplementary Fig. S2 ] [available in the online version]).
Win Ratio Analysis in ABC Pathway Adherence
The Win Ratio comparing adherence to the ABC pathway with nonadherence was 1.57 (95%
CI: 1.33–1.85; p < 0.001). [Fig. 5 ] presents comprehensive results for each outcome. In terms of adherence to components
B and C individually, the corresponding Win Ratios were 1.76 (95% CI: 1.45–2.13; p < 0.001) and 1.46 (95% CI: 1.22–1.71; p < 0.001), respectively.
Fig. 5 Unmatched Win Ratio for the hierarchical composite outcome analysis of all-cause
death, ICH, SSE, non-ICH major bleeding, and MI or HF in ABC pathway adherence versus
nonadherence groups. ABC, Atrial fibrillation Better Care; HF, heart failure; ICH,
intracranial hemorrhage; MI, myocardial infarction; NOACs, non-vitamin K antagonist
oral anticoagulants; SSE, ischemic stroke or transient ischemic attack or systemic
embolism.
When examining the Win Ratio over the course of the study, we noted a slight increase
during the initial 150 days of follow-up, followed by a stable trend throughout the
follow-up period. Notably, the ABC adherence group exhibited a higher total win proportion
in comparison to the nonadherence group, with all-cause death and MI or HF as the
major contributors to wins ([Fig. 4D–F ]).
Discussion
First, our study demonstrates that NOACs have more beneficial effects than warfarin
in patients with NVAF in reducing a hierarchical composite outcome that includes all-cause
death, ICH, ischemic stroke/TIA or SSE, non-ICH major bleeding, and MI or HF. Second,
the Win Ratio and the win proportion change over follow-up, which might suggest the
time-dependent benefits of NOACs compared with warfarin, and the extent of these benefits
varied across each outcome. Third, adherence to the ABC pathway was associated with
favorable win outcomes compared with nonadherence.
By using the Win Ratio method, we simultaneously considered both clinical priorities
and timing of the events when comparing the composite outcome between two groups.
Specifically, this method enables us to focus more on severe events, for example,
all-cause death, over nonfatal and less important events.
According to a meta-analysis of RCTs, NOACs was associated with over 50% greater reduction
in ICH compared with warfarin.[7 ]
[32 ] In our primary analysis, we performed Win Ratio analysis by first considering all-cause
death, followed by ICH, ischemic stroke/TIA or SSE, non-ICH major bleeding, and MI
or HF. The results show that the NOAC group had more wins than the warfarin group
in every hierarchical outcome except for ICH events; however, by analyzing in this
hierarchical order, the benefits of NOACs over warfarin in the ICH outcome could be
obscured by all-cause death. This hypothesis was confirmed by the sensitivity analysis,
in which the NOAC group also had more wins than the warfarin group in ICH events.
The recommendation of anticoagulation for stroke prevention in AF of current guidelines
favors NOACs over warfarin due to its efficacy and safety profile.[3 ]
[4 ] Recently published patient-level meta-analysis of four pivotal trials demonstrated
that NOAC use was associated with an 8% significant reduction of all-cause death,
a 51% significant reduction of ICH, and a 19% significant reduction of stroke or systemic
embolism.[32 ] Our findings concur with those of the meta-analyses in that the NOAC group had larger
win proportions over the entire study time, which might suggest a better effectiveness
than warfarin in reducing the hierarchical outcome. When considering only ICH, SSE,
and non-ICH major bleeding, we found that the win proportions of the NOAC group were
also greater than those of the warfarin group during the whole follow-up period. This
finding is consistent with data from previous studies that showed the greater benefits
of NOACs over warfarin with significant reduction in ICH, ischemic stroke, SSE. Regarding
major bleeding risk, a reduction was observed but its statistical significance was
inconclusive among previous publications.[7 ]
[32 ]
[33 ]
[34 ]
[35 ]
Asian patients with AF who are taking warfarin were reported to have a greater risk
for ICH than non-Asians.[36 ] Meta-analyses of RCTs showed that standard dose of NOACs use in Asians was associated
with a greater reduction in risk for stroke or systemic embolism, ICH, and major bleeding
than non-Asians, when compared with warfarin. The reduction of all-cause mortality
risk was also observed with no significant difference between Asians and non-Asians,
while the reduced risk of MI being nonsignificant.[37 ]
[38 ] However, some analysis of real-world data found significant reductions in MI risk
in patients taking NOACs versus warfarin.[39 ]
[40 ]
In our study, the win proportion of MI or HF in the NOAC group was larger than the
warfarin group, and the contribution to wins of MI or HF was greatest among nonfatal
events. This might suggest the possible benefit of NOACs over warfarin in the MI or
HF outcome; however, further studies are needed to confirm the effect.
Regarding adherence to the ABC pathway, our study shows that the subgroup adherent
to the ABC pathway was associated with greater win benefits compared with the nonadherent
group in reducing the same hierarchical composite outcomes. These benefits are also
observed when comparing adherence to component B and component C to the nonadherence
group. Our findings are consistent with the results from a study that applied the
Win Ratio method to a cluster randomized trial in Asians with AF, comparing the mobile-health
application of the ABC pathway care to the usual care.[41 ]
To the best of our knowledge, this is the first study to apply Win Ratio analysis
to a prospective registry. The concept of Win Ratio has gained more attention in analyzing
composite outcome in cardiovascular trials in recent years,[11 ]
[12 ]
[13 ]
[14 ] whereas the application in cardiovascular registries is limited, and we are unaware
of prior studies in an Asian cohort.
Previous studies utilized only matched pairs approach in the retrospective cohort
analysis.[42 ]
[43 ] The rationale behind selectively applying the matched pairs approach might be the
nature of the retrospective cohort that required matching to mitigate the bias from
a nonrandomized design. We were also aware of this potential bias, so we employed
three methods of Win Ratio application, which were unmatched approach, matched pairs
approach, and conventional Win Ratio on propensity-matched population. The Win Ratio
from all three approaches consistently showed the benefits of NOACs over warfarin
in reducing the hierarchical composite endpoint of NVAF patients. Our approach could
pave the way for further applications of a Win Ratio analysis in cohort studies.
Limitation
There are some limitations to this study. First, our study acquired the data from
the COOL-AF registry, which were collected in Thailand where warfarin was the only
OAC reimbursed by the National Health Security Office. Therefore, the sample size
of the NOAC group was limited. Second, the Win Ratio method was initially designed
for clinical trials, so the application in prospective registry and its interpretation
might be subject to bias. We tried to account for the nonrandomized nature of the
prospective cohort by employing different approaches, including the propensity-matched
population. However, the concordance between our findings and previous results from
meta-analyses and trials comparing NOAC and warfarin in AF may suggest that the unknown
biases are likely to be limited.
Conclusion
This Win Ratio analysis of the COOL-AF registry demonstrated the significant beneficial
effects of NOACs over warfarin in reducing all-cause death, ICH, SSE, non-ICH major
bleeding, and MI or HF in patients with AF. Furthermore, our findings highlight the
association between adherence to the ABC pathway and a reduction in the hierarchical
composite outcome. This study underscores the potential of employing the Win Ratio
method as a novel approach for analyzing time-to-event outcomes in cardiovascular
registry analyses.
What is known about this topic?
When comparing to warfarin, NOAC use in Asians with AF was associated with a greater
reduction in the risk of stroke, ICH, and major bleeding than non-Asians.
AF Better Care (ABC) pathway adherence has been associated with improved outcomes
in AF patients, but the data are limited in Asians.
Win Ratio is a new approach for analyzing composite outcomes considering event timing
and severity. It was developed and applied mainly in cardiovascular trials.
What does this paper add?
In a real-world setting, adherence to the ABC pathway and the use of NOACs were associated
with a greater reduction in the composite outcome of all-cause death, ICH, SSE, non-ICH
major bleeding, and MI or HF compared with ABC pathway nonadherence and warfarin in
Asians.
The application of Win Ratio in a prospective AF registry may be a novel approach
for analyzing time-to-event outcomes in cardiovascular registry.
