Stroke · Journal article

Real-world evidence on off-label underdosing of direct oral anticoagulants in Asian patients with atrial fibrillation: prescribing patterns, determinants, and outcomes.

Future Science Oa · August 3, 2026

Promising Why this matters to you Expert discussion

Encouraging direction, but not yet definitive.

AI Summary

This retrospective cohort of 553 Thai AF patients found that off-label DOAC underdosing occurred in 20.4% of cases, predicted by older age and diabetes, but was not associated with statistically significant differences in ischemic stroke/systemic embolism or bleeding compared to on-label dosing. The wide confidence intervals and small event numbers preclude firm conclusions about safety equivalence and highlight the need for larger prospective studies.

What was studied

Retrospective cohort study. Incident AF patients initiating dabigatran, rivaroxaban, apixaban, or edoxaban at two tertiary hospitals in Thailand; specific eligibility criteria beyond AF diagnosis not detailed in provided text.. Intervention: Off-label underdosed direct oral anticoagulants (lower-than-recommended dose without meeting guideline-specified dose reduction criteria). Compared with: On-label dosing of direct oral anticoagulants. n = 553. Two tertiary hospitals in Thailand.

Key findings

Off-label underdosing occurred in 20.4% of patients, with 73.2% receiving on-label dosing and 6.3% overdosed Rivaroxaban was most frequently underdosed among the four DOACs studied Older age and diabetes mellitus independently predicted off-label underdosing

Safety

Underdosing was not associated with bleeding (aHR 1.06, 95% CI 0.30–3.73; p = 0.924)

Who this may help

The finding that off-label underdosing was not associated with increased stroke or bleeding risk may appear reassuring but is based on small event counts with wide confidence intervals that cannot exclude clinically meaningful differences. Clinicians should not interpret this as evidence supporting routine dose reduction; individualized dosing decisions remain essential, and larger studies are needed.

Plain-language summary

A single-centre retrospective cohort study with modest sample size examining real-world DOAC dosing patterns and outcomes in an underrepresented Asian population; findings on safety equivalence are not powered to exclude clinically important differences.

Study design

As stated by the source record.

MethodsElectronic health records from two tertiary hospitals in Thailand (2015–2023) were reviewed. Incident AF patients initiating one of four DOACs were identified. Doses were classified as on-label, underdosed, or overdosed according to guideline criteria. Propensity score-based inverse probability of treatment weighting (PS-IPTW) with Cox proportional hazards regression was applied to compare ischemic stroke or systemic embolism and bleeding outcomes between underdosed and on-label groups.

DesignRetrospective cohort study

PopulationIncident AF patients initiating dabigatran, rivaroxaban, apixaban, or edoxaban at two tertiary hospitals in Thailand; specific eligibility criteria beyond AF diagnosis not detailed in provided text.

Sample size553

InterventionOff-label underdosed direct oral anticoagulants (lower-than-recommended dose without meeting guideline-specified dose reduction criteria)

ComparatorOn-label dosing of direct oral anticoagulants

SettingTwo tertiary hospitals in Thailand

Reported figures

Quoted from the source exactly as published.

Prevalence of underdosing20.4%

Prevalence of on-label dosing73.2%

Prevalence of overdosing6.3%

Ischemic stroke or systemic embolism adjusted hazard ratio0.48

Ischemic stroke or systemic embolism 95% CI0.11–2.16

Ischemic stroke or systemic embolism p-value0.336

Bleeding adjusted hazard ratio1.06

Bleeding 95% CI0.30–3.73

Clinical bottom line

Caveats

Retrospective design from two Thai hospitals limits generalizability beyond Thailand and Southeast Asia
Small absolute number of clinical events, particularly for ischemic stroke/systemic embolism, resulting in wide confidence intervals that span unity
Unmeasured confounding from indication bias: sicker patients may have been preferentially underdosed, biasing results toward the null
No stratified analysis or effect modification testing by specific underdosing magnitude, renal function, body weight, or individual DOAC agent reported

Evidence assessment

Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.

Study design

moderate

Retrospective cohort using electronic health records with propensity score weighting; observational design limits causal inference.

Directness

strong

Examines actual clinical outcomes (stroke, systemic embolism, bleeding) in real-world AF patients on DOACs.

Consistency

insufficient

No internal consistency data reported; single-centre study in Thailand limits generalizability claims.

Risk of bias

moderate

Retrospective design with potential unmeasured confounding; PS-IPTW mitigates some bias but does not eliminate it.

Sample adequacy

limited

553 total patients with only 20.4% underdosed; small event counts limit precision of hazard ratio estimates.

Source abstract

Background. Off-label underdosing of direct oral anticoagulants (DOACs) is common among Asian patients with atrial fibrillation (AF), partly reflecting concerns about bleeding, yet Southeast Asian data remain limited. This study aimed to characterize DOAC dosing patterns, identify predictors of off-label underdosing, and evaluate associated effectiveness and safety outcomes.Methods. We conducted a retrospective cohort using electronic health records (EHRs) from two tertiary hospitals in Thailand (2015-2023). Incident AF patients initiating dabigatran, rivaroxaban, apixaban, or edoxaban were included. Doses were classified as on-label, underdosed, or overdosed based on guideline criteria. Propensity score-based inverse probability of treatment weighting (PS-IPTW) Cox proportional hazards regression models was used to evaluate ischemic stroke or systemic embolism (ISSE) and bleeding outcomes.Results. Among 553 patients, 20.4% were underdosed, 6.3% overdosed, and 73.2% received on-label dosing. Rivaroxaban was most frequently underdosed. Older age and diabetes mellitus independently predicted underdosing. Compared to on-label dosing, underdosing was not associated with ISSE (adjusted hazard ratio [aHR] 0.48, 95% CI 0.11-2.16; p = 0.336) or bleeding (aHR 1.06, 95% CI 0.30-3.73; p = 0.924).Conclusions. Off-label underdosing occurred in one-fifth of patients but was not associated with significant differences in ISSE or bleeding. Larger, prospective studies in broader Asian and non-Asian populations are warranted.

Follow from this

Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.

Tanaphorn KoonlachadNatrawee BureekamStrokeSirayut PhatthanasobhonPurachai BvornphattanasakulPajaree MongkhonSupisara Pholprasittito