Abstract
Background
Warfarin remains the mainstay of long-term anticoagulation in low-resource settings; however, maintaining optimal international normalized ratio (INR) control is challenging, particularly in countries with limited follow-up services. In Eritrea, patients requiring lifelong anticoagulation often undergo cardiac surgery abroad yet face barriers to consistent INR monitoring upon return.
Objective
To identify demographic, clinical, and treatment-related factors associated with poor anticoagulation control among Eritrean patients on long-term warfarin following cardiac surgery.
Methods
We conducted a retrospective observational cohort study of Eritrean adults (≥18 years) who underwent cardiac surgery at the Salam Centre for Cardiac Surgery in Sudan and returned to Eritrea for follow-up. Data were collected from patients followed between February 2020 and July 2021 (median follow-up: 60 days, IQR: 24–120 days). INR control was assessed using Time in Therapeutic Range (TTR) via the Rosendaal method. Linear mixed-effects modeling identified predictors of INR variability.
Results
Ninety-three patients (37.5% male, median age 45 years) met inclusion criteria. Median TTR was 47.8% (IQR: 33.2–62.9%), with only 25% achieving optimal control (TTR ≥60%). Valvular heart disease was the leading indication (62%), with 67% having metallic valve replacement. Mixed-effects modeling identified mechanical valve replacement (exp(β)=1.26, 95% CI: 1.17–1.37), higher therapeutic target (exp(β)=1.21), COPD/asthma (exp(β)=0.71), and specific co-medications (aspirin, digoxin, enalapril) as significant determinants of INR variability.
Conclusion
In this cohort, anticoagulation control was suboptimal for the majority. Mechanical valve replacement, therapeutic targets, and specific co-medications were significant determinants of INR variability. These findings, while limited by sample size and single-center design, suggest that targeted medication review and individualized therapeutic targets may improve anticoagulation outcomes in similar low-resource settings.
Keywords: warfarin, time in therapeutic range, INR variability, mechanical heart valve, stability-adjusted TTR, Sub-Saharan Africa, low-resource setting, anticoagulation quality
Introduction
Warfarin remains the primary long-term anticoagulant in low-resource settings, particularly for patients with mechanical heart valves or other high-thrombotic-risk conditions. Its affordability and well-established effectiveness make it indispensable; however, warfarin’s narrow therapeutic window and susceptibility to drug-drug and food interactions present serious challenges for maintaining therapeutic International Normalized Ratio (INR) control. In Africa, including sub-Saharan regions, the quality of warfarin anticoagulation is consistently suboptimal. A systematic review and meta-analysis reported that only about 17% of patients achieved a TTR ≥65%, with study-specific values ranging from 10% to 29%, and a pooled therapeutic INR prevalence of around 40%.1,2 Additional studies have documented wide TTR variability across the continent, with mean TTRs ranging from approximately 13.7% to 57.3%.1,3 These findings highlight the pressing need for interventions to improve warfarin control in these settings.
Multiple patient-level and system-level factors contribute to these poor outcomes. Polypharmacy, including drug interactions with frequently used medications in Africa, presents a significant barrier.3 Moreover, poor access to INR testing, limited healthcare infrastructure, and centralized anticoagulation services further impede optimal management.1 To help identify patients at risk of poor anticoagulation control, the SAMe-TT2R2 score was developed, incorporating variables such as female sex, age under 60, comorbidities, interacting treatments, tobacco use, and non-Caucasian race. Higher scores correlate with poorer TTR.4,5 Nevertheless, some reviews suggest the effect size may be modest and not universally clinically impactful.6
Eritrea operates a predominantly public healthcare system with heavily subsidized services. While cardiac surgery is typically sponsored through international partnerships—patients in this cohort underwent surgery at the Salam Centre for Cardiac Surgery in Sudan through charitable support—long-term medication costs and monitoring services are partially borne by patients. The absence of comprehensive health insurance coverage may contribute to inconsistent INR monitoring and medication access, presenting additional challenges for anticoagulation management in this setting.
In many developing countries like Eritrea, post-cardiac surgery patients face especially daunting barriers: limited INR monitoring infrastructure, inconsistent follow-up, and frequent dosing instability, often in the context of polypharmacy and comorbid conditions. This study addresses this critical gap by identifying demographic, clinical, and treatment-related predictors of anticoagulation control among Eritrean adults on long-term warfarin post-cardiac surgery. We also employ stability-adjusted TTR (saTTR), INR Response Index (IRI), and Dose Stability Index (DSI) to capture nuanced variations in INR control beyond those measured by TTR alone.
Methods
Study Design and Setting
We conducted a retrospective observational cohort study of patients on long-term warfarin therapy. While all patients shared a common index exposure (warfarin initiation post-cardiac surgery), we examined heterogeneity in anticoagulation control as a function of patient-level predictors using repeated measures analysis. The categorization of patients into TTR <60% versus ≥60% groups in descriptive tables was performed for presentation purposes only; primary analyses used continuous outcomes. The study was conducted at Orotta National Referral and Teaching Hospital in Asmara, Eritrea, which provides tertiary-level care to patients from both urban and rural areas. Study participants had undergone cardiac surgery at the Salam Centre for Cardiac Surgery in Khartoum, Sudan, and subsequently returned to Eritrea for follow-up care. Data were collected from patients followed between February 1, 2020 and July 21, 2021.
Study Population
Inclusion criteria: Patients aged ≥18 years on warfarin therapy with at least three INR measurements and complete data on key variables (demographics, comorbidities, INR values). Exclusion criteria: Patients with missing or incomplete data on INR measurements, therapeutic range, or key covariates. No matching was performed; confounding was addressed through multivariable mixed-effects modeling.
Data Collection
Data were extracted from clinic logbooks, patient records, and structured interviews. Variables included demographic characteristics (age, gender, residence, education, occupation), clinical history (comorbidities, surgical type, valve involvement), treatment details (warfarin dose, therapeutic range goals), INR measurements, adverse events, and concomitant medications. INR measurements were performed at clinically indicated intervals, with a median inter-measurement interval of 14 days (IQR: 7–28 days). For TTR calculation using the Rosendaal method, a maximum inter-measurement gap of 56 days was permitted, consistent with published methodological standards.7 Measurements separated by >56 days were treated as separate observation periods. Drug-drug interactions between warfarin and co-administered medications were assessed using Lexicomp® Drug Interactions (Wolters Kluwer Clinical Drug Information, Inc.) Each interaction was categorized by severity: A (no known interaction), B (no action needed), C (monitor therapy), and D (consider therapy modification).
Outcome Measures
The primary outcome was Time in Therapeutic Range (TTR), calculated using the Rosendaal linear interpolation method, which estimates the proportion of time spent within the therapeutic range between consecutive INR measurements.7 Poor control was defined as TTR <60%; optimal control as TTR ≥60%. Secondary metrics included: (1) Stability-adjusted TTR (saTTR), which adjusts TTR for INR variability, providing insight into anticoagulation stability independent of time in range; (2) Dose Stability Index (DSI), reflecting the frequency of dose adjustments relative to follow-up duration, with higher values indicating more stable dosing; and (3) INR Response Index (IRI), measuring sensitivity of INR to dose changes. Detailed formulas are provided in Supplementary Appendix S1 (Exemplified in Tables S1–S3).
Statistical Analysis
Continuous variables were summarized using median and interquartile range (IQR); categorical variables were presented as counts and percentages. Between-group comparisons used Mann–Whitney U-tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables. Linear mixed-effects modeling was employed to assess associations between potential determinants and INR control, using log-transformed INR as the dependent variable. This approach accounts for repeated INR measurements within the same patient through patient-specific random intercepts, maximizing statistical power from the 1043 available observations. Time (in days) was included as a fixed covariate. Categorical variables with cell counts <5 were either combined with related categories or excluded from regression models to ensure stable coefficient estimation.
Candidate predictors were first tested in univariate models; those with p<0.20 were considered for multivariable modeling. Backward elimination was used to arrive at the final model, with variance inflation factors >5 indicating multicollinearity. Effect estimates were exponentiated (exp(β)) for interpretation as multiplicative changes in INR. Sensitivity analyses stratified models by target therapeutic range. All analyses were performed in R version 4.4.3.
Ethical Approval
Ethical approval was obtained from the Ministry of Health Research Ethics and Protocol Review Committee (Reference No.: 12/2020). All procedures adhered to the Declaration of Helsinki. Informed consent was obtained from participating patients where applicable; a waiver of consent for retrospective data was granted by the ethics committee.
Results
Demographic and Baseline Characteristics
Of 96 patients screened, 93 met inclusion criteria (Figure 1). The cohort comprised 37.5% males and 62.5% females with a median age of 45 years (IQR: 36–54). Most patients (71%) resided in the Maekel zone. Significant socioeconomic differences were observed between genders: males reported higher median monthly incomes (827 vs 0 Nakfa, p=0.035) and higher educational attainment (56% vs 39% completed secondary school, p=0.026). Approximately 65% of patients had no comorbidities beyond their cardiac condition, while 14% had two or more concurrent conditions. Detailed demographic characteristics are presented in Table 1.
Figure 1.
Flow diagram of the overall patients’ Time in Therapeutic Range (TTR). aSuperscripts – Excluded due to lack of more than 3 consecutive INR measurements.
Table 1.
Demographic Characteristics of the Study Individuals Following in Orotta Cardiac Clinic
| Variable | Overall, N=96a | Male, N=36a | Female, N=60a | p-valueb |
|---|---|---|---|---|
| Age (years) | 45 (36, 54) | 46 (37, 55) | 45 (36, 54) | 0.9 |
| Zonal Address | >0.9 | |||
| Maekel | 68 (71%) | 26 (72%) | 42 (70%) | |
| Debub | 16 (17%) | 7 (19%) | 9 (15%) | |
| Other | 12 (13%) | 3 (8.3%) | 9 (15%) | |
| Educational Level | 0.026 | |||
| No formal/Primary | 20 (21%) | 3 (8.3%) | 17 (28%) | |
| Junior school | 22 (23%) | 6 (17%) | 16 (27%) | |
| Secondary/Above | 53 (56%) | 27 (75%) | 26 (44%) | |
| Monthly income (Nakfa) | 600 (0, 1700) | 827 (0, 1800) | 0 (0, 1375) | 0.035 |
| Comorbidities present | 34 (35%) | 17 (47%) | 17 (28%) | 0.061 |
| Family size | 5 (3, 7) | 6 (4, 7.25) | 4 (3, 6) | 0.11 |
Notes: aMedian (IQR); n (%). bWilcoxon rank sum test; Fisher’s exact test; Pearson’s Chi-squared test. Bold values indicate statistical significance at p < 0.05.
Clinical Indications and Anticoagulation Characteristics
Valvular heart disease was the most common indication for warfarin (62%), followed by valvular disease with atrial fibrillation (25%). The majority of patients (75%) had undergone cardiac surgery, predominantly metallic valve replacement (67%). Mitral valve procedures were more common in females (65% vs 50%), whereas aortic valve interventions were more frequent in males (42% vs 25%). The median duration of warfarin therapy was 60 days (IQR: 24–120 days). INR targets were set at 2.0–3.0 in 43% of patients and 2.5–3.5 in 57% (Table 2).
Table 2.
Clinical Indication and Characteristics of Individuals on Vitamin-K Antagonists
| Characteristic | Overall, N=93a | Male, N=36a | Female, N=57a | p-valueb |
|---|---|---|---|---|
| Indications | 0.3 | |||
| VHD | 58 (62%) | 25 (69%) | 33 (58%) | |
| VHD & AF | 23 (25%) | 7 (19%) | 16 (28%) | |
| AF alone | 8 (8.6%) | 4 (11%) | 4 (7.0%) | |
| DVT & PE | 4 (4.3%) | 0 (0%) | 4 (7.0%) | |
| Interventional status | 0.3 | |||
| Intervened | 70 (75%) | 29 (81%) | 41 (72%) | |
| Not Intervened | 23 (25%) | 7 (19%) | 16 (28%) | |
| Metallic replacement | 62 (67%) | 25 (69%) | 37 (66%) | 0.7 |
| Warfarin Duration (days) | 60 (24, 120) | 76 (21, 132) | 60 (24, 87) | 0.2 |
| Goal Therapeutic Range | 0.13 | |||
| A (2.0–3.0) | 40 (43%) | 19 (53%) | 21 (37%) | |
| B (2.5–3.5) | 53 (57%) | 17 (47%) | 36 (63%) | |
| TTR, median (IQR) | 47.8 (33.2, 62.9) | 55.2 (35.5, 69.9) | 41.7 (31.0, 60.1) | 0.14 |
| Adverse event present | 45 (49%) | 14 (39%) | 31 (55%) | 0.12 |
Notes: an (%); Median (IQR). bFisher’s exact test; Pearson’s Chi-squared test; Wilcoxon rank sum test.
Abbreviations: VHD, Valvular heart disease; AF, Atrial fibrillation; DVT, Deep vein thrombosis; PE, Pulmonary embolism; TTR, Time in therapeutic range.
Medication Profile
Ninety percent of patients received concomitant medications, with a median of 4 medications (IQR: 3–6). Warfarin doses ranged from 2.5 to 15 mg daily (median: 5 mg, IQR: 3.75–7.5 mg). Drug interaction analysis using Lexicomp classifications identified Category C (monitor therapy) interactions in 67% of patients and Category D (consider modification) interactions in 23%. The most common interacting medications were aspirin (58%), digoxin (42%), omeprazole (38%), and enalapril (35%).
Anticoagulation Control
Analysis of 1043 INR measurements revealed a median TTR of 47.8% (IQR: 33.2–62.9%), with only 25% of patients achieving optimal control (TTR ≥60%). Males demonstrated numerically higher TTR (55.2%, IQR: 35.5–69.9%) than females (41.7%, IQR: 31.0–60.1%), though this difference was not statistically significant (p=0.14). INR values showed a bimodal distribution with peaks at 2.3 and 3.1 (Hartigan’s dip test: D=0.022, p=0.0019), corresponding to the two target ranges. Patients managed within the standard range (2.0–3.0) exhibited 2.5-fold greater variability (σ2=0.069) compared to those in the high-intensity range (2.5–3.5; σ2=0.028, p=0.03), suggesting more stable anticoagulation control at higher targets.
Adverse Events
Adverse events were reported in 45 patients (49%), with bleeding events predominating. Bleeding manifestations included ecchymosis (n=22, 24%), gingival bleeding (n=15, 16%), epistaxis (n=8, 9%), hematuria (n=5, 5%), and gastrointestinal bleeding (n=3, 3%). Thrombotic events occurred in 3 patients (3%): ischemic stroke (n=2) and peripheral embolism (n=1). In mixed-effects modeling, ecchymosis was significantly associated with elevated INR (exp(β)=1.14, 95% CI: 1.02–1.27, p=0.023), consistent with its role as a clinical marker of over-anticoagulation.
Determinants of INR Variability
Mixed-effects modeling with log-transformed INR identified several significant predictors of anticoagulation intensity (Table 3). Patients with mechanical valve replacement exhibited 26% higher INR levels (exp(β)=1.26, 95% CI: 1.17–1.37) compared to non-recipients. Mitral valve procedures consistently elevated INR by 22–24%. Patients without surgical intervention maintained 18% lower INR values (exp(β)=0.82, 95% CI: 0.75–0.90). Therapeutic targets significantly influenced INR, with patients targeting 2.5–3.5 demonstrating 21% higher values (exp(β)=1.21, 95% CI: 1.12–1.31). Among comorbidities, COPD/asthma was associated with 29% lower INR (exp(β)=0.71, 95% CI: 0.58–0.86). Regarding medications, aspirin increased INR by 11% (exp(β)=1.11, 95% CI: 1.02–1.20), while digoxin (exp(β)=0.84, 95% CI: 0.76–0.92) and enalapril (exp(β)=0.88, 95% CI: 0.79–0.97) were associated with lower INR values.
Table 3.
Multivariable Mixed-Effect Model-Based Determinants of INR Variability
| Variable | exp(β) [95% CI] | p-value |
|---|---|---|
| Interventional Factors | ||
| Not Intervened (vs Intervened) | 0.82 [0.75–0.90] | <0.001 |
| Metallic Replacement (Yes vs No) | 1.26 [1.17–1.37] | <0.001 |
| Mitral Valve (Yes vs No) | 1.22 [1.13–1.32] | <0.001 |
| MV+AV (vs AV only) | 1.17 [1.02–1.35] | 0.030 |
| MV+AV+TV (vs AV only) | 1.24 [1.02–1.52] | 0.031 |
| Therapeutic Targets | ||
| INR Range B (2.5–3.5 vs A) | 1.21 [1.12–1.31] | <0.001 |
| Comorbidities & Medications | ||
| COPD/Asthma (Yes vs No) | 0.71 [0.58–0.86] | <0.001 |
| Digoxin (Yes vs No) | 0.84 [0.76–0.92] | <0.001 |
| Aspirin (Yes vs No) | 1.11 [1.02–1.20] | 0.019 |
| Enalapril (Yes vs No) | 0.88 [0.79–0.97] | 0.011 |
| Category A Interactions (per one) | 0.95 [0.92–0.99] | 0.006 |
| Demographics | ||
| Military/Service (vs Student) | 1.60 [1.06–2.43] | 0.027 |
| Adverse Events | ||
| Ecchymosis (Yes vs No) | 1.14 [1.02–1.27] | 0.023 |
Note: Bold p-values indicate significant at p values less than or equal to 0.05.
Abbreviations: exp(β), exponentiated coefficient, interpreted as multiplicative change in INR. CI, confidence interval. MV, Mitral valve; AV, Aortic valve; TV, Tricuspid valve; COPD, Chronic obstructive pulmonary disease.
Stability Metrics and Control Phenotypes
Analysis of saTTR against TTR revealed a strong correlation (Pearson’s r=0.89, 95% CI: 0.83–0.92, p<0.001). However, heterogeneity in saTTR values among patients with similar TTR, particularly at higher TTR levels (>70%), suggested that TTR may mask underlying instability. The density heatmap of DSI versus saTTR (Figure 2) revealed four distinct phenotypes: (1) unstable dosing with poor control (low DSI/low saTTR), representing high-risk patients; (2) erratic dosing with paradoxically stable INR (low DSI/high saTTR); (3) stable dosing with poor control (high DSI/low saTTR), suggesting pharmacokinetic mismatches; and (4) stable dosing with excellent control (high DSI/high saTTR), representing optimal management.
Figure 2.
Density heatmap of Stability adjusted Time in Therapeutic Range (saTTR) and Dose Stability Index (DSI).
Discussion
Principal Findings
In this retrospective cohort of 93 Eritrean patients receiving long-term warfarin therapy following cardiac interventions, anticoagulation control was suboptimal. The median TTR of 47.8% falls below the 60% threshold generally considered necessary for optimal outcomes, with only 25% of patients achieving adequate control. Clinical guidelines indicate that vitamin K antagonists deliver maximal efficacy and safety when TTR is maintained at or above 70%, suggesting that a substantial proportion of our cohort may face elevated clinical risk.
Mixed-effects modeling identified mechanical valve replacement, mitral valve involvement, and higher therapeutic targets as associated with increased INR intensity—findings consistent with established guidelines recommending higher targets and closer monitoring in these subgroups. COPD/asthma was associated with lower INR values, potentially reflecting disease-related metabolic changes or concurrent therapies.
The significant associations between specific medications and INR variability have clinical implications. Aspirin’s modest positive effect on INR (11% increase) may reflect anticoagulant synergy and increased bleeding risk. The negative associations with digoxin (16% reduction) and enalapril (12% reduction) align with prior reports of altered warfarin metabolism, though mechanisms remain incompletely understood.8,9 These findings support the need for systematic medication review in this population.
The integration of saTTR and DSI revealed control phenotypes not captured by TTR alone. Patients with stable dosing but poor control (high DSI/low saTTR) may have pharmacokinetic mismatches requiring dose optimization, while those with unstable dosing and poor control represent the highest-risk group requiring intensive intervention.
Comparison with Existing Literature
Our TTR results mirror reports from other low- and middle-income settings, where median TTR values frequently fall below those observed in high-income countries. Studies from Ethiopia3 and pooled analyses across Sub-Saharan Africa2 have described median TTRs ranging from the mid-teens to approximately 30%. Our cohort’s somewhat higher TTR (47.8%) may reflect the specialized nature of post-surgical follow-up at a tertiary referral center. The association between mechanical mitral valves and higher target INRs is well established in contemporary guidelines.10,11
Strengths and Limitations
Strengths of this study include the integration of demographic, clinical, procedural, and medication-related data with both established and novel stability metrics. The use of mixed-effects modeling maximized statistical power by utilizing all 1043 INR observations, and the saTTR-DSI framework captured aspects of anticoagulation control not visible in TTR alone.
Several limitations should be acknowledged. The modest sample size (n=93) limits statistical power and generalizability. While we had 1043 INR observations enabling mixed-effects analysis, the number of patients constrains our ability to detect smaller effect sizes or examine rare subgroups. Post-hoc power analysis indicated 80% power to detect a TTR difference of 12 percentage points between groups (α=0.05), suggesting adequate power for large effects but limited ability to detect smaller clinically meaningful differences. This was a single-center study from a tertiary referral hospital, which may not represent patients managed in peripheral settings. Selection bias may exist, as only patients who survived and returned for follow-up were included. Data on dietary vitamin K intake, medication adherence, renal function, and pharmacogenetic variants (VKORC1, CYP2C9) were not collected; these factors are known to influence warfarin requirements and variability. Patient knowledge data had >20% missingness and were excluded from analysis. Finally, clinical outcomes such as major bleeding, thromboembolism, and mortality were not assessed, precluding direct linkage between observed control phenotypes and patient outcomes.
Implications for Practice
These findings suggest several practical considerations for anticoagulation management in similar settings. Systematic medication review, particularly for patients on aspirin, digoxin, or ACE inhibitors, may help anticipate INR effects. The saTTR-DSI framework could help stratify patients for follow-up intensity: those with stable dosing but poor control require dose optimization, while those with unstable dosing and poor control need comprehensive intervention addressing adherence, diet, and drug interactions.
Conclusions
In this cohort of Eritrean post-cardiac surgery patients, anticoagulation control was suboptimal (median TTR 47.8%), with only 25% achieving optimal control. Mixed-effects modeling identified mechanical valve replacement, higher therapeutic targets, and specific co-medications (aspirin, digoxin, enalapril) as significant determinants of INR variability. The integration of stability-adjusted metrics (saTTR, DSI) revealed distinct control phenotypes not captured by TTR alone. These findings, while limited by sample size and single-center design, suggest that targeted interventions focusing on medication review and individualized therapeutic targets may improve anticoagulation outcomes. Validation in larger, multi-center cohorts is needed before broad implementation, and findings may not be directly generalizable to settings with different healthcare infrastructure or patient populations.
Acknowledgments
We thank the patients, staff, and families associated with Orotta Cardiac Clinic for their cooperation.
Funding Statement
This study did not receive any funding.
Data Sharing Statement
The R scripts used for data cleaning, calculation of TTR, saTTR, IRI, and statistical modeling are publicly available on GitHub at: https://github.com/teklesam/Warfarin_INR_Stability. Individual patient data may be available from the corresponding author upon reasonable request, subject to ethical approval.
Author Contributions
All authors made significant contributions to the work reported, including conception, study design, execution, acquisition of data, analysis and interpretation; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare no conflicts of interest in this work.
References
- 1.Tadesse TA, Tegegne GT, Yadeta D, Chelkaba L, Fenta TG. Anticoagulation control, outcomes, and associated factors in long-term-care patients receiving warfarin in Africa: a systematic review. Thromb J. 2022;20(1):58. doi: 10.1186/s12959-022-00416-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Demsie DG, Asres G, Tegegne GT, et al. Anticoagulation quality through time in therapeutic range in Sub-Saharan Africa: a systematic review and meta-analysis. Front Med. 2025;12:1517162. doi: 10.3389/fmed.2025.1517162 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Teka FS, Legese GL, Wakjira GB, et al. Anticoagulation quality with warfarin therapy, and associated factors among adult outpatients at public hospitals in Nekemte town, western Ethiopia: a retrospective study. Front Pharmacol. 2025;16:1500892. doi: 10.3389/fphar.2025.1544957 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Apostolakis S, Sullivan RM, Olshansky B, Lip GYH. Factors affecting quality of anticoagulation control among patients with atrial fibrillation on warfarin. Chest. 2013;144(5):1555–9. doi: 10.1378/chest.13-0054 [DOI] [PubMed] [Google Scholar]
- 5.Chan PH, Li WH, Sioa-O JY, et al. Use of the SAMe-TT2R2 score to predict good anticoagulation control with warfarin in Chinese patients with atrial fibrillation: relationship to ischemic stroke incidence. PLoS One. 2016;11(3):e0150674. doi: 10.1371/journal.pone.0150674 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Roldán V, Cancio S, Gálvez J, et al. The SAMe-TT2R2 score predicts poor anticoagulation control in AF patients: a prospective ‘real-world’ inception cohort study. Am J Med. 2015;128(11):1237–1243. doi: 10.1016/j.amjmed.2015.05.036 [DOI] [PubMed] [Google Scholar]
- 7.Rosendaal FR, Cannegieter SC, van der Meer FJ, Briët E. A method to determine the optimal intensity of oral anticoagulant therapy. Thromb Haemost. 1993;69(03):236–239. doi: 10.1055/s-0038-1651587 [DOI] [PubMed] [Google Scholar]
- 8.Bhattacharyya A, Bhavnani M, Tymms DJ. Serious interaction between digoxin and warfarin. Br J Cardiol. 2002;9:356–358. [Google Scholar]
- 9.Del Campo M, Roberts G. Changes in warfarin sensitivity during decompensated heart failure and chronic obstructive pulmonary disease. Ann Pharmacother. 2015;49(9):962–968. doi: 10.1177/1060028015590438 [DOI] [PubMed] [Google Scholar]
- 10.Vahanian A, Beyersdorf F, Praz F, et al. 2021 ESC/EACTS guidelines for the management of valvular heart disease. Eur Heart J. 2022;43(7):561–632. doi: 10.1093/eurheartj/ehab395 [DOI] [PubMed] [Google Scholar]
- 11.Brandt RR, Pibarot P. Prosthetic heart valves: part 2 – antithrombotic management. E-J Cardiol Pract. 2022;20:23. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The R scripts used for data cleaning, calculation of TTR, saTTR, IRI, and statistical modeling are publicly available on GitHub at: https://github.com/teklesam/Warfarin_INR_Stability. Individual patient data may be available from the corresponding author upon reasonable request, subject to ethical approval.