Anticoagulation Therapeutic Ranges and Clinical Outcomes in Patients with a Mechanical Heart Valve Treated with Vitamin K Antagonists—a Nationwide Linked-data Dutch Study

Abstract

Aims

To examine the impact of different therapeutic international normalized ratio (INR) ranges on anticoagulation control and clinical outcomes in patients with mechanical heart valves (MHVs) treated with vitamin K antagonists (VKAs) in the Netherlands.

Methods

Data from 17 anticoagulation clinics (2013–2019) were linked to nation-wide data from Statistics Netherlands. Anticoagulation control metrics included significant dose adjustments, INR variance growth rate, and time in therapeutic range. Cause-specific Cox regression models were used to assess associations between therapeutic ranges and clinical outcomes, accounting for death as competing risk. Stratified analyses were performed for significant interactions by type of MHV recipient.

Results

Among 3,473 MHV patients (median age: 67.0 [IQR: 58.0-76.0], 61.7% male, 68.2% acenocoumarol, 26.5% phenprocoumon), patients with lower therapeutic ranges (N = 1,866) (2.0–3.0 for isolated aortic valve without risk factors; 2.5–3.5 for all remaining MHV patients) had poorer anticoagulation control compared to those with higher ranges (N = 1,607) (2.5–3.5 and 3.0–4.0, respectively). No association was found between therapeutic ranges and major/clinically relevant bleeding (fully adjusted hazard ratio [aHR]: 0.80 [95%CI: 0.57–1.1]). However, in patients with a non-aortic valve and/or additional risk factors a lower therapeutic range was potentially associated with increased thromboembolic risk (aHR: 1.3 [95%CI: 0.94–1.9]), while no association was observed in patients with an isolated aortic valve (aHR: 0.71 [95%CI: 0.38–1.3]).

Conclusion

A lower therapeutic range does not apparently increase thromboembolic risk in most MHV patients but may be associated with a higher thromboembolic risk in higher risk patients. Lower therapeutic ranges were not associated with lower bleeding risk.

Keywords

heart valve prosthesis - international normalized ratio - vitamin K antagonists - thromboembolism - bleeding

Introduction

The prevalence of valvular heart disease (VHD) has grown exponentially over the past two decades because of improved survival rates and an aging population.[1] [2] [3] In patients with severe VHD, replacement of the affected valve with a valve prosthesis is essential to improve survival and quality of life.[4] [5] Young patients without contraindications for long-term anticoagulation usually receive mechanical heart valves (MHVs), rather than bioprostheses, and require long-term anticoagulation with vitamin K antagonists (VKAs).[4] [5] Optimizing VKA treatment, measured by the international normalized ratio (INR) and titrated based on the therapeutic INR range, is crucial to balance the risks of thrombosis and bleeding.[6] Different types of VKAs are available, which vary in their half-life; acenocoumarol has the shortest, followed by warfarin and phenprocoumon.[7] [8]

The ideal therapeutic INR range for MHV patients is not fully determined but depends on prosthesis and patient characteristics.[4] [5] Historically, patients in the Netherlands were managed with higher therapeutic ranges compared to current international guidelines. Specifically, therapeutic ranges were set at 2.5 to 3.5 (instead of 2.0–3.0) for patients with an isolated aortic MHV without additional risk factors, and 3.0 to 4.0 (instead of 2.5–3.5) for all other MHV patients. This approach was based on a large observational study showing that a higher therapeutic range resulted in an optimal clinical benefit of VKA treatment.[9] From January 2016 onwards, therapeutic ranges were lowered to 2.0 to 3.0 and 2.5 to 3.5, respectively, to align with international guidelines. Although several randomized controlled trials (RCTs) have compared INR ranges in patients with varying MHV types and positions,[10] [11] [12] [13] [14] the impact of different therapeutic ranges, in particular lower INR ranges, on the quality of anticoagulation control and clinical outcomes in MHV patients remains a topic of debate.[4] [15] [16]

Against this background, we aimed to (1) describe the quality of anticoagulation control in MHV patients treated with high versus low therapeutic ranges and (2) investigate the association between these ranges and clinical outcomes in a large cohort of patients with MHV in the Netherlands.

Methods

Data Sources

Our cohort study used data obtained from 17 Dutch anticoagulation clinics, linked on an individual-level to nation-wide data from Statistics Netherlands (“Centraal Bureau voor de Statistiek,” CBS). Anticoagulation clinics monitor patients receiving VKAs and provide detailed information on VKA therapy, including INR measurements, therapeutic INR ranges, type and dosage of VKA, and treatment indications. A description of the data sources is presented in [Supplementary Text S1] (available in the online version only), and variable identification and extraction methods are outlined in [Supplementary Tables S1]–[S3] (available in the online version only). The study received approval from the Scientific Committee of the Department of Clinical Epidemiology of the Leiden University Medical Centre (A0168). A waiver of participant consent was granted due to the use of pre-existing, de-identified data.

Participants

The source population included patients who received an MHV (i.e., incident MHV recipients) and were treated with VKA managed by one of the participating anticoagulation clinics between January 1, 2013, and December 31, 2019. MHV patients were identified by nationwide hospitalization data with the ICD-9 code V43.3 and the ICD-10 codes Z95.2 and Z95.4. A 3-year look-back period (i.e., 2010 until 2013) was applied to ensure that the first recorded admission for MHV was identified. The discharge date of the first MHV hospitalization was considered the index date.

We validated the obtained cohort by cross-referencing MHV hospitalizations with registered treatment indications from the anticoagulation clinics. Although the exact date of MHV replacement was unavailable, these indications confirmed VKA treatment for an MHV according to clinic records. To minimize misclassification, we included patients identified as having an MHV through both the nationwide hospitalization data and the anticoagulation clinic records. Exclusion criteria were missing therapeutic range at first INR measurement after index date (i.e., discharge date of the first MHV hospitalization), therapeutic range other than 2.0 to 3.0, 2.5 to 3.5, or 3.0 to 4.0, and no INR measurement recorded within 6 weeks post-discharge.

Therapeutic Ranges

The higher therapeutic ranges, as were used in the Netherlands before 2016, were defined as 2.5 to 3.5 for patients with an isolated aortic MHV without additional risk factors and 3.0 to 4.0 for all remaining MHV patients, e.g., aortic valve (AV) recipients with additional risk factors or a higher-thrombogenicity prosthesis, and those with a mitral or tricuspid valve prosthesis.[16] The lower therapeutic ranges, which were already used internationally and implemented from January 2016 onwards in the Netherlands, were defined as 2.0 to 3.0 and 2.5 to 3.5, respectively.[16] Each patient's therapeutic INR range (i.e., higher versus lower) was determined by the range registered at their first INR measurement after index date and treated as a fixed exposure. Patients were censored when their therapeutic range changed or became missing.

Type of MHV Recipient

As information on prosthesis type, model, and implantation site was unavailable, patients were classified in two MHV recipient categories based on the therapeutic INR range at index and the calendar year of the index date. Patients with a range of 2.5 to 3.5 before 2016 or 2.0 to 3.0 from 2016 onwards were classified as isolated AV recipients. Patients with a therapeutic range of 3.0 to 4.0 or 2.5 to 3.5, respectively, were classified as medium-risk MHV recipients.

Covariates

Baseline characteristics included sex, age, standardized household income, migration background (i.e., native Dutch, first or second-generation migration background), prior use of oral anticoagulants, heparins and antiplatelet drugs, registered VKA treatment indications, type of VKA, prior medication use on index date, and presence of comorbidities, including atrial fibrillation (AF) and heart failure ([Supplementary Tables S2] and [S3], available in the online version only).

Outcomes

We studied the quality of anticoagulation control and clinical outcomes. Metrics for the quality of anticoagulation control included percentage of INRs within the therapeutic range, percentage of significant dose adjustments, time spent within, above, and below therapeutic range (TTR/TAR/TBR)[17] and the INR variance growth rate (VGR) according to Cannegieter.[18] For INR measurements recorded before 2016, calculations of the TTR, TAR, TBR, and percentage of INRs within the therapeutic range were based on wider acceptable INR intervals used by anticoagulation clinics, as this reflects clinical practice at that time. A therapeutic INR range of 2.5 to 3.5 corresponded to an acceptable INR interval of 2.0 to 3.5 and a therapeutic range of 3.0 to 4.0 to an interval of 2.5 to 4.0. After 2016, the acceptable INR intervals aligned with the newly implemented lower therapeutic ranges. TTR was calculated according to the Rosendaal method, by which daily INR values between two INR measurements are predicted through linear interpolation,[17] dividing the days with interpolated INR values within the therapeutic range (after 2016) or within the wider acceptable INR interval (before 2016) by the total number of days. TAR and TBR were calculated similarly.

The VGR reflects the variability between consecutive INR measurements, assuming stability when INRs remain around the same value, even if consistently outside the therapeutic INR range.[18] A significant dose adjustment was defined as a change of ≥10% in the average daily prescribed VKA dose at two consecutive INR measurements. These metrics were computed for each 6-month period following the index date for patients with at least two INR measurements within that period.

Clinical outcomes included all-cause mortality, cardiovascular mortality, thromboembolism (TE), and major and clinically relevant bleeding. TE was defined as the first hospitalization with a primary diagnosis of stroke (including transient ischemic attack [TIA]), myocardial infarction (MI), peripheral arterial thromboembolism (ATE), venous thromboembolism (VTE), or death due to TE ([Supplementary Table S2], available in the online version only). Major and clinically relevant bleeding was defined as the first registered hospital admission for any type of bleeding or death from bleeding ([Supplementary Table S2], available in the online version only). Because laboratory values and transfusion data were unavailable, we could not distinguish International Society on Thrombosis and Haemostasis (ISTH)-defined major bleeding from clinically relevant nonmajor bleeding,[19] [20] However, by definition, all hospital-treated bleeding events meet at least the ISTH criteria for clinically relevant nonmajor bleeding. Patients were followed from their index date until the event of interest, a change in therapeutic range, discontinuation of VKA, death, or December 31, 2019, whichever occurred first.

Statistical Analysis

Incidence rates of all-cause death, first major and clinically relevant bleeding, and first TE, and the separate categories of ischemic stroke, MI, and other ATE were calculated as the number of first events divided by the total observation time and expressed per 100 person-years (PY). Events or categories containing fewer than 10 individuals were masked following CBS privacy policy. Median follow-up time was estimated by the Kaplan-Meier estimate, as suggested by Schemper and Smith.[21]

We used multivariable cause-specific Cox regression models, accounting for death from any other cause as a competing risk, to explore associations between therapeutic range and clinical outcomes. Model assumptions are summarized in the directed acyclic graph ([Supplementary Fig. S1] and [Supplementary Text S2] [available in the online version only]). First, we included the different index years in model 1, with 2013 as reference year, assuming that calendar year was the most important determinant of therapeutic INR range. In models 2 to 6, therapeutic range (low vs. high [reference]) was used as exposure, and potential confounders were progressively added: (2) the crude model; (3) model 2 plus sex and age at index; (4) model 3 plus type of MHV recipient (medium-risk MHV vs. isolated AV recipients); (5) model 4 plus migration background and household income; (6) model 5 plus presence of comorbidity and prior antiplatelet drug use. Because household income contained a small proportion of missing values (0.5% of all patients), we performed a complete case analysis for models 5 and 6.

Scaled Schoenfeld residual plots were used to assess the proportional hazard assumption. Natural cubic splines were included into the final model if statistically significant (p-value <0.15), to relax the assumption of linearity between age and clinical outcomes. Interaction terms between sex (model 3 to 6), type of MHV recipient (model 4 to 6) and therapeutic INR range were added to examine the possible multiplicative effect modification by both sex and type of MHV recipient. Subgroup analyses based on sex or MHV recipient type were performed when the individual interaction term was statistically significant (p-value <0.05).

Analyses were performed in IBM SPSS Statistics (version (v.) 25.0, IBM corp.) and the R Statistical Software (v. 4.4.0, R Foundation for Statistical Computing),[22] in the R Studio environment (v. 2024.4.1.748),[23] with the Survival (v. 3.6.4),[24] [25] Splines (v. 3.6.4),[22] EpiR (v. 2.0.74),[26] and prodlim (v. 2023.8.28)[27] packages. Additional packages are listed in [Supplementary Text S3] (available in the online version only).

Sensitivity Analyses

We performed several sensitivity analyses to assess the robustness of our findings. First, hospital admission periods were excluded from the TTR calculation to address potential time-related bias. Second, the censoring period was extended by 7 and 14 days for patients who changed therapeutic ranges or discontinued VKA treatment to capture clinical outcomes occurring shortly after these changes, which could still be associated with the previous recorded therapeutic range. Third, patients identified by the ICD-10 code “Z954 - Presence of other heart-valve replacement,” which is a code potentially given when the type of heart valve is unknown, were excluded to evaluate potential misclassification. Finally, we performed two sensitivity analyses with alternative censoring strategies: (1) therapeutic INR range was modeled as a time-varying exposure (i.e., 2.0–3.0, 2.5–3.5, or 3.0–4.0), with patients censored at VKA discontinuation, death, end of follow-up, event of interest, or a switch to an INR range outside these three main categories; (2) calendar time was used as a proxy for therapeutic range (before vs. after 2016), with patients censored at VKA discontinuation, death, end of follow-up, or event of interest. For both analyses, we used cause-specific Cox models, while accounting for clustering within patients, to assess the crude effects as well as the possible effect modification by both sex and type of MHV recipient for all three outcomes.

Results

Participants

We identified 14,323 VKA users with an MHV hospitalization between 2013 and 2019 who were treated by one of the participating anticoagulation clinics ([Fig. 1]). After cross-validation, 6,975 VKA users were confirmed to have both an MHV hospitalization and a registered MHV treatment indication by the anticoagulation clinic, while 7,348 patients had an MHV hospitalization without corresponding treatment indication in the anticoagulation clinic record. Baseline characteristics of our study population, patients with only an MHV hospitalization and patients with only an MHV treatment indication, are summarized in [Supplementary Table S4] (available in the online version only).

Fig. 1 Flowchart of the patient selection and results of cross-validation. This flowchart displays the cohort selection process. First, we identified hospital admissions with the ICD-10 codes Z952 and Z954, which we merged to the data from the anticoagulation clinics (top left of the flowchart). For the cross-validation step, we identified patients with mechanical heart valve (MHV) as registered vitamin K antagonist (VKA) treatment indication from the anticoagulation clinic records (top right of the flowchart) and merged them to the patients with a hospital admission with one of the ICD codes. This resulted in three groups: patients with an MHV hospitalization but without anticoagulation clinic treatment indication (“only CBS”), patients with an anticoagulation clinic treatment indication but without an MHV hospitalization (“only anticoagulation clinics”), and our final study cohort (“both”). Finally, we applied the eligibility criteria. INR, international normalized ratio. (Created in BioRender. Kruip, M. (2025) https://BioRender.com/o80j477.)

After applying the exclusion criteria, 3,473 patients were included ([Fig. 1]), of whom 1,607 (46.3%) were treated with higher therapeutic ranges (2.5–3.5 or 3.0–4.0) and 1,866 (53.7%) with lower therapeutic ranges (2.0–3.0 or 2.5–3.5) ([Table 1] and [Supplementary Table S5], available in the online version only). The median age was 67.0 (interquartile range [IQR]: 58.0–76.0) and 61.7% were male. At index, 2,370 patients (68.2%) used acenocoumarol, 920 (26.5%) phenprocoumon, and 183 (5.3%) warfarin or an unknown VKA type. The median follow-up was 1.4 (IQR: 0.6–2.5) years ([Supplementary Table S6], available in the online version only). As expected, the distribution of patients per calendar year varied greatly by therapeutic range due to the change in ranges implemented in 2016. However, most baseline characteristics were comparable across therapeutic ranges, also within MHV recipient groups ([Table 1]).

Table 1 Baseline characteristics of the cohort stratified by therapeutic range and type of MHV recipient

		Total (N = 3,473)		Higher therapeutic ranges (N = 1,607)		Lower therapeutic ranges (N = 1,866)
		Medium-risk MHV[a] (N = 2,393)	Isolated AV[a] (N = 1,080)	Medium-risk MHV[a] (N = 1,119)	Isolated AV[a] (N = 488)	Medium-risk MHV[a] (N = 1,274)	Isolated AV[a] (N = 592)
Demographics
Median age in years [Q1, Q3]		68.0 [59.0, 77.0]	65.0 [56.0, 75.0]	68.0 [60.0, 76.0]	65.0 [57.0, 74.0]	69.0 [59.0, 77.0]	65.0 [56.0, 76.0]
Sex, male (n, %)		1,442 (60.3)	702 (65.0)	658 (58.8)	313 (64.1)	784 (61.5)	389 (65.7)
Year of diagnosis
	2013 (n, %)	339 (14.2)	140 (13.0)	339 (30.3)	132 (27.0)	Masked	Masked
	2014 (n, %)	362 (15.1)	212 (19.6)	362 (32.4)	190 (38.9)	Masked	Masked
	2015 (n, %)	367 (15.3)	156 (14.4)	367 (32.8)	149 (30.5)	Masked	Masked
	2016 (n, %)	337 (14.1)	147 (13.6)	29 (2.6)	17 (3.5)	308 (24.2)	130 (22.0)
	2017 (n, %)	328 (13.7)	123 (11.4)	Masked	Masked	325 (25.5)	123 (20.8)
	2018 (n, %)	375 (15.7)	158 (14.6)	Masked	Masked	360 (28.3)	158 (26.7)
	2019 (n, %)	285 (11.9)	144 (13.3)	Masked	Masked	281 (22.1)	144 (24.3)
VKA treatment
Registered indications for VKA therapy[b]
	Biological valve and other heart surgery (n, %)	100 (4.2)	28 (2.6)	57 (5.1)	13 (2.7)	43 (3.4)	15 (2.5)
	Atrial fibrillation and other arrhythmias (n, %)	544 (22.7)	159 (14.7)	285 (25.5)	73 (15.0)	259 (20.3)	86 (14.5)
	Congestive heart failure, cardiomyopathy and other heart defects (n, %)	80 (3.3)	11 (1.0)	44 (3.9)	Masked	36 (2.8)	Masked
	Cerebral vascular disease (n, %)	111 (4.6)	37 (3.4)	63 (5.6)	21 (4.3)	48 (3.8)	16 (2.7)
	Coronary syndrome and vascular interventions (n, %)	175 (7.3)	25 (2.3)	104 (9.3)	11 (2.3)	71 (5.6)	14 (2.4)
	Other (n, %)	72 (3.0)	38 (3.5)	27 (2.4)	19 (3.9)	45 (3.5)	19 (3.2)
Type of VKA
	Acenocoumarol (n, %)	1,616 (67.5)	754 (69.8)	704 (62.9)	315 (64.5)	912 (71.6)	439 (74.2)
	Phenprocoumon (n, %)	650 (27.2)	270 (25.0)	321 (28.7)	138 (28.3)	329 (25.8)	132 (22.3)
	Unknown or warfarin (n, %)	127 (5.3)	56 (5.2)	94 (8.4)	35 (7.2)	33 (2.6)	21 (3.5)
Type of INR monitoring
	At home	780 (32.6)	411 (38.1)	348 (31.1)	178 (36.5)	432 (33.9)	233 (39.4)
	Care home	20 (0.8)	<10	Masked	<10	<10	<10
	Outpatient	1,088 (45.5)	501 (46.4)	531 (47.5)	233 (47.7)	557 (43.7)	268 (45.3)
	Self-dosing/management	241 (10.1)	82 (7.6)	121 (10.8)	36 (7.4)	120 (9.4)	46 (7.8)
	Self-measurement	208 (8.7)	53 (4.9)	84 (7.5)	22 (4.5)	124 (9.7)	31 (5.2)
	Unknown	56 (2.3)	Masked	23 (2.1)	Masked	Masked	Masked
Median time between index date and INR measurement in days [Q1, Q3]		3.0 [2.0, 4.0]	3.0 [2.0, 4.0]	3.0 [2.0, 4.0]	3.0 [2.0, 4.0]	3.0 [1.0, 4.0]	3.0 [2.0, 5.0]
Comorbidities and medication use
≥1 comorbidities present at index date[c]		855 (35.7)	412 (38.1)	346 (30.9)	189 (38.7)	509 (40.0)	223 (37.7)
	Autoimmune disease or immune deficiency (n, %)	53 (2.2)	17 (1.6)	22 (2.0)	Masked	31 (2.4)	Masked
	COPD (n, %)	87 (3.6)	32 (3.0)	41 (3.7)	18 (3.7)	46 (3.6)	14 (2.4)
	History of major and clinically relevant bleeding (n, %)	79 (3.3)	17 (1.6)	28 (2.5)	Masked	51 (4.0)	Masked
	History of MI (n, %)	94 (3.9)	47 (4.4)	49 (4.4)	18 (3.7)	45 (3.5)	29 (4.9)
	Anemia (n, %)	115 (4.8)	49 (4.5)	54 (4.8)	22 (4.5)	61 (4.8)	27 (4.6)
	Heart failure (n, %)	187 (7.8)	76 (7.0)	82 (7.3)	33 (6.8)	105 (8.2)	43 (7.3)
	VHD (n, %)	513 (21.4)	294 (27.2)	240 (21.4)	155 (31.8)	273 (21.4)	139 (23.5)
	Atrial fibrillation and flutter (n, %)	358 (15.0)	122 (11.3)	146 (13.0)	56 (11.5)	212 (16.6)	66 (11.1)
	Atherosclerosis, peripheral artery disease, ATE and stroke (n, %)	255 (10.7)	127 (11.8)	107 (9.6)	57 (11.7)	148 (11.6)	70 (11.8)
	Diabetes mellitus (n, %)	138 (5.8)	64 (5.9)	67 (6.0)	31 (6.4)	71 (5.6)	33 (5.6)
	Hypertension (n, %)	246 (10.3)	123 (11.4)	106 (9.5)	63 (12.9)	140 (11.0)	60 (10.1)
	Kidney and liver disease (n, %)	122 (5.1)	54 (5.0)	42 (3.8)	24 (4.9)	80 (6.3)	30 (5.1)
Medication use at index date[d]
	Antihypertensives (n, %)	2,221 (92.8)	993 (91.9)	1,042 (93.1)	453 (92.8)	1,179 (92.5)	540 (91.2)
	Anti-inflammatory (non-steroidal) (n, %)	285 (11.9)	146 (13.5)	141 (12.6)	68 (13.9)	144 (11.3)	78 (13.2)
	Anti-depressants (n, %)	252 (10.5)	122 (11.3)	119 (10.6)	64 (13.1)	133 (10.4)	58 (9.8)
	Lipid-lowering drugs (n, %)	1,342 (56.1)	619 (57.3)	651 (58.2)	283 (58.0)	691 (54.2)	336 (56.8)
	Steroids (n, %)	458 (19.1)	171 (15.8)	204 (18.2)	78 (16.0)	254 (19.9)	93 (15.7)
	Antacids (n, %)	1,450 (60.6)	649 (60.1)	688 (61.5)	290 (59.4)	762 (59.8)	359 (60.6)
	Polypharmacy[e] (n, %)	2,183 (91.2)	977 (90.5)	1,026 (91.7)	445 (91.2)	1,157 (90.8)	532 (89.9)
Prior medication[f]
	Prior heparin use (n, %)	253 (10.6)	79 (7.3)	113 (10.1)	34 (7.0)	140 (11.0)	45 (7.6)
	Prior DOAC use (n, %)	35 (1.5)	11 (1.0)	Masked	Masked	28 (2.2)	11 (1.9)
	Prior antiplatelet drugs use (n, %)	283 (11.8)	165 (15.3)	154 (13.8)	83 (17.0)	129 (10.1)	82 (13.9)

Abbreviations: ATE, arterial thrombotic event; AV, aortic valve; DOAC, direct oral anticoagulant; IQR, interquartile range; Masked, masked as per privacy policy of CBS; MHV, mechanical heart valve; MI, myocardial infarction; VHD, valvular heart disease; VKA, vitamin K antagonist; VTE, venous thrombotic event.

^a Type of mechanical heart valve recipient groups were defined based on the registered therapeutic INR ranges at index. Patients with an isolated aortic valve (AV) prosthesis without additional risk factors were treated with a therapeutic range of 2.5 to 3.5 before 2016 and 2.0 to 3.0 from January 2016 onwards. All remaining mechanical heart valve (MHV) patients (i.e., medium-risk MHV recipients) were treated with a therapeutic range of 3.0 to 4.0 before 2016 and 2.5 to 3.5 thereafter.

^b All treatment indications for VKA treatment that have been registered until the date of data export and were identified from the Dutch anticoagulation clinics. One or more indications can be present.

^c Comorbidities were identified by examining data on hospitalizations within 3 years before the index date using ICD-10 codes and ICD-9 codes restricting to main or primary diagnosis of hospital admission. One or more comorbidities can be present.

^d Medication use at index date was identified by examining outpatient medication prescriptions covered by the Basic Dutch Health Insurance based on ATC codes in the calendar year of the index date.

^e Polypharmacy was defined as ≥5 different drug types in the calendar year of the index date, at the therapeutic (2nd) level of the ATC classification.

^f Prior use of antithrombotic medication was identified by examining outpatient medication prescriptions covered by the Basic Dutch Health Insurance based on ATC codes and identified as ≥2 prescriptions within 6 months before the index date.

Quality of Anticoagulation Control

In the first 6 months after index, medium-risk MHV recipients who were treated with higher therapeutic ranges exhibited better anticoagulation control than those treated with lower therapeutic ranges ([Fig. 2A] and [Supplementary Table S7], available in the online version only). The mean TTR was 64.5 (95% CI: 63.3–65.8) in patients with a therapeutic range of 3.0 to 4.0 compared to 54.1 (95% CI: 53.0–55.3) in patients with a therapeutic range of 2.5 to 3.5.

Fig. 2 Quality of anticoagulation control in patients with higher and lower therapeutic international normalized ratio (INR) ranges. This figure displays the mean time in therapeutic range (green colors), mean time above range (red colors), and mean time below range (blue colors), per 6 months period, calculated by the Rosendaal method. INR variability is expressed by the variance growth rate of Cannegieter (gray colors). The dotted lines represent the 95% confidence intervals. These metrics are displayed for patients treated with higher versus lower therapeutic ranges among medium-risk mechanical heart valve recipients (A) and among patients with an isolated aortic valve prosthesis without additional risk factors (B). Type of mechanical heart valve recipient groups were defined based on the registered therapeutic INR ranges at index. Patients with an isolated aortic valve prosthesis without additional risk factors were treated with a therapeutic range of 2.5 to 3.5 before 2016 and 2.0 to 3.0 from January 2016 onwards. All remaining mechanical heart valve patients (i.e., medium-risk MHV recipients) were treated with a therapeutic range of 3.0 to 4.0 before 2016 and 2.5 to 3.5 thereafter.

The mean percentage of INRs within therapeutic range was also higher among medium-risk MHV patients treated with higher therapeutic ranges compared to lower ranges (57.3 [95% CI: 56.1–58.4] vs. 47.6 [95% CI: 46.6–48.7]). No significant differences were observed in the VGR or frequency of significant dose adjustments between the therapeutic ranges ([Supplementary Table S7], available in the online version only). Both groups showed improvement in all anticoagulation control metrics during follow-up ([Fig. 2A]). Patients with higher therapeutic ranges maintained a higher TTR and lower TBR until 30 months after index, thereafter the TTR was similar in both groups (61.9 [95% CI: 57.7–66.2] vs. 66.7 [95% CI: 64.4–69.0]).

In the subgroup of isolated AV recipients, patients treated with higher therapeutic ranges similarly exhibited better anticoagulation control than patients with lower therapeutic ranges, although these differences were less pronounced ([Fig. 2B] and [Supplementary Table S7], available in the online version only).

All-cause and Cardiovascular Mortality

Overall, the incidence rates of all-cause and cardiovascular death per 100 PY were similar for the different therapeutic ranges. For all-cause mortality, rates per 100 PY were 5.4 (95% CI: 4.4–6.4) for higher ranges versus 4.9 (95% CI: 4.2–5.8) for lower ranges ([Table 2] and [Supplementary Table S8], available in the online version only). Likewise, no differences in the hazard of all-cause mortality were observed across the different index years compared to 2013 ([Fig. 3] and [Supplementary Table S9], available in the online version only). However, all-cause and cardiovascular mortality rates were lower in the isolated AV subgroup treated with lower therapeutic ranges (i.e., 2.0–3.0) (2.9 per 100 PY for all-cause death [95% CI: 1.9–4.2]) compared to higher ranges (i.e., 2.5–3.5) (5.1 per 100 PY for all-cause death [95% CI: 3.6–7.0]) ([Table 3] and [Supplementary Table S8], available in the online version only).

Table 2 Incidence rates of outcome events stratified by therapeutic range

		Higher therapeutic ranges (n = 1,607)			Lower therapeutic ranges (n = 1,866)
		Events (n)	Person time in years	IR/100 PY[a] (95% CI)	Events (n)	Person time in years	IR/100 PY[a] (95% CI)
All-cause mortality		120	2,238.8	5.4 (4.4–6.4)	146	2,953.3	4.9 (4.2–5.8)
Cardiovascular mortality[b]		51	2,238.8	2.3 (1.7–3.0)	61	2,953.3	2.1 (1.6–2.7)
Major and clinically relevant bleeding		77	2,168.8	3.6 (2.8–4.4)	71	2,876.7	2.5 (1.9–3.1)
Thromboembolic event		76	2,174.6	3.5 (2.8–4.4)	103	2,832.2	3.6 (3.0–4.4)
	Myocardial infarction	24	2,220.4	1.1 (0.7–1.6)	36	2,905.9	1.2 (0.9–1.7)
	Stroke	43	2,201.0	2.0 (1.4–2.6)	55	2,893.3	1.9 (1.4–2.5)
	Other ATE	Masked	Masked	0.35 (0.15–0.71)	Masked	Masked	0.41 (0.21–0.71)

Abbreviations: ATE, arterial embolism; PY, person years.

Note: Number of events <10 are masked as per privacy policy of CBS.

^a Incidence rates are shown per 100 person years and calculated as the number of patients with the outcome event divided by the sum of observation time.

^b Cardiovascular mortality was defined based on the registered ICD-10 code of the primary cause of death, with codes containing “I” referring to cardiovascular causes.

Fig. 3 Forest plots of hazard ratios and 95% confidence intervals for all-cause mortality (A), major and clinically relevant bleeding (B) and thromboembolic events (C). The hazard ratios of therapeutic range (lower vs. higher [reference]) are based on the fully-adjusted model (model 6). The hazard ratios of year of index date are based on model 1, with 2013 as reference. A subgroup analysis by sex or type of mechanical heart valve (MHV) recipient (patients with an isolated aortic valve [AV] without additional risk factors vs. all remaining MHV patients) was performed when the individual interaction term was statistically significant. Type of MHV recipient groups were defined based on the registered therapeutic INR ranges at index. Patients with an isolated AV prosthesis without additional risk factors were treated with a therapeutic range of 2.5 to 3.5 before 2016 and 2.0 to 3.0 from January 2016 onwards. All remaining MHV patients (i.e., medium-risk MHV recipients) were treated with a therapeutic range of 3.0 to 4.0 before 2016 and 2.5 to 3.5 thereafter.

Consistent with this observation, the effect of therapeutic range on all-cause mortality was modified by the type of MHV recipient. In medium-risk MHV recipients, no association between therapeutic range and all-cause mortality was observed (fully adjusted hazard ratio [aHR]: 1.1 [95% CI: 0.81–1.5]), while in isolated AV recipients a lower therapeutic range was potentially associated with lower all-cause mortality (aHR: 0.60 [95% CI: 0.35–1.0]) ([Fig. 3A] and [Supplementary Table S9], available in the online version only). Similar results were found when INR therapeutic range was treated as time-varying exposure: aHR of 1.2 (95% CI: 0.88–1.5) for a range 2.5 to 3.5 versus 3.0 to 4.0 in medium-risk MHV recipients and aHR of 0.68 (95% CI: 0.46–1.0) for a range of 2.0 to 3.0 versus 2.5 to 3.5 in isolated AV recipients ([Supplementary Table S28], available in the online version only).

Major and Clinically Relevant Bleeding

Gastrointestinal bleedings were the most frequently recorded bleeding type (37.2%), followed by intracranial bleeding (20.9%) and hematuria (12.2%) ([Supplementary Table S10], available in the online version only). Incidence rates of bleeding per 100 PY were numerically higher in patients treated with higher therapeutic ranges (3.6 [95% CI: 2.8–4.4]) compared to patients with lower ranges (2.5 [95% CI: 1.9–3.1]) ([Table 2]). Compared to 2013, hazards of major or clinically relevant bleeding seemed to be lower in 2016 (HR: 0.67 [95% CI: 0.36–1.3]) and 2017 (HR: 0.77 [95% CI: 0.42–1.4]), during which the majority of patients were treated with lower therapeutic ranges ([Fig. 3] and [Supplementary Table S9], available in the online version only). These differences between the therapeutic ranges were consistent across types of MHV recipients ([Table 3]). More specifically, bleeding rates were 3.6 (95% CI: 2.7–4.7) in medium-risk MHV and 3.4 (95% CI: 2.2–5.1) in isolated AV recipients treated with higher therapeutic ranges, whereas rates were 2.6 (95% CI: 2.0–3.4) and 2.1 (95% CI: 1.3–3.4), respectively, in patients treated with lower ranges. Nevertheless, no association was found between therapeutic range and major or clinically relevant bleeding (aHR: 0.79 [95% CI: 0.57–1.1]) ([Fig. 3B] and [Supplementary Table S9], available in the online version only). Consistent results were observed for time-varying therapeutic INR range ([Supplementary Table S28], available in the online version only).

Table 3 Incidence rates of outcome events stratified by therapeutic range and type of MHV recipient

		Higher therapeutic ranges in medium-risk MHV recipients[a] (n = 1,119)			Lower therapeutic ranges in medium-risk MHV recipients[a] (n = 1,274)			Higher therapeutic ranges in isolated AV recipients[a] (n = 488)			Lower therapeutic ranges in isolated AV recipients[a] (n = 592)
		Events (n)	Person time in years	IR/100 PY[b] (95% CI)	Events (n)	Person time in years	IR/100 PY[b] (95% CI)	Events (n)	Person time in years	IR/100 PY[b] (95% CI)	Events (n)	Person time in years	IR/100 PY[b] (95% CI)
All-cause mortality		83	1506.9	5.5 (4.4–6.8)	121	2081.6	5.8 (4.8–6.9)	37	732.0	5.1 (3.6–7.0)	25	871.8	2.9 (1.9–4.2)
Cardiovascular mortality[c]		35	1506.9	2.3 (1.6–3.2)	50	2081.6	2.4 (1.8–3.2)	16	732.0	2.2 (1.2–3.5)	11	871.8	1.3 (0.6–2.3)
Major and clinically relevant bleeding		53	1463.6	3.6 (2.7–4.7)	53	2033.9	2.6 (2.0–3.4)	24	705.1	3.4 (2.2–5.1)	18	842.8	2.1 (1.3–3.4)
Thromboembolism		53	1462.6	3.6 (2.7–4.7)	85	1980.8	4.3 (3.4–5.3)	23	712.1	3.2 (2.0–4.8)	18	851.4	2.1 (1.3–3.3)
	Myocardial infarction	Masked	Masked	1.2 (0.71–1.9)	Masked	Masked	1.5 (0.99–2.1)	Masked	Masked	0.83 (0.30–1.8)	Masked	Masked	0.70 (0.26–1.5)
	Stroke	Masked	Masked	1.9 (1.3–2.7)	Masked	Masked	2.3 (1.7–3.0)	Masked	Masked	2.1 (1.2–3.4)	Masked	Masked	1.0 (0.48–2.0)
	Other ATE	Masked	Masked	0.40 (0.15–0.87)	Masked	Masked	2.3 (1.7–3.0)	Masked	Masked	0.27 (0.03–0.99)	Masked	Masked	0.34 (0.07–1.0)

Abbreviations: ATE, arterial embolism; MHV, mechanical heart valve; PY, person years.

Note: Number of events <10 are masked as per privacy policy of CBS.

^a Types of mechanical heart valve (MHV) recipient groups were defined based on the registered therapeutic INR ranges at index. Patients with an isolated aortic valve (AV) prosthesis without additional risk factors were treated with a therapeutic range of 2.5 to 3.5 before 2016 and 2.0 to 3.0 from January 2016 onwards. All remaining MHV patients (i.e., medium-risk MHV recipients) were treated with a therapeutic range of 3.0 to 4.0 before 2016 and 2.5 to 3.5 thereafter.

^b Incidence rates are shown per 100 person years and calculated as the number of patients with the outcome event divided by the sum of observation time.

^c Cardiovascular mortality was defined based on the registered ICD-10 code of the primary cause of death, with codes containing “I” referring to cardiovascular causes.

Thromboembolic Events

Incidence rates of the composite of TE, as well as those for stroke, MI, and peripheral ATE separately, were similar for the different therapeutic ranges: 3.5 per 100 PY (95% CI: 2.8–4.4) for higher therapeutic ranges versus 3.6 per 100 PY (95% CI: 3.0–4.4) for lower therapeutic ranges for the composite outcome of TE ([Table 2]). Similarly, hazards of the composite of TE were comparable across the different index years with 2013 as reference ([Fig. 3] and [Supplementary Table S9], available in the online version only). Stratification by MHV recipient type showed that medium-risk MHV recipients treated with higher therapeutic ranges had a numerically lower incidence of TE than those with lower ranges (3.6 [95% CI: 2.7–4.7] vs. 4.3 [95% CI: 3.4–5.3]) ([Table 3]). In line with this, the effect of therapeutic range on TE was modified by MHV recipient type ([Fig. 3] and [Supplementary Table S9], available in the online version only). A lower therapeutic range was potentially associated with an increased TE risk compared to a higher range in medium-risk MHV recipients (aHR: 1.3 [95% CI: 0.94–1.9]), although not statistically significant. A similar pattern was observed when modeling INR therapeutic as time-varying exposure (2.5–3.5 vs. 3.0–4.0) (aHR: 1.3 [95% CI: 0.95–1.8) ([Supplementary Table S28], available in the online version only). In isolated AV recipients, no association was observed when therapeutic range was treated as a fixed exposure (aHR: 0.71 [95% CI: 0.38–1.3]), while a range of 2.0 to 3.0 versus 2.5 to 3.5 was associated with a lower TE risk in time-varying analysis (aHR: 0.57 [95% CI: 0.35–0.93]) ([Supplementary Table S28], available in the online version only).

Sensitivity Analyses

Results of sensitivity analyses were largely consistent with the main analysis, with all confidence intervals overlapping. Excluding hospital admission periods from the TTR calculation did not quantitatively alter the results ([Supplementary Tables S11] and [S12], available in the online version only). In contrast, the effect of therapeutic range on bleeding was potentially modified by sex after extending the censoring period by 7 (aHR in males: 1.0 [95% CI: 0.66–1.6] vs. females: 0.56 [95% CI: 0.32–0.94]) ([Supplementary Table S16], available in the online version only) and 14 days (aHR in males: 1.0 [95% CI: 0.66–1.6] vs. females: 0.52 [95% CI: 0.31–0.88]) ([Supplementary Table S20], available in the online version only). Notably, the observed effect modification by MHV recipient type on all-cause mortality in the main analysis disappeared after extending the censoring period by 7 and 14 days ([Supplementary Tables S16] and [S20], available in the online version only). Excluding patients with the ICD-10 code “Z954 - Presence of other heart-valve replacement” (n = 522) did not affect the observed effect modification of therapeutic range on all-cause mortality by MHV recipient type ([Supplementary Table S27], available in the online version only). A potential overall association was observed between lower therapeutic ranges and increased TE risk (aHR: 1.4 [95% CI: 1.0–2.0]). However, unlike the main analysis, the interaction between therapeutic range and type of MHV recipient on TE was no longer statistically significant, which might be due to loss of statistical power ([Supplementary Table S27], available in the online version only).

In the calendar time analysis, medium-risk MHV patients treated with lower therapeutic ranges had possible increased risks of all-cause mortality (HR: 1.3 [95% CI: 1.0–1.8]) and TE (HR: 1.3 [95% CI: 0.91–1.8]), while no association was observed in isolated AV recipients for all-cause mortality (HR: 0.95 [95% CI: 0.60–1.5]) and TE (HR: 0.77 [95% CI: 0.44–1.4]) ([Supplementary Table S29], available in the online version only). Furthermore, a lower therapeutic INR range was potentially associated with a lower bleeding risk in isolated AV recipients (HR: 0.56 [95% CI: 0.32–1.0]), but not in medium-risk MHV recipients (HR: 0.99 [95% CI: 0.69–1.4]).

Discussion

In this large-scale study, we evaluated the quality of anticoagulation control in MHV patients treated with high versus low therapeutic INR ranges and assessed associations between these ranges and clinical outcomes. Anticoagulation control was poorer among patients treated with lower therapeutic ranges, particularly among patients with a non-aortic MHV and/or additional thromboembolic risk factors (therapeutic range of 2.5–3.5 vs. 3.0–4.0). In this subgroup, a lower therapeutic range was potentially associated with a higher TE risk. In adjusted analyses, no clear association was observed between therapeutic INR range and bleeding. These results were consistent when INR therapeutic range was modeled as time-fixed or time-varying exposure.

Consistent with previous studies,[15] [16] we observed that patients treated with higher therapeutic ranges (i.e., 2.5–3.5 for isolated AV recipients; 3.0–4.0 for all remaining MHV recipients) had better anticoagulation control compared with patients with lower ranges (i.e., 2.0–3.0 and 2.5–3.5, respectively). This difference likely reflects the wider acceptable INR intervals used before 2016,[28] which were incorporated into the TTR calculations. After 2016, the acceptable intervals were narrowed to align strictly with the newly implemented lower therapeutic ranges, and the TTR values became similar for both groups. Other metrics, such as the frequency of significant dose adjustments and variability between consecutive INRs, were comparable across the different therapeutic ranges. As expected, patients in whom a higher therapeutic INR range was indicated (i.e., patients with a non-aortic MHV and/or additional risk factors) had consistently lower TTR compared to patients with an isolated AV, both in the higher and lower therapeutic range groups.

Although anticoagulation control was poorer in patients treated with lower INR ranges, we did not observe an overall increase in TE risk. This finding aligns with results from several RCTs, such as the LOWERING-IT, ESCAT III, and PROACT trials,[10] [11] [12] [13] which demonstrated that lower INR ranges may be safe in select MHV patients with modern bileaflet prostheses. However, the PROACT Mitral trial could not demonstrate non-inferiority of an INR range of 2.0 to 2.5 compared with 2.5 to 3.5 in patients with an On-X mitral valve prosthesis.[14] Similarly, we observed a possibly higher risk of TE for a range of 2.5 to 3.5 compared with 3.0 to 4.0 in patients with a non-aortic MHV and/or additional risk factors. These high-risk patients may benefit from maintaining a higher therapeutic INR range to mitigate their elevated TE risk.[9]

An unexpected finding was the lower all-cause mortality in isolated AV recipients treated with lower INR ranges. Although this finding could suggest a mortality benefit for lower INR ranges, several factors complicate its interpretation. First, mortality was lower for both cardiovascular and non-cardiovascular causes and was not explained by differences in bleeding. Second, informative censoring may have occurred among patients who discontinued VKA therapy. Indeed, extending the censoring period and using the calendar time as proxy for therapeutic INR range with an alternative censoring approach eliminated the mortality benefit of lower INR ranges. Finally, improved long-term survival over time may have contributed to this finding, as most patients treated with lower therapeutic ranges were included after 2015. However, a similar mortality benefit of lower ranges was not observed in medium-risk MHV patients. Therefore, the lower all-cause mortality in isolated AV recipients treated with lower ranges warrant cautious interpretation and may be attributed to residual confounding.

Although no overall association was found between therapeutic range and risk of major and clinically relevant bleeding, sensitivity analyses using extended censoring periods (7 or 14 days) indicated a lower bleeding risk in women treated with lower therapeutic ranges. By extending the censoring period, we potentially captured more bleeding events, which could still be related to the previous recorded therapeutic INR range. Given that women, especially at reproductive age, may be more vulnerable to bleeding due to menstrual and other gynecological bleeds, a lower therapeutic range might offer greater benefit in this subgroup.

The most important strength of our study is the unique and robust dataset of patients with an MHV diagnosis, which combines nationwide individual-level data with longitudinal INR measurements from anticoagulation clinics. By cross-validating MHV hospital diagnoses with indications from anticoagulation clinic records, we achieved a high level of ascertainment of the presence of an MHV. The uniformity in patient management across Dutch anticoagulation clinics further strengthened the reliability of our findings.

However, several limitations should be acknowledged. First, identification of MHV patients relied on ICD codes, which may have introduced misclassification. To minimize this risk, we cross-validated the MHV ICD codes with treatment indications registered at the anticoagulation clinics, resulting in a more reliable selection of patients. Second, due to the lack of access to primary care records, some comorbidities and events, particularly VTE and clinically relevant bleedings, may have been underreported. However, as the identification strategy for outcomes and confounding factors was employed uniformly across therapeutic range groups, any potential misclassification is unlikely to substantially impact the comparison of clinical outcomes between the therapeutic ranges.

Third, we lacked information on several clinical details, such as valve indication, types, models, and implantation sites. These factors are associated with all-cause mortality, as well as thromboembolic and bleeding risk,[12] [29] [30] [31] and could have confounded or modified the studied associations. Nevertheless, since the therapeutic range was primarily determined by the calendar year of the index date rather than by patient characteristics, the risk of confounding by indication was likely minimal. In addition, the lack of information on hemoglobin levels and transfusion requirements prevented distinguishing major from clinically relevant nonmajor bleeding.

Fourth, censoring strategies may have influenced the observed associations. Most MHV patients who were initially treated with a higher therapeutic range were censored after 2016, when therapeutic INR ranges were lowered nationally. Since this switch was primarily determined by time, we consider this non-informative censoring. Importantly, our sensitivity analyses using alternative censoring approaches yielded broadly similar results. However, informative censoring may still have occurred in patients who discontinued VKA therapy altogether, potentially affecting associations with all-cause mortality. Finally, the relatively low number of events may have reduced the study's statistical power.

Conclusion

Our study findings suggest that lowering therapeutic INR ranges does not lead to an increase of thromboembolic or mortality risk in most MHV patients, but may reduce bleeding risks in some patients, especially in women. However, a higher therapeutic INR range may be needed in complex cases, because of valve position, type, or associated risk factors such as AF or heart failure. Our findings address uncertainties surrounding the optimal therapeutic INR range and are consistent with the current clinical guidelines to achieve a target INR of 2.5 in patients with an aortic MHV without additional thromboembolic risk factors, 3.0 in presence of thromboembolic risk factors or in case of mitral or tricuspid MHV, and 3.5 in patients with a mitral or tricuspid or older generation prosthesis and additional risk factors.[4] [5]

Conflict of Interest

C.V. received a travel award from the International Society on Thrombosis and Haemostasis (ISTH) for attending the ISTH congress 2024. J.S. received a non-restricted research grant from the Dutch Federation of Anticoagulation Clinics. F.K. has received research support from Bayer, BMS, BSCI, AstraZeneca, Angiodynamics, MSD, Leo Pharma, Actelion, Farm-X, The Netherlands Organisation for Health Research and Development, the Dutch Thrombosis Foundation, the Dutch Heart Foundation, and the Horizon Europe Program, all outside this project and paid to his institution. F.L. has received research support from CSL Behring, Takeda, and SOBI and consulting fees from CSL Behring, Biomarin, and Takeda, all outside this project and paid to his institution. A.S. has received speaker and consultancy fees from Amaryn, Astra, Boehringer-Ingelheim, NovoNordisk, Sanofi, and Viatris. R.d.C. received consultancy/speakers fees from Sanofi, Milestone, Daiichi-Sankyo, Janssen, Pfizer, Bristol-Myers Squibb, Menarini, and Amarin, travel support from Amarin and Daiichi-Sankyo, and participated in the Data Safety Monitoring Board of NOAH-AFNET6 (Daiichi-Sankio), all outside this project. H.t.C. received consultancy fees from Astra Zeneca, Novostia, Alveron, and Galapagos. He is shareholder of CoagulationProfile. All revenues are deposited at the CARIM institute for research. Marieke Kruip has received research support from The Netherlands Organisation for Health Research and Development, the Dutch Thrombosis Foundation, and the Horizon Europe Program, and a speaker fee from Roche, all outside this project and paid to her institution. All other authors have nothing to declare.

Acknowledgment

The authors thank the Federation of Dutch Anticoagulation Clinics and Statistics Netherlands for making the data available. Additionally, the authors want to acknowledge the Erasmus MC Graduate School for their academic support and resources throughout the execution of this research.

Data Availability Statement

This study used non-public microdata from Statistics Netherlands and the Federation of Dutch Anticoagulation Clinics. These data cannot be shared directly by the authors. Under certain conditions, these data are accessible for statistical and scientific research. For additional information: [e-mail] and/or [e-mail]

Contributors' Statement

C.V., E.K.K., J.S., Q.C., R.D.C., S.C.C., H.t.C., M.E.W.H., M.K., and M.J.H.A.K. designed the study; C.V. and E.K.K. had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; H.J.A., M.J.B., A.t.C.H., L.M.F., F.A.K., S.J.C.M.v.d.L., M.C.N., R.K.S., A.D.M.S., N.M.W., and Q.C. are responsible for obtaining the data from participating anticoagulation clinics; C.V. and E.K.K. drafted the initial version of the manuscript; J.S., F.W.G.L., A.t.C.H., H.t.C., R.P., M.E.W.H., R.D.C., M.K., and M.J.H.A.K. contributed to the interpretation of the data. All authors critically revised the manuscript and approved the final version of the manuscript.

• References

• 1 Tsao CW, Aday AW, Almarzooq ZI. et al. Heart disease and stroke statistics-2022 update: a report from the American Heart Association. Circulation 2022; 145 (08) e153-e639
• 2 Tsao CW, Aday AW, Almarzooq ZI. et al; American Heart Association Council on Epidemiology and Prevention Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics-2023 update: a report from the American Heart Association. Circulation 2023; 147 (08) e93-e621
• 3 Coffey S, Roberts-Thomson R, Brown A. et al. Global epidemiology of valvular heart disease. Nat Rev Cardiol 2021; 18 (12) 853-864
• 4 Otto CM, Nishimura RA, Bonow RO. et al. 2020 ACC/AHA guideline for the management of patients with valvular heart disease: a report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation 2021; 143 (05) e72-e227
• 5 Praz F, Borger MA, Lanz J. et al. 2025 ESC/EACTS guidelines for the management of valvular heart disease: developed by the task force for the management of valvular heart disease of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS). Eur Heart J 2025; 46 (44) 4635-4736
• 6 Torn M, Cannegieter SC, Bollen WL, van der Meer FJ, van der Wall EE, Rosendaal FR. Optimal level of oral anticoagulant therapy for the prevention of arterial thrombosis in patients with mechanical heart valve prostheses, atrial fibrillation, or myocardial infarction: a prospective study of 4202 patients. Arch Intern Med 2009; 169 (13) 1203-1209 doi
• 7 De Caterina R, Husted S, Wallentin L. et al. Vitamin K antagonists in heart disease: current status and perspectives (Section III). Position paper of the ESC Working Group on Thrombosis—Task Force on Anticoagulants in Heart Disease. Thromb Haemost 2013; 110 (06) 1087-1107
• 8 Hemker HC, Frank HL. The mechanism of action of oral anticoagulants and its consequences for the practice of oral anticoagulation. Haemostasis 1985; 15 (04) 263-270
• 9 Cannegieter SC, Rosendaal FR, Wintzen AR, van der Meer FJ, Vandenbroucke JP, Briët E. Optimal oral anticoagulant therapy in patients with mechanical heart valves. N Engl J Med 1995; 333 (01) 11-17
• 10 Koertke H, Zittermann A, Tenderich G. et al. Low-dose oral anticoagulation in patients with mechanical heart valve prostheses: final report from the early self-management anticoagulation trial II. Eur Heart J 2007; 28 (20) 2479-2484
• 11 Koertke H, Zittermann A, Wagner O. et al; Christ of Huth. Telemedicine-guided, very low-dose international normalized ratio self-control in patients with mechanical heart valve implants. Eur Heart J 2015; 36 (21) 1297-1305
• 12 Torella M, Torella D, Chiodini P. et al. LOWERing the INtensity of oral anticoaGulant Therapy in patients with bileaflet mechanical aortic valve replacement: results from the “LOWERING-IT” trial. Am Heart J 2010; 160 (01) 171-178
• 13 Puskas JD, Gerdisch M, Nichols D. et al; PROACT Investigators. Anticoagulation and antiplatelet strategies after On-X mechanical aortic valve replacement. J Am Coll Cardiol 2018; 71 (24) 2717-2726
• 14 Chu MWA, Ruel M, Graeve A. et al; PROACT Mitral Investigators. Low-dose vs standard warfarin after mechanical mitral valve replacement: a randomized trial. Ann Thorac Surg 2023; 115 (04) 929-938
• 15 Seelig J, Hemels MEW, Xhaët O. et al; GARFIELD-AF investigators. Impact of different anticoagulation management strategies on outcomes in atrial fibrillation: Dutch and Belgian results from the GARFIELD-AF registry. J Thromb Haemost 2020; 18 (12) 3280-3288
• 16 Van't Land RP, Banga JD, van den Besselaar AMHP. Therapeutic quality control in a regional thrombosis center: the effect of changing the target intensity of anticoagulation with vitamin K antagonists. Thromb Res 2021; 203: 85-89
• 17 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
• 18 van Leeuwen Y, Rosendaal FR, Cannegieter SC. Prediction of hemorrhagic and thrombotic events in patients with mechanical heart valve prostheses treated with oral anticoagulants. J Thromb Haemost 2008; 6 (03) 451-456
• 19 Schulman S, Kearon C. Subcommittee on Control of Anticoagulation of the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb Haemost 2005; 3 (04) 692-694
• 20 Kaatz S, Ahmad D, Spyropoulos AC, Schulman S. Subcommittee on Control of Anticoagulation. Definition of clinically relevant non-major bleeding in studies of anticoagulants in atrial fibrillation and venous thromboembolic disease in non-surgical patients: communication from the SSC of the ISTH. J Thromb Haemost 2015; 13 (11) 2119-2126
• 21 Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials 1996; 17 (04) 343-346
• 22 R Core Team. R: A language and environment for statistical computing. Accessed at: https://www.r-project.org/
• 23 Posit team. RStudio: Integrated Development Environment for R. Posit Software, PBC, Boston, MA. Accessed at: http://www.rstudio.com
• 24 Therneau TM. . A Package for Survival Analysis in R. Accessed at: https://CRAN.R-project.org/package=survival
• 25 Therneau TM, Grambsch PM. . Modeling Survival Data: Extending the Cox Model: Springer, New York; 2000
• 26 Stevenson M, Sergeant E. . epiR: Tools for the Analysis of Epidemiological Data. Accessed at: https://CRAN.R-project.org/package=epiR
• 27 Gerds TA. . prodlim: Product-Limit Estimation for Censored Event History Analysis. Accessed at: https://CRAN.R-project.org/package=prodlim
• 28 Federatie van Nederlandse Trombosediensten. Jaarverslag 2016. Accessed at: https://www.fnt.nl/algemeen/jaarverslag-2016
• 29 Cannegieter SC, Rosendaal FR, Briët E. Thromboembolic and bleeding complications in patients with mechanical heart valve prostheses. Circulation 1994; 89 (02) 635-641
• 30 Ruel M, Masters RG, Rubens FD. et al. Late incidence and determinants of stroke after aortic and mitral valve replacement. Ann Thorac Surg 2004; 78 (01) 77-83 , discussion 83–84
• 31 Pastori D, Poli D, Antonucci E. et al; Italian Federation of Anticoagulation Clinics (FCSA). Sex-based difference in anticoagulated patients with mechanical prosthetic heart valves and long-term mortality risk. Int J Clin Pract 2021; 75 (05) e14064

Publication History

Received: 05 September 2025

Accepted after revision: 15 December 2025

Accepted Manuscript online:
02 January 2026

Article published online:
15 January 2026