Abstract
Background
Little is published about warfarin therapy adherence patterns beyond 6 months of initial anticoagulant treatment and their association with effectiveness and safety for patients with venous thromboembolism (VTE).
Objectives
To compare the risks of recurrent VTE and major bleeding during extended treatment between adherence patterns using MarketScan Commercial and Medicare Supplemental databases (2013-2019).
Methods
In a retrospective cohort study, we included patients with incident VTE who completed an initial 6-month anticoagulant treatment and received either warfarin or no extended therapy. Group-based trajectory models were used to identify distinct extended treatment trajectories. Associations between the trajectories and risk of hospitalization due to recurrent VTE and major bleeding were assessed using inverse probability treatment-weighted Cox proportional hazards models.
Results
Compared with no extended treatment, consistently high warfarin adherence was associated with a significantly decreased risk of hospitalization due to recurrent VTE (hazard ratio [HR] = 0.23; 95% CI, 0.12-0.45), but gradually (HR = 0.29; 95CI, 0.08-1.06) or rapidly declining (HR = 0.14; 95% CI, 0.02-1.24) adherence showed no association with the risk of hospitalization due to recurrent VTE. Compared with no extended treatment, warfarin extended therapy was associated with an increased risk of hospitalization due to major bleeding regardless of adherence patterns (consistently high: HR = 2.08; 95% CI, 1.18-3.64, gradually declining: HR = 2.10; 95% CI, 0.74-5.95, and rapidly declining: HR = 9.19; 95% CI, 4.38-19.29). However, compared with rapidly declining adherence, consistently high (HR = 0.23; 95% CI, 0.11-0.47) and gradually declining (HR = 0.23; 95% CI, 0.08-0.64) adherence were associated with decreased risk of hospitalization due to major bleeding.
Conclusion
The findings indicated that consistently high adherence to extended warfarin treatment was associated with a decreased risk of hospitalization due to recurrent VTE but an increased risk of hospitalization due to major bleeding compared with no extended treatment.
Keywords: bleeding, medication adherence, venous thromboembolism, warfarin
Essentials
-
•
Warfarin is used to prevent recurrent venous thromboembolism (VTE) after initial treatment.
-
•
We assessed the association between adherence to extended warfarin use and clinical outcomes.
-
•
Warfarin use was associated with a higher major bleeding risk than no extended treatment.
-
•
Consistently high adherence to warfarin use was associated with decreased recurrent VTE risk.
1. Introduction
Venous thromboembolism (VTE), which includes deep vein thrombosis and pulmonary embolism (PE), affects approximately 900,000 (1-2 per 1000) individuals each year in the United States [1]. Approximately 30% of patients with VTE experience recurrence within 10 years, and the risk of VTE recurrence remains high for 6 to 12 months after the incident VTE [2,3]. An extended phase of anticoagulation treatment is recommended as the standard of care for patients with unprovoked VTE or VTE provoked by a persistent risk factor after 3 to 6 months of initial treatment for incident VTE [[4], [5], [6]]. Although direct oral anticoagulants (DOACs) are recommended during extended treatment, warfarin is one of the most commonly prescribed anticoagulants owing to its relatively low cost, and warfarin remains the first-line therapy for patients with specific clinical conditions [[4], [5], [6], [7]].
Several studies have examined adherence to warfarin treatment and its association with clinical outcomes among patients with VTE. A retrospective cohort study found that 3-month persistence with rivaroxaban or warfarin—defined as consistent refills for 90 days—was associated with a lower 90-day hospital readmission rate than nonpersistence (15% vs 35%; P =.035) [8]. Similarly, among patients who received warfarin treatment, those who discontinued within 3 months were associated with a 1.5-fold increased risk of VTE recurrence compared with patients who discontinued after 3 months [9]. However, using a single adherence/nonadherence measure or single arbitrary cutoff to distinguish adherent vs nonadherent patients may underestimate real-world heterogeneity in dynamic nonadherence behavior, which may have differential associations with recurrent VTE or major bleeding events in clinical practice.
Therefore, we aimed to identify and characterize adherence trajectories to extended warfarin treatment using group-based trajectory models (GBTMs) and to examine the associations between extended treatment trajectories and risk of hospitalization due to recurrent VTE and major bleeding events after 6 months of initial anticoagulant therapy among patients with VTE.
2. Methods
2.1. Data sources
We conducted a retrospective cohort study using data from 2013-2019 MarketScan Commercial and Medicare Supplemental databases. Deidentified person-level information on health service use and enrollment records are available for >80 million individuals in the MarketScan Commercial database and 6 million individuals in the Medicare Supplemental database.
2.2. Study population
Using a previously validated algorithm [10,11], we identified patients aged ≥18 years diagnosed with their first VTE based on receipt of an inpatient diagnosis of deep vein thrombosis or PE using the International Classification of Diseases, Ninth (ICD-9-CM) or Tenth (ICD-10-CM) Revision, Clinical Modification diagnosis codes. Included patients were required to (a) not have a VTE diagnosis or anticoagulant treatment within 12 months before the first VTE; (b) have initiated any oral anticoagulant treatment within 30 days after the first VTE; (c) have anticoagulant treatment for ≥5 months during the 6-month initial treatment period without recurrent VTE or major bleeding events [[12], [13], [14]]; and (d) have either extended treatment with warfarin or no extended treatment. These requirements were assessed by examining prescriptions during 30 days before and after month 6 following the first oral anticoagulant prescription date. We defined the index date for patients receiving warfarin extended therapy as the prescription date of warfarin that was closest to the end of month 6, and for patients with no extended treatment, the day after the last fill date of an anticoagulant prescription plus days of supply. Patients were required to have 12-month continuous enrollment in medical and pharmacy benefits before the index date (baseline period).
2.3. Monthly medication adherence
Monthly medication adherence was defined using the monthly proportion of days covered (PDC), derived by dividing the total number of days covered with warfarin by 30 days each month during the first 6 months of extended treatment after the index date. We calculated monthly PDC until patients were disenrolled from a health insurance plan, switched to a DOAC (or initiation of oral anticoagulants for patients with no extended treatment), or experienced an outcome, whichever occurred first. Their PDC was treated as missing afterward until the end of month 6.
2.4. Outcomes
The effectiveness outcome was time to the first hospitalization due to recurrent VTE, defined as the presence of inpatient ICD-9-CM or ICD-10-CM primary discharge diagnosis codes that have been validated and have a positive predictive value of 83% [10]. The safety outcome was time to the first hospitalization due to major bleeding, defined according to Cunningham’s algorithm, with a positive predictive value of 89% to 99% [15]. The algorithm uses ICD diagnosis and procedure codes to ascertain hospitalizations associated with serious bleeding events, including gastrointestinal bleeding, intracranial hemorrhage, genitourinary bleeding, and bleeding at other sites [15].
For each outcome, we followed up patients until the earliest date of (a) occurrence of an outcome, (b) switch to DOAC (or initiation of oral anticoagulants for patients with no extended treatment), (c) discontinuation of warfarin (>7-day gap between refills), (d) end of enrollment in a health insurance plan, or (e) end of the study period.
2.5. Statistical analysis
We used GBTMs to identify and characterize differential patterns of change in adherence to extended warfarin treatment over time [[16], [17], [18]]. GBTMs identify subgroups of patients with similar medication refilling patterns and provide a trajectory of average medication adherence for each subgroup over time, accounting for both the timing and the extent of medication nonadherence and identifying more heterogeneity in nonadherence behavior than traditional single measures. We used a censored normal probability distribution and a flexible polynomial function of time to identify the number of trajectories that best characterize the data. We included dropout modeling in the GBTM to account for missing PDC [19]. Among the models with 2 to 5 trajectories, we selected the final model based on a combination of (a) the Bayesian information criterion, wherein the largest value indicates the model with the best fit, (b) estimated trajectory group proportions ≥10% to maximize clinical interpretation and utility, and (c) the Nagin criteria to assess the final adequacy of a GBTM [18,20].
After identifying the final trajectory groups, we used the stabilized inverse probability treatment weighting (IPTW) method to balance the differences in baseline patient characteristics among the treatment trajectory groups (covariates in Table 1) [21,22]. A standardized mean difference of >0.1 indicated a statistically significant difference among the groups [23].
Table 1.
Covariates adjusted for inverse probability treatment weighting.
| Covariates | Description |
|---|---|
| Demographic characteristics | Age, sex |
| Comorbidities | Cancer, surgery, trauma, hyperlipidemia, abnormal coagulation, tobacco use, respiratory diseases, liver diseases, chronic kidney disease stages 3 to 5, anemia, alcohol use disorder, drug-use disorder, history of bleeding, ischemic heart disease, myocardial infarction, atrial fibrillation, stroke, heart failure, varicose veins, and thrombocytopenia |
| Prior medication use | Antiplatelet therapy, corticosteroids, nonsteroidal anti-inflammatory drugs, angiotensin-converting enzyme inhibitors, aspirin, β-blockers, calcium channel blockers, selective serotonin reuptake inhibitors, proton pump inhibitors, loop diuretics, potassium-sparing diuretics, thiazide diuretics, vasodilators, estrogens, and cyclooxygenase-2 inhibitors |
| Bleeding risk score | HAS-BLED score calculated by presence of hypertension, abnormal renal or liver function, stroke, bleeding history or predisposition, labile INR (international normalized ratio), age (>65), medication usage predisposing to bleeding, and alcohol use (range, 0-8; high-bleeding risk for scores ≥3) |
| Provoked VTE | Provoked VTE was defined as receiving a cancer diagnosis within 6 mo preceding incident VTE or any of the following within 3 mo preceding incident VTE: pregnancy, trauma, surgery, or hospital admission for at least 3 consecutive days |
VTE, venous thromboembolism.
We reported crude incidence rates for each outcome (number of events per 100 person-years). Following IPTW, we used Cox proportional hazards models to compare recurrent VTE and major bleeding risks among warfarin trajectories, with the no extended treatment trajectory as a reference. We also conducted pairwise comparisons among warfarin-treated groups. Data were analyzed using Stata/MP 16 (StataCorp) and SAS 9.4 (SAS Institute Inc).
3. Results
3.1. Adherence trajectories
We identified 5315 patients with extended warfarin treatment and 5133 patients with no extended treatment. Among adherence trajectories with 2 to 5 groups, a 4-group model was selected based on the Bayesian information criterion, estimated trajectory group proportions >10%, and Nagin’s criteria. The final 4 treatment trajectories comprised no extended treatment (49.1%) and 3 distinct adherence patterns of extended warfarin treatment, including consistently high (24.0%), gradually declining (13.2%), and rapidly declining (13.7%) adherence (Figure). Patient characteristics by adherence trajectory were comparable across the 4 trajectories after IPTW (all maximum standardized mean differences <0.1) (Table 2).
Figure.
Trajectories of adherence to extended treatment with warfarin among patients with venous thromboembolism. Adherence trajectories were identified using a group-based trajectory model assessing the monthly proportion of days covered during the first 6 months of extended treatment. Dotted lines represent 95% CIs.
Table 2.
Patient characteristics by trajectory group based on adherence to extended treatment with warfarin.
| Characteristic | No. (%) of patients |
Maximum SMD before IPTWa | Maximum SMD after IPTW a | ||||
|---|---|---|---|---|---|---|---|
| Total (N = 10,448) | Consistently high adherence (n = 2509) | Gradually declining adherence (n = 1377) | Rapidly declining adherence (n = 1429) | No-extended treatment (n = 5133) | |||
| Percentage of total cohort | 100% | 24.0% | 13.2% | 13.7% | 49.1% | ||
| Age, mean (SD), y | 59.6 (15.3) | 61.7 (14.9) | 61.1 (15.2) | 58.5 (16.3) | 58.5 (15.1) | 0.210 | 0.063 |
| Female | 5289 (50.6) | 1222 (48.7) | 698 (50.7) | 749 (52.4) | 2620 (51.0) | 0.074 | 0.051 |
| Comorbidity | |||||||
| Cancer | 1713 (16.4) | 388 (15.5) | 248 (18.0) | 240 (16.8) | 837 (16.3) | 0.068 | 0.038 |
| Surgery | 2955 (28.3) | 710 (28.3) | 373 (27.1) | 435 (30.4) | 1437 (28.0) | 0.074 | 0.059 |
| Trauma | 824 (7.9) | 196 (7.8) | 138 (10.0) | 143 (10.0) | 347 (6.8) | 0.118 | 0.059 |
| Hyperlipidemia | 4628 (44.3) | 1202 (47.9) | 600 (43.6) | 631 (44.2) | 2195 (42.8) | 0.103 | 0.059 |
| Abnormal coagulation | 2386 (22.8) | 645 (25.7) | 348 (25.3) | 324 (22.7) | 1069 (20.8) | 0.116 | 0.058 |
| Tobacco use | 1070 (10.2) | 274 (10.9) | 140 (10.2) | 152 (10.6) | 504 (9.8) | 0.036 | 0.032 |
| Respiratory tract disease | 3774 (36.1) | 965 (38.5) | 516 (37.5) | 470 (32.9) | 1823 (35.5) | 0.117 | 0.068 |
| Liver disease | 1362 (13.0) | 325 (13.0) | 204 (14.8) | 212 (14.8) | 621 (12.1) | 0.080 | 0.066 |
| Chronic kidney disease | 1476 (14.1) | 416 (16.6) | 249 (18.1) | 220 (15.4) | 591 (11.5) | 0.186 | 0.055 |
| Anemia | 3347 (32.0) | 817 (32.6) | 479 (34.8) | 503 (35.2) | 1548 (30.2) | 0.108 | 0.048 |
| Alcohol use disorder | 277 (2.7) | 62 (2.5) | 30 (2.2) | 47 (3.3) | 138 (2.7) | 0.068 | 0.051 |
| Drug-use disorder | 281 (2.7) | 66 (2.6) | 42 (3.1) | 43 (3.0) | 130 (2.5) | 0.031 | 0.022 |
| History of bleeding | 3474 (33.3) | 854 (34.0) | 473 (34.4) | 498 (34.8) | 1649 (32.1) | 0.058 | 0.018 |
| Ischemic heart disease | 2819 (27.0) | 753 (30.0) | 396 (28.8) | 371 (26.0) | 1299 (25.3) | 0.105 | 0.035 |
| Myocardial infarction | 1038 (9.9) | 274 (10.9) | 147 (10.7) | 138 (9.7) | 479 (9.3) | 0.053 | 0.021 |
| Atrial fibrillation | 1569 (15.0) | 503 (20.0) | 233 (16.9) | 191 (13.4) | 642 (12.5) | 0.205 | 0.065 |
| Stroke | 545 (5.2) | 167 (6.7) | 92 (6.7) | 74 (5.2) | 212 (4.1) | 0.113 | 0.047 |
| Heart failure | 2004 (19.2) | 564 (22.5) | 290 (21.1) | 257 (18.0) | 893 (17.4) | 0.127 | 0.053 |
| Varicose veins | 294 (2.8) | 84 (3.3) | 51 (3.7) | 38 (2.7) | 121 (2.4) | 0.079 | 0.039 |
| Thrombocytopenia | 737 (7.1) | 195 (7.8) | 114 (8.3) | 98 (6.9) | 330 (6.4) | 0.071 | 0.039 |
| HAS-BLED score, mean (SD) | 1.9 (1.2) | 2.0 (1.2) | 2.0 (1.2) | 1.9 (1.2) | 1.8 (1.2) | 0.167 | 0.043 |
| Initial presentation of VTE | |||||||
| DVT and PE | 2458 (23.5) | 688 (27.4) | 400 (29.0) | 323 (22.6) | 1047 (20.4) | 0.202 | 0.096 |
| DVT only | 2163 (20.7) | 580 (23.1) | 325 (23.6) | 381 (26.7) | 877 (17.1) | 0.233 | 0.053 |
| PE only | 5827 (55.8) | 1241 (49.5) | 652 (47.3) | 725 (50.7) | 3209 (62.5) | 0.308 | 0.074 |
| Initial presentation of provoked VTE | 1940 (18.6) | 363 (14.5) | 246 (17.9) | 315 (22.0) | 1016 (19.8) | 0.197 | 0.084 |
| Prior use of medication | |||||||
| Antiplatelet therapy | 671 (6.4) | 184 (7.3) | 97 (7.0) | 98 (6.9) | 292 (5.7) | 0.067 | 0.040 |
| Corticosteroids | 4267 (40.8) | 1005 (40.1) | 563 (40.9) | 589 (41.2) | 2110 (41.1) | 0.024 | 0.036 |
| NSAIDs | 2638 (25.2) | 577 (23.0) | 269 (19.5) | 357 (25.0) | 1435 (28.0) | 0.199 | 0.093 |
| ACE inhibitors | 2783 (26.6) | 768 (30.6) | 395 (28.7) | 368 (25.8) | 1252 (24.4) | 0.140 | 0.058 |
| Aspirin | 341 (3.3) | 85 (3.4) | 35 (2.5) | 47 (3.3) | 174 (3.4) | 0.050 | 0.059 |
| β-blockers | 3773 (36.1) | 1024 (40.8) | 534 (38.8) | 514 (36.0) | 1701 (33.1) | 0.159 | 0.039 |
| Calcium channel blockers | 2476 (23.7) | 654 (26.1) | 365 (26.5) | 312 (21.8) | 1145 (22.3) | 0.109 | 0.061 |
| Selective serotonin reuptake inhibitors | 1925 (18.4) | 430 (17.1) | 282 (20.5) | 259 (18.1) | 954 (18.6) | 0.086 | 0.073 |
| Proton pump inhibitors | 3320 (31.8) | 819 (32.6) | 444 (32.2) | 460 (32.2) | 1597 (31.1) | 0.033 | 0.018 |
| Loop diuretics | 2080 (19.9) | 631 (25.1) | 328 (23.8) | 279 (19.5) | 842 (16.4) | 0.217 | 0.057 |
| Potassium-sparing diuretics | 755 (7.2) | 191 (7.6) | 127 (9.2) | 101 (7.1) | 336 (6.5) | 0.099 | 0.030 |
| Thiazide diuretics | 1168 (11.2) | 311 (12.4) | 176 (12.8) | 137 (9.6) | 544 (10.6) | 0.101 | 0.078 |
| Vasodilators | 561 (5.4) | 145 (5.8) | 88 (6.4) | 62 (4.3) | 266 (5.2) | 0.091 | 0.053 |
| Estrogens | 366 (3.5) | 73 (2.9) | 29 (2.1) | 48 (3.4) | 216 (4.2) | 0.120 | 0.057 |
| COX-2 inhibitors | 315 (3.0) | 67 (2.7) | 27 (2.0) | 38 (2.7) | 183 (3.6) | 0.098 | 0.051 |
ACE, angiotensin-converting enzyme; COX-2, cyclooxygenase-2; DVT, deep vein thrombosis; DOAC, direct oral anticoagulant; IPTW, inverse probability of treatment weighting; NSAIDs, nonsteroidal anti-inflammatory drugs; PE, pulmonary embolism; SMD, standardized mean difference; VTE, venous thromboembolism.
Maximum SMD of 6 SMDs from the 4-group pairwise comparisons, where SMD > 0.1 was considered a nonnegligible difference.
3.2. Recurrent VTE
The crude incidence rates of hospitalization due to recurrent VTE per 100 person-years were 0.42 for consistently high adherence, 0.77 for gradually declining adherence, 0.37 for rapidly declining adherence, and 1.61 for no extended treatment. Compared with no extended treatment, consistently high adherence was associated with a significantly decreased risk of hospitalization due to recurrent VTE (hazard ratio [HR] = 0.23; 95% CI, 0.12-0.45) (Table 3). Patients with gradually declining adherence and rapidly declining adherence seemed to have a lower risk of hospitalization due to recurrent VTE than those with no extended treatment, although the results were not statistically significant (Table 3).
Table 3.
Recurrent venous thromboembolism and major bleeding events by trajectory of adherence to extended treatment with warfarin.
| Clinical outcome | Trajectory of medication adherence |
|||
|---|---|---|---|---|
| Consistently high adherence (n = 2509) | Gradually declining adherence (n = 1377) | Rapidly declining adherence (n = 1429) | No extended treatment (n = 5133) | |
| Recurrent venous thromboembolism | ||||
| Events, no. (%) | 9 (0.36) | 3 (0.22) | 1 (0.07) | 93 (1.81) |
| Time to event, median (IQR), mo | 5.5 (2.6-6.4) | 2.6 (1.0-4.3) | 2.5 (2.5-2.5) | 7.4 (2.9-16.2) |
| Incidence per 100 person-years (95% CI) | 0.42 (0.22-0.81) | 0.77 (0.25-2.39) | 0.37 (0.05-2.61) | 1.61 (1.31-1.97) |
| Follow-up duration, mean (SD), mo | 10.2 (9.6) | 3.4 (2.4) | 2.3 (1.0) | 13.6 (16.1) |
| Comparison with no extended treatment, HR (95% CI)a | 0.23 (0.12-0.45)b | 0.29 (0.08-1.06) | 0.14 (0.02-1.24) | 1 (Reference) |
| Comparison within warfarin-treated groups, HR (95% CI)a | 1.60 (0.17-15.23) 0.78 (0.19-3.25) |
2.05 (0.17-24.63) 1 (Reference) |
1 (Reference) | |
| Major bleeding event | ||||
| Events, no. (%) | 26 (1.04) | 6 (0.44) | 20 (1.40) | 28 (0.55) |
| Time to event, median (IQR), mo | 8.5 (3.0-14.2) | 2.8 (0.5-4.2) | 1.3 (0.7-2.5) | 10.7 (0.8-22.2) |
| Incidence per 100 person-years (95% CI) | 1.22 (0.83-1.80) | 1.54 (0.69-3.43) | 7.41 (4.78-11.48) | 0.49 (0.34-0.71) |
| Follow-up duration, mean (SD), mo | 10.2 (9.5) | 3.4 (2.4) | 2.3 (1.0) | 13.4 (16.0) |
| Comparison with no extended treatment, HR (95% CI)a | 2.08 (1.18-3.64)c | 2.10 (0.74-5.95) | 9.19 (4.38-19.29)b | 1 (Reference) |
| Comparison within warfarin-treated groups, HR (95% CI)a | 0.23 (0.11-0.47)b 0.99 (0.35-2.79) |
0.23 (0.08-0.64)d 1 (Reference) |
1 (Reference) | |
HR, hazard ratio.
Estimated using the inverse probability of the treatment-weighted Cox proportional hazards model.
P < .0001.
P < .05.
P < .01.
3.3. Major bleeding events
For hospitalization due to major bleeding events, crude incidence rates per 100 person-years were 1.22 for consistently high adherence, 1.54 for gradually declining adherence, 7.41 for rapidly declining adherence, and 0.49 for no extended treatment. Compared with no extended treatment, the 3 warfarin trajectory groups were associated with increased risk of hospitalization due to major bleeding events, of which rapidly declining adherence (HR = 9.19; 95% CI, 4.38-19.29) and consistently high adherence (HR = 2.08; 95% CI, 1.18-3.64) showed significantly increased risk of hospitalization due to major bleeding events (Table 3). In pairwise comparisons among the warfarin-treated groups, consistently high adherence (HR = 0.23; 95% CI, 0.11-0.47) and gradually declining adherence (HR = 0.23; 95% CI, 0.08-0.64) were associated with lower risk of hospitalization due to major bleeding events compared with rapidly declining adherence.
4. Discussion
This population-based GBTM analysis provides real-world evidence for trajectories of adherence to extended warfarin treatment and their associated effectiveness and safety outcomes. We identified 4 extended treatment trajectories during 6 months of extended anticoagulant treatment for VTE. Compared with individuals receiving no extended treatment, patients with consistently high treatment adherence were associated with a significantly decreased risk of hospitalization due to recurrent VTE. This association appeared similar for patients with gradually or rapidly declining warfarin adherence, although there was no statistical significance due, in part, to small event sizes. For safety outcomes, the 3 warfarin-treated adherence groups were associated with increased risk of hospitalization due to major bleeding events compared with no extended treatment, and consistently high adherence and gradually declining adherence were associated with decreased risk of hospitalization due to major bleeding events compared with rapidly declining adherence.
Although adherence and persistence of anticoagulation therapy are important factors in preventing VTE recurrence [24], few real-world studies have reported the patterns of warfarin adherence during extended anticoagulant therapy after 3 to 6 months of initial treatment. Among warfarin users at a high risk of recurrent VTE (ie, with cancer or without reversible risk factors for VTE), patients who were medication nonadherent—defined as 1-year PDC <80%—accounted for 77% of the cohort and had a risk of VTE recurrence 3 times as high as patients who were medication adherent [25]. Using a criterion of 80% for PDC would not have differentiated the groups with gradually or rapidly declining medication adherence in our study. Indeed, compared with no extended treatment, gradually declining adherence or rapidly declining adherence—which would have been categorized as medication nonadherent based on PDC <80%—had different recurrent VTE and major bleeding outcomes.
This study showed that consistently high adherence or gradually declining adherence was associated with about a 2-fold increased risk of hospitalization due to major bleeding compared with receiving no extended treatment. In contrast, rapidly declining adherence was associated with a more than 9-fold increased risk of hospitalization due to major bleeding. This finding may be explained by reverse causality, which implies that some patients might discontinue warfarin because of bleeding events, rather than by poor medication adherence causing bleeding events. In fact, the hospitalization due to major bleeding events for 20 patients with rapidly declining adherence occurred in a short time (median time to event, 1.3 [IQR, 0.7-2.5] months). It is also plausible that some patients in this group may have experienced minor bleeding events that led to the discontinuation of anticoagulation therapy.
Regarding major bleeding events, a similar investigation into DOAC adherence, instead of warfarin in the present study, showed that persistent adherence to DOAC was not associated with an increased risk of major bleeding [26]. This can be explained by the use of DOAC being associated with a lower risk of bleeding compared with warfarin in patients with VTE or atrial fibrillation [27].
This study had several limitations. First, we were unable to capture potential confounders, such as international normalized ratio, race or ethnicity, or reasons for treatment discontinuation in claims-based analyses. Second, our PDC calculations may overestimate adherence because we were unable to determine whether patients consumed medications as prescribed. Third, because we consider the occurrence of an outcome when calculating PDC, we cannot rule out reverse causality. Therefore, we could not conclude causality between poor adherence to warfarin and worsened health outcomes. Some patients may have discontinued warfarin because of major or minor bleeding events. Fourth, we were unable to fully explore the associations of adherence trajectories of extended warfarin therapy because of a small number of events during a short follow-up in groups with gradually or rapidly declining adherence. Fifth, although we used previously validated algorithms with high positive predictive values to identify outcomes, incomplete, missing, or miscoded claims may occur; however, coding errors are likely to be nondifferential across groups.
5. Conclusions
This cohort study identified 4 extended treatment trajectories of warfarin adherence during 6 months of extended therapy among patients who completed 6 months of initial VTE treatment. Compared with no extended treatment, the consistent use of warfarin during extended treatment was associated with a decreased risk of hospitalization due to recurrent VTE. Although all warfarin-treated trajectory groups were associated with increased risk for hospitalization due to major bleeding events compared with the no extended treatment group, trajectories with greater medication adherence were associated with significantly decreased risks of hospitalization due to major bleeding. Our findings suggest that remaining consistently highly adherent to extended warfarin treatment was associated with a significantly decreased risk of hospitalization due to recurrent VTE, although warfarin use was associated with an increased risk of hospitalization due to major bleeding compared with no extended treatment.
Acknowledgments
Funding
This research was supported by the BMS/Pfizer Alliance American Thrombosis Investigator Initiated Research Program. The funder was not involved in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.
Ethics statement
The University of Florida Institutional Review Board approved this study, which waived the requirement for obtaining informed consent because only deidentified data were used.
Author contributions
H.-R.K., W.-H.L.-C., C.E.D., E.A.D., and H.P. wrote the manuscript. H.-R.K., W.-H.L.-C., C.E.D., E.A.D., and H.P. designed the research. H.-R.K. and H.P. performed the research. H.-R.K., B.L.J., and P.-L.H. analyzed the data.
Relationship disclosure
W.-H.L.-C. reports receiving research funding from Merck Sharp & Dohme Corp outside the submitted work. E.A.D. reports receiving honoraria for training and education from Bristol-Myers Squibb/Pfizer outside the submitted work. C.E.D. reports being a stockholder of Portola Pharmaceuticals, receiving honoraria for being on the advisory board of Bristol-Myers Squibb outside the submitted work, and currently being employed by Sanofi U.S., although the work in this study was conducted when she was affiliated with the University of Florida. All other authors declare no competing interests.
Data availability
The data that support the findings of this study are available from data vendors under a data use agreement.
Footnotes
Handling Editor: Dr Suzanne Cannegieter
References
- 1.Centers for Disease Control and Prevention Data and Statistics on Venous Thromboembolism. 2020. https://www.cdc.gov/ncbddd/dvt/data.html Available from:
- 2.Heit J.A. Epidemiology of venous thromboembolism. Nat Rev Cardiol. 2015;12:464–474. doi: 10.1038/nrcardio.2015.83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kearon C., Akl E.A., Ornelas J., Blaivas A., Jimenez D., Bounameaux H., et al. Antithrombotic therapy for VTE disease: CHEST guideline and expert panel report. Chest. 2016;149:315–352. doi: 10.1016/j.chest.2015.11.026. [DOI] [PubMed] [Google Scholar]
- 4.Ortel T.L., Neumann I., Ageno W., Beyth R., Clark N.P., Cuker A., et al. American Society of Hematology 2020 guidelines for management of venous thromboembolism: treatment of deep vein thrombosis and pulmonary embolism. Blood Adv. 2020;4:4693–4738. doi: 10.1182/bloodadvances.2020001830. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.National Institute for Health and Care Excellence Venous thromboembolic diseases: diagnosis, management and thrombophilia testing. NICE guideline. 2020. https://www.nice.org.uk/guidance/ng158/resources/venous-thromboembolic-diseases-diagnosis-management-and-thrombophilia-testing-pdf-66141847001797 Available from: [PubMed]
- 6.Stevens S.M., Woller S.C., Baumann Kreuziger L., Bounameaux H., Doerschug K., Geersing G.J., et al. Executive summary: antithrombotic therapy for VTE disease: second update of the CHEST guideline and expert panel report. Chest. 2021;160:2247–2259. doi: 10.1016/j.chest.2021.07.056. [DOI] [PubMed] [Google Scholar]
- 7.Wright J.N., Vazquez S.R., Kim K., Jones A.E., Witt D.M. Assessing patient preferences for switching from warfarin to direct oral anticoagulants. J Thromb Thrombolysis. 2019;48:596–602. doi: 10.1007/s11239-019-01915-9. [DOI] [PubMed] [Google Scholar]
- 8.Patel S.M., Wang T., Outler D.L., Elliott J., Knauss M., Peasah S.K., et al. Low persistence to rivaroxaban or warfarin among patients with new venous thromboembolism at a safety net academic medical center. J Thromb Thrombolysis. 2020;49:287–293. doi: 10.1007/s11239-019-01959-x. [DOI] [PubMed] [Google Scholar]
- 9.Deitelzweig S.B., Lin J., Kreilick C., Hussein M., Battleman D. Warfarin therapy in patients with venous thromboembolism: patterns of use and predictors of clinical outcomes. Adv Ther. 2010;27:623–633. doi: 10.1007/s12325-010-0056-z. [DOI] [PubMed] [Google Scholar]
- 10.Alotaibi G.S., Wu C., Senthilselvan A., McMurtry M.S. The validity of ICD codes coupled with imaging procedure codes for identifying acute venous thromboembolism using administrative data. Vasc Med. 2015;20:364–368. doi: 10.1177/1358863X15573839. [DOI] [PubMed] [Google Scholar]
- 11.White R.H., Garcia M., Sadeghi B., Tancredi D.J., Zrelak P., Cuny J., et al. Evaluation of the predictive value of ICD-9-CM coded administrative data for venous thromboembolism in the United States. Thromb Res. 2010;126:61–67. doi: 10.1016/j.thromres.2010.03.009. [DOI] [PubMed] [Google Scholar]
- 12.Agnelli G., Buller H.R., Cohen A., Curto M., Gallus A.S., Johnson M., et al. Apixaban for extended treatment of venous thromboembolism. N Engl J Med. 2013;368:699–708. doi: 10.1056/NEJMoa1207541. [DOI] [PubMed] [Google Scholar]
- 13.Bauersachs R., Berkowitz S.D., Brenner B., Buller H.R., Decousus H., Gallus A.S., et al. Oral rivaroxaban for symptomatic venous thromboembolism. N Engl J Med. 2010;363:2499–2510. doi: 10.1056/NEJMoa1007903. [DOI] [PubMed] [Google Scholar]
- 14.Couturaud F., Sanchez O., Pernod G., Mismetti P., Jego P., Duhamel E., et al. Six months vs extended oral anticoagulation after a first episode of pulmonary embolism: the PADIS-PE randomized clinical trial. JAMA. 2015;314:31–40. doi: 10.1001/jama.2015.7046. [DOI] [PubMed] [Google Scholar]
- 15.Cunningham A., Stein C.M., Chung C.P., Daugherty J.R., Smalley W.E., Ray W.A. An automated database case definition for serious bleeding related to oral anticoagulant use. Pharmacoepidemiol Drug Saf. 2011;20:560–566. doi: 10.1002/pds.2109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Franklin J.M., Shrank W.H., Pakes J., Sanfélix-Gimeno G., Matlin O.S., Brennan T.A., et al. Group-based trajectory models: a new approach to classifying and predicting long-term medication adherence. Med Care. 2013;51:789–796. doi: 10.1097/MLR.0b013e3182984c1f. [DOI] [PubMed] [Google Scholar]
- 17.Modi A.C., Rausch J.R., Glauser T.A. Patterns of nonadherence to antiepileptic drug therapy in children with newly diagnosed epilepsy. JAMA. 2011;305:1669–1676. doi: 10.1001/jama.2011.506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Nagin D.S. Harvard University Press; Cambridge, Massachusetts: 2005. Group-Based Modeling of Development. [Google Scholar]
- 19.Haviland A.M., Jones B.L., Nagin D.S. Group-based trajectory modeling extended to account for nonrandom participant attrition. Sociolog Methods Res. 2011;40:367–390. [Google Scholar]
- 20.Nagin D.S., Odgers C.L. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010;6:109–138. doi: 10.1146/annurev.clinpsy.121208.131413. [DOI] [PubMed] [Google Scholar]
- 21.Xu S., Ross C., Raebel M.A., Shetterly S., Blanchette C., Smith D. Use of stabilized inverse propensity scores as weights to directly estimate relative risk and its confidence intervals. Value Health. 2010;13:273–277. doi: 10.1111/j.1524-4733.2009.00671.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Austin P.C., Stuart E.A. The performance of inverse probability of treatment weighting and full matching on the propensity score in the presence of model misspecification when estimating the effect of treatment on survival outcomes. Stat Methods Med Res. 2017;26:1654–1670. doi: 10.1177/0962280215584401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Austin P.C. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Simul Comput. 2009;38:1228–1234. [Google Scholar]
- 24.Boutitie F., Pinede L., Schulman S., Agnelli G., Raskob G., Julian J., et al. Influence of preceding length of anticoagulant treatment and initial presentation of venous thromboembolism on risk of recurrence after stopping treatment: analysis of individual participants' data from seven trials. BMJ. 2011;342:d3036. doi: 10.1136/bmj.d3036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Chen S.Y., Wu N., Gulseth M., LaMori J., Bookhart B.K., Boulanger L., et al. One-year adherence to warfarin treatment for venous thromboembolism in high-risk patients and its association with long-term risk of recurrent events. J Manag Care Pharm. 2013;19:291–301. doi: 10.18553/jmcp.2013.19.4.291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kang H.-R., Jones B.L., Lo-Ciganic W.-H., DeRemer C.E., Dietrich E.A., Huang P.-L., et al. Adherence trajectories of extended direct-acting oral anticoagulants and risk of recurrent venous thromboembolism and major bleeding: a retrospective cohort study. Blood. 2021;138:4059. [Google Scholar]
- 27.Novak A.R., Shakowski C., Trujillo T.C., Wright G.C., Mueller S.W., Kiser T.H. Evaluation of safety and efficacy outcomes of direct oral anticoagulants versus warfarin in normal and extreme body weights for the treatment of atrial fibrillation or venous thromboembolism. J Thromb Thrombolysis. 2022;54:276–286. doi: 10.1007/s11239-022-02668-8. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data that support the findings of this study are available from data vendors under a data use agreement.