Abstract
Cancer associated thrombosis (CAT) is a leading cause of death among patients with cancer. It is not clear if non-clinical factors are associated with anticoagulation receipt. We conducted a retrospective cohort study of Optum’s de-identified Clinformatics® Database of adults with cancer diagnosed between 2009 and 2016 who developed CAT, treated with an outpatient anticoagulant (warfarin, low molecular weight heparin (LMWH), or a direct oral anticoagulant (DOAC)). Of 12,622 patients, three months after an episode of CAT, 1,485 (12%) were on LMWH, 1,546 (12%) on DOACs, and 9,591 (76%) were on warfarin. When controlling for other factors, anticoagulant use was significantly associated with socioeconomic factors, region, co-morbidities, type of thrombosis, and cancer subtype. Patients with a bachelor’s degree or greater level of education were less likely to receive warfarin (OR: 0.77; 95% CI: [0.59, 0.99]; p < 0.046) or DOACs (OR: 0.67; 95% CI: [0.55, 0.82]; p < 0.001) compared to LMWH. Patients with higher income levels were more likely to receive LMWH or DOACs compared to warfarin, while patients across all income levels were equally likely to receive LMWH or DOACs. Non-clinical factors including income, education, and region, are associated with anticoagulation receipt three months after an episode of CAT. Sociodemographic factors may result in some patients receiving suboptimal care and contribute to non-guideline concordant care for CAT.
Keywords: Anticoagulants, Drug utilization, Neoplasms, Socioeconomic factors, Venous thrombosis
Introduction
About one in ten cancer patients receiving chemotherapy dies from venous thromboembolism (VTE) which is the second leading cause of death in this group [1]. Cancer associated thrombosis (CAT) is associated with significant morbidity and can interrupt cancer directed therapy. Randomized controlled trials meta-analyses have confirmed a reduction in VTE recurrence with LMWH therapy compared to vitamin K antagonists (VKAs), with a similar complication rate of bleeding [2–7]. As a result, numerous national and international guidelines suggested the use of LMWH over VKA for the management of acute VTE in patients with cancer [6, 7].
Adding further complexity to the management of CAT, recent clinical trials have demonstrated that the direct oral anticoagulants, DOACs (apixaban, edoxaban, and rivaroxaban) may offer an oral anticoagulant option for some patients with CAT [8–11]. Accordingly, DOACs have also been embraced by guidelines for appropriately selected patients with CAT. Unfortunately, nearly a decade of data has shown that half of patients with CAT fail to receive guideline-concordant therapy [12]. It is not well understood why clinical practice differs from guideline recommendations. As treatment options for CAT expand, it is increasingly important to understand the factors associated with anticoagulant use in a non-trial setting to close the gap between ideal and actual care.
It has been hypothesized that cost and regional practice patterns may partly explain the variation in anticoagulant use for CAT [12–17]. This study aimed to determine if non-clinical factors including socioeconomic status (SES) and geographic location were associated with anticoagulation receipt for the treatment of CAT.
Methods
Data source
We used Optum’s de-identified Clinformatics® Data Mart Database, a large commercial and Medicare Advantage claims database containing data on over 61 million privately-insured individuals. The database contains linked, de-identified data on inpatient care, outpatient care, prescription claims, and geographic data.
Patient selection
Data from January of 2009 through June of 2016 (Supplemental Fig. 1 flow diagram) was used. Inclusion criteria included age over 18 years, at least twelve months of continuous data available before initial diagnosis of VTE (allowing up to a 30-day lapse in insurance coverage). A cohort of patients with “active cancer” was identified as those with at least one International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) code for a cancer diagnosis (Supplemental Table 1) with a simultaneous or subsequent cancer directed treatment as defined by Healthcare Common Procedure Coding System Codes (HCPCS) code or National Drug Code (NDC) in any position, from inpatient or outpatient claims data (Supplemental Table 2 and Supplemental Table 3). Both chemotherapy and hormonal therapies were included as cancer treatments. The HCPCS codes and NDC codes [18] were obtained from the National Cancer Institute and reviewed a priori by the study team. Drugs not thought to be consistent with potential cancer treatment were excluded. If patients had more than one code for cancer on or before their later claim for an episode of VTE (“index VTE”) they were categorized in the multiple cancer group; otherwise they were grouped by cancer subtype.
Patients with active cancer were evaluated for VTE (index VTE) identified by a single corresponding ICD-9-CM code (Supplemental Table 4) followed by a new prescription for an anticoagulant. We did not evaluate arterial thrombotic events in this study. Patients were excluded if they had an outpatient claim for VTE over the 12 months prior to the index VTE diagnosis but a code for a history of VTE (V12.51) was permitted. The ICD-9-CM codes listed in Supplemental Table 1 and Supplemental Table 4 were used to determine the type of cancer and type of thrombosis, respectively.
Anticoagulants were identified by NDC codes (Supplemental Table 5). All of the included ICD-9-CM codes have been validated in other studies with high positive and negative predictive values, especially in conjunction with a new anticoagulant prescription [13, 19–23]. Eligible patients had to have an outpatient prescription claim for an anticoagulant within 30 days of the index VTE. Those with no outpatient claims within this timeframe were excluded.
Measures
Patient groups were identified by the most recent outpatient prescription 90 days after the index VTE. To define anticoagulation receipt at 3 months, we started by identifying each patient’s initial outpatient anticoagulant prescription (“Index Anticoagulant”). Patients prescribed LMWH ≤ 10 days from the date of warfarin prescription were considered to be on warfarin, as this was thought to most likely represent bridging anticoagulation. Anticoagulants were otherwise classified based on the class of drug. Once a patient was assigned to an anticoagulant subgroup, three subsequent events could take place: 1) continuing the anticoagulant for the remainder of the study, 2) changing anticoagulants, or 3) stopping anticoagulation. We therefore algorithmically excluded patients who stopped anticoagulation and classified the remaining patients based on their most recently filled anticoagulant class 90 days after index VTE.
Within the limitations of our dataset, “stopping” anticoagulation may be due to various reasons, including death, change of health insurance, or discontinuing the anticoagulant. The definitions used for this process are presented in Supplemental Table 6. Warfarin exposure was defined as validated by Go et al. [24].
Patient characteristics, including age, were assessed at the time of the index VTE, and hospital claims were assessed for recent hospitalization (defined as within the 4 weeks prior to the index VTE). Medical claims over the twelve months prior to index VTE were reviewed to assess relevant co-morbidities [25], tobacco use [26], and concomitant medications by NDC code. A modified Charlson co-morbidity index (modified to exclude cancer as a component) was calculated for each patient [27]. Optum leverages partners that collect data from primary sources, including public records, purchase transactions, census data, and consumer surveys, to provide socioeconomic variables [28], including greatest educational attainment. Prescription co-pay data were analyzed as the most recent prescription cost for the prescription filled at three months.
Statistical analyses
Descriptive statistics were generated to compare variables such as types of malignancy and patient characteristics by anticoagulant class. Patients who were identified as having “stopped” anticoagulation or classified as unknown/other anticoagulant use were excluded from this analysis, as the focus was on patients continuing anticoagulation. We used analysis of variance (ANOVA) for continuous variables and Chi-square tests and likelihood ratio test in loglinear models for categorical variables. For each condition of the categorical variables, we also applied post hoc Chi-square tests to examine differential prescription rates among the three types of anticoagulants in each covariate stratum, providing an overall insight into variables that are potentially different across anticoagulant groups. Eventually, we used multinomial logistic regression models to estimate the effects of baseline variables that may influence anticoagulant selection.
In contrast to prior studies on this topic that focus on the initial anticoagulant prescribed, we felt that the most recent outpatient prescription 90 days after index VTE was more representative of what patients with active cancer were likely receiving long term, given that many patients change anticoagulants early in the treatment course [13, 26].
We conducted a robustness check of the model by assessing patients who had evidence of ongoing anticoagulation at four months after index VTE. All statistical tests were two-sided and p-value ≤ 0.05 was considered statistically significant. We performed all computations using R version 3.5.0. The study was reviewed by the University of Michigan IRB and deemed exempt.
Results
Table 1 presents patient characteristics and the types of malignancy by anticoagulant class. A total of 14,945 individuals with CAT were started on outpatient anticoagulants. After three months, 12,743 (85.2%) had evidence of ongoing anticoagulant use. A total of 9,591 were on warfarin (75.3%), 1,546 on DOACs (12.1%), 1,485 were on LMWH (11.7%), and 121 were unknown (1.0%). For the purposes of analysis, those with unknown anticoagulant use were excluded (final n = 12,622, 84.5%) (Supplemental Fig. 1).
Table 1.
CAT patient characteristics by anticoagulant class after 3 monthsa
| LMWH (n = 1,485) | Warfarin (n = 9,591) | DOACs (n = 1,546) | p-value | |
|---|---|---|---|---|
|
| ||||
| Age (mean (SD)) | 60.7 (12.5) | 68.1 (12.5) | 66.9 (12.9) | < 0.001 |
| Age ranges n (%) | < 0.001 | |||
| 18–34 | 38 (2.6%) | 107 (1.1%) | 25 (1.6%) | < 0.001 |
| 35–54 | 372 (25.1%) | 1,319 (13.8%) | 236 (15.3%) | < 0.001 |
| 55–64 | 538 (36.2%) | 1,996 (20.8%) | 336 (21.7%) | < 0.001 |
| 65–74 | 328 (22.1%) | 2,832 (29.5%) | 488 (31.6%) | < 0.001 |
| 75–84 | 187 (12.6%) | 2,706 (28.2%) | 340 (22.0%) | < 0.001 |
| 85 + | 22 (1.5%) | 631 (6.6%) | 121 (7.8%) | < 0.001 |
| Sex n male (% male) | 652 (43.9%) | 4,351 (45.4%) | 690 (44.6%) | 0.531 |
| Hospitalization n (%) | 263 (17.7%) | 1,629 (17.0%) | 221 (14.3%) | 0.018 |
| History of VTE n (%) | 123 (8.3%) | 857 (8.9%) | 102 (6.6%) | 0.009 |
| ADP Inhibitor n (%) | 40 (0.3%) | 501 (4.0%) | 73 (0.6%) | < 0.001 |
| Malignancy | < 0.001 | |||
| Other b | 376 (25.3%) | 1882 (19.6%) | 338 (21.9%) | |
| Multiple Cancers | 199 (13.4%) | 886 (9.2%) | 152 (9.8%) | < 0.001 |
| Cardiac, neuroendocrine, other thoracic, genitourinary, metastatic or unspecified | 102 (6.9%) | 604 (6.3%) | 115 (7.4%) | 0.200 |
| Sarcoma/soft tissue | 45 (3.0%) | 255 (2.7%) | 60 (3.9%) | 0.025 |
| Other abdominalc | 30 (2.0%) | 137 (1.4%) | 21 (1.4%) | 0.194 |
| Head and neck | 19 (1.3%) | 148 (1.5%) | 30 (1.9%) | 0.327 |
| Gastrointestinal | 294 (19.8%) | 1274 (13.3%) | 203 (13.1%) | < 0.001 |
| Pancreas | 113 (7.6%) | 238 (2.5%) | 46 (2.98%) | < 0.001 |
| Colon | 67 (4.5%) | 516 (5.4%) | 84 (5.4%) | 0.365 |
| Rectum | 51 (3.4%) | 302 (3.2%) | 47 (3.0%) | 0.804 |
| Stomach | 43 (2.9%) | 123 (1.3%) | 16 (1.0%) | < 0.001 |
| Esophageal | 20 (1.4%) | 95 (1.0%) | 10 (0.7%) | 0.151 |
| Breast | 186 (12.5%) | 1,926 (20.1%) | 329 (21.3%) | < 0.001 |
| Lung | 186 (12.5%) | 979 (10.2%) | 136 (8.8%) | 0.003 |
| Hematologic | 148 (10.0%) | 1038 (10.8%) | 174 (11.3%) | 0.496 |
| NHL, TCLd | 64 (4.3%) | 460 (4.8%) | 72 (4.7%) | 0.707 |
| Myeloma | 46 (3.1%) | 320 (3.3%) | 56 (3.6%) | 0.722 |
| Leukemia | 30 (2.0%) | 196 (2.0%) | 37 (2.4%) | 0.660 |
| Hodgkin lymphoma | 8 (0.5%) | 62 (0.7%) | 9 (0.6%) | 0.863 |
| Genitourinary | 120 (8.1%) | 1778 (18.5%) | 242 (15.7%) | < 0.001 |
| Prostate | 54 (3.6%) | 1,151 (12.0%) | 155 (10.0%) | < 0.001 |
| Bladder | 45 (3.0%) | 419 (4.4%) | 52 (3.4%) | 0.016 |
| Kidney | 15 (1.0%) | 166 (1.7%) | 21 (1.4%) | 0.086 |
| Testicular | 6 (0.4%) | 42 (0.4%) | 14 (0.9%) | 0.045 |
| Gynecologic | 98 (6.5%) | 419 (4.4%) | 60 (3.9%) | < 0.001 |
| Other gynecologic | 58 (3.9%) | 266 (2.8%) | 38 (2.5%) | 0.030 |
| Ovarian | 40 (2.7%) | 153 (1.6%) | 22 (1.4%) | 0.006 |
| Brain/CNS | 58 (3.9%) | 147 (1.5%) | 24 (1.6%) | < 0.001 |
| Thrombosis type | < 0.001 | |||
| Multiple VTE codes | 383 (25.8%) | 2,700 (28.2%) | 405 (26.2%) | 0.067 |
| LE DVT and PE | 306 (20.6%) | 1,883 (19.6%) | 348 (22.5%) | 0.028 |
| PE | 295 (19.9%) | 1,413 (14.7%) | 265 (17.1%) | < 0.001 |
| LE DVT | 264 (17.8%) | 2,123 (22.1%) | 335 (21.7%) | < 0.001 |
| Othere | 108 (7.3%) | 843 (8.8%) | 95 (6.1%) | < 0.001 |
| UE DVT | 97 (6.5%) | 502 (5.2%) | 82 (5.3%) | 0.118 |
| Portal vein, IVC, renal vein | 32 (2.2%) | 127 (1.3%) | 16 (1.0%) | 0.018 |
| Co-morbid conditions | < 0.001 | |||
| Hypertension | 853 (57.4%) | 6,896 (71.2%) | 1,048 (67.8%) | < 0.001 |
| Mild liver disease | 466 (31.4%) | 1,809 (18.9%) | 306 (19.8%) | < 0.001 |
| Pulmonary disease | 429 (28.9%) | 3,434 (35.8%) | 512 (33.1%) | < 0.001 |
| Diabetes mellitus | 407 (27.4%) | 3,534 (36.8%) | 488 (31.6%) | < 0.001 |
| Cardiac arrhythmia | 375 (25.3%) | 3,027 (31.6%) | 493 (31.9%) | < 0.001 |
| Anemia | 215 (14.5%) | 1,820 (19.0%) | 250 (16.2%) | < 0.001 |
| Valvular disease | 210 (14.1%) | 1,744 (18.2%) | 264 (17.1%) | < 0.001 |
| Cerebrovascular disease | 191 (12.9%) | 1,605 (16.7%) | 217 (14.0%) | < 0.001 |
| Renal disease | 167 (11.2%) | 1,766 (18.4%) | 238 (15.4%) | < 0.001 |
| Congestive heart failure | 163 (11.0%) | 1,774 (18.5%) | 236 (15.3%) | < 0.001 |
| Thrombocytopenia | 154 (10.4%) | 923 (9.6%) | 142 (9.1%) | 0.529 |
ADP adenosine diphosphate receptor, CAT cancer associated thrombosis, CNS central nervous system, DOAC direct oral anticoagulant, LMWH low molecular weight heparin, NHL non-Hodgkin’s lymphoma, TCL T cell lymphoma, VTE venous thromboembolism
Malignancies or thrombosis subtypes with an incidence < 2% are not shown
Other includes multiple cancer codes, unspecified site, metastatic cancer, cardiac, endocrine, neuroendocrine, sarcoma/soft tissue malignancies, and other thoracic, abdominal, or genitourinary cancers
Other abdominal includes cancer of the small intestine, liver, bile ducts, gallbladder, peritoneum, spleen, and ill-defined sites
Includes related disorders
Other thrombosis types include other or unspecified sites
Patients receiving warfarin tended to be older (mean (SD): 68.1 (12.5) years), followed by patients on DOACs (mean (SD): 66.9 (12.9) years), followed by LMWH (mean (SD): 60.7 (12.5) years) (p < 0.001). The sex of the three groups were similar (p = 0.53, Table 1).
Table 2 shows sociodemographic differences between patients with CAT based on anticoagulant use at 3 months. Patients receiving LMWH had the greatest educational attainment and household income, followed by patients receiving the DOACs, and lastly warfarin. Copayments were greatest for LMWH, followed by DOACs, and then warfarin.
Table 2.
CAT socioeconomic characteristics by anticoagulant class after 3 months
| LMWH n = 1,485 | Warfarin n = 9,591 | DOACs n = 1,546 | p-value | |
|---|---|---|---|---|
|
| ||||
| Education (%) | < 0.001 | |||
| < 12th grade | 3 (0.2%) | 42 (0.4%) | 5 (0.3%) | 0.359 |
| High school diploma | 399 (26.9%) | 3153 (32.9%) | 425 (27.5%) | < 0.001 |
| < Bachelor’s degree | 757 (51.0%) | 4934 (51.4%) | 832 (53.8%) | 0.189 |
| ≥ Bachelor’s degree | 295 (19.9%) | 1189 (12.4%) | 240 (15.5%) | < 0.001 |
| Unknown | 31 (2.1%) | 273 (2.9%) | 44 (2.9%) | 0.245 |
| Household income (%) | < 0.001 | |||
| < $40,000 | 258 (17.4%) | 2843 (29.6%) | 343 (22.2%) | < 0.001 |
| $40,000–49,000 | 103 (6.9%) | 878 (9.2%) | 133 (8.6%) | 0.019 |
| $50,000–59,000 | 103 (6.9%) | 763 (8.0%) | 113 (7.3%) | 0.307 |
| $60,000–74,000 | 165 (11.1%) | 903 (9.4%) | 146 (9.4%) | 0.116 |
| $75,000–99,000 | 188 (12.7%) | 1154 (12.0%) | 208 (13.5%) | 0.256 |
| > $100,000 | 510 (34.3%) | 1911 (19.9%) | 443 (28.7%) | < 0.001 |
| Unknown | 158 (10.6%) | 1139 (11.9%) | 160 (10.4%) | 0.112 |
| Race (%) | < 0.001 | |||
| Asian | 37 (2.5%) | 115 (1.2%) | 28 (1.8%) | < 0.001 |
| Black | 195 (13.1%) | 1258 (13.1%) | 166 (10.7%) | 0.032 |
| Hispanic | 88 (5.9%) | 550 (5.7%) | 79 (5.1%) | 0.560 |
| White | 1088 (73.3%) | 7144 (74.5%) | 1173 (75.9%) | 0.256 |
| Unknown | 77 (5.2%) | 524 (5.5%) | 100 (6.5%) | 0.223 |
| Insurance (%) | < 0.001 | |||
| POS | 760 (51.2%) | 2700 (28.2%) | 555 (35.9%) | < 0.001 |
| HMO | 276 (18.6%) | 3266 (34.1%) | 267 (17.3%) | < 0.001 |
| OTH | 210 (14.1%) | 1914 (20.0%) | 474 (30.7%) | < 0.001 |
| EPO | 109 (7.3%) | 459 (4.8%) | 73 (4.7%) | < 0.001 |
| PPO | 96 (6.5%) | 839 (8.8%) | 119 (7.7%) | 0.008 |
| IND | 34 (2.3%) | 413 (4.3%) | 58 (3.8%) | < 0.001 |
| Co-pay | < 0.001 | |||
| < $10.00 | 448 (30.2%) | 6271 (65.4%) | 404 (26.1%) | < 0.001 |
| $10.00–30.00 | 312 (21.0%) | 2966 (30.9%) | 171 (11.1%) | < 0.001 |
| $30.01–50.00 | 372 (25.1%) | 222 (2.3%) | 618 (40.0%) | < 0.001 |
| $50.01–100.00 | 196 (13.2%) | 63 (0.7%) | 189 (12.2%) | < 0.001 |
| $100.01–250.00 | 69 (4.7%) | 58 (0.6%) | 138 (8.9%) | < 0.001 |
| $250.01–500.00 | 33 (2.2%) | 9 (0.1%) | 25 (1.6%) | < 0.001 |
| > $500.00 | 55 (3.7%) | 2 (0.0%) | 1 (0.1%) | < 0.001 |
CAT cancer associated thrombosis, DOAC direct oral anticoagulant, EPO exclusive provider organization, HMO health maintenance organization, IND indemnity, LMWH low molecular weight heparin, OTH other, POS point of service, PPO preferred provider organization
Figures 1 and 2 show the results of multinomial logistic regression analysis comparing the three types of anticoagulation given all covariates considered here OR (1.56; 95% CI: [1.44–1.70]). Compared to their younger counterpart, older patients were more likely to receive oral anticoagulation (warfarin or DOACs) than LMWH (OR: 1.55; 95% CI: [1.46, 1.65]; p < 0.001; OR: 1.56; 95% CI: [1.44, 1.70]; p < 0.001, respectively). Having a history of VTE was associated with 1.26 (95% CI: [1.03, 1.55]; p < 0.001) times the odds of receiving warfarin rather than LMWH, while having a similar likelihood of receiving a DOAC compared to LMWH (OR: 0.89; 95% CI: [0.67, 1.17]; p = 0.39). Patients with a higher Charlson co-morbidity index were more likely to receive warfarin compared to DOACs or LMWH. Relative to patients who were not recently hospitalized, recently hospitalized patients had higher odds to receive LMWH or warfarin than DOAC, while the odds of choosing LMWH over warfarin were not significantly different (Figs. 1 and 2). When adjusting for other variables, sex was not associated with anticoagulant use.
Fig. 1.
Multinomial Logistic Regression Results Showing the Odds Ratio and 95% Confidence Interval for Each Clinical or Sociodemographic Factor: Warfarin (or DOACs) versus LMWH
Fig. 2.
Multinomial Logistic Regression Results Showing the Odds Ratio and 95% Confidence Interval for Each Clinical or Sociodemographic Factor: DOACs (or LMWH) versus Warfarin
Higher socioeconomic status, assessed through educational attainment, was associated with a higher odds of receiving LMWH and DOACs compared to warfarin. Compared with patients with a high school diploma or lower, those with a bachelor’s degree or higher had 1.17 (95% CI: [0.96, 1.43]; p = 0.125) times the odds of receiving DOAC instead of warfarin, and 1.53 (95% CI: [1.25, 1.87]; p < 0.001) times the odds of receiving LMWH instead of warfarin. Moreover, the odds of receiving DOAC or LMWH instead of warfarin was greater for patients with higher income levels, with this difference being significant for all incomes over $60,000 (Fig. 2). However, income level was not associated with DOAC use relative to LMWH (Fig. 1).
After adjusting for other variables, including SES and race, significant regional variation in practice patterns in anticoagulation use was observed as depicted in Fig. 3 with the associated odds ratios. Patients from the East South Central and Mountain states were significantly more likely to receive warfarin over LMWH and the New England and Mid-Atlantic states more likely LMWH. Considering the DOACs, patients from the East South Central and South Atlantic U.S. were significantly more likely to receive DOACs while those from the New England and West North Central regions were more likely to receive LMWH. The DOACs were more often used relative to warfarin for the Mid-Atlantic and South Atlantic States but less often used for the Mountain, Pacific, and New England States.
Fig. 3.
Multinomial Logistic Regression Results Showing Regional Variation in CAT Anticoagulant Use
A sensitivity analysis repeated four months after the VTE event identified 10,134 (80.2% of our initial cohort) patients with evidence of ongoing anticoagulation use. This analysis showed nearly identical findings.
Discussion
In patients with active cancer who experienced an initial VTE, 12,743 (85.2%) of patients had evidence of continued anticoagulant use after three months, with over 9,000 of these patients (75.2%) on warfarin-based anticoagulation instead of LMWH, as suggested by guidelines during the study period (2009–16). We found that non-clinical factors including household income, educational attainment, and geographic region were significantly associated with anticoagulation receipt. The findings raise concern that some patients may not receive the optimal anticoagulant care due to sociodemographic factors and may partially explain non-guideline concordant care for CAT.
Cost has often been hypothesized [12–16] as a reason that patients receive warfarin instead of LMWH. Our study of insured adults supports that hypothesis by showing an association between socioeconomic status and which anticoagulant a patient receives. Specifically, patients with a greater household income or educational attainment were more likely to receive LMWH than warfarin or a DOAC. Patients with higher income were also more likely to receive a DOAC compared to warfarin. Copayments were significantly higher for LMWH or DOACs than warfarin.
As is often seen, regional variation in practice patterns was identified even after controlling for SES and race. In CAT, some regional variation was previously reported at the time of anticoagulant initiation, but this report did not adjust for other factors and focused on initial anticoagulant use [29]. Such regional variation has been similarly noted in anticoagulation for atrial fibrillation [30, 31] with similar geographic trends in DOAC use in this population [32]. From our data, we are not able to explain the significant regional variation identified and this may be a topic for further study. Regional variation may suggest an opportunity for quality improvement initiatives to better standardize care.
In conjunction with the data supporting the use of the DOACs for CAT [8–11], current guidelines for CAT treatment in the United States (U.S.) [7], advocate the use of LMWH or DOACs due to their improved efficacy relative to VKAs, such as warfarin. These guidelines also highlight the importance of awareness of disparities in health and access to care [32]. To date, there are limited formal published data how factors of health disparities, cost, and geographic region impacts guideline concordant anticoagulation for CAT [7, 12]. Our study found that key measures of socioeconomic status, educational attainment and household income, are associated with which anticoagulant a patient receives. It also highlights copayment variation in anticoagulant drugs for CAT. Patients living in certain regions, with lower education, and lower income may receive suboptimal care. Ideally, patients should be able to access an optimal anticoagulant chosen based on clinical factors, bleeding risk, and their preferences.
Efforts are needed to ensure more equitable access in the U.S. to the spectrum of anticoagulant options that are available for CAT, especially considering significant clinical trial evidence supporting the benefits of LMWH and DOAC use. Interventions targeting provider, health system, and national levels are needed to address disparities in anticoagulation care for CAT.
One strength of this study is that it is one of the first to look comprehensively at socioeconomic factors in the treatment of CAT in a U.S. population. This topic has not been well studied; some related studies have included non-U.S. study populations, largely focused on the first anticoagulant prescribed, and/or have not focused specifically on patients with CAT [7, 17, 29, 32, 33]. Given that patients may change anticoagulants, but generally complete at least three months of anticoagulation, our large sample of patients assessed three months after their thrombotic event is more likely reflective of patients’ chronic CAT therapy. Our sensitivity analysis at 4 months strengthens our findings, with the study conclusions unchanged at the later time point.
Limitations of our study are those inherent to a retrospective insurance claims data analysis. We were not able to adjust for the stage or extent of cancer, insurance claims can be inaccurate, and missing data is always a challenge. In addition, observational studies do not allow for adjustment for unmeasured confounders. Furthermore, there are challenges to defining anticoagulant use over time, potentially resulting in misclassification of patients in the outcome groups. All included patients had private insurance, and therefore the findings may not be generalizable to patients with other types of insurance.
Deciding on an anticoagulant for CAT has become an increasingly complex process that must take into account 1) clinical trial data, 2) clinical factors including the thrombotic event, type of cancer, organ function, and bleeding risk, 3) drug factors including ease of reversal, a patient’s other medications, route of administration, need for monitoring and 4) patient preferences. Until more equitable access to anticoagulation can be achieved nationally, clinicians caring for CAT need to educate patients fully on their condition, including the potential benefits of LMWH or DOACs compared to VKAs, and should help patients explore resources that may facilitate receiving the best anticoagulant for their particular situation. Furthermore, it is important to be vigilant of the high costs that patients may incur in association with their therapy as, unfortunately, such costs may need to be considered when determining a care plan.
Supplementary Material
Supplementary Information The online version of this article (https://doi.org/10.1007/s11239-021-02392-9) contains supplementary material, which is available to authorized users.
Highlights.
Non-clinical factors may be associated with anticoagulant use for cancer associated thrombosis.
We conducted a retrospective cohort study of over 12,000 patients using insurance claims data.
Lower household income was associated with warfarin use over other anticoagulants.
Geographic region and educational attainment were also associated with anticoagulant use.
Footnotes
Conflict of interest Geoffrey Barnes reports honoraria from Pfizer/BMS, Janssen, Portola, and AMAG Pharmaceuticals, research funding from Blue Cross Blue Shield of Michigan, BMS, Pfizer, and NHLBI. Marc Carrier reports research funding from BMS, Pfizer, Leo Pharma and honoraria from Bayer, BMS, Pfizer, Leo Pharma, Sanofi, and Servier. Suman Sood reports consulting for Bayer. Mengbing Li reports research funding from the Michigan Institute for Data Science (MIDAS). Zhenke Wu reports research funding from the National Cancer Institute of the NIH under award number P30CA046592 through the Cancer Center Support Grant (CCSG), Development Funds from the Rogel Cancer Center, and an investigator award from Michigan Precision Health Initiative, and funds from Michigan Institute for Data Science (MIDAS). All other authors have no conflict of interests to declare.
References
- 1.Hisada Y, Geddings JE, Ay C, Mackman N (2015) Venous thrombosis and cancer: from mouse models to clinical trials. J Thromb Haemost 13(8):1372–1382 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Meyer G, Marjanovic Z, Valcke J et al. (2002) Comparison of low-molecular-weight heparin and warfarin for the secondary prevention of venous thromboembolism in patients with cancer: a randomized controlled study. Arch Intern Med 162(15):1729–1735 [DOI] [PubMed] [Google Scholar]
- 3.Lee AYY, Peterso Lee AYY, Levine MN et al. (2003) Low-molecular-weight heparin versus a coumarin for the prevention of recurrent venous thromboembolism in patients with cancer. N Engl J Med 349(2):146–153 [DOI] [PubMed] [Google Scholar]
- 4.Hull RD, Pineo GF, Brant RF et al. (2006) Long-term low-molecular-weight heparin versus usual care in proximal-vein thrombosis patients with cancer. Am J Med 119(12):1062–1072 [DOI] [PubMed] [Google Scholar]
- 5.Lee AYY, Kamphuisen PW, Meyer G et al. (2015) Tinzaparin vs warfarin for treatment of acute venous thromboembolism in patients with active cancer: a randomized clinical trial. JAMA 314(7):677–686 [DOI] [PubMed] [Google Scholar]
- 6.Lee AYY, Peterson EA, Wu C (2016) Clinical practice guidelines on cancer-associated thrombosis: a review on scope and methodology. Thromb Res 140(Suppl 1):S119–127 [DOI] [PubMed] [Google Scholar]
- 7.Key NS, Khorana AA, Kuderer NM, et al. (2019) Venous thromboembolism prophylaxis and treatment in patients with cancer: ASCO Clinical Practice Guideline Update. J Clin Oncol. JCO1901461. [DOI] [PubMed] [Google Scholar]
- 8.Raskob GE, van Es N, Verhamme P, et al. (2017) Edoxaban for the treatment of cancer-associated venous thromboembolism. N Engl J Med [DOI] [PubMed] [Google Scholar]
- 9.Young A, Marshall A, Thirlwall J et al. (2017) anticoagulation therapy in selected cancer patients at risk of recurrence of venous thromboembolism: results of the Select-D Pilot Trial. Blood 130(Suppl 1):62528546143 [Google Scholar]
- 10.McBane RD, Wysokinski WE, Le-Rademacher JG, et al. Apixaban and dalteparin in active malignancy-associated venous thromboembolism: The ADAM VTE trial. J Thromb Haemost. October 2019 [DOI] [PubMed] [Google Scholar]
- 11.Li A, Garcia DA, Lyman GH et al. (2019) Direct oral anticoagulant (DOAC) versus low-molecular-weight heparin (LMWH) for treatment of cancer associated thrombosis (CAT): a systematic review and meta-analysis. Thromb Res 173:158–163 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Mahé I, Chidiac J, Helfer H et al. (2016) Factors influencing adherence to clinical guidelines in the management of cancer-associated thrombosis. J Thromb Haemost 14(11):2107–2113 [DOI] [PubMed] [Google Scholar]
- 13.Khorana AA, Dalal M, Lin J et al. (2013) Incidence and predictors of venous thromboembolism (VTE) among ambulatory high-risk cancer patients undergoing chemotherapy in the United States. Cancer 119(3):648–655 [DOI] [PubMed] [Google Scholar]
- 14.Mahé I, Sterpu R, Bertoletti L et al. (2015) Long-term anticoagulant therapy of patients with venous thromboembolism. What are the practices? PLoS ONE 10(6):e0128741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Wittkowsky AK (2006) Barriers to the long-term use of low-molecular weight heparins for treatment of cancer-associated thrombosis. J Thromb Haemost 4(9):2090–2091 [DOI] [PubMed] [Google Scholar]
- 16.Delate T, Witt DM, Ritzwoller D et al. (2012) Outpatient use of low molecular weight heparin monotherapy for first-line treatment of venous thromboembolism in advanced cancer. Oncologist 17(3):419–427 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Khorana AA, McCrae KR, Milentijevic D et al. (2017) Current practice patterns and patient persistence with anticoagulant treatments for cancer-associated thrombosis. Res Pract Thromb Haemost 1(1):14–22 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Cancer Therapy Look-up Tables. National Cancer Institute. https://crn.cancer.gov/resources/codes.html. Accessed 3 Apr 2019 [Google Scholar]
- 19.Tamariz L, Harkins T, Nair V (2012) A systematic review of validated methods for identifying venous thromboembolism using administrative and claims data. Pharmacoepidemiol Drug Saf 21(Suppl 1):154–162 [DOI] [PubMed] [Google Scholar]
- 20.Sanfilippo KM, Wang T-F, Gage BF et al. (2015) Improving accuracy of International Classification of Diseases codes for venous thromboembolism in administrative data. Thromb Res 135(4):616–620 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Heckbert SR, Kooperberg C, Safford MM et al. (2004) Comparison of self-report, hospital discharge codes, and adjudication of cardiovascular events in the Women’s Health Initiative. Am J Epidemiol 160(12):1152–1158 [DOI] [PubMed] [Google Scholar]
- 22.Fang MC, Fan D, Sung SH et al. (2017) Validity of using inpatient and outpatient administrative codes to identify acute venous thromboembolism: the CVRN VTE study. Med Care 55(12):e137–e143 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.White RH, Garcia M, Sadeghi B et al. (2010) Evaluation of the predictive value of ICD-9-CM coded administrative data for venous thromboembolism in the United States. Thromb Res 126(1):61–67 [DOI] [PubMed] [Google Scholar]
- 24.Go AS, Hylek EM, Chang Y et al. (2003) Anticoagulation therapy for stroke prevention in atrial fibrillation: how well do randomized trials translate into clinical practice? JAMA 290(20):2685–2692 [DOI] [PubMed] [Google Scholar]
- 25.Quan H, Sundararajan V, Halfon P et al. (2005) Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 43(11):1130–1139 [DOI] [PubMed] [Google Scholar]
- 26.Desai RJ, Solomon DH, Shadick N et al. (2016) Identification of smoking using Medicare data—a validation study of claims-based algorithms. Pharmacoepidemiol Drug Saf 25(4):472–475 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Charlson ME, Pompei P, Ales KL et al. (1987) A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40(5):373–383 [DOI] [PubMed] [Google Scholar]
- 28.Hershman DL, Tsui J, Wright JD et al. (2015) Household net worth, racial disparities, and hormonal therapy adherence among women with early-stage breast cancer. J Clin Oncol 33(9):1053–1059 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Khorana AA, Yannicelli D, McCrae KR et al. (2016) Evaluation of US prescription patterns: are treatment guidelines for cancer-associated venous thromboembolism being followed? Thromb Res 145:51–53 [DOI] [PubMed] [Google Scholar]
- 30.Hernandez I, Saba S, Zhang Y (2017) Geographic variation in the use of oral anticoagulation therapy in stroke prevention in atrial fibrillation. Stroke 48(8):2289–2291 [DOI] [PubMed] [Google Scholar]
- 31.Christesen AMS, Vinter N, Mortensen LS et al. (2018) Inequality in oral anticoagulation use and clinical outcomes in atrial fibrillation: a Danish nationwide perspective. Eur Heart J Qual Care Clin Outcomes 4(3):189–199 [DOI] [PubMed] [Google Scholar]
- 32.Nathan AS, Geng Z, Dayoub EJ et al. (2019) Racial, ethnic, and socioeconomic inequities in the prescription of direct oral anticoagulants in patients with venous thromboembolism in the United States. Circ Cardiovasc Qual Outcomes 12(4):e005600. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Noble S, Nelson A, Scott J et al. (2020) Patient experience of living with cancer-associated thrombosis in Canada (PELICANADA). Res Pract Thromb Haemost 4(1):154–160. 10.1002/rth2.12274 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.