Abstract
Rationale & Objective
Patients with kidney failure requiring maintenance dialysis have a high risk of cardiovascular events warranting antithrombotic therapies, including oral anticoagulant (OAC) or antiplatelet therapy (APT). However, chronic use of antithrombotic therapy can increase the bleeding risk in patients receiving dialysis. However, little is known about medication use patterns and risk of bleeding events in real-world clinical practice.
Study Design
Retrospective analysis of data from 2 prospective cohort studies.
Setting & Participants
We included 27,612 patients from the Dialysis Outcomes and Practice Patterns Study (DOPPS) and 5,289 patients from the Peritoneal DOPPS (PDOPPS), international cohorts of hemodialysis (HD) and peritoneal dialysis (PD) patients.
Exposures
Patient demographics and comorbid conditions; OAC and APT use.
Outcomes
OAC and APT use; a bleeding composite outcome including a hospitalization or death because of a major bleeding event.
Analytical Approach
Descriptive analyses to explore OAC and APT utilization and crude rates of the bleeding composite outcome and Kaplan–Meier analyses to estimate medication discontinuation.
Results
Baseline OAC and APT use was 9% and 10% in HD patients and 4% and 7% in PD patients, respectively. Patients prescribed antithrombotic drugs were older and more likely to have a history of cardiovascular disease. After 36 months, the Kaplan–Meier estimated proportions of baseline users who remained on therapy were 57% for OAC and 53% for APT. The composite bleeding rates per 100 patient-years among patients with baseline OAC use versus baseline APT use versus neither were 8.6, 5.6, and 4.1 in HD patients and 12.0, 6.1, and 3.9 in PD patients, respectively.
Limitations
Potential for event misclassification; no over-the-counter medication data; rates unadjusted.
Conclusions
Antithrombotic drugs are infrequently prescribed and often discontinued in patients receiving HD or PD. With major bleeding event rates high among antithrombotic users, new strategies are needed to optimize the risks and benefits of antithrombotic agents in the dialysis setting.
Index Words: Dialysis, bleeding, anticoagulant, antiplatelet
Plain-language Summary
Patients on dialysis face a difficult balance: they are at high risk for blood clots but also prone to serious bleeding. We examined real-world data from over 30,000 patients across multiple countries to understand how often blood thinners are used and what outcomes follow. We found that these medications are prescribed infrequently and often discontinued. Importantly, patients who received them experienced more major bleeding events. These findings highlight the urgent need for safer, more tailored approaches to managing clotting risks in dialysis care.
Patients with chronic kidney disease (CKD) often have an indication for antithrombotic medications to reduce the risk of cardiovascular thrombotic events.1, 2, 3 As kidney function declines, however, both anemia and platelet dysfunction lead to a higher underlying risk of bleeding events in patients with kidney failure requiring maintenance dialysis.4,5 Chronic use of antithrombotic therapy can compound the risk of bleeding, which is particularly relevant to those receiving either hemodialysis (HD) or peritoneal dialysis (PD).6,7 Moreover, among patients receiving HD specifically, the chronic use of heparin during sessions and multiple punctures of the vascular access result in an even higher risk of bleeding.8,9 In such a context, the net benefit of antithrombotic agent use in this population remains uncertain.
This uncertainty is reinforced by the lack of representative randomized controlled trials (RCTs) evaluating the safety and efficacy of antithrombotic agents in patients receiving dialysis.10 Often excluded from atrial fibrillation (AF) and coronary artery disease (CAD) RCTs, patients receiving dialysis remain an understudied population in the context of antithrombotic therapies, despite their high risk of cardiovascular thrombotic events.11
The Kidney Disease: Improving Global Outcomes (KDIGO) and cardiology societies have guidance and recommendations on the management of cardiovascular diseases in CKD.12, 13, 14 Current AF guidelines underscore the inconclusive evidence for routinely prescribing antithrombotic drugs to prevent stroke.15 Meta-analyses have attempted to summarize the risks and benefits of anticoagulation therapy in the dialysis setting, but specific guidelines remain uncertain.16,17 For CAD, most guidelines recommend the routine prescription of 1 or more antiplatelet agents for patients receiving dialysis who have experienced major ischemic events.14 However, the number of agents, type of drug, and duration of antiplatelet therapy for patients receiving dialysis with CAD are debatable.16, 17, 18, 19, 20, 21, 22
Lack of evidence contributes to practice variation for antithrombotic drug prescription and may partially explain variability in rates of bleeding events worldwide. A prior Dialysis Outcomes and Practice Patterns Study (DOPPS) investigation has shown wide variations in the prescriptions of antithrombotic medications and rates of bleeding events in patients receiving in-center HD across regions of the world.6 Variations in practice patterns offer an opportunity to characterize patient populations more likely to be prescribed antithrombotic agents and their risk of experiencing adverse events. The lack of representative clinical trials evaluating antithrombotic therapies in dialysis populations, along with high worldwide variation of practice patterns, makes large observational studies valuable tools to generate real-world evidence to better understanding the risk profile and help design future clinical trials.
In this observational study using data from the DOPPS and peritoneal DOPPS (PDOPPS) studies, we aim first to characterize the dialysis populations who are prescribed anticoagulant and antiplatelet therapies in various regions of the world. Second, we seek to estimate the rates of initiation and discontinuation of both types of agents across dialysis modalities. Finally, we describe the rates of major bleeding events in this population.
Methods
Data Source
This retrospective analysis used data from the DOPPS and PDOPPS.23,24 DOPPS is an international prospective cohort study of adult patients receiving HD, ongoing since 1996, with participants coming from 21 countries. Maintenance patients receiving HD were randomly selected from participating facilities in each country. PDOPPS is an international prospective cohort study, ongoing since 2014, designed to identify optimal practices for individuals treated with maintenance PD in 8 countries. Adult patients were selected at random for enrollment from stratified national samples of PD facilities. DOPPS and PDOPPS approval and patient consent were obtained as required by national and local ethics committee regulations. This analysis included data from DOPPS phases 4-7 (2009-2022) and PDOPPS phases 1-2 (2014-2022) and was restricted to patients with prescription medication data available.
Participating countries included Australia, New Zealand (ANZ), Japan, China, South Korea, Thailand (Other Asia-Pacific), Belgium, France, Germany, Italy, Spain, Sweden, United Kingdom, Russia, Turkey (Europe), Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, United Arab Emirates (Gulf Cooperation Council [GCC] countries), and United States and Canada (North America).
Variables of Interest
The variables of interest were prescription of an oral anticoagulant (OAC) or prescription of an antiplatelet therapy (APT). OAC medications were defined as warfarin, apixaban, betrixaban, edoxaban, rivaroxaban, and dabigatran. APT was defined as dipyridamole, cilostazol, vorapaxar, anagrelide, clopidogrel, prasugrel, ticagrelor, and ticlopidine. Aspirin was not included in this analysis because it is frequently an over-the-counter medication and thus not likely to be recorded in patient prescription data.
Baseline factors of interest included region, age, sex, any history of cardiovascular diseases (CVD), diabetes, and bleeding. CVD was included overall and separated into subcategories including AF, cerebrovascular disease, other atherothrombotic disease, and any other CVD. These patient factors of interest chosen were based on an a priori hypothesis or because they were known to be associated with medication use in the literature.
Outcomes of Interest
We evaluated bleeding as a composite outcome, in which patients were considered to have experienced a bleeding event if hospitalization or death (captured prospectively during follow-up) occurred because of a major bleeding event, defined as follows: hospitalization events included cerebral hemorrhage, subdural hematoma, evacuation of hematoma, gastrointestinal (GI) bleed, epistaxis, abnormal bleeding, hemoptysis, hematuria, and vascular access bleeding. Deaths attributed to bleeding included hemorrhagic stroke, GI hemorrhage, other hemorrhage, and hemorrhage from any of the following: transplant site, vascular access, dialysis circuit, ruptured vascular aneurysm, or surgery. Research staff at participating facilities code hospitalizations using a standardized coding list developed for DOPPS, which includes diagnosis and procedure codes. Research staff use a range of sources to determine these data including patient notes, discharge summaries, and the associated hospital discharge codes.
Baseline factors and medication use were collected at the time of DOPPS/PDOPPS enrollment. Follow-up began at DOPPS/PDOPPS enrollment and ended at the time of death, 7 days after leaving the facility because of transfer or change in modality, loss to follow-up, or administrative end of study phase (whichever occurred first).
Statistical Analysis
We first described patient characteristics of the study population and compared those who were versus those who were not prescribed an OAC or APT at baseline. We also described the patient population stratified by modality (HD vs PD). The percentages of patients prescribed OAC and APT, stratified by patients receiving HD or PD, were reported overall and across the patient subgroups, defined above. Logistic regression models were used to assess the associations between these patient factors of interest and baseline use of OAC or APT. Regression models were run in the combined HD and PD populations and again separately by HD and PD cohorts. The odds ratios and 95% confidence interval (CI) were reported for all covariates of interest, with models also adjusting for dialysis vintage (time since dialysis initiation), vascular access type (HD only), body mass index, and serum albumin levels.
Among patients without prescription of an OAC or APT at study enrollment, the Kaplan–Meier method was used to estimate 36-month cumulative incidence of medication initiation. We similarly estimated the proportions of patients who continued each medication, among those with a baseline prescription. In the phases of DOPPS and PDOPPS included in this analysis, active prescription (yes/no) of OAC and APT was captured at enrollment and every 4 or 6 months during follow-up, giving the Kaplan–Meier plots a stepwise character. Among patients with an active prescription at enrollment, discontinuation was defined as first instance of no active prescription during follow-up. Similarly, among patients with no active prescription at enrollment, initiation was defined as first instance of active prescription during follow-up. Patients without information about medication use after the baseline visit were excluded from the discontinuation and initiation analysis.
The crude composite bleeding rate was calculated per 100 patient-years, overall, by dialysis modality, and by OAC and APT use. For individual event types that contributed the most events to the composite bleeding rate—vascular access related hospitalization, GI bleeding hospitalization, and death because of hemorrhagic stroke—event rates for these specific bleeding types were calculated. The 95% CIs for event rates were estimated using Poisson regression.
Results
This analysis included 27,612 patients receiving HD and 5,289 patients receiving PD (Fig S1). Patient characteristics overall, and by baseline OAC and APT use are summarized in Table 1. The average age was 64 years with a median dialysis vintage of 1.8 years. Baseline OAC and APT were more commonly prescribed in HD (9% and 10%, respectively) than in PD (4% and 7%, respectively). Patients prescribed (vs not prescribed) an OAC tended to be older, more likely to have a history of CVD, and more likely to dialyze with a catheter. Patients prescribed (vs not prescribed) an APT tended to be older and more likely to have a history of CVD and diabetes. The proportion of patients with a history of cardiovascular disease was 63% in the HD cohort and 37% in the PD cohort (Tables S1 and S2).
Table 1.
Patient characteristics, overall and by oral anticoagulant (OAC) and antiplatelet use at baseline.
| Anticoagulant Use |
Antiplatelet Use |
||||
|---|---|---|---|---|---|
| Overall | Yes |
No |
Yes |
No |
|
| (n = 32,901) | (n = 2,821, 8%) | (n = 30,080, 92%) | (n = 3,028, 9%) | (n = 29,873, 91%) | |
| Demographics | |||||
| Age, y, mean ± SD | 64 ± 15 | 69 ± 12 | 63 ± 15 | 68 ± 12 | 64 ± 15 |
| Region, % | |||||
| Australia-New Zealand | 4% | 4% | 4% | 5% | 4% |
| Europe | 45% | 56% | 45% | 44% | 46% |
| GCC | 2% | 1% | 2% | 3% | 2% |
| North America | 16% | 22% | 16% | 19% | 16% |
| Other Asia-Pacific | 32% | 17% | 33% | 29% | 32% |
| Sex (% male) | 61% | 65% | 61% | 65% | 61% |
| Dialysis vintagea, y, median (IQR) | 1.8 [0.3-5.1] | 2.1 [0.4-5.7] | 1.7 [0.3-5.0] | 2.0 [0.4-5.1] | 1.7 [0.3-5.1] |
| Primary cause of kidney failure, % | |||||
| Diabetes | 33% | 30% | 33% | 49% | 31% |
| Glomerulonephritis/vasculitis | 22% | 20% | 22% | 12% | 22% |
| Hypertension/large vessel disease | 17% | 20% | 17% | 18% | 17% |
| Other | 28% | 30% | 28% | 21% | 29% |
| Comorbid condition history, % | |||||
| Prior kidney transplant | 7% | 8% | 7% | 5% | 7% |
| Diabetes | 43% | 45% | 43% | 61% | 41% |
| Hypertension | 88% | 89% | 88% | 92% | 88% |
| Any cardiovascular disease | 59% | 87% | 56% | 87% | 56% |
| Atrial fibrillation | 13% | 55% | 9% | 15% | 12% |
| Cerebrovascular disease | 15% | 22% | 14% | 30% | 13% |
| Other atherothrombotic disease | 41% | 60% | 40% | 74% | 38% |
| Other cardiovascular disease | 33% | 58% | 30% | 44% | 31% |
| Any bleeding | 5% | 7% | 5% | 6% | 5% |
| Gastrointestinal bleeding | 4% | 6% | 4% | 4% | 4% |
| Cerebral hemorrhage | 2% | 2% | 2% | 3% | 2% |
| Dialysis treatments | |||||
| Vascular access type, %b | |||||
| Catheter | 25% | 35% | 24% | 27% | 25% |
| Fistula | 69% | 58% | 70% | 65% | 70% |
| Graft | 6% | 7% | 6% | 9% | 6% |
| Laboratory/biometric markers | |||||
| Body mass index, kg/m2, mean ±SD | 25.4 ± 5.8 | 26.8 ± 6.0 | 25.3 ± 5.8 | 25.8 ± 5.9 | 25.4 ± 5.8 |
| Serum albumin, g/dL, mean ±SD | 3.6 ± 0.5 | 3.6 ± 0.6 | 3.6 ± 0.5 | 3.6 ± 0.6 | 3.6 ± 0.5 |
| CRP, mg/L, median [IQR] | 4.3 [1.4-11.2] | 5.8 [2.4-16.0] | 4.0 [1.3-11.0] | 5.0 [1.8-13.0] | 4.2 [1.3-11.0] |
| HbA1c, % | 6.3 ± 1.5 | 6.3 ± 1.3 | 6.3 ± 1.5 | 6.6 ± 1.5 | 6.3 ± 1.5 |
| Medication use,% | |||||
| OAC therapy | 9% | 100% | -- | 5% | 9% |
| Direct factor Xa inhibitors | 0.1% | 1% | -- | 0.1% | 0.1% |
| Direct thrombin inhibitors | 0% | 0% | -- | 0% | 0% |
| Fondaparinux | 0.1% | 1% | -- | 0.1% | 0.1% |
| Vitamin K antagonists | 8% | 98% | -- | 5% | 9% |
| Antiplatelet agents | 9% | 6% | 10% | 100% | -- |
| Clopidogrel | 9% | 6% | 9% | 97% | -- |
| Other P2Y inhibitors | 0.3% | 0.1% | 0.3% | 3% | -- |
Abbreviations: CRP, C-reactive protein test; GCC, Gulf Cooperation Council; IQR, interquartile range; SD, standard deviation.
Dialysis vintage: time since dialysis initiation.
Among patients receiving hemodialysis.
Overall, 17% of patients were prescribed either an OAC or APT at baseline—19% of HD patients and 11% of PD patients. The use of both OAC and APT was infrequent (0.4% overall; 0.5% in HD and 0.2% in PD). The prevalence rates of baseline OAC and APT use by various strata of interest are shown in Table 2. Use of OAC was highest among patients with a history of AF (37% in HD and 41% in PD), whereas use of APT was most common among patients with cerebrovascular disease (18 % in HD and 19% in PD). Within the OAC group, the proportion of patients using vitamin K antagonists was 98%. Additionally, there was a noticeable increase in medication use with advancing age, especially in those aged 80 years and above. Sex differences were relatively modest, with slightly higher usage in men compared with women.
Table 2.
Prevalence of oral anticoagulant (OAC) and antiplatelet agent use, overall and across relevant strata.
| Counts | HD Participants (DOPPS) (n = 27,612) |
PD Participants (PDOPPS) (n = 5,289) |
||||
|---|---|---|---|---|---|---|
| Anticoagulant Use | Antiplatelet Use | Anticoagulant Use | Antiplatelet Use | |||
| Overall | Overall | (n = 32,901) | 9% | 10% | 4% | 7% |
| Region | Australia-New Zealand | (n = 1,274) | 9% | 14% | 4% | 7% |
| Other Asia-Pacific | (n = 10,545) | 6% | 9% | 2% | 6% | |
| Europe | (n = 14,969) | 11% | 9% | 6% | 8% | |
| GCC | (n = 717) | 6% | 14% | -- | -- | |
| North America | (n = 5,396) | 13% | 12% | 7% | 8% | |
| Age, y | 18-49 | (n = 5,569) | 4% | 4% | 1% | 2% |
| 50-59 | (n = 5,784) | 7% | 9% | 3% | 6% | |
| 60-69 | (n = 8,469) | 9% | 11% | 4% | 8% | |
| 70-79 | (n = 8,407) | 12% | 11% | 6% | 11% | |
| 80+ | (n = 4,630) | 13% | 10% | 10% | 11% | |
| Sex | Male | (n = 20,188) | 10% | 10% | 5% | 7% |
| Female | (n = 12,698) | 9% | 9% | 2% | 6% | |
| Any history of cardiovascular disease | Yes | (n = 19,253) | 13% | 13% | 9% | 14% |
| No | (n = 13,437) | 3% | 3% | 1% | 2% | |
| History of atrial fibrillation | Yes | (n = 4,077) | 37% | 11% | 41% | 9% |
| No | (n = 28,183) | 5% | 9% | 2% | 6% | |
| History of cerebrovascular disease | Yes | (n = 4,863) | 13% | 18% | 11% | 19% |
| No | (n = 27,778) | 9% | 8% | 3% | 6% | |
| History of other atherothrombotic disease | Yes | (n = 13,553) | 13% | 16% | 10% | 19% |
| No | (n = 19,124) | 7% | 4% | 2% | 3% | |
| History of other cardiovascular disease | Yes | (n = 10,639) | 16% | 12% | 11% | 12% |
| No | (n = 22,015) | 6% | 8% | 2% | 5% | |
| History of bleeding | Yes | (n = 1,648) | 12% | 10% | 8% | 11% |
| No | (n = 30,932) | 9% | 10% | 4% | 7% | |
| History of diabetes | Yes | (n = 14,067) | 10% | 14% | 4% | 10% |
| No | (n = 18,537) | 9% | 7% | 4% | 4% | |
Abbreviations: DOPPS, Dialysis Outcomes and Practice Patterns Study; GCC, Gulf Cooperation Council; HD, hemodialysis; PD, peritoneal dialysis; PDOPPS, Peritoneal Dialysis Outcomes and Practice Patterns Study.
The adjusted odds ratios (ORs) of factors associated with baseline OAC and APT use are shown in Table 3. Prescription of OAC was most strongly and positively associated with prior history of AF (OR [95% CI]: 8.78 [7.98-9.65]), whereas prescription of APTs was most strongly and positively associated with the presence of other atherothrombotic diseases (ie, coronary or peripheral artery disease) (OR, 3.74 [3.40-4.11]). Patients with diabetes were more likely to be prescribed APT (OR, 1.58 [1.45-1.72]) and less likely to be prescribed OAC (OR, 0.85 [0.77-0.93]) compared with patients without diabetes. Antithrombotic drug prescriptions varied across regions. Patients in North America were more likely to be prescribed OAC compared with other regions, whereas patients from GCC had higher odds of being prescribed APT than in other regions. These analyses were repeated for each dialysis modality, with results presented in Table S3. Similar to the overall results, patients with AF and other atherothrombotic diseases had the highest chances of being prescribed OAC and APT, respectively. These associations tended to have higher point estimates among patients receiving PD compared with those receiving HD (Table S3).
Table 3.
Adjusted Odds Ratio (95% CI) of Oral Anticoagulant Use and Antiplatelet Agent Use at Baseline.
| Parameter | Anticoagulant Use |
Antiplatelet Use |
|---|---|---|
| Odds Ratio (95% CI) | Odds Ratio (95% CI) | |
| Region | ||
| Australia-New Zealand | 0.55 (0.43-0.70) | 1.10 (0.90-1.34) |
| Europe | 0.71 (0.64-0.80) | 0.83 (0.74-0.93) |
| GCC | 0.59 (0.41-0.83) | 1.45 (1.14-1.85) |
| Other Asia-Pacific | 0.56 (0.49-0.65) | 0.98 (0.86-1.12) |
| North America | REF | REF |
| Age, per 5 years | 1.05 (1.03-1.07) | 1.05 (1.03-1.06) |
| Male | 1.17 (1.07-1.28) | 1.06 (0.97-1.15) |
| Comorbid conditions | ||
| History of atrial fibrillation | 8.78 (7.98-9.65) | 0.75 (0.67-0.85) |
| History of cerebrovascular disease | 1.15 (1.03-1.28) | 1.98 (1.81-2.16) |
| History of other atherothrombotic disease | 1.10 (1.00-1.22) | 3.74 (3.40-4.11) |
| History of other cardiovascular disease | 1.60 (1.46-1.76) | 1.05 (0.96-1.14) |
| History of bleeding | 0.88 (0.74-1.05) | 0.81 (0.68-0.96) |
| History of diabetes | 0.85 (0.77-0.93) | 1.58 (1.45-1.72) |
Models include both hemodialysis (HD) and peritoneal dialysis (PD) patients (n = 32,901). Models adjusted for vintage (time since dialysis initiation), vascular access type [HD only; categorized as no vascular access for PD], body mass index, and serum albumin.
Abbreviations: CI, confidence interval; GCC, Gulf Cooperation Council.
Figure 1A shows the Kaplan–Meier plot for patients who remained on OAC and APT over time among patients prescribed these medications at baseline and with medication history after baseline. After 36 months, the estimated proportions of baseline users who remained on therapy were 57% for OAC and 53% for APT. Figure 1B shows the cumulative incidence plot for OAC and APT initiation among baseline nonusers who had medication history after baseline. The cumulative incidence of medication initiation increased at a slightly faster pace for APT versus OAC. After 36 months, the estimated proportion of baseline nonusers who initiated therapy was 6% for OAC and 10% for APT. The rates of discontinuation of OAC were consistently higher in PD compared with HD (Table S4). Within 12 months, 26% of patients receiving PD and 20% of patients receiving HD discontinued OACs. In contrast, discontinuation rates for APTs were higher in HD, with 24% of patients receiving HD and 20% of patients receiving PD discontinuing APTs during the same period, as shown in Table S4.
Figure 1.
Changes in the prescription of oral anticoagulants and antiplatelet agents in end-stage kidney disease patients. (A) Kaplan–Meier plot of discontinuation among patients prescribed the medication at baseline. (B) Cumulative incidence plot of initiation among patients not prescribed the medication at baseline. Patients lacking medication information beyond baseline were excluded from this analysis.
The composite bleeding rate for participants receiving HD or PD overall and stratified by baseline OAC and APT use are illustrated in Figure 2. The composite bleeding rate per 100 patient-years (95% CI) was 8.6 (7.7-9.7) among patients receiving HD prescribed an OAC at baseline and 4.2 (4.0, 4.5) among patients receiving HD not prescribed an OAC at baseline; in the PD cohort, the bleeding rates were 12.0 (8.3-17.5) versus 4.0 (3.5-4.5) among patients prescribed vs. not prescribed an OAC at baseline. The composite bleeding rate per 100 patient-years (95% CI) was 5.6 (4.9-6.4) among patients receiving HD prescribed an APT at baseline and 4.5 (4.3-4.8) among patients receiving HD not prescribed an APT at baseline. In the PD cohort, the bleeding rates were 6.1 (4.1-8.9) versus 4.1 (3.7-4.7) among patients prescribed versus not prescribed an APT at baseline. The composite bleeding rates per 100 patient-years among patients with neither OAC use nor APT use at baseline was 4.1 (3.9-4.4) in the HD cohort and 3.9 (3.4-4.4) in the PD cohort (not shown in Fig 2). Rates of death because of hemorrhagic stroke, hospitalization because of vascular access bleeding, and GI bleeding requiring hospitalization were higher among those prescribed antithrombotic therapies at baseline and even more so for those on OAC than APT (Fig 3).
Figure 2.
Incidence rates (95% CI) of composite bleeding outcome for patients receiving hemodialysis (HD) or peritoneal dialysis (PD); overall and stratified by prescription of oral anticoagulants and antiplatelets at baseline. Abbreviations: CI, confidence interval; DOPPS, Dialysis Outcomes and Practice Patterns Study; PDOPPS, Peritoneal Dialysis Outcomes and Practice Patterns Study.
Figure 3.
Incidence rates (95% CI) of specific bleeding events for patients receiving hemodialysis (HD) or peritoneal dialysis (PD); overall and stratified by prescription of oral anticoagulants and antiplatelets at baseline. (A) Hospitalization for gastrointestinal (GI) bleeding; (B) Hospitalization for vascular access bleeding; (C) Death because of hemorrhagic stroke. Vascular access hospitalizations were suppressed for the PD population because they are not relevant to the therapy. Abbreviations: CI, confidence interval; DOPPS: Dialysis Outcomes and Practice Patterns Study; PDOPPS: Peritoneal Dialysis Outcomes and Practice Patterns Study.
Discussion
In this international cohort study, we found that patients receiving HD or PD had a low (17% at baseline) prevalence of antithrombotic drug prescriptions given their cardiovascular risks. We also observed low rates of initiation and high rates of discontinuation for antithrombotic agents over time. Compared with patients not prescribed antithrombotic agents, those using APTs or OACs had a higher rate of major bleeding events.
Our study showed that the prescription of antithrombotic therapies in dialysis patients varies by region, dialysis modality, and clinical characteristics, including age and history of CVD and diabetes. The use of antithrombotic agents was higher in HD than in PD, with a larger difference observed for anticoagulant use compared with antiplatelets. Further, the estimated rate of anticoagulation discontinuation was higher in PD compared with HD. Observations might be affected by residual confounding of underlying clinical differences between patients receiving HD and PD, likely driven by modality selection bias. Nephrologists may avoid anticoagulants in patients receiving PD more frequently, in part, because of distinct risk-benefit assessments, with potential variation by region. The threshold for discontinuation of OAC can be lower in PD because the prevalence of risk factors for thromboembolic disease (eg, catheter for vascular access) is generally lower compared with HD.
Similar to a prior DOPPS study, regional variation was clear, and the prevalence of OAC or APT prescriptions was generally higher in North America than in other regions in the study.6 Regional variation in practice patterns reflects in part the distinct risk profiles for cardiovascular events. For instance, in a prior DOPPS study, the rates of stroke varied more than 2-fold between participating DOPPS countries.6 Distinct distribution of risk factors implies that the absolute effect of antithrombotic therapy will vary by region. The higher prevalence of antithrombotic therapy in North America is not surprising considering the generally higher prevalence of cardiometabolic risk factors compared with other regions.
One of the key predictors of antithrombotic prescription in our study was a history of AF, which was strongly and positively associated with the prescription of anticoagulants and negatively associated with the prescription of antiplatelet agents. However, even for the subset of dialysis patients with a strong guideline-directed indication for anticoagulation (ie, AF), still only about 40% of patients with AF in our study were prescribed an OAC—compared with about 60% of those with AF outside the dialysis setting—indicating an unmet need for a safer antithrombotic therapy that does not increase bleeding risk.25
Among patients who were prescribed antithrombotic agents at baseline, the rate of major bleeding events was higher than those not on antithrombotic agents, particularly in the subset of patients receiving PD, in which underlying bleeding risk may be high due to susceptibility in the peritoneal membrane and at the PD catheter site. Among HD patients using anticoagulants, we described rates of major bleeding, the composite of bleeding death or hospitalization, of approximately 9 per 100 patient-years. These event rates are between those reported in the sub-analyses of the ARISTOTLE trial, including patients with kidney disease not receiving dialysis assigned to either the apixaban (3.6% per year) or warfarin arm (5.5% per year), and the RENAL-AF trial, including patients receiving HD, assigned to either the apixaban (11% per year) or warfarin arm (10% per year).26,27 The incidence of bleeding events among dialysis patients randomized to antiplatelet agents is challenging to estimate from the literature. Published trials to date evaluated a broad range of antiplatelet regimens to prevent vascular access thrombosis in dialysis patients; thus, the available data are not comparable to our study.28
To the best of our knowledge, ours is the first study to describe the rate of discontinuation of antithrombotic therapies among dialysis patients, including those receiving PD or HD. Our results suggest that about 20%-26% of patients receiving dialysis discontinued antithrombotic drugs within 12 months after study enrollment and nearly half after 36 months. Since cardiovascular risk does not decrease with dialysis vintage, the decision to discontinue antithrombotic agents can leave patients at a considerable risk of adverse cardiovascular events.29 Such high discontinuation rates likely reflect several factors in the dialysis setting, including high bleeding risk, challenges in managing anemia and periprocedural discontinuations, among others. The observed low rates of initiation and high rates of discontinuation for antithrombotic agents in this study highlight the need for safer antithrombotic therapies to reduce cardiothrombotic risk among dialysis patients. From a risk perspective, prospective studies exploring risk prediction models to further identify patients receiving dialysis at higher risk for bleeding events and antithrombotic therapy discontinuation are warranted. The outcome of antithrombotic therapy discontinuation is largely unknown among patients receiving dialysis and can be studied in future observational studies. From a benefit standpoint, clinical trials assessing the efficacy and safety of antithrombotic therapies, across dialysis modalities and regions, are key next steps to define the trade-offs in prescribing antithrombotic therapies for patients receiving dialysis.
Our study has noteworthy limitations. First, our primary objective in this study was descriptive. As such, our results should not be used to infer causality. Our bleeding rate analyses were not adjusted and therefore can be affected by different patient population characteristics across regions and dialysis modalities. Second, although DOPPS and PDOPPS used standardized data collection processes that harmonized data across countries, we cannot rule out under-reporting or heterogeneity in the classification of bleeding events. Third, we captured only prescribed medications. The total prevalence of antithrombotic use can be higher than reported because of over-the-counter use (eg, aspirin). The exclusion of aspirin could have resulted in misclassification bias when comparing the bleeding rates across groups exposed to antithrombotic agents. Fourth, our rates of discontinuation of antithrombotic agents may have been overestimated because of potential inclusion of transitory discontinuation of such drugs during acute bleeding events or hospitalizations. Patients may have been restarted on antithrombotic therapy if the risk of bleeding was reduced with a medical intervention. Finally, because the relevant medication prescriptions were captured intermittently (every 4-6 months) and not continuously, we were not able to use more complex statistical methods to investigate adjusted associations between anticoagulation use and bleeding events. Our study also has important strengths. We reported on international global studies that were designed to reflect real-world practices. Most of the data were captured using primary data collection under the supervision of site investigators, who locally classified the outcome data, including bleeding events. Lastly, our international representativeness provided strong external validity for the incidence rates of bleeding estimated in this study.
In this international and representative study of dialysis patients, we found that approximately 19% and 11% of patients receiving HD and PD were prescribed antithrombotic agents, respectively. The rates of major bleeding among patients on anticoagulants and antiplatelets were 8.6 (HD) and 5.6 (HD) and 12.0 (PD) and 6.1 (PD) per 100 patient-years, respectively. After 36 months, the Kaplan–Meier estimated proportions of baseline users who remained on therapy were 57% for OAC and 53% for APT. New strategies and treatment options to improve the use of antithrombotic agents and to optimize the balance between the risks and benefits of the therapy among patients receiving dialysis are needed.
Article Information
Authors’ Full Names and Academic Degrees
Murilo Guedes, MD, PhD, Calvin Andrews, MS, Junhui Zhao, PhD, G. Brandon Atkins, MD, PhD, Irina Barash, MD, MS, Lori D. Bash, PhD, Catelyn R. Coyle, PhD, Xuehua Ke, PhD, Dena R. Ramey, PhD, David W. Johnson, MBBS, PhD, Pablo Ureña-Torres, MD, PhD, Pietro Manuel Ferraro, MD, PhD, Mohammed Al Ghonaim, MBBS, Marc P. Bonaca, MD, MPH, Roberto Pecoits-Filho, MD, PhD, and Angelo Karaboyas, PhD
Authors’ Contributions
Research idea and study design: MG, RPF, AK, CC, LDB, GBA, DRR, IB, MPB; data acquisition: RPF, AK; data analysis/interpretation: MG, CA, JZ, RPF, AK, CC, XK, LDB, GBA, DRR, IB, MPB; statistical analysis: CA, JZ; supervision or mentorship: DWJ, PUT, PMR, MAG, RPF, AK, CC, XK, LDB. Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved.
Support
This study received support fro m Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA through a grant to Arbor Research Collaborative. Coauthors from Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA, contributed to the study design, interpretation of data, writing, and decision to submit the report for publication.
Financial Disclosure
Angelo Karaboyas and Calvin Andrews are employees of Arbor Research Collaborative for Health, which administers the DOPPS. Murilo Guedes and Junhui Zhao were employees of Arbor Research Collaborative for Health at the time of the study. Global support for the ongoing DOPPS Programs is provided without restriction on publications by a variety of funders. For details, see https://www.dopps.org/AboutUs/Support.aspx. All support for the DOPPS program is made to Arbor Research Collaborative for Health and not directly to the authors. Roberto Pecoits-Filho has received honoraria (paid to employer) from Astra Zeneca, Boehringer-Lilly, Akebia, Bayer, GSK, and Novo Nordisk for participation in advisory boards and educational activities. Consulting fees Scientific Leadership in clinical trials from the George Clinical. Research grants from Fresenius Medical Care, National Council for Scientific and Technological Development. RPF is employed by Arbor Research Collaborative for health, who runs the DOPPS studies. Global support for the ongoing DOPPS Programs is provided without restriction on publications by a variety of funders. Funding is provided to Arbor Research Collaborative for Health and not to Dr. Pecoits-Filho directly. For details see https://www.dopps.org/AboutUs/Support.aspx. David Johnson has received consultancy fees, research grants, speaker’s honoraria, and travel sponsorships from Baxter Healthcare and Fresenius Medical Care; consultancy fees from Astra Zeneca, Bayer, and AWAK; speaker’s honoraria from ONO and Boehringer Ingelheim & Lilly; and travel sponsorships from ONO and Amgen. He is a current recipient of an Australian National Health and Medical Research Council Leadership Investigator Grant. Pablo Antonio Urena Torres declares receipt of advisory and/or lecture fees from Astellas, Astra Zeneca, GSK, Hemotech, Sanofi, Théradial, and VIFOR-Fresenius-Renal Pharma. Irina Barash, Lori D. Bash, Catelyn R. Coyle, and Xuehua Ke are all full-time employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA. G. Brandon Atkins and Dena Rosen Ramey were full-time employees of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA when the study was conducted. Marc Bonaca is the Executive Director of CPC, a nonprofit academic research organization affiliated with the University of Colorado, that receives or has received research grant/consulting funding between February 2021 and present from the following: Abbott Laboratories; Adamis Pharmaceuticals Corporation; Agios Pharmaceuticals, Inc.; Alexion Pharma; Alnylam Pharmaceuticals, Inc.; Amgen, Inc.; Angionetics, Inc.; ARCA Biopharma, Inc.; Array BioPharma, Inc.; AstraZeneca and Affiliates; Atentiv LLC; Audentes Therapeutics, Inc.; Bayer and Affiliates; Beth Israel Deaconess Medical Center; Better Therapeutics, Inc.; BIDMC; Boston Clinical Research Institute; Bristol-Meyers Squibb Company; Cambrian Biopharma, Inc.; Cardiol Therapeutics, Inc.; CellResearch Corp.; Cook Medical Incorporated; Covance; CSL Behring LLC; Eidos Therapeutics, Inc.; EP Trading Co. Ltd.; EPG Communication Holdings Ltd.; Epizon Pharma, Inc.; Esperion Therapeutics, Inc.; Everly Well, Inc.; Exicon Consulting Pvt. Ltd.; Faraday Pharmaceuticals, Inc.; Foresee Pharmaceuticals Co. Ltd.; Fortress Biotech, Inc.; HDL Therapeutics, Inc.; HeartFlow, Inc.; Hummingbird Bioscience; Insmed, Inc.; Ionis Pharmaceuticals; IQVIA, Inc.; JanOne Biotech Holdings, Inc.; Janssen and Affiliates; Kaneka; Kowa Research Institute, Inc.; Kyushu University; Lexicon Pharmaceuticals, Inc.; LSG Kyushu University; Medimmune Ltd.; Medpace; Merck & Co., Inc., Rahway, NJ, USA; Novartis Pharmaceuticals Corp.; Novate Medical, Ltd.; Novo Nordisk, Inc.; Pan Industry Group; Pfizer, Inc.; PhaseBio Pharmaceuticals, Inc.; PPD Development; LP; Prairie Education and Research Cooperative; Prothena Biosciences Limited; Regeneron Pharmaceuticals, Inc.; Regio Biosciences, Inc.; Rexgenero; Sanifit Therapeutics S.A.; Sanofi-Aventis Groupe; Silence Therapeutics PLC; Smith & Nephew PLC; Stealth BioTherapeutics, Inc.; State of Colorado CCPD Grant; The Brigham & Women's Hospital, Inc.; The Feinstein Institutes for Medical Research; Thrombosis Research Institute; University of Colorado; University of Pittsburgh; VarmX; Virta Health Corporation; WCT Atlas; Worldwide Clinical Trials, Inc.; WraSer; LLC; and Yale Cardiovascular Research Group. Marc Bonaca receives support from the AHA SFRN under award numbers 18SFRN3390085 (BWH-DH SFRN Center) and 18SFRN33960262 (BWH-DH Clinical Project). Pietro Manuel Ferraro received consultant fees and grant/other support from Allena Pharmaceuticals, Alnylam, Amgen, AstraZeneca, Bayer, Gilead, Novo Nordisk, and Otsuka Pharmaceuticals as well as royalties as an author for UpToDate. Mohammed Al Ghonaim has nothing to disclose.
Data Sharing
Data that support the findings of this study are available from Arbor Research Collaborative for Health upon reasonable request to the corresponding author; however, restrictions apply to availability of the DOPPS data, which are not publicly available.
Peer Review
Received December 17, 2024. Evaluated by 1 external peer reviewer, with direct editorial input from an Associate Editor and the Editor-in-Chief. Accepted in revised form July 17, 2025.
Footnotes
Complete author and article information provided before references.
Figure S1: Flow chart diagram of patient inclusion criteria.
Table S1: Characteristics of Patients Receiving Hemodialysis (HD), Overall and by Oral Anticoagulant (OAC) and Antiplatelet Use.
Table S2: Characteristics of Patients Receiving Peritoneal Dialysis (PD), Overall and by Oral Anticoagulant (OAC) and Antiplatelet Agent Use.
Table S3: Adjusted Odds Ratio (95% CI) of Oral Anticoagulant (OAC) and Antiplatelet Use, by Hemodialysis (HD)/Peritoneal Dialysis (PD).
Table S4: Estimated Proportion of Baseline Users that Discontinued Medication and Estimated Proportion of Baseline Nonusers Who Initiated Medication Based on Kaplan–Meier Analyses.
Supplementary Materials
Figure S1; Tables S1-S4.
References
- 1.Tveit D.P., Hypolite I.O., Hshieh P., et al. Chronic dialysis patients have high risk for pulmonary embolism. Am J Kidney Dis. 2002;39:1011–1117. doi: 10.1053/ajkd.2002.32774. [DOI] [PubMed] [Google Scholar]
- 2.Wattanakit K., Cushman M., Stehman-Breen C., et al. Chronic kidney disease increases risk for venous thromboembolism. J Am Soc Nephrol. 2008;19:135–1340. doi: 10.1681/ASN.2007030308. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mahmoodi B.K., Gansevoort R.T., Veeger N.J., et al. Prevention of Renal and Vascular End-stage Disease (PREVEND) Study Group. Microalbuminuria and risk of venous thromboembolism. JAMA. 2009;301:1790–1797. doi: 10.1001/jama.2009.565. [DOI] [PubMed] [Google Scholar]
- 4.Evans E.P., Branch R.A., Bloom A.L. A clinical and experimental study of platelet function in chronic renal failure. J Clin Pathol. 1972;25(9):745–753. doi: 10.1136/jcp.25.9.745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Holden R.M., Harman G.J., Wang M., et al. Major bleeding in hemodialysis patients. Clin J Am Soc Nephrol. 2008;3(1):105–110. doi: 10.2215/CJN.01810407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sood M.M., Larkina M., Thumma J.R., et al. Major bleeding events and risk stratification of antithrombotic agents in hemodialysis: results from the DOPPS. Kidney Int. 2013;84(3):600–608. doi: 10.1038/ki.2013.170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.van Eck van der Sluijs A., Abrahams A.C., Rookmaaker M.B., et al. Bleeding risk of haemodialysis and peritoneal dialysis patients. Nephrol Dial Transplant. 2021;36(1):170–175. doi: 10.1093/ndt/gfaa216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Boccardo P., Remuzzi G., Galbusera M. Platelet dysfunction in renal failure. Semin Thromb Hemost. 2004;30(5):579–589. doi: 10.1055/s-2004-835678. [DOI] [PubMed] [Google Scholar]
- 9.Padberg F.T., Jr., Calligaro K.D., Sidawy A.N. Complications of arteriovenous hemodialysis access: recognition and management. J Vasc Surg. 2008;48(5 Suppl):55S–80S. doi: 10.1016/j.jvs.2008.08.067. [DOI] [PubMed] [Google Scholar]
- 10.Konstantinidis I., Nadkarni G.N., Yacoub R., et al. Representation of patients with kidney disease in trials of cardiovascular interventions: an updated systematic review. JAMA Intern Med. 2016;176:121–124. doi: 10.1001/jamainternmed.2015.6102. [DOI] [PubMed] [Google Scholar]
- 11.Sarnak M.J., Amann K., Bangalore S., et al. Chronic kidney disease and coronary artery disease: JACC state-of-the-art review. J Am Coll Cardiol. 2019;74(14):1823–1838. doi: 10.1016/j.jacc.2019.08.1017. [DOI] [PubMed] [Google Scholar]
- 12.Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int Suppl. 2013;3:1–150. [Google Scholar]
- 13.Wanner C., Herzog C.A., Turakhia M.P. Conference Steering Committee. Chronic kidney disease and arrhythmias: highlights from a Kidney Disease: Improving Global Outcomes (KDIGO) Controversies Conference. Kidney Int. 2018;94(2):231–234. doi: 10.1016/j.kint.2018.05.005. [DOI] [PubMed] [Google Scholar]
- 14.Virani S.S., Newby L.K., Arnold S.V., et al. 2023 AHA/ACC/ACCP/ASPC/NLA/PCNA guideline for the management of patients with chronic coronary disease: a report of the American Heart Association/American College of Cardiology Joint Committee on Clinical Practice Guidelines. J Am Coll Cardiol. 2023;82:833–955. doi: 10.1016/j.jacc.2023.04.003. [DOI] [PubMed] [Google Scholar]
- 15.Joglar J.A., Chung M.K., Armbruster A.L., et al. 2023 ACC/AHA/ACCP/HRS Guideline for the Diagnosis and Management of Atrial Fibrillation: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation. 2024;149(1):e1–e156. doi: 10.1161/CIR.0000000000001193. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kao T.W., Chen Z.W., Lin Y.H. Anticoagulation for patients with concomitant atrial fibrillation and end-stage renal disease: a systematic review and network meta-analysis. J Am Heart Assoc. 2024;13(8) doi: 10.1161/JAHA.123.034176. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kuno T., Takagi H., Ando T., et al. Oral anticoagulation for patients with atrial fibrillation on long-term hemodialysis. J Am Coll Cardiol. 2020;75(3):273–285. doi: 10.1016/j.jacc.2019.10.059. [DOI] [PubMed] [Google Scholar]
- 18.Sarnak M.J., Amann K., Bangalore S., et al. Chronic kidney disease and coronary artery disease: JACC state-of-the-art review. J Am Coll Cardiol. 2019;74:1823–1838. doi: 10.1016/j.jacc.2019.08.1017. [DOI] [PubMed] [Google Scholar]
- 19.Bangalore S., Maron D.J., Fleg J.L., et al. International study of comparative health effectiveness with medical and invasive approaches-chronic kidney disease (ISCHEMIA-CKD): rationale and design. Am Heart J. 2018;205:42–52. doi: 10.1016/j.ahj.2018.07.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Collins A.J., Foley R.N., Gilbertson D.T., et al. United States Renal Data System public health surveillance of chronic kidney disease and end-stage renal disease. Kidney Int Suppl (2011) 2015;5:2–7. doi: 10.1038/kisup.2015.2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Palmer S.C., Craig J.C., Navaneethan S.D., et al. Benefits and harms of statin therapy for persons with chronic kidney disease: a systematic review and meta-analysis. Ann Intern Med. 2012;157:263–275. doi: 10.7326/0003-4819-157-4-201208210-00007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Palmer S.C., Di Micco L., Razavian M., et al. Effects of antiplatelet therapy on mortality and cardiovascular and bleeding outcomes in persons with chronic kidney disease: a systematic review and meta-analysis. Ann Intern Med. 2012;156:445–459. doi: 10.7326/0003-4819-156-6-201203200-00007. [DOI] [PubMed] [Google Scholar]
- 23.Perl J., Davies S.J., Lambie M., et al. The Peritoneal Dialysis Outcomes and Practice Patterns Study (PDOPPS): Unifying efforts to inform practice and improve global outcomes in peritoneal dialysis. Perit Dial Int. 2016;36(3):297–307. doi: 10.3747/pdi.2014.00288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Pisoni R.L., Gillespie B.W., Dickinson D.M., et al. The Dialysis Outcomes and Practice Patterns Study (DOPPS): design, data elements, and methodology. Am J Kidney Dis. 2004;44(5 Suppl 2):7–15. doi: 10.1053/j.ajkd.2004.08.005. [DOI] [PubMed] [Google Scholar]
- 25.Navar A.M., Kolkailah A.A., Overton R., et al. Trends in oral anticoagulant use among 436 864 patients with atrial fibrillation in community practice, 2011 to 2020. J Am Heart Assoc. 2022;11(22) doi: 10.1161/JAHA.122.026723. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hohnloser S.H., Hijazi Z., Thomas L., et al. Efficacy of apixaban when compared with warfarin in relation to renal function in patients with atrial fibrillation: insights from the ARISTOTLE trial. Eur Heart J. 2012;33(22):2821–2830. doi: 10.1093/eurheartj/ehs274. [DOI] [PubMed] [Google Scholar]
- 27.Pokorney S.D., Chertow G.M., Al-Khalidi H.R., et al. Apixaban for patients with atrial fibrillation on hemodialysis: a multicenter randomized controlled trial. Circulation. 2022;146(23):1735–1745. doi: 10.1161/CIRCULATIONAHA.121.054990. [DOI] [PubMed] [Google Scholar]
- 28.Wang Q., Xie X., Xu G. The risk of bleeding for antiplatelet agents in haemodialysis patients: a meta-analysis. BMC Nephrol. 2020;21(1):106. doi: 10.1186/s12882-020-01757-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sumida K., Yamagata K., Iseki K., et al. Different impact of hemodialysis vintage on cause-specific mortality in long-term hemodialysis patients. Nephrol Dial Transplant. 2016;31(2):298–305. doi: 10.1093/ndt/gfv402. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1; Tables S1-S4.