Recurrent thrombosis and major bleeding in children treated for VTE Open Access

Venous thromboembolism (VTE) in children is estimated to be a relatively rare event, with an overall incidence of 0.14 to 0.20 per 10 000 person-years.^1-3 However, the rate of pediatric VTE in hospitalized children continues to increase, and recent data from the Pediatric Health Information System database showed an increase from 46 VTE cases per 10 000 admissions in 2008 to 106 VTE cases per 10 000 admissions in 2019; this is a 130% increase in VTE events.⁴ This rise is speculated to be due to improved advanced medical care in critically ill and medically complex children and improved diagnostic imaging modalities.^4,5 

Despite the increased rate of VTE in hospitalized children, there are limited data on thrombosis recurrence and bleeding outcomes in anticoagulated children with VTE. With the rise in VTE diagnoses in hospitalized children and the availability of new direct oral anticoagulants, which offer ease of administration and less monitoring, a better understanding of which children are at higher risk for recurrent thrombosis and/or major bleeding will guide providers on how to tailor the management of VTE.

The primary objective of this study was to determine the incidence rate of major bleeding on therapeutic anticoagulation and recurrent thrombosis in a cohort of children with VTE. Secondary objectives included describing the clinical characteristics of first-episode thrombosis, risk factors for recurrent thrombosis, and reporting medical conditions associated with major bleeding in this cohort.

We conducted a cohort study of consecutive pediatric patients (aged 0-21 years) with VTE who were treated with therapeutic anticoagulation at our center from 1 January 2015 to 31 December 2022. Cohort members were identified using our center’s clinical anticoagulation database that tracks all noncardiac anticoagulation patients managed by our outpatient thrombosis clinic. All patients on therapeutic anticoagulation seen in our outpatient clinic were included; patients who were only managed on anticoagulation during a hospital admission but did not have follow-up in our outpatient clinic were not captured. Patients with VTE who were not treated with anticoagulation were excluded.

We began using this database on 1 January 2015, and every individual on therapeutic anticoagulation managed in our clinic at that time was registered in our database. Thrombosis nurses regularly update the registry based on data from providers, patients/families, or chart review. Patient charts were reviewed, and patients were included in the cohort if they met study criteria. For each individual, entry into the cohort was defined as the date of initial VTE diagnosis. Individual follow-up times were calculated as the time from cohort entry (initial VTE) to the last hematology clinical encounter, up to 31 December 2022. A nested case-control study (within this cohort) was also conducted to identify risk factors for recurrent thrombosis.

Our clinical database of anticoagulated noncardiac patients includes the following data: demographics, date of thrombosis, site of thrombosis, start and stop dates of anticoagulation therapy, specific anticoagulants used, date and site of any recurrent VTE, date and type of episodes of major bleeding, and risk factors for VTE. Potential risk factors for VTE included the following: the presence of a central venous catheter (CVC), surgery within 6 weeks of VTE event, infection, cancer, congenital thrombophilia (factor V G1691A, factor II G20210A, protein C deficiency, protein S deficiency, and antithrombin deficiency), estrogen-containing contraception (females only), anatomic venous anomaly (May-Thurner syndrome, thoracic outlet syndrome, and inferior vena cava atresia or variants), inflammatory bowel disease (IBD), intestinal failure ([IF; congenital or surgical short bowel syndrome or non-IBD intestinal inflammatory disorders requiring long-term total parenteral nutrition), antiphospholipid syndrome, systemic lupus erythematosus, nephrotic syndrome, and cystic fibrosis. Cohort data were collected via chart review either prospectively or retrospectively. For patients with recurrent VTE, charts were reviewed to assess whether VTE was symptomatic or asymptomatic at index VTE and at the time of recurrence. Symptomatic was defined as having either signs or symptoms of thrombosis that resulted in imaging. Research-specific data were stored in Research Electronic Data Capture software, hosted at our institution. This study was granted a waiver of informed consent by our hospital’s Institutional Review Board.

Major bleeding was defined as fatal bleeding; clinically overt bleeding associated with a decrease in hemoglobin of at least 2 g/dL in a 24-hour period; critical site bleeding such as retroperitoneal, pulmonary, or intracranial; bleeding that requires surgical intervention in an operating or interventional radiology suite or endoscopy; and overt bleeding for which an anticoagulant reversal agent is administered.⁶ 

Patients were identified as having recurrent thrombosis if they had objective findings of new or progressive thrombosis by standard imaging methods, including computed tomography, magnetic resonance imaging, or duplex ultrasonography. Patients could have been symptomatic or asymptomatic from the recurrent VTE. Patients both on and off anticoagulation at the time of recurrence were included. Patients with congenital heart disease were excluded, because our hematology division does not manage anticoagulation in these patients.

Descriptive statistics were used to describe demographic and clinical data in patients with and without recurrent thrombosis, as well as patients with and without major bleeding. Means and ranges were calculated for continuous variables; frequencies and percentages were calculated for categorical variables. Initial assessments of differences between groups were assessed with t tests for continuous data and χ² tests for categorical data. Multivariable logistic regression was used to model the odds of repeat thrombosis based on likely predictors. The initial full model included univariate predictors with P value of ≤ .20. Backward elimination was then used to develop the final model, in which all predictors had P values of <.05. Significance was set a priori at an α level of 0.05. All analyses were performed with Statistical Analysis System (SAS) software, version 9.4 M5 (copyright 2024; SAS Institute, Inc, Cary, NC). The rate of major bleeding was too low to allow for multivariate analysis of bleeding risk factors. Patients were censored at the last hematology clinical encounter; we did not include visits after the end of study (31 December 2022). We lack sufficient data to analyze death as a competing risk, because our data lacked that granularity.

Between 1 January 2015 and 31 December 2022, we identified 632 eligible patients with 744 total episodes of venous thrombosis. Initial VTE occurred before 1 January 2015 in 48 of cohort members (8%), and initial VTE occurred on or after 1 January 2015 in 584 (92%). There were 815.3 patient-years of observation, and the median follow-up time was 357 days. Table 1 summarizes patient demographics and sites of first-time thrombosis. The median age at the time of first thrombosis was 13.9 years. Among these patients, 51.7% were female. There were 43 neonates (aged <2 months) and 36 infants (age <12 months). Predisposing risk factors for the first episode of thrombosis are listed in Table 2, with the most common risk factors for thrombosis being the presence of a CVC (40.7%), surgery within 6 weeks of thrombotic event (25.7%), and infection (18.2%). Among females (n = 327), estrogen-containing hormonal contraception was identified as a risk factor in 25.7% of our cohort. There was no identified risk factor in 56 patients (8.9%). The percentage of first-episode thrombosis that occurred outside of the hospital was 55.6%.

Of the 632 unique patients, 546 had a single episode of VTE. Eighty-six patients (13.6%) had ≥1 episode of recurrent thrombosis, with 112 recurrent thrombosis episodes, with a rate of 13.7 per 100 person-years. Of these 86 patients with recurrent VTE, 53 (8.4% of the cohort) had a repeat episode within 1 year of follow-up. Thirty-four of 112 of recurrent VTE episodes (30%) were asymptomatic. Of those 34 with asymptomatic recurrent VTE, 5 (15%) were at the same site of the index VTE, although deemed to be new due to the thrombus being previously resolved or progressive in extent. Descriptive analysis for patients with recurrent thrombosis is summarized in Table 3. Most episodes of recurrent thrombosis (61.6%) occurred during hospital admission; 30 of recurrent episodes (26.7%) occurred while on therapeutic anticoagulation. There were 257 patients who had a CVC-provoked thrombosis at index VTE. Of these, 38 had a CVC-provoked recurrent VTE (14.8%). The median time from initial VTE to first recurrence was 295 days (interquartile range, 51-707). In univariate analysis, significant risk factors for recurrent thrombosis included IBD (P = .01), the presence of a CVC (P = .02), and IF (P = .03; Table 4).

Because patients with IF and many with IBD may have a CVC, we performed multivariate logistic regression to model the odds of recurrent thrombosis while simultaneously controlling for the presence of CVC, diagnosis of IBD, and diagnosis of IF. We also included the variables anatomic venous abnormality and congenital thrombophilia in the model, which we considered to be clinically and physiologically important. The initial or full model showed that anatomic venous abnormality and IBD were significant risk factors for recurrent thrombosis (Table 5). Then, using an iterative method of backward elimination, we arrived at a final or reduced model in which only anatomic venous abnormality, CVC, and IBD were retained as significant predictors of recurrent thrombosis, with odds ratios of 2.3, 2.0, and 3.1, respectively (Table 5).

We also performed a sensitivity analysis, using the same modeling approach as described above but including only those patients whose first VTE came within the study period (on or after 1 January 2015). Of the 584 patients in this analysis, 65 (11.1%) had ≥1 repeat VTE, compared with 86 (13.6%) from the full cohort. Results are provided in Table 6 and are consistent with the full cohort, although some P values and confidence intervals were larger.

Sixteen patients (2.5%) had major bleeding while on therapeutic anticoagulation; there were 18 major bleeding events in 16 unique patients. The major bleeding rate was 2.21 events per 100 person-years. Descriptive analysis for major bleeding is shown in Table 7. The median total exposure to anticoagulation in patients with major bleeding was 56 days (minimum of 6 days; maximum 2336 days). The median age at the time of the major bleeding event was 12.3 years (range, 0.2-21.8). Up to 76.5% of major bleeding events occurred during hospitalization. Patients met at least 1 criterion for major bleeding, with many meeting >1 criterion. Eleven patients had bleeding with a drop in hemoglobin of at least 2 gm/dL, 6 patients had bleeding requiring going to an operative suite for bleeding control, and 5 patients had bleeding into a critical area. Eleven major bleeding events occurred on enoxaparin (61.1%), 5 on rivaroxaban (27.8%), and 2 on unfractionated hepari (11.1%). The most common site of major bleeding was the gastrointestinal tract in 10 events (55.6%). We identified no fatal bleeding events. In the 86 patients with recurrent thrombosis, 9 had major bleeding (10%); and of the 18 patients with major bleeding, 9 had recurrent thrombosis (56%). Twelve of the 18 major bleeding events (67%) occurred during treatment for a first-time VTE.

In univariate analysis, the medical conditions most associated with major bleeding risk while on anticoagulation were IF in 75% (P = .003) and infection in 46.7% (P = .01; Table 8).

This study represents, to our knowledge, the largest retrospective analysis assessing recurrent VTE in children with index VTE. These data demonstrate that children with VTE have a high rate of recurrence at 13.7 per 100 person-years, approaching the rate in adults whose incidence rate for recurrence is 17.7 per 100 person-years.⁷ Adults with unprovoked thrombosis and persistent risk factors are those that are at the highest risk for recurrence.^8,9 The primary risk factors for VTE recurrence in our cohort were the presence of a CVC, anatomic venous abnormalities, specifically inferior vena cava (IVC) atresia/variant, and an underlying diagnosis of IBD.

Data on pediatric thrombosis have grown significantly, with recent studies of new anticoagulants and studies of the duration of anticoagulation in provoked thrombosis in children. However, there remains a lack of data regarding which children are at the highest risk for recurrent thrombosis and how best and, most safely, to prevent recurrence. Estimating the risk of recurrent VTE in children is time dependent and often imprecise. Reports have estimated that the VTE recurrence rate in children varies from 8% to 18%,^10,11 and the risk of recurrent VTE in children after a provoked thrombosis appears to be low at ∼1%.¹² In a study looking at children with unprovoked thrombosis, considered a high-risk group, the incidence rate was 10.4 per 100 person-years, with risks related to age >12 years and high-risk thrombophilia.¹³ Recurrent thrombosis in the EINSTEIN Jr and DIVERSITY trials also had overall low VTE recurrence rates (1% in rivaroxaban arm and 3% in the standard arm; 4% in dabigatran and 8% in the standard arm).^14,15 However, these 2 randomized trials and the Kids-Duration of Therapy for Thrombosis trial had exclusions that may not fully capture the risk of VTE in the medically diverse population of children. Our data support the conclusion that recurrence rates are high for pediatric patients with VTE. Moreover, our data also highlight the importance of anatomic venous abnormalities and IBD as VTE recurrence risks in children.

CVCs have long been known to be a risk factor for VTE in children. Although there are few mitigating factors to prevent VTE associated with a critically required CVC, a recent single-institution study of CVC-associated thrombosis showed a lower odds ratio of recurrent CVC-associated thrombosis with therapeutic- vs prophylactically-dosed anticoagulation.¹⁶ More data are needed to understand how best to prevent CVC-associated VTE.

Children with venous compression or IVC atresia/variants have also been noted to be at risk for VTE, and this was one of the strongest risk factors for recurrent thrombosis in our cohort. May-Thurner syndrome can often present with unprovoked lower extremity thrombosis in otherwise healthy adolescents,¹⁷ and thoracic outlet obstruction can lead to exertional upper extremity thrombosis.¹⁸ IVC atresia or IVC variants result in altered and slowed venous return and predispose to unprovoked unilateral or bilateral lower extremity deep vein thrombosis and pulmonary embolism.^19-22 IVC atresia/variant was the most common venous abnormality in our cohort of recurrent VTE. Although first rib removal or iliac vein stenting can help with recurrent thrombosis in thoracic outlet obstruction and May-Thurner syndrome, respectively, IVC bypass or reconstructive procedures are not commonly done in children.

Despite the estimated incidence of VTE in IBD being relatively low, varying from 3.7 to 31.2 per 10 000 patient-years,²³ VTE is one of the most common adverse events in pediatric IBD.²⁴ VTE in children with IBD often occurs during an acute IBD flare, with cerebral sinovenous thrombosis being the most frequent site of thrombosis.²⁵ The use of steroids and the presence of CVC have also been found to be independent risk factors for VTE in IBD.²⁶ Our data further support the conclusion that IBD is a major risk factor for recurrent VTE in children.

The incidence rate of major bleeding was 2.5% or 2.21 events per 100 person-years in our population of 632 anticoagulated patients. Historically, major bleeding rates in children on anticoagulation have varied widely depending on the type of anticoagulant used (unfractionated heparin, low molecular, or warfarin) and the treatment setting (outpatient vs critical care unit).^27-29 In recent pediatric VTE multicenter randomized studies comparing rivaroxaban¹⁴ or dabigatran¹⁵ to the standard of care, there were no major bleeding events in the 329 patients who received rivaroxaban and 4 major bleeding events in the 176 patients (2%) who received dabigatran. These rates were similar to the standard-of-care groups: 1.2% in the EINSTEIN-Jr trial and 2% in the DIVERSITY trial; and comparable to the data presented here.^14,15 

Children with IF had the highest risk for major bleeding on anticoagulation in our cohort. IF can result in bleeding without associated anticoagulation due to anastomotic bleeding ulcers or acquired bleeding diatheses from causes such as vitamin K deficiency^30,31; the addition of anticoagulation in these settings would likely worsen any bleeding. This places children with IF in a predicament, because they depend on long-term parenteral nutrition given through a CVC, and CVC thrombosis is the most common VTE risk in these patients.^32,33 In addition, recurrent CVC–associated thrombosis can result in loss of vascular access and is one of the main indications for small bowel transplant.³⁴ In one study of 263 children with IF, a rate of 0.32 VTE per 1000 catheter days was seen; those with the highest risk were children with a higher total number of CVCs, a longer duration of CVC, and those with a history of central line–associated bloodstream infections.³³ Thus, children with IF and CVC are both at high risk for thrombosis and major bleeding.

Taken together, these data support the conclusion that concerns related to bleeding risk should generally not be a barrier to therapeutic anticoagulation in pediatric patients. These data also highlight the need for better anticoagulation strategies in pediatric patients with a high risk of recurrence, such as those with CVCs, anatomic venous abnormalities, or active IBD. Understandably, many pediatric practitioners have a difficult time committing children and adolescents to long-term or indefinite courses of anticoagulation. Adding to the challenges in clinical decision-making is the lack of clinical data regarding outcomes and quality of life in children on long-term anticoagulation. This, coupled with the desire of many of these patients to engage in higher-risk physical activities, makes the decision even more difficult. However, the data presented here suggest that consideration should be made for secondary anticoagulation in pediatric patients with ongoing risk factors, particularly those with CVCs, IBD, and uncorrected anatomic venous anomalies. It is unclear whether therapeutic anticoagulation is sufficient to prevent recurrent VTE; and similarly, data are lacking on the benefits of prophylactically-dosed anticoagulation.

Our study has limitations, including being a retrospective study. One significant limitation is the short median follow-up time of 357 days. Provider-defined unprovoked thrombosis was not captured in our clinical database, despite having data on patients with “no identified risk factor,” and thus, we are not able to report on unprovoked thrombosis in this cohort. Our study also had a selection bias, because we may not have captured hospitalized patients (including neonates and infants) with VTE if they were never seen in our clinic. Because cardiology manages anticoagulation in their patient population in our hospital, our cohort is missing a high-risk group of patients. Clinically relevant nonmajor bleeding was not reported, which can affect treatment choices and, potentially, quality of life. Our study was insufficiently powered to conclude the relative safety and efficacy of the different anticoagulants used in this cohort. For patients on anticoagulation at the time of recurrence, nearly half of children were outpatients. Thus, we cannot assess the impact of poor adherence to therapy as a potential contributor to recurrent VTE.

Many providers are hesitant to use extended or secondary anticoagulation in children due to fear of bleeding complications and the impact on quality of life. Yet, children remain at high risk for recurrent VTE, which could likewise be highly morbid or life threatening. Future randomized pediatric trials of anticoagulation are critically needed to investigate the safety and efficacy of extended and/or secondary pharmacologic thromboprophylaxis in those children at the highest risk for recurrent thrombosis. We also do not know the optimal anticoagulant and dosing that would best prevent recurrent VTE while minimizing bleeding in children. Given the relatively low incidence of pediatric VTE, a multi-institutional approach focused on this area and/or inclusion in cooperative groups are needed.

The authors received funding from Cincinnati Children’s Hospital Medical Center Division of Hematology.

Contribution: S.F.L., C.T., and A.B. conceived and designed the study; S.F.L., C.T., and M.F. interpreted findings; S.F.L., J.S.P., and C.T. wrote the manuscript; M.F. performed biostatistical analysis; S.F.L., A.B., and E.T. performed chart review and data collection; and all authors provided important feedback and reviewed the final manuscript.

Conflict-of-interest disclosure: C.T. reports advisory board fees from Takeda and Sanofi and sits on a study steering committee for Bayer, unrelated to this study. J.S.P. received research funding from Ionis Pharmaceuticals for work that was unrelated to this study. The remaining authors declare no competing financial interests.

Correspondence: Cristina Tarango, Division of Hematology, Cincinnati Children’s Hospital Medical Center, 3333 Burnet Ave, Cincinnati, OH 45229-3039; email: cristina.tarango@cchmc.org.