Risk factors and clinical impact of thrombosis during induction chemotherapy for pediatric acute lymphoblastic leukemia: A report from CYP‐C

Thromboembolism (TE) is associated with reduced survival in pediatric acute lymphoblastic leukemia (ALL). It has been hypothesized that TE might signal leukemic aggressiveness. The objective was to determine risk factors for TE during ALL induction (TEind) therapy and whether TEind is associated with treatment refractoriness. This retrospective cohort study using the population‐based Cancer in Young People Canada (CYP‐C) registry included children <15 years of age diagnosed with ALL (2000–2019) and treated at one of 12 Canadian pediatric centers outside of Ontario. Univariate and multivariable logistic regression models were used to determine risk factors for TEind and whether TEind predicted induction failure and ALL treatment intensification. The impact of TEind on overall and event‐free survival was estimated using Cox proportional hazard regression models. The study included 2589 children, of which 45 (1.7%) developed a TEind. Age (<1 year and ≥10 years vs. 1–<10 years), T‐cell phenotype, high‐risk ALL, and central nervous system involvement were all associated with TEind in univariate analysis. Age and T‐cell phenotype remained independent predictors of TEind in multivariable analysis. Induction failure occurred in 53 patients (2.1%). TEind was not associated with induction failure (OR: not estimable) or treatment intensification (adjusted OR [95% CI]: 0.66 [0.26–1.69]). TEind was independently associated with overall survival (adjusted HR [95% CI]: 2.54 [1.20–5.03]) but not event‐free survival (adjusted HR [95% CI] 1.86 [0.98–3.51]). In this population‐based study of children treated with contemporary chemotherapy protocols, TEind was associated with age and T‐cell phenotype and mortality but did not predict induction failure.

treated with contemporary chemotherapy protocols, TE ind was associated with age and T-cell phenotype and mortality but did not predict induction failure.

| INTRODUCTION
Thromboembolism (TE) is a well-recognized complication of childhood cancer and its treatment. 1Among children, patients with acute lymphoblastic leukemia (ALL) are at particularly high risk. 23][4][5][6][7] TE can lead to complications such as post-thrombotic syndrome, [8][9][10] neuro-developmental impairments, 11,12 chronic thromboembolic pulmonary hypertension, 13,14 and death. 13,14cent data have suggested that the development of a cancerassociated TE negatively impacts survival in children.Decreased overall survival (OS) following cancer-associated TE has been described in adults with multiple types of cancer [15][16][17] and with ALL. 18,19In children, a recent study by Forbrigger et al. found an adverse impact of cancer-associated TE on survival, but the participants with and without TEs had important differences in baseline characteristics including age and underlying cancer diagnosis that may confound differences in cancer outcome. 20Two retrospective studies and one small prospective study did not report a difference in OS 21,22 and EFS [21][22][23] following TE in children with ALL.Rank and colleagues reported a higher risk of death in a large cohort of children up to 18 years old treated as per NOPHO ALL2008 protocol who sustained a symptomatic TE. 24 In their cohort, the small number of relapses did not allow authors to explore differences in event-free survival (EFS) in individuals aged less than 18 years old.
Several hypotheses may explain the adverse effect of TE on leukemia outcomes.Aside from direct mortality following TE, it is possible that the occurrence of a TE delays the administration of anticancer therapy, 24 notably chemotherapy agents such as asparaginase.
Several groups have described that asparaginase discontinuation may lead to reduced EFS in ALL. 25,26Alternatively, TE may be seen with severe and sometimes fatal treatment-related toxicities, such as severe infections or pancreatitis.8][29] Finally, an increasing body of work suggests that there might be a bidirectional relationship between coagulation pathways and cancer biology, whereby the pathological activation of the hemostatic system following cancer both promotes TE and drives cancer progression. 30,31If this last mechanism were true, treatment refractoriness may be more common in children with ALL who developed a TE, especially if it occurred during the early stages of anti-leukemia therapy.Athale and colleagues have recently shown a sixfold increase in the risk of induction failure following TE during induction chemotherapy (TE ind ) in 794 children with ALL enrolled in the Dana Farber Cancer Institute (DFCI) 05-001 trial. 32Our group had recently reported a reduced OS and EFS in Canadian children treated for ALL and who developed a TE, where death was equally attributed to treatment-related mortality and progressive disease, 33 yet we had not explored the differential impact of TE ind and whether TE affected induction outcomes.Thus, to further explore this hypothesis, we performed a population-based retrospective cohort study, using the CYP-C registry.Our objective was to determine the incidence and risk factors of TE ind and to assess whether the development of TE ind was associated with induction failure in children with ALL.

| Data source
The CYP-C cancer registry prospectively collects data pertaining to children with cancer diagnosed and treated in pediatric oncology centers in Canada.Almost all children aged <15 years old are treated in these hospitals. 34In-depth diagnostic, treatment, and outcomes data are collected by trained research assistants or data managers.CYP-C is a collaboration between the Public Health Agency of Canada, the C17 Council, and the Canadian Partnership against Cancer.The collection of data in CYP-C is approved by the Research Ethics Boards of all participating centers outside Ontario, which forms the basis for this analysis.Several methods are used to ensure high-quality data, including education and training of data managers and data monitoring at each site. 35The Research Ethics Board at Centre Hospitalier Universitaire de Québec approved this study.The requirement for informed consent was waived given the use of secondary data and the retrospective nature of the study.

| Study population
We included patients who were (1) between 0 and 14 years of age at Patients with inherited thrombophilia were not excluded, as thrombophilia status is not collected in CYP-C and is not typically tested for most provoked TEs (e.g., catheter-associated TEs).

| Outcomes
Our primary outcome was TE ind , defined as a radiologically proven occlusion of a blood vessel diagnosed within 6 weeks of treatment start.TE grade 3 or above are captured in CYP-C, graded using the Common Terminology Criteria for Adverse Events, version 3.0. 36iefly, grade 3 refers to a TE requiring medical intervention (e.g., symptomatic deep venous thrombosis and pulmonary embolism), grade 4 refers to a TE associated with hemodynamic or neurologic instability, and grade 5 refers to a fatal TE.Grading of complications is performed at each site and adjudicated by the local principal investigator in case of uncertainty.TE was categorized as CVC-related if the thrombus had developed in the region of a central catheter, as per the site investigator, though the location of the TE and the CVC were not recorded.Thus, all TEs, regardless of site (e.g., deep venous thrombosis, pulmonary embolism, cerebral venous sinus thrombosis) are grouped together as the primary outcome.
Secondary outcomes included induction failure, treatment intensification, EFS, and OS.Induction failure was defined as a change or discontinuation of treatment protocol due to "disease progression/no response", using the CYP-C terminology, within 6 weeks of antileukemia treatment start.The exact bone marrow blast count at the end of induction was not routinely collected in CYP-C and minimal residual disease was captured systemically in CYP-C from 2018 onwards.
Treatment intensification was defined as a change from standard-to high-risk ALL therapy, for children treated as per Children's Oncology Group protocols, or from either standard-to high-risk or high-to very high-risk ALL therapy for those treated as per DFCI protocols, occurring within 6 weeks of antileukemia treatment start.This outcome was chosen as a surrogate of treatment response since endof-induction bone marrow blast count and minimal residual disease were variably available in CYP-C.Of note, these secondary outcomes are not mutually exclusive, as children diagnosed with induction failure would likely be subjected to treatment intensification, apart from children already treated with very high-risk ALL therapy.EFS was defined as the time between the date of leukemia diagnosis and the date of relapse, second malignancy, or death (whichever came first), and OS was defined as the time between the date of leukemia diagnosis and death.The cause of death was categorized as being caused by (1) relapsed disease, (2) treatment-related complications, or (3) external causes.The classification was performed at each site by the local principal investigator.In CYP-C, patients are followed prospectively for up to 5 years following the primary neoplasm and any subsequent cancer diagnosis to capture relapse or death.

| Covariates of interest
Covariates of interest included age at ALL diagnosis, sex, race, obesity, leukemia immunophenotype and risk stratification, unfavorable cytogenetics, and extramedullary disease (central nervous system [CNS]   and testicular involvement).Age at ALL diagnosis was categorized as <1 year old, 1-< 10 years old, and ≥10 years old, to allow comparison with preexisting studies 2,5,6 and based on the bimodal distribution of TE in children, with higher rates observed in infants and adolescents. 1,6Obesity was determined in children aged 2 years and above and was defined as BMI >99.9 percentile for age (zBMI >3) in children 2-4.99 years and >97 percentile for age (zBMI >2) in children 5 years and above. 37Initial and final leukemia risk stratification was determined for each patient using baseline characteristics, cytogenetics findings, minimal residual disease (when available), and chemotherapy protocols used (Supplementary Data Table S1).Favorable cytogenetics included any of the following findings among leukemia blasts: hyperdiploidy >50 chromosomes, ETV6::RUNX1 rearrangement, and trisomy of both chromosomes 4 and 10.Conversely, unfavorable cytogenetics included any of the following anomalies: hypodiploidy <44 chromosomes, KMT2A rearrangement, intrachromosomal amplification of chromosome 21 (iAMP21) or BCR::ABL1 rearrangement.CNS involvement represented CNS-2 or CNS-3 disease. 38We recorded whether patients experienced one of the following comorbidities during induction chemotherapy: infection grade 4 or above, and steroid-induced diabetes mellitus requiring long-term insulin or hypoglycemic agents, given the potential associations between these toxicities and TEs.

| Statistical analysis
Population characteristics were summarized descriptively using median and interquartile range (IQR) for continuous variables and frequency and proportion for categorical variables.
The cumulative incidence of TE ind and TE was estimated using the cumulative incidence function, with death considered a competing event.
Univariate and multivariable logistic regression models were performed to explore the predictors of TE ind .We examined Pearson correlation coefficients to evaluate collinearity which guided multivariable models.
Results are presented with odds ratio (OR) and 95% confidence interval (CI).For these analyses, children with TE during post-induction chemotherapy (TE post-ind ) were pooled with children without TE.
Bivariate logistic regression was used to estimate the association between TE ind with induction failure and treatment intensification.
Only TE ind occurring before the outcome of interest was considered.
When the number of events allowed, we performed multivariate analyses adjusted for age at leukemia diagnosis, sex, leukemia immunophenotype and leukemia initial risk stratification (dichotomized between high-or very high-risk vs. standard risk, given the small proportion of children with very high-risk ALL), given the association between these variables and both TEs and leukemia outcomes. 39The Kaplan-Meier survival method was used to estimate the 5-year OS and EFS of children with TE ind compared with those without TE ind , both for the whole cohort and stratified as per ALL risk (standard-vs high-risk).
Univariate and multivariable Cox proportional hazards models were used to determine the hazard ratio (HR) of death (for analysis of overall survival) or first event (for analysis of event-free survival) in children with TE ind .Multivariate analyses were adjusted for age at ALL diagnosis, sex, immunophenotype, and leukemia risk stratification. 39For those  who survived and did not experience an event, respectively, they were censored at their last contact with the healthcare system, 5 years after ALL diagnosis, or January 31st, 2020, whichever occurred first.
We performed two sensitivity analyses.First, we performed a sensitivity analysis where children who developed TE post-ind were removed.Second, we modified the definition of induction failure, where it was defined by a change or discontinuation of treatment protocol due to "disease progression/no response" within 12 weeks, to reflect the slower pattern of treatment response in some leukemias, such as T-cell ALL. 40mplete case analysis was used for all analyses.All tests were two-tailed with a p-value <.05 considered statistically significant.All analyses were conducted using SAS (Version 9.4, Cary, NC, USA).
The characteristics of patients with TE ind and TE post-ind and without TE are presented in Table 1.
In univariate regression analysis, the following variables were asso-  while no statistically significant difference between groups was seen for children with standard-risk ALL (adjusted HR: not estimable).
Almost all deaths (87.5%) were due to toxicities or complications of treatment, the remainder being attributed to relapsed ALL.There was no statistically significant difference in OS between patients with TE ind and TE post-ind (adjusted HR of death, TE post-ind vs. TE ind [95% CI]: 0.64 [0.27-1.50],p = .304).
In the sensitivity analysis where patients with TE post-ind were removed, the results were similar to the main analysis (data not shown).

| DISCUSSION
In this study, 1.7% of children developed a TE ind grade 3 and above during chemotherapy for newly diagnosed ALL.TE ind was independently associated with age <1 year old and >10 years old and T-cell immunophenotype.However, TE ind was not predictive of induction failure or treatment intensification, suggesting that the presence of TE ind was not associated with refractory ALL.
The incidence of TE ind observed in our cohort was slightly lower than that reported elsewhere.It is generally thought that most TEs occur during induction, due to the interaction of multiple risk factors such as more extensive burden of disease, leading to cell apoptosis and contributing to thrombogenicity, placement of CVC, and intensive asparaginase-containing chemotherapy. 41In a meta-analysis of 17 prospective studies, Caruso and colleagues have reported that approximately 4.8% (95% CI: 3.7-6.0%) of patients developed TE ind . 73][44] Our somewhat lower incidence rate might be attributable to the stringent definition of TE (grade 3 and above) used in CYP-C, our study population of children aged up to 14 years old (as the risk of TE increases in adolescents and young adults) as well as factors specific to the Canadian pediatric oncology population, such as approaches to detection methods and cancer treatment protocols.Alternatively, it is possible that some toxicities were not captured during data abstraction but, as CYP-C has restricted its definition to the most severe toxicities, we do not suspect a high rate of missing data.
Few studies have explored predictors of TE ind .Farinasso et al.
have described that catheter-related venous TEs occurring in induction were associated with an increase in catheter size/body surface ratio. 43Furthermore, several authors have reported, in concordance with our results, that TE occurring at any point during ALL treatment was associated with more advanced age, T-cell immunophenotype, and high-risk treatment protocols, 21,23,24,32,44,45 in addition to other risk factors such as prolonged exposure to asparaginase and prednisone (compared to dexamethasone). 7Important factors might contribute to the increased risk of TE ind in these subgroups, including the elevated white cell count at presentation and increased catheter size/ body size ratio for infants, and the use of anthracyclines in children with high-risk and T-cell ALL.While several authors have reported an association between mediastinal mass and TE in children with ALL, 24,45 its impact could not be explored specifically in this analysis as the presence of a mediastinal mass is not consistently captured in CYP-C; it is possible that the increased frequency of mediastinal mass in T-cell ALL partly explains its association with TE ind .
Little data are available regarding the relationship between TE ind and leukemia refractoriness.Athale et al. have reported that among the 794 children enrolled in the DFCI 05-001 trial, TE ind was strongly and independently associated with induction failure, after adjustment for age, sex, initial risk stratification, immunophenotype, presenting white blood cell count, and circulating blasts. 32The DFCI 05-001 trial included adolescents aged up to 21 years old, and TE graded 2 and above were considered.In their cohort, TE was not predictive of survival.Similarly, TE did not predict EFS in a Dutch cohort of children with ALL 23 and did not predict treatment response and relapse in an Israeli cohort of 1191 children with ALL (79 venous TEs). 22It is possible that chemotherapy protocols that rely heavily on the use of asparaginase might be more vulnerable to chemotherapy modifications following TE, as several protocols recommend withholding asparaginase until TE resolution, 25,46 or discontinuing asparaginase for CNS thrombosis. 26Our study findings do not support a relationship between TE ind and treatment refractoriness.Nonetheless, TE ind was independently associated with all-cause death, with most deaths attributable to complications, suggesting patients who developed a TE ind might have unmeasured characteristics making them more vulnerable to toxicities.Further research will be needed to explore whether TE ind could be associated with genetic or acquired susceptibility to chemotherapy-induced toxicities and to describe long-term outcomes in children with TE ind .
Our results provide an important insight into the incidence and risk factors of TE ind in a large, population-based cohort, allowing for unbiased reporting of risk factors and outcomes.Our study also provides important information as researchers continue to untangle the relationship between TE, cancer progression, and cancer outcomes.It will be crucial to capture long-term EFS and OS data in randomized clinical trials exploring the efficacy and safety of thromboprophylaxis during induction chemotherapy for ALL, as thromboprophylaxis likely will alter cancer-and treatment-induced hemostatic abnormalities. 47rther work will also be needed to explore these relationships for children treated with different anti-leukemia protocols and in other jurisdictions.
Strengths of our study include its large sample size, detailed cohort identification, and the use of high-quality, population-based, prospectively collected data.We also restricted our exposure definition to clinically relevant TEs, as grade 1 or 2 TEs (i.e., asymptomatic TEs or TEs that did not lead to an intervention), whose clinical relevance is debated, 48,49 were not included.However, some limitations should be acknowledged.First, the CYP-C registry collected few TErelated variables.The location of TEs is unknown, and the registry did not distinguish between venous and arterial TEs, although we suspect the vast majority of TEs were venous events.Yet, it is possible that all TEs do not carry the same prognostic impact.Second, as data collection spanned over two decades and 12 institutions, representing multiple chemotherapy protocols, no standardized definition was used for induction failure.As the exact blast count was not available, it is possible that some cases of induction failures were not captured, if a high blast count did not prompt a change in treatment protocol.[52] Additionally, we do not expect differential detection of this outcome based on the presence of TE, and we do not expect this could have biased our results.Third, individual patient-level data regarding therapy alteration, for example, intrathecal chemotherapy administrations, were not collected, and whether chemotherapy was delayed or withheld following TE ind or anticoagulation is unknown.This would have been more concerning if a relationship had been identified between TE ind and induction failure.Finally, there was no information regarding the use of antithrombotic agents such as low molecular weight heparin and/or antithrombin replacement before the development of TE.However, primary thromboprophylaxis is currently not standard of care for pediatric cancer patients in Canada.
In summary, approximately 2% of children with ALL sustained a TE ind .TE ind was independently associated with age less than 1 year or more than 10 years and T-cell immunophenotype, in addition to high-risk ALL risk stratification and CNS involvement in univariate analysis.TE ind was not predictive of induction failure or treatment intensification.However, it was independently associated with worse event-free survival in children with high-risk ALL, as well as a twofold increase in the risk of mortality, more commonly attributable to toxicities and complications of treatment.
cancer diagnosis (aligned on the inclusion criteria of the CYP-C database); (2) diagnosed with de novo ALL between January 1, 2001 (inception of the database) and December 31, 2019; (3) diagnosed and treated at one of the 12 pediatric oncology centers in Canada outside Ontario.We excluded children with Burkitt leukemia, given the major differences in treatment protocols.Children diagnosed and treated in hospitals located in Ontario, whose data are provided to CYP-C by the Pediatric Oncology Group of Ontario Network Information System, were excluded as data regarding complications of treatment (including TE) were not collected systematically over the study period.Most children were treated as per Children's Oncology Group (or its legacy groups) or DFCI treatment protocols.Patients typically received a received a 4-week multiagent induction.Children with standard risk ALL received 3-drug induction with steroids, vincristine, asparaginase, and intrathecal chemotherapy.Patients with high-risk ALL additionally received anthracyclines.Based on institutional guidelines, central venous catheters (CVCs) were inserted shortly after diagnosis to allow for safe administration of vincristine and anthracyclines.

T A B L E 1
Patient characteristics a .Patient characteristics Total, N = 2589, n (%) Patients without TE, N = 2440, n (%) Patients with TE in induction, N = 45, n (%) Patients with TE in post-induction, N = 104, n (%) 46 were excluded because of a diagnosis of Burkitt ALL or secondary ALL, leaving 2589 included patients in the study.The median age at F I G U R E 1 Cumulative incidence of thromboembolism in children with newly diagnosed acute lymphoblastic leukemia.Blue-shaded area represents 95% confidence interval around cumulative incidence of thromboembolism.The gray area represents the interval between date of leukemia diagnosis and the end of induction chemotherapy (i.e., 6 weeks after start of ALL therapy).[Color figure can be viewed at wileyonlinelibrary.com]T A B L E 2 Predictors of thromboembolism in induction among children with acute lymphoblastic leukemia.

F I G U R E 2
Event-free and overall survival for children with and without thromboembolism in induction.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 3 Overall survival for entire cohort and stratified by leukemia risk stratification.[Color figure can be viewed at wileyonlinelibrary.com]

Treatment intensification occurred in 5 (
11%) and 429 (16.9%) of children with and without TE ind .TE ind was not significantly associated with treatment intensification in unadjusted (OR [95% CI]: 0.62 [0.24-1.57],p = .311)and adjusted analyses (OR [95% CI]: 0.66 [0.26-1.69],p = .384).As shown in Figure 2, TE ind was associated with a twofold increase in the risk of an event, that is, relapse, second malignancy, or death (HR [95% CI]: 2.03 [1.08-3.81],p = .028),although this was no longer significant in multivariable analysis (adjusted HR [95% CI]: 1.86 [0.98-3.51],p = .057).When analyses were restricted to children with high-risk ALL, TE ind was independently associated with worse EFS (adjusted HR [95% CI]: 2.16 [1.09-4.25],p = .026).Moreover, as shown in Figure 3, TE ind was independently associated with overall survival during the observation period, leading to a two-to threefold increase in the risk of all-cause death within 5 years (adjusted HR [95% CI]: 2.54 [1.20-5.03],p = .014).The adverse impact of TE ind was driven by children with high-risk ALL (adjusted HR [95% CI]: 2.88 [1.39-5.95],p = .004), with TE ind : age <1 or ≥10 years (compared to age 1-<10 years), T-cell immunophenotype, initial high-risk ALL, and CNS disease.In multivariable analysis, only age <1 year and ≥10 years and T-cell ALL were independently associated with the development of TE ind (Table2).Induction failure was seen in 53/2589 patients (2.1%).No children with TE ind were considered to have induction failure (OR: not estimable).When the timeframe used to define induction failure was set at 12 weeks rather than 6 weeks, induction failure was seen in 58 children (2.2%).Again, no patient with TE ind developed induction failure (OR: not estimable), in concordance with the primary analysis.