Venous Thromboembolism in Pediatric Inflammatory Bowel Disease: A Case-Control Study

Abstract

Objectives:

Inflammatory bowel disease (IBD) is associated with increased risk of venous thromboembolism (VTE). Despite this recognized risk, there are limited data and no anticoagulation guidelines for hospitalized pediatric IBD patients. The objectives of this study were to characterize pediatric IBD patients with VTE and determine risk factors.

Methods:

This was a nested case-control study comparing hospitalized children with IBD diagnosed with VTE to those without VTE over a decade at a large referral center. Standard descriptive statistics were used to describe the VTE group. Multivariable conditional logistic regression was used to assess risk factors.

Results:

Twenty-three cases were identified. Central venous catheter (CVC) presence (odds ratio [OR] 77.9; 95% confidence interval [CI]: 6.9–880.6; P<0.001) and steroid use (OR 12.7; 95% CI: 1.3–126.4; P = 0.012) were independent risk factors. Median age at VTE was 17years (interquartile range [IQR] 13.5–18.2), and in 48%, VTE was the indication for admission. Median duration of anticoagulation was 3.8months (IQR 2.3–7.6), and there were no major bleeding events for patients on anticoagulation. There were no patients with known sequelae from VTE, though 22% had severe VTE that required interventions.

Conclusions:

Pediatric patients with IBD are at risk for VTE, although the absolute risk remains relatively low. The safety and efficacy of pharmacologic thromboprophylaxis needs to be further evaluated in this population with attention to risk factors, such as steroid use and presence of CVC.

Patients with inflammatory bowel disease (IBD) are at increased risk for venous thromboembolism (VTE). Studies in adults have shown that IBD is an independent risk factor for VTE, with a 3-fold greater risk overall and up to 15-fold increased risk during disease exacerbation (1,2). Risk of VTE has been shown to be high during hospitalized periods (2,3). Given the significant morbidity and mortality related to VTE, guidelines recommend that hospitalized adult patients with IBD receive thromboprophylaxis (1,4–6).

Although VTE is less common in children, pediatric patients with IBD are also at increased risk, with an estimated incidence of 0.09% to 1.9% (7–10). Previous pediatric studies are limited by sample size, lack of a control group, and inability to gather disease-specific and clinical outcome data. In one population-based study, children hospitalized with IBD flare were estimated to have a 6-fold increased risk for pulmonary embolism and deep vein thrombosis (DVT) as compared with hospitalized patients without IBD (8). In another study, the relative risk for VTE was 2.37 (95% confidence interval [CI]: 2.16–2.61) in patients with Crohn disease (CD) and 1.99 (95% CI: 1.51–2.64) in patients with ulcerative colitis (UC) as compared with hospitalized patients without IBD (7). Risk factors that have been proposed include age at VTE diagnosis, inflammation, malnutrition, dehydration, hypoalbuminemia, surgical intervention, presence of central venous catheter (CVC), total parenteral nutrition, and use of certain medications (7,9,11–13). Despite recognition of this increased risk in pediatric IBD, a survey of pediatric gastroenterologists showed that only 36% reported providing VTE prophylaxis (mechanical or pharmacologic). Barriers to VTE prophylaxis were lack of formal guidelines and supporting evidence (14).

The limited data available about risk factors and outcomes in the pediatric IBD population have resulted in continued questions regarding the scope of the problem and whether VTE prophylaxis should be instituted in children with IBD. The objectives of this study were to characterize IBD patients with VTE in a large pediatric referral center, including symptom-onset and outcomes, and to determine risk factors associated with VTE in hospitalized children and adolescents with IBD.

MATERIALS AND METHODS

Study Design and Selection of Study Population

We performed a nested case-control study comparing hospitalized children and adolescents with IBD diagnosed with VTE to those without VTE over a decade at a single large referral center.

The study population included patients with IBD hospitalized from January 2008 through December 2018 who were 2 to 21years of age at the time of hospitalization. Cases and controls were initially selected by using a structured query language (SQL) algorithm of the electronic health record (EHR). The diagnosis of IBD was based on having a diagnosis code for IBD (Supplement 1, http://links.lww.com/MPG/C233) and an outpatient visit with a gastroenterologist provider at our center. Cases were selected if they met these criteria and had a diagnosis code for VTE (Supplement 2, http://links.lww.com/MPG/C234). Each case was reviewed independently by two physicians, and further adjudicated to ensure that hospitalization date and age were within the pre-defined range, endoscopy had been performed and that the histopathology was consistent with IBD, and that VTE was confirmed by imaging. Controls were matched up to 5 to 1 on sex, race, age at hospitalization (+/− 2years), and date of hospitalization (+/− 2years). Controls were randomly selected and matched. Controls were then validated independently by 2 physicians through chart review, ensuring that hospitalization date, age, sex, and race were correctly matched to cases and that the IBD diagnosis was correct as defined above. If controls were determined to not fit the predefined criteria for matching, new controls were randomly selected through SQL. This was an iterative process that was repeated 3 times before the study population was completed, and there was at least 1 matched control per case. Patients were excluded if they were admitted for elective surgery or hospitalized for a chief complaint that was unrelated to their diagnosis of IBD based on chart review.

Clinical Characteristics

Clinical characteristics were evaluated for both cases and controls by retrospective chart review of the EHR and included date of hospitalization, age at hospitalization, sex, race, ethnicity, IBD type, age at IBD diagnosis, IBD and non-IBD-related past medical and surgical history, IBD and non-IBD-related medications, disease activity, admission duration (in days), number of admissions in the preceding 2months, need for Intensive Care Unit (ICU) care, IBD-related surgery performed during admission, presence of CVC, body mass index (BMI), personal and family history of VTE, tobacco use, birth control use, and history of malignancy. Age of IBD diagnosis was determined based on the date of the first diagnostic endoscopy. IBD type (CD, UC, IC) was assigned based on diagnosis at the time of hospitalization and Paris classification based on the most extensive endoscopy at or before the time of hospitalization (15). BMI was converted to BMI z score using the CDC Growth Reference data (16). To assess disease severity, Pediatric Crohn Disease Activity Index (PCDAI) and Pediatric Ulcerative Colitis Activity Index (PUCAI) scores were calculated using data as close to the time of hospitalization. In order to combine PCDAI and PUCAI into 1 activity score, numerical values were converted to categorical groups of remission (PCDAI and PUCAI <10), mild (PCDAI 10–27.5, PUCAI 10–34), moderate (PCDAI 30–37.5 and PUCAI 35–64), and severe (PCDAI >40 and PUCAI >65) disease activity.

For cases, the location of the VTE, mode of diagnosis, related symptoms, type of anticoagulation used, duration of anticoagulation, VTE resolution, complications on anticoagulation, and recurrent episodes of VTE on subsequent admissions were recorded. Major bleeding and clinically relevant nonmajor bleeding events were defined based on recommendations from the pediatric subcommittee of the International Society of Thrombosis and Hemostasis (ISTH) (17,18). Testing for genetic thrombophilia, if available, was also collected.

Analysis

Stata version 15.1 (College Station, TX) was used for all analyses. Standard descriptive statistics included medians and interquartile ranges for continuous variables and frequencies for categorical variables. Univariate conditional logistic regression analysis was performed for hypothesized risk factors and statistical significance was defined as a P value of <0.05. Multivariable conditional logistic regression analysis was performed using risk factors that were statistically significant in the univariate analysis and including additional biologically plausible risk factors. Effect modification was then evaluated for age and sex with the risk factors that remained statistically significant in the multivariable model using 2-way interactions. Due to performance of multiple comparisons when looking at the interaction terms, a more conservative P value of <0.03 was set using Bonferroni correction.

Ethical Considerations

This study was reviewed by The Children’s Hospital of Philadelphia Institutional Review Board and determined to meet criteria for exemption per 45 CFR 46.104(d) 4(iii).

RESULTS

Proportion of Venous Thromboembolism

There were 23 cases identified over this 10-year period. There were 111 controls given that one of the cases could only be matched to 1 control; the other cases were matched 5:1. The proportion of cases was determined using the denominator of 1869 patients, which was based on the definition outlined above to identify the study population of hospitalized IBD patients during this 10-year period. The proportion of VTE in our IBD population was therefore, 1.3%.

Clinical Characteristics and Course of the Venous Thromboembolism Cases

Table 1 describes the demographic and clinical characteristics of the VTE group. The median age at time of VTE diagnosis was 17years (IQR 13.5–18.2). The median time from IBD diagnosis to diagnosis of VTE was 2.3months (IQR 0.8–41.8). Six patients (26%) had been admitted within the 2 months before VTE diagnosis. Eleven patients (48%) presented with symptoms of clot as the indication for their admission. These symptoms varied based on location of the clot but included extremity swelling and pain, headache, altered mental status, chest pain, palpitations, and hematuria. Nine of these 11 patients had active disease in addition to VTE Mitchel et al symptoms on admission whereas 2 had only mild disease activity but had been recently admitted for a flare. The most common locations for VTE were upper extremity (56%), lower extremity (17%), and lung (13%). Other sites included intracranial, intracardiac, and renal vein thrombosis as demonstrated in Figure 1. All of the patients with an upper extremity clot had a CVC located in the upper extremity near the site of thrombosis. Median time from CVC insertion to VTE event was 10days (IQR 2–14).

TABLE 1.

Description of venous thromboembolism population

Characteristics, N=23	N (%) or median (IQR)
Age at clot in years	17 (13.5–18.2)
Time from IBD to clot diagnosis in months	2.3 (0.8–41.8)
VTE was indication for admission	11 (48)
Admitted in previous 2 months	6 (26)
Admitted to the ICU	8 (35)
Started on anticoagulation	21 (91)
Duration of anticoagulation in months	3.8 (2.3–7.6)*
Occurrence of clot resolution	20 (87)
Time to resolution of clot in months	2.3 (1.4–3.1)
Required additional intervention (ie, thrombolysis)	5 (22)
Bleeding complication on anticoagulation	0 (0)
Recurrence of clot in subsequent encounter	1 (5)

ICU = intensive care unit.

^*Does not include 2 patients who were still on anticoagulation at the end of the study period.

FIGURE 1.

Location of venous thromboembolism in the 23 cases. *Single patient could have more than 1 clot location.

Given that about half of the cases were admitted for the indication of VTE, we compared clinical characteristics and outcomes of patients who were admitted with the indication of VTE to those who were presumed to have developed VTE while admitted, including age at IBD diagnosis, age at VTE diagnosis, time from IBD to VTE diagnosis, admission duration, proportion with ICU admission, CVC presence, steroid use, IBD type, anticoagulation use, anticoagulation duration, time to clot resolution (in those with resolution), and VTE recurrence. There were no statistically significant differences between these groups.

Five cases (22%) had life or limb-threatening thrombosis that required invasive management (thrombolysis, thrombectomy, or emergent colectomy). Clinical characteristics, presenting symptoms, and treatment for these 5 patients are outlined in Table 2. Overall, after intervention and long-term anticoagulation, these patients had minimal known sequelae by the end of follow-up, including postthrombotic pain syndrome, extremity swelling, neurologic deficits, or cardiac insufficiency. Eight patients, including these 5 patients, required ICU admission.

TABLE 2.

Description of 5 cases requiring invasive interventions with life or limb-threatening venous thromboembolism

Age	Sex	IBD type	Thrombus location	Symptoms at presentation	Therapy	Thrombus outcome
16	Female	Crohn disease	Right lower extremity DVT	Lower extremity pain, swelling	Pharmacomechanical thrombolysis and balloon angioplasty with anticoagulation for 6 months	No postthrombotic syndrome
17	Female	Crohn disease	Left iliofemoral DVT	Lower extremity pain, swelling	Pharmacomechanical thrombolysis, balloon angioplasty, and stent placement with anticoagulation for 6 months	No postthrombotic syndrome
13	Male	UC	Cerebral sinovenous thrombosis with infarct	Altered mental status, bilateral lower extremity weakness, headache	Pharmacomechanical thrombolysis with anticoagulation for 6 months	No known sequelae
13	Male	UC	Intracardiac right ventricular thrombus with embolization to right pulmonary artery	Palpitations, chest pain, tachycardia	Surgical thrombectomy and long-term anticoagulation*	No known sequelae
7	Female	UC	Thrombotic storm: bilateral pulmonary, splenic, renal, sinovenous with infarct and upper extremity catheterrelated thromboses	Chest pain, shortness of breath, headache, altered mental status	Plasmapheresis, emergent subtotal colectomy, and long-term Anticoagulation*	No known sequelae

DVT = deep vein thrombosis; UC = ulcerative colitis.

^*Patients sill on anticoagulation at the end of the study period.

Twenty-one (91%) patients were started on anticoagulation. Two patients with CVC present were not started on anticoagulation but instead the CVC was removed and the VTE was followed with serial ultrasound (US) in discussion with hematology. Eighteen of the patients on anticoagulation were started on enoxaparin and 3 on heparin. Four patients were transitioned to warfarin and 1 to rivaroxaban for long-term therapy, whereas the rest were continued on enoxaparin. The median duration of treatment was 3.8months (IQR 2.3–7.6). Resolution of thrombus, based on imaging, was documented in 87% of patients. Three patients did not have resolution, 2 of whom remained on anticoagulation at the conclusion of the study and 1 who had CVC removed; however, all 3 patients had improvement on US. In patients with documented resolution, median time to resolution was 2.3months (IQR 1.4–3.1). There were no episodes of major bleeding or clinically relevant nonmajor bleeding in patients on anticoagulation (17,18). Median time of hematology follow-up after discharge was 6months (IQR 3– 17). One patient had recurrent thrombosis after initial resolution of thrombus and with re-insertion of a CVC while off of anticoagulation. There were no deaths. Of note, none of the controls were started on anticoagulation for thromboprophylaxis.

Thrombophilia Testing

Thrombophilia testing was not uniformly performed. Of the 12 patients who underwent testing, 1 patient was found to be heterozygous for a prothrombin gene mutation. No patient was identified with inherited protein C, S or antithrombin deficiency, antiphospholipid antibody syndrome, or factor V Leiden mutation. Of the 9 patients who had factor VIII activity testing, 6 had levels >150%. There was no significant difference in platelet count at the time of admission between cases and controls (p=0.23).

Risk Factors for Venous Thromboembolism

In Table 3, the results from the univariate and multivariable analyses are presented. In the univariate analysis, presence of a CVC and current steroid use were more common among cases than controls: OR 25.3 (95% CI: 5.68–112.80; P<0.001) and OR 4.5 (95% CI: 1.50–13.72; P=0.01), respectively. ICU stay and admission duration were also significantly more frequent among cases as compared with controls. These two variables, however, were not included in the multivariable model as both were more likely a result of as opposed to a risk factor for VTE. There was a higher proportion of cases with colonic disease, though this was not statistically significant. There was also a higher proportion of cases with recent admission in the preceding 2months, though this was not statistically significant between the two groups. Malignancy and tobacco use were not included as there were no patients with these hypothesized risk factors. In multivariable analysis, CVC presence and current steroid use were independent risk factors for clot: CVC presence; OR 42.4 (95% CI: 6.8–265.8; P<0.001) and steroid use OR 7.5 (95% CI: 1.5–38.2; P= 0.02). CVC presence and steroid use remained statistically significant when controlling for IBD subtype, disease activity, and infliximab exposure. We found a possible interaction with age and presence of CVC (P = 0.04), the younger the patient, the greater the risk of clot in the presence of CVC. This P value, however, did not pass the conservative Bonferroni threshold of 0.03.

TABLE 3.

Univariate and multivariable model to assess risk factors for venous thromboembolism

Factor	Case, N=23	Control, N=111	Univariate OR	(95% CI)	P value*	Multivariable OR	(95% CI)	P value*
Age at hospitalization, years	16.88 (13.51–18.17)	16.87 (14.85–18.11)	–	–	–	–	–	–
Sex
Male	12 (52)	55 (49)	–	–	–	–	–	–
Race
White	20 (87)	97 (87)	–	–	–	–	–	–
Age at IBD diagnosis (years)	13.42 (10.03–17.10)	12.89 (9.12–15.52)	1.08	(0.93–1.24)	0.31	–	–	–
BMI z score	−0.47 (−1.55 to 0.80)	−0.24 (−1.08–0.59)	0.92	(0.66–1.30)	0.65	–	–	–
IBD type
Crohn disease	14 (60.9)	80 (72.1)	–	–	–	–	–	–
Ulcerative colitis	7 (30.4)	24 (21.6)	1.88	(0.63–5.58)	0.26	2.65	(0.35–20.36)	0.35
Indeterminate colitis	2 (9)	7 (6.3)	2.25	(0.33–15.10)	0.41	0.45	(0.00–1517.13)	0.85
Disease activity score
Remission/mild	9 (40.9)	33 (33.7)	–	–	–	–	–	–
Moderate	6 (27.3)	45 (45.9)	0.49	(0.16–1.52)	0.22	0.57	(0.11–2.90)	0.50
Severe	7 (31.8)	20 (20.4)	1.02	(0.27–3.84)	0.98	0.34	(0.46–2.50)	0.30
CVC present	15 (65.2)	10 (9.0)	25.3	(5.68–112.80)	<0.001	77.68	(6.86–880.6)	<0.001
Steroid exposure	17 (74)	51 (46.0)	4.5	(1.50–13.72)	0.01	12.7	(1.28–126.4)	0.03
Infliximab exposure	6 (26.1)	38 (34.2)	0.66	(0.24–1.83)	0.42	0.55	(0.12–2.64)	0.46
Admission duration, days	10 (7, 24)	5 (3, 7)	1.18	(1.07–1.29)	0.001	–	–	–
ICU admission	8 (34.78)	2 (1.80)	18.90	(4.0–89.32)	<0.001	–	–	–
Admission in the preceding 2 months	6 (26.10)	16 (14.7)	2.10	(0.73–6.05)	0.17	–	–	–
Platelet count at admission	335.0 (249.0–445.0)	376.5 (273.5–459.5)	1.00	(0.99–1.00)	0.23	–	–	–
Birth control use	2 (22.22)	15 (29.41)	0.56	(0.10–3.09)	0.51	–	–	–

BMI = body mass index; CI = confidence interval; CVC = central venous catheter; ICU = intensive care unit; OR = odds ratio.

^*P value significance defined as <0.05.

DISCUSSION

This single-center case-control study of pediatric IBD patients with VTE provides important insight into the characteristics and outcomes of these thrombotic events as well as insights regarding potential risk factors. This is the first study in pediatrics to use a case-control design with detailed electronic medical record data available to characterize this population.

The proportion of VTE among hospitalized children and adolescents with IBD in our cohort was 1.3%, which is similar to prior studies (1,7–10). Risk factors that were identified included presence of a CVC and steroid use, and age was found to be a possible effect modifier in the association with CVC: the younger the patient, the greater the risk of clot in the presence of CVC. Among patients with VTE, events occurred close to the time of diagnosis, and about half had VTE as the indication for admission. A significant portion of patients had severe clot that required additional interventions. Overall outcomes in these patients were good with no major bleeding events on anticoagulation or known sequelae, such as postthrombotic pain syndrome, extremity swelling, neurologic deficits, or cardiac insufficiency through the end of study follow-up.

The risk factors identified in this study are consistent with previous studies. Diamond et al (10) showed a high rate of CVC-related thrombosis in pediatric IBD patients and also demonstrated a decrease in incidence of CVC-related thrombosis with use of prophylactic anticoagulation. In addition, using a large inpatient database, an increased risk of thrombosis and infection in patients with IBD receiving parenteral nutrition via CVC was found (19). Steroid use has also been identified as a risk factor for VTE. Experimental studies have shown that steroid treatment increases levels of clotting factors and fibrinogen (12,20). Steroids have been shown to be an independent risk factor for developing VTE in adult IBD patients (21). In one meta-analysis evaluating VTE risk in IBD patients with or without systemic corticosteroid treatment, there was a significantly higher rate of VTE complications in steroid-treated IBD patients compared with IBD patients without steroid exposure (OR 2.2, 95% CI: 1.70–2.86) (12). Colonic disease has previously been described as a risk factor for developing VTE (1,13). There was a higher proportion of colonic disease among cases; however, this was not significantly different from controls, perhaps limited by the small sample size.

Prior literature has described risk of VTE increasing with increasing age (7,8). Consistent with this observation, median age of clot among cases was 17years. In our study, however, there was suggestion of younger age increasing the association between CVC and VTE. We hypothesize that risk of VTE with a CVC increases in younger patients because of the size discrepancy between the size of the catheter and the vessel (22). This finding suggests that particular attention may need to be directed towards younger children with IBD who have a CVC. Of note, we sought to evaluate a specific age cut-off at which the interaction term between CVC and age was significant; however, because of small sample size, we were limited in our analyses, and future exploration is warranted.

The pathogenesis of VTE in patients with IBD has been associated with increased activation of the coagulation cascade, impaired fibrinolysis, increased platelet count, and dysfunction of the endothelium (23,24). Inherited prothrombotic disorders have not been found to be increased in the IBD population as compared with the general population (23,25). This is consistent with the low prevalence of inherited thrombophilia detected in our cases. Interestingly, the majority of cases who had testing for FVIII had elevated levels. FVIII is an acute-phase reactant often elevated in inflammatory states, and has been found to be an independent risk factor for VTE in other cohorts (26–28). As controls did not have FVIII levels measured, it is, however, unclear from this study if elevated FVIII is a risk factor for thrombosis in patients with IBD. Although thrombocytosis has been implicated as a factor leading to VTE, there was no significant difference in platelet count between cases and controls.

Notably, 22% of cases in our study had life or limb-threatening thrombosis that required invasive management including thrombolysis, thrombectomy, or emergency colectomy. This proportion of severe cases is higher than is typically reported in pediatric thrombosis cohorts. Zitomersky et al (9) in a prior pediatric IBD case series described a similarly high proportion of severe VTE (29,30). Despite the associated morbidity in our population, long-term outcomes in these patients were good with no known sequelae described through the duration of the study.

Time from diagnosis to thrombosis event was skewed towards time of diagnosis. We speculate that this may be a result of having more active disease, necessitating hospitalization and steroid use, though this temporal association has not been described previously.

There are several limitations of this study. First, this is a single-center study at one quaternary center, limiting generalizability. Unlike previous studies that utilized insurance-based databases with larger numbers, access to medical records allowed for collection of specific details and an in-depth evaluation of each case. Second, given the retrospective nature of this study, there were certain risk factors that we were interested in but could not measure as they are not routinely documented or tested in patients without VTE, such as family history and degree of mobility. In addition, because of small sample size, we may not have been able to detect a difference in some risk factors whereas for other risk factors, the effect size may have been inflated. Third, because of the matching of certain covariates in this study, we were unable to evaluate them in the regression model, such as age at hospitalization, sex, and timing of IBD diagnosis to hospitalization. We did not match on IBD type and included this as a potential risk factor in the regression model; however, in doing this, we were unable to stratify the analysis by IBD type because of resultant unbalanced matching and small sample size. In addition, true incidence was unable to be determined, given the case-control study design and absence of a time measurement. Finally, we relied on diagnosis codes to define our populations, and while all case and control records were adjudicated, there may have been misclassification of the VTE diagnosis, leading to a potential underestimation of the proportion with VTE.

CONCLUSIONS

Pediatric patients with IBD have increased risk for VTE, although the absolute risk remains relatively low. Given the significant clots reported in some of our patients, the safety and efficacy of pharmacologic thromboprophylaxis based on risk stratification needs to be further evaluated. Our study identifies 2 risk factors that should be considered, CVC presence and steroid use. Nonpharmacologic interventions, such as mobilization, hydration, and serial compression devices should be considered. Future prospective, multicenter cohort studies are necessary to evaluate the risk of VTE in pediatric IBD patients and to inform guidelines for thromboprophylaxis.

What Is Known

• Pediatric patients with inflammatory bowel disease are at increased risk for venous thromboembolism

• Accepted thromboprophylaxis guidelines exist for adult patients with inflammatory bowel disease but do not exist in pediatrics.

What Is New

• Steroid use and presence of central venous catheter are risk factors for venous thromboembolism in pediatric patients with inflammatory bowel disease.

• Pharmacologic thromboprophylaxis should be considered in children and adolescents with inflammatory bowel disease based on risk stratification.