Fifty-six children with pulmonary embolism have been reported in this cohort, equivalent to an incidence of 5·7/10 000 admissions to the Hospital for Sick Children during this 8-year time period and 4·6/100 000 within the catchment population. These rates are significantly higher than those quoted from registry data (Andrew et al, 1994; Van Ommen et al, 2001; Stein et al, 2004) and may be more accurate due to comprehensive reporting. Alternatively, they could reflect the biased population of a tertiary centre.
Risk factors for childhood PE
Acquired risk factors have been shown to play an important role in provoking venous thromboembolic events in childhood. A CVC is the most consistent factor contributing to the occurrence of thrombotic events in childhood, present in 32·8–63·6% of children with thrombosis (Andrew et al, 1994; Van Ommen et al, 2001) and as many as 94% of neonates with thrombosis (Van Ommen et al, 2001). In a study of 244 children with CVC-related venous thrombosis, 16% were found to have symptomatic PE (Massicotte et al, 1998). Most of the subjects were not investigated for PE and the incidence may therefore have been underestimated. The association between CVC-related thrombosis and PE has been further strengthened by a screening study in children receiving parenteral nutrition (Dollery et al, 1994) and an autopsy study of children admitted to an intensive care unit (Derish et al, 1995). Twenty patients in our cohort (35·7%) had a CVC in situ at the time of their PE diagnosis and 12 patients (21·4%) had CVC-related DVT with 50% of these located in the upper venous system.
Immobility, recent surgery and congenital heart disease are further risk factors for thromboembolic disease in children, as shown in previous studies (Andrew et al, 1994; Stein et al, 2004; Rajpurkar et al, 2007) and confirmed by our current study. The presence of malignancy contributes to the occurrence of thrombotic events in a number of ways, including increased thrombin generation due to the malignancy, the use of CVCs for the administration of chemotherapy and blood products, and particular chemotherapeutic agents, such as asparaginase (Lee, 2002). In a study of 452 children treated for haematological malignancy in a single centre, 2·9% developed symptomatic PE (Uderzo et al, 1995). Children with PE represented 0·4% (3/750) of the total number of children treated for haematological malignancy and 0·6% (7/1106) of children treated for solid tumours at our institution during the study period.
Nephrotic syndrome in children has been shown to induce a hypercoagulable state with elevated plasma levels of factor VIII, fibrinogen and lipoprotein (a), a reduction in antithrombin level and hyperaggregability of platelets (Hoyer et al, 1986). The use of V/Q scanning to screen asymptomatic children with nephrotic syndrome has demonstrated changes consistent with PE in 27·9–40% (Hoyer et al, 1986; Huang et al, 2000). Four patients in our cohort had nephrotic syndrome, making it a significant, albeit less common, risk factor.
Thrombus occurring in association with lower limb venous malformation can result in PE. Embolism can be from proximal DVT but can also occur with isolated calf vein thrombosis due to the presence of abnormal communications between the superficial and deep venous systems (Huiras et al, 2005). This occurred in two patients in our cohort, highlighting the importance of identifying and treating distal thrombi in patients with venous malformations.
In contrast to the adult population, symptomatic PE following trauma in paediatric patients is uncommon, occurring in only two in a study of 28 692 unselected trauma patients. Both of these children had spinal cord injuries with associated paraplegia (McBride et al, 1994). This low incidence was supported by our results, identifying only one patient with PE following trauma, in this case a severe head injury. This patient had a CVC-related lower limb DVT.
According to our data, a higher level of suspicion for PE should be entertained in children with congenital heart disease, nephrotic syndrome and in those with malignancy, particularly that which is metastatic or advanced. Idiopathic PE is rare in childhood, the majority of children having multiple risk factors for thromboembolic disease at presentation.
The age distribution of children presenting with PE reflects that of all thromboembolic events in childhood, with a peak in infancy and a later peak in adolescence (Andrew et al, 1994). The probable reason for the fall in incidence occurring after 16 years of age is the transition of care to adult centres.
The clinical features of PE occurring in adolescence have previously been described in a cohort of 18 patients between the ages of 12 and 21 years presenting to a single centre (Bernstein et al, 1986). This study described a predominance of female patients, in three quarters of whom PE complicated elective abortion or the use of an oral contraceptive pill (OCP), although these patients were all 19 years of age or older. Our series showed a more equal sex distribution within the adolescent age group, 54·8% female, and a variety of underlying diagnoses reflecting the overall characteristics of the cohort. Four events were related to the use of an OCP. The major point of note related to the adolescents within our cohort is the identification of an inherited or acquired thrombophilic abnormality in 52% of those tested. All but one of the patients found to have a thrombophilic abnormality were between the ages of 12 and 17 years.
Some guidelines have advised against the screening of children with unselected thrombotic events for inherited thrombophilia (Walker et al, 2001) due to the low yield of positive results, thrombophilia being detected in only 8·8% (Andrew et al, 1994). However, with a thrombophilic abnormality demonstrated in 52% of adolescents tested and 35% overall, these guidelines should perhaps be revised in the setting of PE, particularly in adolescence. In contrast, thrombophilia testing could be deferred in children presenting with PE in the setting of CVC-related DVT as only one of 12 such patients in this cohort had a thrombophilic abnormality (persistently elevated factor VIII level) and this patient was also a teenager. In this cohort, the finding of heterozygosity for the F5 R506Q mutation at a rate no higher than that in the general population (2·5% of those undergoing genetic analysis) supports recent literature suggesting a relatively low risk of PE in these patients compared to those with F2 G20210A mutation (Martinelli et al, 2007).
Clinical and radiological diagnosis of childhood PE
Clinical prediction rules, such as the Well’s clinical probability score, have been validated in adults presenting with symptoms of PE and provide a useful measure of pre-test probability (Wells et al, 1998). The combination of a negative D dimer and low clinical probability score can safely rule out PE, negating the need for diagnostic imaging (Wells et al, 2000). No such clinical probability score has been validated in a paediatric cohort. A negative quantitative D dimer was not able to rule out PE in 4/30 children tested and quantitative D dimer was negative in 36% and 40% of children with PE in two recently reported studies (Rajpurkar et al, 2007; Strouse et al, 2007), a demonstration of the poor negative predictive value of quantitative D dimer assays. With highly sensitive D dimer testing kits, the frequency of underlying illness in children with PE would probably lead to high numbers of false-positive results. Coupled with the masking of symptoms and signs by that of the underlying diagnosis, the role of clinical probability scores and D dimer measurement remains unknown and requires prospective study in children with suspected PE. D dimer measurement at cessation of anticoagulant therapy as a means of predicting recurrence was not utilised in this patient cohort, although has shown potential value in adults with VTE (Kuruvilla et al, 2003).
The predominant diagnostic imaging modalities were V/Q scanning and CTPA. These are standard modalities for the diagnosis of PE in adults although there are no published studies of the sensitivity and specificity of these techniques for PE diagnosis in childhood. It was noted that seven of the nine infants were diagnosed by modalities other than V/Q and CTPA, namely echocardiography and cardiac angiography. In infants, V/Q scanning is limited by the subject’s inability to comply with the aerosol inhalation required for the ventilation phase of the study and CTPA by the poor sensitivity and specificity in detecting peripheral emboli due to the presence of small calibre vessels (Babyn et al, 2005). Due to the lack of ability to utilise conventional imaging modalities and a historically low level of clinical suspicion, one can speculate that PE may be under diagnosed in this age group. Further study is required in order to substantiate this claim and to develop a better imaging strategy for diagnosing PE in children of all ages.
Diagnosis was made at a median of 1 d from onset of symptoms. This is in contrast to the delay in diagnosis reported in a previous cohort of children with PE when average time to diagnosis was 7 d (Rajpurkar et al, 2007). The reason for this difference may reflect that the majority of patients reported in this cohort were inpatients at the time of onset of symptoms.
Complications and outcome of treatment in childhood PE
Major haemorrhage occurred frequently, in 21·4% of patients. Haemorrhage occurred early in treatment, and most often while receiving thrombolytic therapy and/or intravenous unfractionated heparin. The majority of these patients had underlying congenital heart disease, 25% had had recent surgery (within the previous 14 d) and >50% had coagulopathy in addition to the effects of anticoagulant therapy. Only one patient had laboratory evidence to suggest supra-therapeutic anticoagulation at the time of haemorrhage (Table III).
Overall mortality was high: 21·4% of patients died during the follow-up period. Death due to thromboembolism occurred in 8·9%, consistent with previously reported mortality rates of around 10% (Andrew et al, 1994; Van Ommen et al, 2001). The predominance of congenital heart disease and malignancy as underlying diagnoses in those patients who died reflect a poor prognosis group. Excluding these underlying diagnoses, the outcome from PE, in terms of resolution of thrombus and overall survival, was good. The number of deaths due to thromboembolism (8·9%) exceeded that of haemorrhagic complications (3·6%), supporting the role of anticoagulant therapy for PE in childhood.
Recurrence of childhood PE
Recurrent VTE occurred in 12·5% of the patients reported in this study. This outlines the importance of identifying patients in whom recurrence is more likely to occur and considering long-term anticoagulation on the basis of the individual’s risk-benefit profile. Those with persisting risk factors, such as malignancy, and those with intermittent risk factors, such as relapsing-remitting nephrotic syndrome or inflammatory bowel disease, are likely to require long-term anticoagulation or at least prophylaxis at times of relapse of their underlying disease. The presence of antiphospholipid antibodies in children with lupus has been linked to a VTE recurrence rate of 31% (Levy et al, 2003) and it was noted that one of our patients with recurrence had persistently positive anticardiolipin antibodies although they did not fit the criteria for lupus. In addition, a further patient had an elevated factor VIII level, also shown to increase risk of recurrent thrombotic events (Goldenberg et al, 2004). These individual features should be taken into account when deciding on an appropriate duration of anticoagulant therapy. Five of the seven patients with recurrence had recurrent PE, resulting in death in two patients, supporting data from adult studies showing that patients with PE as a first event have a higher risk of recurrent PE and death when compared to patients with unselected VTE (Douketis et al, 1998).