DISCUSSION
To our knowledge, this is the first Portuguese, and one of the few European studies on PE in children, and one of the largest series on number of patients, time period and data analyzed. Furthermore, we describe three incidentally detected PE, a subgroup of patients whose management is particularly unclear, and about whom diagnostic methods, treatment and outcomes details have not been previously reported8. We also found a higher sensitivity for diagnosis of PE applying adults’ scores, compared to pediatric ones.
The number of PE admissions of 9.19/10.000 is similar to the one reported by Rajpurkar and colleagues (5.12-9.22/10.000 admissions). Although we didn´t observe an increase in the diagnosis rate along the 13 years studied, the incidence during 2020 was one of the highest. All four cases tested negative for SARS-COV-2 PCR at admission and none presented suggestive clinical symptoms, but serological tests were not performed.
We found an overall predominance of female patients, especially in the adolescent group, presumably due to the influence of CHC. The mean age at presentation was similar to that described in other reviews8, however, as others7,16, we didn´t find the classical bimodal age distribution described in pediatric PE4,10. PE is probably underdiagnosed, especially in infants, whose diagnosis is even more challenging due to the unspecific clinical presentation. The unexpected predominance of central emboli and massive or submassive PE further corroborates the underdiagnosis of less clinically obvious events, missing the opportunity to identify and treat patients with high thromboembolic risk.
The failure to consider PE as part of the differential diagnosis in pediatrics, even in the presence of typical complains, is a major concern20. An important clue is the presence of thromboembolic risk factors since it is unlikely for PE to occur in pediatric patients in their absence16. We found a higher prevalence of CHC utilization, immunomediated diseases and inherited thrombophilia in PE than previously reported8. Risk factors were different in outpatients versus inpatients. In outpatients PE occurred exclusively in adolescents, mainly female, and CHC intake, obesity, immunomediated diseases or a positive family history were the most important risk factors. Remarkably, all the female patients were on CHC, the majority for less than one year, a period in which the risk of VTE is higher21. Most, however, had other associated risk factors, and these should be considered when first prescribing CHC in teenagers. Besides the classic prothrombotic conditions, as lupus or antiphospholipid syndrome, recently VTE has been increasingly recognized in patients with several systemic immunomediated diseases, as found in our series22-24. In outpatients there was an important delay in diagnosis, although we found a shorter than the mean lag time described8. As reported, other diagnosis had been considered during the previous medical observation, namely infections or psychological distress, emphasizing that the presence of fever or anxiety should not exclude PE.
Inpatients with PE were younger than outpatients, more frequently male, and presented different risk factors, such as CVC, infections and complex chronic diseases. Infections are important triggers and may delay PE diagnosis due to symptoms overlap25. As reported, in long-term hospitalizations, especially in critically ill patients, immobilization could have contributed for PE, emphasizing the importance of mechanical methods for DVT prophylaxis in these patients23. A worsening of their clinical condition, particularly unexplained cardiorespiratory deterioration, must prompt a PE workout. We found three asymptomatic PE (Table 2) in inpatients that were being studied for other reasons. These patients had multiple risk factors and negative D-dimer testing, possibly corresponding to chronic or subacute PE.
Recently, two distinct patterns of pediatric PE have been acknowledged8: classic thromboembolic PE and in situ pulmonary artery thrombosis (ISPAT), which results from local causes, such as congenital heart disease or anomalies of the pulmonary artery26. Two patients from our study may fit into this category: an asymptomatic (table 2) and a 13-years old girl, with a Truncus arteriosus type 1, with a right ventricle-pulmonary artery conduct, hospitalized with endocarditis suspicion. Clinically, decrease in pulmonary sounds and worsening of tricuspid regurgitation prompted CTPA with right central PE diagnosis. She was successfully treated with thrombectomy and conduct substitution surgery and maintained anticoagulation for one year. During follow-up a May-Thurner Syndrome was also detected. Several relevant studies7,10 have included patients with possible septic or tumor thrombus in pediatric PE series, advocating that, in clinical practice, it is often not possible to differentiate septic emboli or tumor thrombus from true thromboembolic PE, as these are independent risk factors for VTE10. Furthermore, in the reported case, culture of the conduct and emboli were negative. Hence, we decided to include this patient in the study, even if this cannot be considered a classic thromboembolic PE, but rather an ISPAT.
ECG and CXR are often performed to investigate thoracic pain or dyspnea. However, the normality of these exams may be misleadingly reassuring, as happened in several of our outpatients. In highly suspicious cases, CTPA, the gold standard method to diagnose PE, should be directly performed. In our series, this happened only in one case, revealing the hesitancy to consider this diagnosis in pediatric patients. In lower-risk patients, other first-line exams can be considered. Although not sensitive, ECG may show unspecific alterations suggestive of PE27, as sinus tachycardia, an alarm sign present in most of our cases. D-dimer testing was reported to lack utility in the diagnosis of childhood PE14, but in a recent multicentric study28 revealed a 100% sensitivity. In our data, D-dimer were positive in all acute PE, and negative only in a subacute PE in an anticoagulated patient and an asymptomatic PE, also antiaggregated. As others, we didn´t find an association between the degree of D-dimer increase and PE severity29.
PE clinical diagnostic prediction models, validated for adult population, have been described as performing poorly in pediatric populations16. In a 36 pediatric PE population, Hennelly et al.15 found Wells criteria to have a sensitivity and specificity of 86% and 58%, and PERC criteria of 100% and a 24%, respectively. Agha et al found the PERC criteria unable to rule out PE in 84% of pediatric emergency room patients30. In contrast, in our study these criteria had a high sensitivity and false negatives were only found in asymptomatic PE. Two pediatric models have been proposed, both derived from single-center studies and including a small number of PE15,16. Hennelly et al. model15performed similarly to Wells criteria in a setting with 36 PE, missing to diagnose four cases of children with clinical findings suggestive of a DVT or a history of malignancy or immobilization. Its application in our PE patients misclassified four patients as non-risk, two of them asymptomatic. A further model developed by Lee et al.16 missed two of our PE patients, one asymptomatic. Both models misclassified a patient with a protein C deficiency, who had a positive family history. We propose that this should be considered in future pediatric models.
In our series, investigation after PE diagnosis was not uniform. May–Thurner Syndrome was diagnosed in a significant proportion of patients, years after PE, due to PTS. Although rare, this syndrome is likely underestimated and should be investigated in patients with unexplained left lower extremity thrombosis31.
A minority of our patients received thrombolysis and two of the massive PE were not eligible due to critical bleeding. Contrastingly, a systematic review8 reported that a third of pediatric PE patients receive pharmacologic thrombolysis as initial treatment, although recent guidelines reserve it for massive PE32. LMWH was used more often to initiate anticoagulation treatment, given its pharmacokinetics profile, while UFH was the first option in high bleeding risk patients33-34. Anticoagulant therapy duration in our study was in average longer than described8,13.
Management of asymptomatic PE lacks evidence, but most physicians claim to treat them similarly to symptomatic PE26. In a literature review8, up to 17% of patients had PE detected incidentally but details on diagnostic methods, treatment and outcomes of this type of PE in pediatric patients had not been reported.
Rajpurkar et al.8 referred a significant mortality rate in pediatric PE (26%), acknowledging, however, a possible reporting bias and that, in many studies, the cause of death might have been attributable to underlying diseases. Indeed, in our series, PE–related mortality 6.9%, while all causes mortality during follow-up reached 17.3%. Most deaths occurred in inpatients with cardiovascular, oncologic or other complex chronic conditions, as well as younger patients, as described in previous studies7,8,10.
Follow up practices after a pediatric PE event are highly variable and data on long-term outcomes, including CTPH, pulmonary function or quality of life, and predictors of outcome are sparse26. Recurrence of VTE was within the expected reported values (0–18.8%). PTS was an important cause of morbidity for several years after PE. Evaluation of residual thrombosis at the end of anticoagulation therapy is not consensual among experts, 60% admitting they routinely re-image although 35% would not change the duration of anticoagulation therapy based in the results.13 More than half our patients had a control CTPA, in a variable period after PE, and CTPH was indirectly accessed by ECHO.
Our study has important limitations: it is a single center retrospective study, and it does not necessarily generalize to other settings or communities. Importantly, autopsy-diagnosed PE were not trackable in a database and thus, reported mortality could be underestimated. Data on the autopsy-diagnosed PE case was not possible to analyze. The application of clinical scores retrospectively can be misleading. Also, due to the lack of a control group, we could not compare risk factors to those of the general population. Due to the 13-years span, and the lack of an institutional protocol, clinical management could have changed along the time, although this wasn´t noticed.
Our data emphasizes the need for a high suspicion index for pediatric PE in outpatients with risk factors, as half the cases were initially misdiagnosed. Towards compatible complains, risk factors, including CHC use, obesity, immunomediated diseases and family history should be explored. In inpatients, risk factors are often inherent to underlying diseases, immobilization, co-infections and CVC presence and thus unexplained cardiorespiratory deterioration should prompt PE investigation. May–Thurner Syndrome is likely underestimated and should be considered. The use of validated adult scores merits to be explored and improved in prospective pediatric multicentric studies.