BRIEF REPORT
The Mortality Benefit Threshold for Patients with Suspected Pulmonary Embolism Jesse M. Pines, MD, MBA, MSCE, Adam L. Lessler, MD, MBA, Michael J. Ward, MD, MBA, and D. Mark Courtney, MD
Abstract Objectives: The mortality benefit for pulmonary embolism (PE) is the difference in mortality between treated and untreated patients. The mortality benefit threshold is the mortality benefit above which testing for a condition should be initiated and below which it should not. To illustrate this concept, the authors developed a decision model to estimate the mortality benefit threshold at several pretest probabilities for low-risk emergency department (ED) patients with possible PE and compare those thresholds with contemporary management of PE in the United States and what is known and not known about treatment benefits with anticoagulation. Methods: The authors built a decision model of a 25-year-old female with suspected PE. Model inputs were obtained from the literature or clinical judgment when data were unavailable. One-way sensitivity analysis was used to derive the mortality benefit threshold at several fixed pretest probabilities, and twoway sensitivity analysis was used to determine drivers of the mortality benefit threshold. Results: At a 15% pretest probability, the mortality benefit threshold was 3.7%; at 10% it was 5.2%; at 5% it was 9.8%; at 2% it was 23.5%; at 1% it was 46.3%; and at 0.5% it was 92.1%. In two-way sensitivity analyses, D-dimer specificity, CT angiography (CTA) ⁄ CT venography (CTV) sensitivity, annual cancer risk, probability of death from renal failure, and probability of major bleeding were major model drivers. Conclusions: The mortality benefit threshold for initiating PE testing is very high at low pretest probabilities of PE, which should be considered by clinicians in their diagnostic approach to PE in the ED. The mortality benefit threshold is a novel way of exploring the benefits and risks of ED-based testing, particularly in situations like PE where testing (i.e., CT use) carries real risks and the benefits of treatment are uncertain. ACADEMIC EMERGENCY MEDICINE 2012; 19:1109–1113 ª 2012 by the Society for Academic Emergency Medicine
From the Departments of Emergency Medicine and Health Policy, George Washington University (JMP), Washington, DC; the Department of Emergency Medicine, Hospital of the University of Pennsylvania (ALL), Philadelphia, PA; the Department of Emergency Medicine, University of Cincinnati (MJW), Cincinnati, OH; and the Department of Emergency Medicine, Northwestern University (DMC), Chicago, IL. Received January 16, 2012; revision received March 22, 2012; accepted April 13, 2012. Presented at the Society for Academic Emergency Medicine annual meeting, Phoenix, AZ, June 2010. The authors have no relevant financial information or potential conflicts of interest to disclose. Supervising Editor: Scott Wilber, MD. Address for correspondence and reprints: Jesse M. Pines, MD; e-mail:
[email protected].
ª 2012 by the Society for Academic Emergency Medicine doi: 10.1111/j.1553-2712.2012.01432.x
ISSN 1069-6563 PII ISSN 1069-6563583
1109
1110
Pines et al.
•
MORTALITY BENEFIT THRESHOLD FOR SUSPECTED PE
El Umbral del Beneficio en Mortalidad Para los Pacientes con Sospecha de Embolia de Pulmón Resumen Objetivos: El beneficio en mortalidad para el embolismo de pulmón (EP) es la diferencia en la mortalidad entre los pacientes tratados y los no tratados. El umbral del beneficio en mortalidad es el beneficio de mortalidad por encima del cual la prueba para una enfermedad debería ser iniciada, y por debajo del cual no se debería iniciar. Para ilustrar este concepto, se desarrolló un modelo de decisión para estimar el umbral del beneficio en mortalidad de varias probabilidades pretest para pacientes de bajo riesgo en el servicio de urgencias (SU) con posible EP, y comparar aquellos umbrales con el manejo actual del EP en Estados Unidos, y con lo que es y lo que no es conocido sobre los beneficios del tratamiento con anticoagulación. Métodos: Se construyó un modelo de decisión para una mujer de 25 años con sospecha de EP. Se obtuvo las entradas al modelo a partir de la literatura o del juicio clínico cuando los datos no estaban disponibles. Se usó un análisis de sensibilidad de una dirección para obtener el umbral del beneficio en mortalidad de varias probabilidades pretest fijadas y una análisis de sensibilidad de dos direcciones para determinar los conductores del límite de beneficio en mortalidad. Resultados: Con una probabilidad pretest del 15%, el umbral del beneficio en mortalidad fue del 3,7%; con una del 10%, fue del 5,2%; con una del 5%, fue del 9,8%; con una del 2%, fue del 23,5%; con una del 1%, fue del 46,3%; y con una del 0,5%, fue del 92,1%. En el análisis de sensibilidad de 2 vías, los principales conductores del modelo fueron la especificidad de los dímeros-D, la sensibilidad de la tomografía computarizada (TC), el riesgo de cáncer anual, la probabilidad de muerte por insuficiencia renal y la probabilidad de sangrado mayor. Conclusiones: El umbral del beneficio en mortalidad para iniciar las pruebas de EP es muy alta cuando las probabilidades pretest de EP son bajas, lo cual debería ser considerado por los clínicos en su abordaje diagnóstico de la EP en el SU. El umbral del beneficio en mortalidad es una manera novedosa para explorar los beneficios y riesgos de las pruebas de los SU, particularmente en situaciones como el EP donde las pruebas (ej: el uso de la TC) conllevan riesgos reales y los beneficios del tratamiento son inciertos.
C
onventional wisdom is that untreated pulmonary embolism (PE) is associated with complications including recurrent PE and mortality risks; however, studies reporting the benefits of anticoagulation for PE are dated and used small samples, and diagnoses were based on clinical, not radiographic grounds.1–3 Recent studies of PE patients not receiving anticoagulation report lower mortality and recurrence rates.4–6 The true risk of untreated PE today is unknown, which is concerning because of increasing PE-protocol computed tomography (CT) scan use and the technical possibility of diagnosing more small clots.7,8 CT also carries risks, including renal failure from contrast, allergic reactions, cancer from radiation, and false-positives.9,10 Newer data show that contrast-related renal failure rates are higher than expected.11 Balancing testing and treatment risks is central to decision-making; one approach is to calculate the pretest probability (i.e., disease prevalence) where risks of testing ⁄ treatment equal risks of not testing. This point of indifference is the testing threshold. Previously, we calculated the testing threshold for PE at 1.4%; however, a major driver was the expected mortality benefit from diagnosing PE.12 At that time, the latest data on contrast nephropathy risks were not included. Using testing thresholds is useful; however, we propose a new concept: the mortality benefit threshold. The mortality benefit is simply the difference in mortality
between untreated and treated PE. For example, if 20% of untreated patients die and 5% of treated patients die, the mortality benefit of testing ⁄ treatment is 15%. The mortality benefit threshold is the mortality benefit above which the test and treat strategy is preferred. In PE, a 15% mortality benefit threshold means that the mortality difference between treated and untreated PE should be 15% or greater to prefer the test ⁄ treat strategy. This can be calculated at specific pretest probabilities or prevalence of disease in a population. The mortality benefit threshold may be preferred to the testing threshold because it reframes the discussion around the expected benefits to the patient, and it may be easier to clinicians to understand than testing thresholds, which are often based on unstructured, highly variable risk estimates. We used decision analytic modeling to determine the mortality benefit threshold for PE at fixed pretest probabilities. We also determined how changes in assumptions alter our mortality benefit threshold estimates. METHODS Study Design We developed a decision model to estimate mortality benefit thresholds at low pretest probabilities for PE. Low-risk PE pretest probabilities were 0.5, 1, 2, 5, 10, and 15% for PE. We used decision analysis software
ACADEMIC EMERGENCY MEDICINE • September 2012, Vol. 19, No. 9
•
(TreeAge Pro 2011, TreeAge Software, Williamstown, MA) and took a societal perspective. No institutional review board approval was required because we did not involve human subjects or medical records.
automatically by the software, represented the mortality benefit threshold. Two-way sensitivity analysis was used to determine which assumptions drove the greatest effects on the mortality benefit threshold. Parameters that changed the threshold by ‡5% were major drivers, moderate drivers from 0.5% to
