The effects of cause of death classification on prognostic assessment of patients with pulmonary embolism

Transcription

1 Journal of Thrombosis and Haemostasis, 9: DOI: /j x ORIGINAL ARTICLE The effects of cause of death classification on prognostic assessment of patients with pulmonary embolism D. SÁNCHEZ,* J. DE MIGUEL, A. SAM,* C. WAGNER,* C. ZAMARRO,* R. NIETO,* L. GARCÍA,* D. AUJESKY,à R. D. YUSEN and D. JIMÉNEZ* *Respiratory Department, Ramón y Cajal Hospital and Instituto Ramón y Cajal de Investigación Sanitaria IRYCIS; Respiratory Department, Gregorio Marañón Hospital, Madrid, Spain; àdivision of General Internal Medicine, Bern University Hospital, Bern, Switzerland; Divisions of Pulmonary and Critical Care Medicine and General Medical Sciences, Washington University School of Medicine, St Louis, MO, USA; and Medicine Department, Alcalá de Henares University, Madrid, Spain To cite this article: Sánchez D, de Miguel J, Sam A, Wagner C, Zamarro C, Nieto R, García L, Aujesky D, Yusen RD, Jiménez D. The effects of cause of death classification on prognostic assessment of patients with pulmonary embolism. J Thromb Haemost 2011; 9: Summary. Background: Although previous studies have provided evidence that the majority of deaths following an acute pulmonary embolism (PE) directly relate to the PE, more recent registries and cohort studies suggest otherwise. Methods: We assessed the cause of death during the first 30 days after the diagnosis of acute symptomatic PE in a consecutive series of patients. We also assessed the prognostic characteristics of the simplified Pulmonary Embolism Severity Index (spesi) and cardiac troponin I (ctni) obtained at the time of PE diagnosis. Results: During the first 30 days after diagnosis, 127 of the 1291 patients died (9.8%; 95% confidence interval [CI], ). Sixty patients (4.6%; 95% CI, ) died from definite or possible PE, and 67 (5.2%; 95% CI, ) died from other causes (cancer 25, infection 18, hemorrhage 7, heart failure 7, chronic obstructive pulmonary disease 5, renal failure 1, seizures 1, unknown 3). The spesi predicted all-cause (odds ratio [OR], 5.97; 95% CI, ; P < 0.01) and PE-associated mortality (OR, 8.79; 95% CI, ; P = 0.04). ctni only predicted PE-associated mortality (adjusted OR, 2.39; 95% CI, ; P < 0.01). For all-cause mortality, the spesi low-risk strata had a negative predictive value of 98.8% (95% CI, ) in comparison with 91.3% (95% CI, ) for the ctni. Conclusions: Within the first 30 days after the diagnosis of acute symptomatic PE, death due to PE and death due to other causes occur in a similar proportion of patients. As ctni only predicted PE-associated mortality, lowrisk spesi had a higher negative predictive value for all-cause mortality compared with ctni. Correspondence: David Jiménez Castro, Respiratory Department and Medicine Department, Ramo n y Cajal Hospital and Alcalá de Henares University, Madrid, Spain. Tel.: ; fax: djc_69_98@yahoo.com Received 20 April 2011, accepted 10 August 2011 Keywords: mortality, prognosis, pulmonary embolism. Introduction Despite the advances in the diagnosis and management of venous thromboembolism (VTE), pulmonary embolism (PE) remains a major cause of morbidity and mortality [1]. Investigators have thus focused on accurate risk-stratification in such patients. A high-risk subgroup of hemodynamically stable patients with acute symptomatic PE might benefit from intensive care or thrombolytic treatment. Alternatively, patients at very low risk for short-term adverse events might undergo treatment entirely at home, and this may lead to increased patient satisfaction and decreased healthcare costs [2]. Ideally, risk scores would accurately identify both types of patients. Understanding the causes of death after acute PE helps with the assessment of the strengths of the various prognostic tools. All-cause mortality serves as an appropriate outcome when aiming to identify a low-risk group of patients that could safely undergo initial PE therapy at home. On the other hand, PEspecific mortality functions as an appropriate outcome when aiming to identify high-risk patients who might benefit from more aggressive PE-specific therapy. Although some previous studies have provided evidence of PE-associated right ventricular (RV) dysfunction as the most common cause of death during the first 30 days after diagnosis of PE, more recent registries and cohort studies suggest that many patients die of conditions not specifically related to the acute PE event [3 5]. In those studies in which PE most commonly caused death [6], cardiac biomarkers (i.e. troponin and brain natriuretic peptides [BNP]) and cardiac imaging testing (i.e. transthoracic echocardiography and computed tomography [CT] scan) had high predictive values for overall mortality. However, the biomarkers and imaging testing had poor predictive value in studies that had a lower proportion of deaths due to PE [7]. To ascertain the most common causes of death within the first 30 days after the diagnosis of acute symptomatic PE, we

2 2202 D. Sa nchez et al retrospectively analyzed data from a large consecutive series of patients that participated in a prospective registry. We also assessed the effects of the cause of death classification on the prognostic characteristics of the simplified PESI (spesi) and the cardiac troponin I (ctni). In particular, we concentrated on the ability of either prediction tool to identify accurately lowrisk patients with acute PE who might be candidates for treatment in the outpatient setting. Methods Study design For this retrospective cohort study, we used data from our ongoing registry that has prospectively gathered data from consecutively eligible patients with confirmed acute PE since 1 January Patients in this cohort underwent enrollment through to 30 June Patients, setting and eligibility criteria We screened outpatients presenting with symptoms of acute PE at the Emergency Department of Ramo n y Cajal Hospital, Madrid, Spain (Fig. 1). This study required patients to have the acute symptomatic PE confirmed by objective testing. The study did not exclude patients that received thrombolysis, inotropic support or inferior vena cava filter therapy. The Institutional Review Board approved the data collection. Diagnosis of PE Available troponin (n = 799 [61.9%]) Patients with clinical suspicion of PE (n = 6084) Elegible for study (n = 1291 [96.5%]) PE objectively confirmed (n = 1338 [22.0%]) Unavailable troponin (n = 492 [38.1%]) PE excluded (n = 4746 [78.0%]) Not elegible for study (n = 47 [3.5%]) Consent denied: 23 Unavailable for follow-up: 24 Fig. 1. Patient flow diagram. PE, pulmonary embolism; ctni, cardiac troponin I. For confirmation of the diagnosis of PE, we required: an intraluminal filling defect on PE-protocol contrast enhanced helical chest CT [8]; a high probability ventilation-perfusion scan according to the criteria of the Prospective Investigation of Pulmonary Embolism Diagnosis [9]; or an indeterminate ventilation-perfusion lung scan and confirmed lower limb deep vein thrombosis (DVT) on venous ultrasound [10]. Boardcertified radiologists interpreted the ventilation-perfusion scans and CT scans. Cardiac troponin I assay The study protocol strongly recommended troponin levels to be measured within 24 h of PE diagnosis. The hospital laboratory measured cardiac troponin I (ctni) levels quantitatively using a microparticle enzyme immunoassay (Abbot, Abbot Park, IL, USA). The analytical sensitivity for this ctni is 0.08 ng ml )1 and represents the lowest measurable concentration of ctni that can be distinguished from zero. With this assay, we regarded ctni concentrations of > 0.1 ng ml )1 as indicating myocardial injury (ctni positive). The study staff did not reveal the troponin levels to the clinicians throughout the hospital stay. Calculation of the prediction rule Using the prospectively collected baseline data at the time of PE diagnosis, we retrospectively determined the simplified Pulmonary Embolism Severity Index (spesi) and classified each patient into the low- or the high-risk category [11]. Anticoagulant therapy Clinicians initially treated patients with therapeutic doses of parenteral anticoagulants and the oral vitamin K antagonist acenocoumarol. After successful completion of the initial ÔoverlapÕ treatment period, the clinicians continued the patients on dose-adjusted oral vitamin K antagonist therapy (target INR of 2.5 [therapeutic range ]). The responsible attending physicians used thrombolysis and/or inotropic support in patients with hemodynamic instability as deemed appropriate. Study endpoints The study used all-cause mortality and mortality due to PE as the primary and secondary endpoints, respectively. The study assessed the endpoints through the first 30 days after diagnosis. Study investigators assessed mortality with patient or proxy interviews, and/or hospital chart review. Two investigators (D.S. and C. Z.) independently adjudicated the cause of death as [1] definite fatal PE, [2] possible fatal PE, or [3] death from other causes. For deaths confirmed by autopsy or those following a clinically severe PE, either initially or shortly after an objectively confirmed recurrent event, in the absence of any alternative diagnosis, the adjudicators judged death to be due to definite fatal PE. Severe PE included arterial hypotension, cardiogenic shock, circulatory collapse, or echocardiographic findings signifying acute right heart failure at presentation. Possible fatal PE consisted of death in a patient who died suddenly or unexpectedly. In cases of disagreement between the reviewers, a third reviewer (D.J.) established the final cause of death by consensus with the other reviewers.

3 Cause of death in PE 2203 Statistical analysis We used descriptive statistics and calculated 95% confidence intervals (CI) from the binomial distribution. For nonnormally distributed (Kolmogorov Smirnov test) continuous data, we used the median and 25th and 75th percentiles for descriptive statistics and the Mann Whitney U-test for group comparisons. For categorical data, we described baseline characteristics with counts and proportions and used chi-squared or FisherÕs exact tests for group comparisons. We used the kappa statistic to test for agreement beyond chance among the raters for adjudication on the cause of death. We used logistic regression to assess if the variables of ctni (i.e. elevated and not elevated) and spesi score (i.e. low-risk group and high-risk group) had an independent association with 30-day all-cause and PE-related mortality (i.e. either definite or possible PE), after adjusting for other variables that had a P value of < 0.05 in the univariate analyses. We used a manual backward stepwise approach to develop the multivariate models. For construction of the full models, we considered variables with imbalance between the groups at baseline for inclusion. During model construction, we did not remove variables that showed evidence of confounding (i.e. the coefficient of the variable changed by more than 10% when removed from the full model). Because a proportion of the patients had missing troponin (38%) data, we performed a sensitivity analysis that used imputed data. In order to evaluate whether troponin testing influenced the study results, we imputed this variableõs missing data by using the multiple imputations by chained equation method, which resulted in 10 imputed datasets. We independently analyzed each of the 10 datasets. We averaged estimates of the variables to give a single mean estimate according to the Rubin rules [12]. To assess the test and performance characteristics of the ctni-positive vs. ctni-negative and spesi high-risk vs. spesi low-risk groups, we estimated sensitivity, specificity and positive and negative predictive values and likelihood ratios for predicting all-cause and PE-specific mortality. For all analyses, we used a two-sided P < 0.05 to indicate statistical significance. We used Statistical Package for Social Sciences (SPSS, release 15.0, Chicago, IL, USA) for analyses. Results Baseline characteristics The eligible 1291 patients (Fig. 1; Table 1), 579 men and 712 women, had a mean age of 69 ± 16 years. A previous study that developed the spesi clinical prediction rule utilized 995 (77%) of these patients [10]. Seven hundred and ninety-nine patients had ctni levels measured, and 492 did not. Compared with patients who did not have ctni levels measured, patients that had ctni levels measured were younger in age and less likely to have an active malignancy, Table 1 Clinical symptoms and relevant findings at presentation in 1291 consecutive patients with acute symptomatic pulmonary embolism Characteristic Patients, n (%) Demographic factors Median (25th to 75th percentile) age (years) 74 (61 80) Age > 65 years 871 (67) Male sex 579 (45) Comorbid illness Cancer* 301 (23) Heart failure 85 (6.6) Chronic lung disease 101 (7.8) Previous VTE 128 (9.9) Immobilization 262 (20) Previous surgery à 127 (9.8) Clinical findings Pulse 110 beats min )1 246 (19) Systolic blood pressure < 100 mmhg 109 (8.4) Respiratory rate 30 breaths min )1 90 (7.0) Temperature < 36ºC 141 (11) Altered mental status 14 (1.1) Arterial oxyhemoglobin saturation (SaO 2 ) < 90% 320 (24.8) Simplified PESI (10) Low-risk strata 407 (32) High-risk strata 884 (68) Laboratory findings Presence of DVT detected by CCUS 571 (44) Cardiac troponin I > 0.1 ng ml )1 247 (31) CCUS, complete lower limb ultrasound testing; DVT, deep vein thrombosis; PESI, Pulmonary Embolism Severity Index; SD, standard deviation; VTE, venous thromboembolism. *Active or under treatment in the last year. Immobilized patients are defined in this analysis as non-surgical patients who had been immobilized (i.e. total bed rest with bathroom privileges) for 4 days in the month prior to PE diagnosis. à In the 2 months prior to PE. Defined as confusion, disorientation or somnolence. In the subgroup of 799 patients who had ctni tested. concomitant DVT (assessed by lower limb ultrasound testing) and hypotension. They also had worse arterial oxygen saturation values, higher heart rates and higher respiratory rates. About one-third (247 of 799; 30.9%; 95% CI, ) of patients had elevated serum ctni levels (ctni-positive group), while approximately two-thirds (552 of 799; 69.1%; 95% CI, ) of patients had normal serum ctni levels (ctni-negative group). All patients had spesi scores calculated. Although we retrospectively calculated the spesi scores for this study, study investigators prospectively collected all data to calculate them at the time of presentation. About one-third (407 of 1291; 31.5%; 95% CI, ) of patients had low-risk spesi scores, while approximately two-thirds (884 of 1291; 68.5%; 95% CI, ) had high-risk spesi scores. About 3% of patients received thrombolytic therapy (44 of 1291; 3.4%; 95% CI, ), and approximately 7% (three of 44 patients; 6.8%; 95% CI, ) of these patients died. About 10% (124 of 1247 patients; 9.9%; 95% CI, ) of those who did not receive thrombolytic therapy died (absolute difference, 3.1%; 95% CI of the absolute difference, )4.5% to 10.7%; P = 0.68).

4 2204 D. Sa nchez et al Classification and time dependence of death risk All patients had the primary and secondary outcomes assessed. During the first 30 days after diagnosis, 127 of 1291 patients died (9.8%; 95% CI, ). Sixty of the 127 patients (47.2%; 95% CI, ) died from definite (n = 16) or possible PE (n = 44), and 67 of the 127 patients (52.8%; 95% CI, ) died from other causes (cancer 25, infection 18, hemorrhage 7, heart failure 7, chronic obstructive pulmonary disease 5, renal failure 1, seizures 1, unknown 3) (Fig. 2). The adjudicators of the cause of death had very good agreement (j = 0.84). One hundred and two of the 1291 patients (7.9%; 95% CI, ) died during their hospital stay. Half of these patients (51 of 102; 50.0%; 95% CI, ) died from definite or possible PE, and the other half died from other causes (cancer 17, infection 16, hemorrhage 5, heart failure 4, chronic obstructive pulmonary disease 4, renal failure 1, seizures 1, unknown 3). During the first 7 days of treatment, 69 of the 1291 patients died (5.3%; 95% CI, ). Thirtyeight of the 69 patients (55.1%; 95% CI, ) died from definite or possible PE, and 31 of the 69 (44.9%; 95% CI, ) died from other causes (cancer 10, infection 5, hemorrhage 5, heart failure 4, chronic obstructive pulmonary disease 2, renal failure 1, seizures 1, unknown 3). Table 2 shows the clinical characteristics of the patients who died from definite or possible PE or non-pe-related causes. Compared with patients with a non-pe-related death, those with a definite or possible PE-related death had a higher frequency of signs of clinical severity (tachycardia, hypotension and hypoxemia) and elevated troponin levels and a lower frequency of a diagnosis of cancer at presentation. Ability of spesi and ctni to identify low-risk patients The 31.5% (407/1291) of patients classified as low risk by the spesi had a 30-day mortality of 1.7% (95% CI, ), compared with the 13.6% (95% CI, ) in the high-risk group. The 69.1% (552/799) of patients classified as low risk by the ctni had a 30-day mortality of 8.7% (95% CI, ), compared with 11.3% (95% CI, ) in the high-risk group. For the 799 patients that had ctni levels measured, multivariate logistic regression analysis showed an independent association between severity of illness assessed by spesi (OR, 5.97; 95% CI, , P < 0.01) and the primary outcome of 30-day all-cause mortality, after adjusting for the other significant baseline covariates of age, cancer, systolic blood pressure and arterial oxyhemoglobin saturation (Table 3). Regression did not detect an association between elevated ctnl and all-cause mortality. Multivariate analysis also showed an independent association between severity of illness using the spesi (8.79; 95% CI, ; P = 0.04) and the secondary outcome of 30-day PE-specific mortality, after adjusting for the other significant baseline covariates of age, systolic blood pressure, heart rate and arterial oxyhemoglobin saturation at the time of acute PE diagnosis. Multivariate logistic regression Table 2 Clinical characteristics of 127 patients who died Characteristic PE-related death (n = 60) Non-PE-related death (n = 67) P Demographic factors Median (25th to 75th percentile) age (years) 80 (69 84) 76 (63 83) 0.46 Age > 65 years 49 (82%) 48 (72%) 0.18 Male sex 23 (38%) 33 (49%) 0.22 Comorbid illness Cancer* 22 (37%) 35 (52%) 0.08 Heart failure 6 (10%) 10 (15%) 0.40 Chronic lung disease 6 (10%) 10 (15%) 0.40 Previous VTE 5 (8.3%) 2 (3.0%) 0.19 Immobilization 24 (40%) 18 (27%) 0.12 Previous surgery à 3 (5.0%) 4 (6.0%) 0.81 Clinical findings Pulse 110 beats min )1 21 (35%) 13 (19%) 0.05 Systolic blood pressure < 100 mmhg 16 (27%) 7 (10%) 0.02 Respiratory rate 30 breaths min )1 4 (6.7%) 6 (8.9%) 0.63 Temperature < 36 C 9 (15%) 7 (10%) 0.44 Altered mental status Arterial oxyhemoglobin saturation (SaO 2 ) < 90% 28 (47%) 20 (30%) 0.05 Simplified PESI (10) Low-risk strata 2 (3.3%) 5 (7.5%) 0.31 Laboratory findings Presence of DVT detected by CCUS 26 (43%) 32 (48%) 0.19 Cardiac troponin I > 0.1 ng ml )1 22 (37%) 6 (8.9%) CCUS, complete lower limb ultrasound testing; DVT, deep vein thrombosis; PESI, Pulmonary Embolism Severity Index; SD, standard deviation; VTE, venous thromboembolism. *Active or under treatment in the last year. Immobilized patients are defined in this analysis as non-surgical patients who had been immobilized (i.e. total bed rest with bathroom privileges) for 4 days in the month prior to PE diagnosis. à In the 2 months prior to PE. Defined as confusion, disorientation or somnolence. In the subgroup of 799 patients who had ctni tested.

5 Cause of death in PE 2205 Table 3 Unadjusted and adjusted odds ratios for all-cause mortality after study entry with acute symptomatic pulmonary embolism Risk factor Unadjusted OR (95% CI) P value Adjusted OR (95% CI) P value Age, per year 1.05 ( ) < Male gender 0.80 ( ) 0.37 Cancer* 2.37 ( ) < 0.01 Immobilization 2.40 ( ) < 0.01 SBP < 100 mmhg 3.03 ( ) < Arterial oxyhemoglobin saturation < 90% 1.91 ( ) < 0.01 Pulse 110 beats min ) ( ) 0.13 Simplified PESI high-risk strata ( ) < ( ) < ctni > 0.1 ng ml ) ( ) ( ) 0.48 à CI, confidence interval; ctni, cardiac troponin I; OR, odds ratio; PESI, Pulmonary Embolism Severity Index; SBP, systolic blood pressure. *Active or under treatment in the last year. Immobilized patients are defined in this analysis as non-surgical patients who had been immobilized (i.e. total bed rest with bathroom privileges) for 4 days in the month prior to PE diagnosis. à Adjusted by systolic blood pressure. confirmed that elevated ctni (OR 2.39; 95% CI, ; P < 0.01) had an independent association with PE-specific mortality, after adjusting for the other significant baseline covariate of systolic blood pressure. After imputation of missing data, logistic regression confirmed that elevated ctni (OR, 1.82; 95% CI, ) had an independent association with PE-specific mortality, though it was not associated with all-cause mortality (OR, 1.2; 95% CI, ). In the cohort of patients that had ctni values measured (n = 799), the spesi low-risk strata classification had a negative predictive value for 30-day all-cause mortality of 98.8% (95% CI, ), in comparison to a negative predictive value of 91.3% (95% CI, ) for a negative ctni (< 0.1 ng ml )1 ). For 30-day PE-related mortality in the same cohort, the spesi low-risk strata classification had a negative predictive value of 98.8% (95% CI, ), and ctni had a negative predictive value of 96.6% (95% CI, ). For the outcome of 30-day definite fatal PE, the spesi low-risk strata had a negative predictive value of 100%, and a negative ctni had a negative predictive value of 99.3%. Table 4 shows negative predictive value and other test characteristics of ctni and spesi. Discussion PE remains one of the leading causes of cardiovascular morbidity and mortality [13 15]. A better understanding of the etiology of death soon after diagnosis of acute symptomatic PE appears critical for evaluation of prognostic tools [16]. This study showed that approximately one out of 10 patients with acute symptomatic PE that presented to an emergency department died during the first 30 days of follow-up. Definite or possible PE caused approximately half of the deaths, while other etiologies caused the other half. A high-risk spesi independently predicted 30-day all-cause and PE-specific mortality, while an elevated ctni only independently predicted PE-specific mortality. The etiology of deaths influenced the prognostic performance of the cardiac biomarker ctni, though it did not significantly affect the performance of the clinical model spesi. For the prediction of patients at low risk of Table 4. Prediction rule test characteristics in patients with acute symptomatic PE that underwent ctni testing (n = 799) Simplified PESI parameter (95% CI) ctni parameter (95% CI) All-cause mortality Sensitivity, % 96.1 ( ) 36.8 ( ) Specificity, % 33.9 ( ) 69.6 ( ) Positive 13.4 ( ) 11.3 ( ) Negative 98.8 ( ) 91.3 ( ) Positive likelihood ratio 1.45 ( ) 1.21 ( ) Negative likelihood ratio 0.11 ( ) 0.91 ( ) PE-related mortality* Sensitivity, % 97.6 ( ) 53.7 ( ) Specificity, % 32.6 ( ) 70.2 ( ) Positive 7.4 ( ) 8.9 ( ) Negative 99.6 ( ) 96.5 ( ) Positive 1.45 ( ) 1.80 ( ) likelihood ratio Negative likelihood ratio 0.07 ( ) 0.66 ( ) CI, confidence interval; ctni, cardiac troponin I; PE, pulmonary embolism; PESI, Pulmonary Embolism Severity Index. *Definite or possible PE. death, a low-risk spesi had a higher negative predictive value for 30-day all-cause and PE-specific mortality in comparison to a non-elevated ctni. The outcomes assessed by prognostic tools for patients with acute PE should have a relationship to the therapeutic options. For example, prognostic tools should ideally predict all-cause mortality, recurrent venous thromboembolism (proximal DVT or PE) and major bleeding soon (e.g. in the first 7 14 days) after diagnosis in patients being considered for outpatient therapy for PE. Prognostic tools should focus on the prediction of PE-related death and non-fatal PE-related complicated course (e.g. hemodynamic collapse or recurrent PE) in patients being considered for escalated disease-specific therapy such as thrombolysis. Thus, studies that assess the usefulness of

6 2206 D. Sa nchez et al Observed proportions of death by PE or by other causes Non PE-related death PE-related death Time (days) Fig. 2. Observed proportions of death due to definite or suspected PE or other causes among 1291 patients with acute symptomatic PE. different tools for risk stratification of patients with acute PE should describe the proportion of deaths that have an association with PE. Studies have shown conflicting data regarding the incidence of death and the proportion of deaths associated with PE during short-term follow-up after the diagnosis of PE [6,17,18]. In the Management and Prognosis in Pulmonary Embolism Trial (MAPPET), 69 of 719 (9.6%; 95% CI, ) patients with acute PE died during the hospital stay, and PE caused the majority (94.2%) of deaths [6]. However, the study only included patients with acute right heart failure or pulmonary hypertension due to PE, or both. In contrast, a larger study of 2442 patients with acute PE in the International Cooperative Pulmonary Embolism Registry (ICOPER) ascribed 45.1% of deaths to PE during the 3-month follow-up period [17]. Data from the Registro Informatizado de la Enfermedad TromboEmbo lica (RIETE) registry further supported the ICOPER findings [18]. Of 6599 patients in RIETE with acute PE, 417 (6.3%; 95% CI, ) died within 30 days; 153 (36.7%) died of the initial PE, 29 (6.9%) of recurrent PE, and 235 (56.4%) of other diseases. Our studyõs large sample size adds strength to the body of evidence regarding the pathogenesis of death soon after diagnosis of acute symptomatic PE. Interestingly, the proportion of PE-related deaths in our study remained around 50% for periods of follow-up < 30 days (e.g. 7 days or inhospital stay). Some meta-analyses and registries support the notion that cardiac troponins may predict adverse outcomes [19]. However, conflicting conclusions about the prognostic ability of such markers from a recently published meta-analysis [20] and a large cohort study [7] have prompted further debate. Data from these studies in conjunction with our findings may explain the variable prognostic findings of markers of myocardial injury. ctni appeared to have better predictive ability in studies that had higher rates of PE-related deaths, while the spesi predicted both all-cause and PE-related mortality. We acknowledge that a prediction tool that predicts PE-specific mortality rather than all-cause mortality may appear more relevant to clinicians. A prognostic tool that attempts to identify high-risk patients who may benefit from specific treatments, such as thrombolysis, should accurately predict PE-specific mortality. Because the spesi aims to identify eligible low-risk patients with acute symptomatic PE for outpatient care [11,21], the use of overall rather than PEspecific mortality as the primary outcome provides advantage in this setting. Some study methodology limitations affect the findings and interpretation of this study. This single-center study of patients that presented to a tertiary care urban emergency department may not generalize to other settings. Regarding the accuracy of the cause of death determination, only a small percentage of the patients who died (eight of 127, 6%) had an autopsy, and the autopsies did not follow a protocol related to this study. As the negative predictive value for a nonelevated ctni increased when we excluded non-pe-related deaths and possible PE-related deaths from the denominator, these findings support that the insufficient negative predictive value is driven by non-pe-related deaths. Furthermore, misclassification of cause of death would probably have resulted in small changes in the negative predictive value of ctni for PE-specific mortality, and it would not be likely to affect our key finding of the insufficient ability of ctni to identify accurately patients with acute PE at low risk for allcause mortality who might be candidates for treatment in the outpatient setting. Regarding missing test result data, the analyses included only patients who had ctni testing performed around the time of the PE diagnosis. Comparison of data in those with and those without ctni testing suggested that selection bias occurred. However, we used imputation for missing data to further address concerns of bias and to assess the robustness of the study findings. The fact that this method yielded similar conclusions further strengthened the soundness of the results. In addition, this study used a conventional troponin assay. Recent data suggest that the use of a highly sensitive troponin T (hstnt) assay may improve risk stratification of PE [22]. Finally, the patients used to derive the spesi study composed a large proportion of this study cohort (77%), and this made it more likely that this study would further validate the spesi. Though the study did not aim to validate the spesi and the ctni, our results demonstrated the suboptimal performance of the ctni in identifying low-risk patients with acute symptomatic PE, and it showed the superiority of the spesi in comparison to ctni for this endeavor. In conclusion, a significant proportion (one out of 10) of patients with acute symptomatic PE presenting to an emergency department died during the first 30 days of follow-up. Definite or possible PE caused approximately half of the deaths, while other etiologies caused the other half. The insufficient negative predictive value of ctni for all-cause mortality underscores the need for clinical prognostic models (i.e. spesi) for decision-making regarding outpatient therapy.

7 Cause of death in PE 2207 Addendum Study concept and design: D. Sa nchez, J. de Miguel, D. Aujesky, R. D. Yusen, D. Jime nez. Acquisition of data, analysis and interpretation of data and statistical analysis: D. Sa nchez,a.sam,c.wagner,c.zamarro,r.nieto,l.garcı a, R. D. Yusen, D. Jiménez. Drafting of the manuscript: J. de Miguel, D. Aujesky,R. D. Yusen, D. Jime nez. Critical revision of the manuscript for important intellectual content: D. Sa nchez, J. de Miguel, A. Sam, C. Wagner, C. Zamarro, R. Nieto, L. Garcı a, D. Aujesky, R. D. Yusen, D. Jiménez. Study supervision: D. Jiménez, R. D. Yusen. The corresponding author, D. Jime nez, had full access to all the data in the study and had final responsibility for the decision to submit for publication. Acknowledgements This work has been supported in part by grants FIS 08/0200 and SEPAR This work has been supported in part by an unrestricted grant by ROVI. Disclosure of Conflict of Interests The authors state that they have no conflict of interest. References 1 White RH. The epidemiology of venous thromboembolism. Circulation 2003; 107: I Aujesky D, Smith KJ, Cornuz J, Roberts MS. Cost-effectiveness of low-molecular-weight heparin for treatment of pulmonary embolism. Chest 2005; 128: Laporte S, Mismetti P, De cousus H, Uresandi F, Otero R, Lobo JL, Monreal MRI. Clinical predictors for fatal pulmonary embolism in 15,520 patients with venous thromboembolism: findings from the RIETE registry. Circulation 2008; 117: Sanchez O, Trinquart L, Caille V, Couturaud F, Pacouret G, Meneveau N, Verschuren F, Roy PM, Parent F, Righini M, Perrier A, Lorut C, Tardy B, Benoit MO, Chatellier G, Meyer G. Prognostic factors for pulmonary embolism: the prep study, a prospective multicenter cohort study. Am J Respir Crit Care Med 2010; 181: Nijkeuter MSM, Tick LW, Kamphuisen PW, Kramer MH, Laterveer L, van Houten AA, Kruip MJ, Leebeek FW, Bu ller HR, Huisman MV, Christopher SI. The natural course of hemodynamically stable pulmonary embolism: clinical outcome and risk factors in a large prospective cohort study. Chest 2007; 131: Kasper W, Konstantinides S, Geibel A, Olschewski M, Heinrich F, Grosser KD, Rauber K, Iversen S, Redecker M, Kienast J. Management strategies and determinants of outcome in acute major pulmonary embolism: results of a multicenter registry. J Am Coll Cardiol 1997; 30: Jime nez D, Dı az G, Molina J, Martí D, Del Rey J, Garcı a-rull S, Escobar C, Vidal R, Sueiro A, Yusen RD. Troponin I and risk stratification of patients with acute nonmassive pulmonary embolism. Eur Respir J 2008; 31: Remy-Jardin M, Remy J, Wattinne L, Giraud F. Central pulmonary thromboembolism: diagnosis with spiral volumetric CT with the single-breath-hold-technique-comparison with pulmonary angiography. Radiology 1992; 185: PIOPED investigators. Value of ventilation/perfusion scan in acute pulmonary embolism: results of the prospective investigation of the pulmonary embolism diagnosis (PIOPED). JAMA 1990; 263: Kearon C, Ginsberg JS, Hirsh J. The role of venous ultrasonography in the diagnosis of suspected deep venous thrombosis and pulmonary embolism. Ann Intern Med 1998; 129: Jime nez D, Aujesky D, Moores L, Go mez V, Lobo JL, Uresandi F, Otero R, Monreal M, Muriel A, Yusen RD, for the RIETE investigators. Simplification of the Pulmonary Embolism Severity Index for prognosticating patients with acute symptomatic pulmonary embolism. Arch Intern Med 2010; 170: Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med 1991; 10: Tapson VF. Acute pulmonary embolism. NEnglJMed2008; 358: Konstantinides S. Clinical practice. Acute pulmonary embolism. N Engl J Med 2008; 359: Kearon C, Kahn SR, Agnelli G, Goldhaber S, Raskob GE, Comerota AJ; American College of Chest Physicians. Antithrombotic therapy for venous thromboembolic disease: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th edition). Chest 2008; 133: 454S 545S. 16 Jime nez D, Yusen RD, Otero R, Uresandi F, Nauffal D, Laserna E, Conget F, Oribe M, Cabezudo MA, Díaz G. Prognostic models for selecting patients with acute pulmonary embolism for initial outpatient therapy. Chest 2007; 132: Goldhaber SZ, Visani L, De Rosa M. Acute pulmonary embolism: clinical outcomes in the International Cooperative Pulmonary Embolism Registry (ICOPER). Lancet 1999; 353: Otero R, Trujillo-Santos J, Cayuela A, Rodriguez C, Barron M, Martin JJ, Monreal M, and the RIETE investigators. Haemodynamically unstable pulmonary embolism in the RIETE registry. Systolic blood pressure or shock index? Eur Respir J 2007; 30: Becattini C, Vedovati MC, Agnelli G. Prognostic value of troponins in acute pulmonary embolism. Circulation 2007; 116: Jime nez D, Uresandi F, Otero R, Lobo JL, Monreal M, Martı D, Zamora J, Muriel A, Aujesky D, Yusen RD. Troponin-based risk stratification of patients with nonmassive pulmonary embolism: systematic review and metaanalysis. Chest 2009; 136: Sam A, Sanchez D, Gomez V, Kopecna D, Zamarro C, Moores L, Aujesky D, Yusen RD, Jimenez D. Usefulness of the Shock Index and the Simplified Pulmonary Embolism Severity Index for identification of low-risk patients with acute symptomatic pulmoanry embolism. Eur Respir J 2011; 37: Lankeit M, Friesen D, Aschoff J, Dellas C, Hasenfuss G, Katus H, Konstantinides S, Giannitsis E. Highly sensitive troponin T assay in normotensive patients with acute pulmonary embolism. Eur Heart J 2010; 31: