Methods

We performed a case-control study of 206 consecutive patients (mean age 70.3 ± 15.6 years, 41.3% men) with an “International Diseases Classification, 10th revision,” discharge code of PE from January 2007 to December 2010. A clinical review of each case was performed to ensure that the “International Diseases Classification, 10th revision,” coding referred to acute PE. Baseline data were collected at the PE diagnosis, and the prognostic scores (Geneva prognostic score, sPESI, ESC, Shock Index, and GRACE) were assessed retrospectively. The primary outcome measure was all-cause 30-day mortality assessed through hospital medical records revision or proxy interviews. No patient was lost to follow-up (30 days from presentation). The human subjects committee of Coimbra’s Hospital Centre and University approved the study protocol.

All patients were recruited from the emergency department of a tertiary care hospital, and all patients provided informed consent. Eligibility for the present study required that patients had acute symptomatic PE confirmed by multidetector computed tomography. Patients with PE during hospitalization for other clinical reasons were not included.

The creatinine clearance was estimated using the Modification of Diet in Renal Disease equation: glomerular filtration rate (ml/min/1.73 m ² ) = 186 × (plasma creatinine) − 1.154 × (age) − 0.203 × (0.742 if female gender) × (1.210 if Afro-Caribbean). Troponin I was considered elevated if >0.034 ng/ml. Right ventricular failure was defined as an increase in the right ventricular diameter and/or the occurrence of right ventricular pressure overload in the echocardiographic assessment or N-terminal pro-B-type natriuretic peptide >4,500 pg/ml.

The patients were risk classified according to previously reported cutoff points. Regarding the Geneva prognostic score, the patients were classified as low (score ≤2) or high (score >2) risk. The sPESI dichotomized subjects into low (score <1) and high (score ≥1) risk. The ESC prediction rule stratified patients as low risk if no clinical or right ventricular dysfunction/myocardial injury markers were present, intermediate if either right ventricular dysfunction and/or myocardial injury markers were present, and high risk if shock or persistent hypotension were documented. The Shock Index, defined as the heart rate divided by the systolic blood pressure, divided patients into low risk (score <1) and high risk (score ≥1). For the GRACE score, we distributed patients into 3 risk strata using the criteria recommended for acute coronary syndromes (low risk, score <109; intermediate risk, score 109 to 140; and high risk, score >140) using the cutoff points of the receiver operating characteristic curve that ensured the most favorable compromise between sensitivity and specificity. This score was termed the “PE risk score.”

Statistical analysis was performed using SPSS, version 17.0 (SPSS, Chicago, Illinois) for the comparison of nominal variables with the chi-square test and continual variables with the Student’s t test. Univariate analysis was performed with the chi-square test, determining the respective relative risk with the 95% confidence interval (CI). Results with p <0.05 were regarded as statistically significant. We assessed the overall discriminatory power of each risk score (Geneva prognostic score, sPESI, ESC, Shock Index, and GRACE) by calculating the area under each receiver operating characteristic curve. Discrimination, measured in terms of the area under the curve, refers to a model’s ability to assign a greater probability to nonsurvivors than to survivors. However, it has been widely reported that the area under the curve cannot always measure a clinically meaningful quantity or demonstrate the value of new markers or models useful for prediction. Recently, other measures of incremental value have been proposed that examine the extent to which a new marker or model reclassifies subjects, such as the net reclassification improvement index and the integrated discrimination improvement. The net reclassification improvement method, described by Pencina and D’Agostino, states that a positive and significant net reclassification improvement translates a net overall successful reclassification of subjects into a more appropriate risk category. The integrated discrimination improvement, which can be seen as a continuous form of the net reclassification improvement method, assesses the improvement in risk discrimination by estimating the change in the difference of the average of the predicted probabilities of an event between those with and without the event under consideration. It is a more appropriate measure of risk reclassification when comparing scores with considerably different risk categorization (e.g., sPESI stratifies patients into 2 risk strata and GRACE into 3 risk strata). The GRACE score was also tested for calibration. Calibration refers to the agreement between the predicted and true probabilities and is most often measured using Hosmer-Lemeshow goodness-of-fit statistics, suggesting a good fit when associated with p >0.05.

Comparisons using receiver operating characteristics curve analysis and a measure of risk reclassification (the integrated discrimination improvement index) were performed between the GRACE model and the currently available PE risk scores (Geneva prognostic score, Shock Index, sPESI, and ESC). Furthermore, GRACE goodness of fit was assessed with the modified Hosmer-Lemeshow chi-square statistic.

Results

The patients’ baseline clinical and analytical and imaging characteristics are listed in Table 1 . In our population sample, 39 patients (18.9%) died within 30 days of presentation, with 33 corresponding to in-hospital mortality. No patient was lost to follow-up.

Hemodynamic shock was documented in 15 patients (7.3%), and no cases of cardiac arrest at admission were reported. Regarding specific medical interventions, thrombolysis was used in 19 patients (9.2%), mechanical ventilation in 7 (3.4%), and inotropic support in 14 (6.8%). Major hemorrhagic complications after thrombolysis were seen in 4 patients. In 2 of those patients, bleeding was considered related to the cause of death. A group comparison according to the occurrence of all-cause 30-day mortality is listed in Table 1 , and the predictors of death on univariate analysis (categorical variables) are listed in Table 2 .

Using the previously defined criteria and the previously described scores, we risk stratified our cohort ( Table 3 ). The Geneva prognostic score and Shock Index were the scores with a greater proportion of patients in the low-risk group (score ≤ 2; Geneva prognostic score, 161 of 206 patients [78.2%]; Shock Index, 153 of 206 patients [74.3%]) compared to the sPESI (score <1; 46 of 206 patients [22.3%]) and ESC (low-risk class; 54 of 206 patients [26.2%]). The primary end point was reached in the Geneva prognostic score low-risk group (69.2% of deaths; 27 of 39 deaths), in Shock Index low-risk group (56.4% of deaths; 22 of 39 deaths), and in the sPESI and ESC low-risk groups (10.3%, 4 of 39 deaths, and 7.7%, 3 of 39 deaths, respectively).

The receiver operating characteristic curve and the area under the curve for the 4 algorithms for all-cause 30-day mortality prediction are shown in the Figure 1 and listed in Table 4 . The receiver operating characteristic curve for the Geneva prognostic score (area under the curve 0.623, 95% CI 0.53 to 0.71; p = 0.016), receiver operating characteristic curve for Shock Index (area under the curve 0.639, 95% CI 0.55 to 0.73; p = 0.008), receiver operating characteristic curve for sPESI (area under the curve 0.705, 95% CI 0.61 to 0.80; p <0.001), and receiver operating characteristic curve for ESC (area under the curve 0.662, 95% CI 0.57 to 0.76; p = 0.002). No statistically significant differences were found between the area under the curve for the 4 scores (p >0.05; Table 4 ). The performance of all risk scores in our cohort is listed in Table 3 . We used the stratification criteria proposed by the original investigators: Geneva prognostic score >2, Shock Index ≥1, sPESI ≥1, and ESC (intermediate or more). ESC’s prediction rule had greater sensitivity and a greater negative predictive value than the remaining PE scores for predicting all-cause 30-day mortality.

PE scores and GRACE receiver operating characteristic curves. Gray line indicates ESC; blue line, Geneva risk score; red line, GRACE risk score; yellow line, Shock Index; and green line, sPESI.

Table 4

Receiver operating characteristic curve analysis for prediction of 30-day mortality using each risk score

	ROC AUC for Prediction of All-cause 30-d Mortality
	Geneva	Shock Index	sPESI	ESC	GRACE
Area under curve	0.623	0.639	0.705	0.662	0.715
95% Confidence interval	0.53–0.71	0.55–0.73	0.61–0.80	0.57–0.76	0.63–0.80
p Value	0.016	0.008	<0.001	0.002	<0.001
Cross-tabulation of area under receiver operating characteristic curve among different scores
Geneva	Geneva	Area differences = 0.023; p = 0.721	Area differences = 0.066; p = 0.181	Area differences = 0.002; p = 0.980	Area differences = 0.076; p = 0.271
Shock Index	Area differences = 0.023; p = 0.721	Shock Index	Area differences = 0.043; p = 0.454	Area differences = 0.021; p = 0.752	Area differences = 0.053; p = 0.351
	Area differences = 0.066; p = 0.181	Area differences = 0.043; p = 0.454	sPESI	Area differences = 0.064; p = 0.353	Area differences = 0.010; p = 0.860
European Society of Cardiology	Area differences = 0.002; p = 0.980	Area differences = 0.021 P value = 0.752	Area differences = 0.064 P value = 0.353	ESC	Area differences = 0.074 P value = 0.212
Global Registry of Acute Coronary Events	Area differences = 0.076; p = 0.271	Area differences = 0.053; p = 0.351	Area differences = 0.010; p = 0.860	Area differences = 0.074; p = 0.212	GRACE

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Influence of Gender on Ischemic Times and Outcomes After ST-Elevation Myocardial Infarction
2. Effects of Niacin on Glucose Levels, Coronary Stenosis Progression, and Clinical Events in Subjects With Normal Baseline Glucose Levels (<100 mg/dl): A Combined Analysis of the Familial Atherosclerosis Treatment Study (FATS), HDL-Atherosclerosis Treatment Study (HATS), Armed Forces Regression Study (AFREGS), and Carotid Plaque Composition by MRI During Lipid-Lowering (CPC) Study
3. Brugada Phenocopy: Redefinition and Updated Classification
4. ANCHOR Trial Conclusions Regarding the Effects of Pure Eicosapentaenoic Acid on Low-Density Lipoprotein Cholesterol

Stay updated, free articles. Join our Telegram channel

Join