Scintigraphic Lung Scans and Clinical Assessment in Critically Ill Patients With Suspected Acute Pulmonary Embolism*

Transcription

1 Scintigraphic Lung Scans and Clinical Assessment in Critically Ill Patients With Suspected Acute Pulmonary Embolism* Jerald W. Henry, MS; Paul D. Stein, MD, FCCP; Alexander Gottschalk, MD, FCCP; Bruce Relyea, MD; and Kenneth V. Leeper, Jr, MD, FCCP critically ill patients. The positive predictive values of the clinical assessments did not differ to a statistically significant extent from noncritically ill patients. Clinical assessment, when concordant with the lung scan interpretation, usually increased the positive predictive value for pulmonary embolism. Conclusion: Scintigraphic lung scans and clinical assessment retain their diagnostic value, even in critically ill patients. (CHEST 1996; 109:462-66) Cl=confidence interval; PE=pulmonary embolism; PIOPED=Prosp«?ct;ive Investigation of Pulmonary Embolism Diagnosis; VIQ=ventilation/perfusion Key words: pulmonary embolism; thromboembolic disease; ventilation/perfusion lung scan Purpose: The purpose of this investigation was to evaluate the diagnostic accuracy of radionuclide scintigraphic lung scans and clinical assessment in critically ill patients with suspected acute pulmonary embolism. Materials and methods: Critically ill patients were defined as follows: (1) patients who were hypoxemic on room air, and not given ventilatory support (n=89); (2) patients given ventilatory support (n=46); and (3) patients in ICUs, but not given ventilatory support (n=85), and hypotensive patients who were not hypoxemic or given ventilatory support (n=3). Comparisons were made with patients who had none of these characteristics of critically ill patients (n=627). Data are from the Prospective Investigation of Pulmonary Embolism Diagnosis. Results: The sensitivities, specificities, and positive predictive values of high probability lung scans among each of the four categories of critically ill patients were not statistically significantly lower than values in nonpulmonary embolism (PE) is a potentially lifethreatening condition and is particularly likely to be fatal in critically ill patients if the condition remains undetected. 1 Unfortunately, the clinical diagnosis of acute PE in an ICU is exceedingly difficult_! Ventilation/perfusion (V/Q) lung scans are useful in the assessment of the general population of patients with suspected acute PE, 2 in patients with prior cardiopulmonary disease, 3 and in patients with COPD. 4 The utility of lung scans in critically ill patients, however, is unknown. The ability to obtain a technically sound ventilation lung scan is difficult in patients who are given ventilatory support, 5 although some have succeeded in obtaining inhalation radioaerosollung scans of excellent quality in such patients. 6 *From the Henry Ford Heart and Vascular Institute, Detroit (Mr. Henry and Drs. Stein and Relyea); Michigan State University School of Medicine, (Dr. Gottschalk) East Lansing; and The Umversity of Tennessee College of Medicine (Dr. Leeper), Memphis, Tennessee. Manuscript received July 10, 1995; revision accep.ted September 19. Reprint requests: Dr. Stein, Henry Ford Hospital, Room 1107, New Center Pavilion, 2921 West Grand Boulevard, Detroit, MI To assess the diagnostic accuracy of lung scans and clinical assessment in critically ill patients with suspected acute PE, data from the Prospective Investigation of Pulmonary Embolism Diagnosis (PI 0 PED) were evaluated. 2 MATERIALS AND METHODS Data were evaluated in 850 patients in PIOPED in whom the diagnosis of PE was made or excluded by p u l ~ oangiography. n a y In all patients, a clinical assessment of the likelihood of PE had been made. With the exception of patients given ventilatory support, all patients had V/Q scans, and they also had measurements of blood gases while breathing room air. Patients were in either of two arms of PIOPED. Patients in one arm consented to randomization to obligatory pulmonary angiography as described in the primary PI OPED report. 2 Patients in the other arm ofpioped were referred for pulmonary angiography by their attending physicians. These patients were not described in the primary PIOPED report. The methods used for obtaining and interpreting pulmonary angiograms and V/Q scans were previously described. 2 The definitions of high, intermediate, low, and near normal/normal probability V/Q lung scans also have been previously described. 2 The near normal/normal category included readings of very low probability by one central V/Q scan reader and low probability by the other, very low probability by both readers, or very low probability by one and normal by the other. 2 The category "normal" V/Q scan required an

2 Table!-Sensitivities and Specificities of High Probability Lung Scans Among Critically Ill and Noncritically Ill Patients* Hypoxemic, not on vent Ventilatory support ICU not on vent Hypotensive, not hypoxemic Noncritically ill Sensitivity n!n (%) 17/27 (63) 1 4/12 (33) 8/31 (26) 0/1 (0) 87/227 (38) Specificity n1/n1 (%) 60/62 (97) 34/34 (100) (98) 385/400 (96) *VENT=ventilatory support; n=number of patients with high probability V/Q scans; N=number of patients with PE; n 1 =number of patients with nonhigh probability V/Q scans; N 1 =number of patients with nope. tp<0.02 hypoxemic vs not critically ill. interpretation of normal by both central V/Q scan readers. 2 Among the 850 patients studied, 223 were critically ill, as defined below, and 627 were not critically ill. Among the critically ill: ( l ) 89 were hypoxemic on room air (Pa02 <50 mm Hg) and not on ventilatory support; (2) 46 patients were given ventilatory support prior to the suspected PE; (3) 85 patients were in ICUs but not given ventilatory support; and ( 4) 3 were hypotensive, but not hypoxemic, and not given ventilatory support. One of 89 hypoxemic patients who was not given ventilatory support was also hypotensive. Fourteen of 89 hypoxemic patients who were not given ventilatory support and 1 of 3 hypotensive patients who were not given ventilatory support, were in ICUs. Thirty of 46 patients on ventilatory support were in ICUs. None of the patients on ventilatory support were hypotensive. The 627 patients who were not critically ill had a systolic blood pressure ~ 8mm 0 Hg, Pa02 on room air ~ 5mm 0 Hg, were not given ventilatory support, and were not in ICUs. Among these patients, 227 had PE and 400 did not have PE. Among the 46 patients who were given ventilatory support, 36 had V/Q scans and 10 had only perfusion lung scans. Perfusion scans alone were read according to the same criteria as V/Qscans, except the location of perfusion defects was related to the location of abnormalities on the plain chest radiograph 7 Patients were excluded if they had no clinical assessment of the likelihood of PE. A clinical assessment of the likelihood of PE was made either before the lung scan or without knowledge of the lung scan. These clinical assessments were made on the basis of bedside evaluation, chest radiograph, electrocardiogram, and routine blood tests (not D-dimers). Assessments were based on clinical judgement; there were no specific criteria for the likelihood of PE. Statistical Methods A x 2 test was used to compare the positive predictive values, sensitivities, and specificities of the various critically ill groups with noncritically ill patients. Positive predictive value was defined as the percentage of patients with a particular clinical assessment or diagnostic interpretation of the lung scan who had PE. Sensitivity was defined as the proportion of patients with PE who had a high probability interpretation of the lung scan. Specificity was defined as the proportion of patients who did not have PE in whom the lung scan was interpreted as nonhigh probability. The 95% confidence inte1vals (CI) were determined on the basis of the exact binomial distribution. RESULTS Comparison of Randomized Patients vs Referred Patients Among the 223 patients in the 4 categories of critically ill patients, 159 were randomized for obligatory pulmonary angiography and 64 were referred for pulmonary angiography. Among the 627 noncritically ill patients, 431 were randomized for obligatory pulmonary angiography and 196 were referred for pulmonary angiography. Comparisons of patients randomized for obligatory angiography with patients referred for pulmonary angiography showed no statistically significant differences of the sensitivity and specificity of a high probability interpretation of the lung scan in any of the four categories of critically ill patients or in the noncritically ill patients. There were no statistically significant differences of the positive predictive value of high, intennediate, low, or near normal/normal interpretations of the lung scan between randomized and referred patients in any of these groups. Also, the positive predictive values of the clinical assessment did not differ to a statistically significant extent in any group between randomized and referred patients. Because there were no statistically significant differences between randomized and referred patients, we considered it valid to maximize the database by combining both groups. Only data from the combined group are reported. Patients With Hypoxemia Not on Ventilatory Support Among 89 patients who were hypoxemic and not on ventilatory support, 27 (30%) had PE and 62 (70%) did Table 2-Positive Predictive Value ofv!q Lung Scan Interpretation and Clinical Assessment in Hypoxemic Patients Not on Ventilatory Support Clin Prob % V/Q Scan PE+/Test (%) Clin Prob 20-79% Clin Prob 0-19% High 6/6 (100) Intermediate 010 (-) Low 010 (- ) Near Normal/Normal 0/2 (0) 6/8 (75) ll/13 (85) 6/29 (21) 3/18 (17) 0/1 (O) 20/61 (33) 0/13 (0) 1/6 (17) 0/l (0) l/20 (5) 17/19 (89) 6/42 (14)* 4/24 (17) 0/4 (0) 27/89 (30) *p<o.ol hypoxemic vs noncritically ill patients (see Table 3). CHEST / 109 / 2 / FEBRUARY,

3 Table 3-Positive Predictive Value ofv!q Lung Scan Interpretation and Clinical Assessment in Patients Who Were Not Critically Ill V/Q Scan PE ftest (%) Clin Prob % Clin Prob 20-79% Clin Prob 0-19% PE+/Test (%) High 13/14 (93) Intermediate 22/29 (76) Low 418 (50) Near Normal/Normal 1/ l (100) (77) 69/81 (85) sn(n ) 87/102 (85) 60/177 (34) (21) 96/272 (35) 29/122 (24) 4/62 (6) 37/192 (19) 5/38 (13) 1/22 (5) 7/61 (ll) 163/418 (39) (15) 227/627 (36) not have PE. The sensitivity of a high probability interpretation of the V/Q lung scan was 17 of27 (63%) (95% CI %) (Table 1). This sensitivity was higher than the sensitivity of a high probability interpretation of the V/Q lung scan in noncritically ill patients, 87 of 227 (38%) (95% CI 32-45%) (p<0.02) (Table 1). The specificity of high probability V/Q lung scans in hypoxemic patients was 60 of 62 (97%) (95% CI %) (Table 1). V/Q was 17 of 19 (89%) (95% CI 67-99%) (Table 2). Positive predictive values of high, low and near normal! normal interpretations of V/Q lung scans among hypoxemic patients did not differ to a statistically significant extent from positive predictive values in noncritically ill patients (Tables 2 and 3). Intermediate probability interpretations of the V/Q lung scan, however, showed a lower positive predictive value for PE in hypoxemic patients than among noncritically ill patients (p<0.01) (Tables 2 and 3). 100% likelihood ofpe, 6 of8 (75%) (95% CI 35-97%) had PE (Table 2). When the clinical assessment was 0 to 19% likelihood of PE, PE was present in only 1 of 20 (5%) (95% CI 0%-25%) (Table 3). The positive predictive values of high and low likelihood clinical assessments did not differ to a statistically significant extent between hypoxemic patients and noncritically ill patients (Tables 2 and 3). Patients Given Ventilatory Support Among 46 patients who were given v entilatory sup- port, 12 (26%) had PE and 34 (74%) did not have PE. The sensitivity of a high probability lung scan (V/Q scan or perfusion lung scan alone) was 4 of 12 (33%) (95% CI 10-65%) (Table 1). This sensitivity was not statistically significantly different from the sensitivity of a high probability V/Q scan in patients who were not critically ill, 87 of227 (38%) (95% CI 32-45%) (Table 1). The specificity of a high probability lung scan (V/Q scan or perfusion lung scan alone) was 34 of 34 (100%) (95% CI %) (Table 1). lung scan was 4 of 4 (100%) (95% CI %) (Table 4). Positive predictive values of high, intermediate, low, and near normal/normal interpretations of lung scans among patients on ventilatory support did not differ to a statistically significant extent from positive predictive values ofv/q scans in patients who were not critically ill. 100% likelihood ofpe, 3 of 4 (75%) (95% CI 19-99%) had PE (Table 4). When the clinical assessment was 0 to 19% likelihood ofpe, PE was present in 0 of 11 (0%) (95% CI 0-28%) (Table 4). The positive predictive values of high and low likelihood clinical assessments did not differ to a statistically significant extent between patients on ventilatory support and non critically ill patients (Tables 3 and 4). Patients in ICUs Not on Ventilatory Support Among 85 patients who were in ICU s and not given ventilatory support, 31 (36%) had PE and 54 (64%) did not have PE. Patients in ICUs not given ventilatory Table 4-Positive Predictive Value of Lung Scan Interpretation and Clinical Assessment in Patients on Ventilatory Support High V/Q Scan Intermediate Low Near Normal/Normal Clin Prob % PE ftest (%) l /1 (100) 0/1 (0) 3/4 (75) Clin Prob 20-79% PE ftest (%) 5/16 (31) 2/13 (15) 9/31 (29) Clin Prob 0-19% on(o) 0/3 (0) 0/l (0) 0/ll (0) 4/4 (100) 6/24 (25) 2/17 (12) Oil (0) 12/46 (26) 464

4 Table 5-Positive Predictive Value ofv!q Lung Scan Interpretation and Clinical Assessment in Patients in ICUs Who Were Not On Ventilatory Support Clin Prob % Clin Prob 20-79% Clin Prob 0-19% V/Q Scan PE+!fest+ (%) PE+!fest+ (%) PE+!fest+ (%) PE+!fest+ (%) High 212 (100) Intermediate 415 (80) Low 1/1 (100) Near Normal/Normal 7/8 (88) 6/6 (100) 0/1 (0) 8/9 (89) 14/34 (41 ) 0/3 (O) 18/42 (43) 4/22 (18) 017 (O) 5/30 (17) 0/4 (0) 010 (- ) 0/4 (0) 24/66 (36) 0/ll (0) (36) support were from either the medical ICU (n=62) or the surgical ICU (n=23). The sensitivity of a high probability V/Q lung scan was 8 of 31 (26%) (95% CI 12-45%) (Table 1). This sensitivity was not statistically significantly different than the sensitivity of a high probability V/Q lung scan in patients who were not critically ill, 87 of227 (38%) (95% CI 32-45%) (Table 1). The specificity of high probability V/Q lung scans was 53 of 54 (98%) (95% CI %) (Table 1). V/Q lung scan was 8 of 9 (89%) (95% CI %) (Table 5). Positive predictive values of high, intermediate, low, and near normal/normal interpretations of V/Q lung scans among patients in ICUs and not given ventilatory support did not differ to a statistically significant extent from noncritically ill patients. 100% likelihood of PE, 7 of 8 (88%) (95% CI 47%-100%) had PE (Table 5). When the prior clinical assessment was 0 to 19% likelihood of PE, PE was present in 0 of 11 (0%) (95% CI 0%-28%). The positive predictive value of high and low likelihood clinical assessments also did not differ to a statistically significant extent between patients in ICUs and noncritically ill patients (Tables 3 and 5). Hypotensive Patients Not Hypoxemic and Not on Ventilatory Support Data were sparse on patients who were hypotensive and not hypoxemic and not given ventilatory support (Table 1). DISCUSSION A high probability interpretation of the V/Q lung scan (or perfusion scan alone in patients given ventilatory support) in critically ill patients showed a sensitivity for PE that was not statistically significantly lower than the sensitivity in noncritically ill patients. The specificity of high probability lung scans among critically ill patients was also no lower than in noncritically ill patients. Even patients given ventilatory support, who are likely to be difficult to study, showed a sensitivity and specificity of a high probability V/Q scan or perfusion scan alone that was at least as high as in noncritically ill patients. Among critically ill patients, the positive predictive value of a high probability interpretation of the V/Q scan was as high as in noncritically ill patients. In critically ill patients in whom there was sufficient data for evaluation (hypoxemic patients not on ventilatory support, patients on ventilatory support, and patients in ICUs not on ventilatory support), when physicians thought that the likelihood of PE was 80 to 100%, they were correct in 75 to 88% of the patients. This was comparable with the positive predictive value in noncritically ill patients (77%). Conversely, if physicians assessed the clinical likelihood of PE as 0 to 19%, only 0 to 5% had PE, which was comparable to the frequency of PE in non critically ill patients with a low likelihood clinical assessment ( 15%). There were few critically ill patients in whom physicians were confident of the diagnosis of PE. Among each of the three categories of critically ill patients in whom meaningful data were available, a high likelihood clinical assessment was made in only 9%. The fraction of patients with a high likelihood clinical assessment in noncritically ill patients was comparable, 8%. Among these categories of critically ill patients, physicians were uncertain of the clinical diagnosis in 67% to 78% of patients, and in noncritically ill patients they were uncertain in 67%. When the clinical likelihood of PE and the lung scan interpretation were concordant, the positive predictive value of PE was usually enhanced. Although experience suggests that PE is particularly difficult to diagnose in patients in I CU s, 1 our data show that the V/Q scan (or perfusion scan alone in patients on ventilatory support) and clinical assessment is as accurate as in noncritically ill patients. Ten of 46 patients who were given ventilatory support had perfusion lung scans without ventilation lung scans because the ability to obtain a technically sound ventilation lung scan is difficult in patients on ventilatory support. 5 In a random sample of patients from PI OPED, a high probability interpretation of a perfusion lung scan alone had a positive predictive value for PE of 93%, which was virtually the same as the positive predictive value of a high probability V/Q scan, CHEST I 109 I 2 I FEBRUARY,

5 94%. 7 The sensitivity and specificity of high probability interpretations of perfusion scans alone did not differ to a statistically significant extent from sensitivity and specificity ofv/q scans. A higher percentage of patients with perfusion lung scans alone, however, had intermediate (indeterminate) interpretations than among patients who had V/Q scans. 7 In conclusion, scintigraphic lung scans and clinical assessment retain their diagnostic utility in critically ill patients and are as accurate as among noncritically ill hospitalized patients. REFERENCES 1 Cowen JC, Kelley MA. An organized approach to detecting pulmonary embolism in the critically ill. J Crit Illness 1994; 9: The PIOPED investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism: Results of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED). JAMA 1990; 263: Stein PD, Coleman RE, Gottschalk A, et al. Diagnostic utility of ventilation/perfusion lung scans in acute pulmonary embolism is not diminished by preexisting cardiac or pulmonary disease. Chest 1991; 100: Lesser BA, Leeper KV, Stein PD, eta!. The diagnosis of pulmonary embolism in patients with chronic obstructive pulmonru.y disease. Chest 1992; 102: Davis LP, Fink-Bennett D. Nuclear medicine in the acutely ill patient-i. Crit Care Clin 1994; 10: Butler SP, Alderson PO, Greenspan RL, et al. The utility of Technetium-99m DTPA inhalational scans in artificially ventilated patients. J Nucl Med 1990; 31: Stein PD, Terrin ML, Gottschalk A, eta!. Value of ventilation/ perfusion scans compared to perfusion scans alone in acute pulmonary embolism. Am J Cardiol1992; 69: