The accuracy of natriuretic peptides (brain natriuretic peptide and N-terminal pro-brain natriuretic) in the differentiation between transfusion-related acute lung injury and transfusion-related circulatory overload in the critically ill
Contribution: TK and GAW performed data collection and management; GL and MK analyzed results and assembled the manuscript; CED, JLW, SBM, and OG designed the research and revised the paper.
This work was supported by the National Heart, Lung, and Blood Institute HL78743, HL81027 and the grant from National Blood Foundation.
Guangxi Li, Division of Pulmonary Department of Guanganmen Hospital China Academy of Chinese Medical Science, Beijing, China; e-mail: email@example.com.
BACKGROUND: The diagnostic workup of transfusion-related acute lung injury (TRALI) requires an exclusion of transfusion-associated circulatory overload (TACO). Brain natriuretic peptide (BNP) and N-terminal pro-brain natriuretic (NT-pro-BNP) accurately diagnosed TACO in preliminary studies that did not include patients with TRALI.
STUDY DESIGN AND METHODS: In this prospective cohort study, two critical care experts blinded to serum levels of BNP and NT-pro-BNP determined the diagnosis of TRALI, TACO, and possible TRALI based on the consensus conference definitions. The accuracy of BNP and NT-pro-BNP was assessed based on the area under the receiver operating curve (AUC).
RESULTS: Of 115 patients who developed acute pulmonary edema after transfusion, 34 were identified with TRALI, 31 with possible TRALI, and 50 with TACO. Median BNP was 375 pg per mL (interquartile range [IQR], 123 to 781 pg/mL) in TRALI, 446 pg per mL (IQR, 128 to 743 pg/mL) in possible TRALI, and 559 pg per mL (IQR, 288 to 1348 pg/mL) in TACO patients (p = 0.038). The NT-pro-BNP levels among patients with TRALI, possible TRALI, and TACO differed significantly with a median value of 1559 pg per mL (IQR, 629 to 5114 pg/mL), 2349 pg/mL (IQR, 919 to 4610 pg/mL), and 5197 pg/mL (IQR, 1695 to 15,714 pg/mL; p = 0.004), respectively. The accuracy of BNP and NT-pro-BNP to diagnose TACO was moderate with an AUC of 0.63 (95% confidence interval [CI], 0.51-0.74) and 0.70 (95% CI, 0.59 to 0.80).
CONCLUSIONS: Natriuretic peptides are of limited diagnostic value in a differential diagnosis of pulmonary edema after transfusion in the critically ill patients.
area under the curve
body mass index
brain natriuretic peptide
estimated creatinine clearance
receiver operator characteristic curves
transfusion-associated circulatory overload.
Transfusion-related acute lung injury (TRALI) is an underreported and underdiagnosed but potentially severe complication of blood transfusion in the critically ill.1 Diagnostic workup for TRALI is tedious and expensive and requires exclusion of transfusion-associated circulatory overload (TACO).2,3 However, the differential diagnosis of cardiogenic (TACO) and permeability pulmonary edema (TRALI) after blood product transfusion poses a diagnostic challenge.4,5 Despite recent clinical and laboratory advancements, diagnosis or exclusion of cardiogenic pulmonary edema is difficult and limited by both the nonspecific nature of clinical findings and the invasive nature of hemodynamic monitoring.6
B-type natriuretic peptide (BNP) and N-terminal pro-BNP (NT-pro-BNP) are cardiac neurohormones specifically secreted from the ventricles in response to volume expansion and pressure overload5 and may represent an attractive and noninvasive clinical test to diagnose or exclude cardiogenic pulmonary edema (TACO) after transfusion. Indeed, BNP and NT-pro-BNP have demonstrated clinical utility in the diagnosis of heart failure in patients presenting to the emergency room,7,8 and preliminary studies suggested both BNP and NT-pro-BNP may be useful for diagnosis of TACO.9,10 These studies, however, did not include patients with TRALI, the most important alternative diagnostic consideration in patients with pulmonary edema after transfusion. To determine the diagnostic value of BNP and NT-pro-BNP in the differentiation between TACO and TRALI, we conducted a prospective cohort study of patients who developed pulmonary edema after transfusion in four intensive care units of a tertiary care hospital.
MATERIALS AND METHODS
The Mayo Clinic Institutional Review Board approved the study protocol. Adult patients (>17 years old) admitted to one of four intensive care units (medical, cardiac, surgical, and mixed) over a period of 2 years (2005-2007) with symptoms and signs of acute pulmonary edema developing within 6 hours of blood product transfusion were considered for enrollment. Patients were followed until discharge from the hospital. Patients with pulmonary edema before transfusion, those who died within the first 6 hours after transfusion, and those who refused research authorization were excluded.
Blood samples were collected using standard venipuncture techniques into tubes containing ethylenediaminetetraacetic acid (EDTA) as soon as feasible after onset of pulmonary edema and within 48 hours in all patients. A BNP test (Triage, Biosite, Inc., San Diego, CA) and an analyzer (Elecsys 2010, Roche Diagnostics, Indianapolis, IN) were used to measure BNP and NT-pro-BNP, respectively. The samples before transfusion were collected from the “waste” blood (residual plasma from EDTA tube ordered by bedside providers less than 48 hr before the transfusion and refrigerated at 4°C). Owing to a lack of stability of BNP, only NT-pro-BNP was determined in pretransfusion samples. Pretransfusion BNP data were collected in a subset of patients who had the test done as a part of their clinical care.
A trained study coordinator prospectively collected clinical characteristics from the electronic medical records. For quality control, all entries were audited and inaccuracies corrected by one of the investigators (GL). TRALI was defined according to a recent Canadian Consensus Conference definition as a syndrome of inflammation and increased pulmonary capillary permeability that is associated with a constellation of clinical, radiologic, and physiologic abnormalities of acute pulmonary edema that cannot be explained by left atrial hypertension, that is, acute lung injury (ALI) developing within 6 hours of transfusion.2,3,11,12 The term “possible TRALI” was used in the presence of other known risk factors of ALI including sepsis, pneumonia, trauma, and aspiration, although the transfusion itself may have had a contributory or even primary role in the development of the ALI.4 TACO was defined by a combination of clinical signs (gallop, jugular venous distension, systolic hypertension), radiographic (cardiothoracic ratio >0.53 and vascular pedicle width >65 mm),13 electrocardiographic (new ST segment and T wave changes), laboratory (elevated troponin T > 0.1 ng/mL), hemodynamic (PAOP >18 mmHg, CVP > 12), echocardiographic findings: the ratio of mitral peak velocity of early filling to early diastolic mitral annular velocity (E/E′ ratio) greater than 15 and/or ejection fraction (EF) of less than 45 percent, presence of severe left-sided valvular heart disease (aortic or mitral stenosis or regurgitation), and the prompt response to appropriate therapy: preload/afterload reduction, treatment of ischemia, or inotropic agents.14,15 Blood transfusion was defined as infusion of any plasma-containing blood product including red blood cells (RBCs), platelets (PLTs), fresh-frozen plasma, or cryoprecipitate. Cellular blood products (e.g., RBCs and PLTs) were not washed or volume-reduced. Sepsis was defined according to the standard criteria.16 Aspiration was defined as witnessed or strongly suspected aspiration of gastric contents into the airway.17 Pneumonia was defined as a new infiltrate on chest x-ray with compatible clinical features (at least two symptoms of acute lower respiratory tract infection and auscultatory findings of altered breath sounds or localized rales).18
Two expert intensivists (CD, OG), blinded to BNP and NT-pro-BNP results, reviewed clinical data from these patients post hoc (at the end of hospitalization) and assigned the diagnosis of TRALI, possible TRALI, or cardiogenic pulmonary edema according to a previously published algorithm.4 Pulmonary edema was confirmed by an independent evaluation of portable chest radiographs, arterial blood gases, and respiratory monitoring (respiratory rate, pulse oximetry, ventilator settings). Dynamic changes in patient's history, physical exam, echocardiogram, electrocardiogram, cardiac enzymes, digital portable chest radiographs, fluid balance, and data from invasive monitoring (central venous pressure and pulmonary artery occlusion pressure) were integrated with the clinical course and response to therapy to determine whether the onset of acute pulmonary edema after transfusion in these patients was principally due to cardiogenic (TACO) or noncardiogenic (TRALI or possible TRALI) causes. According to Consensus Conference Criteria,4 the presence of antibodies to human leukocyte antigens (HLA) and human neutrophil antigens (HNA) were not required in assigning the clinical diagnosis of TRALI. Disagreements were resolved by consensus. This methodology provided a moderate interobserver agreement (kappa value between 0.6 and 0.7) in the two previous studies.19,20
Demographics, baseline characteristics, the risk factors for ALI and for cardiogenic edema, pre- and posttransfusion NT-Pro-BNP and BNP levels were compared among patients who developed TACO, TRALI, and possible TRALI. Data were expressed as mean values ± standard deviation (SD), median, or proportions and were compared among groups using analysis of variance and Kruskal-Wallis test, if appropriate. A post hoc analysis comparing TRALI and TACO (excluding possible TRALI) was performed and annotated in tables. Spearman coefficients were used to measure correlations between variables and linear regression analysis was performed to describe relationship between BNP and NT-pro-BNP levels and other variables (without Bonferroni correction).21,22 Receiver operator characteristic curves (ROC) were constructed and area under the curve (AUC) calculated to determine the accuracy of the biomarkers and the best threshold levels that accurately differentiate TACO from TRALI and possible TRALI. Sensitivity, specificity, positive and negative likelihood ratios, and 95 percent confidence intervals (CIs) were calculated for the cutoff values. Computer software (JMP 7.0.1, SAS Institute, Inc., Cary, NC) was used for all analyses. A probability of 0.05 or less was taken as significant.
Of 115 patients with acute pulmonary edema after transfusion enrolled in this prospective cohort study, 34 were determined to have TRALI, 31 possible TRALI, and 50 TACO. Median time from the recognition of pulmonary edema to the time of testing was 21.5 hours (interquartile range [IQR], 9.1 to 27.2 hr) for BNP and 11.7 hours (IQR, 6.6 to 24.0 hr) for NT-pro-BNP. Of six patients with clinical TRALI who underwent donor testing, HLA antibodies were present in 5 and HNA antibodies in 2.
Table 1 provides clinical characteristics of critically ill patients with TACO, TRALI, and possible TRALI. APACHE III scores were similar among the three groups. Patients with TACO were more likely to be older, have a history of congestive heart failure, and have worse kidney function but lower 24-hour fluid balance. Patients with TRALI had lower hemoglobin (Hb) values and lower hospital mortality (Table 1).
Table 1. Clinical characteristics of patients with TACO, TRALI, and possible TRALI
|Age (years)*||72.5 ± 13.2||60.5 ± 16.9||59.6 ± 14.4¶||<0.001|
|APACHE III at admission†||64.5 (53.0-88.8)||71.0 (53.0-94.0)||67.0 (46.8-93.8)||0.770|
|PaO2/FiO2†||152.0 (94.6-226.3)||128.0 (95.7-192.0)||162.5 (122.5-223.5)||0.295|
|BMI†||27.1 (24.1-31.1)||25.4 (21.9-28.4)||24.3 (21.5∼32.1)||0.630|
|History of congestive heart failure||11 (22.0)||1 (3.2)||4 (11.8)||0.055|
|Chronic lung disease (%)||9 (18.0)||6 (19.4)||4 (11.8)||0.665|
|Sepsis (%)||7 (14.0)||15 (48.4)||‡||<0.001|
|Pneumonia (%)||8 (16.0)||5 (16.1)||‡||0.886|
|Aspiration (%)||3 (6.0)||3 (9.7)||‡||0.482|
|Trauma (%)||12 (24.0)||17 (54.8)||‡||0.002|
|Postoperative (%)||32 (64.0)||16 (51.6)||26 (76.5)||0.151|
|Minimal Hb within 48 hr§ (mg/dL)†||8.7 (8.1-10.3)||8.5 (7.7-9.2)||8.3 (7.1-9.3)¶||0.010|
|Renal failure (%)||34 (68.0)||13 (41.9)||14 (41.2)¶||0.019|
|Renal replacement therapy (%)||5 (10.0)||0||1 (2.9)||0.112|
|24-hr fluid balance (L)†||3.15 (1.04 to 5.00)||5.02 (0.81 to 8.50)||4.81 (1.9 to 10.0)¶||0.020|
|CrCl|| (mL/min)†||47.0 (30.78 to 67.67)||78.2 (47.2 to 102.9)||68.4 (45.4 to 96.1)¶||0.008|
|In-hospital death (%)||16 (32.0)||15 (48.4)||4 (11.8)¶||0.006|
|Mechanical ventilation (%)||22 (44.0)||17 (54.8)||19 (55.9)||0.479|
|Mechanical ventilation before pulmonary edema||12 (24.0)||10 (32.3)||7 (20.6)||0.538|
|New mechanical ventilation after pulmonary edema||10 (20.0)||7 (22.6)||12 (35.3)||0.264|
Table 2 provides hemodynamic characteristics of patients with TACO, TRALI, and possible TRALI. TACO patients had significantly lower systolic EF, but no significant differences in systolic blood pressure, mean artery pressure, troponin T, estimated right ventricular systolic pressure, CVP, PCWP, or proportion of patients with acute ischemia and tachydysrhythmia. Both BNP and NT-pro-BNP levels were significantly higher in patients with TACO compared to TRALI and possible TRALI (Table 3). However, diagnostic accuracy of both markers was only moderate (Table 4). The AUC for BNP was 0.63 and for NT-pro-BNP 0.70. The results were similar when patients with possible TRALI and those in whom the two experts disagreed were excluded from the analysis (AUCBNP = 0.62, AUCpro-BNP = 0.73). The diagnostic performance of BNP and NT-pro-BNP at several cutoff points is presented in Table 4. For example, at a cutoff point 11,120 pg per mL or more, NT-pro-BNP had a specificity of 90 percent and a positive predictive value of 74 percent for the diagnosis of TACO. At a cutoff point 999 pg per mL or more, BNP had a specificity of 89 percent and a positive predictive value of 78 percent for the diagnosis of TACO.
Table 2. Hemodynamic characteristics of patients with TACO, TRALI, or possible TRALI
|Systolic blood pressure (mmHg)*||122 (106-138)||113 (104-133)||120 (110-131)||0.369|
|Mean arterial pressure (mmHg)*||80 (71-93)||78 (68-83)||79 (73-88)||0.250|
|Electrocardiogram acute ischemic changes (%)||9 (18.0)||2 (6.5)||4 (11.8)||0.264|
|Electrocardiogram tachydysrhythmia present (%)||19 (38.0)||16 (51.6)||11 (32.4)||0.210|
|Troponin T (ng/dL), n = 98*||0.08 (0.03-0.071)||0.045 (0.01-0.083)||0.05 (0.01-0.06)||0.292|
|EF, % (n = 59)*||44 (27.8-54.8)||64.5 (57.3-72.3)||60.0 (58.0-68.0)†||<0.001|
|Estimated right ventricular systolic pressure (%), n = 37*||48 (41-54)||52.5 (36-62)||33 (31-42.8)||0.058|
|Pulmonary artery diastolic pressure (cmH2O), n = 21*||25 (22-28)||23 (12-26)||13 (10-18)||0.135|
|Central venous pressure, cmH2O (n = 63)*||14.5 (11.3-17.0)||10 (7-14.3)||11 (7.5-14.5)||0.322|
|Pulmonary capillary wedge pressure, cm H2O (n = 11)*||20 (11.5-29)||14.5 (13-16)||11.5 (9.5-12.8)||0.209|
Table 3. Comparison of BNP and NT-pro-BNP between TACO, TRALI, and possible TRALI
|BNP before transfusion, pg/mL (n = 23)*||521.5 (143-2180.3)||85 (49-291)||170.5 (41-407.3)||0.128|
|BNP after transfusion (n = 73), pg/mL*||559 (287.8-1347.8)||446 (128-743.3)||375 (122.5-780.5)‡||0.038|
|NT-pro-BNP before transfusion (n = 61), pg/mL*||3410 (686-11951.5)||948 (232-2352)||664 (138.5-2402)‡||0.024|
|NT-pro-BNP after transfusion, pg/mL (n = 84)*||5197 (1695-15714)||2349 (919-4610)||1558.5 (628.5-5114)‡||0.004|
|NT-pro-BNP ratio† (n = 61)*||1.3 (1.0-3.8)||3.0 (1.4-9.0)||2.0 (1.3-5.9)||0.257|
Table 4. Performance characteristics of BNP and NT-pro-BNP at various cutoff points
|BNP|| || || || || || || |
| ≥999||0.37 (0.22-0.54)||0.89 (0.73-0.96)||0.78 (0.52-0.93)||0.57 (0.43-0.70)||3.31 (1.2-9.1)||0.71 (0.55-0.91)||0.63 (0.51-0.74)|
| ≥587||0.45 (0.29-0.62)||0.64 (0.46-0.79)||0.57 (0.38-0.74)||0.52 (0.37-0.67)||1.24 (0.71-2.17)||0.86 (0.63-1.19)|| |
| ≥176||0.84 (0.68-0.93)||0.28 (0.15-0.45)||0.55 (0.42-0.68)||0.63 (0.36-0.84)||1.17 (0.91-1.49)||0.57 (0.24-1.36)|| |
|NT-pro-BNP|| || || || || || || |
| ≥11,120||0.40 (0.24-0.58)||0.90 (0.78-0.96)||0.74 (0.49-0.90)||0.69 (0.56-0.79)||4.08 (1.62-10.30)||0.67 (0.51-0.88)||0.70 (0.59-0.80)|
| ≥2,920||0.6 (0.42-0.76)||0.61 (0.46-0.74)||0.51 (0.35-0.67)||0.69 (0.53-0.81)||1.53 (0.99-2.37)||0.66 (0.43-1.01)|| |
| ≥283||1 (0.88-1.00)||0.12 (0.05-0.25)||0.44 (0.33-0.55)||1 (0.52-1.00)||1.13 (1.03-1.25)||0|| |
Eighty (69.6%) patients survived to hospital discharge. BNP levels were higher in nonsurvivors (562 pg/mL; range, 262-1,247.5 pg/mL) than in survivors (450 pg/mL; range, 199.5-899.5 pg/mL) but the difference was not significant (p = 0.388). NT-pro-BNP levels were also higher in nonsurvivors (4,242 pg/mL; range, 1,326.8-22,325 pg/mL) than in survivors (2533 pg/mL; range, 942.3-6,422 pg/mL) without significant difference (p = 0.119).
Correlations between the levels of BNP and NT-pro-BNP with clinical and hemodynamic characteristics of transfused critically ill patients were calculated with Spearman test. It showed that EF (rs = −0.471, p = 0.004), estimated creatinine clearance (CrCl; rs = −0.285, p = 0.014), and body mass index (BMI; rs = −0.229, p = 0.049) correlated with BNP while NT-pro-BNP was significantly correlated with EF (rs = −0.338, p = 0.033), CrCl (rs = −0.493, p < 0.001), BMI (rs = −0.256, p = 0.017), E/E′ (rs = 0.310, p = 0.025), and fluid balance (rs = −0.228, p = 0.035).
In this single-center prospective cohort study of critically ill medical and surgical patients presenting with acute pulmonary edema after transfusion, serum natriuretic peptides (BNP and NT-pro-BNP) levels were higher in patients who developed posttransfusion cardiogenic pulmonary edema (TACO) compared to those who developed ALI (TRALI and possible TRALI) but proved to be of limited diagnostic value due to a large overlap among the observed values in the critically ill patients. Both peptides correlated with systolic dysfunction, BMI, and estimated creatinine clearance, while NT-pro-BNP also correlated with diastolic dysfunction and fluid balance. BNP and NT-pro-BNP levels did not predict short-term mortality.23
Our results confirm the previous findings demonstrating natriuretic peptide levels could not adequately discriminate between hydrostatic pulmonary edema and ALI in clinical settings other than transfusion.19,24,25 This may be due to increased levels of BNP and NT-pro-BNP related to a direct effect of inflammatory mediators on cardiac myocytes,26 shock,27 the presence of renal dysfunction,19 or severe hypoxemia.28 BNP and NT-pro-BNP levels are frequently elevated in acute respiratory distress syndrome, a more severe form of ALI.28,29 One reason is the common occurrence of acute right heart dysfunction in setting of severe hypoxemia regardless of its cause.30 Another probable reason is related to a direct stimulatory effect of acute hypoxemia on BNP and NT-pro-BNP secretion.31 Khan and colleagues32 reported that hypoxemia alone may be sufficient to increase steady-state levels of NT-pro-BNP protein which may be adaptive, preserving cardiomyocyte function during hypoxemia.
The relationship between hypoxemia and elevated levels of BNP and NT-pro-BNP provides the likely explanation for the discrepant findings between our study and the two preliminary studies, which suggested that both BNP and NT-pro-BNP may be useful in the diagnosis of TACO.9,10 Specifically, the control groups of the previous studies did not include patients with hypoxemia and performance of these tests in groups defined by the presence of hypoxemia, that is, TACO and TRALI, is uncertain and not supported by our data. Zhou and colleagues9 measured BNP before and after transfusion of RBCs in 21 patients with clinical suspicion of TACO and in 19 controls that did not have symptomatic dyspnea or hypoxemia. In Zhou's study, the increase in BNP after transfusion had a sensitivity of 81 percent and specificity of 89 percent for the diagnosis of TACO. Since none of the control patients developed TRALI, the value of BNP in differentiating between TRALI and TACO could not be determined. Tobian and coworkers10 suggested that NT-pro-BNP is both a sensitive and a specific marker for TACO and can be helpful in confirming transfusion-associated fluid overload. This study design was similar in design to Zhou's study which excluded TRALI patients in the control group. Our study was specifically designed to investigate the utility of BNP and NT-pro-BNP in hypoxemic patients with TACO or TRALI and found both biomarkers of limited diagnostic value in this setting. Clearly, bedside clinicians and transfusion specialists can not solely rely on elevated BNP or NT-pro-BNP to exclude TRALI in patients presenting with dyspnea and hypoxemia after transfusion. However, BNP and NT-pro-BNP testing may still be useful for risk stratification, perhaps combined with other clinical variables. A model combining biomarker levels and key clinical variables may ultimately be the most efficient way to provide an accurate differential diagnosis of syndromes such as TACO and TRALI. Of note, our results do not support the use of a commonly used clinical test—calculated fluid balance for decision making with regard to the presence or absence of “fluid overload” and TACO. In fact, in our study calculated fluid balance was lower in patients with TACO compared to those with TRALI and possible TRALI despite higher cardiac filling pressures and worse cardiac and renal function in the former group. The accuracy of fluid balance in the critically ill patient is questionable at best and its use may not be accurate in differential diagnosis of pulmonary edema after transfusion.
In our study, natriuretic peptide values were not predictive for intensive care unit or in-hospital mortality, in patients with TRALI or in patients with TACO. Our findings are similar to those reported by Jefic and colleagues,25 but in contrast to reports from two other studies.27,33 This may in part be related to low mortality of our TRALI and TACO patients. In contrast to the previous reports,34,35 natriuretic peptides correlated poorly with Hb value in our cohort. This could in part be explained by the increased complexity of patients in intensive care unit settings.
In addition to hypoxia, other factors such as age,36 sex,36 diuretics and angiotensin-converting enzyme inhibitors,37,38 BMI,39 anemia,34 high output states (for example, sepsis, hyperthyroidism), and renal failure4 have all been reported to modify the levels of natriuretic peptides in critically ill patients which may contribute to the limitation of natriuretic peptides. Especially the renal function is very important factor that could affect the clearance of natriuretic peptides and limit their diagnostic value.19 Similar to other cohorts of critically ill patients, a significant number of patients in our study had acute renal impairment. When stratified by the renal function (CrCl > 45), the performance of NT-pro-BNP improved (AUC, 0.80) with only a small sample size left (n = 29).
The strengths of our study stem from its prospective design and blinding of the outcome assessors, minimizing bias in the interpretation of natriuretic peptide results. Our study design is, however, limited by the existing clinical definitions and somewhat overlapping nature of cardiogenic (TACO) and permeability (TRALI) pulmonary edema. The concurrent presentation of both entities may confound accurate differentiation and thus diagnostic test performance. The accuracy of performance might be affected by the time of sample collection which contributed to the variation of the natriuretic peptides concentration after the presence of pulmonary edema although we tried to circumscribe it within 24 hours. Moreover, in critically ill patients, natriuretic peptides level may be influenced by numerous coexisting conditions further limiting their usefulness in differential diagnosis of TACO and TRALI.
In conclusion, BNP and NT-pro-BNP testing did not reliably distinguish TACO from TRALI and possible TRALI in a cohort of transfused critically ill patients. In particular, high levels of BNP and NT-pro-BNP cannot be used to obviate the need for diagnostic workup of TRALI.