Nilam J. Soni, MD

History and Physical Exam

Pleural effusions are most commonly associated with heart failure, pneumonia, cancer, pulmonary embolism, viral disease, coronary artery bypass surgery, and cirrhosis with ascites. The most common symptoms related to pleural effusion are nonspecific and often indistinguishable from those of the underlying disease process, including cough, dyspnea, and pleuritic chest pain.[5]

Diagnostic accuracy of a physical examination to detect pleural fluid is highly dependent on the size of the effusion and is unlikely to detect effusions <300 mL. A systematic review found the most accurate physical exam findings to rule in a pleural effusion were dullness to percussion (positive likelihood ratio [LR]: 8.7; 95% CI: 2.2‐33.8) and asymmetric chest expansion (positive LR: 8.1; 95% CI: 5.2‐12.7). Normal tactile vocal fremitus was the most accurate physical exam finding to rule out a pleural effusion (negative LR: 0.21; 95% CI: 0.12‐0.37).[6] A major limitation of all these studies is that physical exam was compared to chest radiography as the reference standard, and posterior‐anterior chest radiographs are not sensitive for detection of pleural effusions until 200 mL of fluid has accumulated.[7] Further, chest percussion penetrates to a maximum depth of 6 cm, and its utility is limited in obese patients.[8] Characteristics of pleural fluid that can change management, such as loculations, cannot be detected by physical exam.

Chest Radiography

Chest radiography has traditionally been used to diagnose pleural effusions. Free‐flowing pleural fluid collects in the most dependent portions of the thorax, initially in the subpulmonic space followed by the costophrenic recesses. Pleural fluid is detectable in the costophrenic recesses on lateral upright chest radiograph after 50 mL has accumulated. On standard posterior‐anterior chest radiograph, blunting of the costophrenic recesses and obliteration of the hemidiaphragm are seen when >200 mL and >500 mL of pleural fluid have accumulated, respectively.[7] However, upright chest radiographs can miss a considerable number of effusions, including as many as 10% of parapneumonic effusions large enough to indicate need for drainage.[9] Supine anterior‐posterior chest radiographs can miss a significant proportion of large effusions seen on chest CT,[10] ultrasound,[11] and lateral decubitus radiographs.[12] Pleural effusions are frequently mistaken for parenchymal opacities on portable anterior‐posterior chest radiographs.[10]

Computerized Tomography

Chest CT serves as the reference standard in most modern diagnostic accuracy studies. Limitations of chest CT include difficulty distinguishing small effusions from pleural thickening, dependent atelectasis, or tumor; lower sensitivity for detecting pleural fluid septations compared to ultrasound[13]; exposure of patients to approximately 7 mSv of ionizing radiation (the equivalent radiation dose of 350 chest radiographs)[14]; high cost; and need to transport patients to radiology departments where CT scanners are located.

Pleural Ultrasonography

Ultrasound can rapidly differentiate conditions that demonstrate a nonspecific, radiopaque appearance of lower lung fields on chest radiographs, including pleural effusions, pneumonia, atelectasis, elevated hemidiaphragm, and lung or pleural masses. Advantages of point‐of‐care ultrasound include the ability of hospitalists to acquire and interpret images at the bedside and integrate findings into clinical decision making immediately. Multiple studies have demonstrated superior diagnostic accuracy of ultrasound compared to chest radiography for detection of pleural effusions. Pleural ultrasound can detect physiologic amounts of pleural fluid (5 mL),[15] but a minimal volume of 20 mL is more reliably detected,[16] and ultrasound is 100% sensitive for effusions >100 mL.[17] In a prospective study of critically ill patients with acute respiratory distress syndrome, the diagnostic accuracy of ultrasound for pleural effusions was superior (93%) compared to auscultation (61%) and anterior‐posterior chest radiograph (47%), using chest CT as the reference standard.[18] A meta‐analysis of 4 studies calculated a pooled sensitivity and specificity of ultrasound for detection of pleural effusions as 93% (95% CI: 89%‐96%) and 96% (95% CI: 95%‐98%), respectively.[18, 19, 20, 21, 22] Ultrasound has the additional benefit of characterizing underlying lung parenchyma, which is well described in the literature but beyond the scope of this review.[23]

Sensitivity and specificity of chest radiography and ultrasonography to detect a pleural effusion are displayed in Table 1.[9, 10, 11, 12, 18, 20, 21, 22, 24, 25, 26]

Pleural Fluid Volume

Quantification of pleural fluid volume has been proposed using formulas with sonographic measurements.[30, 31, 32] These formulas are most accurate for moderate‐sized effusions but have not been validated beyond individual study cohorts. The largest study (n = 150) found a strong correlation between calculated and actual volumes drained by thoracentesis (r² = 0.79; P < 0.001) using the formula (Volume [mL] = 16 parietal to visceral pleura distance (mm) at the mid‐diaphragm).[31] Although an accurate quantitative pleural fluid volume assessment may be possible, these formulas are not commonly used in clinical practice. A qualitative assessment is adequate for most clinical decision making using categories of minimal, small, moderate, or large volume.

Simple Versus Complex Effusions

Based on its sonographic appearance, pleural effusions are categorized as simple or complex. Simple pleural effusions are anechoic and usually transudative. Complex pleural effusions are subcategorized as homogeneously or heterogeneously echogenic, with or without septations, and are more often exudative.[33]

Effusions with heterogeneous echogenicity with swirling echoes suggest high cellular content that is often associated with malignancy.[34] Fibrinous stranding, septations, and loculations also suggest an exudative effusion (Figure 4A), and are more readily identified and characterized on lung ultrasound than CT scan.[35]

Figure 4

Homogenously echogenic effusions are most often due to hemothorax or empyema.[36] The high cell count of a hemothorax creates a layering effect in costophrenic recesses (hematocrit sign). Empyemas develop from complex effusions that organize into collections of pus and usually have a homogeneously echogenic, speckled appearance (Figure 4B). Sonographic evidence of septations in the presence of empyema predicts the need for intrapleural fibrinolytic therapy, longer duration of drainage, and possible surgical intervention.[37]

Isolated dense loculations may be challenging to differentiate from peripheral lung or pleural lesions, such as abscess or tumor.

Pleural Thickness

Normal visceral and parietal pleura are apposed and 0.2 to 0.3 mm thick.[38] Pleural effusions with parietal pleural thickness >10 mm, pleural nodularity, and diaphragmatic thickness >7 mm predicted underlying malignancy with high specificity and positive predictive value in 1 study.[39] As many as 20% of anechoic lesions of the pleura are solid rather than fluid. Color flow Doppler ultrasound can differentiate small pleural effusions from solid pleural abnormalities with sensitivity and specificity of 89% and 100%, respectively.[40]

History and Physical Exam

Pleural effusions are most commonly associated with heart failure, pneumonia, cancer, pulmonary embolism, viral disease, coronary artery bypass surgery, and cirrhosis with ascites. The most common symptoms related to pleural effusion are nonspecific and often indistinguishable from those of the underlying disease process, including cough, dyspnea, and pleuritic chest pain.[5]

Diagnostic accuracy of a physical examination to detect pleural fluid is highly dependent on the size of the effusion and is unlikely to detect effusions <300 mL. A systematic review found the most accurate physical exam findings to rule in a pleural effusion were dullness to percussion (positive likelihood ratio [LR]: 8.7; 95% CI: 2.2‐33.8) and asymmetric chest expansion (positive LR: 8.1; 95% CI: 5.2‐12.7). Normal tactile vocal fremitus was the most accurate physical exam finding to rule out a pleural effusion (negative LR: 0.21; 95% CI: 0.12‐0.37).[6] A major limitation of all these studies is that physical exam was compared to chest radiography as the reference standard, and posterior‐anterior chest radiographs are not sensitive for detection of pleural effusions until 200 mL of fluid has accumulated.[7] Further, chest percussion penetrates to a maximum depth of 6 cm, and its utility is limited in obese patients.[8] Characteristics of pleural fluid that can change management, such as loculations, cannot be detected by physical exam.

Chest Radiography

Chest radiography has traditionally been used to diagnose pleural effusions. Free‐flowing pleural fluid collects in the most dependent portions of the thorax, initially in the subpulmonic space followed by the costophrenic recesses. Pleural fluid is detectable in the costophrenic recesses on lateral upright chest radiograph after 50 mL has accumulated. On standard posterior‐anterior chest radiograph, blunting of the costophrenic recesses and obliteration of the hemidiaphragm are seen when >200 mL and >500 mL of pleural fluid have accumulated, respectively.[7] However, upright chest radiographs can miss a considerable number of effusions, including as many as 10% of parapneumonic effusions large enough to indicate need for drainage.[9] Supine anterior‐posterior chest radiographs can miss a significant proportion of large effusions seen on chest CT,[10] ultrasound,[11] and lateral decubitus radiographs.[12] Pleural effusions are frequently mistaken for parenchymal opacities on portable anterior‐posterior chest radiographs.[10]

Computerized Tomography

Chest CT serves as the reference standard in most modern diagnostic accuracy studies. Limitations of chest CT include difficulty distinguishing small effusions from pleural thickening, dependent atelectasis, or tumor; lower sensitivity for detecting pleural fluid septations compared to ultrasound[13]; exposure of patients to approximately 7 mSv of ionizing radiation (the equivalent radiation dose of 350 chest radiographs)[14]; high cost; and need to transport patients to radiology departments where CT scanners are located.

Pleural Ultrasonography

Ultrasound can rapidly differentiate conditions that demonstrate a nonspecific, radiopaque appearance of lower lung fields on chest radiographs, including pleural effusions, pneumonia, atelectasis, elevated hemidiaphragm, and lung or pleural masses. Advantages of point‐of‐care ultrasound include the ability of hospitalists to acquire and interpret images at the bedside and integrate findings into clinical decision making immediately. Multiple studies have demonstrated superior diagnostic accuracy of ultrasound compared to chest radiography for detection of pleural effusions. Pleural ultrasound can detect physiologic amounts of pleural fluid (5 mL),[15] but a minimal volume of 20 mL is more reliably detected,[16] and ultrasound is 100% sensitive for effusions >100 mL.[17] In a prospective study of critically ill patients with acute respiratory distress syndrome, the diagnostic accuracy of ultrasound for pleural effusions was superior (93%) compared to auscultation (61%) and anterior‐posterior chest radiograph (47%), using chest CT as the reference standard.[18] A meta‐analysis of 4 studies calculated a pooled sensitivity and specificity of ultrasound for detection of pleural effusions as 93% (95% CI: 89%‐96%) and 96% (95% CI: 95%‐98%), respectively.[18, 19, 20, 21, 22] Ultrasound has the additional benefit of characterizing underlying lung parenchyma, which is well described in the literature but beyond the scope of this review.[23]

Sensitivity and specificity of chest radiography and ultrasonography to detect a pleural effusion are displayed in Table 1.[9, 10, 11, 12, 18, 20, 21, 22, 24, 25, 26]

Pleural Fluid Volume

Quantification of pleural fluid volume has been proposed using formulas with sonographic measurements.[30, 31, 32] These formulas are most accurate for moderate‐sized effusions but have not been validated beyond individual study cohorts. The largest study (n = 150) found a strong correlation between calculated and actual volumes drained by thoracentesis (r² = 0.79; P < 0.001) using the formula (Volume [mL] = 16 parietal to visceral pleura distance (mm) at the mid‐diaphragm).[31] Although an accurate quantitative pleural fluid volume assessment may be possible, these formulas are not commonly used in clinical practice. A qualitative assessment is adequate for most clinical decision making using categories of minimal, small, moderate, or large volume.

Simple Versus Complex Effusions

Based on its sonographic appearance, pleural effusions are categorized as simple or complex. Simple pleural effusions are anechoic and usually transudative. Complex pleural effusions are subcategorized as homogeneously or heterogeneously echogenic, with or without septations, and are more often exudative.[33]

Effusions with heterogeneous echogenicity with swirling echoes suggest high cellular content that is often associated with malignancy.[34] Fibrinous stranding, septations, and loculations also suggest an exudative effusion (Figure 4A), and are more readily identified and characterized on lung ultrasound than CT scan.[35]

Figure 4

Homogenously echogenic effusions are most often due to hemothorax or empyema.[36] The high cell count of a hemothorax creates a layering effect in costophrenic recesses (hematocrit sign). Empyemas develop from complex effusions that organize into collections of pus and usually have a homogeneously echogenic, speckled appearance (Figure 4B). Sonographic evidence of septations in the presence of empyema predicts the need for intrapleural fibrinolytic therapy, longer duration of drainage, and possible surgical intervention.[37]

Isolated dense loculations may be challenging to differentiate from peripheral lung or pleural lesions, such as abscess or tumor.

Pleural Thickness

Normal visceral and parietal pleura are apposed and 0.2 to 0.3 mm thick.[38] Pleural effusions with parietal pleural thickness >10 mm, pleural nodularity, and diaphragmatic thickness >7 mm predicted underlying malignancy with high specificity and positive predictive value in 1 study.[39] As many as 20% of anechoic lesions of the pleura are solid rather than fluid. Color flow Doppler ultrasound can differentiate small pleural effusions from solid pleural abnormalities with sensitivity and specificity of 89% and 100%, respectively.[40]

History and Physical Exam

Pleural effusions are most commonly associated with heart failure, pneumonia, cancer, pulmonary embolism, viral disease, coronary artery bypass surgery, and cirrhosis with ascites. The most common symptoms related to pleural effusion are nonspecific and often indistinguishable from those of the underlying disease process, including cough, dyspnea, and pleuritic chest pain.[5]

Diagnostic accuracy of a physical examination to detect pleural fluid is highly dependent on the size of the effusion and is unlikely to detect effusions <300 mL. A systematic review found the most accurate physical exam findings to rule in a pleural effusion were dullness to percussion (positive likelihood ratio [LR]: 8.7; 95% CI: 2.2‐33.8) and asymmetric chest expansion (positive LR: 8.1; 95% CI: 5.2‐12.7). Normal tactile vocal fremitus was the most accurate physical exam finding to rule out a pleural effusion (negative LR: 0.21; 95% CI: 0.12‐0.37).[6] A major limitation of all these studies is that physical exam was compared to chest radiography as the reference standard, and posterior‐anterior chest radiographs are not sensitive for detection of pleural effusions until 200 mL of fluid has accumulated.[7] Further, chest percussion penetrates to a maximum depth of 6 cm, and its utility is limited in obese patients.[8] Characteristics of pleural fluid that can change management, such as loculations, cannot be detected by physical exam.

Chest Radiography

Chest radiography has traditionally been used to diagnose pleural effusions. Free‐flowing pleural fluid collects in the most dependent portions of the thorax, initially in the subpulmonic space followed by the costophrenic recesses. Pleural fluid is detectable in the costophrenic recesses on lateral upright chest radiograph after 50 mL has accumulated. On standard posterior‐anterior chest radiograph, blunting of the costophrenic recesses and obliteration of the hemidiaphragm are seen when >200 mL and >500 mL of pleural fluid have accumulated, respectively.[7] However, upright chest radiographs can miss a considerable number of effusions, including as many as 10% of parapneumonic effusions large enough to indicate need for drainage.[9] Supine anterior‐posterior chest radiographs can miss a significant proportion of large effusions seen on chest CT,[10] ultrasound,[11] and lateral decubitus radiographs.[12] Pleural effusions are frequently mistaken for parenchymal opacities on portable anterior‐posterior chest radiographs.[10]

Computerized Tomography

Chest CT serves as the reference standard in most modern diagnostic accuracy studies. Limitations of chest CT include difficulty distinguishing small effusions from pleural thickening, dependent atelectasis, or tumor; lower sensitivity for detecting pleural fluid septations compared to ultrasound[13]; exposure of patients to approximately 7 mSv of ionizing radiation (the equivalent radiation dose of 350 chest radiographs)[14]; high cost; and need to transport patients to radiology departments where CT scanners are located.

Pleural Ultrasonography

Ultrasound can rapidly differentiate conditions that demonstrate a nonspecific, radiopaque appearance of lower lung fields on chest radiographs, including pleural effusions, pneumonia, atelectasis, elevated hemidiaphragm, and lung or pleural masses. Advantages of point‐of‐care ultrasound include the ability of hospitalists to acquire and interpret images at the bedside and integrate findings into clinical decision making immediately. Multiple studies have demonstrated superior diagnostic accuracy of ultrasound compared to chest radiography for detection of pleural effusions. Pleural ultrasound can detect physiologic amounts of pleural fluid (5 mL),[15] but a minimal volume of 20 mL is more reliably detected,[16] and ultrasound is 100% sensitive for effusions >100 mL.[17] In a prospective study of critically ill patients with acute respiratory distress syndrome, the diagnostic accuracy of ultrasound for pleural effusions was superior (93%) compared to auscultation (61%) and anterior‐posterior chest radiograph (47%), using chest CT as the reference standard.[18] A meta‐analysis of 4 studies calculated a pooled sensitivity and specificity of ultrasound for detection of pleural effusions as 93% (95% CI: 89%‐96%) and 96% (95% CI: 95%‐98%), respectively.[18, 19, 20, 21, 22] Ultrasound has the additional benefit of characterizing underlying lung parenchyma, which is well described in the literature but beyond the scope of this review.[23]

Sensitivity and specificity of chest radiography and ultrasonography to detect a pleural effusion are displayed in Table 1.[9, 10, 11, 12, 18, 20, 21, 22, 24, 25, 26]

Pleural Fluid Volume

Quantification of pleural fluid volume has been proposed using formulas with sonographic measurements.[30, 31, 32] These formulas are most accurate for moderate‐sized effusions but have not been validated beyond individual study cohorts. The largest study (n = 150) found a strong correlation between calculated and actual volumes drained by thoracentesis (r² = 0.79; P < 0.001) using the formula (Volume [mL] = 16 parietal to visceral pleura distance (mm) at the mid‐diaphragm).[31] Although an accurate quantitative pleural fluid volume assessment may be possible, these formulas are not commonly used in clinical practice. A qualitative assessment is adequate for most clinical decision making using categories of minimal, small, moderate, or large volume.

Simple Versus Complex Effusions

Based on its sonographic appearance, pleural effusions are categorized as simple or complex. Simple pleural effusions are anechoic and usually transudative. Complex pleural effusions are subcategorized as homogeneously or heterogeneously echogenic, with or without septations, and are more often exudative.[33]

Effusions with heterogeneous echogenicity with swirling echoes suggest high cellular content that is often associated with malignancy.[34] Fibrinous stranding, septations, and loculations also suggest an exudative effusion (Figure 4A), and are more readily identified and characterized on lung ultrasound than CT scan.[35]

Figure 4

Homogenously echogenic effusions are most often due to hemothorax or empyema.[36] The high cell count of a hemothorax creates a layering effect in costophrenic recesses (hematocrit sign). Empyemas develop from complex effusions that organize into collections of pus and usually have a homogeneously echogenic, speckled appearance (Figure 4B). Sonographic evidence of septations in the presence of empyema predicts the need for intrapleural fibrinolytic therapy, longer duration of drainage, and possible surgical intervention.[37]

Isolated dense loculations may be challenging to differentiate from peripheral lung or pleural lesions, such as abscess or tumor.

Pleural Thickness

Normal visceral and parietal pleura are apposed and 0.2 to 0.3 mm thick.[38] Pleural effusions with parietal pleural thickness >10 mm, pleural nodularity, and diaphragmatic thickness >7 mm predicted underlying malignancy with high specificity and positive predictive value in 1 study.[39] As many as 20% of anechoic lesions of the pleura are solid rather than fluid. Color flow Doppler ultrasound can differentiate small pleural effusions from solid pleural abnormalities with sensitivity and specificity of 89% and 100%, respectively.[40]

Ascites

Ultrasound is the gold standard for diagnosing ascites and can detect as little as 100 mL of ascitic fluid.[3] When ascites is not immediately evident, hospitalists can apply the principles of the FAST (Focused Assessment with Sonography in Trauma) examination to detect small amounts of ascites by evaluating the most dependent areas of the abdominopelvic cavity, the hepatorenal, left subdiaphragmatic, and rectovesicular or rectouterine spaces.[1] When ascites is identified and paracentesis is indicated, ultrasound guidance for site selection reduces bleeding complications.[4]

Aortic Aneurysm

Novice providers with limited ultrasound training can accurately screen patients for abdominal aortic aneurysm (AAA). Multiple studies from emergency departments have shown that point‐of‐care ultrasound can be used to accurately detect AAA, and a recent meta‐analysis of 7 high‐quality studies demonstrated a sensitivity of 99% (95% confidence interval [CI]: 96%‐100%) and a specificity of 98% (95% CI: 97%‐99%).[5] Hospitalists could use ultrasound to rapidly detect AAA in patients with acute abdominal pain, monitor the size in patients with known AAA, and possibly screen high‐risk patients.[6]

Hydronephrosis

Once detected, relief of postrenal obstruction usually results in rapid reversal of acute kidney injury. Although diagnostic accuracy studies of detection of hydronephrosis have yet to be conducted with hospitalists, studies of other frontline providers with limited training in renal ultrasonography have revealed sensitivities of 72% to 87% and specificities of 73% to 82% in patients with renal colic.[7, 8]

Pneumonia

Normally aerated lung parenchyma generates A‐lines, horizontal hyperechoic lines that are artifacts due to repeated reflections, or reverberations, between the highly reflective pleura and transducer.[1] These normal A‐lines disappear with pneumonia due to accumulation of interstitial fluid and cellular exudate in consolidated alveoli. A meta‐analysis of 9 studies of lung ultrasound to diagnose pneumonia reported pooled sensitivity of 97% (95% CI: 93%‐99%) with specificity of 94% (95% CI: 85%‐98%).[13]

Pleural Effusion

Half of patients with community‐acquired pneumonia have a pleural effusion, yet chest x‐ray often cannot differentiate pneumonia from pleural effusion, especially along the lower lung fields. Ultrasound can accurately differentiate consolidated lung from pleural effusion and is more sensitive than a chest x‐ray for detecting small pleural fluid volumes (100% vs 71%).[14] Serial monitoring of size and character of a pleural effusion can distinguish free flowing from loculated pleural effusions. Drainage of pleural effusions with ultrasound guidance is associated with a lower rate of postprocedure pneumothorax and lower total hospital costs.[15]

Pneumothorax

Lung ultrasound can accurately and rapidly detect pneumothorax after lung and pleural procedures, including thoracentesis, bronchoscopy, and transthoracic biopsy.[2] Multiple studies have demonstrated that lung ultrasound is superior to chest x‐ray. Three recent meta‐analyses reported near‐perfect specificity for both ultrasound and x‐ray. But the sensitivity of ultrasound (79%95%) was far better than that of x‐ray (40%52%) to detect pneumothorax.[16, 17]

The hallmark ultrasound findings of pneumothorax include absence of lung sliding, absence of B‐lines, and a stratified pattern using M‐mode ultrasonography (stratosphere sign). Both lung sliding and B‐lines rule out pneumothorax with a negative predictive value of 100%.[18] Absence of either finding, however, does not rule in pneumothorax with similar strength. Absent lung sliding is seen in other conditions, such as pleurodesis, mainstem intubation, and massive atelectasis; absent B‐lines are most suggestive of the normal lung (see below).[1]

Pulmonary Edema

The classic ultrasound finding of acute pulmonary edema is bilateral anterior B‐lines. In contrast to horizontal A‐lines, B‐lines are vertical, laser‐like reverberations that originate from the pleura and are due to interlobular septal edema. A linear correlation has been shown between the quantity of B‐lines and radiographic lung water score (r=0.78; P<0.01).[19] Yet B‐lines are not specific for high pulmonary capillary wedge pressure because interstitial edema can be caused by a variety of etiologies. Nonetheless, visualization of multiple B‐lines in a single intercostal space corresponds with a sensitivity of 86% to 100% and specificity of 92% to 98% for either high‐ or low‐pressure pulmonary edema.[20, 21]

Central Venous Volume

The physiologic relationship between central venous volume and central venous pressure (CVP) is complex. Initially, there is upward stepwise progression to the stressed volume threshold, and then the relationship becomes curvilinear with the steepness of the slope dependent on the stiffness or tone of the central veins.[22]

The complexity of this relationship may explain the variable diagnostic accuracy of inferior vena cava (IVC) measurements to determine CVP, with measurements best reflecting CVP at extreme values. An IVC maximal diameter >2.0 cm predicted CVP >10 mm Hg (sensitivity 82% and specificity 84%) and pulmonary capillary wedge pressure >16 mm Hg (sensitivity 75% and specificity 83%) in 1 study.[23] Adding measurement of the collapsibility of the IVC with respiration may improve diagnostic accuracy, particularly with intermediate ranges of CVP and is recommended by current echocardiography guidelines.[24]

Nonetheless, in patients with acute dyspnea, a dilated, noncollapsing IVC may differentiate acute decompensated heart failure (ADHF) from primary pulmonary disease.[25, 26] IVC measurements may guide fluid removal in hemodialysis and heart failure patients.[27, 28] In 2 studies of patients hospitalized with ADHF, lack of improvement of IVC collapsibility index at the time of discharge was associated with higher rates of readmission.[29, 30] A follow‐up study comparing diuresis guided by IVC collapsibility to usual care in patients hospitalized with ADHF showed a reduction in hospital readmission rates (4% vs 30%, P=0.03) without an increase in hospital length of stay or renal dysfunction.[31] Patients with small, collapsed IVCs can be administered intravenous fluids safely, particularly in the setting of hypovolemic or septic shock, and the response to this fluid resuscitation can be assessed by serially measuring the change in IVC diameter.[32]

Thromboembolism

Multiple studies have shown that point‐of‐care ultrasound can accurately diagnose deep venous thrombosis (DVT) with a pooled sensitivity of 96% and specificity of 96% based on a recent meta‐analysis of 19 studies.[33] In symptomatic patients with a lung ultrasound pattern showing A‐lines, positive and negative predictive values of DVT in predicting pulmonary embolism (PE) were 94% and 98%, respectively.[34] A diagnostic accuracy study to diagnose PE using lung ultrasound to detect pleural‐ or subpleural‐based lesions yielded a sensitivity of 90%, specificity of 60%, positive predictive value of 80%, and negative predictive value of 78%.[35] In a study of 96 patients with suspected PE who underwent computed tomography pulmonary angiogram (CTPA), a focused ultrasound exam of the heart, lungs, and lower extremity veins was able to detect DVT (2.1%) or an alternative diagnosis (56.2%) in the majority of these patients, potentially obviating the need for CTPA in 58.4% of patients.[36] In addition, point‐of‐care cardiac ultrasound may reveal direct findings, such as free‐floating thrombus in the pulmonary artery, or indirect findings, such as right ventricular dilation and systolic dysfunction, septal bowing, McConnell's sign, or IVC dilation.[1] Cardiac abnormalities are more specific (88%94%) than sensitive (31%77%), and absence of cardiac abnormalities rules out massive PE, justifying withholding thrombolytic medications in most patients.[37]

Ascites

Ultrasound is the gold standard for diagnosing ascites and can detect as little as 100 mL of ascitic fluid.[3] When ascites is not immediately evident, hospitalists can apply the principles of the FAST (Focused Assessment with Sonography in Trauma) examination to detect small amounts of ascites by evaluating the most dependent areas of the abdominopelvic cavity, the hepatorenal, left subdiaphragmatic, and rectovesicular or rectouterine spaces.[1] When ascites is identified and paracentesis is indicated, ultrasound guidance for site selection reduces bleeding complications.[4]

Aortic Aneurysm

Novice providers with limited ultrasound training can accurately screen patients for abdominal aortic aneurysm (AAA). Multiple studies from emergency departments have shown that point‐of‐care ultrasound can be used to accurately detect AAA, and a recent meta‐analysis of 7 high‐quality studies demonstrated a sensitivity of 99% (95% confidence interval [CI]: 96%‐100%) and a specificity of 98% (95% CI: 97%‐99%).[5] Hospitalists could use ultrasound to rapidly detect AAA in patients with acute abdominal pain, monitor the size in patients with known AAA, and possibly screen high‐risk patients.[6]

Hydronephrosis

Once detected, relief of postrenal obstruction usually results in rapid reversal of acute kidney injury. Although diagnostic accuracy studies of detection of hydronephrosis have yet to be conducted with hospitalists, studies of other frontline providers with limited training in renal ultrasonography have revealed sensitivities of 72% to 87% and specificities of 73% to 82% in patients with renal colic.[7, 8]

Pneumonia

Normally aerated lung parenchyma generates A‐lines, horizontal hyperechoic lines that are artifacts due to repeated reflections, or reverberations, between the highly reflective pleura and transducer.[1] These normal A‐lines disappear with pneumonia due to accumulation of interstitial fluid and cellular exudate in consolidated alveoli. A meta‐analysis of 9 studies of lung ultrasound to diagnose pneumonia reported pooled sensitivity of 97% (95% CI: 93%‐99%) with specificity of 94% (95% CI: 85%‐98%).[13]

Pleural Effusion

Half of patients with community‐acquired pneumonia have a pleural effusion, yet chest x‐ray often cannot differentiate pneumonia from pleural effusion, especially along the lower lung fields. Ultrasound can accurately differentiate consolidated lung from pleural effusion and is more sensitive than a chest x‐ray for detecting small pleural fluid volumes (100% vs 71%).[14] Serial monitoring of size and character of a pleural effusion can distinguish free flowing from loculated pleural effusions. Drainage of pleural effusions with ultrasound guidance is associated with a lower rate of postprocedure pneumothorax and lower total hospital costs.[15]

Pneumothorax

Lung ultrasound can accurately and rapidly detect pneumothorax after lung and pleural procedures, including thoracentesis, bronchoscopy, and transthoracic biopsy.[2] Multiple studies have demonstrated that lung ultrasound is superior to chest x‐ray. Three recent meta‐analyses reported near‐perfect specificity for both ultrasound and x‐ray. But the sensitivity of ultrasound (79%95%) was far better than that of x‐ray (40%52%) to detect pneumothorax.[16, 17]

The hallmark ultrasound findings of pneumothorax include absence of lung sliding, absence of B‐lines, and a stratified pattern using M‐mode ultrasonography (stratosphere sign). Both lung sliding and B‐lines rule out pneumothorax with a negative predictive value of 100%.[18] Absence of either finding, however, does not rule in pneumothorax with similar strength. Absent lung sliding is seen in other conditions, such as pleurodesis, mainstem intubation, and massive atelectasis; absent B‐lines are most suggestive of the normal lung (see below).[1]

Pulmonary Edema

The classic ultrasound finding of acute pulmonary edema is bilateral anterior B‐lines. In contrast to horizontal A‐lines, B‐lines are vertical, laser‐like reverberations that originate from the pleura and are due to interlobular septal edema. A linear correlation has been shown between the quantity of B‐lines and radiographic lung water score (r=0.78; P<0.01).[19] Yet B‐lines are not specific for high pulmonary capillary wedge pressure because interstitial edema can be caused by a variety of etiologies. Nonetheless, visualization of multiple B‐lines in a single intercostal space corresponds with a sensitivity of 86% to 100% and specificity of 92% to 98% for either high‐ or low‐pressure pulmonary edema.[20, 21]

Central Venous Volume

The physiologic relationship between central venous volume and central venous pressure (CVP) is complex. Initially, there is upward stepwise progression to the stressed volume threshold, and then the relationship becomes curvilinear with the steepness of the slope dependent on the stiffness or tone of the central veins.[22]

The complexity of this relationship may explain the variable diagnostic accuracy of inferior vena cava (IVC) measurements to determine CVP, with measurements best reflecting CVP at extreme values. An IVC maximal diameter >2.0 cm predicted CVP >10 mm Hg (sensitivity 82% and specificity 84%) and pulmonary capillary wedge pressure >16 mm Hg (sensitivity 75% and specificity 83%) in 1 study.[23] Adding measurement of the collapsibility of the IVC with respiration may improve diagnostic accuracy, particularly with intermediate ranges of CVP and is recommended by current echocardiography guidelines.[24]

Nonetheless, in patients with acute dyspnea, a dilated, noncollapsing IVC may differentiate acute decompensated heart failure (ADHF) from primary pulmonary disease.[25, 26] IVC measurements may guide fluid removal in hemodialysis and heart failure patients.[27, 28] In 2 studies of patients hospitalized with ADHF, lack of improvement of IVC collapsibility index at the time of discharge was associated with higher rates of readmission.[29, 30] A follow‐up study comparing diuresis guided by IVC collapsibility to usual care in patients hospitalized with ADHF showed a reduction in hospital readmission rates (4% vs 30%, P=0.03) without an increase in hospital length of stay or renal dysfunction.[31] Patients with small, collapsed IVCs can be administered intravenous fluids safely, particularly in the setting of hypovolemic or septic shock, and the response to this fluid resuscitation can be assessed by serially measuring the change in IVC diameter.[32]

Thromboembolism

Multiple studies have shown that point‐of‐care ultrasound can accurately diagnose deep venous thrombosis (DVT) with a pooled sensitivity of 96% and specificity of 96% based on a recent meta‐analysis of 19 studies.[33] In symptomatic patients with a lung ultrasound pattern showing A‐lines, positive and negative predictive values of DVT in predicting pulmonary embolism (PE) were 94% and 98%, respectively.[34] A diagnostic accuracy study to diagnose PE using lung ultrasound to detect pleural‐ or subpleural‐based lesions yielded a sensitivity of 90%, specificity of 60%, positive predictive value of 80%, and negative predictive value of 78%.[35] In a study of 96 patients with suspected PE who underwent computed tomography pulmonary angiogram (CTPA), a focused ultrasound exam of the heart, lungs, and lower extremity veins was able to detect DVT (2.1%) or an alternative diagnosis (56.2%) in the majority of these patients, potentially obviating the need for CTPA in 58.4% of patients.[36] In addition, point‐of‐care cardiac ultrasound may reveal direct findings, such as free‐floating thrombus in the pulmonary artery, or indirect findings, such as right ventricular dilation and systolic dysfunction, septal bowing, McConnell's sign, or IVC dilation.[1] Cardiac abnormalities are more specific (88%94%) than sensitive (31%77%), and absence of cardiac abnormalities rules out massive PE, justifying withholding thrombolytic medications in most patients.[37]

Ascites

Ultrasound is the gold standard for diagnosing ascites and can detect as little as 100 mL of ascitic fluid.[3] When ascites is not immediately evident, hospitalists can apply the principles of the FAST (Focused Assessment with Sonography in Trauma) examination to detect small amounts of ascites by evaluating the most dependent areas of the abdominopelvic cavity, the hepatorenal, left subdiaphragmatic, and rectovesicular or rectouterine spaces.[1] When ascites is identified and paracentesis is indicated, ultrasound guidance for site selection reduces bleeding complications.[4]

Aortic Aneurysm

Novice providers with limited ultrasound training can accurately screen patients for abdominal aortic aneurysm (AAA). Multiple studies from emergency departments have shown that point‐of‐care ultrasound can be used to accurately detect AAA, and a recent meta‐analysis of 7 high‐quality studies demonstrated a sensitivity of 99% (95% confidence interval [CI]: 96%‐100%) and a specificity of 98% (95% CI: 97%‐99%).[5] Hospitalists could use ultrasound to rapidly detect AAA in patients with acute abdominal pain, monitor the size in patients with known AAA, and possibly screen high‐risk patients.[6]

Hydronephrosis

Once detected, relief of postrenal obstruction usually results in rapid reversal of acute kidney injury. Although diagnostic accuracy studies of detection of hydronephrosis have yet to be conducted with hospitalists, studies of other frontline providers with limited training in renal ultrasonography have revealed sensitivities of 72% to 87% and specificities of 73% to 82% in patients with renal colic.[7, 8]

Pneumonia

Normally aerated lung parenchyma generates A‐lines, horizontal hyperechoic lines that are artifacts due to repeated reflections, or reverberations, between the highly reflective pleura and transducer.[1] These normal A‐lines disappear with pneumonia due to accumulation of interstitial fluid and cellular exudate in consolidated alveoli. A meta‐analysis of 9 studies of lung ultrasound to diagnose pneumonia reported pooled sensitivity of 97% (95% CI: 93%‐99%) with specificity of 94% (95% CI: 85%‐98%).[13]

Pleural Effusion

Half of patients with community‐acquired pneumonia have a pleural effusion, yet chest x‐ray often cannot differentiate pneumonia from pleural effusion, especially along the lower lung fields. Ultrasound can accurately differentiate consolidated lung from pleural effusion and is more sensitive than a chest x‐ray for detecting small pleural fluid volumes (100% vs 71%).[14] Serial monitoring of size and character of a pleural effusion can distinguish free flowing from loculated pleural effusions. Drainage of pleural effusions with ultrasound guidance is associated with a lower rate of postprocedure pneumothorax and lower total hospital costs.[15]

Pneumothorax

Lung ultrasound can accurately and rapidly detect pneumothorax after lung and pleural procedures, including thoracentesis, bronchoscopy, and transthoracic biopsy.[2] Multiple studies have demonstrated that lung ultrasound is superior to chest x‐ray. Three recent meta‐analyses reported near‐perfect specificity for both ultrasound and x‐ray. But the sensitivity of ultrasound (79%95%) was far better than that of x‐ray (40%52%) to detect pneumothorax.[16, 17]

The hallmark ultrasound findings of pneumothorax include absence of lung sliding, absence of B‐lines, and a stratified pattern using M‐mode ultrasonography (stratosphere sign). Both lung sliding and B‐lines rule out pneumothorax with a negative predictive value of 100%.[18] Absence of either finding, however, does not rule in pneumothorax with similar strength. Absent lung sliding is seen in other conditions, such as pleurodesis, mainstem intubation, and massive atelectasis; absent B‐lines are most suggestive of the normal lung (see below).[1]

Pulmonary Edema

The classic ultrasound finding of acute pulmonary edema is bilateral anterior B‐lines. In contrast to horizontal A‐lines, B‐lines are vertical, laser‐like reverberations that originate from the pleura and are due to interlobular septal edema. A linear correlation has been shown between the quantity of B‐lines and radiographic lung water score (r=0.78; P<0.01).[19] Yet B‐lines are not specific for high pulmonary capillary wedge pressure because interstitial edema can be caused by a variety of etiologies. Nonetheless, visualization of multiple B‐lines in a single intercostal space corresponds with a sensitivity of 86% to 100% and specificity of 92% to 98% for either high‐ or low‐pressure pulmonary edema.[20, 21]

Central Venous Volume

The physiologic relationship between central venous volume and central venous pressure (CVP) is complex. Initially, there is upward stepwise progression to the stressed volume threshold, and then the relationship becomes curvilinear with the steepness of the slope dependent on the stiffness or tone of the central veins.[22]

The complexity of this relationship may explain the variable diagnostic accuracy of inferior vena cava (IVC) measurements to determine CVP, with measurements best reflecting CVP at extreme values. An IVC maximal diameter >2.0 cm predicted CVP >10 mm Hg (sensitivity 82% and specificity 84%) and pulmonary capillary wedge pressure >16 mm Hg (sensitivity 75% and specificity 83%) in 1 study.[23] Adding measurement of the collapsibility of the IVC with respiration may improve diagnostic accuracy, particularly with intermediate ranges of CVP and is recommended by current echocardiography guidelines.[24]

Nonetheless, in patients with acute dyspnea, a dilated, noncollapsing IVC may differentiate acute decompensated heart failure (ADHF) from primary pulmonary disease.[25, 26] IVC measurements may guide fluid removal in hemodialysis and heart failure patients.[27, 28] In 2 studies of patients hospitalized with ADHF, lack of improvement of IVC collapsibility index at the time of discharge was associated with higher rates of readmission.[29, 30] A follow‐up study comparing diuresis guided by IVC collapsibility to usual care in patients hospitalized with ADHF showed a reduction in hospital readmission rates (4% vs 30%, P=0.03) without an increase in hospital length of stay or renal dysfunction.[31] Patients with small, collapsed IVCs can be administered intravenous fluids safely, particularly in the setting of hypovolemic or septic shock, and the response to this fluid resuscitation can be assessed by serially measuring the change in IVC diameter.[32]

Thromboembolism

Multiple studies have shown that point‐of‐care ultrasound can accurately diagnose deep venous thrombosis (DVT) with a pooled sensitivity of 96% and specificity of 96% based on a recent meta‐analysis of 19 studies.[33] In symptomatic patients with a lung ultrasound pattern showing A‐lines, positive and negative predictive values of DVT in predicting pulmonary embolism (PE) were 94% and 98%, respectively.[34] A diagnostic accuracy study to diagnose PE using lung ultrasound to detect pleural‐ or subpleural‐based lesions yielded a sensitivity of 90%, specificity of 60%, positive predictive value of 80%, and negative predictive value of 78%.[35] In a study of 96 patients with suspected PE who underwent computed tomography pulmonary angiogram (CTPA), a focused ultrasound exam of the heart, lungs, and lower extremity veins was able to detect DVT (2.1%) or an alternative diagnosis (56.2%) in the majority of these patients, potentially obviating the need for CTPA in 58.4% of patients.[36] In addition, point‐of‐care cardiac ultrasound may reveal direct findings, such as free‐floating thrombus in the pulmonary artery, or indirect findings, such as right ventricular dilation and systolic dysfunction, septal bowing, McConnell's sign, or IVC dilation.[1] Cardiac abnormalities are more specific (88%94%) than sensitive (31%77%), and absence of cardiac abnormalities rules out massive PE, justifying withholding thrombolytic medications in most patients.[37]

Study Question

In selected populations of patients with suspected local or systemic infection, what are the effects of using procalcitonin measurement plus clinical criteria for infection to guide initiation, intensification, and/or discontinuation of antibiotic therapy when compared to clinical criteria for infection alone?

Search Strategy

MEDLINE and EMBASE were searched from January 1, 1990 through December 16, 2011, and the Cochrane Controlled Trials register was searched with no date restriction for randomized and nonrandomized comparative studies using the following search terms: procalcitonin AND chronic obstructive pulmonary disease; COPD; critical illness; critically ill; febrile neutropenia; ICU; intensive care; intensive care unit; postoperative complication(s); postoperative infection(s); postsurgical infection(s); sepsis; septic; surgical wound infection; systemic inflammatory response syndrome OR postoperative infection. In addition, a search for systematic reviews was conducted in MEDLINE, the Cochrane Database of Systematic Reviews, and Web sites of the National Institute for Clinical Excellence, the National Guideline Clearinghouse, and the Health Technology Assessment Programme. Gray literature, including databases with regulatory information, clinical trial registries, abstracts and conference papers, grants and federally funded research, and manufacturing information was searched from January 1, 2006 to June 28, 2011.

Study Selection

A single reviewer screened abstracts and selected studies looking at procalcitonin‐guided antibiotic therapy. Second and third reviewers were consulted to screen articles when needed. Studies were included if they fulfilled all of the following criteria: (1) randomized, controlled trial or nonrandomized comparative study; (2) adult and/or pediatric patients with known or suspected local or systemic infection, including critically ill patients with sepsis syndrome or ventilator‐associated pneumonia, adults with respiratory tract infections, neonates with sepsis, children with fever of unknown source, and postoperative patients at risk of infection; (3) interventions included initiation, intensification, and/or discontinuation of antibiotic therapy guided by procalcitonin plus clinical criteria; (4) primary outcomes included antibiotic usage (antibiotic prescription rate, total antibiotic exposure, duration of antibiotic therapy, and days without antibiotic therapy); and (5) secondary outcomes included morbidity (antibiotic adverse events, hospital and/or intensive care unit length of stay), mortality, and quality of life.

Studies with any of the following criteria were excluded: published in non‐English language, not reporting primary data from original research, not a randomized, controlled trial or nonrandomized comparative study, not reporting relevant outcomes.

Data Extraction and Quality Assessment

A single reviewer abstracted data and a second reviewer confirmed accuracy. Disagreements between reviewers were resolved by group discussion among the research team and final quality rating was assigned by consensus adjudication. Data elements were abstracted into the following categories: quality assessment, applicability and clinical diversity assessment, and outcome assessment. Quality of included studies was assessed using the US Preventive Services Task Force framework[17] by at least 2 independent reviewers. Three quality categories were used: good, fair, and poor.

Data Synthesis and Analysis

The decision to incorporate formal data synthesis in this review was made after completing the formal literature search, and the decision to pool studies was based on the specific number of studies with similar questions and outcomes. If a meta‐analysis could be performed, subgroup and sensitivity analyses were based on clinical similarity of available studies and reporting of mean and standard deviation. The pooling method involved inverse variance weighting and a random effects model.

The strength of evidence was graded using the Methods Guide,[16] a system based on the Grading of Recommendations Assessment, Development and Evaluation Working Group.[18] The following domains were addressed: risk of bias, consistency, directness, and precision. The overall strength of evidence was graded as high, moderate, low, or insufficient. The final strength of evidence grading was made by consensus adjudication among the authors.

Adult ICU Patients: Procalcitonin‐Guided Antibiotic Discontinuation

Five studies[19, 20, 21, 22, 23] (N=938) addressed procalcitonin‐guided discontinuation of antibiotic therapy in adult ICU patients. Four studies conducted superiority analyses for mortality with procalcitonin‐guided therapy, whereas 1 study conducted a noninferiority analysis. Absolute procalcitonin values for discontinuation of antibiotics ranged from 0.25 to 1 ng/mL. Physicians in control groups administered antibiotics according to their standard practice.

Antibiotic Usage

The absolute reduction in duration of antibiotic usage with procalcitonin guidance in these studies ranged from 1.7 to 5 days, and the relative reduction ranged from 21% to 38%. Meta‐analysis of antibiotic duration in adult ICU patients was performed (Figure 2A).

Figure 2

Adult ICU Patients: Procalcitonin‐Guided Antibiotic Intensification

Two studies[24, 33] (N=1272) addressed procalcitonin‐guided intensification of antibiotic therapy in adult ICU patients. The Jensen et al. study[33] was a large (N=1200), high‐quality study that used a detailed algorithm for broadening antibiotic therapy in patients with elevated procalcitonin. The Jensen et al. study also educated physicians about empiric therapy and intensification of antibiotic therapy. A second study[24] was too small (N=72) and lacked sufficient details to be informative.

Adult Patients With Respiratory Tract Infections

Eight studies[25, 26, 27, 28, 29, 30, 34, 35] (N=3492) addressed initiation and/or discontinuation of antibiotics in adult patients with acute upper and lower respiratory tract infections, including community‐acquired pneumonia, acute exacerbation of chronic obstructive pulmonary disease, and acute bronchitis. Settings included primary care clinics, emergency departments, and hospital wards. Physicians in control groups administered antibiotics according to their own standard practices and/or evidence‐based guidelines. All studies encouraged initiation of antibiotics with procalcitonin levels >0.25 ng/mL, and 4 studies strongly encouraged antibiotics with procalcitonin levels >0.5 ng/mL.

Antibiotic Usage

Procalcitonin guidance reduced antibiotic duration, antibiotic prescription rate, and total antibiotic exposure. Absolute reduction in antibiotic duration ranged from 1 to 7 days, and relative reductions ranged from 13% to 55%. Four of the 8 studies reported sufficient details to be pooled into a meta‐analysis (Figure 3A) with a statistically significant pooled mean difference of 2.35 days favoring procalcitonin (95% CI: 4.38 to 0.33). Procalcitonin guidance also reduced antibiotic prescription rate with absolute reductions ranging from 2% to 7% and relative reductions ranging from 1.8% to 72%. Meta‐analysis of prescription rates from 8 studies (Figure 3B) yielded a statistically significant pooled risk difference of 22% (95% CI: 41% to 4%). Total antibiotic exposure was consistently reduced in the 4 studies reporting this outcome.

Figure 3

Neonates With Sepsis

One study[31] (N=121) evaluated procalcitonin‐guided antibiotic therapy for suspected neonatal sepsis. Neonatal sepsis was suspected on the basis of risk factors and clinical signs and symptoms. Antibiotic initiation or discontinuation was based on a procalcitonin nomogram. Antibiotic therapy in the control group was based on the physician's assessment. The quality of this study was rated good, and strength of evidence was rated moderate for antibiotic usage and insufficient for morbidity and mortality outcomes.

Children Ages 1 to 36 Months With Fever of Unknown Source

One study[36] (N=384) evaluated procalcitonin‐guided antibiotic therapy for fever of unknown source in children 1 to 36 months of age, but the overall strength of evidence was judged insufficient to draw conclusions.

Adult Postoperative Patients at Risk of Infection

One study[32] (N =250) monitored procalcitonin in consecutive patients after colorectal surgery to identify patients at risk of infection who might benefit from prophylactic antibiotic therapy. Two hundred thirty patients had normal procalcitonin levels. Twenty patients with elevated procalcitonin levels (>1.5 ng/mL) were randomized to receive prophylactic antibiotic therapy with ceftriaxone or no antibiotics. The strength of evidence was judged insufficient to draw conclusions from this study.

Summary of the Main Findings

Diagnosis of sepsis or other serious infections in critically ill patients is challenging because clinical criteria for diagnosis overlap with noninfectious causes of the systemic inflammatory response syndrome. Initiation of antibiotic therapy for presumed sepsis is necessary while diagnostic evaluation is ongoing, because delaying antibiotic therapy is associated with increased mortality.[37, 38, 39] Our review found that procalcitonin guidance significantly reduced antibiotic usage in adult ICU patients by reducing the duration of antibiotic therapy, rather than decreasing the initiation of antibiotics, without increasing morbidity or mortality.

In contrast, the use of procalcitonin as an indicator of need for intensification of antibiotic therapy in adult ICU patients should be discouraged because this approach was associated with increased morbidity. The large, well‐designed study by Jensen[33] showed that antibiotic intensification in response to elevated procalcitonin measurement was associated with increased morbidity: a longer ICU LOS, an increase in days on mechanical ventilation, and an increase in days with abnormal renal function. The authors concluded that the increased morbidity could only be explained by clinical harms of increased exposure to broad‐spectrum antibiotics.

Clinical and microbiological evaluations are neither sensitive nor specific for differentiating bacterial from viral respiratory tract infections. Procalcitonin can guide initiation of antibiotic therapy in adults with suspected bacterial respiratory tract infection. Our review showed that procalcitonin guidance significantly reduced antibiotic usage with respect to antibiotic prescription rate, duration of antibiotic therapy, and total exposure to antibiotic therapy in adult patients with respiratory tract infections.

The role of procalcitonin‐guided therapy in other populations is less clear. One study in postoperative colorectal surgery patients reported that elevated procalcitonin levels may identify patients at risk for infection who benefit from prophylactic antibiotic therapy.[32] Patients with elevated procalcitonin levels who received prophylactic antibiotic therapy had a significant decrease in the incidence and severity of systemic infections, whereas patients with normal procalcitonin levels did not require any additional surgical or medical therapy. Although these findings are promising, more data in postoperative patients are needed.

The utility of procalcitonin in pediatric settings is a significant gap in the present literature. One study[31] in neonates with suspected sepsis showed a significant decrease in the proportion of neonates started on empiric antibiotic therapy and a decrease in the duration of antibiotic therapy with procalcitonin guidance. However, there was insufficient evidence that procalcitonin guidance does not increase morbidity or mortality.

Comparison to Other Systematic Reviews

Six systematic reviews of procalcitonin guidance in the management of patients with infections were published prior to our review.[9, 10, 11, 12, 13, 14] Our systematic review differs from past reviews in the number of studies included and the pooling of studies according to patient population, type and severity of infection, and different uses of procalcitonin measurements, either for initiation, discontinuation, or intensification of antibiotic therapy. Previous systematic reviews included 7 to 14 studies, whereas ours included 18 randomized, controlled trials. One previous review[13] included and pooled the Jensen et al. study[33] with other studies of adult ICU patients. We evaluated the Jensen et al. study separately because it uniquely looked at procalcitonin‐guided antibiotic intensification in adult ICU patients, in contrast to other studies that looked at procalcitonin‐guided antibiotic discontinuation. We addressed pediatric populations separately from adult patients, and recognizing that there are distinct groups within the pediatric population, we separately grouped neonates and children ages 1 to 36 months. Despite these differences, our review and other systematic reviews, we came to similar conclusions: procalcitonin‐guided antibiotic decision making compared to clinical criteria‐guided antibiotic decision making reduces antibiotic usage without increasing morbidity or mortality.

Limitations

An important limitation of this review was the uncertainty about the noninferiority margin for morbidity and mortality in adult ICU patients. Only the Bouadma et al. study[23] did a power analysis and predefined a margin for noninferiority for 28‐ and 60‐day mortality. Meta‐analysis of all 5 ICU studies showed a pooled point estimate of 0.43% in mortality and a 95% CI of 6% to 5% for difference in mortality between procalcitonin‐guided therapy versus standard care. A 10% noninferiority margin for mortality has been recommended by the Infectious Diseases Society of America and American College of Chest Physicians, but there is concern that a 10% margin for mortality may be too high. Presently, 2 large trials are in progress that may yield more precise estimates of mortality in the future.

Differences in reporting of total antibiotic exposure and morbidity outcomes limited our ability to pool data. Total antibiotic exposure is conventionally reported as mean days per 1000 days of follow‐up, but some studies only reported relative or absolute differences. Likewise, morbidity was reported with different severity of illness scales, including Sepsis‐Related Organ Failure Assessment, Simplified Acute Physiology (SAP) II, SAP III, and Acute Physiology and Chronic Health Evaluation II, which limited comparisons across studies.

Research Gaps

We identified gaps in the available literature and opportunities for future research. First, the safety and efficacy of procalcitonin‐guided antibiotic therapy needs to be studied in patient populations excluded from current randomized controlled studies, such as immunocompromised patients and pregnant women. Patients who are immunocompromised or have chronic conditions, such as cystic fibrosis, account for a significant percentage of community‐acquired respiratory tract infections and are often treated empirically.[29, 30] Second, standardized reporting of antibiotic adverse events and emergence of antibiotic resistance is needed. Strategies to reduce antibiotic usage have been associated with reductions in antibiotic adverse events, such as Clostridium difficile colitis and superinfection with multi‐drug resistant Gram‐negative bacteria.[37, 40, 41] Few studies in our review reported allergic reactions or adverse events of antibiotic therapy, [25, 27, 34] and only 1 reported antibiotic resistance.[19] Third, procalcitonin guidance should be compared to other strategies to reduce antibiotic usage, such as structured implementation of practice guidelines and antibiotic stewardship programs.[42] One single‐arm study describes how procalcitonin can be used in antibiotic stewardship programs to decrease the duration of antibiotic therapy,[43] but additional studies are needed. Finally, generalizing results from those studies that were conducted primarily in Europe would depend on similar use of and adherence to study‐based algorithms. Newer observational studies have demonstrated reduced antibiotic usage with implementation of procalcitonin algorithms in real‐life settings in Europe, but algorithm adherence was significantly less in the United States.[44, 45]

In summary, our systematic review found that procalcitonin‐guided antibiotic therapy can significantly reduce antibiotic usage in adult ICU patients without affecting morbidity or mortality. Procalcitonin should not be used to guide intensification of antibiotic therapy in adult ICU patients because this approach may increase morbidity. In adults with respiratory infections, procalcitonin guidance can significantly reduce antibiotic usage without adversely affecting morbidity or mortality. There is insufficient evidence to recommend procalcitonin‐guided antibiotic therapy in neonates with sepsis, children with fever of unknown source, or postoperative patients at risk for infection.

Study Question

In selected populations of patients with suspected local or systemic infection, what are the effects of using procalcitonin measurement plus clinical criteria for infection to guide initiation, intensification, and/or discontinuation of antibiotic therapy when compared to clinical criteria for infection alone?

Search Strategy

MEDLINE and EMBASE were searched from January 1, 1990 through December 16, 2011, and the Cochrane Controlled Trials register was searched with no date restriction for randomized and nonrandomized comparative studies using the following search terms: procalcitonin AND chronic obstructive pulmonary disease; COPD; critical illness; critically ill; febrile neutropenia; ICU; intensive care; intensive care unit; postoperative complication(s); postoperative infection(s); postsurgical infection(s); sepsis; septic; surgical wound infection; systemic inflammatory response syndrome OR postoperative infection. In addition, a search for systematic reviews was conducted in MEDLINE, the Cochrane Database of Systematic Reviews, and Web sites of the National Institute for Clinical Excellence, the National Guideline Clearinghouse, and the Health Technology Assessment Programme. Gray literature, including databases with regulatory information, clinical trial registries, abstracts and conference papers, grants and federally funded research, and manufacturing information was searched from January 1, 2006 to June 28, 2011.

Study Selection

A single reviewer screened abstracts and selected studies looking at procalcitonin‐guided antibiotic therapy. Second and third reviewers were consulted to screen articles when needed. Studies were included if they fulfilled all of the following criteria: (1) randomized, controlled trial or nonrandomized comparative study; (2) adult and/or pediatric patients with known or suspected local or systemic infection, including critically ill patients with sepsis syndrome or ventilator‐associated pneumonia, adults with respiratory tract infections, neonates with sepsis, children with fever of unknown source, and postoperative patients at risk of infection; (3) interventions included initiation, intensification, and/or discontinuation of antibiotic therapy guided by procalcitonin plus clinical criteria; (4) primary outcomes included antibiotic usage (antibiotic prescription rate, total antibiotic exposure, duration of antibiotic therapy, and days without antibiotic therapy); and (5) secondary outcomes included morbidity (antibiotic adverse events, hospital and/or intensive care unit length of stay), mortality, and quality of life.

Studies with any of the following criteria were excluded: published in non‐English language, not reporting primary data from original research, not a randomized, controlled trial or nonrandomized comparative study, not reporting relevant outcomes.

Data Extraction and Quality Assessment

A single reviewer abstracted data and a second reviewer confirmed accuracy. Disagreements between reviewers were resolved by group discussion among the research team and final quality rating was assigned by consensus adjudication. Data elements were abstracted into the following categories: quality assessment, applicability and clinical diversity assessment, and outcome assessment. Quality of included studies was assessed using the US Preventive Services Task Force framework[17] by at least 2 independent reviewers. Three quality categories were used: good, fair, and poor.

Data Synthesis and Analysis

The decision to incorporate formal data synthesis in this review was made after completing the formal literature search, and the decision to pool studies was based on the specific number of studies with similar questions and outcomes. If a meta‐analysis could be performed, subgroup and sensitivity analyses were based on clinical similarity of available studies and reporting of mean and standard deviation. The pooling method involved inverse variance weighting and a random effects model.

The strength of evidence was graded using the Methods Guide,[16] a system based on the Grading of Recommendations Assessment, Development and Evaluation Working Group.[18] The following domains were addressed: risk of bias, consistency, directness, and precision. The overall strength of evidence was graded as high, moderate, low, or insufficient. The final strength of evidence grading was made by consensus adjudication among the authors.