Diagnostic accuracy of lung ultrasonography combined with procalcitonin for the diagnosis of pneumonia: a pilot study

Study subjects and design

This is a prospective accuracy study, in which patients enrolled in a registry from December 2011 to August 2012 were included if PCT level at admission was available [8]. Patients were recruited in the ED with an annual census of 130,000 visits. The study was approved by the local institutional review board. Written informed consent was obtained for inclusion in the study.

Consecutive patients aged >18 years with at least one unexplained respiratory complaint among dyspnea, chest pain, cough or hemoptysis with or without fever, for which the attending emergency physician ordered a chest CT were considered in the study. PCT was ordered by the attending emergency physician, independent from the study protocol, when a infective cause was suspected.

Methods

Lung ultrasonography was performed before and within 3 h from chest CT by one of eight sonographer investigators who participated to the study. The investigators were four internal and emergency medicine staff physicians with at least 5 years experience on point-of-care emergency ultrasonography and four resident physicians (two in Internal Medicine and two in Emergency Medicine) with at least 6 months training in emergency ultrasound. The investigators were aware of the presenting symptoms and the evident physical signs, but were blinded to all the other general clinical information including any radiologic finding and laboratory results. The following multi-probe machines were used: MyLab30 Gold (Esaote, Genova, Italy) and HD7 (Philips, Amsterdam, Holland). LUS was performed by using a 4–8 MHz linear probe or a 3.5–5 MHz curved array probe. The lung was examined by longitudinal and oblique scans on anterior, lateral and posterior chest. The patient was preferably examined in the sitting position. When this latter position could not be maintained, due to critical clinical conditions or poor compliance, the examination was performed in the supine or semi-recumbent position. The posterior areas were scanned in the sitting position or, when not feasible, by turning the patient in the lateral decubitus on both sides. The LUS examination was targeted to the detection of typical subpleural lung consolidations with tissue-like or anechoic pattern and blurred, irregular margins with associated focal B-lines [9]. The consolidations due to infection usually show dynamic air bronchograms (branching echogenic structures with centrifuge movement with breathing) or multiple hyperechogenic lentil-sized spots, due to air trapped in the small airways. LUS was considered positive when at least one consolidation showing the above-described features was detected. Pleural effusion could be associated with consolidation and has been interpreted as an auxiliary sign, but its detection was not considered individually as sign of suspected pneumonia.

Chest radiography was performed using a Practix 300 Bucky diagnostic equipment (Philips Medical Systems, Hamburg, Germany) by posterior-anterior and lateral views in the upright patients and anterior-posterior view in the supine patients, following standardized hospital diagnostic protocol. The film was digitally reviewed by an expert radiologist blinded to the results of LUS and CT. The radiologist had the possibility to review also previous CXR, when available, as part of standard clinical care. The radiologist was asked to detect and locate the opacities that might correlate with the diagnosis of pneumonia. CXR was considered positive when at least one typical consolidation was visualized.

Procalcitonin, requested by the treating physician at ED presentation for clinical reason, was measured using an automated Enzyme-Linked Fluorescent Assay (ELFA): VIDAS^® B.R.A.H.M.S PCT™ (Vidas, BioMerieux). Samples were processed by personnel blinded from any patient data. We considered two PCT cut-off of 0.25 and 0.5 ng/ml based on previous studies [6, 7, 10–12]. These studies suggested that a PCT level <0.25 ng/ml can be used to rule out bacterial pneumonia in patients with suspected respiratory tract infections and a PCT level >0.5 ng/ml is considered to rule in bacterial pneumonia and to support antibiotic therapy. Patients that didn’t consent or didn’t undergo LUS within the time limit were excluded from the study. Patients in which PCT was not requested in ED were also excluded.

Chest CT was performed by one Somatom Definition AS128 (Siemens, Erlangen, Germany), only for clinical purposes independent from the study protocol. An expert radiologist, other than the attending radiologist and blinded to LUS and CXR results, reviewed the studies investigating one or more consolidations related to pneumonia, defined as homogeneous increase in pulmonary parenchymal attenuation obscuring the margins of vessels and airway walls, with or without air bronchograms [13].

The final diagnoses of pneumonia were determined by two expert internal medicine physicians, blinded to CXR, LUS and PCT results, who independently reviewed all available clinical data including CT data and medical records for hospitalized patients. In case of discordance, a third senior physician adjudicated the diagnosis.

Statistical analysis

The sample size was calculated considering a prevalence of pneumonia of 50% among suspected patients in ED and a sensitivity of LUS for the diagnosis of pneumonia when chest CT was used as gold standard of 80% [8]. Wanting to prove that PCT increased the sensitivity of 10% and considering a type I error of 5% and a power (1-ß) of 80%, it was estimated that this would require approximately 120 patients.

Data points are expressed as mean ± standard deviation (SD). The unpaired Student’s t test was used to compare normally distributed data. Fisher’s exact test was used for the comparison of non-continuous variables expressed as proportions. p < 0.05 indicates statistical significance. All p values are two sided. The diagnostic performance of LUS, CXR and PCT and the combination of LUS and CXR with PCT was assessed by calculating sensitivity, specificity, positive predictive value, negative predictive value and likelihood ratios. The extended McNemar and the McNemar tests were used to evaluate if there was a significant difference in the accuracy, sensitivities and specificities of LUS, CXR, PCT and the combination of imaging tests with PCT [14]. Receiver operating characteristic (ROC) curves were constructed to assess the sensitivity and specificity of white blood cells count (WBC), PCT and multivariable models to compare the ability of LUS and CXR combined with PCT to predict pneumonia. Calculations were performed by using SPSS statistical package (version 17.0; IBM).

Baseline characteristics

A total of 308 patients with respiratory complains underwent chest CT in the ED during the study period. Four patients did not consent to participate. In 19 patients LUS was not performed before chest CT or within the time limit. In 157 patients PCT was not requested by the attending emergency physician at presentation in the ED. Thus, 128 patients were included in the final analysis (Fig. 1). These patients had a mean age of 70.7 ± 15.8 years (range 23–100) and 64 (50%) were female. Pneumonia was the final diagnosis in 61 patients (47.7%). The patients’ characteristics according to the presence or absence of pneumonia are shown in Table 1. Alternative diagnoses in patients without pneumonia are reported in Table 2.

	Pneumonia (n = 61)	No pneumonia (n = 67)	p value
Mean age ± SD, y	69.8 ± 18.8	71.6 ± 12.6	0.51
Women	25 (41%)	39 (58.2)	0.076
Medical history of lung disease
Chronic obstructive pulmonary disease	7 (11.5)	10 (14.9)	0.611
Lung cancer	2 (3.3)	5 (7.5)	0.444
Heart failure	11 (18)	14 (20.9)	0.824
Previous pulmonary embolism	1 (1.6)	3 (4.5)	0.621
Signs and symptoms at presentation
Dyspnea	44 (72.1)	49 (73.1)	1
Cough	26 (42.6)	12 (31.6)	0.003
Pleuritic chest pain	8 (13.1)	8 (11.9)	1
Shock/hypotension	17 (27.9)	13 (19.4)	0.3
SBP ± SD, mmHg	121.7 ± 30.5	124.9 ± 37.3	0.6
Heart rate >100 bpm	42 (68.9)	30 (44.8)	0.008
Respiratory rate/min	26.1 ± 8.2	23.3 ± 6.8	0.2
PaO2/FiO2 ± SD	292.6 ± 83.9	309.5 ± 94	0.29
Temperature >37.8 °C	31 (50.8)	22 (32.8)	0.049
White blood cells count/mm3	12,494 ± 10,130.3	11,906 ± 6657	0.7
Mechanical invasive ventilation	4 (6.6)	6 (9)	0.747

Data are given as no. (%) or mean ± standard deviation (SD)

SD standard deviation; SBP systolic blood pressure; PaO ₂ arterial oxygen pressure express in mmHg; FiO ₂ fraction of inspired oxygen; bpm beats/min; ED emergency department

Diagnosis	No (%)
Pneumonia	61 (47.7)
Pulmonary embolism	21 (16.4)
Heart failure	9 (7)
Pleural effusion	8 (6.3)
Sepsis with no pulmonary involvement	8 (6.2)
Chronic obstructive pulmonary disease	4 (3.1)
Acute coronary syndrome	3 (2.3)
Tachyarrhythmia	1 (0.8)
Respiratory failure in pulmonary malignancy	1 (0.8)
Pericardial effusion/pericarditis	1 (0.8)
Miscellaneous	11 (8.6)

Data are given as no. (%) unless otherwise indicated

Main results

Lung ultrasonography was feasible in all patients, but in 11 cases (8.2%), included in the study, pulmonary examination was limited only to the anterior-lateral areas mainly due to uncooperative patients. LUS was positive for at least one consolidation in 60 patients (46.9%). Fifty-two (86.7%) out of 60 patients with positive LUS had a final diagnosis of pneumonia (Fig. 1). LUS was false positive in eight cases (6.2%) and false negative in nine patients (7%). In three out of nine (33%) false negative cases pulmonary examination was limited only to the anterior-lateral areas. The sensitivity and the specificity of LUS were 85.2% (95% CI 73.8–93%) and 88.1% (95% CI 77.8–94.7%) respectively.

Procalcitonin median in patients with and without pneumonia were 0.3 and 0.1 ng/ml respectively (p = 0.05). The sensitivity and the specificity of PCT ≥ 0.25 ng/ml were 73.8% (95% CI 61–84.2%) and 47.8% (95% CI 35.4–60.3%) respectively, whereas the sensitivity and the specificity of PCT > 0.5 ng/ml were 42.6% (95% CI 30–55.9%) and 67.2% (95% CI 54.6–78.1%) respectively.

Chest radiography was performed with the patient in the up-right position in 29 (23%) cases. Thirty-seven out of 45 patients (82.2%) with positive CXR had a final diagnosis of pneumonia. CXR showed false positive examination in 8 (6.2%) patients and false negative examination in 24 (18.7%) patients. The sensitivity and the specificity of CXR were 60.7% (95% CI 47.3–72.9%) and 88.1% (95% CI 77.8–94.7%) respectively.

In 38 cases (29.7%) we observed the combination of negative LUS and negative PCT (<0.25 ng/ml). A false negative combination of LUS/PCT test was observed in 2 cases (1.6%). In one of these two patients the final diagnosis was infective endocarditis and heart failure associated to pneumonia. The other patient had a final diagnosis of pneumonia complicated by heart failure and non-ST elevated myocardial infarction. The derived sensitivity of negative LUS/PCT test was 96.7% (95% CI 88.6–99.5%) with a significant superiority to LUS alone and PCT alone (p < 0.05 for both). Negative predictive value (NPV) and negative likelihood ratio (LR) were 94.7% (95% CI 88.2–99.2%) and 0.06 (95% CI 0.02–0.24) respectively. In 24 patients (18.8%) we observed the combination of positive LUS and positive PCT (>0.5 ng/ml). Out of these, four were false positive (3.1%). The derived specificity, positive predictive value (PPV) and positive LR were 94% (95% CI 85.4–98.3%), 83.3% (95% CI 62.6–95.2%) and 5.49 (95% CI 1.99–15.17) respectively. Specificity of the LUS/PCT test was not significantly different when compared to LUS alone (p = 0.125).

In 46 patients (35.9%), CXR was negative for pneumonia and PCT was <0.25 ng/ml. False negative cases where 12 (9.4%). The derived sensitivity, NPV and negative LR were 80.3% (95% CI 68.1–89.4%), 73.9% (95% CI 58.9–85.7%) and 0.39 (95% CI 0.22–0.68) respectively. In 16 patients (12.5%) CXR was positive for pneumonia and PCT was >0.5 ng/ml (16, 12.5%), with one false positive case (0.8%). The derived specificity, PPV and positive LR were 98.5% (95% CI 91.9–99.7%), 93.7% (95% CI 69.7–98.9%) and 16.48 (95% CI 2.24–121.05%) respectively.

According to ROC curve, the diagnostic accuracy of LUS associated with PCT (88.8% ± 3.1) was higher than the total count of white blood cells (WBC) (49.1% ± 5.2), PCT alone (60.7% ± 5) and CXR associated with PCT (77.4% ± 4.2) (Fig. 2).

Study limitations

A questionable characteristic of our protocol is that patients were not enrolled on the basis of a direct suspicion for pneumonia. We concentrated our study on patients with respiratory complains, who demanded CT scan for ambiguous diagnosis. In case of suspected uncomplicated lung infection, where the diagnosis is mainly based on clinical findings and CXR, it is unusual to undergo CT scans. Thus, our specific patient population does not allow generalization of our results. However, a systematic application of CT studies in all suspected low respiratory tract infection is not feasible because it does not represent a standard of care, and it is not ethical for its radiation burden and not sustainable for high cost. Another limitation of our study, is that the final diagnosis of pneumonia was based on the detection of a consolidation at chest CT, while the possibility of interstitial pneumonia was not considered. We have also to consider that it is not known if the consolidation identified at chest CT was due to bacterial, viral or other germ infection. Regarding the comparison between LUS and CXR, a consideration should be done on the criteria of enrolment. In fact, the patients enrolled in the study underwent chest CT once the physician was aware of CXR results, and this could have selected a population with not diagnostic CXR. Thus, this criteria probably affected the low sensitivity of CXR. However, in the real practice, the role of LUS is often to confirm negative chest films in patients with high suspicion of pneumonia. Thus, the difference in sensitivity between the two methodologies is still confirmed. Indeed, the fact that in our study the majority of CXR were not performed in the up-right position, may have influenced more significantly the comparison between the two methods. The anterior-posterior radiographic view is undoubtedly less accurate for the study of the lungs. However, the evaluation of sub-optimal chest films is largely representative of the real daily practice and the main cause of inconclusive CXR studies. Finally, the lack of systematic and consecutive PCT assays in the population studied may have biased our results. Indeed, a PCT assay was only based on the personal judgement of the attending physician facing challenging diagnostic situations and was independent from the study protocol. However, these situations are exactly those deserving more accurate and sometimes alternative diagnostic tools, that is the possible future role of a combined LUS/PCT protocol.