Patients
This is prospective cross-sectional study conducted in the Emergency Department (ED) of three University Hospitals (Fattouma Bourguiba University Hospital, Sahloul University Hospital, and Farhat Hached University Hospital, Tunisia) from January 2016 to October 2017.
A convenience sampling approach, including all patients admitted to the ED for acute dyspnea as chief complaint, was used. Exclusion criteria were: age less than 18 years, impossibility to give consent to participate in the study, post-traumatic dyspnea, pregnant women, and need for endotracheal intubation or inotropic drugs patients who were deemed too unstable for sonography by the treating team were also excluded.
Methodology
All eligible patients underwent a complete physical examination. Blood pressure, heart rate, and pulse oximetry were measured and oxygen was delivered by face mask as needed. Research associates collected the following data: name, age, sex, previous medical history, ongoing treatment, and physical examination findings. The following additional tests were performed for all included patients: blood gas, hemoglobin, serum creatinine, BNP, electrocardiogram, chest X-ray, and echocardiogram. Lung ultrasonography was performed by EM residents using two ultrasound machines (Philips EnVisor C, Nederland; SonoSite M-Turbo, Sonosite Inc., Bothell, WA) and broadband curved array probes (3.5–5 MHz). The study period overlapped one and half academic year in three university hospitals, so a total of 40 residents were eligible to participate. ED residents were appointed to carry out this examination less than 4 h following patients’ admission. None of the ED residents used LUS for the assessment of B-lines prior to the study. All participating residents were previously attended a 2-h training session with at least 10 clinical tests supervised by a certified emergency physicians who had accomplished a full mentoring program for “Ultra-Sound Life Support”. The first 30 min of the training course included basic ultrasound physics, use of ultrasound equipment, probe positioning, and lung ultrasound interpretation (A-lines, B-lines, consolidation, lung sliding, lung pulse, and miscellaneous artifacts). In the second 30 min, real-time LUS was performed in healthy volunteers describing the technique and findings. The rest of the training was hands-on training on actual patients. Trainees had to identify the presence of lung sliding, A-lines, B-lines and consolidation.
For each patient, two LUS tests were performed by two independent residents who were not aware of patient's clinical data and did not participate in the patient’s management. We recorded the ED residents’ interpretation and images were recorded for each LUS study for later expert review. To not break the blind protocol, patients were asked to not provide information on their medical history to the operators during LUS. Patients were placed in a semi-recumbent or supine position depending on their respiratory tolerance. For each side of the chest, 4 zones have to be assessed (Fig. 1): 2 anterior and 2 lateral. The anterior chest wall was delineated from the sternum to the anterior axillary line and was subdivided into upper and lower halves (approximately from clavicle to the second–third intercostal spaces and from the third space to diaphragm). The lateral chest was delineated from the anterior to the posterior axillary line and was subdivided into upper and basal halves. The operator was asked to calculate the B-lines score which is the sum of the B-lines found in both sides (8 zones) [14]; the intercostal space with the greatest number of B-lines within each zone was used for scoring. B-line was defined as a vertical bright echogenic bundle with a narrow basis, spreading from the transducer to the deepest part of the screen (Fig. 2). For B-lines that were wide or confluent, the score was determined by assessing the percentage of the rib space occupied by B-lines and dividing it by ten [10].
According to the study of Gargani et al. the B-lines score is suggestive of CHF when it is ≥ 15 [15]. The probability of CHF was also expressed according to the following ordinal scale: unlikely if B-lines score < 15, likely if B-lines score is between 16 and 29, and very likely if B-lines score ≥ 30. The operator also had to assess the presence or absence of B-profile pattern which is suggestive of CHF according to Lichtenstein criteria [8]. B-profile pattern was defined as such if two or more lung zones per side were positive. A lung zone was positive if three or more B-lines were identified. The final leading diagnosis of dyspnea was assessed by two independent senior EM physicians after reviewing the entire medical record of each patient it was based on: (1) the clinical presentation (severe shortness of breath, worsening dyspnea, orthopnea, paroxysmal nocturnal dyspnea, coughing up or wheezing with white or pink blood-tinged phlegm, foamy mucus), and the physical exam findings (pulmonary congestion and/or peripheral edema, rales, crackles); (2) the diagnostic tests’ results including chest X-ray (pulmonary venous congestion, pleural effusion, interstitial or alveolar edema and cardiomegaly), echocardiography (structural or functional cardiac abnormalities), brain natriuretic peptide (BNP > 300 pg/mL, or NT-proBNP > 1200 pg/mL), the saved images of LUS study, treatment, and outcome [4]. In case of a disagreement, a third senior physician was consulted and adjudicated the case. All senior physicians participating in the study were masked to LUS results. Informed consent was obtained in all the patients before the start of the protocol.
Statistical analysis
Prior to enrollment, a power analysis was performed to determine the sample size needed. Assuming an alpha of 0.05 and a desired precision of 0.07, we calculated a sample size of 502 patients required if we considered that the estimated prevalence of CHF is 25% and the targeted sensitivity and specificity would both be 0.80.
After analysis of normality distribution, variables were expressed by the arithmetic mean and standard deviation (SD) or the median and the 95% confidence interval (or interquartile range). Comparison between patients with CHF (HF group) and those without CHF (non-HF group) was performed by Student’s t-test for continuous variables and Chi-2 test for categorical variables. The difference was considered statistically significant for values of p ≤ 0.05. Discrimination power of the assessed models was studied by the area under the receiver operating characteristic (ROC) curve. An area under curve (AUC) = 1 represents a perfect test; an area of 0.5 represents a worthless test (random prediction), and an area greater than 0.70 means that accuracy of the diagnostic test is at least fair. For the assessment of diagnostic accuracy of B-lines, the scanning order was randomly determined according to an electronic randomization. Agreement between residents’ interpretation was assessed by kappa agreement index for qualitative indices (B-lines score as ordinal scale, and B-profile pattern recorded dichotomously as present or absent). Agreement was considered “low” when kappa value was less 0.40, “fair” from 0.41 to 0.60, “good” from 0.61 to 0.80 and “excellent” from 0.81 to 1. For the B-lines score, the Bland and Altman plot was constructed. A good match was defined when the differences between B-lines score pairs is around the average line and between the lines of − 2 and + 2 SD. The data obtained in this study were collected, recorded and analyzed using SPSS computer software version 18.0 (Chicago, IL).