POCUS in dyspnea, nontraumatic hypotension, and shock; a systematic review of existing evidence

We searched the PubMed and Embase databases for all original studies published between March 2002 and March 2022 on adult patients with dyspnea, nontraumatic hypotension, and shock who were assessed using point-of-care ultrasound. In the PubMed database we used MeSH terms for our search: («Ultrasonography»[Mesh] OR «Ultrasonography»[tiab] OR «Echocardiography»[Mesh] OR «Echocardiography»[tiab] OR «lung ultrasound»[tiab] OR «ultrasound»[tiab] OR «LuCUS»[tiab] OR «Point-of-Care Systems»[Mesh] OR «Point-of-Care Testing»[Mesh] OR «POCUS»[tiab] OR «point-of-care-ultrasound»[tiab]) AND («Dyspnea»[Mesh] OR «Tachypnea»[Mesh] OR «Shock»[Mesh] OR «dyspnea»[tiab] OR «tachypnea»[tiab] OR «shock»[tiab] OR «hypotension»[tiab] OR «respiratory failure»[tiab] OR «pneumonia»[tiab] OR «sepsis»[tiab]) AND («Emergency Service, Hospital»[Mesh] OR «Intensive Care Units»[Mesh] OR «Patients’ Rooms»[Mesh] OR «Medical ward»[tiab] OR «general ward»[tiab] OR «Internal medicine»[tiab] OR «Internist»[tiab] OR «Emergency department»[tiab] OR «ED»[tiab] OR «Intensive Care Unit»[tiab] OR «ICU»[tiab] OR «MET»[tiab] OR «Medical Emergency Team»[tiab])

In Embase we used the terms point of care ultrasound/exp AND dyspnea/di OR point of care ultrasound/exp AND hypotension/di. Thirdly, we reviewed the reference lists of included papers.

We included English language prospective and retrospective clinical trials and observational studies. Diagnostic accuracy was the primary outcome measure. Secondary outcome measures were mortality, admission to ICU, length of stay, duration of treatment, and narrowing the differential diagnosis. Two reviewers (BK, DW) independently screened articles for inclusion, extracted data, and assessed the quality of included studies. Disagreements on study selection and gathering of data were discussed between the reviewers until consensus was reached.

Both reviewers assessed the risk of bias of all selected studies using the Cochrane tool for randomized controlled trials (Risk of Bias 2) and non-randomized studies (ROBINS-I). Risk assessment was discussed between reviewers until consensus was reached. Risk of bias assessment is tabulated in appendix I.

We reported our literature search in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and this is displayed in appendix II. This systematic review was registered.

We will discuss the identified articles based on dyspnea in different settings and nontraumatic hypotension and shock in different settings.

22 studies have investigated POCUS in undifferentiated dyspnea. Diagnostic accuracy was the primary outcome in 18 studies; in six studies secondary outcomes were assessed. The data are summarized in Table 1.

Zanobetti et al. assessed the accuracy of POCUS compared with conventional workup in 2683 patients with acute dyspnea. In this prospective observational cohort study the scanning physicians had expertise in POCUS and were not involved in patient management. They were not blinded to findings in the primary assessment. The reference standard was the final diagnosis based on medical record review including POCUS findings as decided by two independent physicians. The POCUS diagnosis had a good concordance with final diagnosis and was superior in diagnosing acute heart failure. For pneumonia, acute myocardial infarction, and pleural fluid no superiority was found. Patients with asthma and COPD were less often correctly diagnosed by POCUS.

Bekgoz et al. assessed if a modified BLUE protocol was capable of detecting the causes of acute dyspnea in this setting. They conducted a prospective observational cohort study with blinding of the POCUS trained physicians to clinical findings and patient management. The reference standard was the final diagnosis based on medical record review decided by one attending physician. The diagnostic accuracy for pneumonia, heart failure and pneumothorax was good. Correct identification of asthma/COPD was shown to be less successful. Comparable studies have been conducted by six study groups and a seventh in patients aged 75 years and above. All found similar results but had methodological shortcomings in blinding and patient involvement of the scanning physicians. Laursen et al. (two studies) demonstrated that POCUS increased the rate of correct presumptive diagnosis from 63.7 to 88% and could rule in and rule out life threatening conditions which they found in 14% of patients. All scans were performed by the same scanning physician who was blinded to clinical information.

Papanagnou et al.  found no improvement in diagnostic accuracy of POCUS when compared to final diagnosis. However, the scanning physicians in this study were the treating physicians and not blinded to clinical information.

Guttikonda et al. Investigated the use of POCUS in a resource-limited setting. They found a diagnostic accuracy of 88%, a change in diagnosis and therapeutic management in 44.3 and 53.6% respectively, and an increase in diagnostic accuracy when POCUS was added to the conventional workup. In all three studies POCUS experienced physicians conducted or supervised the scanning. Only in the study by House et al. blinding to clinical information was undertaken whilst in the study by Guttikonda et al. the scanning physician was involved in patient treatment.

Buhumaid et al., concluded that POCUS had a comparable accuracy to CXR. Its superior sensitivity in case of a normal lung ultrasound made the additional value of a CXR negligible. Only the research by Lamsam et al. ensured blinding to clinical information.

Buhumaid et al and Stewart et al concluded that POCUS narrowed the differential diagnoses. In the former study scanning physicians were blinded to clinical information whilst in the latter they were not. Zanobetti et al. found an important time benefit of POCUS. Its results were available instantaneously while the interpretation of a chest X-ray by a radiologist lasted 95 min on average. Comparable findings were made by Gaber et al. randomized 211 patients between blinding and unblinding of POCUS findings to the treating physicians and examined the impact of POCUS on concordance with final diagnosis. The primary outcome in both groups was comparable but the patients whose POCUS findings were unblinded to the treatment team received appropriate therapy more often with an absolute increase of 13,5%. Also the proportion of patients discharged within 1 day was larger (0.40 versus 0.24).

21 POCUS studies focused on dyspnea caused by acute heart failure. Diagnostic accuracy was the primary outcome in all. The data are summarized in Table 1.

Pivetta et al.  investigated twice if lung ultrasound could differentiate cardiogenic from other causes of dyspnea. The first study examined a convenience sample of 1005 patients and scanning physicians were trained in POCUS and involved in patient management and thus not blinded to clinical information. The second randomized 518 patients between POCUS versus no POCUS and blinded the scanning physicians to NT-proBNP and CXR. Lung ultrasound alone showed good diagnostic accuracy but the optimal result was obtained when lung ultrasound and conventional workup were combined.

Baker et al. investigated if lung ultrasound practiced by novices contributed to the diagnostic accuracy for diffuse interstitial syndrome. 442 patients were randomized between the control group receiving conventional workup or the intervention group receiving an additional lung ultrasound. All scans were reviewed by a POCUS expert unaware of clinical information. Lung ultrasound had a small effect on diagnostic accuracy. This study demonstrated that lung ultrasound by novices was safe, not time consuming, nor costly.

Anderson et al. conducted a combination of cardiac and lung ultrasound by POCUS proficient physicians who were blinded to clinical information and concluded that the specificity of POCUS for acute heart failure was excellent. Similar conclusions were drawn by Glöckner et al. (two studies), Russell et al., and Sforza et al. whose studies ensured blinding to clinical information. Other studies by Cibinel et al., Dehbozorgi et al., Gargani et al., and Öhman et al. had equal findings but lacked blinding hypothesized that the high rate of cardiac and renal disfunction in this patient category impaired the discriminative role of brain natriuretic peptides whilst not affecting POCUS findings. They were not blinded to clinical information. Liteplo et al. favored the use of POCUS because of its faster results. They together with Wang et al. also stated that congruent ultrasound and NT-proBNP results have the best diagnostic ability. In both studies blinding was ensured by review of all scans by a POCUS expert blinded to scan interpretation and clinical information. Nazerian et al., Prosen et al, and Sartini et al. concluded that a combination of clinical criteria, NT-proBNP, and lung ultrasound yielded the best diagnostic results. The research group of Prosen et al. were involved in patient management. Nakao et al. compared POCUS and CXR and found better accuracies in the former in identifying heart failure. In this study the scans were reviewed by the ultrasound team blinded to clinical information.

Miller et al. investigated if scanning the inferior vena cava could differentiate cardiac dyspnea from other causes. The scanning physicians were not blinded to clinical information and patient with pulmonary hypertension or tricuspid regurgitation were excluded.

Six studies have assessed POCUS for dyspnea caused by pneumonia. All had diagnostic accuracy as their primary outcome. The data are summarized in Table 1.

Bourcier et al.  compared the diagnostic accuracies of CXR and POCUS in 144 patients with a suspected pneumonia. POCUS had a superior sensitivity. In a subgroup of 44 patients with pneumonia who had symptoms for less than 24 h CXR found positive findings in 13 versus 43 using POCUS. In this study the scanning physician was involved in patient management. Cortellaro et al. found comparable results between lung ultrasound and CXR. They concluded that lung ultrasound was feasible in all patients and completed faster. The latter group also found that CXR taken from two orthogonal sites was only feasible in about two thirds of dyspneic patients. In both studies the scanning physicians were blinded to CXR but not to clinical findings.

Javaudin et al. concluded that lung ultrasound improved diagnostic accuracy from 27% to 77%. The scanning physicians had varying POCUS and were blinded to clinical information. Sezgin et al. found similar results.

Nazerian et al. showed that the combination of procalcitonin and lung ultrasound further improved the sensitivity to identify pneumonia when using chest CT as the reference test. The scanning physicians were not involved in patient management and had expertise in lung ultrasound.

Eleven studies investigated POCUS in patients with acute respiratory failure. Diagnostic accuracy was the primary outcome in eight. Other outcomes studied were physician’s diagnostic confidence, change in patient management, time to diagnosis, and cumulative fluid administration. The data are summarized in Table 2.

Lichtenstein et al. introduced the bedside lung ultrasound in emergency (BLUE) protocol to determine the etiology of respiratory failure in 260 ICU patients. The correct diagnosis was made in 90.5% of patients including acute heart failure, pneumonia, pneumothorax, asthma/COPD, and presence of venous thrombo-embolism as a proxy for pulmonary embolism. The scanning physicians were not involved in patient management and the reference standard was the final diagnosis as decided by the treating physicians unaware of the ultrasound results. Dexheimer Neto et al. who classified themselves as non-experts in POCUS, repeated the BLUE protocol in a final study population of only 37 patients with acute respiratory failure. The scanning physicians were blinded to clinical information and obtained a high diagnostic accuracy of 84% with POCUS, which was better than CXR, but not significantly better than clinical gestalt alone which was 65%.

Other researchers have modified the BLUE protocol by adding a focused cardiac ultrasound or leaving out VTE assessment. Barman et al. showed that POCUS changed the initial diagnosis, added a diagnosis, and had higher concordance with final diagnosis. Bataille et al. compared lung ultrasound and the combination of cardiac and lung ultrasound and concluded that latter resulted in higher diagnostic accuracies for heart failure and pneumonia. Although they were not involved in patient management they were not blinded to obvious clinical clues. Sekiguchi et al. applied a likewise approach and demonstrated this was able to better differentiate between etiologies of acute respiratory failure. However, the POCUS physicians were involved in patient management. Silva et al. conducted a multiorgan scanning protocol and was able to demonstrate comparable diagnostic accuracies to Lichtenstein et al. In this study no strict blinding to obvious clinical clues was applied but as a reference standard final diagnosis was based on medical record review including chest CT findings in 75% and echocardiography in 68% of cases. Smit et al. compared lung ultrasound with chest CT in 79 patients. The scanning physicians had a heterogenous level of POCUS experience and were involved in patient management. They found 147 respiratory conditions (most commonly consolidation, interstitial syndrome, and pleural fluid) and emphasized that multiple respiratory conditions can be present in one patient.

Tierney et al. assessed the ability of lung ultrasound by experienced physicians and CXR to identify lung pathology and the affected lobe. Agreement of POCUS and chest CT was 87% versus 62% of CXR. Scanning physicians were not involved in patient management but not blinded to obvious clinical signs. In this study it is concluded that POCUS in the ICU is a better alternative than CXR.

The studies by Wallbridge et al., Wang et al, and Xirouchaki et al. focused on other outcomes of POCUS. Wallbridge et al.  found an increase in the physician’s diagnostic confidence by 44%. In this study no strict blinding to obvious clinical clues was applied. Wang et al. showed that acute pulmonary edema was diagnosed faster and patient were given less intravenous fluids. They were not involved in patient management but not blinded to obvious clinical signs. In a study by Xirouchaki et al. on mechanically ventilated patients lung ultrasound resulted in a direct change in management half of the times (number of lung ultrasounds was 253). 53 lung ultrasounds revealed findings not suspected by the primary physician such as pneumothorax, atelectasis, significant pleural effusion, consolidation, and diffuse interstitial syndrome. One scanning physician who was not blinded to obvious clinical signs conducted all the scans.

Seven studies investigated POCUS in ward patients with undifferentiated dyspnea. The data are summarized in Table 3. Diagnostic accuracy was the primary outcome in six, change in primary diagnosis and time to initiation of appropriate therapy in one. The largest study was conducted in a short stay unit by Volpicelli et al. who assessed the potential of lung ultrasound in diagnosing alveolar interstitial syndrome in 295 patients. Scanning physicians were trained in POCUS and blinded to clinical findings. Interstitial syndrome was detected with a sensitivity of over 85% and specificity of 97%. Others concluded that lung ultrasound identified patients with heart failure better than chest X-ray and was a more useful instrument to monitor treatment effect in this patient category than repeat measurements of NT-proBNP levels.

Sen et al. examined the BLUE protocol in patients with respiratory failure and showed a good overall diagnostic accuracy of 84%, applied by experienced POCUS users, which was higher than for clinical diagnosis.

Mearelli et al. showed in a study of 315 patients with a suspected community acquired pneumonia that application of lung ultrasound by experts resulted in an excellent accuracy. The scanning physicians were blinded to clinical information and demonstrated to outperform the diagnostic accuracy of chest X-ray assessed by a radiologist.

Ben-Baruch Golan et al. found that the main effects of application of a cardiac and lung ultrasound in patients admitted for dyspnea, chest pain, or worsening peripheral edema were twofold. The first was a change in diagnosis in 28%. The second a faster initiation of appropriate therapy which was five hours with POCUS versus 24 h without. The patient population was randomized between an intervention and control group and the scanning physicians blinded to clinical findings.

Sixteen studies to investigate the role of POCUS in patients with nontraumatic hypotension and shock have been conducted, including five studies on POCUS and sepsis, and two studies assessing both circulatory and respiratory complaints. The data are summarized in Table 4. Diagnostic accuracy was the primary outcome in eleven, whilst the other studies focused on clinical decision making, disposition time, and 30-day mortality.

Javali et al. showed that POCUS increased the percentage of correct presumptive diagnosis from 45% to 89%. The single scanning physician was blinded to clinical information. Comparable results were found by Shokoohi et al. but only the first study group ensured blinding of the scanning physicians. Sasmaz et al. also found similar results but the scanning physicians were involved in patient care.

In addition to an increase in diagnostic accuracy Ahn et al. found that POCUS was an effective tool to help physicians narrow the differential diagnoses and to increase their confidence on diagnosis and treatment strategy. Moore et al. concluded that they found a good correlation between physicians trained in POCUS and cardiologists in estimation of cardiac ejection fraction in patients with symptomatic hypotension. In both studies no blinding of the scanning physicians occurred but they were not involved in patient care.

A retrospective analysis by Jones et al. found that a hyperdynamic left ventricular function on POCUS was highly specific for sepsis as the etiology of shock. An approximate 40 percent of the patient population was excluded due to insufficient image interpretability.

Becker et al. concluded that the diagnostic accuracy of a POCUS incorporated clinical approach to identify the cause of shock or respiratory distress was superior to conventional workup in a resource-limited setting. The scanning physicians were involved in patient management.

Cortellaro et al. studied the diagnostic accuracy of POCUS in identifying the focus of infection in patients with sepsis. POCUS was conducted by the treating physician within ten minutes after the primary assessment and improved the diagnostic accuracy from 52.5% to 75%. Secondly, it was able to immediately identify the source of infection in 73% whilst standard workup took 6 h to do so.

Hall et al. found a beneficial effect of POCUS on disposition time of almost an hour. Importantly, the scanning physicians were involved in patient care.

Coen et al. found that POCUS to guide fluid administration reduced the necessity of invasive monitoring (i.e. central venous catheter) whilst Haydar et al. concluded that POCUS resulted in a change in the treatment plan of intravenous fluid administration in more than 40% of the patients. In the former study the scanning physicians had involvement in patient care whilst in both studies the investigators were not blinded to clinical signs.

Atkinson et al. assessed the effect of POCUS on survival until discharge or 30 days. A prospective convenience sample of 270 adult patients with undifferentiated nontraumatic hypotension was randomized between assessment with or without using POCUS by physicians proficient in its use. They were involved in patient management. They found no difference in survival nor in the secondary outcomes fluid administration, inotropes, rate of admission to ICU, and length of stay. Only patients with truly undifferentiated shock were included, patients with a high suspicion of acute abdominal aneurysm and a clear mechanism of shock were excluded. Musikatavorn et al. found no positive effect on 30-day mortality of patients with sepsis who received an inferior vena cava assessment to guide intravenous fluid management. They were involved in patient management.

Five studies to assess the role of POCUS in patients with shock have been conducted. Diagnostic accuracy was the primary outcome in two, 28-day mortality in two, and clinical outcome after 48 h in one. The data are summarized in Table 5.

Vaidya et al. studied the ability of POCUS to classify the cause of shock and found excellent agreement for obstructive shock but less so for other causes due to overlap in ultrasound findings. POCUS expertise, involvement in patient management and blinding are not specified.

Sekiguchi et al. conducted a pilot study to examine the consequence of POCUS on diagnosis, treatment plan, and physician’s confidence in patients with severe sepsis or septic shock. The result was modest with no change on primary diagnosis, but a change in treatment in approximately 25% and an increase in confidence for a given treatment. Scanning physicians were not involved in patient management and all images were reviewed by a cardiologist blinded to clinical information.

The largest study to assess the role of POCUS on 28-day survival in patients with undifferentiated shock was conducted by Kanji et al. They assessed a prospective cohort of 220 consecutive patients who were assessed with or without POCUS to aid in recommendations on intravenous fluid and inotropes. They found a positive change in 28-day survival of ten percent, less stage three kidney injury and less need of renal replacement therapy. POCUS resulted in less fluid prescription and more inotropic agents. Scanning physicians were POCUS experts, not involved in patient management but not blinded to obvious clinical signs. Li et al. conducted an outcome study in 94 patients of whom 49 received POCUS and found no difference in 28-day mortality. The clinical effect might have been subtle because of a small number of patients with abnormal ultrasound findings. Investigators were involved in patient care.

Lanspa et al. compared POCUS and early goal directed therapy in a small pilot study and found no benefit. An explanation for this could have been that resuscitation with substantial amounts of intravenous fluids was started before randomization thereby reducing the size of any possible effect. The investigators were involved in patient care.

Two studies to assess the role of POCUS in patients with shock have been conducted which both focus on diagnostic accuracy as the primary outcome. The data are summarized in Table 6.

Blans et al. demonstrated an increase in diagnostic accuracy by 27% and an increase in physician’s certainty. In the study by Zieleskiewicz an increase in diagnostic accuracy of 14% and earlier start of first treatment were found. In both studies the scanning physicians were involved in patient care.

Randomized controlled trials included in this systematic review were all evaluated with some concern or high risk of bias. The non-randomized trials were assessed as having a moderate to serious risk of bias. The reasons to assess the risk of bias as substantial are lack of blinding of the scanning physician to the intervention, selection of the scanning physicians, selection of study participants, and absence of an objective gold standard as the reference test in diagnostic accuracy studies. In addition, very few negative results have been published and therefore publication bias must be considered as well.

This systematic review provides a large body of indirect evidence and outweighs the lack of direct evidence. As has been reported in the literature, executing well designed controlled trials on clinical outcome remains difficult if not unfeasible at all. Therefore we doubt whether high quality studies will be added to the evidence base.

As stated in the risk of bias assessment, all studies were prone to bias, often in the study design, the execution of the study, and generalizability of the results. In a minority of papers we considered the risk of bias as high making their findings less robust to be translated into clinical practice.