Lung ultrasound allows for earlier diagnosis of bronchiolitis than auscultation: an animal experiment and human case series

Abstract

Purpose

Early diagnosis of bronchiolitis in infants allows for risk stratification for central apnea, and, when available, the timely initiation of antiviral treatment. An animal model could demonstrate if earlier diagnosis is possible with ultrasound than with clinical exam. Even if possible, translating this to pediatrics would require observations from undifferentiated human infants.

Methods

We used serial daily clinical and lung ultrasound exams in a bovine calf model (Bos taurus) of respiratory syncytial virus bronchiolitis. Ultrasound and clinical examiners were blinded to each other’s findings and the treatments used in 24 calves. Time to diagnosis was compared using Kaplan–Meier curves. A case series of human infants with upper respiratory tract infections, without clinical signs of bronchiolitis, and in whom lung ultrasound was performed, was extracted from hospital records.

Results

In the bovine model, lung ultrasound findings emerged earlier and lasted later than auscultatory findings. Relying on auscultation, 5/24 (21%) of animals were diagnosed by post-inoculation day 5 whereas 24/24 (100%) were diagnosed by ultrasound.

We identified seven infants in whom lung ultrasound was used to diagnose bronchiolitis before adventitial lung sounds emerged. Three of these subsequently developed typical clinical findings of bronchiolitis in the hospital. Two had alternative explanations for their abnormal lung ultrasounds (both required surgical intervention). Two were discharged and required no further medical attention.

Conclusion

Lung ultrasound allowed earlier diagnosis of bronchiolitis than clinical exam in the bovine model. In the human case series this was also true, but alternative causes of abnormal ultrasound were frequent.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40477-021-00648-x.

Introduction

Bronchiolitis, an inflammatory lung disease characterized by increasing airway obstruction following a period of prodromal cough and rhinorrhea, is one of the most common causes of hospital admission in infants [1]. Historically, treatment has been supportive and has been limited to ensuring hydration, managing nasal secretions, selective use of bronchodilators for associated bronchospasm, and positive pressure airway support for increased work of breathing. Consequently, there has been little interest in early diagnosis. This will change with development of antiviral drugs for respiratory syncytial virus (RSV), the most common cause of bronchiolitis. Some have already published data in general interest medical journals [2, 3], and many others e.g. including presatovir (GS-5806), ziresovir (AK0529), lumicitabine (ALS-008176), JNJ-53718678, JNJ-64417184, and EDP-938 are in development [4]. A common feature of RSV antivirals will likely be that they need to be given as early as 3–5 days after inoculation be effective [2, 5–7]. Initiating treatment this early, while avoiding unnecessary treatment of infants who without treatment would have only had an upper respiratory tract infection, requires very early diagnosis. This is challenging because in clinical practice the timing of inoculation is uncertain. Some infants may be brought for medical care in that period where the RSV could be demonstrated on nasal swabs but with scant or no lung findings on auscultation and where the future course of the infection (simple upper respiratory tract infection or bronchiolitis) cannot be deduced from the clinical exam.

The second scenario where early diagnosis of bronchiolitis is useful is when the physician is confronted with an infant who is vulnerable to central apnea. Central apnea complicates 1% of bronchiolitis and typically occurs early in the disease, when the characteristic auscultatory findings may be missed or have not manifest [8]. Infants who are at risk of central apnea, typically younger than six weeks of age, premature, or with comorbidities, may be admitted to the hospital for fear that apnea may occur when the parents are sleeping resulting in an unexpected potentially avoidable infant death [9–11].

Point-of-care lung ultrasonography provides a potential solution. Lung ultrasound has been shown to predict more severe lung disease in infants, i.e. prior to severe auscultatory findings developing [12–17]. However, lung ultrasound has not yet been proven to allow earlier detection of bronchiolitis in infants. The exact time of RSV inoculation in a human infant is rarely if ever known. Therefore, a concerning lung ultrasound in the absence physical findings may alert the clinician to the presence of early bronchiolitis and to consider the risk of central apnea when deciding whether to admit or discharge the infant from the emergency department.

We hypothesized that point-of-care lung ultrasound would allow for earlier diagnosis of bronchiolitis than auscultation. Because we cannot deliberately infect infants with RSV we used an experimental approach with a bovine model (which we describe in article 1 in this series [18]) and an observational approach in human infants seeking out cases where our pediatric emergency physicians had attempted to use ultrasound for early diagnosis.

Bovine subjects

Methods

The methods are as described the first article [18] in this series. Briefly, we experimentally infected 24 5-week-old pre-ruminant Holstein bull calves (Bos taurus) with bovine RSV [7]. We performed physical examinations and lung ultrasounds daily until Day 10 when they were euthanized.

Bovine subjects: history and physical examination

Physical examination including lung auscultation was performed daily by a veterinarian and was recorded by an assistant on a different case report form to that used by the ultrasonographers, was transferred to a separate database, and the data not merged until the final analysis. This was done to ensure independence of the auscultation and ultrasound findings. Clinical scoring was performed using a previously described modification of the score described by Collie as per Appendix 1 [19, 20]. 

Bovine subjects: ultrasound examinations

The investigators and directly supervised animal science undergraduate volunteers performed six-window ultrasounds of the calves using either a Sunbright 150 ultrasound (Focus Technology Co., Ltd., Nanjing, Jiangsu, China) or a Clarius first generation hand-held ultrasound machine (Clarius Mobile Health, Burnaby, Canada). The results were read and documented on paper in real time by one or both of the pediatric emergency physicians on the study team.

Bovine subjects: data analysis

The data were collected on paper case report forms and transferred to customized FilemakerPro 12 databases (Filemaker Inc, Santa Clara, CA). Data were exported to Stata 16.1 statistical software (Statacorp LLP, College Station, TX) for data management and analysis. Ultrasound scores were calculated as described in the first article in this series and Table 1 [18].  We plotted the standardized clinical score excluding temperature component, and the standardized respiratory rate against the standardized ultrasound scores for each animal for each day. We plotted Kaplan–Meier curves for time to first abnormal ultrasound, first abnormal auscultatory lung exam, and first time to either being abnormal. We present these results graphically. We include daily viral load on this plot to provide a reference point for the reader.

Table 1.

Description and ordinal scoring of ultrasound findings in order of increasing severity (most severe at the bottom of the table)

Normal	Abnormal	Ordinal score
A lines		0
B lines ≤ 3 per intercostal space		1
	B lines > 3 per intercostal space	2
	Pleural thickening	3
	Air bronchograms	3
	Multiple linear air bronchograms suggesting atelectasis	4
	Concentric air bronchograms or hepatization suggesting consolidation	5
	Abscess	5
	Effusion	5

Where multiple findings were present the most severe was scored

Human subjects

This was a nested retrospective case series of seven infants younger than six months of age who were felt to be at risk of early bronchiolitis and associated central apnea, and who had had a point of care lung ultrasound performed. We searched hospital medical records from RSV season in 2015 and 2019. Unlike the first article in this series [18] the comparative rarity of infants in whom a physician had performed a point-of-care lung ultrasound seeking evidence of bronchiolitis in the absence of auscultatory findings required a more extensive electronically assisted search of our hospital's medical records over a longer period of time.

Identification of human subjects

The SQL code used to identify charts is in Appendix 2. Sentiment analysis and regexm functions were used to identify charts where the treating physician had recorded an interpretation of a point of care lung ultrasound  [21]. Our case series was small because we were attempting only to determine if cases that demonstrated the characteristics being sought exist, not estimate prevalence.

Humans subjects: history and physical examination

Physical examination including lung auscultation was performed by a pediatric or general emergency physician, resident physician, or physician extender. The presence of cough, apnea, parental reported noisy breathing or wheezing, and fever were recorded. Vital signs and pulse oximetry were extracted, lung auscultation was coded as showing wheezing/ronchi/coarse breath sounds or crackles/rales. Data was abstracted from the medical record by an investigator (PW) using an electronic template.

Humans subjects: ultrasound examinations

Ultrasounds were performed by the treating pediatric emergency physician or resident using a Zonare Z One Pro Ultrasound machine using either a 10–5 MHz or a 14.0 MHz probe (Mindray North America Inc, Mahwah, NJ). The physicians performing the point of care lung ultrasound typically imaged the posterior acoustic windows by running the US probe down the patient’s back midway between the scapula and vertebral column. Axillary and anterior windows were typically interrogated with single views of each. Physicians sometimes chose either to not interrogate or to not document their findings for all windows if they had already reached or documented their diagnosis.

Human subjects: data management

Study data were collected and managed using REDCap electronic data capture tools hosted at Sutter Health, Palo Alto, CA. Sentiment analysis was performed using the sentimentr package in R (R Foundation for Statistical Computing, Vienna, Austria). Data were de-identified prior to analysis with Stata 16.1 (Statacorp LLP, College Station, TX) statistical software.

Human subjects: data analysis

We followed the patients’ medical record longitudinally to determine if they subsequently developed adventitial breath sounds. We classified patients as supporting our belief that lung ultrasound can lead to earlier diagnosis if the initial physician exam did not demonstrate bronchiolitis, but the ultrasound did, and the child subsequently developed physical exam findings consistent with bronchiolitis documented by the inpatient pediatricians. We classified patients as not supporting our hypothesis if the initial physician exam did not demonstrate bronchiolitis, but the ultrasound did, and the child subsequently did not develop clinical evidence of bronchiolitis, or had an alternative diagnosis made as an inpatient. We classified patients as neither supporting nor refuting our hypothesis if the initial physician exam did not demonstrate bronchiolitis, but the ultrasound did, and the child subsequently was discharged from the emergency department by the inpatient team thereby precluding subsequent examination. As this was a case series no statistical analysis was performed.

Ethical considerations

Our institutional review board and animal care use committees approved this research.

Results

In the animal model, lower airways involvement was often, but not always detected earlier by ultrasound than auscultation. Lung ultrasound correlated most closely with physical exam around the period of peak clinical illness, but ultrasound findings typically emerged earlier and lasted later than auscultatory findings. The day by day correlation between lung ultrasound and exam scores are shown in Fig. 1. Relying on auscultation alone, 5/24 (21%) of animals would have been diagnosed on or before post inoculation day 5 whereas if relying on ultrasound alone, all the animals 24/24 (100%) would have been diagnosed within that time frame. Diagnosis would have been made by Day 3 in 5/24 (21%) relying on auscultation alone, and 13/24 (54%), relying on ultrasound alone. Using both ultrasound and auscultation had the best results and would have allowed first diagnosis by 15/24 (63%) by day 3. Figure 2 shows the time to first abnormal lung exam by ultrasound, auscultation, or either.

Fig. 1.

Plot showing boxplots of respiratory rate and clinical and ultrasound scores for each day. The diagnosis of RSV infection however was based on the presence of excess B-lines and pleural thickening rather than a specific score. (Scores were not calculated while the calf was being scanned.)

Fig. 2 .

Kaplan Meier curves for time to diagnosis. Time to first abnormal lung exam consistent with RSV infection by ultrasound, auscultation, or either. Log of viral load is superimposed

In the human infant case series, we identified only seven cases where lung ultrasound was used in the context of diagnosing suspected bronchiolitis prior to the development of adventitious lung sounds. Six of these were admitted to the hospital. Three of these six subsequently developed typical clinical findings of bronchiolitis. Two were ultimately found to have alternative explanations for their abnormal lung ultrasounds and two were discharged in less than 12 h of arrival, thereby allowing insufficient time to determine if clinically apparent bronchiolitis would have evolved.

Patient #1 presented at 11 days of age with one day of sneezing, rhinorrhea, but without apnea and with a normal lung auscultation. Point-of-care lung ultrasound demonstrated excessive short and long coalescent B lines and was reported as consistent with bronchiolitis. Longitudinal follow up was informative. She later developed an episode of apnea with hypoxia (Sa02 88%) that responded to stimulation and was started on supplemental oxygen. She subsequently developed wheezing during her inpatient stay. During her hospitalization an echocardiogram was performed. This showed a patent foramen ovale, possible left coronary artery to atrial fistula, normal cardiac function and no suggestion of congestive heart failure. Subsequent echocardiograms were normal and the initial concern for vessel fistula discounted.

Patient #2 presented at 10 days of age with cough and nasal congestion. On exam she had rhinorrhea, but auscultation was normal. Her lung ultrasound was consistent with bronchiolitis. During the first day of her inpatient stay developed abdominal breathing, sneezing, and minor oxygen desaturations. On the second day she developed mildly increased work of breathing but did not require oxygen had no further apneic episodes and was discharged (Figs. 3, 4).

Fig. 3.

Sequential lung ultrasounds from a Holstein calf taken at days 2, 3, 4, 5, 6, and 9 after experimental infection with bovine RSV

Fig. 4.

Lung ultrasounds taken from infants in whom evolving bronchiolitis was suspected but whose lung fields were clear to auscultation

Patient #3 was a 25-day old male who presented with nasal congestion and rhinorrhea. Initial exam was normal apart from rhinorrhea; however, his ultrasound was consistent with bronchiolitis. He was admitted for observation for risk of central apnea. During his stay he developed increasing intercostal and subcostal retractions, and purulent nasal secretions but did not have episodes of apnea or oxygen desaturation.

In two cases alternative diagnoses were made. One had a retro-laryngeal hemangioma that caused mixed obstructive and central apneic episodes. In this case we suspect that the excess coalescent B-lines on lung ultrasound represented negative pressure pulmonary edema rather than bronchiolitis. The third patient was diagnosed with moderate to large atrial and ventricular septal defects and in this case too it seems likely that the B-lines observed on the ultrasound reflected cardiac pathology rather than bronchiolitis.

Two other cases were discharged within 12 h of ED arrival with a diagnosis of evolving bronchiolitis based on lung ultrasound. No follow up was available to confirm or refute the diagnoses of bronchiolitis but neither child either re-presented or died within 72 h of their discharge (Table 2).

Table 2.

Summaries of patients in whom lung ultrasound was performed because of diagnostic uncertainty

Demographics				Vital signs				Auscultation		Work of breathing	Virus	Outcome	Bedside ultrasound report
ID	Age (days)	Gender	Days sick	HR	Temp (°C)	Sa02	RR	Wheeze	Crackles
Allowed early diagnosis
#1	11	Female	1	162	36.4	95	80	Absent	Absent	Mildly increased	RSV	Transient desaturation to 88% while boarding in the ED. Developed mild wheezing during hospital stay. CXR suggestive of pulmonary vascular congestion but echocardiogram was normal for age	14.5 MHz probe. More than 3 long and short B lines per intercostal space. No consolidation/effusion
#2	10	Male	1	149	36.8	98	58	Absent	Absent	Normal	Rhino	Day 1 of admission developed abdominal breathing, lots of sneezing. Oxygren saturations mid to high 90 s. Day 2 of admission Had slight intercostal retractions mild WoB congestion 98–100%	LUS 14.5 MHZ probe > 3 long and > 3 short B lines per ICS on several windows. No PTX. No effusion. No consolidation. Impression: excess fluid c/w bronchiolitis
#3	25	Male	2–3	161	37.1	100	31	Absent	Absent	Normal	RSV	Day 2 of admission developed mild retractions and large amounts thick yellow secretions but no apnea	14.5 MHz Excess short B lines bilaterally and excess long B lines left base and occasional pleural thickening and some linear air bronchograms. Impression: likely bronchiolitis
Alternative diagnosis
#4	10	Male	7	156	37.4	98	44	Absent	Absent	Increased between Apneic spells	NEG	Desaturation to 80% during ED evaluation. Ultimately diagnosed with mixed central and obstructive apnea and retro laryngeal hemangioma requiring surgical intervention	14.5 MHz Excess long B lines right > left but diffusely in posterior axillary windows. Impression: Bronchiolitis v Diffuse pneumonitis
#5	49	Female	1	189	35.8	100	60	Absent	Absent	Normal after apnea resolved	NEG	Admitted. No further events. Mod to large ASD and VSD	10.5 MHz Impression DDx includes Covid, Negative pressure Pulmonary edema, pneumonitis, bronchiolitis
Led to early diagnosis of bronchiolitis that did not subsequently require intervention
#6	26	Male	2–3	177	36.9	98	60	Absent	Absent	Normal	Rhino	Discharge <12 hours from ward	10.5 MHz Probe. lots of excess long B lines bilaterally c/w bronchiolitis
#7	55	Female	1	148	37.2	98	30	Absent	Absent	Normal	Flu A Rhino	Discharge	POC Lung Ultrasound: Mild scattered B lines bilaterally. Normal lung sliding

ASD atrial septal defect, c/w consistent with, DDx differential diagnosis, HR heart rate, Temp measured temperature, RSV respiratory syncytial virus, Rhino rhinovirus Flu influenza, LUS lung ultrasound, MHz megahertz, Mod moderate, NEG no virus detected on polymerase chain reaction testing, POC point of care ultrasound, PTX pneumothorax, v versus, VSD ventricular septal defect

Discussion

We demonstrated that in an experimental model with known timing of RSV infection that lung ultrasound findings generally, but not always, preceded auscultatory findings. Combining both auscultation and lung ultrasound led to the earlier detection of bronchiolitis than either method alone and would lead to diagnosis within 3 days of infection, the optimal timing for antiviral therapy in 63% and within 5 days in 100% of subjects, the longest duration from infection where antiviral therapy may still be effective.

Experimental animal models represent the optimal setting in which to prove ultrasound allows for earlier diagnosis than auscultation. The timing and amount of viral inoculum are carefully planned to ensure the subsequent development of clinically apparent illness. Although animal models allow proof of concept research that would be unethical in humans, animal models alone are insufficient. Our human case series showed that while ultrasound can, as in the animal model, lead to earlier diagnosis of bronchiolitis than auscultation, lung ultrasound can also lead to false positives. In our series, the false positives were admitted to the hospital and had significant pathology requiring surgical intervention.

Our human case series also pointed another potential limitation of lung ultrasound, namely that it may be overly sensitive and enable the diagnosis of bronchiolitis that might not subsequently have become clinically apparent had it never been diagnosed. This may be less of a problem than it appears; currently physicians are accustomed to providing only supportive treatment as needed. However, this calculus may change as anti-RSV drugs come to market in which case overly sensitive diagnosis would have cost implications. For infants at risk of apnea emergency physicians will welcome the additional sensitivity in diagnosis, whereas admitting services may not.

Our work is broadly consistent with previous studies where lung ultrasound better predicted subsequent deterioration and need for respiratory support in infants with bronchiolitis than physical exam or chest X-ray [12, 17, 22]. Additional evidence supporting the validity of lung ultrasound in bronchiolitis comes from work showing improved clinical and ultrasound scores as bronchiolitis resolved [13, 22, 23]. These studies have the advantage of larger numbers and a purely human methodology. These studies also lent themselves to a prospective methodology. This ensured better recording of findings, more consistent imaging practices, and the use of scoring systems.

Our results differ from previous work with respect to false positives. Two of the seven patients thought to have bronchiolitis on point-of-care lung ultrasound in fact had other diagnoses. This differs from the very high diagnostic accuracy (97%) in diagnosing bronchiolitis [15] and even at distinguishing between respiratory etiologies reported elsewhere [14].

However, clinical series that demonstrated very high accuracy included children who had already presented with clinically apparent bronchiolitis, pneumonia, or both. This differs from our design where we were trying to retrospectively find infants whose lungs had been imaged prior to the development of clinically apparent bronchiolitis. Consequently, we were left with only electronic medical record accounts of the human patients, their ultrasound findings, and their subsequent course. Moreover, in both cases where bronchiolitis was incorrectly diagnosed it was clear that the emergency physician was seeking bronchiolitis, a common diagnosis, and in a case of ‘the eye sees what the mind looks for’ did not consider rarer diagnoses, such as cardiac etiology in one case, and a proximal mass causing intermittent negative pressure pulmonary edema in the other. Overall, our work is among other things a lesson on the importance of conditioning the value of a diagnostic test on the population it is going to be used in.

Limitations

We did not have an uninfected control group in the animal arm if the study. Maintaining the blind on such a group would have been impossible as they would have had to be housed separately from the infected animals because RSV is as highly contagious in calves as it is in human infants. Consequently, all the calves could be expected to develop ultrasound and auscultation findings at some point. This was mitigated because the calves were part of a larger study of various combinations of antivirals and antipyretics (to which the investigators were blinded) and their days of inoculation were staggered. We also used only six acoustic windows. This approach was necessary given the time-consuming nature of restraining and imaging each animal while also completing other experiment related tasks.

Our human case series was very small. This reflects the relatively small number of infants who were (1) younger than 6 weeks of age, where (2) the apparent upper respiratory tract symptoms were thought to potentially be the prodrome for bronchiolitis, and (3) who were also thought to be at risk of central apnea, and (4) in whom a positive ultrasound was obtained and acted on, and (5) in the absence of adventitial lung sounds on auscultation. Such children's care reflects clinical management leading published research, a relatively common phenomenon in acute pediatrics where almost all drugs are used off-label, and some types of acute prospective research are often limited or prohibited outright.

Future research

One potential benefit of this increased sensitivity of lung ultrasound in the diagnosis of bronchiolitis may be the ability to defer more invasive testing for a source of infection once lung ultrasound has identified a source. Lung ultrasound may help resolve the diagnostic dilemma that occurs when an infant with a viral URI and cough and likely rhinovirus must still be managed as if the rhinovirus was not present, whereas if bronchiolitis, which is often caused by rhinovirus, is present the risk of concomitant invasive bacterial illness is often sufficiently low to allow deferral of invasive testing [24–26] While such a treatment algorithm would be attractive to clinicians it must be prospectively validated.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Funding for this research was from The United States Department of Agriculture, National Institute for Food Animals—Grant #2016-11003 awarded to LG and PW (Dual purpose, dual benefit), by UL1 TR001860 from the National Institutes of Health’s National Center for Advancing Translational Sciences (HB) and The Pediatric Emergency Medicine Research Foundation.

Declarations

Conflicts of interest/competing interests

None.

Availability of data and material

The data files for the calves are attached. The medical records of the patients involved in the study are not. For auditing purposes patient identifiers are accessible to the authors for the duration of the ongoing studies and the institutional review board indefinitely via a Sutter approved data steward. See Appendices 2 and 3

Code availability (software application or custom code)

Code to replicate the analyses from the cleaned data file is attached.

Ethics approval

Ethics approval was obtained from the institutional review boards, (Sutter Medical Center), institutional animal care and use committee and biological use authorization for live virus use (UC Davis, veterinary school).

Consent to participate

A waiver of consent was obtained for human patients from the institutional review and privacy boards, (Sutter Medical Center).

Consent for publication

All authors agree to the publication of this manuscript.