**Sincere apologies for the delay in this post (due 16th Nov, delayed by technical problems)**
Where can I find this paper?
What is this paper about (what is the research question)?
How good is ultrasound at diagnosing elbow fractures in children?
Summary of the Paper
Design: prospective, observational, multi-centre diagnostic study
Objective: to determine the performance characteristics of ED ultrasound by PEM physicians for diagnosis of elbow fractures in children
Test under evaluation: ED (point-of-care) ultrasound of the elbow (scanning protocol pre-determined) to determine presence of lipohaemarthrosis OR elevated posterior fat pad
Reference standard: plain radiographs reported by radiologists blinded to ultrasound findings. Clinical follow-up with notes review or structured telephone follow-up; radiologist report used if follow-up imaging performed.
Primary outcome: determination of test characteristics
Population: convenience sample of patients aged up to 21yrs presenting to either of two urban EDs when a trained study physician was available, between Sept 2010 and Dec 2011.
- Inclusion: possible elbow fracture requiring radiographic assessment
- Exclusion: elbow radiographs already obtained, previous confirmed elbow fracture, open wound at the elbow, unstable vital signs, severe life threatening injuries requiring resuscitation.
Results: 130 patients enrolled, fracture present on initial ED radiography in 38/130. 5 additional fractures diagnosed on follow-up radiography (23 patients).
Elevated posterior fat pad was seen on initial XR in 57/130 patients (44%): 36 with fracture, 21 without (of whom, 4 patients had fracture identified on follow-up).
Elevated posterior fat pad OR lipohaemarthrosis was seen on ED US in 68/130 patients.
Elevated posterior fat pad OR lipohaemarthrosis gave the following test characteristics: sensitivity 98% (95% CI 88-100%), specificity 70% (95% CI 60-79%), PPV 0.62 (95% CI 0.50-0.72), NPV 0.98 (95% CI 0.91-1.0), LR+ 3.3 (95% CI 2.4-4.5), LR- 0.03 (95% CI 0.01-0.23)
With focused musculoskeletal ultrasonographic training, novice PEM sonologists were able to attain the skills necessary to perform point-of-care elbow ultrasonography to evaluate for fracture by assessing the posterior fat pad for elevation and lipohaemarthrosis.
On the study design
Despite including “children” up to the age of 21 (!), there are many positive points about the methodology in this paper. The authors involved two centres over 16 months and although the sample was obtained as a convenience sample, the wide inclusion and few exclusions mean that there’s a good chance that the population is representative of patients seen in my ED.
There are reasonable attempts to minimise bias:
- Standardised training
- Blinding of radiologists to US results/clinical findings
- Standardised data collection sheet for clinical/examination findings
- Composite reference standard (telephone follow-up)
In addition, it seems pragmatic – with only 1hr training plus 5 practice scans PEM clinicians could achieve an agreement of 0.94 with an experienced PEM sonologist! What concerns me slightly is how unclear it is who these PEM clinicians were. We can all think of the colleagues who would jump at the chance to take part in a study like this; the fact that only one sonologist had prior experience of elbow ultrasound does NOT mean that the others were inexperienced with US in general. Would these results be reproducible in the hands of a not-quite-sure-which-end-of-the-probe-goes-where EM clinician – even with an hour of training? I’m not so sure…
However, the study is nicely designed and would be relatively easy to replicate in your own department.
What were the results and what does this mean?
For diagnostic studies, it can be helpful to draw a 2×2 table. This allows the calculation of test characteristics. Below is a 2×2 table for the presence of posterior fat pad OR lipohaemarthrosis on elbow US.
Sensitivity of 98% is not terrible, although the 95% confidence interval (88-100%) is pretty wide. This means the “true” sensitivity of elbow US in ED could be as low as 88%. Remember, high sensitivity means that when the test is negative we can effectively rule out the disease (fracture). Would we be happy to rule out at 88%? I don’t think so.
What’s most interesting is that the LR+ for elevated posterior fat pad AND lipohaemarthrosis is 5.8; that is, the proportion of patients who have a fracture and these findings is 5.8 times the proportion of patients with these findings who do not have a fracture. We can surmise that finding both elevated PFP and LH would be highly suggestive of an underlying fracture; but would this prevent an x-ray? Unlikely (pesky orthopods)! We don’t know quite how this correlates with the degree of required intervention (were these the fractures which needed to go to theatre; that were horrible looking supracondylar fractures from the outset?). And the LR- is not particularly brilliant when we look at the confidence intervals.
Not all patients who were followed-up had repeat XRs, but this reflects clinical practice and I think it is pretty reasonable not to XR a child who is pain-free with normal range of movement at their follow-up appointment. In fact, the radiation would be hard to justify. It is worth noting that four patients could not be followed-up; all had negative US and negative XR at their initial visit and were included in the “no fracture” group. The authors made reasonable attempts to follow-up these patients; how would the results be altered if we assumed the worst for these cases? And does it matter?
What can we take from this paper into clinical practice?
With appropriate training and practice, EM clinicians could use ultrasound to reliably exclude elbow fractures in children. However, a larger study is needed to tighten those confidence intervals; further training might also have this effect.
If both lipohaemarthrosis and elevated posterior fat pad are identified on US it is very likely that there is an underlying fracture, but would this change clinical practice? Probably not – yet
More questions to ask
- Would more training/greater patient numbers narrow the confidence intervals (so we can effectively rule-out elbow fracture with ultrasound)?
- Would the results be reproducible with a joe-bloggs ED clinician who’s a bit wobbly with an US probe (like me)?
- Could we ever persuade orthopods to manage patients on the basis of US-diagnosed fractures without corresponding radiographs?
Follow us on twitter: @PEMLit