The most recent episode of the SGEM hot of the press series came out today. The article we reviewed was looking at the utility of point of care ultrasound in pediatric forearm fractures:
Poonai N, Myslik F, Joubert G et al. Point-of-care Ultrasound for Nonangulated Distal Forearm Fractures in Children: Test Performance Characteristics and Patient-centered Outcomes. Acad Emerg Med. 2017.
The best way to hear about the details of this study is to head over to the SGEM podcast. As a quick summary, these authors compared point of care ultrasound to x-ray for children 4-17 years of age presenting to the emergency department with a suspected non-angulated forearm fracture. (The physicians performing the ultrasounds were all certified expert scanners.) The primary outcome was sensitivity of ultrasound, but secondary outcomes included pain, caregiver satisfaction, and procedure duration.
Ultrasound was good, but not great, at identifying fractures in this study. The key test characteristics were:
- Sensitivity 94.7% (95% CI; 89.7% to 99.8%)
- Specificity 93.5% (95% CI; 88.6 to 98.5%)
- PPV 92.3 % (95% CI; 86.4% to 98.2%)
- NPV 95.6% (95% CI; 91.4 to 99.8%)
- +LR 14.6
- -LR 0.6
Pain scores were statistically, but probably not clinically significantly lower with ultrasound: median 1 versus 2 (median difference: 0.5; 95% CI: 0.5, 1; p<0.001).
A full critical appraisal, including responses from the lead author, is available on the SGEM podcast. I just wanted to raise 2 points about this study.
1. X-ray as a gold standard
I think we all know that x-ray, despite being our clinical test of choice, has some issues as a gold standard. It will miss many injuries that might show up if a different modality, such as CT or MRI, had been used. There were 6 patients in this study who had a fracture identified by ultrasound that wasn’t seen on x-ray. These were interpreted as false positives, resulting in the specificity of 94%. However, in essentially every ultrasound modality, we know that ultrasound identifies injuries that aren’t seen on x-ray. Without a true gold standard, there is no way to know, but I would guess that some of these might have been real fractures that were simply not seen on x-ray.
2. Clinically important injuries
There were 4 injuries missed in this study: 1 buckle fracture and 3 non displaced ulnar styloid fractures. The question is whether the x-ray truly identified anything clinically important here? We already know that buckle fractures don’t require plaster casts, but instead can be safely managed with tensor bandages or removable velcro splints. In fact, more than 80% of all fractures identified in this study were buckle fractures, which would lead me to believe that there were very few patients in this cohort who actually required intervention (which would be my reason for ordering an x ray). The ulnar styloid is more tricky. Radiographically, these often seem like minor injuries, but I am not aware of any long term outcome studies that would tell us how these injuries would progress without a cast. (The fact that 75% of the misses were in one location does indicate that future ultrasound protocols should probably emphasize adequate views of the ulnar styloid.)
Combined, these two points illustrate my most consistent concern with diagnostic studies: our over-reliance on sensitivity and specificity as end points. This is not a specific criticism of this paper. We are early in our understanding of the use of ultrasound to diagnose fractures in the emergency department, and comparing its performance to current practice (using x-ray) is important. However, clinical outcomes are even more important. The value of these tests can only be determined by measuring patient important outcomes. We need to follow patients clinically. What if the “false positive” ultrasounds in this study were having ongoing pain and limited function at 6 weeks? How many of the missed fractures really required an intervention? Might they have been OK (or perhaps better off) in a removable velcro splint?
In order to truly understand the value of any test, we still require randomized control trials that measure patient important outcomes. For now, I think ultrasound is promising, and that the sensitivity and specificity can probably be better than we saw here, but our x-ray techs should be comfortable in their short term job security.