Podcast 63 set off some expected controversy given my take that plain films are a dead imaging modality for c-spine injuries. I wanted to briefly outline my impression of the existing evidence:
Worst Case Scenario for Sensitivity
Mathen R, Inaba K, et al. (J Trauma 2007;62:1427)
Showed a sensitivity of 45% for plain films.
Prospective study of trauma patients who could not be cleared by NEXUS. Got 3-view plain films and CT. Gold standard was evidence of injury during entire hospitalization.
Post NEXUS Prevalence was ~10%, so probably a mix of moderate and high risk patients.
Best Case Scenario for Sensitivity
Mower WR, Hoffman JR, et al. Use of Plain Radiography to Screen for Cervical Spine Injuries (Ann Emerg Med 2001;38(1):1)
It is a reanalysis of the NEXUS Data (NEJM 2000;343(2):94)
818 Patients with 1496 c-spine injuries
Missed 320 and found 498 of the c-spine injuries in those 818 patients
Of the 320 misses, 237 were deemed inadequate plain films
So 498 out of 581 patients with adequate plain films
So sensitivity of the exam is 85%; We’ll assume a specificity of 100%
If you evaluate the performance by fracture instead of patient, the numbers become worse
I will say in the Mower paper, they tried to exclude SCIWORA patients, but from what I can glean from this paper (J Trauma 2002;53:1-4), these patients had their MRI without CT scans preceding it. CT may have picked up most of these injuries.
Now how can we get away with such a crappy sensitivity
The reason quoted is the NPV is excellent, they state 99.6% NPV. But NPV is a really crappy number, why…
Because as you change the prevalence, the NPV changes.
So now we need to go to a second enormous study…
Let’s look at the Canadian C-Spine Studies (JAMA 2001;286(15):1841 & NEJM 2003;349(26):2510), why? Because their entry criteria are exactly the patients we want to discuss–namely, acute trauma with alert mental status, an injury within the past 48 hours, and in stable condition. The prevalence of c-spine injuries in these patients was ~2% and in the NEXUS trial it was 2.4% So now we have some numbers for a low risk cohort. However, after you get a group of patients who could not be excluded by CCR, the prevalence of the group increases to ~4%. I would argue these patients are now moderate risk. If you pursue plain film strategy in this group, from the best numbers I can gather, you will miss 1 in 100 c-spine injuries and half of these will be clinically significant injuries.
75% of your plain films will be inadequate and require a CT scan
Plain films read as normal but which have loss of lordosis or soft tissue swelling were interpreted as abnormal by NEXUS folks and demand CT scan, this will account for patients going on to CT as well
Finally, patients with persistent midline pain probably deserve a CT prior to d/c in a collar as well
Let’s Put it all Together
The authors of this paper from the journal Medical Physics (Med Physics 2009;36(10):4461) attempted to take all the variables: radiation risk, cancer, missed injuries, etc. and evaluate whether plain films or CT is a better strategy. The results…in all risk levels, CT was the smarter move. This was with factoring in the putative cancer risks.
What about MRI for patients with persistent Midline Tenderness
BF asked about this in the comments
(Ann Emerg Med 2011;58:521)
44% of patients with persistent pain had an MRI abnormality
3/20 patients who were alert, oriented but with persistent neck pain after negative CT had MRI findings
Of relevance to Oli Flower’s comments from the previous podcast, a whopping 24% of clinically unevaluable patients had injuries found on MRI.
For a delightful review of the literature on this question, see Adventure of the Crooked Man by EMNERD.
Now, on to the podcast…