Article Text

Download PDFPDF

2296 Predicting recovery in patients with mild traumatic brain injury and a normal CT using diffusion tensor imaging
  1. Sophie Richter1,
  2. Stefan Winzeck1,
  3. Endre Czeiter2,
  4. Evgenios Kornaropoulos3,
  5. Dan Whitehouse1,
  6. Kevin Wang4,
  7. Andras Buki5,
  8. Andrew Maas6,
  9. Marta Correia1,
  10. David Menon1,
  11. Virginia Newcombe1
  1. 1University of Cambridge
  2. 2University of Pecz
  3. 3Lund University
  4. 4Morehouse School of Medicine
  5. 5Orebro University
  6. 6Antwerp University


Aims and Objectives UK emergency departments conduct 200,000 CTs for mild traumatic brain injury (TBI) annually.

Whilst >90% of CTs are normal, 30-50% of patients report long-term functional deficits. This study aims to assess if advanced MRI (diffusion tensor imaging, DTI) improves existing models for the prediction of 3-month functional recovery in patients with mild TBI and a normal CT.

Method and Design This formal prognostic study followed the TRIPOD statement and used data from the prospective multi-center CENTER-TBI study. Patients aged ≥16 years were included if they had a Glasgow Coma Scale score >12, a normal CT, and DTI within 31 days of injury. Complete recovery at 3 months was defined as an extended Glasgow Outcome Scale score of 8.

DTI data was harmonized and age-corrected using matched healthy controls (n = 157). The current best prognostic models for an incomplete recovery (UPFRONT-PLUS, UPFRONT-ED and HeadSMART) were fitted with and without DTI information and internally validated using bootstrapping.

Where available (n = 107), we assessed if serum neurofilament light (NFL) can select patients for DTI.

Results and Conclusion The study included 153 patients, aged 20-70 (mean 44) years, 108 (71%) male, 70 (46%) with an incomplete recovery. The best model without DTI (UPFRONT-PLUS) explained 10% (-6-26) of the variation in outcome (ViO) with an area under the curve (AUC) of 0.57 (0.47-0.68). Adding DTI raised the ViO to 74% (66-82) and the AUC to 0.79 (0.77-0.81), p <0.001.

NFL could have avoided 36% of DTI with a sensitivity of 0.81 (0.67-0.89) if sampled at initial presentation, or 24% with a sensitivity of 0.95 (0.81-0.98) if sampled at the time of DTI.

This suggests that NFL and DTI could help select mild TBI patients at risk of incomplete recovery for interventional trials and clinical follow up, pending external validation and a health economics assessment.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.