Saliva RNA biomarkers predict concussion duration and detect symptom recovery: a comparison with balance and cognitive testing

Gregory Fedorchak,Rebekah Mannix,Andrea C Loeffert,Jayson Loeffert,Samantha Devita,Mohammad N Haider,Thomas Uhlig,John Leddy,Cayce Onks,Frank Middleton,Timothy Lee,Scott L Zuckerman,Robert P Olympia,Aakanksha Rangnekar,Keith Owen Yeates,Steven Hicks,Jessica Rieger,Aaron Roberts,Chuck Monteith

doi:10.1007/s00415-021-10566-x

Abstract

ObjectiveThe goals of this study were to assess the ability of salivary non-coding RNA (ncRNA) levels to predict post-concussion symptoms lasting ≥ 21 days, and to examine the ability of ncRNAs to identify recovery compared to cognition and balance.MethodsRNA sequencing was performed on 505 saliva samples obtained longitudinally from 112 individuals (8–24-years-old) with mild traumatic brain injury (mTBI). Initial samples were obtained ≤ 14 days post-injury, and follow-up samples were obtained ≥ 21 days post-injury. Computerized balance and cognitive test performance were assessed at initial and follow-up time-points. Machine learning was used to define: (1) a model employing initial ncRNA levels to predict persistent post-concussion symptoms (PPCS) ≥ 21 days post-injury; and (2) a model employing follow-up ncRNA levels to identify symptom recovery. Performance of the models was compared against a validated clinical prediction rule, and balance/cognitive test performance, respectively.ResultsAn algorithm using age and 16 ncRNAs predicted PPCS with greater accuracy than the validated clinical tool and demonstrated additive combined utility (area under the curve (AUC) 0.86; 95% CI 0.84–0.88). Initial balance and cognitive test performance did not differ between PPCS and non-PPCS groups (p > 0.05). Follow-up balance and cognitive test performance identified symptom recovery with similar accuracy to a model using 11 ncRNAs and age. A combined model (ncRNAs, balance, cognition) most accurately identified recovery (AUC 0.86; 95% CI 0.83–0.89).ConclusionsncRNA biomarkers show promise for tracking recovery from mTBI, and for predicting who will have prolonged symptoms. They could provide accurate expectations for recovery, stratify need for intervention, and guide safe return-to-activities.

Highlights

Guidelines from the Centers for Disease Control and Prevention (CDC) Pediatric Mild Traumatic Brain Injury Workgroup recommend that clinicians inform patients and families that some factors predict risk for persistent post-concussion symptoms (PPCS), but individual recovery from mTBI is unique [1]
We investigated longitudinal ncRNA levels and medical/demographic factors among 112 individuals with mTBI at a minimum of two time-points following injury, including an initial saliva sample collected no later than 14 days post-injury and a follow-up sample used to define presence/absence of PPCS, beginning ≥ 21 days post-injury
The heat map of gradient-boosted machine (GBM)-ranked “recovery” features in Fig. 4a demonstrates that individuals with PPCS have poorer balance and cognitive scores at ≥ 21 days post-injury compared to non-PPCS counterparts

Summary

Introduction

Guidelines from the Centers for Disease Control and Prevention (CDC) Pediatric Mild Traumatic Brain Injury (mTBI) Workgroup recommend that clinicians inform patients and families that some factors predict risk for persistent post-concussion symptoms (PPCS), but individual recovery from mTBI is unique [1]. In ambulatory clinics, where patient visits may last only 15 min, rapid, objective measures that do not require specialist interpretation are urgently needed Such tools would improve care for patients with PPCS in two ways: (1) individuals who receive education about prognosis have improved outcomes [4, 5]; and (2) identifying those at risk for PPCS provides an opportunity for early intervention prior to the development of prolonged and debilitating symptoms [1]. We investigated longitudinal ncRNA levels and medical/demographic factors among 112 individuals with mTBI at a minimum of two time-points following injury, including an initial saliva sample collected no later than 14 days post-injury and a follow-up sample used to define presence/absence of PPCS, beginning ≥ 21 days post-injury. Refinement and validation of the algorithm could promote objective anticipatory guidance, facilitate safer return-to-play decisions, and foster effective therapeutics based on individual biologic responses to mTBI

Methods

Participants

Participants characteristics and symptoms

Discussion

Strengths and limitations

Conclusions