Effect of Transcranial Low-Level Light Therapy vs Sham Therapy Among Patients With Moderate Traumatic Brain Injury

Key Points Question Is near-infrared low-level light therapy (LLLT) feasible and safe after moderate traumatic brain injury, and does LLLT affect the brain and exhibit neuroreactivity? Findings In this randomized clinical trial including 68 patients with moderate traumatic brain injury who were randomized to receive LLLT or sham therapy, 28 patients completed at least 1 LLLT session without any reported adverse events. In the late subacute stage, there were statistically significant differences in the magnetic resonance imaging–derived diffusion parameters of the white matter tracts between the sham- and light-treated groups, demonstrating neuroreactivity of LLLT. Meaning The results of this clinical trial show that transcranial LLLT is feasible, safe, and affects the brain in a measurable manner.

This supplementary material has been provided by the authors to give readers additional information about their work. amount of light provided by the helmet was based on our prior animal experiments and is below the American National Standards Institute (ANSI) limit for maximum skin exposure of 0.1W/cm2 for 800nm NIR light. 1,2 Even though the wavelength composition of sun light is quite different, for comparison, the approximate solar irradiance at the Earth's surface is 0.1373 W/cm 2 . 3 A removable and disposable molded plastic piece isolated the patient form direct LED contact. Fans located on the exterior of the helmet dissipated any heat generated by the LEDs and the control electronics. The helmet was recalibrated after use on 10 subjects in order to ensure that the light intensity did not exceed the prespecified threshold of 36mW/cm 2 +/-20%. The calibration was performed using a custom-built light monitoring device.
The LLLT helmet was used by a trained coordinator who also monitored and recorded the core temperature of each subject before and after each light therapy session. A control unit connected to the helmet by a 5foot long cable was used to control the helmet state: On/Off and Active/Control. The control unit was kept hidden from the subject during the light therapy session.

Structural Data
We used the Freesurfer developmental version longitudinal stream to perform automated segmentation and cortical parcellation of the T1-weighted volumetric images. 4 The longitudinal stream generates an unbiased, within-subject template for all three time-points of each subject using a robust, inverse consistent registration. 5 Several processing steps, such as skull stripping, Talairach transforms, atlas registration as well as spherical surface maps and parcellations were then  initialized with common information from the within-subject template, significantly increasing reliability and statistical power. All volumes were inspected for accuracy and minor manual edits were performed where needed by a trained operator. These were usually restricted to removal of non-brain tissue included within the cortical boundary. 4 The longitudinal stream from the software TRActs Constrained by UnderLying Anatomy (TRACULA) 6 was used to automatically delineate following 18 major white matter tracts in each subject: The anatomical landmarks are segmented by the Freesurfer longitudinal stream.

Complete Study Protocol
The complete study protocol, as disclosed to the ClinicalTrials.gov, is provided in Appendix A.

Revised Power Calculation
Before the start of this clinical trial, the sample size was calculated using the data from a cohort of 12 moderate and severe TBI patients scanned using quantitative MRI (from Singh et al., 2010, Novel diffusion tensor imaging methodology to detect and quantify injured regions and affected brain pathways in traumatic brain injury. MRI 28 22-40). For this cohort, four DTI parameters ---namely, the fractional anisotropy (FA), the mean diffusivity (MD), the axial diffusivity L1 and the radial diffusivity (Lt) ---were measured in 16 regions of the brain. For this cohort of patients, the most variable DTI parameter was the FA: the coefficient of variation (CV), defined as the ratio of mean to the standard deviation, was 53.45%.  Our current assessment of CV using the blinded data from the cohort we have enrolled so far reveals that the variability in our own cohort is in fact less than 40% that was used for estimating the cohort size at the time of grant submission. For the cerebral blood flow (CBF), measured using ASL, this value is about 25%. For the DTI parameters, the variability is even lower. Reduction in the parameter variability is most likely a result of using the same scanner by the same operator for all subjects. We also exercise extreme caution in care in positioning all patients in the same manner and making sure that all the parameters of the scan are tightly controlled.
If we substitute 0.25 for CV, instead of the original 0.40 obtained from other scans, the cohort size n reduces to 32. In order to be safe, and to account for data that may not be usable for various reasons (e.g., movement artifacts), we targeted a cohort of 40 patients before breaking the blind.

Effect of LLLT on Individual White Matter Tracts
In addition to the linear mixed effect model that tries to discern of effect of global variables (time, treatment, time x treatment), we also performed tract-by-tract analysis for each tract. The results are summarized in Figure S2.  In individuals in the light-treatment group, axial diffusivity (AD) was lower in 16/18 tracts at 2 weeks, and higher in 14/18 tracts at 3 months, compared to control group (binomial test, respectively, p<0.01 and p = 0.03). Similarly, at 2-weeks and 3-months, respectively, radial diffusivity (RD) was lower in 13/18 tracts and greater in 13/18 (p=0.01 for both), mean diffusivity (MD) was lower in 14/18 tracts and greater in 16/18 (p=0.03 and p<0.01), and functional anisotropy (FA) was lower in 14/18 tracts, and remained lower in 13/18 (p=0.03 and p=0.10) At the level of individual tracts, light had a small impact on diffusion parameters. The light induced effect did not reach statistical significance in most individual tracts even though the direction change was consistent across all tracts ( Figure S2).

Random Effects Associated with the LME Model
The eTable 3 shows that variance components of the linear mixed effect model: population covariation (σ 2 ) between random factors and the dependent variable, and the variation (τ) and intraclass correlation coefficient (ICC) within random factors. N denotes the number of individual tracts and patients (ID