The primary outcomes are as follows: 1) safety: biomarkers of systemic inflammation (C-reactive protein); lipids (total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides); sodium; liver enzymes (alanine aminotransferase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), albumin, bilirubin, total protein, nutritional ketosis (beta-hydroxybutyrate) and glycemic status (fasting blood glucose and insulin); 2) neurological recovery (AIS) motor and sensory scores and electrical perceptual thresholds); 3) functional independence (SCIM); 4) microbiome composition; and 5) serum biomarkers (targeted proteomics).
In compliance with goals of the National Institute of Neurological Disorders and Stroke (NINDS) Common Data Elements (CDE) initiative, we are utilizing portions of the SCI CDE that are expected to increase the efficiency and effectiveness of our clinical research study and diet intervention, increase data quality, and facilitate data sharing and comparisons across studies. For data analyses and results reporting, we adhere to the guidelines for reporting results using the International Spinal Cord Core Data Set (28).
Minimally acceptable intervention adherence rates
Adherence to diet will be closely monitored by the study team by measuring serum beta-hydroxybutyrate levels once a week. Blood concentration of 0.5–3 mM (nutritional ketosis) will be considered safe and acceptable. Adherence will also be assessed using food records. The food record will be a simple check list to indicate that supplied foods were consumed, with space to add the quantity of consumption and any deviations from the protocol. In addition, the study dietician will communicate directly with each participant at least weekly to improve adherence.
Clinical procedures to measure safety
Fasting blood samples are collected to determine the levels of biomarkers of systemic inflammation, lipids, sodium, liver enzymes, and glycemic status. In addition, blood samples are collected for measurement of blood ketone (beta-hydroxybutyrate) levels three times a week, 2 h after the first meal of the day, to confirm nutritionally induced ketosis (blood concentration, 0.5–3.0 mM) in the KD group. All serum samples will be obtained in the UAB Hospital Acute Care Unit and then stored in freezers within the PI’s (Yarar-Fisher) research laboratory at -80 °C until analyzed by core personnel. Levels of sodium, beta-hydroxybutyrate, lipids, C-reactive protein, glucose, and insulin are assessed via enzymatic assays in the Diabetes Research Center Human Physiology Core. The concentration of Low-density lipoprotein is calculated using the Friedewald formula. In addition, levels of liver enzymes are measured via enzymatic assays at the UAB Hospital Outreach Laboratory.
Procedures to measure neurological recovery
Sensory and motor function: The motor and sensory examinations are performed according to the ISNCSCI standards (29–31) and described by AIS. The scales used for the neurological tests for this proposal were selected because they are standard, widely accepted tools and can be readily administered in the acute SCI setting. The AIS Motor Index Score is used for measuring motor function. This index uses standard manual muscle testing on a six-point scale (graded 0–5). The maximum motor score for upper and lower extremity is 50. For the sensory examination, each dermatome is tested for both sharp (pin-prick) and light-touch sensation and sensory status is graded on a three-point scale. Numerically, the sensory scores total 116 points.
Electrical perception thresholds (EPTs) are recorded according to established standards (23). In brief, electrical stimuli are delivered at 3 Hz at key points within cervical dermatomes (C4 to C8). Stimulation commences at 0 mA and progressively increase at 0.1-mA intervals until the participant perceives a stimulus. The stimulus is then turned down until perception is lost. This process is repeated three times to generate an average electrical perception threshold for each dermatome.
Procedures to measure functional independence
The SCIM is used for measuring functional independence (24, 25) and is, at present, the only comprehensive rating scale that measures the ability of patients with SCI to perform everyday tasks according to their value for the patient. It requires no manual testing and the range of the total score is 0–100. Although one American Spinal Injury Association (ASIA) In-Step certified clinician performs all examinations, inter-rater reliability between the primary examiner and back-up examiners are assessed periodically as a quality assurance measure for consistency and a precautionary measure in the event that a back-up examiner might be required. Examiners are blinded to the diet assignment. Neurological examinations are routine both in the acute and in-patient care (in the SRC) facilities under the supervision of physiatrist.
Procedures to determine gut microbiome composition
Microbiome collection and analysis are performed using established protocols (18). Briefly, stools samples are collected in ParaPak vials (Meridian Biosciences, Inc; Cincinnati, OH) and consequently diluted to 0.1 mg/ml in Cary-Blair medium for a total volume of 20 mL with 10% glycerol (by volume). Aliquots of 5 mL are stored at − 80 °C in cryovial tubes until DNA extraction. Stool bacterial DNA is extracted using a Zymo Research Fecal DNA isolation kit (Zymo Research; Irvine, CA) per the manufacturer’s instructions. Polymerase chain reaction (PCR) is then used to amplify the V4 region of the 16S rRNA gene. The PCR products are then retrieved after separated on an agarose gel using electrophoresis, excised from the gel, and subsequently purified with a QIAquick Gel Extraction Kit (Qiagen; Germantown, MD). The Illumina MiSeq DNA sequencing platform is used to sequence 250 base paired-end kits.
Bioinformatics analysis. Microbiome analyses will be performed using the Quantitative Insights into Microbial Ecology (QIIME) bioinformatics software and QWRAP program as previously described (18). Briefly, quality assessment and filtering low-quality data will be performed using the FASTQC and FASTX toolset, respectively. A combination of tools within the QIIME suite will be utilized for clustering reads into operational taxonomy units (OTU) (uclust), taxa assignment (RDP classifier using the Greengenes 16S rDNA database), and as necessary, alignment and phylogenetic inference (using PyNAST and FastTree). These procedures will allow us to quantitatively assess the microbiome population down to the genus, if not species level. For comparative analyses, indices of alpha (Chao1, Observed species, Phylogenetic distance, Simpson’ index, and Shannon’s index) and beta diversity (Bray–Curtis dissimilarity, unweighted UniFrac and weighted UniFrac) will be computed (18).
Procedures to quantify serum biomarkers that may be linked to improvements in neurological recovery and functional independence
a) Sample Processing: Serum samples are processed using disposable affinity-depletion cartridges (NorGen Biotek Corp) that remove the major serum proteins albumin, alpha-antitrypsin, transferrin, and haptoglobin (32). Protein samples are evaporated and resuspended in 6 M urea, 100 mM Tris buffer at 10 mg/mL. Aliquots of the affinity-depleted serum are reduced using dithiothreitol (25 mM) at 50 °C for 30 min followed by alkylation of free thiol groups with iodoacetamide (55 mM) for 30 min in the dark at room temperature. Solutions are then diluted 1/10 in double-distilled water (ddH2O) and overnight digestion is carried out using mass spectrometry (MS)-grade trypsin (0.2 µg/µL). The digests are evaporated to dryness in a Speedvac and then resuspended in 50 µL of ddH2O with 0.1% formic acid.
b) Targeted protein analysis: Peptides representing fibrinogen, extracellular signal-regulated kinase 1/2, CD11b/CD18 integrin receptor, and the epidermal growth factor is selected from either preliminary data (fibrinogen) or using Skyline, a targeted proteomics software tool (33). Predicted product ions from each peptide are used to set up a multiple reaction monitoring liquid chromatography tandem MS (LC-MS/MS) assay carried out on a SCIEX 6500 Qtrap using micro-flow LC. This instrument is very sensitive and is capable of detecting peptides in the amol range. The tryptic peptides are resolved on a 5–50% linear gradient of acetonitrile in 0.1% formic acid mobile phase before entering the electrospray ionization (ESI) interface of the mass spectrometer. Synthetic peptides will be used to obtain quantitative data. To obtain absolute quantitative data on the selected peptide(s), 13C/15N-isotopically labeled versions of the peptide(s) will be prepared (34).
c) Protein Library Generation: An aliquot (5 µL) of each digest is loaded onto a Nano cHiPLC 200 µm × 0.5 mm ChromXP C18-CL 3-µm 120-Å reverse-phase trap cartridge (Eksigent, Dublin, CA) at 2 µL/min coupled to an Eksigent 415 LC system autosampler. After washing the cartridge for 10 min with 0.1% formic acid in ddH2O, the bound peptides are flushed onto a Nano cHiPLC column (200 µm ID × 15 cm ChromXP C18-CL 3 µm 120 Å, Eksigent) with a 100-min linear (5–50%) acetonitrile gradient in 0.1% formic acid at 1000 nL/min using an Eksigent Nano1D + LC. The column then is washed with 90% acetonitrile-0.1% formic acid for 5 min and reequilibrated with 5% acetonitrile-0.1% formic acid for 15 min. A SCIEX 5600 TripleTof mass spectrometer(SCIEX, Toronto, Canada) is used to analyze the protein digest. The ionization spray voltage is set at 2300 V and the declustering potential set at 80 V. The ionization spray and curtain gases are set at 10 psi and 25 psi, respectively. The interface heater temperature is 120 °C. In each 1.25-s duty cycle, eluted peptides are subjected to a 250-ms time-of-flight survey scan from m/z 400–1250 to determine the top 20 most intense ions for MS/MS analysis. Product ion time-of-flight scans at 50 ms are carried out to obtain the tandem mass spectra of the selected parent ions over the range from m/z 400–1000. Spectra is centroided and de-isotoped by Analyst software, version 1.7 TF (SCIEX). A β-galactosidase trypsin digest is used to establish and confirm the mass accuracy of the mass spectrometer. The MS/MS data is processed to provide protein identifications using an in-house Protein Pilot 4.5 search engine (SCIEX) using the UniProt Homo sapiens protein database and a trypsin digestion parameter. Proteins are included in the SWATH library on the criteria of having at least two peptides detected with a confidence score of 95% or greater using the Paradigm method imbedded in the Protein Pilot software.
d) SWATH-MS Analysis (35): Individual samples are analyzed by the same nanoLC-ESI-MS/MS method described above. SWATH data are collected in 1.8-s duty cycles - a 200-ms time-of-flight survey scan from m/z 400–1250 followed by thirty-two, 50-ms successive tandem mass spectra using 25 m/z mass windows. The MS/MS data are de-convoluted and processed using PeakView™ 1.2 with the SWATH Application (SCIEX). This software uses a pre-generated protein identification library to construct individual peptide signatures with fragmentation patterns after analyses for protein identifications. In addition, the areas of the peptide peaks in the MS/MS spectra are used for protein quantification. Identification of proteins associated with the phenotypes being studied will be carried out by univariate (Volcano plots) and multivariate (principal components and partial least squares-discriminant analysis) analyses and associated heat-maps.
Power calculations were performed using nQuery Advisor + nTerim 3.0, and assume a two-sided statistical test and a significance level of 5%. We obtained estimates of the standard deviation for the motor score of 23.5 and for the light touch score of 33 (both based on our preliminary data). With a final sample size of 24 participants per group, and also assuming a two-group t-test and the prior assumptions, we have 80% power to detect between-group differences of 19.5 in the motor score and 27.3 in the light touch score as being statistically significant. With a final sample size of 24 participants per group, and assuming a paired t-test and the prior assumptions, we have 80% power to detect within-group changes of 14.1 in the motor score and 19.8 in the light touch score as being statistically significant. All of these differences are approximately at the same levels as those that Kramer et al. (2012) detected as being statistically significant and clinically meaningful (36). We believe that these estimates are conservative since we are performing our primary statistical analyses for between-group and within-group comparisons simultaneously, using statistical methods that are more sophisticated than those that are assumed here. Given our history in SCI patient recruitment and the large SCI patient base at the UAB SRC, we fully expect to meet our recruitment goal.
Randomization and Blinding
Randomization is performed using the block randomization method, with a block size of 4 (Study statistician). A randomization list generated, and the random assignments were placed into closed envelopes. Each study participant opens one of these envelopes to learn of his/her group assignment. Study patients will inevitably know their group assignment, as we are limited by difficulties inadequate masking of the ketogenic diet, which is very restrictive and requires avoidance of usual foods like bread, pasta, rice, potatoes, and a wide variety of fruits. In order to avoid a potential placebo effect or a participant biasing results due to knowledge of their diet, during the consenting process we simply explain that the study is designed to examine the impact of KD and SD upon outcomes. In addition, individuals performing clinical and laboratory tests and the study statistician were blinded to the study interventions.
To protect privacy, each study participant is assigned a unique 4-digit identification number that cannot be traced to any protected health information (PHI). All data forms, participant information, and biological specimens are coded using these ID numbers. No personally identifiable information (PHI) appears on these materials; instead, the keys linking participants’ identities to their unique identification numbers are stored separately in a secure software system designed for clinical trials, which meets both HIPAA and 21 CFR Part 11 requirements.
Data Capture, Verification, and Disposition. All study data is managed in a central database using REDCap, a secure web software system designed for clinical trials, which meets both HIPAA and 21 CFR Part 11 requirements. All questionnaires and daily surveys are coded into REDCap and administered electronically. Data from serum assays, resting energy expenditure measurements, and sensory measures via dermatomal somatosensory evoked potentials and electrical perception thresholds are outputted from the respective technical equipment in electronic form and then uploaded directly into the REDCap database. However, data on adverse events that are collected by the medical team are recorded on paper forms, stored in the Clinical Research Unit, and then manually entered into the REDCap database. Neurological data (assessed by ISNCSCI standards) and food intake data are captured on paper forms and double entered into the database.
Paper documents are scanned and saved on UAB's secure network and/or stored in locked cabinets in the research coordinator’s office. The PI and post-doctoral trainee periodically check the database for missing data and document all such data and the reasons for absence. The study statistician also periodically performs a quality check of the database, which includes error checks using expected ranges for data values. After the study has completed, all data quality will be checked by the statistician, PI, and the post-doctoral trainee in year 3; tasks are divided among the three personnel according to their respective responsibilities. In compliance with UAB Institutional Review Board (IRB) policies, all data and records will be kept for at least 3 years after the study is completed; all PHI for participants will be deleted 3 years after the trial is completed, while the de-identified final dataset will be retained indefinitely and published online for other scientists to benefit from.