Caffeine Enhances Intermittent Hypoxia-Induced Gains in Walking Function for People with Chronic Spinal Cord Injury

Abstract

Incomplete spinal cord injury (iSCI) often results in lifelong walking impairments that limit functional independence. Thus, treatments that trigger enduring improvement in walking after iSCI are in high demand. Breathing brief episodes of low oxygen (i.e., acute intermittent hypoxia, AIH) enhances breathing and walking function in rodents and humans with chronic iSCI. Pre-clinical studies found that AIH also causes the accumulation of extracellular adenosine that undermines AIH-induced functional plasticity. Pharmacologically blocking adenosine A2a receptors (A2aR) prior to AIH resulted in a dramatic improvement in motor facilitation in rodents with iSCI; however, a similar beneficial effect in humans is unclear. Thus, we conducted a double-blind, placebo-controlled, crossover randomized study to test the hypothesis that a non-selective A2aR antagonist (i.e., caffeine) enhances AIH-induced effects on walking function in people with chronic (≥1yr) iSCI. We enrolled 12 participants to receive daily (5 days) caffeine or placebo (4 mg/kg) 30 min before breathing 15, 1.5-min low oxygen (AIH; F_IO₂ = 0.10) or SHAM (F_IO₂ = 0.21) episodes with 1-min intervals. We quantified walking function as the change in the 10-meter walk test (speed) and 6-min walk test (endurance) relative to baseline, on Day 5 post-intervention, and on follow-up Days 12 and 19. Participants walked faster (Day 19; p < 0.001) and farther (Day 19; p = 0.012) after caffeine+AIH and the boost in speed persisted more than after placebo+AIH or caffeine+SHAM (Day 19; p < 0.05). These results support our hypothesis that a caffeine pre-treatment to AIH training shows promise as a strategy to augment walking speed in persons with chronic iSCI.

Introduction

There is no cure for spinal cord injury (SCI), and no single treatment restores lost walking ability. While most injuries are incomplete (iSCI) with some spontaneous recovery,¹ functional deficits persist and limit the quality of life of those affected. New combinatorial treatments that trigger neural plasticity are needed to restore walking ability and functional independence.

Breathing brief, moderate bouts of low oxygen (acute intermittent hypoxia, [AIH]) is emerging as a treatment that elicits spinal plasticity and improves motor function.² AIH causes episodic serotonin release³ that leads to increases in brain-derived neurotrophic factor (BDNF) and tropomyosin receptor kinase B (TrkB) activation,⁴ which results in increased synaptic strength⁵ and motor output.⁶ This mechanism translates to not only improved respiratory but also non-respiratory motor function. Seven consecutive days of AIH nearly restored ladder walking ability in rodents with cervical SCI⁶ and daily (5 days) AIH enhanced overground walking speed and endurance in humans with chronic iSCI.²

During episodes of breathing low oxygen, the accumulation of extracellular adenosine may undermine the efficacy of AIH-induced motor facilitation.⁷ While repetitive AIH induces motor facilitation predominantly via serotonin-dependent mechanisms,⁸ even mild hypoxia causes ATP and adenosine release from glia and other sources in the nervous system.⁷ Consequently, blocking adenosine A2a receptors (A2aR) enhances AIH-induced respiratory motor plasticity by nearly 50% in rodents with iSCI.⁹ The present study determined if A2aR antagonism also enhances AIH-induced motor function in humans with spinal injury. Specifically, we tested the hypothesis that caffeine, a non-selective adenosine receptor antagonist¹⁰ augments AIH-induced walking recovery in people with chronic iSCI.

Methods

Participants

Twelve ambulatory adults with chronic (≥ 1 year post-injury) iSCI as defined using International Standards for Neurological Classification of SCI¹¹ participated in this randomized, double-blinded crossover study (Table 1). Participants were recruited via ClinicalTrials.gov (NCT02323698) and within the Mass General Brigham healthcare system. We included participants having injuries between levels C₃ and L₁, no joint contractures, some voluntary ankle, knee, and hip movements, and the ability to ambulate without human assistance. Participants were encouraged to use assistive devices as needed for safe ambulation during the walking assessments. We excluded participants with progressive SCI or severe concurrent medical conditions.² Treatments and assessments occurred at Spaulding Rehabilitation Hospital as detailed in Table S1 in the Supplementary Material. Participants provided written informed consent to enroll in the study and the institutional review board approved the study protocol at Mass General Brigham. This study followed the Consolidated Standards of Reporting Trials (Supplementary Material).

Table 1.

Participant Demographics

Participant	Age (years)	Gender	LOI/ ISNCSCI Scale	TSI (years)	Walking Aid	CYP1A2*1F SNP
S01	54	M	C3/D	2	RW	C/A
S02	63	M	C3/D	16	None	C/A
S03	37	F	C5/C	15	Crutches	A/A
S04	66	M	C5/D	43	Cane x 1	C/A
S05	33	F	T12/C	3	Cane x 2	C/A
S06	53	M	C5/C	1	RW	C/A
S07	59	M	C5/D	5	RW	A/A
S08	53	M	T11/D	5	None	C/A
S09	64	F	C5/D	1	Cane x 2	–
S10	41	F	L1/C	1	None	C/A
S11	71	M	C4/D	2	RW	C/A
S12	29	M	T2/D	2	RW	A/A
Mean ± SD	52 ± 14			8 ± 12

ISNCSCI, International Standards for Neurological Classification of SCI⁷; LOI, level of injury; SD, 1 standard deviation; TSI, time since injury; RW, rolling walker; CYP1A2, cytochrome P450 family 1, subfamily A, member 2; CYP1A2^*1F SNP, single nucleotide polymorphism, rs762551.

Intervention

For 5 consecutive days, participants received ∼4 mg/kg caffeine (C) or placebo (P) capsules 30 min prior to AIH (15 episodes/day: 1.5 min of F_IO₂ = 0.10 ± 0.02; 1 min normoxic intervals) or SHAM (F_IO₂ = 0.21 ± 0.02).² We administered the AIH and SHAM treatments using an automated gas mixing system as reported previously (Fig. 1).¹² The caffeine dosage, similar to 16-20 oz (∼300-400 mg) brewed coffee, ensured A2aR antagonism with minimal risk of adverse side effects.¹³ Participants were instructed to completely abstain from caffeine during all treatment days. C+AIH, P+AIH, and C+SHAM intervention arms were assigned using a random number generator. The order of interventions was balanced across participants with at least 2-week washout periods. We administered placebo for caffeine and a sham for AIH as control conditions in our randomized crossover trial.

FIG. 1.

(A) Acute intermittent hypoxia (AIH) treatment setup. During AIH or SHAM, we monitor F_IO₂, the fraction of inspired oxygen. BP, blood pressure; HR, heart rate; SpO₂, peripheral blood oxygen saturation. (B, top) A single session consisted of caffeine or placebo (∼4 mg/kg) consumption 30 min before a breathing protocol of 15 episodes of 1.5 min of low oxygen (F_IO₂ = 0.10) alternating with 1 min of room air (F_IO₂ = 0.21). The protocol timeline (B, bottom) illustrates the assessment and treatment days; the vertical lines correspond to baseline (BL), treatment Days 1-5 (D, bold), and follow-up Days 12 and 19.

We monitored for pain, dizziness, altered vision, respiratory distress, autonomic dysreflexia, and change in mental function during and after treatment. Cardiovascular responses also were monitored (Radical-7, Masimo Corp, USA) before, during, and immediately following C+AIH, P+AIH, and C+SHAM treatments (Fig. 1). We required participants' heart rate to be between 40 and 160 BPM, systolic pressure between 85 mm Hg and 160 mm Hg, and SpO₂ to remain above 75% throughout the duration of treatment sessions.²

Blood samples were drawn before and immediately after treatment on Days 1 (D₁) and 5 (D₅) of each arm to test for C-reactive protein (CRP) levels as an inflammatory marker and to confirm caffeine levels in blood serum (Table S2 in the Supplementary Material). Participants also provided a saliva sample once during the study for genetic testing of CYP1A2 polymorphisms relevant to caffeine metabolism.

Outcome measures

Evaluators blinded to both the breathing and pharmacological interventions assessed walking ability at baseline (BL) and within 60 min post-AIH/SHAM on the 5th treatment day (D₅). The median 10-meter walk test (10MWT) baseline times did not differ between the treatment rounds (Median Test, p = 0.638). Follow-up walking assessments occurred 1 week (D₁₂) and 2 weeks (D₁₉) post-intervention (Fig. 1B). We quantified the primary outcome measure walking speed using the 10MWT. Using the 6-min walk test (6MWT), we quantified the secondary outcome measure walking distance. Participants used preferred assistive devices for all testing when necessary for safe ambulation (Table 2). These walking assessments have high inter-rater and intra-rater reliability.¹⁴

Table 2.

Participant Treatment Order, Baseline Walking Speeds, and Handheld Walking Aids across Treatment Rounds

Participant	Round 1			Round 2			Round 3
	Treatment	10MWT (m/sec)	Walking aid	Treatment	10MWT (m/sec)	Walking aid	Treatment	10MWT (m/sec)	Walking aid
S01	P+AIH	0.2	RW	C+AIH	0.2	RW	C+SHAM	0.3	RW
S02	P+AIH	0.9	none	C+SHAM	0.9	none	C+AIH	0.9	none
S03	C+SHAM	0.1	Crutches	P+AIH	0.1	Crutches	C+AIH	0.2	Crutches
S04	C+AIH	1.1	Cane x 1	C+SHAM	1.1	none	P+AIH	-	n/a
S05	C+SHAM	1.3	Cane x 2	C+AIH	0.9	Cane x 2	P+AIH	1.0	Cane x 2
S06	C+AIH	0.5	RW	P+AIH	0.7	RW	C+SHAM	0.7	RW
S07	C+AIH	0.4	RW	P+AIH	0.4	RW	C+SHAM	0.5	RW
S08	C+SHAM	2.2	none	P+AIH	-	n/a	C+AIH	1.9	none
S09	C+SHAM	-	n/a	C+AIH	-	n/a	P+AIH	-	n/a
S10	P+AIH	-	n/a	C+SHAM	-	n/a	C+AIH	-	n/a
S11	P+AIH	0.1	RW	C+AIH	0.1	RW	C+SHAM	0.1	RW
S12	C+AIH	0.5	RW	C+SHAM	0.3	Crutches	P+AIH	0.5	Crutches
Median Speed (m/sec)		0.5			0.4			0.5

P+AIH, placebo and acute intermittent hypoxia (AIH); C+AIH, caffeine and AIH; C+SHAM, caffeine and sham AIH; 10MWT, 10-meter walk test at baseline; RW, rolling walker.

Statistical analysis

We used SPSS 28 (IBM Inc.) for statistical analyses. Results were considered significant at p < 0.05 and reported as mean differences (MDs) with 95% confidence intervals (CIs). Using a linear mixed model with fixed effects (LMM),¹⁵ we compared differences in 10MWT time (primary outcome measure) and 6MWT distance (secondary outcome measure) from baseline and between treatments. The LMM is sensitive to within-participant changes and robust against variability, small sample sizes, and missing/unbalanced data sets as compared with other statistical models.^15,16 Time and treatment corresponded to the main effects. We found no difference in variance between groups at D₁₉ (Levene Test, p > 0.05).¹⁷ Since the CYP1A2 enzyme metabolizes caffeine,¹⁸ we explored the association between CYP1A2 gene mutation (rs762551 single nucleotide polymorphism) and caffeine-related treatment effects on walking speed and endurance at Day 5. Specifically, we compared treatment effects between participants classified as “slow” (C/C and C/A genotypes) and “fast” (A/A genotype) metabolizers¹⁹ to determine if the rate of caffeine metabolism influenced outcomes (Table 1).

Results

Participants receiving caffeine (C) or AIH (n = 10) improved overground walking ability. However, C+AIH increased walking speed more than either treatment alone (Fig. 2). 10MWT time improved with C+AIH versus baseline at D₅ (MD 4.9 sec, CI 1.1-8.8 sec, p = 0.012), D₁₂ (MD 5.5 sec, CI 1.7-9.3 sec, p = 0.005), and D₁₉ (MD 7.5 sec, CI 3.7-11.3 sec, p < 0.001). At D₁₉, walking speed was greater for C+AIH versus P+AIH (MD 8.2 sec, CI 3.0-13.5 sec, p = 0.002) and C+SHAM (MD 7.09 sec, CI 2.0-12.2 sec, p = 0.007). Participants with CYP1A2*1F A/A alleles (n = 3) improved their 10MWT time by 9.8 ± 3.6 sec versus 2.6 ± 1.1 sec for C/A heterozygotes at D₅. Walking endurance also increased after daily C+AIH, but this improvement did not differ from P+AIH at D₅, D₁₂, and D₁₉. The 6MWT distance improved with C+AIH versus baseline at D₅ (24.1 m, CI 6.1-42.1 m, p = 0.009), D₁₂ (23.7 m, CI 6.9-40.6 m, p = 0.006), and D₁₉ (22.4 m, CI 5.0-39.8 m, p = 0.012). 6MWT distance also improved with P+AIH versus baseline at D₅ (17.2 m, CI 0.8-33.5 m, p = 0.040) and D₁₉ (24.3 m, CI 7.2-41.3 m, p = 0.006). Improvements persisted at D₁₉ for C+AIH (26.9 m, CI 3.9-49.8 m, p = 0.022) and P+AIH (25.1 CI, 1.6-48.5 m, p = 0.037) compared with C+SHAM (Fig. 2).

FIG. 2.

(A) Bars represent mean ± one standard error changes in the 10-meter walk test (10MWT) time (sec) across all subjects (n = 10) at each time-point for daily (5 days) placebo (P) combined with AIH (P+AIH, white), caffeine (C) combined with AIH (C+AIH, black), and caffeine combined with normoxia (C+SHAM, gray). Decreases in time correspond to increased walking speed relative to baseline (BL). Participant changes in 10MWT times (sec) relative to BL at Days 5 (D₅), 12 (D₁₂), and 19 (D₁₉) after daily C+AIH (B). (C) Bars represent mean ± 1 standard error changes in the 6-min walk test (6MWT) distance (m) across all subjects at each time-point for daily P+AIH, C+AIH, and C+SHAM. Participant changes in 6MWT distance (m) relative to BL at D₅, D₁₂, and D₁₉ following daily C+AIH (D). Single asterisks (*) indicate significance relative to BL, and brackets with single asterisks indicate significant differences between interventions at p < 0.05 level. Hash marks (#) indicate outcomes for participants with the CYP1A2*1F A/A genotype.

We quantified the integrity of our blinding conditions by questioning participants on whether they could identify the treatments they received (caffeine or placebo, SHAM or AIH) at the end of treatment D₅. Although the imbalance of our treatment arms (two of three treatments included caffeine, two of three treatments included AIH) gave participants a 67% chance of guessing the treatment correctly, the study cohort's guess of placebo vs caffeine (specificity = 0.46) or SHAM vs AIH (specificity = 0.50) were no better than chance.

Participants did not report any unexpected discomfort before, during, or after treatments. Participants received each treatment at a consistent time window (±1 h) across the three rounds.² Participants also received at least 2-week washout periods between rounds with an average of 10 ± 13 weeks (median = 5 weeks). We found no difference in washout duration between the three treatment arms (Mann-Whitney U-test, p = 0.824). Of the 12 participants who enrolled, one subject (S09) was unable to complete the study due to COVID-19 restrictions. Baseline assessment of S08 was identified as an outlier using a labeling multiplier²⁰ of 2.2 and was excluded from subsequent analysis; this appeared to be due to excessive pain unrelated to the study. We found no differences in heart rate, blood pressure, or SpO₂ between baseline and post-treatment on D₅ in any intervention (p > 0.05). No serious adverse events occurred during this study.

Discussion

Our data provide preliminary evidence that consuming caffeine prior to AIH holds promise as a safe, effective intervention to restore walking function in persons with chronic iSCI. With C+AIH, the primary benefit was increased speed. All participants improved their walking ability on the 10MWT or 6MWT, and improvements persisted for at least two weeks in 9 of 10 participants. Caffeine prior to AIH boosted walking speed more than either alone (i.e., C+SHAM, P+AIH). These results appear to complement prior pre-clinical findings that an A2aR antagonist enhances AIH-induced motor function after iSCI.^7,21

Current walking therapies for persons with iSCI require up to 12 weeks, yet effect sizes and clinically meaningful improvements remain limited.²² The brief (5-day) combinatorial treatment studied here decreased 10MWT times by 22% at 14 days post-treatment. Seven of 10 participants demonstrated an increase in walking speed that exceeded the minimal clinically important difference of 0.06 m/sec at 2 weeks post-treatment.²³ The persistence of this improvement was the most striking difference between the combinatorial therapy and the control arms. The changes in walking speed following C+AIH are profound considering the short time course (one week) as compared with several weeks of contemporary gait training.²⁴

We did not predict that 5 consecutive days of C+AIH would dramatically change functional walking behaviors (e.g., reduced reliance on walking aids) in this study cohort due, in part, to no gait training protocol. Prior studies in humans and rodents with SCI found that the largest gains in AIH-induced walking occur when the “primer” precedes locomotor/walking training.^2,6,9,25 We speculate that the participants who relied on handheld walking aids (n = 7) may benefit from walking training. Since the study did not involve skilled walking training, we avoided encouraging participants to use less restrictive devices during the walking assessments. Despite this safety precaution, two participants independently reduced reliance on their handheld walking aids after Round 1 (Table 2). Participant S04 converted from a single-point cane to no walking aid and participant S12 converted from a rolling walker to crutches. Future studies may help determine if combining C+AIH with task-specific gait training to minimize the use of handheld walking aids may lead to greater functional gains in overground walking.

Caffeine prior to AIH did not result in greater endurance gains than AIH alone (i.e., P+AIH). While the improvement in walking endurance as measured by the 6MWT had a smaller effect size than walking speed, it was notable that only in the AIH conditions were statistically significant changes found. This was an expected outcome that matches the results of a prior study showing that AIH alone had a similar average improvement in 6MWT distance.² Studies pairing AIH with walking training showed a significant improvement in walking endurance,^2,25 which could indicate that participants had increased home and community ambulation despite the lack of formal walking training in our study.

We identified five participants who improved their 10MWT time, but they showed little to no change in their 6MWT distance. Although there are several competing factors that may contribute to this outcome, we suspect that improvements in walking speed may be limited for those who are high functioning due to a possible ceiling effect in the 10MWT. For instance, the greatest speed gains appear to correspond to those participants who walked slowest and who relied on bilateral hand-held walking aids (S01, S03, S11), while those participants who walked fastest and who required little to no walking aides (S04, S08) showed the largest improvement in 6MWT distance. This result appears to align with a recent result that showed daily AIH combined with walking practice elicited the greatest improvements in 10MWT time for participants who relied on bilateral arm-driven walking aids.²⁵

The lack of a significant acute impact of caffeine alone on 6MWT distance on D₅ seems to contradict evidence in able-bodied athletes showing a modest improvement in endurance following caffeine ingestion.²⁶ The methodology used in these studies (e.g., time trials or time-to-exhaustion) is intended to measure the extremes of endurance which prevent a direct comparison to the 6MWT which aims to represent activities of daily living. However, a time trial-based study of spinal cord injured athletes also showed that caffeine had no effect on endurance,²⁷ which may suggest that the pathway through which caffeine acutely enhances endurance is altered by SCI.

Further study of A2aR antagonists in rehabilitation treatments is needed. More significant effects in participants with faster caffeine metabolism may be due to faster A2aR binding or greater selectivity. We acknowledge that caffeine's multiple physiological targets may contribute to the varying degree of walking improvements seen following caffeine pre-treatment. Caffeine is a safe but non-selective A2aR antagonist.²⁸ At low plasma concentrations (1-10 μM) after a single serving of coffee, caffeine has a high affinity for adenosine A1, A2a, and A2b receptors²⁸ as well as triggering norepinephrine release.²⁹ Thus, a selective A2aR antagonist, such as istradefylline, may offer more effective adenosine receptor inhibition to improve AIH as a therapeutic modality.⁷

Study limitations

We acknowledge several study design limitations in our randomized clinical trial that reduce the generalizability of our findings. These include small sample size, large response variability, and potential for washout biases. We implemented a balanced, cross-over study design to minimize the influence of these factors as well as to overcome our recruitment barriers. Indeed, response variability and recruitment challenges are common limitations with SCI clinical trials.³⁰ While our preliminary study showed caffeine and AIH elicited the greatest effect on walking speed, the effect sizes varied with dramatic effects from a few study participants (Fig. 2). The crossover design limited our ability to assess the potential long-term impact of C+AIH on walking recovery. The results do not show a clear washout period and prior studies do not indicate the half-life of the daily AIH intervention. The potential influences of genetic (CYP1A2 gene variants) and demographic factors (age, gender, SCI level, and severity) on the C+AIH effect size are probable, but the study design limitations weaken our interpretation of their influences. Important clinical questions remain concerning the interactions between these factors and dose on the combinatorial treatment's long-term effectiveness for improving walking ability in persons with chronic SCI.

Conclusions

Safe and effective combinatorial treatments that improve walking function in people with chronic iSCI are in high demand. In this preliminary clinical trial, we found combining a non-selective A2aR antagonist (caffeine) with a plasticity-promoting treatment (AIH) may offer a rapid, non-invasive boost in overground walking speed for persons with lifelong paralysis after SCI. Removing adenosine-dependent constraints on AIH shows promise as a pre-treatment to enhance functional mobility after chronic iSCI.

Transparency, Rigor, and Reproducibility Summary

The cross-over study design and analysis plan were preregistered on ClinicalTrials.gov (NCT02323698). Prespecified sample size was n = 17, yielding statistical power of 0.70 for detection of a change in 10MWT time (f = 0.7, F_3,12 = 3.5; p = 0.4, α = .05). During enrollment, subjects were assigned to C+AIH, P+AIH, and C+SHAM intervention arms using a random number generator. The order of interventions was balanced across subjects. We consented 34 individuals, 22 lost during allocation due to relocation, change in medical status, and COVID-19 pandemic-related restrictions. Ten participants completed the study (see Supplemental Material). In the present study, we had n = 9 cross-over participants complete daily (5 consecutive days) of C+AIH, C+SHAM, and P+AIH. Although the calculated variances are larger than the planned value, the observed change in 10MWT is larger than expected. That leads to an observed effect size of 0.64 using the change of 10MWT at D₁₉ from baseline in the C+AIH group with n = 10, which is close to the planned effect size of 0.7. A post hoc power calculation with multivariate T² test for the 4-period cross-over design, a sample size of n = 10 would be required to have the statistical power of 80% at α = 0.05 when the effect sizes of change from baseline are estimated to be 0.5 (D₅), 0.55 (D₁₂), 0.65 (D₁₉) from the current study, and the correlation between outcomes from different periods is 0.8. A minimum of a 2-week washout yielded intervention arms with baseline 10MWT medians the same across the intervention rounds (Median Test, p = 0.638). The distribution of baseline 10MWT also is the same across the intervention rounds (Kruskal-Wallis Test, p = 0.731). All primary outcomes were assessed by investigators blinded to the intervention arm assignment. All materials required to perform the assessments and interventions may be available upon request. The primary outcome measure (10-meter walk test) and secondary outcome measure (6-minute walk test) are standard in the field and have high test-retest reliability. We used the Kolmogorov-Smirnov normality test for the primary outcome measure at D₁₉ for each intervention arm: P+AIH (p = 0.200), C+AIH (p = 0.088), and C+SHAM (p = 0.200). Correction for multiple comparisons was performed for the outcome measures. The findings have not yet been replicated or externally validated. Data are available upon request and the manuscript is open access.

Authors' Contributions

Randy D. Trumbower, P.T., Ph.D. was the principal investigator for this study. He contributed to the development of study concept and design and data analyses and interpretation, literature search, figures, and manuscript revisions. As the corresponding author, he had full access to all data and final responsibility for the decision to submit for publication. Stella Barth, B.S., Christopher Tuthill, Ph.D., and Chloe Slocum, M.D., M.P.H. contributed to study design, participant recruitment, data collection, analyses, and interpretation. Ms. Barth collected blood and saliva samples for testing. Drs. Trumbower, Shan, Barth, and Tuthill performed statistical analyses. Drs. Trumbower, Barth, Tuthill, Slocum, Shan, Zafonte, and Mitchell wrote the first draft of the manuscript and contributed to manuscript revisions and approval of the final submission. Randy D. Trumbower, P.T., Ph.D., Stella Barth, B.S., Ross Zafonte, D.O., and Chloe Slocum, M.D., M.P.H. contributed to participant recruitment, screening, blind rating, data collection, and analyses, as well as approval of final submission; Stella Barth, B.S., Randy D. Trumbower, P.T., Ph.D., and Gordon S. Mitchell, Ph.D. contributed to the development of study concept and design, data interpretation, literature search, figures, and manuscript revisions and approval of the final submission.

Funding Information

This study was supported through the Wings for Life Foundation (WFLUS02614) and the National Institutes of Health (HD081274, HL147554). The funding sources had no role in study design, collection, patient recruitment, or interpretation of data.

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Table S1

Supplementary Table S2