Hydrogen inhalation is associated with a transient rightward shift in prefrontal oxyhemoglobin asymmetry and autonomic modulation

Abstract

Molecular hydrogen (H₂) has selective antioxidant and anti-inflammatory properties, yet its immediate effects on human cerebral oxygenation and autonomic function remain unclear. In this study, we evaluated acute central and autonomic responses to a single 30-minute session of hydrogen inhalation in healthy adults, using 99.9% hydrogen delivered via nasal cannula at a fixed flow rate of 300 mL/min. Cerebral oxygenation was assessed using time-domain near-infrared spectroscopy (TD-NIRS) to quantify the concentrations of oxyhemoglobin (oxy-Hb) and deoxyhemoglobin (deoxy-Hb) in the bilateral prefrontal cortex (PFC), and to calculate interhemispheric asymmetry indices before hydrogen inhalation, immediately after the end of inhalation, and at 30 and 90 min thereafter. Autonomic activity was assessed via continuous electrocardiography (ECG) to derive heart rate, R-R interval, and frequency-domain heart rate variability metrics (low-frequency (LF), high-frequency (HF) and LF/HF ratio). Hydrogen inhalation elicited robust, transient increases in the right-PFC asymmetry of the oxy-Hb concentration. Concurrently, the LF/HF ratio increased during inhalation, indicating sympathetic activation, followed by decreases in the heart rate after inhalation, consistent with parasympathetic recovery. These parallel cerebral and autonomic responses suggest a coordinated neurovascular–autonomic coupling in response to hydrogen inhalation. Our findings show that acute hydrogen inhalation transiently modulates PFC oxygenation lateralization and autonomic tone, suggesting potential relevance to cognitive and cardiovascular regulation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-36599-8.

Introduction

Molecular hydrogen (H₂) has emerged as a biologically active gas with selective antioxidant, anti-inflammatory, and anti-apoptotic properties, attracting significant attention as a potential therapeutic agent for various pathological conditions. Since the seminal study by Ohsawa et al., which demonstrated that inhaled hydrogen selectively scavenges hydroxyl radicals and attenuates infarct size in a rat model of cerebral ischemia–reperfusion injury¹, a growing body of evidence has implicated hydrogen in reducing oxidative stress, modulating inflammatory responses, and preserving mitochondrial integrity^2–4. These pleiotropic effects have prompted exploration of hydrogen-based interventions in neurological, cardiovascular, and metabolic disorders, including stroke⁵, myocardial infarction⁶, and neurodegenerative diseases such as Parkinson’s and Alzheimer’s^7,8.

Despite this expanding interest, the acute physiological effects of hydrogen gas in humans, particularly at the level of cerebral and autonomic function, remain incompletely understood. MRI-based pilot studies have suggested that hydrogen inhalation may alleviate cerebral infarct severity or improve neuronal integrity^8,9, yet the precise hemodynamic dynamics underlying these effects have not been clearly described. In this context, time-domain near-infrared spectroscopy (TD-NIRS) represents a powerful noninvasive modality that provides absolute quantification of oxyhemoglobin (oxy-Hb) and deoxyhemoglobin (deoxy-Hb) concentrations in cortical tissue, leveraging photon time-of-flight measurements to isolate absorption from scattering^10,11. TD-NIRS thus offers unique potential for capturing subtle, temporally resolved changes in cerebral oxygenation associated with gaseous interventions.

Concurrently, molecular hydrogen has been postulated to modulate autonomic nervous system activity. Animal and human studies using electrocardiography (ECG) have reported enhanced heart rate variability (HRV), reduced sympathetic dominance, and increased vagal tone following hydrogen administration^12,13. However, the immediate autonomic responses to acute hydrogen inhalation, particularly in relation to heart rate (HR), R-R interval (RRI), and frequency-domain HRV indices, have not yet been systematically characterized in conjunction with cerebral hemodynamic data.

Accordingly, we hypothesized that hydrogen inhalation would induce specific modulations in prefrontal cortex (PFC) hemodynamics, particularly in interhemispheric asymmetry, accompanied by transient shifts in autonomic tone. By simultaneously employing TD-NIRS to monitor PFC cortical oxygenation and ECG-derived metrics to assess autonomic modulation, we aimed to elucidate the central–autonomic interplay evoked by hydrogen inhalation. Particular emphasis was placed on temporal patterns of interhemispheric asymmetry and HRV dynamics, which may offer mechanistic insights into the brain–body coupling processes potentially influenced by molecular hydrogen. Crucially, this study was designed to delineate acute physiological responses rather than to evaluate clinical efficacy and therefore did not include a sham gas control.

Materials and methods

Participants

Fifteen healthy adult volunteers (8 males, 7 females; mean age ± standard deviations (SD): 53.3 ± 12.1 years) were recruited. Participants were recruited and studied between June and July 2024. This study was designed as an exploratory pilot study. No formal sample size calculation was conducted, and the sample size was determined based on feasibility and consistency with previous work using similar experimental paradigms. All participants were free from known neurological, cardiovascular, or psychiatric disorders and were not taking any medications known to affect cerebral hemodynamics or autonomic function. Participants abstained from all medications, over-the-counter drugs, dietary supplements (including functional and health foods), and vigorous physical activity for ≥ 72 h prior to testing, and from alcohol, caffeine, and nicotine for ≥ 24 h before each experimental session. They fasted (water only) for ≥ 12 h from 21:00 h on the evening before testing, and on the day of testing a standardized light meal and 500 mL of water were provided 90 min before data acquisition. All experiments were conducted in a dedicated outpatient examination room arranged for physiological measurements. All measurements were performed in a temperature-controlled room (22-24 °C), and participants rested quietly in a seated position for 10 min before baseline recordings. All methods were carried out in accordance with relevant guidelines and regulations. Written informed consent was obtained from the participate and the study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Ueno Asagao Clinic (approval number: 2024-09). The study was prospectively registered in the UMIN Clinical Trials Registry (UMIN000054586; https://center6.umin.ac.jp/cgi-open-bin/ctr_e/ctr_view.cgi? recptno=R000062375), a WHO International Clinical Trial Registry Platform–recognized primary registry, on June 22, 2024. This manuscript was prepared in accordance with the CONSORT 2025 reporting guideline. The trial protocol and statistical analysis plan are provided as a separate document submitted with the manuscript and are available for peer review.

Hydrogen inhalation protocol

Participants underwent a single, 30-minute session of hydrogen inhalation using 99.9% hydrogen at the source, delivered via nasal cannula at a fixed flow rate of 300 mL/min with a commercial hydrogen generator (AQY-300, CNB Medical Research Institute, Japan). Hydrogen was delivered via nasal cannula at a fixed flow rate while participants breathed room air spontaneously. Accordingly, the inspired hydrogen fraction was not directly measured and likely varied across participants depending on individual ventilation (tidal volume and respiratory rate). Respiratory parameters were not recorded and the inspired gas composition was not sampled, so the inspired hydrogen fraction or hydrogen partial pressure could not be quantified on a per-participant basis. Peripheral oxygen saturation (SpO₂) was not monitored, and inspired oxygen concentration was not sampled during the session. The hydrogen flow rate was calibrated before each session. Heart rate was continuously monitored via ECG throughout the inhalation session, and participants were visually observed by the investigator before, during, and after hydrogen inhalation to ensure their safety. Time zero in the time axis (t in min) was defined as the start time of hydrogen inhalation.

Cerebral hemodynamic measurements (TD-NIRS)

Cerebral oxygenation was assessed using a TD-NIRS system (TRS-20, Hamamatsu Photonics, Japan). The system employs three pulsed laser diodes (wavelengths: 760, 800, and 840 nm) and time-resolved photon detection to quantify absolute concentrations of oxy-Hb and deoxy-Hb. Optodes were bilaterally attached to the forehead with a fixed source–detector separation of 3.0 cm, targeting the left and right PFC. To leverage the TRS system’s capability of absolute quantification, key features of time-domain measurement, we employed a curve fitting of the analytical solution of the pulsed light propagation to the measured distribution of time-of-flight (DTOF). By this fitting, the absorption and reduced scattering coefficients at the three wavelengths were obtained under the assumption of homogeneous tissue for the human forehead. The concentrations of oxy-Hb and deoxy-Hb were calculated from the absorption coefficients at the three wavelengths using the absorption spectra of oxy-Hb and deoxy-Hb. Hereafter for simplicity, “oxy-Hb” and “deoxy-Hb” refer to the concentrations of oxy-Hb and deoxy-Hb, respectively.

TD-NIRS data were acquired for 3 min at four time points: before inhalation (t = −15 min, T1) as the baseline, immediately after the end of inhalation (t = 30 min, T2), 30 min after the end of inhalation (t = 60 min, T3), and 90 min after inhalation (t = 120 min, T4) as shown Fig. 1. Signals were averaged over the 3-minute periods for each hemisphere and time point.

Fig. 1.

Experimental protocol showing the time points of measurements of TD-NIRS, ECG and MBP. Gray shading represents the 30-min H₂ inhalation period (t = 0 min). MBP (mean blood pressure) was measured at five time points: t = − 15, 15, 30, 60, and 120 min (blue bars B1–B5). ECG was recorded continuously from t = − 20 min to t = 110 min, and HRV/R–R interval analyses were performed in the four segments: S1 (t = − 15 to − 5 min), S2 (t = 0 to 30 min), S3 (t = 30 to 40 min), and S4 (t = 60 to 70 min) (double-headed arrows). TD-NIRS measurements (3 min each time) were performed at four time points: t = at − 15 min (T1), 30 min (T2), 60 min (T3), and 120 min (T4) (red bars).

Asymmetry index calculation

To quantify interhemispheric balance, we calculated an asymmetry index (AI) for both the oxy-Hb and deoxy-Hb at each time point^14,15. The AI for the oxy-Hb concentration was calculated using the following formula:

where “Right” and “Left” refer to the measured oxy-Hb averaged over 3 min at the right and left PFC, respectively. This metric yields a normalized value between − 1 and 1, with positive values indicating right-dominant oxygenation. The same formula was applied to the measured deoxy-Hb to assess asymmetries in deoxygenation.

Exploratory data quality control

Thirty-three time-domain NIRS–derived features were extracted from the measured data, particularly the DTOFs obtained at the three wavelengths (WLs), all the time points and both sides of the forehead. These features included peak times of the DTOFs at the 3 WLs, peak values of the DTOFs at the 3 WLs, full width at half maximum of the DTOFs at the 3 WLs, integrated photon counts of the DTOFs at the 3 WLs and their sum over the 3 WLs, optical pathlengths at the 3 WLs, absorption coefficients at the 3 WLs, reduced scattering coefficients at the 3 WLs, oxy-Hb, deoxy-Hb and total-Hb (= oxy-Hb + deoxy-Hb) with their standard deviations, oxygen saturation of the tissue with its standard deviation (Supplementary Fig. 1), and time from the start of the measurement was added.

Principal component analysis (PCA) was performed on a total of the 66 TD-NIRS features from both the left and right forehead. Outliers were identified as the participants whose first three PCs (PC1 to PC3) were placed at the greatest Euclidean distances from the origin in the 3D PCA space (Supplementary Fig. 2). This method captures multivariate deviation rather than univariate outliers along individual PCs, and is equivalent to a multivariate outlier detection approach. Based on this criterion, four participants were excluded from further analysis. All analyses of the TD-NIRS outcomes were performed in participants who completed the experimental session and met the predefined data quality control criteria. There were no missing data in the TD-NIRS outcome measures included in the analyses.

Cardiac and autonomic measurements

Electrocardiographic (ECG) signals were continuously recorded from 15 min before to 90 min after hydrogen inhalation using a wearable monitor (myBeat, Union Tool Co., Tokyo, Japan) secured at the epigastrium. QRS detection enabled beat-to-beat RRI (in ms) calculation. Raw signals were visually inspected, and RRIs < 300 ms or > 2,000 ms, or those deviating by > 30 bpm from an eight-beat running average, were removed. HR was calculated as 60,000/RRI. R-R interval is the primary beat-to-beat signal used for HRV computation and artifact screening. Therefore, we report both HR and R-R interval for transparency, while explicitly acknowledging their inverse mathematical relationship and focusing interpretation on HRV indices and the direction of chronotropic change.

Although ECG was continuously recorded up to 90 min after inhalation, HRV frequency analysis was limited to segments aligned with the TD-NIRS acquisition time points up to T3 (t = 60 min) to ensure data quality and comparability. The analyzed segments, defined relative to the inhalation start time, were as follows: S1 (t = − 15 to 0 min, baseline), S2 (t = 0–30 min, during inhalation), S3 (t = 30–40 min, immediately after the end of inhalation), and S4 (t = 60–70 min, 60 min after inhalation) as shown in Fig. 1.

Within each segment, the device’s standard frequency-domain algorithm yielded low-frequency (LF; 0.04–0.15 Hz) and high-frequency (HF; 0.15–0.40 Hz) power. The values of the power less than zero were replaced with 0.1× (the smallest positive value in that segment), then log-transformed prior to averaging. Mean HR and RRI were calculated directly per segment, whereas log-LF, log-HF, and log (LF/HF) were averaged on the log scale. Mean blood pressures (MBP) were also measured at five time points of t = −15 (B1), 15 (B2), 30 (B3), 60 (B4) and 120 (B5) min as shown in Fig. 1.

Statistical analysis

All data preprocessing, asymmetry index calculation, and visualization were performed using custom scripts in Python 3.11. Statistical analyses were conducted using GraphPad Prism 8 and 9 (GraphPad Software, San Diego, CA, USA).

For time-course comparisons of cerebral oxygenation (oxy-Hb and deoxy-Hb), the Friedman test was used to detect main effects of time across the four TD-NIRS time points (T1–T4). When a significant main effect was found, post hoc pairwise comparisons were conducted using Dunn’s test with Bonferroni correction. Hemispheric asymmetry was quantified by calculating the AI at each time point, and the AI values were compared between the time points using the Wilcoxon signed-rank test. Baseline-referenced differences of the oxy-Hb and deoxy-Hb at T2, T3, T4 (Δoxy-Hb, Δdeoxy-Hb) were calculated relative to T1 for each hemisphere and left–right differences were evaluated using the Wilcoxon signed-rank test for paired samples. Additionally, for the time points where statistically significant hemispheric differences were observed, the magnitude of the effect was quantified using Cohen’s d for paired samples. To assess the robustness of the primary hemispheric asymmetry contrast, we performed prespecified sensitivity analyses focusing on ΔAI = AI(T2) − AI(T1): (1) leave-one-out (LOO) analyses repeating the Wilcoxon comparison and effect-size estimation after excluding each participant in turn, (2) a re-analysis excluding the two most influential participants defined as those whose exclusion produced the largest LOO p-values, (3) a nonparametric bootstrap procedure (B = 20,000 resamples) to compute a 95% confidence interval for the median ΔAI, and (4) a sign test to quantify directional consistency of ΔAI across participants.

To improve robustness, outlier values at each time point were excluded using a ± 3 SD criterion applied within each subject. For the cardiovascular and autonomic metrics (MBP, HR, RRI, LF, HF, LF/HF), phase effects (phases B1 to B5 and S1 to S4) were assessed using the Friedman test because data did not meet normality assumptions. Here, “phase” means “before, during and after the hydrogen inhalation”. When statistically significant, post hoc pairwise comparisons were performed using paired t-tests or Wilcoxon signed-rank tests with Bonferroni correction for multiple comparisons. Baseline-referenced changes (Δ values) were also analyzed to evaluate phase-specific effects. For comparisons yielding statistical significance in HR, RRI, and spectral indices, Cohen’s d for paired samples was calculated to estimate the magnitude of the effect. HR was used as the primary chronotropic descriptor, and RR interval was presented in parallel mainly to support HRV computation and data-quality transparency.

To explore potential age-dependent modulation, Spearman rank correlation analyses were performed to examine whether participant age was associated with the magnitude of changes in both NIRS-derived cerebral oxygenation and ECG-derived cardiovascular/autonomic parameters.

All tests were two-tailed and p-values < 0.05 were considered statistically significant. Adjusted p-values are reported where applicable.

Results

Time-course of cerebral oxygenation during hydrogen inhalation

To characterize the time-course of cortical oxygenation during the hydrogen inhalation session (99.9% at the source, 30 min), we first plotted time series of oxy-Hb and deoxy-Hb in the left and right PFC using violin plots over the four time points (T1–T4) as shown in Fig. 2A, where each data point represents a 3-minute average. Individual data points were color-coded by participant age (25–72 years), as indicated in the color bar. Although the Friedman test revealed a significant main effect of the time point on the left hemisphere oxy-Hb signal (χ²₍3₎ = 8.35, p = 0.039), post hoc Dunn’s pairwise comparisons after Bonferroni correction did not identify significant differences between specific time points. No significant main effects were observed for the right hemisphere oxy-Hb (χ²₍3₎ = 2.67, p = 0.445) or for the deoxy-Hb signal in either hemisphere (left: χ²₍3₎ = 4.42, p = 0.220; right: χ²₍3₎ = 5.84, p = 0.120). These data indicate substantial inter-individual variability in the magnitude and direction of oxy-Hb changes across time points.

Fig. 2.

Time-course and lateralization of prefrontal cortex hemodynamics during hydrogen inhalation. (a) Violin plots of the oxy-Hb (left panel) and deoxy-Hb (right panel) in the left (warm/red or cool/blue shades) and right (yellow or light-blue shades) PFC at the four time points (T1–T4) averaged over 3-min measurements. Data points are overlaid on the plots and color‐coded by age as in the color bar. Grey shading indicates the 30-min inhalation period. (b) Box‐and‐scatter plots of the AI for the oxy-Hb (left panel) and deoxy-Hb (right panel) across T1–T4. Lines connect paired observations; * p < 0.05 vs. T1 (Wilcoxon signed-rank test). (c) Violin plots of baseline-referenced changes (changes from T1), Δoxy-Hb (left panel) and Δdeoxy-Hb (right panel) at T2–T4. Data points (filled = left; open = right) are overlaid. * p < 0.05 for Δoxy-Hb at T2 (paired Wilcoxon); all other hemispheric comparisons n.s. (not statistically significant).

Hemispheric asymmetry shifts induced by hydrogen

Next, we quantified lateralization by calculating the AI for oxy-Hb and deoxy-Hb at each time point. As shown in Fig. 2B (left panel), the median AI values for oxy-Hb shifted from slightly negative at baseline (T1: − 0.031 [–0.063–0.001]) to positive during inhalation (T2: 0.003 [–0.015–0.041]), and returned closer to the baseline levels after inhalation (T3: − 0.016 [–0.058– −0.007]; T4: 0.011 [–0.048–0.046]). A Wilcoxon signed-rank test indicated that the AI at T2 was significantly higher than that at the baseline (T1 vs. T2: p = 0.014; d = 0.790), with corresponding mean ± SD values of − 0.031 ± 0.059 at T1 and 0.054 ± 0.138 at T2. Because AI is a relative left–right measure, this rightward shift can reflect a right-sided increase, a left-sided decrease, or a combination of both, rather than an absolute increase in the right hemisphere alone. In contrast, no significant differences were observed between T1 and T3 (p = 0.102) or between T1 and T4 (p = 0.054). To assess the robustness of this T2–T1 asymmetry shift against influential participants, we performed sensitivity analyses on the primary contrast ΔAI = AI(T2) − AI(T1). In the full sample (n = 11), the T2 vs. T1 shift remained statistically significant (p = 0.0137) with a large effect size (r = 0.724; d = 0.790). LOO analyses showed that the comparison remained significant after excluding any single participant (p range: 0.00195–0.02734), with consistently large effect sizes across iterations (r range: 0.693–0.886; d range: 0.694–0.973) (Supplementary Table 1). When excluding the two most influential participants (n = 9) defined as those yielding the largest LOO p-values, the direction of the effect was preserved (median ΔAI = 0.037 [0.022–0.079]), although statistical significance was attenuated (p = 0.055; r = 0.652; d = 0.775), indicating sensitivity to influential-case exclusion in this modest sample. To complement p-values, we additionally computed a nonparametric bootstrap 95% confidence interval for the median ΔAI, which excluded zero (95% CI: 0.022–0.164), and confirmed directional consistency (ΔAI > 0 in 10/11 participants; p = 0.012) (Supplementary Fig. 3). Supplementary Fig. 3 visualizes participant-level ΔAI alongside age (color) and BMI category (dot size) to facilitate transparent appraisal of inter-individual variability and potential demographic confounding.

For the deoxy-Hb asymmetry index (Fig. 2B, right panel), the median values remained positive at all the time points (T1: 0.060 [0.016–0.119]; T2: 0.055 [0.027–0.127]; T3: 0.049 [−0.008–0.077]; T4: 0.021 [0.006–0.118]). Wilcoxon signed-rank tests revealed no significant differences between T1 and T2 (p = 0.638), T1 and T4 (p = 0.898), and between T2 and T3 (p = 0.175). These results demonstrate that hydrogen inhalation is associated with a transient rightward shift in oxy-Hb hemispheric asymmetry, most evident during and immediately after the inhalation session, without a clear change in deoxy-Hb asymmetry.

Baseline-referenced changes and lateralization

To further evaluate the hemisphere-specific effects, we calculated the baseline-referenced changes, Δoxy-Hb and Δdeoxy-Hb, at the time points (T2–T4) as shown in Fig. 2C. A Wilcoxon signed-rank test was used for all comparisons between hemispheres. At T2, the right hemisphere exhibited a significantly greater Δoxy-Hb compared to the left hemisphere (Left: − 2.00 [–3.96––0.01] µM; Right: 1.27 [–0.85–3.29] µM; p = 0.017; d = − 0.858), with corresponding mean ± SD values of − 1.99 ± 6.55 µM for the left and 2.45 ± 5.82 µM for the right hemisphere. In contrast, no significant hemispheric differences were observed at T3 (Left: − 0.73 [–1.94–0.77] µM; Right: − 0.88 [–3.61–2.33] µM; p = 0.875) and T4 (Left: − 3.00 [–5.92––0.66] µM; Right: − 1.74 [–2.62–2.31] µM; p = 0.077) (Fig. 2C, left panel). Δdeoxy-Hb showed no significant hemispheric differences at any time points after inhalation. At T2 (Left: 0.68 [–0.25–2.44] µM; Right: 1.21 [0.51–2.65] µM; p = 0.396), T3 (Left: 2.05 [0.76–2.59] µM; Right: 1.45 [–1.26–2.04] µM; p = 0.256), and T4 (Left: 1.03 [–0.42–1.93] µM; Right: 0.77 [–0.07–1.80] µM; p = 0.865), the median Δdeoxy-Hb values remained comparable between hemispheres (Fig. 2C, right panel). Because the comparisons are made between hemispheres at each time point, the results primarily inform hemispheric differences in baseline-referenced changes. These findings indicate a transient between-hemisphere difference in baseline-referenced oxy-Hb changes at T2, with a greater Δoxy-Hb in the right than in the left PFC.

In addition, exploratory Spearman correlation analyses did not reveal consistent monotonic relationships between the participant age and the magnitude of these baseline-referenced changes in either the oxy-Hb or deoxy-Hb (Supplementary Fig. 3).

Dynamic changes in cardiovascular and autonomic parameters

Finally, we examined systemic autonomic effects by analyzing the ECG-derived metrics before hydrogen inhalation (S1), during the entire 30-minute inhalation period (S2), immediately after the end of inhalation (S3), and late after inhalation (S4). The MBPs were also examined from B1 to B5. Continuous time-series analysis revealed that MBPs remained stable throughout the measurement period from B1 to B5 (Fig. 3, top). In contrast, a gradual reduction in HR and a corresponding increase in RRI were observed during and after hydrogen inhalation (Fig. 3, second and third panels from the top). Frequency-domain indices (LF and HF power) fluctuated within typical ranges without sustained elevation or suppression, while the LF/HF ratio exhibited transient peaks during inhalation (Fig. 3, bottom two panels).

Fig. 3.

Continuous cardiovascular and autonomic dynamics before, during, and after hydrogen inhalation. Time-series of mean blood pressure (MBP, top), heart rate (HR, second), R–R interval (RRI, third), and frequency‐domain components (LF and HF power, fourth; LF/HF ratio, bottom) from 15 min before inhalation (t = −15 min) to the end of measurement (t = 110 min). Colored ribbons denote ± 1 SD about the group mean traces (solid lines). Shaded blue bar indicates the 30-min inhalation period.

Mean MBP did not change significantly across the phases (B1 vs. B2: p = 0.108; B1 vs. B3: p = 0.118; B1 vs. B4: p = 0.288). In contrast, HR significantly decreased relative to the baseline (S1 vs. S2: p = 0.009, d = − 0.635; S1 vs. S3: p = 0.037; S1 vs. S4: p = 0.015, d = −0.746), with a corresponding increase in RRI compared to the baseline (S1 vs. S2: p = 0.034; S1 vs. S3: p = 0.029; S1 vs. S4: p = 0.008, d = 0.831). These trends were consistently observed in both the absolute values (Fig. 4) and the baseline-referenced differences shown in Fig. 5. In particular, ΔHR and ΔRRI demonstrated significant changes from the baseline at 30 min after inhalation (ΔHR: p = 0.014, d = −0.746; ΔRRI: p = 0.007, d = 0.831), while earlier time points did not reach significance after Bonferroni correction. For the frequency-domain indices, the log-transformed LF power showed a modest reduction during inhalation (S1 vs. S2: p = 0.011, d = 0.669), whereas HF power and the LF/HF ratio did not demonstrate significant changes. Similarly, the baseline-referenced differences in ΔLF and ΔHF were not significant, while the ΔLF/HF ratio transiently increased during inhalation (S1 vs. S2: p = 0.023 unadjusted, p = 0.070 adjusted, d = 0.745) but returned to the baseline afterward (Fig. 3). Although MBP did not show significant group-level changes, Spearman correlation analyses indicated that age was negatively correlated with MBP changes during inhalation and at the end of inhalation (ΔMBP (B2-B1): ρ = − 0.674, p = 0.023; ΔMBP (B3-B1): ρ = − 0.665, p = 0.025), suggesting that older participants tended to show smaller MBP reductions. In contrast, age was not significantly associated with HR, RRI, LF, or HF changes at any phase (all phases: p > 0.3). The LF/HF ratio showed a moderate positive correlation with age during inhalation (ρ = 0.512, p = 0.074), although this did not reach statistical significance. Because baseline HR varied across participants, we repeated post hoc comparisons after excluding the participant with the highest baseline HR in the Pre window (mean HR 88.8 bpm). The direction of HR change and effect sizes were similar to the full-sample analysis, although the Bonferroni-adjusted p-values exceeded 0.05 (Pre vs. During, adjusted p = 0.051; Pre vs. 30 min, adjusted p = 0.098).

Fig. 4.

Phase-wise bar plots of absolute and baseline‐referenced MBP and ECG‐derived metrics. Bar + scatter plots (mean ± SEM bars; individual circles) for MBP, HR, RRI, LF and HF power, and LF/HF ratio at the four phases: before inhalation (B1 and S1), during inhalation (B2 and S2), immediately after the end of inhalation (B3 and S3), late after inhalation (B4 and S4). Asterisks denote significant differences from the baselines measured before inhalation (paired t-test or Wilcoxon signed-rank, Bonferroni-corrected): * p < 0.05.

Fig. 5.

Baseline-referenced changes (Δ) and age correlations in MBP, cardiovascular and autonomic parameters. Violin plots of the baseline-referenced changes (Δ values) for MBP, HR, RRI, log-LF and log-HF powers, and Δ LF/HF ratio during inhalation (B2 and S2), immediately after the end of inhalation (B3 and S3), late after inhalation (B4 and S4). Individual data points are color‐coded by age (grey scale). Symbols above each violin indicate significant phase effects (# p < 0.05 unadjusted; * p < 0.05 adjusted).

Taken together, oxy-Hb hemispheric asymmetry showed a transient rightward shift at T2, while deoxy-Hb asymmetry did not show a clear change. Autonomic indices showed time-dependent changes, including a transient increase in LF/HF during inhalation. For the TD-NIRS outcomes, four participants were excluded based on predefined data quality control criteria, leaving eleven participants for the TD-NIRS analyses. ECG-derived analyses were conducted in participants with analyzable ECG recordings across all prespecified time windows, and the sample size therefore differs from the TD-NIRS analyses. No adverse events were observed during the study.

Discussion

In this study, we have characterized for the first time in healthy humans that a single 30-minute session of hydrogen inhalation was associated with a transient shift in prefrontal oxy-Hb hemispheric asymmetry without a concomitant change in the deoxy-Hb, accompanied by modest shifts in autonomic tone. Although the overall oxy-Hb and deoxy-Hb did not change significantly across the four measurement time points (T1–T4), the oxy-Hb asymmetry index showed a transient increase from T1 to T2, returning toward the baseline at T3 (Fig. 2B). The between-hemisphere difference in baseline-referenced oxy-Hb change at T2 (Δoxy-Hb, right vs. left) (Fig. 2C) is compatible with a lateralized alteration in oxygenation, while deoxy-Hb remained stable. To our knowledge, this is the first report of a transient rightward shift in prefrontal oxy-Hb hemispheric asymmetry during a hydrogen inhalation session in healthy adults.

One plausible functional interpretation is that hydrogen inhalation transiently enhances a rightward shift in prefrontal oxygenation lateralization. Neuroimaging studies in younger adults have repeatedly shown that tasks requiring sustained attention, working memory, or novel information processing preferentially activate right-hemispheric prefrontal networks, which correlates with faster reaction times and improved accuracy^16,17. By contrast, aging is associated with a reduction in task‐related hemispheric asymmetry, the HAROLD phenomenon, where older adults recruit bilateral prefrontal regions to compensate for declining neural efficiency^18,19. In our cohort of middle‐aged to older individuals, the brief shift toward a more right-lateralized asymmetry pattern might reflect a temporary re‐engagement of youthful lateralization, although we did not assess cognitive performance or arousal directly. Given the marked inter-individual variability, we further explored whether the magnitude of the asymmetry shift was associated with age or BMI; however, no clear relationship was observed in this modest sample (Supplementary Fig. 3). Larger, sham-controlled studies will be required to determine whether demographic or anthropometric factors systematically modulate the strength of this response.

Alternatively, and perhaps concurrently, these hemodynamic shifts may reflect autonomic-vascular coupling. Our ECG analyses showed that heart rate slightly but significantly decreased during inhalation, while the LF/HF ratio transiently increased. This pattern is indicative of a temporary shift toward sympathetic predominance (Fig. 3). Prior work has linked right PFC activation to sympathetic arousal and anxiety states¹⁴, suggesting that the act of inhaling a novel gas stimulus could elicit mild anticipatory tension, particularly during the early inhalation phase. Indeed, single‐session hydrogen interventions have been shown both to acutely heighten alertness, possibly due to transient sympathetic activation. With longer or repeated exposures, to reduce baseline sympathetic tone, improve heart‐rate variability, and lower stress markers^12,20. Our findings of a transient asymmetry shift coupled with sympathetic upticks may thus reflect an interplay between cortical arousal and vascular regulation rather than purely metabolic demand.

Mechanistically, these vascular changes may be mediated by enhanced nitric oxide (NO) bioavailability. Molecular hydrogen selectively scavenges hydroxyl radicals (•OH), which otherwise degrade endothelial NO, a potent vasodilator¹. By neutralizing ·OH, hydrogen could preserve NO levels, promoting cerebral vasodilation and increase oxy-Hb without altering deoxy-Hb since oxygen consumption remains unchanged²¹. This NO-mediated pathway offers a biochemical basis for our TD-NIRS observations. The observed rightward shift in hemispheric asymmetry is compatible with the possibility that hydrogen may modulate resting neural networks with right-lateralized hubs, such as the salience network²² or components of the default-mode network (DMN)²³, both implicated in internal-external monitoring and autonomic regulation. Alternatively, right PFC itself serves as a critical node within the central–autonomic network (CAN), coordinating cortical and subcortical autonomic centers^24,25. The act of hydrogen inhalation may thus evoke an orienting response in right-prefrontal CAN nodes, translating cortical activation into downstream sympathetic modulation. Ultimately, our findings are best interpreted within the CAN framework, wherein cortical and autonomic phenomena constitute a unified response. The co-occurrence of the asymmetry shift and the transient sympathetic predominance (LF/HF rise) likely reflect two facets of hydrogen’s modulation of brain-body coupling. In this view, hydrogen inhalation acts as a subtle modulator of the CAN, initiating in the right PFC and cascading through autonomic effector pathways. This integrative perspective not only elucidates the acute physiological actions of molecular hydrogen but also broadens its therapeutic potential as a regulator of central–autonomic interactions.

Several limitations warrant consideration. First, the sample size was modest, and the study lacked a placebo (sham gas) control group, limiting our ability to draw firm causal inferences about treatment effects. In particular, both the acute changes observed during or immediately after inhalation and the delayed changes observed up to 120 min thereafter could reflect nonspecific time/position effects and time-on-task factors, as well as normal within-day physiological variation, rather than hydrogen-specific effects. Importantly, this investigation was intended to characterize short-term physiological responses to hydrogen inhalation rather than to assess its therapeutic efficacy. Consequently, future work employing a randomized, placebo-controlled design will be essential to confirm that the observed cerebral and autonomic responses are specifically attributable to hydrogen. Second, because the primary asymmetry contrast was modest and inference was sensitive to influential-case exclusion, we complemented the main analysis with prespecified sensitivity analyses (LOO, bootstrap confidence interval, and directional consistency), and therefore interpret the asymmetry finding as exploratory. Third, hydrogen was delivered via nasal cannula at a fixed source flow rate, and neither respiratory parameters nor the inspired gas composition were measured; therefore, the inspired hydrogen fraction/partial pressure could not be quantified on a per-participant basis. This limits dose–response interpretation and may contribute to inter-individual variability in physiological responses. Future studies should quantify exposure by measuring inspired hydrogen concentration or estimating it from ventilation and, ideally, standardize delivery using a controlled inspiratory setup. In addition, peripheral oxygen saturation and inspired oxygen concentration were not monitored, and brief subclinical desaturation cannot be excluded. Fourth, we restricted measurements to the PFC and did not acquire direct measures of cognitive performance or subjective arousal. Future research incorporating whole-brain imaging, task-based paradigms, and psychometric assessments will be crucial to clarify the functional significance of this right-lateralized response and its network-level impact. Fifth, although TD-NIRS offers distinct methodological advantages over conventional continuous-wave NIRS (CW-NIRS), including reduced susceptibility to probe-related artifacts, it also has inherent limitations. In TD-NIRS, the absorption and reduced scattering coefficients (i.e., optical properties) are estimated from the DTOF under the assumption that the head is optically homogeneous. However, in reality, the scalp, skull, and brain exhibit distinct optical properties. Consequently, the calculated concentrations of oxy-Hb and deoxy-Hb likely reflect a mixture of signals originating from both intracerebral and extracerebral compartments. While TD-NIRS is relatively robust against signal contamination caused by probe removal or reattachment, which is particularly advantageous in repeated-measurement designs like ours, this method does not enable definitive separation of cortical and superficial hemodynamic changes. Nevertheless, extracerebral hemodynamics primarily reflect systemic circulation and are unlikely to exhibit strong lateralization across the forehead. In contrast, cerebral responses are often regionally specific and may differ between hemispheres. Therefore, the observed hemispheric asymmetry in oxygenation is more likely attributed to hemispheric differences in cerebral hemodynamics than to scalp-derived effects. In addition, with further methodological improvements in extracting the absorption coefficients from the DTOF, the separation of the extracerebral and intracerebral absorption coefficients may become feasible²⁶, and this will be addressed in future work.

Conclusion

This integrative NIRS–ECG study provides the first detailed evidence that acute hydrogen inhalation transiently modulates PFC oxygenation asymmetry and autonomic balance in humans. The observed transient rightward shift in oxy-Hb hemispheric asymmetry, accompanied by sympathetic activation, suggests a coordinated neurovascular–autonomic response to hydrogen inhalation. These findings establish a physiological basis for future investigations exploring repeated or prolonged hydrogen interventions, with potential relevance to cognitive enhancement, mitigation of age-related asymmetry decline, and modulation of cardiovascular risk via neurovascular pathways.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

MM, YD, and KS conceptualised and designed this work. MM acquired the data. MM, KO, and YY analysed the data. MM and KS interpreted the data. MM and YY drafted the manuscript. All authors contributed to critical revision of the manuscript for important intellectual content. All authors approved the final version of the manuscript and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

This work was partly supported by the JSPS-KAKENHI grant number 23K25233 (K.O.) and by CX Wellness Inc.

Data availability

The datasets generated and/or analysed in this study are available from the corresponding author on reasonable request.

Declarations

Competing interests

Y.D. is the representative director of CNB Medical Research Institute, which manufactures and distributes hydrogen inhalation devices. This study received support from CX Wellness Inc. The remaining authors declare no competing interests.