Mild to Moderate Sleep Apnea Is Linked to Hypoxia-induced Motor Recovery after Spinal Cord Injury

To the Editor:

Acute intermittent hypoxia (AIH) is a novel treatment to enhance respiratory and nonrespiratory motor function after chronic incomplete spinal cord injury (SCI) (1–3). Despite promising findings, uncertainties in optimal AIH delivery to a heterogeneous population of persons with SCI remain a challenge to clinical translation. Beneficial AIH responses are variable between individuals, and we lack useful biomarkers to determine which individuals benefit most from treatment. AIH-induced functional benefits rely on mechanisms of AIH-induced neuroplasticity. Multiple factors modify AIH-induced plasticity in rodent models, including intermittent hypoxia preconditioning and systemic inflammation, among others (4, 5). One factor not accounted for in human AIH trials, to date, is the incidence of sleep-disordered breathing (SDB). Many individuals with SCI exhibit mild to moderate SDB (6), leading to extended periods of nocturnal high-dose intermittent hypoxia. Because chronic intermittent hypoxia (CIH) elicits both neuroplasticity and inflammation (5), SDB after SCI may contribute to between-person variations in response to AIH therapy.

Here, we report associations between SDB and AIH-induced motor recovery in individuals with chronic incomplete SCI. Specifically, we conducted a blinded continuing analysis in 20 participants who performed baseline sleep assessments and one of three clinical trials with similar experimental designs (1–3), including treatment methods, randomization (AIH/sham treatment), blindness, and washout period. Participants who missed an evaluation session (n = 5), were enrolled in more than one of these studies (n = 5) (in which case, we evaluated only the first study completed), or had <4 hours of sleep data duration (n = 8) were not included. Participants performed five consecutive (daily) AIH breathing treatment sessions (or a single-day session; n = 3). A manually controlled air delivery system (Hypoxico, Inc.) was used with 15 sequences of 60–90 seconds of low-oxygen breathing (AIH; Fi_O2 = 0.10 ± 0.02) or room air (sham; Fi_O2 = 0.21 ± 0.02) alternated with 60-second intervals of breathing room air. We compiled the apnea–hypopnea index, lowest O₂ pulse saturation (average pulse oxygen saturation [Sp_O2] nadir), inspiratory flow limitations, snoring counts, and pulse frequency (ApneaLink; Resmed Corp.) and quantified associations between these metrics and AIH-induced improvements in motor function. We computed effect sizes to standardize analyses across trials with different functional outcome measures (limb strength, walking ability, or hand function). Within-subject Cohen’s d scores corresponded with a participant’s effect size after treatment and were calculated as follows: Cohen’s d = [average change from baseline with AIH] − [average changes from baseline with sham]/baseline pooled SD (7).

Mild to moderate sleep apnea (5 < apnea–hypopnea index <30/h) occurred in 50% of study participants. The participants with mild to moderate apnea (n = 10) had the same characteristics as those without apnea (n = 10, Table 1). Participants with mild to moderate SDB demonstrated greater motor gains with AIH therapy (Cohen’s d median: 3.6; range: −0.5 to 13.6) versus those without apnea (1.7; range: −6.6 to 4.9; P = 0.043) (Figure 1). Cohen’s d positively correlated with the oxygen desaturation index (ODI-4%) (r = 0.74; P < 0.0001) and negatively correlated with the average Sp_O2 nadir (r = −0.69; P = 0.001). Moreover, the time spent during sleep with Sp_O2 below 90% also correlated to Cohen’s d (r = 0.51; P = 0.024); in a multiple regression analysis (r = 0.73; P = 0.002), ODI-4% was an independent predictor of Cohen’s d (P = 0.007), whereas time spent during sleep with Sp_O2 below 90% was not (P = 0.56), suggesting that the AIH response mostly related to repetitive nocturnal O₂ desaturations. Conversely, pulse frequency (sympathetic activity) or snoring events were not related to AIH response. Finally, Cohen’s d did not correlate with Sp_O2 during AIH therapy (first and fifth sessions) (r < 0.30; P > 0.25).

Table 1.

Patient Characteristics, Sleep Parameters, and Functional Gain after AIH Therapy (Cohen’s d)

	No Apnea (n = 10)	Mild to Moderate Apnea (n = 10)	P Value*	Mild Apnea 5 ≤ AHI < 15 (n = 7)	Moderate Apnea 15 < AHI < 30 (n = 3)
AHI, no./h†	1.5 (0–4)	9.0 (6–29)	<0.001	6.5 (6–10)	23 (17–29)
Cohen’s d	0.2 ± 4.0	4.2 ± 3.8	0.043	2.4 ± 2.3	7.8 ± 4.6
Age, yr	46 ± 14	45 ± 8	0.90	42 ± 15	53 ± 14
Level of injury†	C5 (C2–T4)	C5 (C4–T8)	0.26	C5 (C5–T8)	C5 (C4–T3)
Time since injury, yr†	7 (3–38)	7 (3–25)	0.64	7.5 (6–18)	11 (7–24)
BMI, kg/m2†	20 (19–30)	24 (20–33)	0.23	24 (20–25)	28 (24–33)
Sleep parameters
Evaluation duration, h/night	6 ± 2	7 ± 2	0.42	6 ± 2	8 ± 1
Apnea index, no./h†	0 (0–2.5)	2.5 (0.5–27)	0.001	1.5 (0.5–4.3)	14 (0.5–27)
Hypopnea index, no./h	1.1 ± 0.9	5.6 ± 2.1	<0.001	5.1 ± 1.0	5.6 ± 3.7
Oxygen desaturation index, no./h†	1 (0–6)	10 (4–34)	<0.001	8.5 (4–13)	23 (4–34)
Average SpO2, %	93 ± 4	91 ± 4	0.048	93 ± 2	89 ± 7
Average SpO2 nadir, %	91 ± 5	84 ± 6	0.005	87 ± 2	76 ± 3
Time with SpO2 ≤ 90%, min†	2 (0–36)	67 (1–137)	0.002	48 (1–110)	85 (1–137)
Average pulse frequency, beats/min	65 ± 9	65 ± 10	0.95	65 ± 12	65 ± 6
Highest pulse frequency, beats/min	130 ± 41	114 ± 32	0.77	102 ± 29	140 ± 42
Snoring events, no./h	578 ± 673	653 ± 760	0.47	353 ± 411	1,353 ± 1,023

Definition of abbreviations: AHI = apnea–hypopnea index; AIH = acute intermittent hypoxia; BMI = body mass index; Sp_O2 = pulse oxygen saturation.

Data are mean ± SD except where otherwise noted.

^*Comparison between “no apnea” and “apnea” (mild to moderate apnea) (t test or Mann-Whitney rank test comparisons).

^†Median (range) for nonparametric variables.

Figure 1.

Effect of sleep-disordered breathing (as assessed by the apnea–hypopnea index [AHI]) on the response to acute intermittent hypoxia (AIH) therapy. Shown are individual values and box plot (median, 10th, 25th, 75th, and 90th percentiles with error bar) of Cohen’s d in individuals with incomplete spinal cord injury and sleep apnea (mild to moderate apnea 5 < AHI < 30/h) or no apnea (AHI < 5/h).

Our results demonstrated that mild to moderate SDB is associated with more significant AIH-induced therapeutic benefits in people with chronic incomplete SCI. SDB (and associated CIH) may precondition AIH therapy, enabling larger responses to subsequent AIH treatments as reflected in improved AIH-induced performance gains. The dose–response between nocturnal O₂ desaturation and AIH-induced motor gains supports the hypothesis that a minimum hypoxia severity is necessary to induce motor adaptations (4). Based on available literature in rodent models, this effect may arise via CIH-preconditioning effects on carotid body chemoreceptors, enabling long-term sensory facilitation in carotid sinus nerve activity and amplification of chemoafferent integration within the central nervous system (8). However, given the lack of correlation between Sp_O2 during AIH therapy (an inverse indicator of respiratory chemoreflex) and AIH-induced motor gains, it seems unlikely that chemosensory plasticity would explain our results. Alternatively, spinal mechanisms may enhance the capacity for AIH-induced motor plasticity (4). For example, repetitive serotonin receptor activation may increase BDNF (brain-derived neurotrophic factor) protein synthesis and BDNF/TrkB (tropomyosin receptor kinase B) signaling, causing the functional synaptic plasticity that underlies beneficial effects of therapeutic AIH (9). In humans, reminder doses of three times/week maintain increases in walking speed and endurance for up to 5 weeks after 1 week of daily AIH (10). The independent effect of the repetitive occurrence of sleep apnea (ODI) on functional gain would support this hypothesis. Hence, plasticity in somatic motor output could result from cumulative repeated CIH/AIH.

Our results also suggest that SDB may serve as a “biomarker” that indicates the persons most likely to benefit from AIH therapy. It would be interesting to know if SDB itself (via mild to moderate CIH) is directly associated with better functional outcomes during normal rehabilitative therapeutics. However, because severity of hypoxia (dosage) and repetitive intermittent hypoxia are key components of metaplasticity, mild to moderate amounts of CIH may more likely precondition AIH therapy (4).

Participants accepted into this study did not have severe SDB; more severe SDB may lead to deleterious proinflammatory responses that inhibit AIH therapy–induced gains (5). Given the potential benefits of mild or modest CIH versus the known pathological impact of obstructive and central SDB on health, this study does not guide whether individuals with SCI need SDB treatment. Understanding the history of hypoxic exposures merits further consideration when evaluating therapeutic AIH dosing and efficiency.

Overall, our study demonstrates that AIH therapy–induced functional gains are more robust in people with chronic incomplete SCI and mild to moderate SDB. Repetitive nocturnal desaturation may precondition spared neural circuitry involved in functional tasks, enabling them to respond to subsequent AIH therapy more positively. If confirmed, these findings strongly suggest a need to refine AIH delivery protocols and that assessment of SDB provides valuable insight into individual responsiveness to AIH therapy. Further larger and dedicated studies are needed to confirm our findings and pursue the mechanisms by which mild to moderate SDB may precondition AIH therapy and favor motor recovery in SCI.