Skip to main content
NCBI home page
Sleep logoLink to Sleep
. 2025 Oct 4;49(4):zsaf310. doi: 10.1093/sleep/zsaf310

Respiratory-related leg movement as a predictor of mortality: the Osteoporotic Fractures in Men Sleep Study

Neda Esmaeili 1, Sayaka Aritake 2, Susan Redline 3, John W Winkelman 4,5,2, Ali Azarbarzin 6,2,✉
PMCID: PMC13089395  PMID: 41045147

Abstract

Study Objectives

Respiratory event-related leg movements (RRLMs) in obstructive sleep apnea may reflect event severity and predict adverse outcomes. This study examines the association between RRLM frequency and mortality.

Methods

A sample of 571 men (median [IQR] age: 76.0 [72.0–81.0] years; BMI: 28.0 [25.0–30.0] kg/m2) with an apnea–hypopnea index (AHI, ≥3 per cent desaturation) ≥10 events/h was selected based on available incident data and annotations of RRLMs, from the Osteoporotic Fractures in Men Studies . RRLMs were scored if they occurred within ±5 s of respiratory event termination lasting 0.5–10 s; the average number of RRLMs per hour of sleep respiratory-related leg movement index (RRLMI) was measured. For comparison, the number of periodic leg movements of sleep (PLMS) per hour of sleep (PLMI) was also quantified. Cox regression models estimated the adjusted hazard ratios (aHR) for mortality per SD increase in RRLMI and PLMI. Secondary analyses assessed the relationship between RRLM duration (i.e. a measure of RRLM strength) and measures of event severity (i.e. desaturation area, arousal, heart rate response, and event duration).

Results

There were 287 all-cause deaths during 11.4 [6.9;12.5] years of follow-up. RRLMI and PLMI were 6.0 [3.4;9.3] and 22.0 [2.0;54.0] events/h, respectively. In an adjusted model, every 1SD increase in RRLMI was associated with all-cause mortality (aHR: 1.17 [95% CI: 1.01–1.34], p-value = .03). In contrast, PLMI was not associated with mortality (aHR: 1.04 [95% CI: 0.92–1.17], p-value = .5). Furthermore, RRLM duration was linked to measures of event severity.

Conclusions

In a sample of older men with obstructive sleep apnea, RRLMI was independently associated with all-cause mortality, whereas PLMI was not, though confirmatory studies are needed.

Keywords: respiratory-related leg movement, sleep apnea, all-cause mortality, cardiovascular

Statement of Significance Respiratory-related leg movements (RRLMs) are common in obstructive sleep apnea and may reflect the severity of respiratory events. Previous research suggests that RRLMs may be associated with adverse health outcomes, but their long-term associations with mortality and cardiovascular diseases remain unclear. This study demonstrates that RRLM frequency is independently associated with increased all-cause mortality in older men with obstructive sleep apnea. Additionally, RRLM duration is associated with markers of respiratory event severity.

Graphical Abstract

Graphical Abstract.

Graphical Abstract

Open in a new tab

Introduction

Obstructive sleep apnea (OSA) is associated with various adverse outcomes, such as hypertension (HTN), cardiovascular diseases (CVDs), and mortality [1–5]. OSA is characterized by frequent complete or partial obstructions of upper airway during sleep, associated with oxygen desaturation, arousal, and/or increase in heart rate and blood pressure [6], which are thought to be key physiological consequences of airway obstruction, leading to adverse health outcomes [7]. In addition, termination of respiratory events is often accompanied by leg movements (i.e. respiratory-related leg movement, RRLM) [6]. However, RRLMs are rarely quantified on a sleep study and their association with outcomes remains poorly understood.

Periodic leg movements of sleep (PLMS) are characterized by stereotyped, repetitive, and involuntary leg movements during sleep outside of respiratory events. PLMS often occur in association with restless legs syndrome (RLS) and are considered the sleep-related motor sign of that disorder, though they appear with high frequency in other contexts as well. PLMS have been reported to be associated with CVD and mortality, although evidence is derived from selected or comorbid populations [8–11]. RRLMs have been also suggested as an independent marker of increased CVD risk in OSA [6, 12, 13]. The biological plausibility for this relationship comes from studies showing that there are greater heart rate changes at the termination of respiratory events when accompanied by RRLMs, indicating that RRLMs may indicate events with increased autonomic nervous system activation- a pathway that could lead to increase CVD risk [6, 12]. In addition, previous studies have shown that RRLMs were more likely to occur at the end of obstructive apneas than hypopneas [14, 15], in the presence of event-related arousals, and in participants with high-PLMS [15]. These findings suggest that RRLM may be an easily identified surrogate of more severe respiratory events. Beyond the simple presence of RRLM, the relationships between the duration of these leg movements (as an RRLM severity marker) and quantitative measures of respiratory event severity, including event duration, severity of oxygen desaturation (i.e. hypoxic burden [HB] per event [16, 17]), and arousal, heart rate response to events (ΔHR) [18, 19] have not been well characterized.

Our first objective is to test the hypothesis that increased number of RRLMs per hour of sleep respiratory-related leg movement index (RRLMI) is associated with adverse outcomes in OSA, independent of the overall apnea–hypopnea index (AHI). Additionally, we examine the number of PLMS per hour of sleep (PLMI) as an alternative predictor using the same statistical models. Second, we hypothesize that respiratory events ending with a leg movement are more severe, and that longer leg movement durations indicate greater severity. Specifically, the duration of RRLMs may correlate with the severity of preceding respiratory events, including their duration, the extent of oxygen desaturation, the presence of arousals, and heart rate surges. To test these hypotheses, manually scored RRLMs from the sleep studies in a subset of the Osteoporotic Fractures in Men Study (MrOS) cohort were used.

Materials and Methods

Participants

Between 2000 and 2002, the MrOS baseline examination enrolled 5994 men, aged 65 or older, from six US clinical centers. The MrOS Sleep Study, an ancillary study of the parent MrOS cohort, conducted a comprehensive sleep assessment for 3135 participants between December 2003 and March 2005, of which 2911 underwent in-home single-night polysomnography (PSG). Ethical approval was obtained from local institutional review boards, and all participants provided informed consent [20, 21]. Of these, 1634 had an AHI ≥ 10 (≥3 per cent desaturation criteria), acceptable leg movement recordings, achieved with bilateral anterior tibialis piezoelectric sensors, and had available CVD outcome data [13]. A total of 600 participants (300 with incident CVD and 300 without incident CVD over 6 years of follow-up) were randomly selected to investigate the predictive value of RRLM for CVD, and their RRLMs were visually annotated [13]. Eventually, 575 participants had sufficient PSG data and were included in the final dataset of the original study [13]. In this study, we used the same dataset which provided manually annotated RRLMs.

Polysomnography

Sleep recordings were performed using a Type II unattended overnight PSG using the Compumedics (Abbotsford, Australia) Safiro sleep monitoring system. Trained staff members visited the participants to attach the sensors and electrodes. The recording montage included two central electroencephalograms (EEG), bilateral electrooculograms, bilateral chin electromyogram, a bipolar electrocardiogram, nasal flow via pressure transducer and nasal cannula and nasal-oral thermistor, abdominal and chest respiratory inductance plethysmography, finger pulse oximetry, bilateral leg movement detection by piezoelectric sensors, and body position [22, 23]. Certified sleep technicians scored sleep and respiratory events according to the standard criteria [24]. A decrease in flow amplitude of more than 90 per cent for a minimum of 10 s was scored as apnea. A reduction of at least 30–90 per cent in nasal pressure or inductance bands lasting at least 10 s was considered hypopnea. Similar to the original study, AHI was defined as total number of apneas plus hypopneas associated with ≥3 per cent desaturation [13] per hour of sleep.

Leg movement scoring

The duration and latency criteria for distinguishing RRLM from PLMS have changed over time, supported by several research studies [12, 15, 25], which have shown the guidelines of the International RLS Study Group and the World Association of Sleep Medicine (IRLSSG/WASM) 2006 or the American Academy of Sleep Medicine (AASM) underestimate the number of RRLMs. The WASM 2016 standards have since recommended longer latency (2 s before to 10.25 s after the termination of a respiratory event) [26] compared to the WASM 2006 guidelines, which suggest a latency of 0.5 s before to 0.5 s after the end of the respiratory event [27]. After incorporating these guidelines, and consistent with our previous report that used the same data [13], RRLMs were defined if they occurred within ±5 s of the respiratory event termination and lasted for 0.5–10 s. The RRLMI was calculated as the number of RRLMs per hour of sleep.

Additional PSG parameters

The extent of total respiratory event-related hypoxemia over the sleep period was quantified by the sleep apnea-related HB, defined as the total area under the respiratory event-related desaturation curve during a subject-specific search window [16, 17]. Individual event-related oxygen desaturation was defined by HB per event (i.e. individual desaturation area under the respiratory event-related desaturation curve; HBEV [28]). Both HB and HBEV were log-transformed consistent with previous studies [16, 17]. Consistent with previous studies [18, 19], the heart rate response to apneas and hypopneas (ΔHR) was defined as the difference between a maximum pulse rate during a subject-specific search window and the minimum pulse rate during apneas and hypopneas [19]. Arousals were scored according to the AASM criteria [29] and their duration was quantified.

Comorbidities and outcomes

All participants completed a questionnaire at the time of the sleep visit, which included questions about medical history [30]. Prevalent CVD was defined as any history of physician-diagnosed coronary heart disease, cerebrovascular disease, and peripheral arterial disease reported during sleep studies. In this study, these subjects were excluded for incident CVD outcome. HTN and diabetes mellitus (DM) were defined based on self-report of physician-diagnosed condition. After the baseline visit, MrOS sleep participants underwent regular surveillance for potential incident cardiovascular or clinically relevant arrhythmia events by phone contact and/or postcard every 4 months with >99 per cent response rate. For fatal events, the death certificate and hospital records from the time of death were collected. For fatal events that occurred when participants were not hospitalized, a proxy interview with next of kin and hospital records from the most recent hospitalization in the prior 12 months were obtained. Only events confirmed by the adjudicator were included for analysis [16, 30].

Statistical analyses

The demographic and baseline characteristics of participants with low (< 6.0 RRLMs/h) or high RRLMI (≥ 6.0 RRLMs/h) were compared with 6.0 RRLMs/h serving as the median threshold. We used the Kaplan–Meier survival curves to assess the probability of all-cause mortality for low or high RRLMI.

To examine the association of RRLMI with all-cause death, CVD-related death, and incident CVD, five different Cox proportional hazards regression models with different levels of adjustment were built to estimate the adjusted hazard ratios (aHR) for outcomes. The proportional hazards assumption was tested using Schoenfeld residuals test. Cofounders and covariates were defined based on their associations in prior studies [12, 31, 32]. Absence of multicollinearity was confirmed by a variance inflation factor of <5. General characteristics of age, race, body mass index (BMI), alcohol consumption, and smoking status were included in Model 1. Physical activity rate, defined between 1 and 4 by subjects through questionnaire, antidepressant, and benzodiazepine use as well as prevalent HTN and DM were added for Model 2. In fully-adjusted Model 3, AHI and prevalent CVD (i.e. previous cerebrovascular, peripheral arterial, and coronary heart diseases) were added to covariates of Model 2 (for incident CVD only AHI was added). AHI was substituted with HB in exploratory Model 4, or with the ΔHR in Model 5 to explore which may serve as a better or alternative predictor of the outcomes compared to AHI. Given the observed sample size and event proportion, we had ~86 per cent power to detect an aHR of 1.20 per 1SD increase in RRLMI for all-cause death and ~70 per cent power to detect an aHR of 1.27 per 1SD increase in RRLMI (two-sided α = 0.05) for CVD death. Secondary analyses were done (1) with the same models for PLMI as the number of PLMS per hour of sleep, instead of RRLMI in the whole sample and subgroup of participants with RRLMI ≥ 6 and RRLMI < 6 events/h (PLMS were defined according to AASM 2006 PLMS criteria). (2) We quantified the extent to which the duration of RRLMs was explained by the duration of respiratory events, HBEV, ΔHR, and presence of arousal related to respiratory events, by using mixed-effect linear regression models with a random effect for each participant. To statistically quantify the goodness-of-fit of fixed and random effects, we measured the marginal and conditional R2.

Furthermore, consistent with prior studies that have applied thresholds such as ≥5 events/h to define clinically relevant subgroups, sensitivity analyses were performed with the same regression models in those with RRLMI ≥ 5 and ≥10 events/h. We also constructed an alternative leg movement index (LMI), including PLMS and all other scored leg movements regardless of their temporal association with respiratory events to compare the predictive value of the RRLMI with the overall LMI. Significance level was set at p < .05 and all statistical analyses were conducted using the R statistical packages (e.g. survival, lme4) (http://www.r-project.org).

Results

The summary of the demographic and baseline polysomnographic characteristics of all participants and categorized RRLMI subgroups are reported in Table 1. Over 11.4 [6.9–12.5] years of follow-up, mortality status was ascertained for 571 subjects. There were 287 all-cause deaths and 116 CVD-related deaths. The sample had an average age (median [interquartile range]) of 76.0 [72.0;81.0] years and an average BMI of 28.0 [25.0;30.0] kg/m2. After excluding those with history of previous CVD, there were 135 incident CVDs over 6.1 [2.6–10.3] years of follow-up. Baseline AHI, HB, ΔHR, and RRLMI are shown in Table 1. A total of 336 (58.8 per cent) participants had RRLMI ≥ 5 events/h and 27 (4.7 per cent) had RRLMI ≥ 15 events/h. Spearman correlation between AHI and RRLMI was 0.51 (Figure 1). Baseline demographic and clinical characteristics of the analytic sample were comparable to those of the broader MrOS sleep cohort, with similar distributions of age, BMI, race/ethnicity, smoking status, alcohol use, diabetes, and HTN (Supplementary Table S1).

Table 1.

Demographic and baseline characteristics of all participants and those in low or high RRLMI category

Baseline Characteristics All (N = 571)
336 (58.8%): RRLMI ≥ 5;
27 (4.7%): RRLMI≥15 (events/h)		 RRLMI < 6
(N = 286)		 RRLMI ≥ 6
(N = 285)
Age (years) 76.0 [72.0;81.0] 75.0 [72.0;79.0] 77.0 [73.0;82.0]
BMI (kg/m2) 28.0 [25.0;30.0] 27.0 [25.0;30.0] 28.0 [25.0;30.0]
Race
White 533 (93.3%) 263 (92.0%) 270 (94.7%)
African American 10 (1.8%) 9 (3.1%) 1 (0.4%)
Asian 15 (2.6%) 8 (2.8%) 7 (2.5%)
Hispanic 10 (1.8%) 5 (1.7%) 5 (1.8%)
other 3 (0.5%) 1 (0.3%) 2 (0.7%)
Smoking status, n (%) N = 570
Never 228 (40.0%) 111 (38.8%) 117 (41.2%)
Former 330(57.9%) 168 (58.7%) 162 (57.0%)
Current 12 (2.1%) 7 (2.4%) 5 (1.8%)
Alcohol consumption 380 (66.9%) 189 (66.3%) 191 (67.5%)
Diabetes, n (%) 78 (13.7%) 38 (13.3%) 40 (14.1%)
Hypertension, n (%) 303 (53.2%) 151 (52.8%) 152 (53.5%)
AHI (events/h) 21.0 [14.0;33.5] 17.0 [13.0;23.0] 27.0 [19.0;41.0]
Hypoxic burden (%min/h) 68.5 [47.3;104.4] 53.9 [38.8;74.1] 89.6 [61.2;137.4]
ΔHR (bpm) 6.6 [4.8;9.5] 6.4 [4.7;9.1] 6.8 [4.8;9.9]
RRLMI (events/h) 6.0 [3.4;9.3] 3.4 [2.3;4.6] 9.3 [7.6;11.8]
PLMI (events/h) 22.0 [2.0;54.0] 9.5 [1.0;32.0] 39.0 [11.0;72.0]
Open in a new tab

Data presented as median [interquartile] for continuous variables and the number (%) for categorical variables. AHI (apnea–hypopnea index): All apneas + hypopneas with ≥30 per cent nasal cannula reduction and ≥3 per cent oxygen desaturation per hour of sleep. Hypertension was defined as participant-reported use of antihypertensive medications, or an average systolic blood pressure measurement of ≥140 mmHg or a diastolic blood pressure measurement of ≥90 mmHg, obtained during a research clinic examination. RRLMI is the number of respiratory related leg movements (RRLMs) per hour of sleep. PLMI is the number of periodic leg movements of sleep (PLMS) per hour of sleep.

Figure 1.

Figure 1

Open in a new tab

Scatterplot of respiratory-related leg movement index (RRLMI) versus apnea–hypopnea index (AHI).

Unadjusted Kaplan–Meier curves showed a probability of all-cause (p = .002; Figure 2A) or CVD-related (p = .01; Figure 2B) mortality in participants with a high versus low RRLMI. As shown in Table 2, RRLMI was significantly associated with all-cause mortality in all adjusted models (aHR of 1.22 [95% CI: 1.08–1.37] for fully-adjusted Model 3). (see Supplementary Table S2, for the subgroup of those without a history of CVD). The RRLMI was significantly associated with CVD-related mortality in the first two models (aHR of 1.28 [95% CI: 1.07–1.53] for Model 2). Adding AHI in Model 3 or replacing AHI with HB in Model 4 resulted in non-significant associations, while replacing AHI with ΔHR in Model 5 resulted in a significant association of RRLMI with CVD-related mortality (aHR of 1.28 [95% CI: 1.06–1.55], Table 2). Finally, RRLMI was not significantly associated with incident CVD in any of the models (Table 2). Results of subgroup analyses with the same models restricted to participants with ≥5 or ≥10 events/h were added as Supplementary Tables S3 and S4. In the subgroup of those with RRLMI ≥ 5 events/h, RRLMI was associated with both all-cause and CVD mortality in all adjusted models, whereas in those with RRLMI ≥ 10 events/h, associations were only significant for all-cause mortality.

Figure 2.

Figure 2

Open in a new tab

(A) Unadjusted Kaplan–Meier curves showed different probabilities of all-cause mortality in participants with low RRLMI (<6 events/h) versus high RRLMI (≥6 events/h). (B) Unadjusted Kaplan–Meier curves showed different probabilities of CVD-related mortality in participants with low RRLMI (<6 events/h) versus high RRLMI (≥6 events/h).

Table 2.

The Number of RRLMs Per hour of sleep (RRLMI) was significantly associated with mortality while associations with incident CVD was not consistent

Models aHR [95% CI]
All-cause mortality		 aHR [95% CI]
CVD-related mortality		 aHR [95% CI]
Incident CVD
Model 1 1.22 [1.09–1.37]***
(N = 559)		 1.26 [1.05–1.51]*
(N = 568)		 1.06 [0.89–1.25], p = .52
(N = 311)
Model 2 1.22 [1.09–1.36]***
(N = 559)		 1.27 [1.07–1.53]**
(N = 568)		 1.06 [0.89–1.25], p = .53
(N = 311)
Model 3 1.17 [1.01–1.34]*
(N = 542)		 1.17 [0.94–1.46], p = .16
(N = 551)		 1.06 [0.87–1.30], p = .55
(N = 311)
Model 4 1.19 [1.04–1.37]*
(N = 542)		 1.16 [0.93–1.45], p = .17
(N = 551)		 1.07 [0.88–1.29], p = .49
(N = 311)
Model 5 1.21 [1.08–1.37]**
(N = 542)		 1.28 [1.06–1.55]**
(N = 551)		 1.05 [0.88–1.24], p = .59
(N = 311)
Open in a new tab

CVD: cardiovascular disease; covariates in Model 1: age, race, smoking, alcohol consumption, and BMI. Model 2: Model 1 + hypertension, prevalent DM, physical activity, antidepressant use, benzodiazepine use. Model 3: Model 2 + AHI for incident CVD, + prevalent CVD (any history of physician-diagnosed coronary heart disease, cerebrovascular disease, and peripheral arterial disease) for both mortality outcomes. Model 4: Model 3 + sleep apnea specific hypoxic burden (HB) instead of AHI, defined as the total area under desaturation curve. Model 5: Model 3 + the heart rate response to respiratory events(ΔHR), instead of AHI.

Significance codes: p < .1, *p < .05; **p < .01; ***p < .001.

In sensitivity analyses, the LMI was not associated with outcomes in this sample (unadjusted model for all-cause mortality: HR = 1.01, 95% CI: 0.89–1.13, p = .92; CVD mortality: HR = 0.95, 95% CI: 0.79–1.14, p = .57; incident CVD: HR = 1.11, 95% CI: 0.93–1.33, p = .27) or in the larger MrOS sleep cohort (Supplementary Table S5). Similarly, PLMI was not associated with these outcomes in all subjects (Table 3) or subgroups of high or low RRLMI (Supplementary Table S6). The mixed-effect regression analysis showed that the duration of RRLMs was significantly associated with presence of arousal (Estimates: 0.98 [95% CI: 0.92–1.04; p-value < .001]), increase in event duration (Estimates: 0.57 [95% CI: 0.48–0.66; p-value < .001]), HBEV (Estimates: 0.35 [95% CI: 0.31–0.38; p-value<0.001]), and ΔHR (Estimates: 0.38 [95% CI: 0.34–0.41; p-value < .001]; Table 4). Figure 3 shows the scatter plot of event duration and RRLM duration.

Table 3.

The number of PLMS Per Hour of Sleep (PLMI) was not associated with outcomes

Models aHR [95% CI]
All-cause mortality		 aHR [95% CI]
CVD-related mortality		 aHR [95% CI]
Incident CVD
Model 1 1.04 [0.93–1.17], p = .45
(N = 559)		 1.00 [0.82–1.21], p = .98
(N = 568)		 1.10 [0.94–1.30], p = .23
(N = 311)
Model 2 1.03 [0.92–1.16], p = .58
(N = 559)		 1.00 [0.82–1.21], p = .98
(N = 568)		 1.09 [0.93–1.29], p = .28
(N = 311)
Model 3 1.04 [0.92–1.17], p = .53
(N = 542)		 0.98 [80–1.19], p = .81
(N = 551)		 1.09 [0.93–1.29], p = .29
(N = 311)
Model 4 1.04 [0.92–1.18], p = .49
(N = 542)		 0.98 [0.80–1.20], p = .85
(N = 551)		 1.09 [0.93–1.29], p = .29
(N = 311)
Model 5 1.05 [0.93–1.18], p = .45
(N = 542)		 0.99 [0.81–1.21], p = .89
(N = 551)		 1.08 [0.92–1.28], p = .22
(N = 311)
Open in a new tab

CVD: cardiovascular disease; covariates in Model 1: age, race, smoking, alcohol consumption, and BMI. Model 2: Model 1 + hypertension, prevalent DM, physical activity, antidepressant use, benzodiazepine use. Model 3: Model 2 + AHI for incident CVD, + prevalent CVD (any history of physician-diagnosed coronary heart disease, cerebrovascular disease, and peripheral arterial disease) for both mortality outcomes. Model 4: Model 3 + sleep apnea specific hypoxic burden (HB) instead of AHI, defined as the total area under desaturation curve. Model 5: Model 3 + the heart rate response to respiratory events (ΔHR), instead of AHI.

Table 4.

Mixed-effect linear regression with duration of rrlms as a dependent variable, HBEV, ΔHR, presence of EEG arousal, and event duration as fixed effects, and subjects as random effects

Duration of RRLM
Predictors Estimates [95% CI] P-value Random effects
Presence of arousal 0.98 [0.92–1.04] <.001 σ 2: 3.74
M/C R2: 0.05/0.21
HBEV, per 1SD 0.35 [0.31–0.38] <.001 σ 2: 3.86
M/C R2: 0.03/0.19
ΔHR, per 1SD 0.38 [0.34–0.41] <.001 σ 2:3.85
M/C R2: 0.03/0.20
Event duration, per 1SD 0.57 [0.48–0.66] <.001 σ 2:3.90
M/C R2: 0.01/0.18
Open in a new tab

Random effects: σ2: between-subjects variability; M/C R2: Marginal R2 (i.e. proportion of the total variance in duration of RRLMs attributable to fixed effects)/Conditional R2 (i.e. the total variance in duration of RRLMs). HBEV: Event-related oxygen desaturation defined by hypoxic burden per event; ΔHR: the heart rate response to apneas and hypopneas.

Figure 3.

Figure 3

Open in a new tab

Scatterplot of averaged RRLM duration versus respiratory event duration.

Discussion

We found significant associations between the frequency of respiratory-related leg movements and risk of mortality in elderly persons with OSA. The magnitude of the association did not change meaningfully after accounting for relevant confounders and risk factors, including medication use, prevalent CVD diseases, AHI, and novel OSA-specific risk factors, such as HB and heart rate response to respiratory events. In addition, we showed that increased duration of respiratory-related leg movements is a marker of a more severe respiratory event, characterized by increased event duration and oxygen desaturation, presence of EEG arousal following termination of events, and elevated heart rate response to events. Similar significant associations were found for intensity (defined as the amplitude of the leg movement signal over the duration) of leg movements (see online Supplementary Method and Supplementary Table S7). These findings suggest that quantifying respiratory-related leg movements could be a practical method for identifying a more severe form of OSA. Such a method could potentially help in predicting adverse health outcomes, thereby improving patient management and treatment strategies.

The findings that the frequency of respiratory-related leg movements predict mortality (while AHI didn’t precisely predict mortality) in the context of OSA may be explained by two alternative explanations: (1) RRLMs are a “composite” marker of more severe respiratory events that better identify a high-risk version of OSA and/or (2) frequency of RRLM is a risk factor for adverse outcomes independent of OSA. Our findings support both of these explanations. The observation that the duration of leg movements is partly explained by severity of events, arousals, and surges in heart rate may suggest that by counting leg movements, we may be probing the more severe respiratory events with greater sympathetic activity [6, 12]. It is well-known that increases in heart rate in association with respiratory events are mediated by activation of the autonomic nervous system, which has been suggested as a key mechanism underlying the link between OSA and CVD [6, 33, 34]. These results are in line with past studies [6] demonstrating that respiratory events with leg movements lead to a notably higher increase in heart rate [6] compared to events occurring without leg movements. More severe respiratory events have also been shown to induce a greater heart rate surge [19, 35]. Alternatively, the frequency of RRLMs may be considered an independent motor phenomenon that predicts adverse outcomes. Our findings, at least to some extent, support this hypothesis. Indeed, the magnitude of association between the RRLMI and all-cause and CVD-related mortality as well as incident CVD does not change meaningfully after adjusting for known OSA-related risk factors, such as HB and heart rate response to events [16, 17, 19, 36] (Table 2). However, our results should be interpreted with caution, given potential residual confounding. In addition, respiratory-related leg movements tend to be more prevalent in individuals with high PLMS [15]. PLMS have been shown to predict adverse outcomes independent of OSA [8–11]. For example, they have been associated with reduced sleep quality and more fragmented sleep in the MrOS Sleep Study [37]. Koo et al. showed in the same dataset that PLMI ≥ 30 events/h was associated with incident CVDs, but these associations were not statistically significant in fully adjusted models [32]. Consistently, we didn’t find significant associations with incident CVDs in our fully adjusted models (with the same level of adjustments). Differences from Koo et al. regarding hazard ratios of PLMI and incident CVD likely reflect differences in study design and analytic approach. Koo et al. analyzed the full MrOS sleep cohort (N = 2911) with shorter follow-up (study conducted in 2011) and compared high versus low PLMI groups, whereas we examined a smaller subset (N = 571), modeled PLMI continuously, and excluded participants with prevalent CVD. Notably, the older PLMS definition excluded only RRLMs occurring within 0.5 s of respiratory events, which may also partly account for the observed association with CVD. Recently, Zinchuk et al. conducted a study that showed PLMI is a risk factor for incident CVD or death among veterans who had AHI < 30 events/h [31]. In our study, PLMI was not significantly associated with death. The potential factors may explain the discrepancy between our findings and those of Zinchuk et al. including differences in study populations (our dataset consists of older males with a mean ± SD age of 76.8 ± 5.4 years and theirs included 93.4 per cent males with a mean ± SD age of 57.3 ± 11.7 years), baseline OSA severity (AHI, median [IQR] of 21.0 [14.0;33.5] in our study vs. 5.5 [1.6–13.4] events/h in theirs), baseline PLMS severity (PLMI, median [IQR] of 22.0 [2.0;54.0] in our study vs. 1.0 [0.0–9.0] PLMS/h in theirs), and less statistical power in our study (our sample size is 571 vs. theirs: 1049).

RRLMs occur in temporal relation to airflow reduction, and we found their duration or intensity is linked to the severity of the respiratory events. The association of RRLMI with mortality, unlike PLMI, may be explained by more severe airflow reduction, a higher probability of cortical arousals, and consequent sympathetic activation, which leads to heart rate increase. Although PLMS can trigger sympathetic activation and transient heart rate elevations, their occurrence outside of respiratory events means they do not capture these pathophysiologic stresses to the same extent, which may explain their lack of association with mortality. Regardless of these mechanistic differences, quantifying respiratory-related leg movements may improve risk prediction in sleep apnea. This is important given that OSA, defined by AHI, has shown inconsistent associations with health outcomes [7, 38, 39]. Although there are several novel metrics such as HB, ventilatory burden, arousal burden, and heart rate response to events predict, these adverse outcomes better than AHI [16, 17, 19, 25, 36, 40], measuring them in a clinical setting may be challenging, however, counting leg movements may provide a simple and predictive marker of OSA severity and health outcomes.

The existing guidelines for scoring PLMS do not exclude RRLMs. Nonetheless, there has been substantial discourse on distinguishing between PLMS and RRLMs. The AASM 2012 rule defines RRLMs as leg movements occurring during 0.5 s before to after the event [13, 15, 25]. WASM 2006 guidelines defined RRLMs as leg movements within the interval of ±0.5 s around the end of the respiratory events which underestimate the number of RRLM [12]. One evidence-based research has shown that in people with OSA, leg movement activity increases mostly around the end of events. They defined a much wider interval, i.e. from 2 s before to 10.25 s after the end of the respiratory event [41] (recommended by WASM 2016 rules in collaboration with the IRLSSG [26, 41]). However, there is no consensus on the best rules defining RRLMs [26]. In the present study, RRLMs were scored if they were lasting 0.5–10 s with onset within 5 s of the respiratory event termination [13]. This choice was consistent with leg movement rescoring approaches used by Aritake et al., 2015 [13] in the same subgroup of the MrOS Sleep Study. Using the 0–5 s window, they found a median RRLM% of ~23 per cent in this group of elderly men with OSA, indicating that many respiratory events have leg movements in this range. Back-averaging analyses, done by Manconi et al. 2015, demonstrated that the peak probability of RRLM onset is approximately 2.5 s after termination, with a distribution extending several seconds before and after [41]. These observations provide support for using a broader window than the AASM or WASM 2006 definitions, as a strictly ±0.5 s cutoff likely underestimates physiologically relevant RRLMs. However, future research is needed to define optimal scoring criteria for RRLMs and to explain the mechanism generating RRLMs.

Limitations

Our sample included a subset of older white males with OSA from the MrOS Sleep Study; therefore, our findings may not generalize to women, non-white individuals, or younger men. Second, our dataset only included PSGs with available manually-scored RRLMs [13], resulting in a relatively small sample of OSA individuals which may explain the lack of significant association between the RRLMI and incident CVD (only 316 of 571 participants had no previous CVD). Third, in this sample, RRLMs were scored if they were within 5 s of event termination [13]. The updated WASM guidelines recommend a −2 to +10.25 s window around the end of events to score an RRLM [26]. Work is now ongoing to both automate the identification of leg movements, regardless of their association with events, and identify the optimal definition that provides strong prediction of outcomes. Fourth, it has been suggested that the occurrence of RRLM might be influenced by the presence of RLS [42]. In this study, we did not have adequate information to test this hypothesize; however, future work is needed to investigate the effect of RLS on the relationship between RRLMs and health outcomes. Fifth, we scored leg movements using piezoelectric sensors rather than anterior tibialis EMG electrodes. While EMG is the gold standard for PLM recording, a validation analysis in the MrOS cohort demonstrated a strong correlation (r = 0.81) between PLMS scored with the MrOS piezoelectric sensors and simultaneous anterior tibialis EMG [43]. Although this data supports the comparability of piezo-derived and EMG-derived leg movement indices, they remain two distinct methodologies. Therefore, caution is warranted when generalizing our results to studies that employ anterior tibialis EMG as the standard for movement detection. Sixth, RRLMI may serve as a potential marker or predictor of mortality, but causality cannot be inferred from the present analysis. Seventh, this sample includes only apneas and hypopneas with at least a 3 per cent oxygen desaturation, and respiratory effort-related arousals (RERAs) were not included.

Interpretation

In conclusion, our study demonstrates significant associations between the frequency of respiratory-related leg movements and increased risk of all-cause and CVD-related mortality in persons with OSA. These associations remained robust even after adjusting for confounders and novel OSA-specific risk factors. Additionally, the duration of these leg movements serves as a marker for severe respiratory events. Monitoring respiratory-related leg movements could thus be a practical approach for identifying severe OSA and predicting adverse health outcomes, enhancing patient management and treatment strategies.

Supplementary Material

ONLINE_SUPPLEMENT_zsaf310
online_supplement_zsaf310.docx (168.1KB, docx)

Acknowledgments

The National Heart, Lung, and Blood Institute provided funding for the ancillary MrOS Sleep Study, “Outcomes of Sleep Disorders in Older Men,” under the following grant numbers: R01 HL071194, R01 HL070848, R01 HL070847, R01 HL070842, R01 HL070841, R01 HL070837, R01 HL070838, and R01 HL070839. The National Sleep Research Resource was supported by the National Heart, Lung, and Blood Institute (R24 HL114473 and 75N92019R002). A.A. was supported by the National Institutes of Health (R21 HL161766 and R01 HL153874).

Contributor Information

Neda Esmaeili, Division of Sleep and Circadian Disorders, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United States.

Sayaka Aritake, Faculty of Health and Social Services, Saitama Prefectural University, Saitama, Japan.

Susan Redline, Division of Sleep and Circadian Disorders, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United States.

John W Winkelman, Departments of Psychiatry and Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States; Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States.

Ali Azarbarzin, Division of Sleep and Circadian Disorders, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United States.

Funding

N.E. reports funding from American Academy of sleep Medicine (AASM, 335-FP-24), A.A. reports funding from NIH (R01 HL153874, R21 HL161766), AHA (19CDA34660137), and the AASM (188-SR-17, SR-2217).

Disclosure statement

Financial disclosure: The authors declare no financial conflicts of interest with respect to the current article.

Non-financial disclosure: N.E. has no conflict of interest. A.A. serves as a consultant for Respicardia, Amgen, Eli Lilly, Inspire, Cerebra and Apnimed. A.A. also serves as an advisory board member for Incannex Healthcare. A.A.’s interests were reviewed by Brigham and Women’s Hospital and Mass General Brigham in accordance with their institutional policies. S.R. has been an unpaid member of the Apnimed Scientific Advisory Committee and a consultant for Eli Lilly Inc. and Amgen Inc.

Data availability

Deidentified signals, covariates, and outcomes data were obtained under agreement from the parent study. Individual data presented in the current study can be obtained by request but may require a three-way data use agreement with parent cohort investigators.

Artificial intelligence use statement: Artificial intelligence wasn’t used for this manuscript.

References

  • 1. Ettehad  D, Emdin  CA, Kiran  A, et al.  Blood pressure lowering for prevention of cardiovascular disease and death: a systematic review and meta-analysis. Lancet. 2016;387(10022):957–967. 10.1016/S0140-6736(15)01225-8 [DOI] [PubMed] [Google Scholar]
  • 2. Peppard  PE, Young  T, Palta  M, Skatrud  J. Prospective study of the association between sleep-disordered breathing and hypertension. N Engl J Med. 2000;342:1378–1384. 10.1056/NEJM200005113421901 [DOI] [PubMed] [Google Scholar]
  • 3. Kendzerska  T, Gershon  AS, Hawker  G, Leung  RS, Tomlinson  G. Obstructive sleep apnea and risk of cardiovascular events and all-cause mortality: a decade-long historical cohort study. PLoS Med. 2014;11:e1001599. 10.1371/journal.pmed.1001599 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Yeghiazarians Y, Jneid H, Tietjens JR, et al.  Obstructive sleep apnea and cardiovascular disease: a scientific statement from the American Heart Association. Circulation. 2021;144:e56–e67. 10.1161/CIR.0000000000000988 [DOI] [PubMed] [Google Scholar]
  • 5. Lewington S, Clarke R, Qizilbash N, Peto R, Collins R, Prospective Studies Collaboration . Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet. 2002;360:1903–1913. 10.1016/s0140-6736(02)11911-8 [DOI] [PubMed] [Google Scholar]
  • 6. Yang  C-K, Jordan  AS, White  DP, Winkelman  JW. Heart rate response to respiratory events with or without leg movements. Sleep. 2006;29:553–556. 10.1093/sleep/29.4.553 [DOI] [PubMed] [Google Scholar]
  • 7. Azarbarzin  A, Labarca  G, Kwon  Y, Wellman  A. Physiological consequences of upper airway obstruction in sleep apnea. Chest. 2024;166(5):1209–1217. 10.1016/j.chest.2024.05.028 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Koo  BB, Blackwell T, Ancoli-Israel S, Stone KL, et al.  Association of incident cardiovascular disease with periodic limb movements during sleep in older men: outcomes of sleep disorders in older men (MrOS) study. Circulation. 2011;124:1223–1231. 10.1161/CIRCULATIONAHA.111.038968 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Al-Alawi  A, Mulgrew  A, Tench  E, Ryan  CF. Prevalence, risk factors and impact on daytime sleepiness and hypertension of periodic leg movements with arousals in patients with obstructive sleep apnea. J Clin Sleep Med. 2006;2(3):281–287. 10.5664/jcsm.26587 [DOI] [PubMed] [Google Scholar]
  • 10. Zinchuk  A, Srivali N, Qin L, et al.  Association of periodic limb movements and obstructive sleep apnea with risk of cardiovascular disease and mortality. J Am Heart Assoc. 2024;13(3):e031630. 10.1161/JAHA.123.031630 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Yumino D, Wang H, Floras JS, et al.  Relation of periodic leg movements during sleep and mortality in patients with systolic heart failure. Am J Cardiol. 2011;107:447–451. 10.1016/j.amjcard.2010.09.037 [DOI] [PubMed] [Google Scholar]
  • 12. McCall  CA, Winkelman  JW. Respiratory-related leg movements of sleep are associated with serotonergic antidepressants but not bupropion. J Clin Sleep Med. 2018;14:1569–1576. 10.5664/jcsm.7338 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Aritake S, Blackwell T, Peters KW, et al.  Prevalence and associations of respiratory-related leg movements: the MrOS sleep study. Sleep Med. 2015;16:1236–1244. 10.1016/j.sleep.2015.06.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Schipper MH, Alvarez-Estevez D, Jellema K, et al.  Sleep-related leg movements in obstructive sleep apnea: definitions, determinants, and clinical consequences. Sleep Med. 2020;75:131–140. 10.1016/j.sleep.2020.06.004 [DOI] [PubMed] [Google Scholar]
  • 15. Fulda  S, Heinzer  R, Haba-Rubio  J. Characteristics and determinants of respiratory event–associated leg movements. Sleep. 2018;41(2). 10.1093/sleep/zsx206 [DOI] [PubMed] [Google Scholar]
  • 16. Azarbarzin  A, Sands SA, Stone KL, et al.  The hypoxic burden of sleep apnoea predicts cardiovascular disease-related mortality: the Osteoporotic Fractures in Men Study and the Sleep Heart Health Study. Eur Heart J. 2019;40:1149–1157. 10.1093/eurheartj/ehy624 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Azarbarzin A, Sands SA, Taranto-Montemurro L, et al.  The sleep apnea-specific hypoxic burden predicts incident heart failure. Chest. 2020;158:739–750. 10.1016/j.chest.2020.03.053 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Azarbarzin  A, Ostrowski  M, Hanly  P, Younes  M. Relationship between arousal intensity and heart rate response to arousal. Sleep. 2014;37:645–653. 10.5665/sleep.3560 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Azarbarzin A, Sands SA, Younes M, et al.  The sleep apnea–specific pulse-rate response predicts cardiovascular morbidity and mortality. Am J Respir Crit Care Med. 2021;203:1546–1555. 10.1164/rccm.202010-3900OC [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Blank JB, Cawthon PM, Carrion-Petersen ML, et al.  Overview of recruitment for the osteoporotic fractures in men study (MrOS). Contemp Clin Trials. 2005;26(5):557–568. 10.1016/j.cct.2005.05.005 [DOI] [PubMed] [Google Scholar]
  • 21. Orwoll  E, Blank  JB, Barrett-Connor  E, et al.  Design and baseline characteristics of the osteoporotic fractures in men (MrOS) study—a large observational study of the determinants of fracture in older men. Contemp Clin Trials. 2005;26(5):569–585. 10.1016/j.cct.2005.05.006 [DOI] [PubMed] [Google Scholar]
  • 22. Blackwell  T, Yaffe  K, Ancoli-Israel  S, et al.  Associations between sleep architecture and sleep-disordered breathing and cognition in older community-dwelling men: the Osteoporotic Fractures in Men Sleep Study. J Am Geriatr Soc. 2011;59(12):2217–2225. 10.1111/j.1532-5415.2011.03731.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Zhang GQ, Cui L, Mueller R, et al.  The National Sleep Research Resource: towards a sleep data commons. J Am Med Inform Assoc. 2018;25:1351–1358. 10.1093/jamia/ocy064 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Rechtschaffen A, Kales A. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. Washington DC, Public Health Service. US Government Printing Office, 1968.
  • 25. Lee  S-A, Lee  Y-S, Cho  C-W, Yang  H, Im  K. Different scoring rules for respiratory event-related leg movements: effects on the prevalence of periodic limb movements during sleep and their association with depressive symptoms in patients with obstructive sleep apnea. Sleep Med. 2021;85:204–210. 10.1016/j.sleep.2021.07.020 [DOI] [PubMed] [Google Scholar]
  • 26. Ferri  R, Fulda  S, Allen  RP, et al.  World Association of Sleep Medicine (WASM) 2016 standards for recording and scoring leg movements in polysomnograms developed by a joint task force from the International and the European Restless Legs Syndrome Study Groups (IRLSSG and EURLSSG). Sleep Med. 2016;26:86–95. 10.1016/j.sleep.2016.10.010 [DOI] [PubMed] [Google Scholar]
  • 27. Iber C, Ancoli-Israel S, Chesson AL, Quan SF. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology, and Technical Specification. Westchester, IL, American Academy of Sleep Medicine. 2007.
  • 28. Hajipour M, Beaudin AE, Raneri JK, et al. Association between respiratory event-related hypoxemia, autonomic responses to respiratory events and chronic kidney disease in obstructive sleep apnea. Am J Respir Crit Care Med 2024;209:A4733. 10.1164/ajrccm-conference.2024.209.1_MeetingAbstracts.A4733 [DOI] [Google Scholar]
  • 29. Berry RB, Budhiraja R, Gottlieb DJ, et al. Rules for scoring respiratory events in sleep: update of the 2007 AASM manual for the scoring of sleep and associated events: deliberations of the sleep apnea definitions task force of the American Academy of Sleep Medicine. Journal of clinical sleep medicine. 2012;8(5):597–619. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Baumert  M, Linz  D, Stone  K, et al.  Mean nocturnal respiratory rate predicts cardiovascular and all-cause mortality in community-dwelling older men and women. Eur Respir J. 2019;54(1):1802175. 10.1183/13993003.02175-2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Zinchuk A, Srivali N, Qin L, et al.  Association of periodic limb movements and obstructive sleep apnea with risk of cardiovascular disease and mortality. J Am Heart Assoc. 2024;13(3):e031630. 10.1161/JAHA.123.031630 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Koo  BB, Blackwell  T, Ancoli-Israel  S, et al.  Association of incident cardiovascular disease with periodic limb movements during sleep in older men: outcomes of sleep disorders in older men (MrOS) study. Circulation. 2011;124(11):1223–1231. 10.1161/CIRCULATIONAHA.111.038968 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Shepard  JW  Jr. Hypertension, cardiac arrhythmias, myocardial infarction, and stroke in relation to obstructive sleep apnea. Clin Chest Med. 1992;13:437–458. 10.1016/S0272-5231(21)00873-X [DOI] [PubMed] [Google Scholar]
  • 34. Winkelman  JW. The evoked heart rate response to periodic leg movements of sleep. Sleep. 1999;22:575–580. 10.1093/sleep/22.5.575 [DOI] [PubMed] [Google Scholar]
  • 35. Azarbarzin  A, Ostrowski  M, Moussavi  Z, Hanly  P, Younes  M. Contribution of arousal from sleep to postevent tachycardia in patients with obstructive sleep apnea. Sleep. 2013;36:881–889. 10.5665/sleep.2716 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Esmaeili  N, Labarca  G, Hu  WH, et al.  Hypoxic burden based on automatically identified desaturations is associated with adverse health outcomes. Ann Am Thorac Soc. 2023;20(11):1633–1641. 10.1513/AnnalsATS.202303-248OC [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Claman  DM, Ewing  SK, Redline  S, et al.  Periodic leg movements are associated with reduced sleep quality in older men: the MrOS sleep study. J Clin Sleep Med. 2013;09(11):1109–1117. 10.5664/jcsm.3146 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Peker  Y, Zinchuk  A, Celik  Y, Sands  SA, Azarbarzin  A. Hypoxic burden but not AHI predicts risk of cardiovascular events: a secondary analysis of the RICCADSA clinical trial. Am J Respir Crit Care Med. 2024;A4730. 10.1164/ajrccm-conference.2024.209.1_MeetingAbstracts.A4730 [DOI] [Google Scholar]
  • 39. Redline  S, Azarbarzin  A, Peker  Y. Obstructive sleep apnoea heterogeneity and cardiovascular disease. Nat Rev Cardiol. 2023;20:560–573. 10.1038/s41569-023-00846-6 [DOI] [PubMed] [Google Scholar]
  • 40. Butler  MP, Emch  JT, Rueschman  M, et al.  Apnea–hypopnea event duration predicts mortality in men and women in the sleep heart health study. Am J Respir Crit Care Med. 2019;199(7):903–912. 10.1164/rccm.201804-0758OC [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Manconi  M, Zavalko  I, Fanfulla  F, Winkelman  JW, Fulda  S. An evidence-based recommendation for a new definition of respiratory-related leg movements. Sleep. 2015;38:295–304. 10.5665/sleep.4418 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Manconi  M, Zavalko  I, Bassetti  CL, Colamartino  E, Pons  M, Ferri  R. Respiratory-related leg movements and their relationship with periodic leg movements during sleep. Sleep. 2014;37(3):497–504. 10.5665/sleep.3484 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43. Claman  DM, Ewing  SK, Redline  S, et al.  Periodic leg movements are associated with reduced sleep quality in older men: the MrOS sleep study. J Clin Sleep Med. 2013;09(11):1109–1117. 10.5664/jcsm.3146 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ONLINE_SUPPLEMENT_zsaf310
online_supplement_zsaf310.docx (168.1KB, docx)

Data Availability Statement

Deidentified signals, covariates, and outcomes data were obtained under agreement from the parent study. Individual data presented in the current study can be obtained by request but may require a three-way data use agreement with parent cohort investigators.

Artificial intelligence use statement: Artificial intelligence wasn’t used for this manuscript.

RESOURCES