The relationship between sleep duration and arterial stiffness: A meta-analysis

Abstract

Chronically short (<7 h) and long (>9 h) sleep duration may increase cardiovascular disease (CVD) risk relative to the recommended sleep duration (7–9 h). The objective of this study was to evaluate the effects of short and long sleep duration on arterial stiffness, a marker of CVD risk, in adults. Eleven cross-sectional studies were reviewed with a total sample size of 100,050 participants (64.5% male). Weighted mean differences (WMD) and 95% confidence intervals (95% CI) were calculated and pooled using random effects models, and standardized mean differences (SMD) were calculated to determine effect size magnitude. Compared to the recommended sleep duration, both short (WMD = 20.6 cm/s, 95% confidence intervals (CI): 13.8–27.4 cm/s, SMD = 0.02) and long sleep duration (WMD = 33.6 cm/s, 95% CI: 20.0–47.2 cm/s, SMD = 0.79) were associated with higher (detrimental) pulse wave velocity (PWV). The associations between short sleep and higher PWV in adults with cardiometabolic disease, and long sleep and higher PWV in older adults, were also significant in sub-group analysis. These findings indicate short and long sleep duration may contribute to subclinical CVD.

1. Introduction

Nearly half of the U.S. population regularly obtains short (40%) or long (5%) sleep [1], both of which are associated with overt [2–5] and subclinical cardiovascular disease (CVD). For example, relative to those sleeping 7–9 h, short [6–8] and long [9] sleepers (defined using National Sleep Foundation criteria [10]) are reported to have greater central pulse wave velocity (PWV). PWV reflects arteriosclerosis and is the gold-standard non-invasive measure of macrovascular health [11]. Because PWV assesses subclinical rather than clinical outcomes, it can be used as a continuous indicator of potential CVD outcomes [12,13], providing more information for risk prediction. Currently, there is a scarcity of comprehensive analysis that investigates the association between sleep duration and PWV of both short and long sleepers simultaneously. Clarifying the association between sleep duration and PWV will allow for more accurate prediction of risk for all potential CVD outcomes.

1.1. Objective

The purpose of this meta-analysis was to review the available literature and determine the association of short (<7 h) and long (>9 h) sleep duration with central PWV, each relative to a reference group (7–9 h).

2. Methods

This meta-analysis was carried out in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [14].

2.1. Data sources and searches

Electronic databases (PubMed and SPORTDiscus) were searched by two authors (PPL, AP) utilizing the keywords: “((sleep)) AND (arterial stiffness OR pulse wave velocity)” and “sleep AND (arterial stiffness OR pulse wave velocity)”, respectively. The reference lists of all identified studies and relevant reviews or editorials were also examined. The search was limited to English language studies published between inception and August 2022.

2.2. Article selection

Two researchers (PPL, AP) completed the study selection independently. For the purpose of this meta-analysis, the term “study” is the published manuscript, and “effect size” is the unit of analysis. A given study may have resulted in more than one eligible “effect size” if the article included more than one observed group. In studies with multiple observation arms and a single control group, the sample size of the control group was divided by the number of treatment groups to avoid over-inflation of the sample size [15]. Repeated publications for the same studies were excluded. Initially, study titles and abstracts were screened for relevance. The full-text of potentially eligible articles were obtained to review eligibility for inclusion. The following criteria were used to select studies for inclusion in the review: (i) English language articles; (ii) differentiated between short and/or long sleep duration, and a reference group of 7–9 h; (iii) related sleep duration and a measure of central PWV; and (iv) participants were aged 18 years or older.

2.3. Data extraction and quality assessment

Data extraction was completed by two researchers (AP, PPL). Data extracted for each eligible study and effect size included bibliographic information (author, publication year), baseline participant characteristics, details of intervention(s), and results of reported outcomes.

Study quality was assessed using the National Heart, Lung, and Blood Institute (NHLBI) Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (range 1–3, with 1 being low quality, 2 being fair quality and 3 being highest quality), which includes items related to sample size justification, likelihood of sample bias, and quality of statistical test choice [16]. This tool was chosen due to the cross-sectional nature of all studies. Quality assessment was independently completed by two researchers (AP, PPL) and scores verified by a sleep study expert (CEK). If there was a scoring discrepancy, the sleep study expert (CEK) would re-score the article, and that would be the final score.

2.4. Data synthesis

Two authors (LS and PPL) conducted the data synthesis. Mean and standard deviation values for the outcomes of interest were entered into a spreadsheet. When the required data for a given publication were incomplete, i.e., missing mean or standard deviation, the corresponding author was contacted. If the author did not respond or was unable to provide the data, the missing values were estimated using methods from the Cochrane Handbook for Systematic Reviews of Interventions [15]. For studies reporting multiple intervals of short or long sleep, a weighted mean based on sample size of all groups that were short or long sleep groups were used for analysis of short sleep and long sleep duration for that study. For example, if an article contained a group that slept 5–6 h and another that slept 6–7 h, a weighted mean was calculated from both groups, and categorized as short sleep.

2.5. Data analysis

Two authors (LS and PPL) conducted the data analysis. All statistical analyses and ancillary analyses were performed using the metafor-package [17] for the R statistical environment (R version 3.6.3). Outcome measures were calculated as weighted mean differences (WMDs) as well as the standardized mean difference (SMD). The SMD was used to determine the magnitude of the effect, where <0.2, 0.2, 0.5, and 0.8 were defined as trivial, small, moderate, and large, respectively [18]. Effects sizes were pooled using random-effects modelling (DerSimonian-Laird). Random-effects were used as it allows for potential heterogeneity within- and between-effect size variance [19]. Subsequently, we examined the robustness of the pooled results, the potential for publication bias, and heterogeneity. To test robustness, sensitivity analyses were performed by excluding one effect size at a time. Measures of publication bias included visual inspection of Begg’s funnel plots [20] and inspection of Doi plots with The Luis Furuya Kanamori (LFK) index. Doi plots are reported to be more objective than the qualitative Begg’s funnel plot, and the LFK index has been reported to be a more sensitive quantitative measure of asymmetry and potential bias than the Egger’s regression intercept test [14]. Using the LFK index, <1 indicates no asymmetry, 1 to 2 suggests minor asymmetry, and >2 indicates major asymmetry [14]. When asymmetry was detected, Duval and Tweedie’s trim-and-fill analysis [21] was utilized to determine if the relationship varied as a result of sample size.

Statistical heterogeneity was assessed using the I² statistic, where <25%, 25%, and 75% represent low, moderate, and considerable heterogeneity, respectively [15]. Heterogeneity of ≥25% was assumed to indicate that effect sizes could not be treated as estimates of one common effect size, justifying a priori-specified moderator analyses. Potential moderators included: sleep duration assessment method (objective method such as accelerometery or polysomnography, or subjective method such as a survey or diary), method for measuring central PWV (brachial-ankle PWV [baPWV], carotid-femoral PWV [cfPWV], or aortic PWV [aPWV]), age category of the sample based on mean sample age (mean age <65 years [young] or ≥65 years [old]), cardiometabolic disease (CMD) status (no [healthy], yes [diseased]). A sample was considered to have CMD if the study population had at least one CVD risk factor (hypertension, atrial fibrillation, sleep apnea, tobacco smoking, chronic kidney disease) or a metabolic disease (pre-diabetes or diabetes, dyslipidemia, metabolic syndrome, fatty liver disease). In the event of a significant moderator effect, at p < 0.1, sub-group analyses were conducted using dummy-coded variables.

3. Results

3.1. Literature search and effect size selection

A flow chart detailing the article selection process is provided as Fig. 1. A total of 571 potentially eligible studies were identified. Following screening of abstracts and titles, 555 were excluded because they did not meet selection criteria. Sixteen cross-sectional studies were assessed for eligibility. Five studies [22–24] were excluded because the manuscript was either a review article (n = 2), did not explicitly report sleep duration (n = 2), or did not have a PWV measure (n = 1). A total of 11 studies [25–35] yielding 23 effect sizes were included in the final analyses.

Fig. 1. Study Selection Process.

Flow chart including inclusion and exclusion criteria. The term ‘article’ is used synonymously with ‘study’, and ‘effect size’ is the unit included in the meta-analysis.

3.2. Study setting and participants

Included study characteristics are summarized in Table 1. The studies were carried out in China (n = 2), Netherlands (n = 1), Taiwan (n = 2), Japan (n = 3), US (n = 1), and Britain (n = 1). The number of participants in each study ranged from 27 to 14350. Seven effect sizes included only female participants and seven effect sizes only included male participants. The mean age of the participants ranged from 40.7 years to 70.3 years. Three studies examined clinical populations with CMD, including one effect size looking at a population with at least one CVD risk factor (hypertension, pre-diabetes or diabetes, dyslipidemia, tobacco smoking, chronic kidney disease, atrial fibrillation, metabolic syndrome, or sleep apnea) and one that only evaluated hypertensive individuals [30,32]. Another study included patients with non-alcoholic fatty liver disease, which is a risk factor for CMD [35]. Twelve effect sizes reported ethnicity; two of these effect sizes recruited Dutch, South-Asian, African Surinamese, Ghanaian, Turkish, or Moroccan adults, respectively [35].

Table 1.

Dataset characteristics.

Author	Year	Total Sample (n)	Short Sleep Duration (n)	Long Sleep Duration (n)	Average Age	Sleep Duration Assessment	Cardiometabolic Disease Status	PWV Method
Yoshioka et al. [26]	2011	4,268	1538	865	48.0 ± 6.9	Self-report	No CM Disease	baPWV
Tsai et al. [25]	2014	3,508	492	174	42.3 ± 9.6	Self-report	No CM Disease	baPWV
Kim et al. [34]	2015	18,106	8739	3216	41.8 ± 7.3	Self-report	No CM Disease	baPWV
Cao et al. [31]	2016	15,372	1022	4354	45.6 ± 10.1	Self-report	No CM Disease	baPWV
Niijima et al. [32]	2016	2,304	716	775	64.7 ± 11.3	Self-report	CM Disease	baPWV
Zonoozi et al. [28]	2017	1,429	209	95	78.3 ± 4.5	Self-report	No CM Disease	cfPWV
Logan et al. [33]	2018	908	252	104	68.4 ± 9.1	Accelerometry	No CM Disease	aPWV
Anujuo et al. [35]	2019	10,994	3601	839	40.7 ± 0.8	Self-report	No CM Disease	aPWV
Del Brutto et al. [29]	2019	303	89	27	70.3 ± 7.8	Self-report	No CM Disease	cfPWV
Hu et al. [30]	2020	14,485	621	5197	64.4 ± 7.4	Self-report	CM Disease	baPWV
Liu et al. [27]	2020	10,808	6210	254	50.9 ± 11.7	Self-report	No CM Disease	baPWV

Data represented as means ± standard deviations. CM, cardiometabolic; baPWV, brachial-ankle pulse wave velocity; cfPWV; carotid-femoral pulse wave velocity; aPWV, aortic pulse wave velocity.

3.3. Methodological quality assessment

The methodological assessment of included effect sizes is summarized in Table 2. All studies were rated with a score of “2”, indicating “fair” quality.

Table 2.

The association of short sleep vs. recommended sleep duration with age, CM status, and pulse wave velocity (PWV).

	Effect Size	Sample	Quality	Pooled Effect				Asymmetry	Heterogeneity
	(n)	(n)	/3	WMD	95%CI	SMD	P	LFK	Q	P	I2
All	23	72795	2.0	20.6	(13.80, 27.4)	0.02	<0.001	8.20	471.5	<0.001	95
Age			2.0
<65	19	68852	2.0	23.1	(15.9, 30.2)	0.03	<0.001		453.3	<0.001	96
>65	4	3943	2.0	−3.9	(−48.9, 41.0)	0.02	0.834		17.5	<0.001	83
CM Status			2.0
Healthy	21	61978	2.0	22.6	(15.6, 29.6)	0.03	<0.001		459.0	<0.001	96
Diseased	2	10817	2.0	−10.2	(−85.8, 65.5)	−0.03	0.792		12.6	<0.001	92
PWV method			2.0
cfPWV	2	1610	2.0	30.0	(7.9, 52.1)	0.18	0.008		0.0	1.000	0
baPWV	8	60226	2.0	11.3	(2.9, 19.6)	−0.20	0.008		363.6	<0.001	98
aPWV	13	10959	2.0	33.4	(20.1, 46.8)	0.18	<0.001		25.6	0.012	53

SMD: trivial, small, moderate and large effect sizes are defined as <0.2, 0.2, 0.5, and 0.8, respectively. LFK:<1 indicates no asymmetry, 1–2 suggests minor asymmetry, and >2 indicates major asymmetry. I²: 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively. WMD, weighted mean difference; LCI, lower confidence interval; UCI, upper confidence interval; SMD, standardized mean difference; LFK, Luis Furuya-Kanamori Index; CM, cardiometabolic disease.

3.4. Synthesis of the results

3.4.1. Short sleep duration

Short sleep duration was associated with a small, but significantly higher (detrimental) PWV compared to the recommended sleep duration (WMD = 20.6 cm/s, 95% confidence interval (CI): 13.8–27.4 cm/s, SMD = 0.02) (Fig. 2). Sensitivity analysis did not indicate an undue effect by any effect size estimate. Visual inspection of the funnel plot revealed substantial asymmetry, which was supported by an LFK index of 8.2 (major asymmetry). Trim and fill analysis decreased the WMD to −0.4 cm/s (95%CI: −7.1 to 6.4 cm/s) (Supplemental Fig. 1). There was considerable heterogeneity (I² = 96%, p < 0.001), which may be partially explained by differences in sleep duration assessment methods, age category, CMD status, and PWV measurement modality.

Fig. 2. Effect of short sleep duration and pulse wave velocity (PWV) method.

The effect of short sleep duration on PWV by method used to obtain PWV. WMD, weighted mean difference; n, sample size; Q, Cochran’s Q; aPWV, aortic PWV; baPWV, brachial-ankle PWV; cfPWV, carotid-femoral PWV; 95% CI, 95% confidence interval.

Moderator analyses indicated that age category (p = 0.031) and CMD status (p = 0.025) were moderators when evaluated individually. Further, moderator analysis for PWV method indicated that baPWV (p < 0.001), compared to aPWV, was a moderator, but not cfPWV. In the simultaneous model including age category, CMD status, and PWV method, all but CMD status were moderators. Therefore, we conducted sub-group analyses by age category, CMD status, and PWV method since there was evidence that these factors were modifiers when evaluated individually or in a simultaneous model. Sub-group analyses are reported in Table 2. For brevity we will discuss the significant sub-group associations here. Statistically significant associations were found between short sleep duration and PWV in the younger adult sub-group (WMD = 23.1 cm/s, 95%CI: 15.9–30.2 cm/s, p < 0.001, SMD = 0.03), in healthy adults (WMD = 22.6 cm/s, 95%CI: 15.6–29.6 cm/s, p < 0.001, SMD = 0.03), in studies using cfPWV (WMD = 30.0 cm/s, 95%CI: 7.9–52.1 cm/s, p < 0.01, SMD = 0.18), in those using baPWV (WMD = 11.3 cm/s, 95%CI: 2.9–19.6 cm/s, p < 0.01 SMD = −0.20), and in studies using aPWV (WMD = 33.4 cm/s, 95%CI: 20.1–46.8 cm/s, p < 0.001, SMD = 0.18). The mean and standard deviation for the younger adult and older adult age groups were 48.6 ± 7.72 y and 70.5 ± 8.07 y respectively.

3.4.2. Long sleep duration

Compared to the recommended sleep duration, long sleep duration was associated with a higher PWV (WMD = 33.6 cm/s, 95% CI: 20.0–47.2 cm/s, p < 0.001) (Fig. 3). The SMD was 0.79, indicating a large effect size. Sensitivity analysis indicated that none of the effect sizes unduly influenced the pooled effect. Visual inspection of the funnel plot and the LFK index (−1.56) both indicated minor asymmetry. Additionally, the trim and fill analysis did not change the overall outcome (WMD = 33.6 cm/s, 95%CI: 20.0–47.2 cm/s) (Supplemental Fig. 2). There was considerable heterogeneity (I² = 94%, p < 0.001), which may be partially explained by differences in sleep duration assessment methods, age category, CMD status, and PWV measurement.

Fig. 3. Effect of long sleep duration and pulse wave velocity (PWV) method.

The effect of long sleep duration on PWV by method used to obtain PWV. PWV is in units of cm/s. WMD, weighted mean difference; n, sample size; Q, Cochran’s Q; aPWV, aortic PWV; baPWV, brachial-ankle PWV; cfPWV, carotid-femoral PWV; 95% CI, 95% confidence interval.

Moderator analyses indicated that age category (p = 0.061) and CMD status (p = 0.008) were moderators when evaluated individually. When entered in a simultaneous model with age category, CMD status, and PWV method, only age category (p = 0.023) and cfPWV compared to aPWV (p = 0.097) were moderators. Therefore, we conducted sub-group analyses by age, CMD status, and PWV method since there was evidence that these factors were modifiers when evaluated individually or in a simultaneous model. Subgroup analyses are reported in Table 3; for brevity we will discuss the significant sub-group analyses here. Sub-group analysis revealed stronger associations for populations: older adult subgroup (WMD = 79.5 cm/s, 95% CI: 1.2–157.8 cm/s, p < 0.05, SMD = 0.27), younger adult sub-group (WMD = 28.2 cm/s, 95% CI: 13.8–42.6 cm/s, p < 0.05, SMD = 0.90), adults with CMD (WMD = 86.8 cm/s, 95% CI: 28.6–145.0 cm/s, p < 0.01, SMD = 0.25), healthy adults (WMD = 26.7 cm/s, 95% CI: 12.1–41.3 cm/s, p < 0.01, SMD = 0.86), and in studies using baPWV (WMD = 45.8 cm/s, 95% CI: 27.9–63.8 cm/s, p < 0.001, SMD = 1.98). The mean and standard deviation for the younger adult and older adult age groups were 48.6 ± 7.72 y and 70.5 ± 8.07 y respectively.

Table 3.

The association of long sleep vs. recommended sleep duration with age, CM status, and pulse wave velocity (PWV).

	Effect Size	Sample	Quality	Pooled Effect				Asymmetry	Heterogeneity
	(n)	(n)	/3	WMD	95%CI	SMD	P	LFK	Q	P	I2
All	23	65206	2.0	33.6	(20.0, 47.2)	0.79	<0.001	−1.56	342.9	<0.001	94
Age			2.0
<65	19	61528	2.0	28.2	(13.8, 42.6)	0.90	<0.001		294.5	<0.001	94
>65	4	3678	2.0	79.5	(1.2, 157.8)	0.27	0.046		38.9	<0.001	92
CM Status			2.0
Healthy	21	49754	2.0	26.7	(12.1, 41.3)	0.86	<0.001		318.8	<0.001	94
Diseased	2	15452	2.0	86.8	(28.6, 145.0)	0.25	0.004		9.8	0.002	90
PWV method			2.0
cfPWV	2	1434	2.0	36.2	(−41.5, 113.8)	0.25	0.361		3.2	0.073	69
baPWV	8	55723	2.0	45.8	(27.9, 63.8)	1.98	<0.001		268.4	<0.001	97
aPWV	13	8049	2.0	21.1	(−2.8, 45.0)	0.11	0.084		25.1	0.014	52

SMD: trivial, small, moderate and large effect sizes are defined as<0.2, 0.2, 0.5, and 0.8, respectively. LFK:<1 indicates no asymmetry, 1–2 suggests minor asymmetry, and >2 indicates major asymmetry. I²: 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively. WMD, weighted mean difference; LCI, lower confidence interval; UCI, upper confidence interval; SMD, standardized mean difference; LFK, Luis Furuya-Kanamori Index; CM, cardiometabolic disease.

4. Discussion

The aim of this meta-analysis was to synthesize the existing literature investigating the association between sleep duration (short and long) and PWV, a well-validated marker of arterial stiffness and an established CVD risk biomarker [36]. The main findings were: 1) both short and long sleep duration were associated with higher (worse) PWV (short: WMD = 20.6 cm/s; long: WMD = 33.6 cm/s), and 2) the magnitude of the association was stronger for long sleep duration. These results provide high-level evidence for an association between sleep duration and vascular health.

4.1. Limitations

Although this meta-analysis was conducted with high rigor, there were several limitations to consider. All studies included were cross-sectional, preventing derivation of causal relationships or inferring details about specific relationships (linear or non-linear in nature). Among the studies, some further limitations exist due to some methodological differences between studies. All but one study used the Pittsburgh Sleep Quality Index (PSQI) or a modified version to assess sleep duration. While the PSQI is a validated tool for assessing self-reported sleep duration, the correlation between the subjective sleep duration question and objective sleep measures is modest, particularly in individuals 45 years old and above [37,38]. Further, this meta-analysis is likely not generalizable to populations with regularly shifting or other abnormal sleep patterns, such as shift workers. While some studies captured individuals with abnormal sleep duration (e.g., <4 h or >10 h) [26,34], these groups were very small. Lastly, only a priori subgroups were investigated to account for heterogeneity in the study results. While data for additional lifestyle factors in the analyzed studies was sparse, it is possible untested lifestyle factors could account for some of the observed heterogeneity. Such factors could include diet, physical activity, psychological distress levels, and depression, which have been related to both short and long sleep duration [39].

4.2. Comparison with other studies: short sleep duration

Our findings for short sleep differ from the findings of a prior systematic review [24] on the associations of PWV with short and long sleep duration. Whereas the previous review [24] identified a mixed relationship (i.e., some positive and some negative associations) between short sleep and PWV, our study found short sleep was associated with higher PWV (WMD = 20.6 cm/s), albeit with a small effect size (SMD = 0.02). One likely reason for the divergent findings is the greater number of studies (11 vs 4) included in the current study, thereby providing a clearer picture of the relationship between short sleep and PWV. Even still, there was considerable heterogeneity in the results. Part of the heterogeneity could be explained by participant characteristics. Contrary to the findings for long sleep duration, subgroup analyses revealed that the association between PWV and short sleep was stronger for adults categorized as younger (WMD = 23.1 cm/s) and as healthy (WMD = 22.6 cm/s). The reason for this difference could be related to naturally-occurring reductions in sleep duration with age, with older adults more frequently reporting regular short sleep duration [40]. While short sleep in older adults may be reflective of a natural reduction in sleep duration or changes in sleep quality parameters (such as wake after sleep onset), short sleep in younger adults may be more likely related to factors such as stress or deliberate sleep restriction, which could have cardiovascular consequences manifesting as higher PWV.

4.3. Comparison with other studies: long sleep duration

Our study found that long sleep was significantly associated with higher PWV (WMD = 33.6 cm/s), with a large effect size (SMD = 0.79). This finding is consistent with a systematic review published in 2017, which reported an association between long sleep and PWV in two out of the three included trials [24]. In the current study we were able to support and extend the findings of the previous review [24] by incorporating more trials (11 studies, yielding 23 effect size comparisons) and objectively quantifying potential sources of variation (i.e., age category, CMD status, and PWV methodology) using moderator analysis. The association between PWV and sleep duration was maintained in adults categorized as older (WMD = 79.5 cm/s) and those categorized as younger (WMD = 28.2 cm/s). Furthermore, the association held in adults with CMD (WMD = 86.8 cm/s) as well as those without CMD (WMD = 26.7 cm/s). While causality is yet to be determined, the consistent associations found across these subgroups indicate there may be additional sources of variation in the relationship other than age and CM status. Long sleep duration has been associated with reduced sleep quality and high stress [41], both of which may influence the association between sleep duration and PWV. We also found that the association between long sleep and higher PWV was stronger in studies using baPWV (WMD = 45.8 cm/s) compared to studies using cfPWV (WMD = 36.2 cm/s) or aPWV (WMD = 21.1 cm/s). These PWV methodology-dependent findings may, at least partially, be attributable to differences in the arterial pathway measured. While cfPWV and aPWV are measures of central (aortic) arterial stiffness, baPWV also incorporates the peripheral arteries of the leg. Long sleep duration may be particularly damaging to peripheral arteries in the leg, which would not be detected by cfPWV or aPWV. However, it is more likely that the findings are explainable by measurement accuracy and/or precision. The aPWV is estimated from pressure waveforms collected on the upper arm and may not truly reflect central arterial stiffness. The cfPWV is the gold-standard measure of central arterial stiffness but can be technically challenging to capture. Conversely, the baPWV is simple to conduct, is predominantly automated (using blood pressure cuffs on the upper arm and ankle), and the reported precision is superior to cfPWV [42,43], with one community-based study reporting a between-day intra-class correlation of 0.70 for cfPWV and 0.84 for baPWV [43]. The superior precision and lesser operator-dependency may, at least partially, explain the stronger relationship between long sleep with baPWV compared to cfPWV.

4.4. Implications

The implications of this study are summarized in Table 4. Public health organizations such as the American Heart Association have called for further characterization of the relationship between sleep duration and CVD risk [44], going so far as to add sleep to “Life’s Simple 7” to become the “Life’s Essential 8” [45]. Overall, our findings indicate that short sleep duration had a small association with PWV, and long sleep had a large association with PWV. These findings support current sleep recommendations for minimum hours of sleep by current public health organizations [46], while reinforcing the limited evidence for the association between long sleep duration and higher PWV. This analysis also builds on prior research by specifically indicating short and long sleep duration are contraindicated for those with existing CMD. While the association between long sleep and PWV was stronger than that for short sleep and PWV, both short and long sleep duration should be targeted to mitigate the effects of insufficient or too much sleep respectively on CVD risk accrual. Future considerations should address gaps in the literature surrounding interactions between sleep duration and other less-investigated sleep-related factors that may be particular to at-risk groups, such as sleep quality, and subclinical CVD for a comprehensive understanding of CVD risk.

Table 4.

Summarized results.

What did we know prior to this study?

• Short sleep and long sleep are associated with subclinical arteriosclerosis and overt cardiovascular disease (CVD).

What didn’t we know prior to this study?

• The exact nature of the relationship between sleep duration and arteriosclerosis (measured by pulse wave velocity).

What does this study add?

• Evidence for the association of sleep duration and arterial stiffness, including moderator and sub-group analysis.

• Short sleep (<7 h): WMD = 20.6 cm/s, 95% Confidence intervals (CI): 13.8–27.4 cm/s, SMD = 0.02

• Long sleep (>9 h): WMD = 33.6 cm/s, 95%CI: 20.0–47.2 cm/s, SMD = 0.79.

How do we use this new information?

• This study provides a higher level of evidence of the association of subclinical CVD and short or long sleep.

• This information supports public health policy focused on healthy sleep guidelines for regularly short and long sleep, including for populations at high risk for CVD.

What needs to happen next to move the field forward?

• Longitudinal studies are needed to assess any causal relationship between sleep duration and arterial stiffness.

• Investigation into the role of other aspects of sleep, such as sleep quality, on CVD, which could influence the relationship between sleep duration and CVD.

5. Conclusions

Relative to recommended sleep duration, short and long sleep duration are associated with higher PWV, indicating that arterial stiffness could contribute to the association between sleep duration and CVD. The association between short sleep and higher PWV primarily comes from adults who are younger than 65 years of age or have a CMD, and the association of long sleep with higher PWV is pervasive across all subgroups of adults. Our recommendation is for all adults to adhere to current public health sleep duration recommendations. Furthermore, young adults with short sleep duration should specifically targeted by public health recommendations related to sleep duration to reduce subclinical CVD.

Practice points.

This meta-analysis gives public health policymakers and clinicians evidence for.

1. Increase in a marker of cardiovascular disease risk with short (<7 h) and long (>9 h) sleep duration in a general population.

2. Adults younger than 65 years old, with and without cardiometabolic disease, as sub-groups particularly affected by short sleep duration.

3. Long sleep duration as a risk for all adults.

Research agenda.

Future research may be able to explore longitudinal relationships between sleep duration and arterial stiffness. Furthermore, the influence of additional factors, such as sleep quality or sleep timing in the 24-h day, can be explored as they relate to sleep duration and arterial stiffness.

Glossary of terms

Arterial Stiffness

Fundamental mechanical behavior or rigidity of the material properties of the artery wall, determined by both structural and functional components

Cardiovascular Disease

A general term for pathologies affecting the heart or blood vessels, including cerebrovascular disease, hypertension, myocardial infarction, coronary heart disease, peripheral artery disease

Cardiometabolic Disease

Includes all cardiovascular diseases (above) and metabolic diseases (e.g. diabetes and pre-diabetes, metabolic syndrome)

Abbreviation

aPWV

Aortic pulse wave velocity

baPWV

Brachial-ankle pulse wave velocity

cfPWV

Carotid-femoral pulse wave velocity

CMD

Cardiometabolic disease

CVD

Cardiovascular disease

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PWV

Pulse wave velocity

SMD

Standardized mean difference

U.S.

United States of America

WMD

Weighted mean difference