Sleep Medicine 16 (2015) 811–819
Contents lists available at ScienceDirect
Sleep Medicine j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / s l e e p
Review Article
Daytime napping and mortality from all causes, cardiovascular disease, and cancer: a meta-analysis of prospective cohort studies Guochao Zhong a, Yi Wang b, TieHong Tao c, Jun Ying d, Yong Zhao e,* a
Department of Hepatobiliary Surgery, The Second Affiliated Hospital of Chongqing Medical University, Chongqing, China Department of Respiratory Medicine, The Second Affiliated Hospital of Chongqing Medical University, Chongqing, China Department of Cardiothoracic Surgery, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China d Department of Environmental Health, University of Cincinnati College of Medicine, Cincinnati, OH, USA e School of Public Health and Management, Chongqing Medical University, Chongqing, China b c
A R T I C L E
I N F O
Article history: Received 20 October 2014 Received in revised form 5 January 2015 Accepted 7 January 2015 Available online 21 April 2015 Keywords: Napping Death Mortality Cancer Cardiovascular disease Meta-analysis
A B S T R A C T
Objectives: The association between daytime napping and mortality remains controversial. We conducted a meta-analysis to examine the associations between daytime napping and the risks of death from all causes, cardiovascular disease (CVD), and cancer. Methods: PubMed and Embase databases were searched through 19 September 2014. Prospective cohort studies that provided risk estimates of daytime napping and mortality were eligible for our metaanalysis. Two investigators independently performed study screening and data extraction. A randomeffects model was used to estimate the combined effect size. Subgroup analyses were conducted to identify potential effect modifiers. Results: Twelve studies, involving 130,068 subjects, 49,791 nappers, and 19,059 deaths, were included. Our meta-analysis showed that daytime napping was associated with an increased risk of death from all causes [n = 9 studies; hazard ratio (HR), 1.22; 95% confidence interval (CI), 1.14–1.31; I2 = 42.5%]. No significant associations between daytime napping and the risks of death from CVD (n = 6 studies; HR, 1.20; 95% CI, 0.96–1.50; I2 = 75.0%) and cancer (n = 4 studies; HR, 1.07; 95% CI, 0.99–1.15; I2 = 8.9%) were found. There were no significant differences in risks of all-cause and CVD mortality between subgroups stratified by the prevalence of napping, follow-up duration, outcome assessment, age, and sex. Conclusions: Daytime napping is a predictor of increased all-cause mortality but not of CVD and cancer mortality. However, our findings should be treated with caution because of limited numbers of included studies and potential biases. © 2015 Published by Elsevier B.V.
1. Introduction Daytime napping is a common practice in many countries, especially in those where daytime napping is a part of the cultural norm, such as China, Latin American, and Mediterranean countries. In Latin American countries, the estimated prevalence of daytime napping is 29.2% [1] among adults aged ≥40 years. In China, the prevalence reaches as high as 68.6% among adults ≥45 years of age [2]. The frequency of napping usually increases with age [3–6], and it is higher in men than that in women [6,7]. Traditionally, daytime napping is usually considered a healthy habit, and it is often linked
Abbreviations: CI, confidence interval; CVD, cardiovascular disease; EPIC, European Prospective Investigation into Cancer; HR, hazard ratios. * Corresponding author. School of Public Health and Management, Chongqing Medical University, No.1, Yixueyuan Road, Yuanjiagang, Chongqing 400016, China. Tel.: +86 18723147857; fax: 86 023 68485008. E-mail address: [email protected] (Y. Zhao). http://dx.doi.org/10.1016/j.sleep.2015.01.025 1389-9457/© 2015 Published by Elsevier B.V.
with the low incidence of coronary heart disease and high tendency of longevity [8] through a hypothetical “stress relief” mechanism [9] in Latin American and Mediterranean countries. The Chinese believe that daytime napping is complementary to nighttime sleep, and hence a beneficial behavior. Despite implications from these intuitive perceptions, the associations between daytime napping and later health outcomes are poorly understood, and whether daytime napping is beneficial, neutral, or deleterious is still in debate. Recently, daytime napping has been linked with a series of adverse health outcomes, including nonalcoholic fatty liver disease [10], diabetes mellitus [2,11], breast cancer [12], and mortality [9,13]. The napping–mortality association has received significant attention, but it remains controversial. Several studies identified that daytime nappers were at an elevated risk of all-cause mortality [9,13,14], whereas others failed to observe a positive association between daytime napping and all-cause mortality [15,16]. A definite demonstration of the napping–mortality association is crucial to understand the effects of napping on health outcomes. However, to the best of our knowledge, a meta-analysis on this topic is not
812
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
available at present. Therefore, we conducted a meta-analysis to examine the associations of daytime napping with risks of death from all causes, cardiovascular disease (CVD), and cancer. Moreover, we investigated underlying effect modifiers of the napping– mortality association where possible. 2. Methods 2.1. Search strategy The study was performed in adherence to the Preferred Reporting Items for Systemic Reviews and Meta-Analysis statement [17]. We performed a systematic electronic search of PubMed and Embase databases through 19 September 2014, with no restriction. The following search terms were used: “siesta,” “napping,” “nap,” “mortality,” and “death.” To include extra citations, we also checked the reference lists of pertinent articles manually. We did not attempt to contact original authors for extra data parameters through e-mails.
and I2 < 50.0%, low heterogeneity. To examine the stability of pooled results and possible explanations of heterogeneity, we performed sensitivity analyses using the following methods: omitting a single study in turn, repeating analyses via a fixed-effects model, and using various eligibility criteria. To investigate potential effect modifiers, we performed subgroup analyses stratified by countries with prevalent napping, length of follow-up, outcome assessment, sex, and age. A P-value for interaction between subgroups was calculated through meta-regression. We did not perform sensitivity and subgroup analyses for the association between daytime napping and cancer mortality because of few included studies. For studies that reported HRs separately for men and women [8,16,23–25], or for different napping duration ranges [9,16,19], we employed a randomeffects model to yield overall estimates for our meta-analysis. Publication bias was assessed by the Begg rank correlation test [26] and the Egger linear regression test [27]. Data synthesis and analysis were performed via STATA software (version12.0, StataCorp, College Station, TX, USA). The statistical significance level was set at P < 0.05 under a two-sided test unless otherwise specified.
2.2. Study selection Studies were included if they (1) had a prospective cohort study design and were published in English or Chinese; and (2) provided adjusted risk estimates of daytime napping and mortality from all causes, CVD, or cancer. Two investigators (G.C.Z. and Y.W.) independently performed study screening. We first excluded obviously irrelevant studies based on the titles and abstracts, and then we excluded unrelated studies by reviewing the full text. Any discrepancy regarding eligibility of studies was resolved by consensus. 2.3. Data extraction The following information was extracted from the included studies: first author, publication date, study location, mean age of participants at baseline, mean follow-up duration, number of deaths, comparison of napping, number of nappers, sample size, exposure assessment, outcome assessment, the most fully adjusted risk estimates with corresponding 95% confidence interval (CI), and adjustment variables. Two investigators (G.C.Z. and Y.W.) independently conducted data extraction. The final results were compared, and any disagreement was resolved by discussion. 2.4. Quality assessment Two investigators (G.C.Z. and Y.W.) independently assessed the quality of included studies on the basis of the Newcastle–Ottawa quality assessment scale [18]. This tool is composed of eight items, but only seven of them can be applicable to the current metaanalysis. After judging three domains (selection, comparability, and outcome), a maximum of eight stars (three stars for selection, two stars for comparability, and three stars for outcome) could be awarded to each study. We described the quality of included studies as high (six to eight stars), moderate (four to five stars), and low (zero to three stars). Any discrepancy was resolved by discussion. 2.5. Statistical analysis We employed a random-effects model to estimate the combined effect size. Hazard ratio (HR) was used as a common measure of the napping–mortality relationship. Risk odds ratios [19,20] were directly regarded as HRs. Our meta-analysis used data parameters from the report with the longest follow-up duration when multiple reports [19,21] originated from the same population. The I2 statistic was used to quantify the between-study heterogeneity [22]. We interpreted I2 statistic as the following criteria: I2 > 75.0%, substantial heterogeneity; 50.0% ≤ I2 ≤ 75.0%, moderate heterogeneity;
3. Results 3.1. Literature search We identified 268 relevant articles after removing duplicates. A total of 249 articles were further excluded after reading titles and abstracts. The remaining 19 articles were assessed in more detail for eligibility by reviewing the full text. Of these, seven were excluded. Three studies [19–21] were derived from the same study population, but two of them provided different end points (one [21] provided all-cause mortality, another [20] provided CVD and cancer mortality); the remaining one [19] provided additional information for subgroup analysis stratified by sex. Thus, 12 studies were included in our review (Fig. 1). 3.2. Study characteristics and quality assessment The main characteristics of the included studies are summarized in Table 1. Our study involved 130,068 individuals, consisting of 51,341 men (39.5%) and 78,727 women (60.5%). Of these, 49,791 subjects reported daytime napping, and the prevalence of daytime napping ranged from 10.8% [14] to 70.9% [25] across studies. During the follow-up period ranging from 4.0 [28] to 19.0 years [24], a total of 19,059 deaths occurred. All except three studies [9,13,25] recruited individuals aged ≥65 years as their participants. Eleven studies recruited both men and women, whereas one study [14] recruited only women. The most frequent comparison of napping was daily versus never. Only one study [16] adopted face-to-face interview to assess exposure, and all of the remaining studies adopted questionnaire. Half of the included studies [9,13,14,24,25,28] relied on death certificates for the ascertainment of causes of death, and another half [8,16,19–21,23] relied on death registry. Of the included studies, nine studies reported HRs for all-cause mortality, six for CVD mortality, and only four for cancer mortality. The quality of included studies was generally good, with 10 studies showing high quality (Table 2). Three exposed groups [14,23,25] had poor representativeness. Nevertheless, all nonexposed groups were drawn from the same population as exposed groups. Five studies [9,13,14,23,24] did not use a secure record or a structured interview to ascertain napping behavior. All included studies controlled the most important confounder (age or sex) for the napping– mortality association, and six [8,9,13,20,21,24] of them further controlled a second important confounder (sleep duration). Five included studies [8,9,13,21,24] had adequate follow-up duration (≥10 years), and the loss to follow-up rate was 0.05). Likewise, all stratified factors did not modify the relationship between daytime napping and CVD mortality (all P for interaction >0.05). 3.7. Sensitivity analyses The results of sensitivity analyses are summarized in Table 4. Sensitivity analyses that ignored a single study did not materially alter the initial association between daytime napping and all-cause mortality. In addition, changing eligibility criteria and repeating the
analysis via a fixed-effects model did not alter the association between daytime napping and all-cause mortality, either. The pooled HR of CVD mortality ranged from 1.11 (95% CI, 0.85– 1.45) [14] to 1.29 (95% CI, 1.05–1.57) [25] when we omitted a single study in turn, and we observed a significant combined result (HR, 1.27; 95% CI, 1.19–1.36) when repeating the analysis via a fixedeffects model. The exclusion of two studies [9,25] being components of European Prospective Investigation into Cancer (EPIC) [29] resulted in a higher risk (HR, 1.40; 95% CI, 1.19–1.65), with low heterogeneity (I2 = 33.2%). The exclusion of three studies [9,13,25] with a large sample size (>10,000) also produced a higher risk (HR, 1.55; 95% CI, 1.19–2.02), with low heterogeneity (I2 = 10.7%).
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
815
Table 2 The results of quality assessment. Source (reference)
Selection Representativeness of exposed group ☆
Selection of nonexposed group ☆
Leng [9] Lee [23] Jung [24] Tanabe [13] Stone [14] Naska [25] Lan [16] Bursztyn [21] Burazeri [8] Bursztyn [19] Bursztyn [20] Hays [28]
☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆
☆ ☆ ☆ ☆ ☆ ☆
Comparability
Outcome
Exposure as certainment ☆
Comparable on confounders ☆☆
Outcome assessment ☆
Adequate follow-up (≥10 years) ☆
Loss to follow-up rate (≤20%) ☆
☆ ☆
☆
☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆ ☆ ☆ ☆
☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
Total score
☆ ☆
☆ ☆
7 5 6 7 4 6 6 7 7 6 6 7
Fig. 2. Meta-analysis for daytime napping and all-cause mortality.
3.8. Publication bias No evidence of publication bias was found for any association by Begg’s test and Egger’s test (all P > 0.05). 4. Discussion Findings from our meta-analysis showed that daytime napping was associated with an increased risk of all-cause mortality but not of CVD and cancer mortality. On the basis of limited numbers of included studies, subgroup analyses identified that the prevalence of napping, follow-up duration, outcome assessment, age, and sex could not modify the associations between daytime napping and all-cause and CVD mortality. Recent studies indicated that both excessive and inadequate nighttime sleep durations were significant predictors of mortality [30–34]. Several studies supported an association between daytime napping and nighttime sleep duration [15,35–37]. Thus, it is possible that the interaction between daytime napping and nighttime
Fig. 3. Hazard ratios (HR) of all-cause mortality by napping duration ranges. Each marker (circle, square, diamond, triangle, and plus sign shape) represents the midpoint of the lower and upper boundaries of each napping duration range. The area of each marker is proportional to the inverse varice of log HR.
816
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
Fig. 4. Meta-analysis for daytime napping and cardiovascluar disease mortality.
Fig. 5. Meta-analysis for daytime napping and cancer mortality.
sleep duration occurs in the pathway to death. Of the included studies, only three [9,16,20] addressed this critical issue, and all of them observed that there was no significant difference in the risk of all-cause mortality for nappers across different nighttime sleep duration ranges. However, Cohen-Mansfield and colleagues [15] found that the association between short nighttime sleep duration and mortality changed from nonsignificant in the overall population to significant in the nappers. Taken together, we could not rule out the modification effect of nighttime sleep duration on our results. A U-shaped pattern is identified for the association between nighttime sleep duration and risk of death from all causes [32,33,38]. Generally, sleepers with a duration of 7–8 h have the lowest risk of death [33,38]. Likewise, we speculate that risks from daytime napping may dynamically change with napping duration in a certain pattern. In fact, findings from several observational studies supported our speculation. Cao and colleagues found that afternoon nappers with a longer duration experienced a higher risk of developing hypertension [39]. Meanwhile, Chen and colleagues also observed that daytime nappers with a longer duration tended to
present a higher risk of developing osteoporosis [40]. In this meta-analysis, five included studies [8,9,16,19,24] investigated the potential association between napping duration and all-cause mortality. However, they presented mixed results. The inconsistency among these results raised a concern with regard to the true shape of the association between napping duration and mortality. Previous studies [41] indicated that a 10-min daytime nap was the most effective choice of producing napping-associated benefits. Unfortunately, due to uncertainty about the shape of the association between napping duration and mortality, determining the napping duration that presents the lowest risk of death appears to be impossible at present. To clarify the shape of the association between napping duration and mortality, more studies are therefore warranted. Our results showed that the adverse effect of daytime napping was statistically significant for all-cause mortality but not for CVD and cancer mortality. A biologically possible explanation for this phenomenon is that increased all-cause mortality is attributable to increased non-CVD and noncancer mortality. Indeed, it is found that daytime napping is associated with increased respiratory disease
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
817
Table 3 Subgroup analyses of daytime napping and all-cause and cardiovascular mortality. Subgroup
All studies Countries with prevalent nappingc Yes No Length of follow-up ≥10 years
Contents lists available at ScienceDirect
Sleep Medicine j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / s l e e p
Review Article
Daytime napping and mortality from all causes, cardiovascular disease, and cancer: a meta-analysis of prospective cohort studies Guochao Zhong a, Yi Wang b, TieHong Tao c, Jun Ying d, Yong Zhao e,* a
Department of Hepatobiliary Surgery, The Second Affiliated Hospital of Chongqing Medical University, Chongqing, China Department of Respiratory Medicine, The Second Affiliated Hospital of Chongqing Medical University, Chongqing, China Department of Cardiothoracic Surgery, The First Affiliated Hospital of Chongqing Medical University, Chongqing, China d Department of Environmental Health, University of Cincinnati College of Medicine, Cincinnati, OH, USA e School of Public Health and Management, Chongqing Medical University, Chongqing, China b c
A R T I C L E
I N F O
Article history: Received 20 October 2014 Received in revised form 5 January 2015 Accepted 7 January 2015 Available online 21 April 2015 Keywords: Napping Death Mortality Cancer Cardiovascular disease Meta-analysis
A B S T R A C T
Objectives: The association between daytime napping and mortality remains controversial. We conducted a meta-analysis to examine the associations between daytime napping and the risks of death from all causes, cardiovascular disease (CVD), and cancer. Methods: PubMed and Embase databases were searched through 19 September 2014. Prospective cohort studies that provided risk estimates of daytime napping and mortality were eligible for our metaanalysis. Two investigators independently performed study screening and data extraction. A randomeffects model was used to estimate the combined effect size. Subgroup analyses were conducted to identify potential effect modifiers. Results: Twelve studies, involving 130,068 subjects, 49,791 nappers, and 19,059 deaths, were included. Our meta-analysis showed that daytime napping was associated with an increased risk of death from all causes [n = 9 studies; hazard ratio (HR), 1.22; 95% confidence interval (CI), 1.14–1.31; I2 = 42.5%]. No significant associations between daytime napping and the risks of death from CVD (n = 6 studies; HR, 1.20; 95% CI, 0.96–1.50; I2 = 75.0%) and cancer (n = 4 studies; HR, 1.07; 95% CI, 0.99–1.15; I2 = 8.9%) were found. There were no significant differences in risks of all-cause and CVD mortality between subgroups stratified by the prevalence of napping, follow-up duration, outcome assessment, age, and sex. Conclusions: Daytime napping is a predictor of increased all-cause mortality but not of CVD and cancer mortality. However, our findings should be treated with caution because of limited numbers of included studies and potential biases. © 2015 Published by Elsevier B.V.
1. Introduction Daytime napping is a common practice in many countries, especially in those where daytime napping is a part of the cultural norm, such as China, Latin American, and Mediterranean countries. In Latin American countries, the estimated prevalence of daytime napping is 29.2% [1] among adults aged ≥40 years. In China, the prevalence reaches as high as 68.6% among adults ≥45 years of age [2]. The frequency of napping usually increases with age [3–6], and it is higher in men than that in women [6,7]. Traditionally, daytime napping is usually considered a healthy habit, and it is often linked
Abbreviations: CI, confidence interval; CVD, cardiovascular disease; EPIC, European Prospective Investigation into Cancer; HR, hazard ratios. * Corresponding author. School of Public Health and Management, Chongqing Medical University, No.1, Yixueyuan Road, Yuanjiagang, Chongqing 400016, China. Tel.: +86 18723147857; fax: 86 023 68485008. E-mail address: [email protected] (Y. Zhao). http://dx.doi.org/10.1016/j.sleep.2015.01.025 1389-9457/© 2015 Published by Elsevier B.V.
with the low incidence of coronary heart disease and high tendency of longevity [8] through a hypothetical “stress relief” mechanism [9] in Latin American and Mediterranean countries. The Chinese believe that daytime napping is complementary to nighttime sleep, and hence a beneficial behavior. Despite implications from these intuitive perceptions, the associations between daytime napping and later health outcomes are poorly understood, and whether daytime napping is beneficial, neutral, or deleterious is still in debate. Recently, daytime napping has been linked with a series of adverse health outcomes, including nonalcoholic fatty liver disease [10], diabetes mellitus [2,11], breast cancer [12], and mortality [9,13]. The napping–mortality association has received significant attention, but it remains controversial. Several studies identified that daytime nappers were at an elevated risk of all-cause mortality [9,13,14], whereas others failed to observe a positive association between daytime napping and all-cause mortality [15,16]. A definite demonstration of the napping–mortality association is crucial to understand the effects of napping on health outcomes. However, to the best of our knowledge, a meta-analysis on this topic is not
812
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
available at present. Therefore, we conducted a meta-analysis to examine the associations of daytime napping with risks of death from all causes, cardiovascular disease (CVD), and cancer. Moreover, we investigated underlying effect modifiers of the napping– mortality association where possible. 2. Methods 2.1. Search strategy The study was performed in adherence to the Preferred Reporting Items for Systemic Reviews and Meta-Analysis statement [17]. We performed a systematic electronic search of PubMed and Embase databases through 19 September 2014, with no restriction. The following search terms were used: “siesta,” “napping,” “nap,” “mortality,” and “death.” To include extra citations, we also checked the reference lists of pertinent articles manually. We did not attempt to contact original authors for extra data parameters through e-mails.
and I2 < 50.0%, low heterogeneity. To examine the stability of pooled results and possible explanations of heterogeneity, we performed sensitivity analyses using the following methods: omitting a single study in turn, repeating analyses via a fixed-effects model, and using various eligibility criteria. To investigate potential effect modifiers, we performed subgroup analyses stratified by countries with prevalent napping, length of follow-up, outcome assessment, sex, and age. A P-value for interaction between subgroups was calculated through meta-regression. We did not perform sensitivity and subgroup analyses for the association between daytime napping and cancer mortality because of few included studies. For studies that reported HRs separately for men and women [8,16,23–25], or for different napping duration ranges [9,16,19], we employed a randomeffects model to yield overall estimates for our meta-analysis. Publication bias was assessed by the Begg rank correlation test [26] and the Egger linear regression test [27]. Data synthesis and analysis were performed via STATA software (version12.0, StataCorp, College Station, TX, USA). The statistical significance level was set at P < 0.05 under a two-sided test unless otherwise specified.
2.2. Study selection Studies were included if they (1) had a prospective cohort study design and were published in English or Chinese; and (2) provided adjusted risk estimates of daytime napping and mortality from all causes, CVD, or cancer. Two investigators (G.C.Z. and Y.W.) independently performed study screening. We first excluded obviously irrelevant studies based on the titles and abstracts, and then we excluded unrelated studies by reviewing the full text. Any discrepancy regarding eligibility of studies was resolved by consensus. 2.3. Data extraction The following information was extracted from the included studies: first author, publication date, study location, mean age of participants at baseline, mean follow-up duration, number of deaths, comparison of napping, number of nappers, sample size, exposure assessment, outcome assessment, the most fully adjusted risk estimates with corresponding 95% confidence interval (CI), and adjustment variables. Two investigators (G.C.Z. and Y.W.) independently conducted data extraction. The final results were compared, and any disagreement was resolved by discussion. 2.4. Quality assessment Two investigators (G.C.Z. and Y.W.) independently assessed the quality of included studies on the basis of the Newcastle–Ottawa quality assessment scale [18]. This tool is composed of eight items, but only seven of them can be applicable to the current metaanalysis. After judging three domains (selection, comparability, and outcome), a maximum of eight stars (three stars for selection, two stars for comparability, and three stars for outcome) could be awarded to each study. We described the quality of included studies as high (six to eight stars), moderate (four to five stars), and low (zero to three stars). Any discrepancy was resolved by discussion. 2.5. Statistical analysis We employed a random-effects model to estimate the combined effect size. Hazard ratio (HR) was used as a common measure of the napping–mortality relationship. Risk odds ratios [19,20] were directly regarded as HRs. Our meta-analysis used data parameters from the report with the longest follow-up duration when multiple reports [19,21] originated from the same population. The I2 statistic was used to quantify the between-study heterogeneity [22]. We interpreted I2 statistic as the following criteria: I2 > 75.0%, substantial heterogeneity; 50.0% ≤ I2 ≤ 75.0%, moderate heterogeneity;
3. Results 3.1. Literature search We identified 268 relevant articles after removing duplicates. A total of 249 articles were further excluded after reading titles and abstracts. The remaining 19 articles were assessed in more detail for eligibility by reviewing the full text. Of these, seven were excluded. Three studies [19–21] were derived from the same study population, but two of them provided different end points (one [21] provided all-cause mortality, another [20] provided CVD and cancer mortality); the remaining one [19] provided additional information for subgroup analysis stratified by sex. Thus, 12 studies were included in our review (Fig. 1). 3.2. Study characteristics and quality assessment The main characteristics of the included studies are summarized in Table 1. Our study involved 130,068 individuals, consisting of 51,341 men (39.5%) and 78,727 women (60.5%). Of these, 49,791 subjects reported daytime napping, and the prevalence of daytime napping ranged from 10.8% [14] to 70.9% [25] across studies. During the follow-up period ranging from 4.0 [28] to 19.0 years [24], a total of 19,059 deaths occurred. All except three studies [9,13,25] recruited individuals aged ≥65 years as their participants. Eleven studies recruited both men and women, whereas one study [14] recruited only women. The most frequent comparison of napping was daily versus never. Only one study [16] adopted face-to-face interview to assess exposure, and all of the remaining studies adopted questionnaire. Half of the included studies [9,13,14,24,25,28] relied on death certificates for the ascertainment of causes of death, and another half [8,16,19–21,23] relied on death registry. Of the included studies, nine studies reported HRs for all-cause mortality, six for CVD mortality, and only four for cancer mortality. The quality of included studies was generally good, with 10 studies showing high quality (Table 2). Three exposed groups [14,23,25] had poor representativeness. Nevertheless, all nonexposed groups were drawn from the same population as exposed groups. Five studies [9,13,14,23,24] did not use a secure record or a structured interview to ascertain napping behavior. All included studies controlled the most important confounder (age or sex) for the napping– mortality association, and six [8,9,13,20,21,24] of them further controlled a second important confounder (sleep duration). Five included studies [8,9,13,21,24] had adequate follow-up duration (≥10 years), and the loss to follow-up rate was 0.05). Likewise, all stratified factors did not modify the relationship between daytime napping and CVD mortality (all P for interaction >0.05). 3.7. Sensitivity analyses The results of sensitivity analyses are summarized in Table 4. Sensitivity analyses that ignored a single study did not materially alter the initial association between daytime napping and all-cause mortality. In addition, changing eligibility criteria and repeating the
analysis via a fixed-effects model did not alter the association between daytime napping and all-cause mortality, either. The pooled HR of CVD mortality ranged from 1.11 (95% CI, 0.85– 1.45) [14] to 1.29 (95% CI, 1.05–1.57) [25] when we omitted a single study in turn, and we observed a significant combined result (HR, 1.27; 95% CI, 1.19–1.36) when repeating the analysis via a fixedeffects model. The exclusion of two studies [9,25] being components of European Prospective Investigation into Cancer (EPIC) [29] resulted in a higher risk (HR, 1.40; 95% CI, 1.19–1.65), with low heterogeneity (I2 = 33.2%). The exclusion of three studies [9,13,25] with a large sample size (>10,000) also produced a higher risk (HR, 1.55; 95% CI, 1.19–2.02), with low heterogeneity (I2 = 10.7%).
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
815
Table 2 The results of quality assessment. Source (reference)
Selection Representativeness of exposed group ☆
Selection of nonexposed group ☆
Leng [9] Lee [23] Jung [24] Tanabe [13] Stone [14] Naska [25] Lan [16] Bursztyn [21] Burazeri [8] Bursztyn [19] Bursztyn [20] Hays [28]
☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆
☆ ☆ ☆ ☆ ☆ ☆
Comparability
Outcome
Exposure as certainment ☆
Comparable on confounders ☆☆
Outcome assessment ☆
Adequate follow-up (≥10 years) ☆
Loss to follow-up rate (≤20%) ☆
☆ ☆
☆
☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
☆ ☆ ☆ ☆ ☆
☆
☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆ ☆
Total score
☆ ☆
☆ ☆
7 5 6 7 4 6 6 7 7 6 6 7
Fig. 2. Meta-analysis for daytime napping and all-cause mortality.
3.8. Publication bias No evidence of publication bias was found for any association by Begg’s test and Egger’s test (all P > 0.05). 4. Discussion Findings from our meta-analysis showed that daytime napping was associated with an increased risk of all-cause mortality but not of CVD and cancer mortality. On the basis of limited numbers of included studies, subgroup analyses identified that the prevalence of napping, follow-up duration, outcome assessment, age, and sex could not modify the associations between daytime napping and all-cause and CVD mortality. Recent studies indicated that both excessive and inadequate nighttime sleep durations were significant predictors of mortality [30–34]. Several studies supported an association between daytime napping and nighttime sleep duration [15,35–37]. Thus, it is possible that the interaction between daytime napping and nighttime
Fig. 3. Hazard ratios (HR) of all-cause mortality by napping duration ranges. Each marker (circle, square, diamond, triangle, and plus sign shape) represents the midpoint of the lower and upper boundaries of each napping duration range. The area of each marker is proportional to the inverse varice of log HR.
816
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
Fig. 4. Meta-analysis for daytime napping and cardiovascluar disease mortality.
Fig. 5. Meta-analysis for daytime napping and cancer mortality.
sleep duration occurs in the pathway to death. Of the included studies, only three [9,16,20] addressed this critical issue, and all of them observed that there was no significant difference in the risk of all-cause mortality for nappers across different nighttime sleep duration ranges. However, Cohen-Mansfield and colleagues [15] found that the association between short nighttime sleep duration and mortality changed from nonsignificant in the overall population to significant in the nappers. Taken together, we could not rule out the modification effect of nighttime sleep duration on our results. A U-shaped pattern is identified for the association between nighttime sleep duration and risk of death from all causes [32,33,38]. Generally, sleepers with a duration of 7–8 h have the lowest risk of death [33,38]. Likewise, we speculate that risks from daytime napping may dynamically change with napping duration in a certain pattern. In fact, findings from several observational studies supported our speculation. Cao and colleagues found that afternoon nappers with a longer duration experienced a higher risk of developing hypertension [39]. Meanwhile, Chen and colleagues also observed that daytime nappers with a longer duration tended to
present a higher risk of developing osteoporosis [40]. In this meta-analysis, five included studies [8,9,16,19,24] investigated the potential association between napping duration and all-cause mortality. However, they presented mixed results. The inconsistency among these results raised a concern with regard to the true shape of the association between napping duration and mortality. Previous studies [41] indicated that a 10-min daytime nap was the most effective choice of producing napping-associated benefits. Unfortunately, due to uncertainty about the shape of the association between napping duration and mortality, determining the napping duration that presents the lowest risk of death appears to be impossible at present. To clarify the shape of the association between napping duration and mortality, more studies are therefore warranted. Our results showed that the adverse effect of daytime napping was statistically significant for all-cause mortality but not for CVD and cancer mortality. A biologically possible explanation for this phenomenon is that increased all-cause mortality is attributable to increased non-CVD and noncancer mortality. Indeed, it is found that daytime napping is associated with increased respiratory disease
G. Zhong et al./Sleep Medicine 16 (2015) 811–819
817
Table 3 Subgroup analyses of daytime napping and all-cause and cardiovascular mortality. Subgroup
All studies Countries with prevalent nappingc Yes No Length of follow-up ≥10 years
