The relationship between duration and quality of sleep and upper respiratory tract infections: a systematic review

Abstract Background Upper respiratory tract infections (URTIs) are common, mostly self-limiting, but result in inappropriate antibiotic prescriptions. Poor sleep is cited as a factor predisposing to URTIs, but the evidence is unclear. Objective To systematically review whether sleep duration and quality influence the frequency and duration of URTIs. Methods Three databases and bibliographies of included papers were searched for studies assessing associations between sleep duration or quality and URTIs. We performed dual title and abstract selection, discussed full-text exclusion decisions and completed 50% of data extraction in duplicate. The Newcastle–Ottawa Quality Assessment Scale assessed study quality and we estimated odds ratios (ORs) using random effects meta-analysis. Results Searches identified 5146 papers. Eleven met inclusion criteria, with nine included in meta-analyses: four good, two fair and five poor for risk of bias. Compared to study defined ‘normal’ sleep duration, shorter sleep was associated with increased URTIs (OR: 1.30, 95% confidence interval [CI]: 1.19–1.42, I2: 11%, P < 0.001) and longer sleep was not significantly associated (OR: 1.11 95% CI: 0.99–1.23, I2: 0%, P = 0.070). Sensitivity analyses using a 7- to 9-hour baseline found that sleeping shorter than 7–9 hours was associated with increased URTIs (OR: 1.31, 95% CI: 1.22–1.41, I2: 0%, P < 0.001). Sleeping longer than 7–9 hours was non-significantly associated with increased URTIs (OR: 1.15, 95% CI: 1.00–1.33, I2: 0%, P = 0.050, respectively). We were unable to pool sleep quality studies. No studies reported on sleep duration and URTI severity or duration. Conclusions Reduced sleep, particularly shorter than 7–9 hours, is associated with increased URTIs. Strategies improving sleep should be explored to prevent URTIs.


Introduction
Upper respiratory tract infections (URTIs) are typically viral infections (1) of the URT, including the nose, sinuses, pharynx and larynx. Adults experience ~two to three URTIs per year (2), and with the total direct and indirect cost of URTIs on the UK economy surpassing £76 million (3), URTIs place a large burden on the economy and medical services. Due to the issue of growing antibiotic resistance, in 2019 the UK government set a goal to reduce UK antibiotic use in humans by 15% by 2024 (4). A recent study assessed the contribution of URTIs to primary care antibiotic prescribing rates in England, using data recorded between 2013 and 2015 from UK primary care records. It was found that of the 69% of antibiotic prescriptions linked to a clinical condition or body system, 10.4% were for cough symptoms, 7.68% for a sore throat and 6.67% for URTIs (5). URTIs are mostly self-limiting, but are significant because they are common, and for their impact on antibiotic prescribing.
'Poor sleep' is commonly believed to increase susceptibility to infection. UK government advice states poor sleep and catching a cold or the flu could be related and recommends most individuals need 8 hours sleep a night (6). The National Sleep Foundation suggests normal sleep for 18-to 64-year olds is 7-9 hours, but acknowledges 6-11 hours may be appropriate for some 18-to 25-year olds, and 6-10 hours for some 26-to 64-year olds (7). The impact of sleep on immunity has been systematically reviewed: studying sleep in laboratory settings showed sleep deprivation produces a diminished cytokine response to lipopolysaccharide, a component of gram-negative bacteria (8). One study found that sleep deprivation reduces the efficacy of the hepatitis A vaccine (9).
Studies on sleep and URTI occurrence have conflicting results. Viral challenge studies showed that short sleep and sleep disturbance are associated with increased URTIs (10)(11)(12), yet a study in Sweden found that sleep duration and quality were not associated with increased URTIs (13). We therefore aimed to bring together the entirety of the clinical evidence in the first systematic review of sleep and URTIs.

Registration
A protocol was prospectively registered in PROSPERO (CRD42018097466).

Outcomes
The primary outcome was symptomatic URTI, expressed as the rate of occurrence of URTI in participants, or the proportion of patients who experienced ≥1 URTIs. The secondary outcomes were the severity and duration of the URTIs, and all clinically relevant outcomes reported.

Exposures
We compared the study defined 'normal sleep' duration with longer or shorter durations. 'Short' sleep was defined as sleep durations lower than study defined 'normal', and 'long' sleep was defined as sleep durations higher than study defined 'normal'. To overcome study variability in defining 'normal sleep duration', we pre-planned two sensitivity analyses using 7-8 and 7-9 hours of sleep as the reference group, based on the National Sleep Foundation's (7) definition of 'normal sleep' for adults aged ≥65 years and 18-65 years, respectively.

Eligibility criteria
Studies were eligible for inclusion if they examined the association between sleep quality or duration and URTIs. We included adults, aged ≥18 or as the study defined, of any sex, in any setting. We included studies that diagnosed URTIs via clinician assessment, laboratory techniques or self-report. We included studies measuring sleep subjectively or objectively. There were no restrictions on language or year of publication. For interventional studies involving infection with viruses, a minimum follow-up period of 5 days was pre-specified, as the common cold incubation period is 12 hours to 5 days (14). Exclusion criteria: (i) studies looking solely at populations with sleep or chronic disorders, (ii) studies looking solely at children, (iii) studies where sleep duration, sleep quality and number of URTIs were measured but an effect could not be calculated, (iv) patient follow-up rate below 80%, (v) protocolonly publications or (vi) case series and case reports.

Key Messages
• This is the first systematic review of sleep duration and quality on upper respiratory tract infections (URTIs).
• We included nine studies in meta-analyses out of 5146 titles. • Sleeping less than study defined 'normal' is associated with increased URTIs. • Sleeping for less than 7-9 hours is associated with increased URTIs. • Sleeping for more than 7-9 hours is associated with increased URTIs. • Data from studies on sleep quality and URTIs are lacking.
lists of included articles were reviewed for extra citations. The search strategy (Supplementary Table S1) was developed in consultation with an information specialist (NR). Search terms were reviewed by a sleep researcher, Nick Meyer, at King's College London; the URTI terms were developed from those used in a study by Merlin Wilcox, Clinical Lecturer at the University of Southampton, and reviewed by Oliver van Hecke, Clinical Lecturer at the University of Oxford; and the patient and public involvement group for the Nuffield Department of Primary Care Health Sciences, to ensure all possible search terms were included. The database searches were updated to include any studies published between 1 January 2018 and 10 January 2020.

Study selection, data collection and quality assessment
Titles and abstracts from the first database searches on 31 May 2018 were screened for eligibility by two reviewers (CR and IJ) independently using Rayaan software (15). Two reviewers (CR and JL) double screened the studies' titles and abstracts from the database searches that ran on 10 January 2020. Discrepancies were resolved by discussion. Duplicates were removed (CR). Studies were imported into Mendeley version 1.19.3, a reference manager. CR led a full-text review and discussed exclusion decisions with the study team (CA and JL). CR extracted relevant data. The first five papers were extracted and quality assessed in duplicate (CR and IJ) for accuracy. Disagreements were resolved by discussion. The Newcastle-Ottawa Quality Assessment Scale (NOS) (16) was used to assess study bias of cross-sectional and cohort studies. The results from the NOS were converted into the Agency for Healthcare Research and Quality standards of 'good', 'fair' and 'poor' quality. As no explicit conversion guidance exists, conversion thresholds from a prior publication were used (17).

Statistical analysis and reporting
Data analysis followed the Cochrane Handbook for systematic reviews of interventions (18). Random effects meta-analysis was performed where possible; if not, a narrative synthesis of included studies was performed. We estimated pooled odds ratios (ORs) and 95% confidence intervals (CIs) for URTI occurrence in each sleep duration group. The I 2 statistic assessed statistical heterogeneity, which was explored with sensitivity analyses where appropriate. Funnel plots and Egger's tests, if appropriate, were planned to assess publication bias if appropriate. We conducted sensitivity analyses using 7-8 and 7-9 hours of sleep per night as the reference group. We assessed the impact of studies with a high risk of bias in sensitivity analyses excluding their data. Analysis was conducted using Review Manager 5.3 software (19). We used the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) checklist (20).

Study selection and characteristics
The searches ran on 31 May 2018 found 5146 studies (Fig. 1). We removed 1699 duplicates. Citation searching gave one extra study. Title and abstract screening removed 3392 studies, full-text review excluded 45 studies, resulting in 11 eligible studies (10-13,21-27), with nine included in the meta-analysis (10)(11)(12)(13)(23)(24)(25)(26)(27). Two eligible studies (21,22) were not included in the meta-analysis, as they did not report sufficient quantitative information. One study (15) included in the meta-analysis presented pooled data from three similar studies from the same research group: Prather 2017a was referred to as 'The Pittsburgh Cold Study 2' (PCS2), Prather 2017b as 'The Pittsburgh Cold Study 3' (PCS3) and Prather 2017c as 'The Pittsburgh Mind-Body Center Study' (PMBC) in the Prather 2017 paper (11). Prather 2015 (10) and Cohen 2009 (12) report on the same participants as in Prather 2017b and Prather 2017c, respectively. They were included in this review as they presented additional data that was not reported in the Prather 2017 data obtained from the authors. Statistical analyses include data from Prather 2017 or one or both of Prather 2015 and Cohen 2009, to prevent double inclusion of participants. Table 1 presents the characteristics of the included studies in the qualitative synthesis. The papers present results from 66 229 patients, across five countries. Three cross-sectional studies (23)(24)(25) and six cohort studies (10)(11)(12)(13)26,27) were included in the meta-analysis. There were no randomized controlled trials. The database searches to 10 January 2020 found 946 studies. Two hundred fifty-eight duplicates were removed. Title and abstract screening removed 688 studies and no additional studies were included in the review from this update.

Risk of bias within studies
Tables 2 and 3 present the risk of bias assessment for included cross-sectional and cohort studies. No study received the highest possible mark of all nine stars: each was judged to have a risk of bias or lack of clarity in methodological reporting. Results of the NOS were converted to AHRQ standards: four studies were good (10)(11)(12)22), two were fair (23,25) and five were poor (13,21,24,26,27). Only one study (23) was truly representative of the general population; the other studies looked at select groups such as military recruits, mothers of young children, university students or volunteers responding to advertisements. All studies addressed confounders, with seven studies (10)(11)(12)(13)22,25,27) addressing what we considered most important: being chronically ill or immunocompromised.

Sleep quality
Eight studies (10)(11)(12)(13)(22)(23)(24)26) measured sleep quality, but data was not used in the meta-analysis because measures were too heterogeneous. Four studies (12,(22)(23)(24) reported associations between poor sleep quality and URTIs and four (10,11,13,26) found no association. Sleep quality was measured in 11 different ways (Supplementary  Table S2). Self-reported sleep efficiency (the proportion of time in bed spent sleeping described as a percentage) was measured in four studies (10)(11)(12)22). Two studies found associations (12,22) between lower self-reported sleep efficiency and increased risk of developing colds, and two found no associations (10,11). Meta-analysis was not possible to investigate sleep efficiency and URTI occurrence, as the sleep efficiency groups across studies were too heterogeneous to be pooled. 'Abnormal' sleep efficiency was the only sleep efficiency group reported in more than one study. Data from the two studies (12,22) reporting 'abnormal' sleep efficiency could not be combined as we considered these individuals likely had disorders beyond sleep efficiency, as their reported sleep efficiencies were much lower than the reported mean, and so were outside the realms of 'normality'. Subjective sleep quality was measured in three studies: one measured it on a 0-3 scale (11), one study (13) measured it as one of two categories (quite good/good, or neither bad nor good/quite bad/ bad) and one study (22) did not report its measurement. None of the studies found an association between subjective sleep quality and cold/URTI risk. The additional sleep quality measures reported in papers were unique to their study. One study (10) measured sleep quality objectively using wrist actigraphy and did not show an association between sleep quality and URTIs.

Secondary outcomes
The severity or duration of URTIs was measured in three studies (21,24,27), but sufficient information was not reported to calculate their association with sleep. No other clinically relevant outcomes were reported.

Risk of bias across studies
None of the meta-analyses included 10 or more studies; therefore, we could not reliably assess publication bias using funnel plots or Egger's tests (18).
We performed two pre-specified sensitivity analyses removing four studies assessed as 'poor' for risk of bias from the metaanalyses calculated using OR and CIs. Removing these studies (13,24,26,27) from the meta-analysis with 'shorter sleep' as the The study presented data on respiratory tract infections including head and chest colds, influenza, pneumonia and ear infections. Our review only included the head and chest colds data to ensure we met our URTI inclusion criteria.  comparator caused I 2 to drop from 11% to zero and changed the OR from 1.30 (1.19-1.42, P < 0.00001) to 1.33 (1.25-1.42, P < 0.00001). Removing poor quality studies (13,26,27) from the meta-analysis with 'longer sleep' as the comparator, kept I 2 at 0% and changed the estimates from 1.11 (0.99-1.16, P = 0.07000) to 1.11 (0.98-1.25, P = 0.09000).

Main findings
The study defined 'short sleep' was associated with increased URTIs, whereas 'long sleep' was not when comparing against the study defined 'normal sleep'. From the sensitivity analyses using 7-to 8-hour  and 7-to 9-hour reference groups, our findings suggest that sleeping for shorter than 7-9 hours per night could increase the occurrence of URTIs. Sleeping longer than 7-9 hours was non-significantly associated with increased URTIs (P = 0.050). The 7-to 8-hour and 7-to 9-hour reference group sensitivity analyses were calculated by pooling two and three studies, respectively, with one study in each analysis significantly dominating the weighting. The quality of studies was mixed, with only one study (23) awarded a star for being truly representative of the general population. Additional sensitivity analyses supported conclusions from meta-analyses. There is little evidence addressing the association between sleep quality and URTI occurrence, and we did not meta-analyse it due to variable measurements of sleep quality. Apart from sleep efficiency, subjective sleep quality was the only other quality assessment measured across studies (11,13,22) and meta-analysis was not possible. No significant association was found between subjective sleep quality and cold/URTI risk, suggesting subjective sleep quality does not influence URTI occurrence. No studies directly investigated the relationship between sleep duration and URTI severity or duration.

Strengths and limitations
The strengths of this review are that we brought together all published findings in a systematic review, following PRISMA guidelines, and our protocol was published prospectively on PROSPERO. The comprehensive search strategy was developed in consultation with an information specialist and subject experts, so is unlikely to have missed studies that would change the results. To our knowledge, this is the first systematic review looking at the effect of sleep on URTI occurrence and it includes all the available evidence. The included studies present data from 66 229 people across five countries, so results are broadly applicable.
The systematic review was limited by the available evidence. Quality was variable and the small number of included studies meant assessment of publication bias as planned was not possible. The lack of a clearly defined 'normal' sleep duration and quality led to variability in measurements reported, limiting meta-analysis. We were able to address this with sensitivity analyses. The association between sleep quality and URTI occurrence could not be quantitatively assessed, as outcome measures were too heterogeneous. Included studies had very limited data on secondary outcomes, so an effect could not be calculated. Many studies included self-diagnoses of URTIs. This is a limitation, but self-diagnosis of colds is usually accurate (28) and falsepositive influenza reports are rare (29), so subjective outcome measurement is likely to be adequate. The outcome measured was ≥1 URTIs which is dependent on patient follow-up. All the included cohort studies had a follow-up length of 13 weeks or shorter; with the exception of one study (26) which did not report patient follow-up length and one study (13) with a follow-up length of 9 months; however, neither study was included in the sensitivity analyses with 7-8 hours and 7-9 hours sleep as reference. The different follow-up times could have increased heterogeneity of results, but heterogeneity was low. To check we performed a post hoc meta-regression of effect sizes against follow-up time for seven studies examining short sleep (11,13,23,24,27). There was no association between effect size and follow-up. We were unable to perform a similar analysis for studies of long sleep as there were only three studies (13,23,27).

Comparison with existing literature and implications for practise
This is the first systematic review examining sleep quality and duration and URTI occurrence. It has been previously established that sleep has a regulatory role on the immune system (30,31). Immune parameters in human blood show systematic fluctuations; the influence of sleep on these temporal changes has been separated from those of circadian processes in two studies (30,32). Studies show that sleep deprivation activates the hypothalamus-pituitary-adrenal axis and sympathetic nervous system (33), which results in diminished immune response: reduced T-cell proliferation (34,35), T-helper cell 1 cytokine production (34,36) and natural killer cell cytotoxicity (37). A systematic review and meta-analysis of cohort and experimental sleep deprivation studies found that sleep disturbance and long sleep duration are associated with increased systemic inflammatory markers (interleukin-6 and C-reactive protein) (38). However, whether sleep's role on the immune system influences URTI occurrence had not been systematically reviewed. Despite these relationships between sleep and markers of the immune system, the mechanism through which sleep duration may influence URTI occurrence is unknown. One study in our systematic review (11) investigated whether nasal inflammation was a plausible pathway through which sleep influences cold occurrence: the data suggested that nasal cytokines and inflammation do not play a significant role. We found less evidence as to whether longer sleep influences infections. One study with over 56 000 people found that self-reported sleep ≥9 hours increased the risk of pneumonia (39), supporting a possible relationship between long sleep and URTI occurrence.
The direct clinical application of our findings is limited, but they can inform discussions about sleep between patients and their primary care clinicians, and may help facilitate discussions around the broader health implications of short sleep. The commonly held belief that short sleep is associated with URTIs is supported by our review, which is in line with the consensus statement from the American Academy of Sleep Medicine and Sleep Research Society that 7-9 hours of sleep is important for other health conditions (40). URTIs may represent a common opportunity for these discussions, particularly as there is evidence that family practitioners may be more likely to include health-promotion messages in a consultation when they have immediate relevance to their presenting complaint (41).

Implications for future research
Our review has identified gaps in the evidence base and should prompt examination of sleep association and causality in the occurrence of more clinically serious infections. Future research should explore the role of longer and poor quality sleep on respiratory infections and should use objective measures of sleep quality and duration. For example, studies should consider using the Pittsburgh Sleep Quality Index (42), which was used by only two studies in this review (11,12). Randomized trials of sleep-improvement interventions for the prevention of URTIs could support or provide more evidence for a causal link and inform clinical practice. These could address both prevention and treatment of infections. Future studies would also benefit from expanding outcomes measured to include URTI duration and severity.

Conclusions
Our findings suggest that sleeping for shorter than 7-9 hours per night could increase URTI occurrence. Sleeping longer than 7-9 hours was non-significantly associated (P = 0.05). This review will inform discussions with patients in primary care around sleep and should prompt further research on the broader health implications of short sleep, in particular the association in occurrence of more serious infections, such as SARS-CoV-2 or pneumonia.

Supplementary material
Supplementary material is available at Family Practice online.