Correlations in sleeping patterns and circadian preference between spouses

Spouses may affect each other’s sleeping behaviour. In 47,420 spouse-pairs from the UK Biobank, we found a weak positive phenotypic correlation between spouses for self-reported sleep duration (r = 0.11; 95% CI = 0.10, 0.12) and a weak inverse correlation for chronotype (diurnal preference) (r = −0.11; −0.12, −0.10), which replicated in up to 127,035 23andMe spouse-pairs. Using accelerometer data on 3454 UK Biobank spouse-pairs, the correlation for derived sleep duration was similar to self-report (r = 0.12; 0.09, 0.15). Timing of diurnal activity was positively correlated (r = 0.24; 0.21, 0.27) in contrast to the inverse correlation for chronotype. In Mendelian randomization analysis, positive effects of sleep duration (mean difference=0.13; 0.04, 0.23 SD per SD) and diurnal activity (0.49; 0.03, 0.94) were observed, as were inverse effects of chronotype (−0.15; −0.26, −0.04) and snoring (−0.15; −0.27, −0.04). Findings support the notion that an individual’s sleep may impact that of their partner, promoting opportunities for sleep interventions at the family-level.

In this manuscript, Richmond et al. report on their investigation of sleeping patterns and circadian preference between cohabiting spouses.They leveraged two very large datasets, from UK Biobank (47,420 couples) and 23andMe, Inc. (127,035 couples), and applied a very neat study design.First, they reported a weak positive correlation for sleep duration and a weak negative correlation for chronotype between spouses.These correlations were replicated in both UK Biobank and 23andMe.Then, they used accelerometer data in the UK Biobank and observed similar results to the self-report, with the time of diurnal activity showing a correlation.
The authors then extended the Mendelian randomisation design to investigate the causal effects of one individual on another, what they describe as social or indirect genetic effects, and tested 9 sleep traits.They also performed other sensitivity analyses to test the robustness of the MR, and assessed the role of age, time of wear and birth location.
This study is by far the largest of its kind, and represents an interesting and solid contribution to our understanding of how cohabitation influences sleep behaviour.It is very well designed and executed and written clearly, so it is easy for the reader to follow.It will make a great addition to the literature.
I have a few thoughts that the authors might want to consider, either as part of this paper or for their future work: -How does occupation and socioeconomic status affect the correlations?(e.g., if the couple members are retired or working) -How do lifestyle variables and substance use (alcohol and tobacco) patterns affect sleep behaviour correlations in spouses?-How do BMI and chronic disease status affect sleep behaviour in spouses?Also, the authors should discuss how their analysis are impacted by the fact that the previous GWAS they used to estimate polygenic risk scores had adjusted for sex differences.
Reviewer #2 (Remarks to the Author): Weak inverse effects for chronotype were found between spouses in the UK Biobank and 23andMe.One explanation could be that they can better handle the 24-h challenge of caring for toddlers.Do you see an association between the chronotype difference between spouses and the number of children?
Sleep in a relationship may be affected by others living in the same household (e.g., pets or children).Furthermore, imposed work schedules can affect the time when people go to sleep.Do these factors mediate or mitigate some of the tested associations?Do you know whether couples lived in an apartment building or a one-family house?Any differences in how concordantly spouses slept?Does it matter whether you live in an urban or rural area for how well spouses sleep concordantly?What about sleep associations among homosexual couples?Similar associations like those observed between heterosexual couples?Variance in sleep is not only a matter of genes; it highly depends on the state in which a person or a couple is.Thus, I'm curious whether sleep overlaps between spouses may be more or less pronounced in certain life situations, such as low or high socioeconomic deprivation, taking care of a child, or before and after retirement.Likewise, it may be interesting to see whether couples spending more time together or have similar activity patterns during wake time (which would mean that they may share similar wake experiences resulting in a comparable homeostatic sleep pressure) are more likely to sleep concordantly than those who live their "own" lives.
Reviewer #3 (Remarks to the Author): Thanks for inviting me for reviewing this paper.Richmond et al. conducted a Mendelian randomization (MR) study exploring correlations in sleeping patterns and circadian preference between spouses, and heir partner.Although the paper is well written, this is an obvious observation and the justification that this warrants an MR seems very weak.Public health implications of this study are also not clear and convincing (i.e., providing insights into human sleep interactions).Novelty is also a major concern.
I have some more specific comments, as below: 1.In introduction, the authors are suggested to provide stronger argument why determine the existence and directionality in correlated sleep patterns between spouses matters.Will the results have any possibility in changing the current or future clinical or public health practice.Or by clarifying this research question, can we gain better understanding about etiology?2. All sleep questions in UK Biobank are very simple and not validated.Also, no information regarding the reliability of these questions has been provided.3. The interpretation of correlations are not clear.Need to explain why not used beta coefficients, which is more commonly used in MR in this study.4. Interpretation of SEs in some tables (i.e., Table 2) is difficult because of the lack of units for some variables and reference groups. .

COMMSBIO-23-0514-T: Response to reviewers
Editor comments: Your manuscript en tled "Correla ons in sleeping pa erns and circadian preference between spouses" has now been seen by 3 referees, whose comments are appended below.You will see from their comments copied below that while they nd your work of poten al interest, they have raised quite substan al concerns that must be addressed.In light of these comments, we cannot accept the manuscript for publica on, but would be interested in considering a revised version that addresses these serious concerns.
We hope you will nd the referees' comments useful as you decide how to proceed.Should further experimental data or analysis allow you to address these cri cisms, we would be happy to look at a substan ally revised manuscript.However, please bear in mind that we will be reluctant to approach the referees again in the absence of major revisions.
In par cular, we ask that you inves gate the impact of other socioeconomic or lifestyle factors as outlined by Reviewers #1-2, and address the contextual and sta s cal concerns men oned by Reviewer #3.
Thank you for considering our work for publica on in Communica ons Biology and for permi ng us a generous amount of me to revise the manuscript and address the referee comments in full.We have now extended the e ect modica on analysis to consider other socio-economic and lifestyle factors which might a ect the rela onships between sleep traits among spouses, including an assessment of employment status, children living in household, type of household, socio-economic posi on, rural/urban living and day me ac vity ming, as suggested by Reviewers #1 and #2.As requested by Reviewer #3, we have also provided more ra onale for the study and poten al public health implica ons in the Background and Discussion sec ons of the manuscript, have claried the sta s cal queries and have made the sta s cal code publicly available via a GitHub repository (h ps://github.com/rcrichmond/spousal_sleep).

Reviewers' comments:
Reviewer #1 (Remarks to the Author): In this manuscript, Richmond et al. report on their inves ga on of sleeping pa erns and circadian preference between cohabi ng spouses.They leveraged two very large datasets, from UK Biobank (47,420 couples) and 23andMe, Inc. (127,035 couples), and applied a very neat study design.First, they reported a weak posi ve correla on for sleep dura on and a weak nega ve correla on for chronotype between spouses.These correla ons were replicated in both UK Biobank and 23andMe.Then, they used accelerometer data in the UK Biobank and observed similar results to the selfreport, with the me of diurnal ac vity showing a correla on.
The authors then extended the Mendelian randomisa on design to inves gate the causal e ects of one individual on another, what they describe as social or indirect gene c e ects, and tested 9 sleep traits.They also performed other sensi vity analyses to test the robustness of the MR, and assessed the role of age, me of wear and birth loca on.
This study is by far the largest of its kind, and represents an interes ng and solid contribu on to our understanding of how cohabita on inuences sleep behaviour.It is very well designed and executed and wri en clearly, so it is easy for the reader to follow.It will make a great addi on to the literature.
We thank the reviewer for their posi ve appraisal of our manuscript.
I have a few thoughts that the authors might want to consider, either as part of this paper or for their future work: 1) How does occupa on and socioeconomic status a ect the correla ons? (e.g., if the couple members are re red or working) We have extended our sensi vity analyses to assess a number of family-level e ect modiers which may a ect the magnitude of the causal rela onships between sleep traits among spouses.This includes an assessment of employment status (both employed, one employed and one unemployed/re red, both unemployed/re red) and socioeconomic posi on (Townsend depriva on index ter les).Neither employment status nor socioeconomic posi on appeared to strongly inuence the causal e ects which were revealed in the main analysis, i.e. chronotype, sleep dura on, snoring and L5-ming (Supplementary Figure 8).
We have updated the Methods and Results to incorporate this addi onal analysis (Methods, pages 23-24, lines 722-747 and Results, page 9, lines 253-280).
2) How do lifestyle variables and substance use (alcohol and tobacco) pa erns a ect sleep behaviour correla ons in spouses?How do BMI and chronic disease status a ect sleep behaviour in spouses?
We have included an assessment of how family-level factors may inuence the magnitude of the causal e ects of sleep traits between spouses, as suggested by Reviewers #1 and #2.This includes an assessment of employment status, children living in household, type of household, socio-economic posi on, rural/urban living and day me ac vity ming (Supplementary Figure 8; Methods, pages 23-24, lines 722-747 and Results, page 9, lines 253-280).How individual-level traits such as alcohol, smoking, BMI and chronic disease status also a ect sleep behaviour correla ons in spouses if very worthy of inves ga on.We are currently conduc ng as part of future work which will also assess these traits in a causal framework to evaluate whether they are causes or consequences of sleep concordance/discordance between spouses.
3) Also, the authors should discuss how their analysis are impacted by the fact that the previous GWAS they used to es mate polygenic risk scores had adjusted for sex di erences.
While previous GWAS have adjusted for sex, the polygenic risk scores derived from the GWAS summary sta s cs strongly associated with sleep traits in both males and females in UK Biobank.This is demonstrated in the assessment of instrument strength shown in Table 3.For some of the accelerometer-derived traits, F-sta s cs were small which could indicate weak instrument bias.This was par cularly the case for accelerometer traits in females, sugges ng that there may be some sex di erences in the gene c contribu on to the sleep traits.However, there would be a necessary tradeo with lower sample sizes for the sex-specic rather than sex-combined GWAS, which may reduce sta s cal power.This is highlighted in the Discussion (page 15, lines 447-452).
Reviewer #2 (Remarks to the Author): 1) Weak inverse e ects for chronotype were found between spouses in the UK Biobank and 23andMe.One explana on could be that they can be er handle the 24-h challenge of caring for toddlers.Do you see an associa on between the chronotype di erence between spouses and the number of children?
We have included an assessment of how family-level factors may inuence the magnitude of the causal e ects of sleep traits between spouses, as suggested by Reviewers #1 and #2.This includes an assessment of employment status, children living in household, type of household, socio-economic posi on, rural/urban living and day me ac vity ming (Supplementary Figure 8).We did not nd evidence to suggest that e ects in chronotype were substan ally di erent among those spouses who had children living in the household than those who didnt.However, we did nd evidence to suggest that diurnal ac vity e ects (based on accelerometer-derived L5 ming) were stronger in spouses when there were no children in the household (mean di erence = 0.90; 0.20, 1.60 SD per SD with no children vs. -0.11;-0.82, 0.62 SD per SD with one or more child in the household; I2 = 74%; PHet=0.05).
2) Sleep in a rela onship may be a ected by others living in the same household (e.g., pets or children).Furthermore, imposed work schedules can a ect me when people go to sleep.Do these factors mediate or mi gate some of the tested associa Our assessment of how family-level factors may inuence the magnitude of the causal e ects of sleep traits between spouses now includes an inves ga on of the number of children living in the household as well as employment status.While we found some evidence to suggest that number of children in the household might inuence concordance in diurnal ac vity between spouses (as described in response to Reviewer #2, comment #1), we did not nd that employment status strongly inuenced the e es mates for chronotype, sleep dura on, snoring and L5-ming (Supplementary Figure 8).We have updated the Methods and Results to incorporate this addi onal analysis (Methods, pages 23-24, lines 722-747 and Results, page 9, lines 253-280).
3) Do you know whether couples lived in an apartment building or a one-family house?Any di erences in how concordantly spouses slept?
Our assessment of how family-level factors may inuence the magnitude of the causal e ects of sleep traits between spouses now includes an inves ga on of the type of household in which the spouses were dwelling during the rst UK Biobank assessment.We did not nd evidence that type of household (house/bungalow of at/maisone e/apartment) inuenced the e ect es mates for chronotype, sleep dura on, snoring and L5-ming (Supplementary Figure 8).We have updated the Methods and Results to incorporate this addi onal analysis (Methods, pages 23-24, lines 722-747; Results, page 9, lines 253-280).
4) Does it ma er whether you live in an urban or rural area for how well spouses sleep concordantly?
Our assessment of how family-level factors may inuence the magnitude of the causal e ects of sleep traits between spouses now includes an assessment of the urban/rural loca on of the spouses households at recruitment into the study.We did not nd evidence that rural/urban loca on inuenced the e ect es mates for chronotype, sleep dura on, snoring and L5-ming (Supplementary Figure 8).We have updated the Methods and Results to incorporate this addi onal analysis (Methods, pages 23-24, lines 722-747 and Results, page 9, lines 253-280).

5) What about sleep associa ons among homosexual couples? Similar associa ons like those observed between heterosexual couples?
Unfortunately we were unable to assess sleep associa ons among homosexual couples as the spouse pairs in UK Biobank were determined using a previous approach for which exclusions were made if couples were of the same sex (1).Similarly, in 23andMe, spouses were determined based on parento spring trios and so spouses were mother-father pairs.Since our analysis inves gated sex di erences in sleep pa erns, and since di erences in sleep concordance between homosexual and heterosexual couples was not the focus of the paper, we would prefer not to inves gate this as part of the current manuscript but agree that this would be interes ng to evaluate in future studies.
6) Variance in sleep is not only a ma er of genes; it highly depends on the state in which a person or a couple is.Thus, I'm curious whether sleep overlaps between spouses may be more or less pronounced in certain life situa ons, such as low or high socioeconomic depriva on, taking care of a child, or before and a er rement.
Our assessment of how family-level factors may inuence the magnitude of the causal e ects sleep traits between spouses includes an inves ga on of employment status, children living in household, type of household, socio-economic posi on, rural/urban living and day me ac vity ming.
No large di erences in e ects were found by socio-economic, demographic and lifestyle factors, or accelerometer characteris cs.This was except for the accelerometer-derived ac vity ming, where later diurnal ac vity of one spouse had a larger e ect on the other if spouses lived in households without children.In addi on, for sleep dura on, larger posi ve e ects observed at older ages, sugges ng a convergence in sleep dura on between spouses over me.
7) Likewise, it may be interes ng to see whether couples spending more me together or have similar ac vity pa erns during wake me (which would mean that they may share similar wake experiences resul in a comparable homeosta c sleep pressure) are more likely to sleep concordantly than those who live their "own" lives.
We have now inves gated whether spouses were more likely to sleep concordantly if they had similar ac vity pa erns during wake me.This was done using the accelerometer-derived measure of M10ming, which is the most ac ve 10 hours of the day.We assessed whether sleep ming between spouses (based on L5-ming) was a ected by the di in day me ac vity (based on M10ming) between spouses, by assessing heterogeneity in the causal es mates between three groups experiencing the lowest, middle and highest levels of di erence in ac vity pa erns during wake me, based on M10-ming (Supplementary Figure 8).We found no clear evidence that spouses were likely to sleep more concordantly if they had similar ac vity pa erns during the day.

Reviewer #3 (Remarks to the Author):
Thanks for invi ng me for reviewing this paper.Richmond et al. conducted a Mendelian randomiza on (MR) study exploring correla ons sleeping pa erns and circadian preference between spouses, and concluded that individuals sleep may impact that of their partner.
1) Although the paper is well wri en, this is an obvious observa on and the jus ca on that this warrants an MR seems very weak.
While it may be an cipated that sleep traits within a couple are correlated, un l now only small-scale studies have inves gated the extent of this interdependence.Within the UK Biobank, we inves gated the correla ons between nine sleep traits (5 self-report and 4 accelerometer-derived measures), with replica on in 23andMe.A number of correla ons were found, including some which were not expected e.g. an inverse correla on between chronotype.To further understand the nature of these correla ons, we performed Mendelian randomiza on analysis in order to i) determine whether they represent true causal e ects or confounding by shared environmental factors and ii) establish direc onality in any causal e ects between spouses.MR analysis provided support for e ects of an individuals sleep trait on that of their spouse for chronotype, diurnal ac vity, sleep dura on and The jus ca on for both the large-scale observa onal analysis and Mendelian randomiza on followup are emphasised in the following paragraph in the Introduc on (page 4, lines 78-85).
Previous studies using ac graphy measures to inves gate spousal sleeping pa erns have been limited in terms of sample size.While larger studies have inves gated self-reported sleep traits among spouses (2, 3), they may su er from bias due to individuals percep on and recall of sleeping pa erns, which may di er between men and women.Previous observa onal studies inves ga ng both self-reported and objec vely-assessed sleeping pa erns between spouses may also be biased by confounding (i.e. by shared socioeconomic and lifestyle factors) and it can be di cult to determine the direc onality in correlated sleep pa erns between spouses (i.e. the extent to which one spouse inuences the sleep pa erns of the other, and vice versa).
2) Public health implica ons of this study are also not clear and convincing (i.e., providing insights into human sleep interac ons).
In the Introduc on, we state that Within cohabi ng couples, it is of interest to inves gate the interdependence in sleep pa erns since this may exacerbate sleep problems, which could have social, psychological, and physical health implica ons (page 3, lines 54-56).We now elaborate on this by sta ng: Establishing whether the sleep habits of a spouse could serve as a risk factor for an individuals own poor sleep would enhance our understanding of the familial impacts on sleep and sleep-related ill health, and would promote opportuni es for interven ons aimed at the family level.(page 3, lines 56-58).
Based on the results of our study, we state in the Discussion that Our ndings provide insights into sleep behaviour among cohabi ng spouses which may impact on rela onship factors and could have further downstream physical and mental health consequences.We now elaborate on this by sta ng: Our ndings provide insights into sleep behaviour among cohabi ng spouses and highlight how certain sleep traits can be inuenced by the sleep of a persons spouse.This helps us gain a be er understanding of the e ology of poor sleep, which may in turn impact on rela onship factors and could have further downstream physical and mental health consequences (4)(5)(6)(7).Results of this study promote further inves on into the familial impacts of sleep and sleep-related ill health and raise possible opportuni es for sleep interven ons aimed at the family level.However, the magnitude of sleep e was small and whether this level of correla on between spouses contributes towards disease risk, as indicated in (8), requires further inves ga on.(page 16, lines 489-496).
While the abstract word limit restricts an extensive discussion of the public health implica ons of our ndings, we have now replaced the phrase providing insights into human sleep interac ons with promo ng opportuni es for sleep interven ons at the family-level (page 2, lines 37-38).
3) Novelty is also a major concern.
We disagree that our study is not novel.Exis ng studies in the sleep eld which assessing concordance in sleep behaviours between spouses are scarce and those that exist have typically inves gated a very small number of par cipants (n<100 couples) (9)(10) (11).The present study therefore serves as a novel addi on to the sleep literature by conduc ng an assessment of sleep behaviour in 47,420 spouse pairs in UK Biobank and 127,035 spouse pairs in 23andMe.
Within the Mendelian Randomiza on eld, while previous studies have inves gated spousal concordance and indirect gene c e ects in rela on to a range of socio-economic, lifestyle and behavioural phenotypes (12,13), this has not included an assessment of sleep traits.Furthermore, while the majority of these studies have iden ed posi ve correla ons in traits between spouses (representa ve of assorta ve ma ng, social homogamy or interac ons a er partnership), the nding of an inverse correla on (and conrmed causal e ect) in chronotype between spouses is novel and of interest, since it goes against the plethora of evidence indica ng widespread es (rather than di erences) of spouses for many phenotypes.
I have more specic comments, as below: 4) In introduc on, the authors are suggested to provide stronger argument why determine the existence and direc onality in correlated sleep pa erns between spouses ma ers.Will the results have any possibility in changing the current or future clinical or public health prac ce.
Or by clarifying this research ques on, can we gain be er understanding about e ology?
We thank the reviewer for this sugges on.As men oned in rela on to the reviewers comment #2, in the Introduc on, we now emphasise the poten al importance of our research ques on with the following statement: Establishing whether the sleep habits of a spouse could serve as a risk factor for an individuals own poor sleep would enhance our understanding of the familial impacts on sleep and sleep-related ill health, and would promote opportuni es for interven ons aimed at the family level.(page 3, lines 56-58).
5) All sleep ques ons in UK Biobank are very simple and not validated.Also, no informa on regarding the reliability of these ques ons has been provided.
The reviewer raises an important point about the validity of the sleep ques onnaire measures which we have now addressed in the Limita ons sec on of the Discussion (page 14, lines 411-420): Self-reported measures of the sleep traits in UK Biobank and 23andMe were based on a limited number of ques ons which may crudely assess underlying sleep traits and/or be subject to bias.However, there is a trade-o between the number of individuals with sleep phenotypes obtained from easy-to-administer ques onnaires and more objec ve or clinically-derived sleep measures, which are typically available on a smaller number of par cipants.The replica on of gene c associa ons iden ed in rela on to the self-reported sleep traits in UK Biobank with comparable phenotypes in independent studies including 23andMe (14-16), and with objec ves measures i.e. from the accelerometer assessment in UK Biobank (16,17), serves as an indirect means of valida on of the self-report measures (18).Further to this, we also directly assessed spousal concordance using accelerometer-derived measures in a subset of the UK Biobank.
6) The interpreta on of correla ons are not clear.Need to explain why not used beta coe cients, which is more commonly used in MR in this study.
E ect es mates from the MR analysis represent the mean di erence in the spouses sleep trait (in SD) per SD increase in an individuals own sleep trait, with the excep on of snoring for which es mates represent a risk di erence (since this is a binary trait).This interpreta on is highlighted in Figure 4 and also in the Results.Corresponding mean di erences from both mul variable and MR analysis are directly comparable to phenotypic par al correla ons when both exposure and outcome are measured in SD units.The is highlighted in the Methods, Standardized variables are presented..to allow for direct comparisons with the correla on coe cients es mated for the phenotypic correla on.(page 23, lines 698-699).
7) Interpreta on of SEs in some tables (i.e., Table 2) is di cult because of the lack of units for variables and reference groups.
As above, e ect es mates (and corresponding SEs) represent the mean di erence in the spouses sleep trait (in SD) per SD increase in an individuals own sleep trait, with the excep on of snoring for which es mates represent risk di erence.This is described in the Table 2 footnote.
8) Its also suggested to provide the R code of the data analysis for reviewers and readers to check.
The Stata code used to es mate phenotypic and gene c correla ons, and to perform the main onesample Mendelian randomiza on analysis in UK Biobank, is available via GitHub (h ps://github.com/rcrichmond/spousal_sleep).