Body mass index associates with sleep phase delay and low daytime skin temperature despite similar 24-hour patterns of physical activity in adults

Introduction. Circadian disruption is a complex state of desynchronized circadian rhythms, either due to external misalignment between the intrinsic clock and the environment, or to internal desynchronization. Examples of external misalignment are social jet lag, light at night, and shift-work (Roenneberg et al, 2012; Gooley et al, 2011; Vetter, 2020). Internal desynchronization is more likely to occur at an older age and in the presence of disease or pre-disease (Bass, 2012; Gubin, 2013; Gubin & Weinert, 2015; Gubin et al, 2016; Xie et al, 2019). In other words, disorganized circadian hygiene can be an underlying cause of disease, and a misaligned circadian phenotype can facilitate the onset and progression of pathology. Vice versa, disease can manifest itself as a deviant circadian pattern, alterations observed in terms of period, phase, amplitude, and/or overall pattern of variability.

Circadian disruption is considered a risk factor for numerous chronic diseases (Agadzhanian & Gubin, 2004; Arble et al, 2015). They include metabolic state (Maury, 2019; Noh, 2018; Panda, 2016; Park et al, 2019; Zimmet et al, 2019), obesity (Noh, 2018; Broussard & Van Cauter, 2016; McHill & Wright, 2017; Park et al, 2019; Zimmet et al, 2019), and diabetes (Arble et al, 2015; Gubin et al, 2017b; Javeed & Matveyenko, 2018; Mason et al, 2020). Excessive artificial light in the late evening or night disrupts melatonin production and signalling, sleep, thermoregulation, blood pressure, glucose homeostasis (Gooley et al, 2011; Qian & Scheer, 2016; Touitou et al, 2017; Gubin et al, 2017a; Poggiogalle et al, 2018), and lipid metabolism (Gooley, 2016; Poggiogalle et al, 2018).

Specific mechanisms of intrinsic vs. extrinsic circadian disruption, and genetic factors predisposing individual susceptibility of circadian clock to disrupting factors, however, remain largely unknown. For example, the over-50-fold individual difference in melatonin production sensitivity in response to the same evening light exposure remains unexplained (Phillips et al, 2019).

Modern monitoring techniques are useful tools for an in-depth analysis of circadian physiology, not just in the laboratory, but also in everyday life. In this study, we investigate overt circadian rhythms of physical activity, skin temperature and metabolism, simultaneously with sleep parameters, monitored during three consecutive days (72-hours) in adults followed-up for metabolic reasons in Tyumen University clinic.

Methods. Thirteen adults (seven women), followed-up for their risk of developing metabolic disorders at the Tyumen Medical University’s clinic, underwent a full 72-hour outpatient monitoring of sleep, activity, skin temperature and metabolism during the 2019-2020 winter season. All data were obtained with the same monitor Sense Wear, Arm Band Pro3 , Pittsburgh, USA, which

BMI <27;      BMI >27

Figure 1. Circadian rhythms of sleep, skin temperature, physical activity, energy expenditure, heat flux and step counts in groups with higher or lower body mass index (BMI)

has proven to be a reliable device for scientific research on sleep physiology, activity, and metabolism (Almeida et al, 2011; Koehler et al, 2017). Only data from records (n=13) with more than 90% reliable measurements were analyzed. Patients were subdivided into two groups according to their body mass index (BMI): either >27 kg/m2 (n=6, 4 women) or <27 kg/m2 (n=7, 3 women). Patients in the two groups had a similar average age (60.8±3.3 vs. 55.4±2.9; p=0.250).

Hourly means and standard deviations of each patient were computed from the original data of sleep time, lying time, step counts, energy expenditure, physical activity, heat flux, and skin temperature. The circadian patterns between the two groups were compared by Cosinor and ANOVA analysis of the hourly mean values.

Statistical analyses were performed using the software packages Excel, STATISTICA 6 and SPSS 23.0. The Shapiro-Wilk’s W-test was applied to check for normality. When variables were normally distributed (W-test’s p-value >0.05), a 1-way ANOVA was used, with Tukey’s post hoc correction for multiple testing. Otherwise, the Kruskal-Wallis and the Mann-Whitney post hoc tests were used.

The level of statistical significance was set at 5%.

Each 72-hour time series was fitted with a 24-hour cosine curve by least squares to yield estimates of the MESOR (M, Midline Estimating Statistics Of Rhythm, a rhythm-adjusted mean), 24-hour amplitude (A) and acrophase (Φ phase angle of the peak of the cosine curve fitted to the data in reference to local midnight). Results were summarized by population-mean cosinor (Cornelissen, 2014). Rhythm parameters were compared using Bingham’s parameter tests (Bingham et al, 1982). Exact p-values are listed in the text.

Results. Groups with higher vs. lower BMI differed significantly in their 24-hour patterns of sleep (F_(23,1659) =3.25; p<0.0001) and skin temperature (F_(23,1659) =2.82; p<0.0001), Fig.1. The higher BMI group also tended to have a later sleep phase (cosinor-estimated mean phase delay of about 1h:44min, p=0.10) and a later circadian phase of skin temperature (by about 2h:40min, p=0.09), Fig.2. The higher BMI group had a lower daytime skin temperature, but a night-time skin temperature similar to that of the lower BMI group. No differences in mean value, MESOR, or circadian amplitude were found for either sleep amount or skin temperature. The circadian

Figure 2. The group with a higher body mass index tended to have a delayed circadian phase of skin temperature. A: BMI > 27 kg/m²; B: BMI < 27 kg/m^2.

Population-mean cosinor analysis:

BMI < 27 kg/m2: Mean phase = 01:00 hh: min;

BMI > 27 kg/m2: Mean phase = 03:40 hh: min;

Bingham-test for acrophase difference: F(1, 11) =3.316; p=0.09.

pattern of skin temperature differed between men and women.

Contrary to sleep and skin temperature, the two groups had similar 24-hour patterns of physical activity (p=0.243), step counts (p=0.663), heat flux (p=0.143), and energy expenditure (p=0.055), Fig.1. There were no differences in the mean values, amplitudes or corresponding circadian phases or in the hourly averages of these variables. A “gating” effect was apparent for physical activity and energy expenditure, as differences between the two groups were only present for a short time in the morning and evening, Fig.1.

Discussion. Until recently, research on the bidirectional relationship between BMI and sleep focused mainly on its association with sleep duration. Most studies concentrated on adolescents and young adults, which yielded results generally different from those found in adults. In adults, the relationship between BMI and sleep duration was close to be U-shaped. An increase in BMI could be linked to short and especially long sleep duration (Grandner et al, 2015).

Not only altered sleep duration, but also a shift in the circadian phase of sleep and temperature can serve as an indicator of metabolic risk, and can be specific for certain disorders. Herein we found that daytime skin temperature was lower in adults with a higher BMI. These results are in agreement with the previously reported gradual increase in body temperature in prediabetes and type 2 diabetes mellitus during the night (Gubin et al, 2017b), since the circadian patterns of skin temperature and body temperature are inversely related.

The main finding from this pilot study is the association between a phase-delayed circadian rhythm of temperature and sleep with a metabolic indicator, independently of any differences in their 24-hour mean values or in the circadian characteristics of physical activity or energy expenditure. Objective estimates of phase are often overlooked in currently undertaken surveys. Differences in 24-hour temperature in relation to BMI, however, may not be evident in the general population (Heikens et al, 2011). Differences in temperature have been attributed to gender and menopausal stage in women (Bastardot et al, 2019).

The circadian phase is a reliable and relatively stable endpoint, which can be used in prospective studies to detect early signs of metabolic disorders. Indeed, the circadian phase, the intrinsic circadian period, and chronotype were found to be fairly reproducible over several months (Kantermann & Eastman, 2018). A diminished circadian amplitude of body or wrist temperature can be used as a criterion of metabolic disorder (Gubin et al, 2017b; Harfmann et al, 2017). In cases when monitoring is performed during several consecutive days, a diminished temperature stability of the circadian temperature variation can serve the same purpose (Harfmann et al, 2017). These endpoints can also correlate with certain biochemical variables, such as triglycerides (Harfmann et al, 2017), leptin (Gubin et al, 2017b), or a disrupted 24-hour profile of fasting blood glucose (Gubin et al, 2017b).

A later chronotype (evening preferences in diurnal activity) has been repeatedly shown to be associated with a higher overall morbidity and mortality (Knutson & von Shantz, 2018). Specifically, a higher risk of overweight (Maukonen et al, 2016; Mazri et al, 2020; Sun et al, 2020), prediabetes and diabetes (Reutrakul et al, 2013; Knutson & von Shantz, 2018; Afroz-Hossain et al, 2019; Hudec et al, 2020) has been linked to an evening chronotype. A higher BMI in adult women correlates with an evening chronotype (Sládek et al, 2020). A later chronotype also associates with higher HbA1c determinations in patients with prediabetes, independently of social jetlag and other sleep disturbances in a Thai population (Anothaisintawee et al, 2017).

Since chronotype is a reflection of the intrinsic clock (Duffy et al, 2001; Emery et al, 2009; Gubin & Kolomeychuk, 2019) and has heritability of about 50% (Kalmbach et al, 2017), certain genetic factors are expected to be shared between specific circadian phases and the risk of metabolic disorder. For example, the melatonin receptor MTNR1B rs10830963 G-allele associates with reduced pancreatic beta-cell function and higher diabetes mellitus risk due to an increase in fasting blood glucose (Prokopenko et al, 2009; Sparsø et al, 2009). The MTNR1B G-allele also associates with prolonged endogenous melatonin secretion (Lane et al, 2016). The combination of an evening chronotype with the CG (though not GG) genotype potentiates diabetes risk (Tan et al, 2020). A personalized approach is thus recommended, since the association between evening chronotype and metabolic consequences may have genetic specifics, such as racial differences (Sun et al, 2020).

Social jet-lag is another established factor underlying the risk of metabolic disorder (Roenneberg et al, 2012), and can be associated with a decreased circadian amplitude of wrist temperature (Polugrudov et al, 2016). A multidirectional relationship exists among social jet lag, chronotype, photoperiod, mood, and food addiction (Borisenkov et al, 2020a, b).

Not only chronotype, sleep phase and duration, but also the habitual exposure to artificial light during the night may exert an independent effect leading to a higher BMI and an increased risk of obesity (Park et al, 2019). In mice, even relatively short-term exposure to low levels of light-at-night (LAN) alters metabolism and the circadian rhythm of core temperature, and increases body mass (Borniger et al, 2014). LAN not only leads to weight gain, but also increases glucose intolerance and insulin resistance, that can be reversed by return of darkness. LAN exposure also shortens survival of mice with type 2 diabetes (Russart et al, 2019). Thus, individual differences in diurnal vs. nocturnal exposure to light (deficient natural diurnal light / excessive artificial nocturnal light) and chronotype should be considered in research aimed at evaluating metabolic risk.

The timing of habitual food intake is another important factor that modulates both overt circadian rhythms and metabolic consequences. For example, skipping breakfast was recently reported to be an independent lifestyle habit associated with persistently increased arterial stiffness in patients with type 2 diabetes mellitus in Japan (Mita et al, 2020). Diet-induced thermogenesis and the ability to maintain temperature homeokinesis can be subject to individual genetic-based differences (Reinhardt et al, 2016; Vinales et al, 2019). Therefore, future studies should include reliable tools to assess circadian patterns in food intake that estimate both the amount and timing of meals.

Conclusion. Profound changes in the circadian pattern and its phase of physiological functions may occur in the absence of any difference in mean value. This pilot study indicates that suboptimal metabolic state can be predicted in terms of specific changes in the circadian phase of sleep and temperature that are not observed in their 24-hour mean value or amplitude, and not in the circadian patterns of physical activity and energy expenditure. Further in-depth analyses of circadian rhythm parameters are needed to unmask undesirable behavioural factors and unravel metabolic risk in order to provide personalized recommendations for a healthy lifestyle and well-being.