Habitual Sleep/Wake Patterns in the Old Order Amish: Heritability and Association with Non-Genetic Factors

INTRODUCTION

The preferred wake time and bedtime, also referred to as the preferred sleep phase or the diurnal preference, is regulated by at least two factors: the endogenous circadian pacemaker and environmental stimuli.¹ The endogenous circadian pacemaker, commonly named the biological clock, resides in the suprachiasmatic nucleus (SCN) of the hypothalamus.² Environmental stimuli, such as light, exercise, social contact, and strict schedules, entrain the endogenous pacemaker to the environment.³ Light, the most potent entraining factor, regulates the SCN by exciting retinal photoreceptors that connect to the SCN via the retinohypothalamic tract.²

Many diseases display a circadian pattern of risk and have been shown to be associated with sleep disturbance.¹ Epidemiologic studies have shown that both long and short sleep duration is highly prevalent in the United States⁴ and is associated with many common diseases, such as hypertension,⁵ obesity,⁶ diabetes,⁷ and coronary heart disease.⁸ Shift workers who experience a misalignment between their endogenous circadian pacemaker and behavioral cycles face an increased risk for cardiovascular disease and cancer.⁹ Recently, circadian misalignment has been directly manipulated in a small sample of people in a laboratory setting, resulting in adverse cardiometabolic effects.¹⁰ Therefore, identifying biological and environmental factors that affect sleep/wake patterns may lead to interventions that reduce the incidence of many common disorders.

Much of what is known about the molecular mechanisms of the circadian pacemaker comes from studies of model organisms and Mendelian human sleep disorders. Two functionally conserved gene families, Period and Cryptochrome, play a central role in the control of the mammalian circadian pacemaker by engaging in an autoregulatory negative feedback loop.² In humans, mutations in components of the well-known circadian rhythm pathway have been found to result in rare Mendelian sleep disorders such as advanced or delayed sleep phase syndrome, which are estimated to affect less than 1% of the adult population.^2,11 Much less is known, however, about the genetic regulation of less extreme variation in sleep phase among community-dwelling individuals. The few studies that have been conducted rely on self-administered questionnaires or sleep diaries to determine sleep phase. Based on these instruments, a substantial degree of heritability (h²) is suggested (h² estimates for diurnal preference, sleep length, and daytime sleepiness are 0.23–0.52,^12–17 0.4–0.44,^12,18 and 0.38–0.48,^19–21 respectively). To date, a single genome-wide association study of daytime sleepiness, usual bedtime and usual sleep duration assessed by self-administered questionnaires in a community-dwelling population has been performed and evidence for significant heritability and genetic association for all three traits was found.²²

In order to study the association between genetic/non-genetic factors and habitual sleep patterns, it is important to objectively measure sleep phase for multiple days in a large number of community-dwelling individuals. This poses the formidable challenge of accurately measuring a person's true sleep phase efficiently without being so invasive as to disrupt normal sleep habits. To this end, we explored the usefulness of Actical physical activity monitors (formerly Mini Mitter, Bend, OR, now Philips Respironics) to determine habitual bedtime and wake time. While the Actical device is mechanically and electronically similar to actigraphy devices used to measure sleep,²³ its software is optimized for the measurement of daytime physical activity level.^24,25 Moreover, our study participants wore the Actical on the hip, the recommended placement for physical activity measurement, as opposed to the wrist, the standard placement for sleep assessment.²³ Accelerometers worn on the hip to measure physical activity level are increasingly being used in large epidemiological studies, including the National Health and Nutrition Examination Survey (NHANES) for the 2003-2004 cycle and later cycles. Demonstrating the utility of such measurements for the assessment of sleep could potentially lead to the development of valuable resources for sleep researchers.

Our use of the Actical accelerometer to determine habitual bedtime, wake time, and sleep duration, a use for which it was not designed, compelled us to validate these sleep parameters using sleep diaries. Sleep diaries have been extensively used in sleep research since the 1960s and have been shown to provide reliable estimates of bedtime and wake time.^26–29 Upon observing strong correlations between Actical-based and diary-based sleep parameters, we sought to determine the extent to which Actical-based sleep parameters are heritable and associated with non-genetic factors. Study participants also completed the Horne-Östberg Morningness-Eveningness Questionnaire (MEQ)³⁰ to estimate diurnal preference and a modified version of the Epworth Sleepiness Scale (ESS)³¹ to estimate daytime sleepiness. Both the categorical diurnal types and the continuous scores from the MEQ have been shown to be correlated with the endogenous circadian period.^32–35 We examined the contribution of genetic and non-genetic factors on these questionnaire scores as well as the relationship between questionnaire scores and Actical-based sleep parameters.

Factors that accompany urbanization and industrialization are likely to affect sleep.^3,36–39 We attempted to avoid much of the influence from these factors by conducting our study in the Old Order Amish (OOA), a conservative Christian sect residing in rural areas of Lancaster County, Pennsylvania. While the avoidance of modern electrical conveniences by the OOA likely reduces non-genetic influences on sleep, the OOA are still bound to schedules by occupational and social obligations. Little is known about sleep in rural populations,^40,41 and even less in populations that avoid modern technology.¹⁴ This community-based study of the OOA offers a rare examination of sleep without the influence of industrialization. In addition to the traditional agrarian lifestyle of the OOA, the homogeneity of socioeconomic status and lifestyle is expected to facilitate the identification of genetic factors contributing to variation in habitual sleep by reducing sources of non-genetic influences. The heritability estimation of habitual sleep parameters was made possible by the use of data collected from extended pedigrees.⁴²

METHODS

RESULTS

There were slightly more males (388) than females (335) in the study population of 723 community-dwelling adults (P = 0.05) (Table 1). Combining both sexes, the mean age was 43.3 ± 13.8 years, and the range was from 20 to 80 years. Male subjects were significantly younger than female subjects (men: 41.9 ± 13.2; women: 44.9 ± 14.3 years; P = 0.003) (Table 1). Very few participants classified themselves as retired; 87% of men were farmers, carpenters, or mechanics, while 79% of women were housewives (Table 1). Napping was common in the study population, with 54% of men and 62% of women reporting that they made time in the day to take a nap at least once a week (Table 1). Total activity level was significantly higher in men than in women (men: 525.2 × 10³ ± 269.6 × 10³; women: 377.5 × 10³ ± 210.2 × 10³ counts/day; P < 0.0001), but sleep activity level did not differ by sex (Table 1).

Sleep parameters measured by Actical and sleep diaries were compared in 164 people from whom both sets of data were available. Wake time obtained from the 2 instruments was highly correlated, as was bedtime (wake time: r = 0.82; bedtime: r = 0.72; both P < 0.0001; Figure 2 A, B). On average, the diary wake times and bedtimes were only 6 and 11 minutes earlier, respectively, than those measured by Actical, but these differences were significant (P = 0.0004 and P < 0.0001, respectively). Bland-Altman plots indicate that there is no obvious relationship between the difference and the mean of the 2 measurements (Figure 2 C, D).

The mean Actical-based habitual wake time, bedtime, and sleep duration when both sexes were combined was 5:06 ± 00:44, 22:00 ± 00:44, and 7:05 ± 00:51, respectively. The Actical-based habitual wake time was significantly earlier for men (4:59 ± 45 min) than for women (5:14 ± 43 min, P < 0.0001) (Table 1). Actical-based habitual bedtime was not significantly different by sex, and consequently, habitual sleep duration was significantly shorter in men (7:01 ± 52 min) than in women (7:11 ± 49 min, P = 0.01) (Table 1). Based on the MEQ where lower scores indicate evening diurnal types, only 1% of the participants were classified as moderately evening types, and none were classified as definitely evening types (Table 1). Given the narrow range of diurnal types in the study population, continuous MEQ scores, which were not significantly different by sex (Table 1), were used in regression analysis. Based on the ESS where higher scores indicate higher levels of daytime sleepiness, men had higher levels of daytime sleepiness than women (men: 9.1 ± 3.4; women: 7.6 ± 3.2; P < 0.0001) (Table 1).

The association between sleep parameters and factors known to be associated with sleep phase and duration, including age, sex, season of the year, occupation, and activity level, was evaluated in regression models conditioned on the relatedness of the study participants (Table 2).^3,49 In general, factors associated with an unequal shift in wake time and bedtime were associated with a shift in sleep duration. The observation that sleep duration was significantly different by sex in univariate analysis (Table 1) but not multivariate analysis (Table 2) most likely reflects the high degree of collinearity between sex and the housewife occupation. Indeed, in multivariate analysis, the housewife occupation was associated with longer sleep duration (18.0 ± 7.2 min, P = 0.02), indicating that the sex-associated variance was attributed to occupation in some of the models (Table 2). Summer, the season with the latest sunset, also had the latest bedtime compared with winter, the baseline season (43.8 ± 4.2 min, P < 0.0001, Table 2).

Higher total activity level was significantly associated with earlier wake time (−18.6 ± 4.2 min, P < 0.0001), but not bedtime (−3.0 ± 3.6 min), and accordingly, was significantly associated with shorter sleep duration (−13.8 ± 4.8 min, P = 0.004) (Table 2). Although sleep activity level was not significantly associated with wake time or bedtime, it was significantly associated with longer sleep duration (6.0 ± 3.0 min, P = 0.04, Table 2). The effects of BMI and napping frequency were not significantly associated with any of the outcomes in multivariate analyses, and were not included in final models.

From the analysis of outcomes obtained from questionnaires, increasing age was associated with a higher MEQ score (1.7 ± 0.2, P < 0.0001), corresponding to an increasing tendency towards the morning diurnal type, but no other factor was significantly associated (Table 2). ESS score was significantly associated with age, sex, and occupation, but not with season of the year or activity level (Table 2).

It was investigated whether Actical-based sleep parameters were associated with MEQ or ESS scores by individually adding these Actical-based parameters to the final regression models analyzing MEQ and ESS scores adjusted for age, sex, season, occupation, and activity level (Table 2). As expected, Actical-based wake time and bedtime later in the day were significantly associated with lower MEQ scores, corresponding to more evening diurnal types (wake time: −2.0 ± 0.4, P < 0.0001; bedtime: −1.2 ± 0.4, P = 0.004; Table 2). Actical-based sleep duration was not associated with diurnal preference measured by MEQ scores (−0.3 ± 0.4, Table 2). Consistently, Actical-based wake time and bedtime showed significant, negative correlation with MEQ scores (r = −0.18 and −0.15, respectively, both P < 0.0001), but Actical-based sleep duration did not. Actical-based wake time later in the day was significantly associated with lower ESS scores (−0.6 ± 0.2, P = 0.001), corresponding to less daytime sleepiness, and later Actical-based bedtime was significantly associated with higher ESS scores (0.4 ± 0.2, P = 0.01), corresponding to more daytime sleepiness (Table 2). As expected, longer sleep duration was significantly associated with lower ESS scores (−0.7 ± 0.2, P < 0.0001), corresponding to less daytime sleepiness (Table 2).

After adjustment for non-genetic factors and the effects of a shared household, variance component analysis revealed that wake time, MEQ score, and ESS score were significantly heritable (all P < 0.01), but not bedtime or sleep duration (Table 3). Additive genetic influences explained 20%, 21%, and 17% of the residual variance in Actical-based wake time, MEQ score, and ESS score, respectively (Table 3). There was a significant genetic correlation between Actical-based wake time and MEQ score (ρ_G = −0.60 ± 0.26, P = 0.05), but not between Actical-based wake time and ESS score, or between MEQ and ESS score. The effects of a shared household were significantly associated with all three Actical-based sleep parameters, but not with MEQ score or ESS score (Table 3). There was no evidence for sex-specific effects on heritability.

DISCUSSION

In this study, we used a physical activity monitor, the Actical, to measure habitual wake time, bedtime, and sleep duration. Upon observing that sleep parameters from Actical and sleep diaries were highly correlated and in agreement, we evaluated the contribution of genetic and non-genetic factors on Actical-based sleep parameters, as well as on questionnaire-based diurnal preference and daytime sleepiness. As would be expected from an informative measure of sleep phase and duration, we found Actical-based sleep parameters to be associated with factors known to be associated with sleep phase, such as age, occupation, and season of the year. Also, we found Actical-based wake time and bedtime to be associated with diurnal preference and shortened Actical-based sleep duration to be associated with increased daytime sleepiness. The measurement of physical activity level and sleep parameters with the Actical revealed an association between greater amounts of total physical activity with earlier wake time and shorter sleep duration in this study population. After adjusting for the effect of non-genetic factors and shared households, Actical-based wake time, diurnal preference, and daytime sleepiness showed significant heritability. Furthermore, diurnal preference had a significant genetic correlation with Actical-based wake time, suggesting a shared genetic influence or a similarity in these traits.

To study habitual sleep phase in large community-dwelling populations, actigraphy provides a reasonable alternative to the gold-standard method of polysomnography.²³ The Actical device used in this study is similar in design to sleep actigraphy devices, but instead of being worn on the wrist as is standard for sleep studies, the Actical was worn on the hip to optimize the detection of physical activity. In addition, a validated computer algorithm was not used to determine sleep parameters from Actical data. Despite these limitations, we observed Actical-based wake time and bedtime to be highly correlated with sleep diary measures. Consistent with a previous comparison between actigraphy and sleep diaries,²⁹ we observed weaker agreement between sleep diary and Actical-based bedtime than for wake time, possibly reflecting reduced accuracy for diary-based bedtime. A previous study determined wake time and bedtime using the ActiGraph (formerly CSA) accelerometer model 7164 worn on the hip without using a validated computer algorithm.⁵⁰ Taken together, these data suggest that activity monitors worn on the hip can provide useful information on habitual wake time and bedtime.

It has been posited that urbanization has contributed to the lack of sleep that is so common in modern industrialized societies, and that time spent watching television before bedtime can possibly be exchanged for sleep time.^38,39 Thus, the freedom from modern technology among the OOA might be expected to result in longer habitual sleep duration, but in fact, habitual sleep duration in our OOA study population was similar to that found in other study populations with similar age distributions.^22,38,51,52 In addition to sleep duration, it was of interest to compare sleep phase in our OOA study population with other study populations. While sleep phase is generally early in middle-aged study populations such as ours, we found habitual sleep phase in our study population to be slightly earlier than that found in other study populations with similar age distributions.^22,33,51,53 Consistent with this finding, the diurnal preference was similar, if not slightly earlier, to other middle-aged study populations.^33,53 Even after applying revised MEQ score cutoffs designed for middle-aged populations, we only classified 7% of the OOA study participants as evening diurnal types, potentially indicating our study population has a slightly earlier average diurnal preference relative to other middle-aged populations.⁵³ Consistent with our findings, previous studies have found rural populations to have earlier chronotypes compared to urban populations.^36,37 ESS scores that measure daytime sleepiness also appear to be higher in our study population compared with other populations.^22,54 Considering the omission of two items from the modified Amish ESS (see Methods), the level of daytime sleepiness in our study population might be underestimated. This high level of daytime sleepiness is unlikely to be a result of short sleep duration, as sleep duration in the OOA is similar to what has been reported for other populations. Perhaps other factors associated with high ESS scores are prevalent in our OOA study population.

The early sleep phase of the OOA apparently did not mask known associations with sleep phase. Actical-based wake time was earlier with increasing age, consistent with previous findings.^33,49 It has been suggested that the influence of the solar light/dark cycle on sleep timing^3,49 is stronger in smaller towns compared with larger cities.³⁶ Consistently, Actical-based sleep parameters in this agrarian OOA study population were significantly associated with the season of the year. Other seasonal factors, such as temperature and/or harvest activities, could also explain this observed association. Consistent with previous findings that diurnal preference is more strongly associated with sleep phase than sleep duration,^49,55 we observed MEQ score to be associated with Actical-based wake time and bedtime but not with sleep duration.

Our use of Actical accelerometers provided us with the opportunity to explore the complex relationship between sleep patterns and total physical activity level. This relationship has been shown to be potentially modulated by many factors, such as age, sex, and the type and timing of the exercise.^56,57 Perhaps not surprisingly, various studies have found inconsistent associations between chronic physical activity and sleep duration, but in general, increased levels of chronic physical activity are associated with more sleep.^56,57 In contrast, we observed a negative correlation between activity level per waking hour and sleep duration in the OOA. Perhaps the direction of the association between total physical activity and sleep duration changes at different levels of physical activity. In fact, it has been postulated that strenuous exercise might stimulate high concentrations of pro-inflammatory cytokines that can increase wakefulness.⁵⁷ The OOA appear to be a very physically active population; Actical physical activity counts are much higher in our study population compared with a previous study of non-Amish participants fitted with the Actical.⁵⁸ A pedometer-based study of an OOA population in Canada also concluded that physical activity levels in the Amish are much higher than in other populations.⁵⁹ Further work will be needed to determine if total physical activity and sleep duration are negatively correlated in other highly active populations, or if this finding is specific to the OOA.

Our heritability estimates of various sleep-related traits in the Amish are similar to estimates obtained from other populations. We found Actical-based wake time, but not Actical-based bedtime or sleep duration, to be moderately heritable (h² = 0.20 ± 0.09). Our heritability estimate for diurnal preference based on MEQ score (h² = 0.21 ± 0.09) was similar to previous estimates in other populations.^12–16 A study of the Hutterites,¹⁴ a rural population with a traditional lifestyle similar to the OOA, estimated the heritability of diurnal preference using a modified version of the MEQ and the heritability of wake time and bedtime using the Sleep Timing Questionnaire. Their heritability estimates for sleep parameters were very similar to our estimates, but they did not adjust for shared household effects (wake time: h² = 0.12 ± 0.06; bedtime: h² = 0.10 ± 0.10; diurnal preference: h² = 0.24 ± 0.09).¹⁴ Individuals within a household not only have a shared environment, but are also more likely to be related. Thus, our heritability estimates should be considered conservative as the effects of a shared household were taken into account. Without adjusting for the effects of a shared household, our heritability estimate for Actical-based wake time increases to 0.32 ± 0.09. The absence of heritability for Actical-based bedtime was primarily due to the strong influence from household effects (72% variance explained). Daytime sleepiness measured by the ESS was also significantly heritable, but the genetic influence on ESS appeared to be distinct from the genetic influence on diurnal preference or Actical-based wake time based on our genetic correlation analyses.

In conclusion, findings from our study suggest that Actical could be a promising alternative for the ascertainment of information on habitual sleep patterns. Further validation studies are needed to compare sleep measurements recorded by Actical with measurements from established methods such as polysomnography. In addition to information on sleep, Actical also measures the level of total physical activity. As physical activity, a modifiable factor, is observed to be strongly associated with sleep patterns in the Amish, a better understanding of its relationship with sleep in other populations may lead to effective preventive measures for sleep-related disorders. In addition, the observation that wake time, diurnal preference, and daytime sleepiness are heritable provides justification for further genetic studies of these traits in the OOA.

DISCLOSURE STATEMENT

This was not an industry supported study. Dr. Perlis has consulted for Gerson Lehrman Group, Clinical Advisors, MedaCorp/Leerink Swann, Actelion, SleepEasily, Sanofi-Aventis, and L.E.K. Consulting LLC. He has received research support from Sanofi-Aventis, Cephalon, and Vanda. He participated in speaking engagements with Sanofi-Aventis and the SMEI, the AASM, UR Sponsored, Respironics, Sepracor, Cephalon, Merck, Neurocare, Phillips, and Internet Didactic Services. He has received honoraria or consultant fees from Internet Didactic Services, Actelion, Takeda, Gerson Lehrman, Clinical Advisors, LEK, and Meda Corp. He has received other remunerations and/or use of materials from Internet Data Portal, Minimitter/Respironics, and Lifelines. The other authors have indicated no financial conflicts of interest.