Sleep and cognitive performance of African-Americans and European-Americans before and during circadian misalignment produced by an abrupt 9-h delay in the sleep/wake schedule

Participants were recruited with on-line ads and flyers from September 2014 through June 2016. An initial questionnaire excluded a majority of these individuals as they did not meet inclusion criteria: having a BMI less than 35, being a non-smoker, not being Hispanic or Latino and both biological parents being either Black/African-American or White/European-American. Follow up phone calls and in person interviews excluded other participants who did not meet more stringent inclusion criteria such as a clean drug screen and no reported health or sleep issues. We did not keep track of how many people applied to be in the study. There were 53 people who were enrolled and who signed consent forms three to four days before the start of the 14-day laboratory study which occurred between September 2014 and June 2016. Of these, 47 started the 14-day study, 46 completed the study and 45 were included in the current analyses. One participant (male) self-identified as Black/African-American and indicated that both parents and all four grandparents were Black/African-American. But we did not use his data because his ancestry DNA results (see below for methods) indicated he was 30% Sub-Saharan African, 56% European and 14% Indigenous American, and thus could not be included in the African-American group or the European-American group. The 45 participants were between 18 and 44 years old. There were 18 people who participated in the first study [19] and 9 to 33 months later participated in the second study [20].

Prior to entry to the study, participants completed a Family/Ancestor Questionnaire (). This questionnaire asks participants to mark all of the following categories that applied to them, their biological parents, and all four grandparents: White, Black or African-American, Asian, Hispanic or Latino, European, Middle Eastern, Far East Asian, Indian Subcontinent, North African, Afro-Caribbean, American Indian or Alaska native, Native Hawaiian or other Pacific Islander, Other, Don’t Know. Participants self-identified as being either African-American (n = 23) or white (n = 22) and none self-identified as being “Hispanic or Latino”. S1 Appendix

DNA samples were collected from Buccal (cheek) swabs and were analyzed (Ancestry by DNA, DNA Diagnostics Center, Fairfield, OH) to confirm self-reported ancestry. Greater detail about the Family/Ancestor Questionnaire and DNA samples is reported in Eastman et al. [20]. This company performed biogeographical ancestry estimates based on 176 ancestry informative markers, also known as population-specific alleles, which show large frequency differences between populations [24, 25]. Results were returned several weeks later with percentages for each subject in four categories: European, Sub-Saharan African, East Asian and Indigenous American (See the first table in [20]).

Participant demographics are shown in Table 1. Participants completed the Morningness-Eveningness Questionnaire (MEQ) [26] and the Munich Chronotype Questionnaire (MCTQ) [27]. As shown in Table 1, African-Americans were more morning-type (higher scores on the MEQ) than European-Americans (p<0.05). As shown in the first table of [20], there were 12 morning types and one evening type in the African-American group and five morning types and two evening types in the European-American group.

Subjective socioeconomic status (SES) was determined using a socioeconomic status ladder [28] which consists of a 10-rung ladder. The top of the ladder represents those with the most money, most education and best jobs. The bottom represents those who have the least money, least education and worst or no job. Participants were instructed to place themselves on the ladder in terms of where they think they fell. This position was then translated into a score between 1 and 10, where 1 = bottom rung (i.e., lowest SES) and 10 = top rung (i.e., highest SES). In this way, higher scores are representative of a higher perceived SES.

All participants were physically and psychologically healthy as determined by a Health Information Questionnaire (S2 Appendix), a For Women Only Questionnaire (S3 Appendix), the Beck Depression Inventory [29] and the Minnesota Multiphasic Personality Inventory (MMPI-2) [30]. Participants also completed the Pittsburg Sleep Quality Index (PSQI) [31], the Epworth Sleepiness Scale (ESS) [32], the Berlin Questionnaire [33] and the Insomnia Severity Index [34] to determine if they had sleep disturbances.

Participants were non-smokers and were free from medication with the exception of six women (3 African-Americans and 3 European-Americans) who were taking oral contraceptives. Participants were low to moderate consumers of caffeine (≤ 300mg caffeine/day; equivalent to < 2cups coffee/day) and alcohol (≤ 3 standard drinks/day). Participants did not work night shifts in the month prior to the study. Participants were asked to abstain from caffeine in the four days preceding the study.

The study was approved from the Rush University Medical Center Institutional Review Board in accordance with standards set by the Declaration of Helsinki. All participants gave written informed consent to participate in the study. All were made aware that participation was completely voluntary and that they could withdraw at any time. Upon completion participants were given a financial compensation for their participation.

The current analysis is of data from our study which involved a delay of the sleep/wake schedule [20], however, all methods are the same as our study which involved an advance in the sleep/wake schedule [19]. Both studies took place at the Biological Rhythms Research Laboratory in Chicago, USA. Participants completed the protocol in groups of two or three, usually with a mixture of African- and European-Americans in each group. The current report focuses on the last 10 days (labeled as days 1–10, Fig 1) of the larger 14-day study with the delay of sleep [20]. To see the protocol diagram for the full 14 days see Eastman et al. [20]. Participants remained in the laboratory for 14 days, with the exception of one 8-h break following the first baseline sleep (Fig 1, day 2) during which participants were free to leave the laboratory if they wished. While participants remained in the laboratory, they were continuously monitored by research staff. Caffeine and alcohol were not permitted while in the laboratory or during the 8-h break. Participants were reminded not to take any naps during the 8-h break. Upon re-entry to the laboratory participants were given a urine drug screen and were breathalyzed. While participants were not required to leave the laboratory, a majority of participants chose to do so.

Participants were scheduled to four baseline days with 8h time in bed (TIB) (Days 2–5 Fig 1). Baseline sleep schedules were tailored for each participant to best match their habitual schedule measured with sleep logs completed prior to the start of the study. Starting on day 6, the sleep-wake (light-dark) schedule was shifted 9 h later (i.e., delayed) as though participants had flown from Chicago to Japan which is 9 time zones west (Fig 1).

During baseline and delayed days, participants remained in a bedroom/testing suite. Participants had their own bedrooms, with separate, externally controlled lighting. Light levels were set to the maximum level (maximum ~ 500 lux, median; 66 lux at angle of gaze) during the first 10 h of wakefulness and dimmed to the lowest level (maximum < 100 lux, median; 17 lux) for the last 6 h of wakefulness. Rooms were completely dark during the 8-h sleep episodes. This simulates, as much as possible in our laboratory, the experience of people who may be exposed to brighter light for the first 10 h that they are awake, and then during the 6 h after sunset (when they are still awake) are only exposed to indoor artificial lighting which is less intense. Temperature was maintained at a consistent level (73 ± 2°F or 23 ± 1°C). Participants were allowed access to cell phones, electronic devices (e.g., laptops, tablets), and time pieces (e.g., watches) during waking episodes. All devices were turned off during performance testing and were removed from rooms during sleep episodes. Each bedroom had a wall clock indicating the ‘local’ time of either Chicago (baseline days) or Japan (delayed days).

Meals were served at regular intervals relative to each participant’s waking time starting on day 3; 1 h (breakfast), 6 h (lunch), and 12 h (dinner) after waking. Participants were allowed up to two small snacks (≤160 calories each) each day. Beginning 2 h after waking, participants completed a test battery every 3 h (Fig 1). Performance testing (described below) was conducted relative to each individual’s wake-time. There were five tests per day which were given during the last two baseline days and the first two advanced days (Fig 1). The schedule (meal timing, testing relative to waking time) remained the same during baseline and delayed days (Fig 1).

The method and results pertaining to circadian phase were reported in our previous publication [20].

Circadian phase markers are only presented here (Fig 2) to illustrate the enormous amount of circadian misalignment produced in this study. The method for determining circadian phase is described briefly below. On days 5 and 9 (Fig 1) participants remained seated in comfortable recliners under dim light conditions (<5 lux) during which the dim light melatonin onset (DLMO) was assessed. Saliva samples were collected every 30 mins using Salivettes (Starstedt, Newton, NC, USA). Samples were centrifuged, frozen, and later sent to Solid Phase Inc. (Portland Maine, USA) to be radioimmunoassayed for melatonin [35]. Based on prior work showing that the core body temperature minimum (Tmin) occurs approximately 7 h after the DLMO [36–38], Tmin was estimated by adding 7 h to the DLMO for illustration purposes (Fig 2).

Sleep was measured via wrist actigraphy and sleep diaries. Participants wore activity monitors (Actiwatch Spectrum, Philips Respronics, Bend Oregon, USA) on their non-dominant wrist for the duration of the study. Sleep was recorded in 1-min epochs and data analyzed using the Philips Actiware-6 software package. Participants were required to complete their sleep diaries within 10 min of waking. Sleep diaries included information about sleep onset and offset times, and any wakefulness during the scheduled sleep episode. Early Morning Awakening (EMA) was the duration of time between the final awakening and scheduled wake time. In instances where a participant had been awake for longer than 2 h during the second half of the sleep episode, the EMA was manually calculated, even if they fell back asleep before the scheduled wake time. There were five occasions for four participants (3 African-Americans, 1 European-American) where the EMA was calculated manually. The primary outcome measure for both actigraphy and sleep diaries was total sleep time (TST). The use of actigraphy as a measure of TST has been validated against polysomnography (PSG) [39–41].

Participants completed the Automated Neuropsychological Assessment Metrics (ANAM) [42] test battery. The ANAM test battery, which was administrated on desktop computers, lasted approximately 20–30 min and consisted of nine different tasks. The ANAM tasks were completed in the following order: subjective sleepiness, mood, simple reaction time, code substitution learning, procedural reaction time, mathematical processing, matching to sample, code substitution delayed recognition, and Go/No-Go. All tasks within the ANAM battery, except the simple reaction time task which was a timed task, involved a set number of trials with the task ending upon completion of all trials. For tasks where a correct response was required (procedural RT, code substitution learning and delayed recognition, mathematical processing, and the matching to sample task) only correct responses were included in data analyses.

Subjective sleepiness was assessed using the Stanford Sleepiness Scale [43] which is a 7-point Likert scale where 1 = “feeling very alert, wide awake, and energetic” and 7 = “very sleepy and cannot stay awake much longer.” Participants selected the statement that best matched their current feelings of sleepiness. Participants rated their mood using an abbreviated 7-dimension mood scale [44], containing a set of 24 items. Using a scale of 0–6, where 0 = “not at all”, 3 (midpoint) = “somewhat”, and 6 = “very much”, participants rated each item based on their current state. Scores were grouped into seven mood dimensions: vigor (high energy level), happiness (positive disposition), depression (dysphoria), anger (negative disposition), fatigue (low energy level), anxiety (anxiety level), and restlessness (motor agitation).

Sustained attention and reaction time (RT) were measured using a 10-min simple RT task, akin to the psychomotor vigilance task (PVT) [45], with an interstimulus interval of 2–10 seconds. Participants were required to respond as quickly as possible (pressing the left mouse button) to a visual stimulus (asterisk) displayed in the center of a blank screen. Lapses were defined as RT > 500ms. Participants also completed a procedural RT task to assess processing speed and visuomotor RT when following a set of rules. In this basic block version of the task, a single digit between 2 and 5 was presented in the center of the screen. Participants were required to indicate whether the number presented was ‘low’ (2 or 3) with a left mouse click, or ‘high’ (4 or 5) with a right mouse click. Slow responses (SR)–comparable to lapses–were defined as responses exceeding the 90^th percentile of the cumulative distribution of each participant’s baseline responses [46]. Main outcome measures for the simple RT and the procedural RT tasks were the median RT and number of lapses or slow responses.

Two versions of the code substitution test were administered non-consecutively. The first version, code substitution learning, was similar to the Digit Symbol Substitution Test (DSST)[47]. In this test, a single digit-symbol pair was presented at the bottom of the screen. Participants were required to indicate whether the pair was correct (left mouse click) or incorrect (right mouse click) relative to a set of 9 defined digit-symbol pairs (i.e., the key) displayed at the top of the screen. Feedback was provided following each response. The second version, code substitution delayed recognition, was identical to the first, however the key was not displayed at the top of the screen. This test was presented several minutes after the learning version, after three intervening tests. Participants were required to determine whether the displayed digit-symbol pair was correct, based on the key presented earlier in the learning version. These tasks assessed sustained attention, visual scanning, associative learning and visual memory.

A mathematical processing task assessed basic computational skills and working memory. Participants were presented with a simple arithmetic problem (e.g., 4 + 8–5) and were required to indicate whether the answer was greater than (right mouse click) or less than (left mouse click) 5. Visuospatial working memory and processing was measured with a matching to sample task where a pattern 4 x 4 grid pattern with light and dark shaded cells was presented. Following a brief delay (5 sec) during which the screen was blank, two comparison grids were shown side-by-side. Participants indicated with a left or right mouse click, which grid matched the preceding grid. The primary outcome measure for the two code substitution tasks, mathematical processing, and matching to sample was the percent correct responses (number correct responses/number of trials*100).

Response inhibition was assessed with a Go/No-Go (GNG) task. One of two stimuli (“x” or “o”) were presented and participants were required to respond as quickly as possible to the “x” stimuli (i.e., “go”) but to do nothing, or inhibit the response (i.e., “no-go”), in response to the “o” stimuli. The number of correct responses (i.e., “hits”), incorrect responses (i.e., “false alarms”) and incorrect non-responses (i.e., “misses”) were extracted and a d-prime discriminability value was calculated (d’ = Z(hit rate)–Z(false alarm rate)). The d’ value, which was the primary outcome measure for this task, reflects the overall ability of the participant to discriminate between the go and no-go stimuli [48, 49].

Participants also completed a Columbia Jet Lag scale [50] each day prior to each sleep episode. This scale contains a set of nine items relating to sleepiness, fatigue, daytime alertness, and concentration. Participants rated how they had felt during the entire wake episode for each item using a scale of 0–4 where 0 = “not at all” and 4 = “extremely”. A total jet lag score was calculated from the sum of all the items.

All data were analyzed using SPSS v.23 for Windows. Separate mixed model ANOVAs were performed to assess the main and interaction effects of condition (baseline or delayed) and ancestry (African-American or European-American) on sleep and performance. Regardless of main and interaction effects, separate mixed model ANOVAs were performed to assess the effects of ancestry on each day (sleep and Columbia Jet Lag Scale) or each hour after baseline wake time (cognitive performance measures). These additional analyses were performed to investigate how cognitive performance varied across hours of wakefulness or across days in the study. All models were performed on all variables (sleep and cognitive performance measures). As many of these variables were nearly identical to each other, only a subset of commonly used cognitive performance measures are reported in the results. Due to device error, actigraphy data was lost for two participants (1 African-American, 1 European-American). Data on the simple RT task for three individuals (all African-Americans) were excluded due to non-compliance. All models included participant ID as a random effect. Significance was assumed at p < 0.05.

Following baseline days, DLMOs and Tmins were in a normal phase relationship to sleep. With the exception of three participants, all DLMOs occurred before scheduled bedtime (Fig 2, top panel) and the estimated Tmins occurred within the sleep episode with the exception of one participant (Fig 2, bottom panel). In contrast, following the three delayed days on Japan time, DLMOs and Tmins for all participants were misaligned relative to the sleep episodes. While all the DLMOs occurred before the delayed sleep period, these occurred several hours earlier than usual (Fig 2, top panel). Further, most of the Tmins occurred before the scheduled sleep episode, i.e. at the wrong circadian phase (Fig 2, bottom panel). The magnitude of the phase delay for each individual is shown in Eastman et al. [20]. Most participants delayed between 0 h and approximately 4 h, with the largest delay being around 6 h. None of the participants delayed the whole 9 h, which would be required for complete circadian alignment with the delayed sleep/wake schedule. It should be noted that these phase delays were measured more than 24 hours after the last performance test. That is, the final phase assessment occurred after a full day with no performance measures. Therefore the phase delays that occurred during the time of performance testing and sleep measures would have been even less than that shown in Fig 2, or the fourth figure in [20].

Fig 3 shows sleep duration for the last three baseline days (days 3–5) and the three delayed days (days 6–8). There were no main effects of ancestry or the interaction between ancestry and condition on any of the sleep variables; however, there were main effects of condition on all sleep variables (Tables 2 and 3). Compared to baseline/aligned days, TST was shorter, while EMA and Columbia Jet Lag Scale scores were higher during delayed/misaligned days (Tables 2 and 3). Across days, there were no differences between African-Americans and European-Americans except for Columbia Jet Lag Scale scores on day 7 (Fig 2) where European-Americans reported feeling more symptoms of jet lag.

Fig 4 shows cognitive performance outcomes and Fig 5 shows subjective sleepiness and fatigue. There were no differences between African-Americans and European-Americans on any of the cognitive performance or subjective variables except for happiness (Table 2). On average, African-Americans reported feeling happier than European-Americans (Table 3). Compared to baseline/aligned days, cognitive performance and mood worsened on delayed/misaligned days on almost all measures (Tables 2 and 3).

During baseline/aligned days, cognitive performance remained relatively stable across hours of wakefulness, with slight decrements as the day progressed (Fig 4, left panels). Similarly, during baseline/aligned days, there was a slight increase in fatigue and sleepiness ratings as the day progressed (Fig 5, left panels). During delayed/misaligned days, however, cognitive performance worsened even more as the day progressed and was most impaired at times corresponding to the end of the baseline sleep episode (Fig 4, right panels). Likewise, during delayed/misaligned days, subjective sleepiness and fatigue also increased even more as the day progressed peaking around times corresponding to the end of the baseline sleep episode (Fig 5, right panel). There were no differences between African-Americans and European-Americans during either baseline/aligned or delayed/misaligned days on any of the performance measures (Fig 4). In contrast, during baseline/aligned days, European-Americans reported feeling slightly sleepier and more fatigued compared to African-Americans at several time points (Fig 5, left panels). There were no differences between the two groups in subjective sleepiness and fatigue during the delayed/misaligned days (Fig 5, right panels).

All relevant data are within the paper and its Supporting Information files.