The participants in this randomized, crossover study were 15 healthy adult females (mean age ± standard deviation [SD], 21.7 ± 0.5 years; body mass index, 19.81 ± 2.45 kg/m2) recruited between August and September 2018. None of the subjects had any previous night shift experience, and none was identified as morning type or evening type according to the morningness–eveningness questionnaire . All of the participants were current nonsmokers, in the luteal phase of their menstrual cycle [12,13], and had normal sleep patterns (habitual sleep ranging between 7 and 9 h). No participants were currently taking any prescribed medications. The required sample size was determined to be 14 (actual power 81.0%) based on an effect size, a error, and power (1-b) of 0.25, 0.05, and 0.8, respectively. The power calculation in this study was carried out using G*Power 3 .
All participants consumed one or two meals between 16:00 and 09:00. Under both the Snack and Skipping conditions, all participants ate a meal (containing 708 kcal, 19.4 g protein, 17.9 g fat, 112.4 g carbohydrates, and 3.5 g sodium) at 19:30 (not eating at night), and under the Snack condition only, participants ate a snack at 03:30 (eating at night). The meal provided at 03:30 consisted of two rice balls (containing 352 kcal, 5.8 g protein, 1.8 g fat, 75.8 g carbohydrates, and 1 g sodium) and two slices of yellow pickled radish. Many night shift workers prefer a high-energy diet rich in carbohydrates , such as the rice balls that are widely available at any convenience store in Japan. All participants were given 20 min to finish their meals, and they were encouraged to eat everything on their plate.
A flowchart of this study is shown in Figure 1. The measurements were conducted over 2 consecutive days between 16:00 and 09:00. Each experiment day had 5 participants who were randomly assigned to one of two conditions using counterbalancing. All participants were instructed to refrain from strenuous physical exercise and not to consume caffeine or alcohol for 24 h prior to and during each study day. Two days before the experiment began, all participants wore an actigraphy monitoring device (ActiGraph; Ambulatory Monitoring Inc., Ardsley, NY, USA) on their non-dominant wrist and recorded their sleep and activity levels in a diary.
On the day of the experiment, all participants arrived at the laboratory at 15:00. Until 16:00, they carried out practice assessments, including the Uchida-Kraepelin test (UKT) and the psychomotor vigilance test (PVT). The Snack condition involved consuming a meal and a snack at 19:30 and 03:30, whereas the Skipping condition involved consuming a meal at 19:30 only. At the start of each experiment, the participants were fitted with a heart rate variability (HRV) sensor (GMS Inc., Tokyo, Japan). For each hour throughout the experiment, they were given 10 min to record their sublingual temperature once and complete the visual analog scale for sleepiness, fatigue, and hunger, 10 min to perform the UKT, and 10 min to measure the PVT. The next 20 min were considered free time, and the remaining 10 min were considered a rest period. The participants spent their free time reading, drawing, or drinking water. During the 10-min rest period, they sat on chairs and chatted with the other participants. The same meal amounts and contents were given to the participants between 19:30 and 19:50 in each experimental period. The HRV sensor was removed at the end of each experiment, but the participants were asked to continue wearing the actigraphy monitoring device until they woke up the next day. During all waking times, the participants remained awake in the laboratory and were continuously monitored by the researchers.
All participants stayed in a windowless and sound-insulated laboratory for 2 consecutive days (1 night) (Figure 1). The laboratory environment was maintained at 26 ± 2 °C  and 50% relative humidity under indoor illumination on the table at 200 lx.
The sleep parameters measured were total sleep time, sleep efficiency (total sleep time / time in bed ´ 100), sleep onset latency, and wake after sleep onset. All parameters were measured using the actigraphy monitoring device, and they were analyzed using the AW2 software package (Ambulatory Monitoring Inc.).
The circadian rhythm of body temperature is one of the most frequently used indicators of circadian rhythmicity , and body temperature has been shown to be related to sleepiness, fatigue, and performance of a single-digit mental arithmetic task . Sublingual temperature, which is considered an index of internal body temperature , was measured hourly using an oral thermometer (MC-612; Omron Inc., Kyoto, Japan) to assess changes in circadian modulation during the night.
Subjective assessment of sleepiness, fatigue, and hunger
A visual analog scale was used for the subjective assessment of sleepiness, fatigue, and hunger . The participants rated their sleepiness, fatigue, and hunger every hour on a 100-mm line, with values ranging from 0 mm (not sleepy, tired, or hungry at all) to 100 mm (extremely sleepy, tired, or hungry).
Uchida-Kraepelin Test (UKT)
The UKT (Nisseiken, Tokyo, Japan), a serial mental arithmetic task, was used to measure cognitive performance. This test is a questionnaire that requires intense concentration and effort, and it has been used as a tool to induce mental stress . The test material consisted of a pre-printed paper with 20 rows of 115 random, single-digit figures. The subjects’ task was to add adjacent figures horizontally, and then write the one-digit answer between the 2 figures; they were asked to proceed along each row as quickly and as accurately as they could in a 1-min period. On being given the first cue, the subjects began calculating from the first row. Then, when the second cue was given after 1 min, the subjects were required to begin a new row, without regard to their position on the current row. This procedure was repeated 8 more times, for the total duration of 10 min. The sum of the correct answers for each 1-min period over the 10-min task was used as the value for the analysis.
Psychomotor vigilance test (PVT)
The PVT is a reaction time task considered to be a sensitive measure for assessing the effects of sleep loss . In this study, a precise computer-based version of the 10-min PVT was used to avoid problems of uncertainty with regard to the accuracy of the test platform timing . All participants were instructed to look at a computer monitor and press a response button when a white circular edge appeared on the screen; pressing the response button stopped the counter and displayed the response time (in milliseconds) for a 1-s period. The PVT measures response times to visual stimuli randomly occurring at 2- to 10-s intervals over a 10-min period . The outcome measures for the PVT include the median response time, number of lapses (response time > 500 ms), and total errors (incorrect responses), as well as mean response time.
Autonomic nervous system activity
For the purposes of the present study, HRV was obtained through autoregressive analysis of R–R intervals measured between 16:00 and 09:00. All data were analyzed offline after analog-to-digital conversion of 250-Hz R–R waves. HRV was measured every 5 min during each hour and then averaged; these measurements were used to monitor autonomic nervous system activity throughout the night . High-frequency (HF) and the low-frequency/high-frequency (LF/HF) are used as indicators of cardiac parasympathetic and cardiac sympathetic nervous activity, respectively [25,26]. The LF/HF power ratio indicates the balance between sympathetic and parasympathetic outflows .
All results are shown as mean ± SD or standard error of the mean. All sleep variables measured the day before the experiment were analyzed using the t-test.
To test the effects of consuming a snack on neurobehavioral and physiological outcomes during the early morning measurement periods, a fully saturated, linear mixed-effects analysis of variance was carried out , with a between-participant fixed effect of condition and a within-participant fixed effect of time (at 03:00 vs. from 04:00 to 09:00) and a random intercept. Within-condition comparisons were used to minimize the effect of individual differences. Multiple comparisons were assessed using the Bonferroni correction to evaluate patterns of change under the two conditions. As a secondary analysis, between-condition hourly comparisons from 16:00 to 09:00 were analyzed using the Mann-Whitney U test.
To assess the postprandial effect of the meal throughout the testing time, the net incremental area under the curve (niAUC), calculated from pre- (at 03:00) and postprandial time points, was tested. All statistical analyses were conducted using SPSS (version 22.0J; IBM, Tokyo, Japan). The hypothesis rejection level for all tests was set at p < .05, and a notable trend was set at p < .1.
This study was approved by the Center for Integrated Medical Research of Hiroshima University (study protocol ID No.: C-252). Written, informed consent was obtained from all participants before the first examination. The study protocol conformed to the Declaration of Helsinki guidelines. This study was registered with the University Hospital Medical Information Network-Clinical Trials Registry (UMIN-CRT registry ID: UMIN 000034345) after the enrollment of participants had begun. The authors confirm that all ongoing and related trials for this intervention have been registered.