Effects of a Single Bout of Exercise on Memory and Attention Following One Night of Sleep Loss: Results of a Randomised Controlled Pilot Study


 BackgroundSleep-loss is a severe problem in night-shift workers. It causes fatigue and a decrease in awareness that may be counter-acted by exercise. This randomised controlled study in 22 university students investigated the effects of exercise on cognitive and physical performance following sleep deprivation. MethodsWe were comparing a single bout of a 20-minutes circuit training to control in an experimental setting of overnight sleep loss. Outcomes included memory, cognitive tasks, and physical parameters. The occurrence of false memories was considered being the main outcome.ResultsExercise did not exert significant effects on false memories (p = 0.456). We could detect a trend to significanxe (p < 0.01) assessing cognitive dimensions, i.e. selective and sustained attention, and visual scanning speed. This revealed strong effects of exercise on attention (p = 0.091; Cohen’s d = 0.76; ∆14%), cognitive performance, performance speed, and perceived sleepiness (p = 0.008; d = 0.60; ∆2.4 cmVAS). ConclusionsThis study failed to show effects of exercise on memory function. Still, medium to strong effects on attention and consciousness can be considered clinically relevant. The results of this study encourage further research to determine its practicability and meaningfulness among night-shift workers.Trial registrationGerman Clinical Trials Register, DRKS00010655, registered 21. June 2016.

cognitive performance of nurses [10], and shooters show a degraded marksmanship performance [11]. A study in polysomnography technicians showed one 12 hours night shift to be already su cient to increase perceived sleepiness, reaction time and the number of lapses of attention [12].
Exercise has been emphasized being an approach to motivate tired populations and to improve the quality of sleep. Among workers with work-related fatigue, a 6 weeks exercise intervention improves employees' well-being, and reduces the degree of emotional exhaustion [13]. This is in line with a trial showing effects on study-related fatigue among university students [14]. Exercise and physical activity seem suitable to improve the quality of sleep in a variety of speci c indications, e.g. pregnancy [15], concerned parents of children with cancer [16], or female university students [17]. There is one study, showing 16 weeks of aerobic exercise to improve daytime function and reduce depressive symptoms besides improving sleep in older adults with insomnia [18]. Finally, in the elderly, physical activity has been suggested to improve cognition [19] and to prevent people from cognitive decline [20].
In summary, literature suggests general bene ts of exercise on symptoms following sleep loss [21]. Exercise is easily applicable and would thus be a valuable option for shift-workers to be performed after the shift before being exposed to decision-making and cognitive tasks in their daily life. This exploratory pilot study was designed to determine if a single exercise intervention is able to address both, functional and mental skills in healthy subjects following sleep loss.

Study Design
A single-centre randomised controlled study investigating the effects of a 20-minutes moderate-tovigorous circuit training on parameters of cognitive performance when compared to a passive control.
The study took place in the summer term 2016 at the Sports Campus of the Goethe-University Frankfurt am Main, Germany. The study was approved by the Ethics Committee FB05 of the Goethe-University of Frankfurt, Germany (reference 2016-23) and is in agreement with the Declaration of Helsinki (Version Fortaleza 2012). The study was registered on June 21st 2016 at the German Clinical Trials Register (DRKS00010655). Written informed consent was achieved by all participants Participants checked in at 18:00 the day of the experiment, starting with a collective standardised Mediterranean dinner (grilled chicken, salad, vegetables, water). Eating was stopped at 20:00 and after two hours of digestion, baseline parameter were assessed at 22:00 (see Fig. 1). In the following night, sleep loss of the participants was assured by the study team. The participants watched two mainstream movies, were allowed to talk, to play cards or other board games that did not require physical activity or high intellectuality. Slices of vegetables (cucumber, peppers, and tomatoes) were offered till 4:00. At 6:30 participants were randomised into two groups with the "exercise group" leaving to a nearby gym where they participated in an exercise program, whereas the "control group" remained passively seated in the common area. At 7:30 all outcomes were re-assessed and the study ended thereafter.

Participants
Participants (> 18 < 36 years) with good sleep (6 to 8 hours/night) and no abuse of substances were assessed for study eligibility using the following exclusion criteria: pregnancy or lactation, intake of neuroenhancers, ability for smoking cessation less than 24 hours, suffering from diabetes, severe illness in uencing the quality of life, or severe illness in uencing the mental or physical performance. All participants signed a written informed consent prior to enrolment.
All participants have been recruited advertising at the Campus two weeks prior to the experiment.
Randomised treatment allocation, blinding and sample-size estimation Participants were randomly assigned to one of the two study groups using the smartphone-based application Certi ed True Randomizers (Integer Generator, Random.org, Dublin, Ireland).
Sample size was estimated a priori using the software G*Power (Version 3.15, University of Düsseldorf, Germany). With α set at 5%, 22 participants are required to have 90% power to detect a decreased number of false memories in the intervention group (estimated 10%) as described for the use of caffeine. [22] Exercise Intervention The exercise intervention was performed after sleep deprivation at approx. 7:00. Participants exercised together and performed a 20-minutes lasting circuit training under the guidance of two trainers. First, Warm-Up was performed for 5 minutes and consisted of a series of ve movements, each for 60 seconds, i.e. marching on the spot, side to side weight stabilising, step touches, double step touches, and V-steps.
Two of the included elements derived from Yoga, i.e. sun salutation and the warrior, the other three were rotation of the vertebra, and stretching of chest and the quadriceps muscles. An internal assessment of this program among Master students of exercise physiology prior to this study revealed the exercise to be of vigorous intensity (60-80% HRR).

Outcome measures
Demogra cs included information on smoking, drinking behaviour (alcohol, coffee, caffeine, energy drinks), the extent of physical activity, and the quality of sleep (Pittsburgh Sleep Quality Inventory PSQI, German version) [23]. The PSQI ranges from 0 to 21 with higher scores indicating a decreased quality of sleep, and a PQSI > 5 being a cut-off value for good and bad sleepers.

Main Outcome Measure
Based on the sample size calculation the rate of false memories dating from the Desse, Roediger and McDermott (DRM) paradigm [24] were chosen as primary outcome, and implemented as an experimental False Memory Task by Diekelmann et al. [22]. Participants learned 18 DRM lists the evening prior to sleep loss. Each list consisted of 15 semantically associated words (e.g. apple, vegetables, orange, juice, …). The list words were presented sequentially by a human female voice with a delay of ten seconds between lists and 750 ms between words, at the end of the baseline assessments in the evening previous to sleep deprivation. Retrieval took place the next morning in the common area, 108 words were projected in white letters on a black background at the wall. Half of these words originated from the evening before, 36 words were distractors not matching the list topics, and 18 words were theme words that had never been presented before (so called false memories). Participants gave an old/new judgment for each word (i.e. to indicate whether the word had been presented during learning or not), a con dence rating for their answer on a 4-point scale ranging from 1 (''I had to guess'') to 4 (''absolutely sure''), and a Remember/Know/Guess judgment for words judged as "old". The ratios of false memories, true recognition and false alarms were calculated according to Diekelmann [22].

Secondary outcome measure
Secondary outcomes included psychologic as well as physical parameters. The d2 test of attention was used to assess selective and sustained attention, and visual scanning speed [25]. It consists of a general attention score (total errors per total task, in %), with lower scores indicating higher performance. Subindices include performance speed (total number processed, max. 299), and concentration performance (speed minus error count), with higher scores indicating improvement.
The severity of sleep deprivation was assed using the Stanford Sleepiness Scale, ranging from 0 to 10 cm (with 0 cm being "fully awake", and 10 cm "sleeping") [26].
The German version of the Multidimensional Mood Questionnaire (MDBF; [27]) was used to assess three different dimensions of stress and anxiety, i.e. good-bad mood, awake-tired and calm-nervous, which are supposed to be expressions of mood, mental fatigue, and feelings of restlessness. Higher scores indicate a positive attitude.
Physical parameters included grip strength using a manual dynamometer (JAMAR, Homecraft Ltd, Nottinghamshire, UK), pressure pain threshold (PPT) with a mechanical pressure algometer (pdt, Rome, Italy; range 2-20 kg/cm2, diameter 1 cm) upon the belly of the transversal trapezius muscle, as well as the Balance Error Scoring System (BESS; [28]). The BESS is a standardised battery investigating balance in three different stances (double-leg, single leg, and tandem stance) on rst a rm and second a foam surface (range 0-60 points, with 60 indicating worst balance).
The sequence of assessments was in the order as described above, with the exception that the D2-test and the test for false memories were performed as the last two assessments.

Statistical Analysis
Statistical analysis was conducted for comparison of the primary and secondary outcome measures between the two study groups. No evidence was found that the parametric tests used were inappropriate. Baseline characteristics and single time point data were analysed with unpaired t-test (for continuous measures) and chi-square test (for nominal data) to assess for differences among the two study groups. Longitudinal data were analysed using a time x group repeated measures ANOVA (2 × 2), followed by an unpaired t-test of change scores for post-hoc comparisons. Effect sizes according to Cohen were calculated for all relevant outcomes. Signi cance with α < 10% was considered showing a statistical trend (pilot study), whereas < 5% indicated signi cant differences. Data are presented as mean ± standard deviation.
Data analysis was performed with the SPSS statistical software system, version 24.0 (SPSS Inc., Chicago, IL, USA).

Results
Twenty-two participants (8 female and 14 male, age 29.0 ± 3.0 years, weight 71.6 ± 10.7 kg, height 175.6 ± 8.7 cm) were included in the study and no dropouts occurred during the trial. After randomisation we could observe in each group three left-handed participants, and three participants with a PSQI score > 5 (range 6-11) indicating mild sleeping disorders. There were no differences between groups regarding the intake of alcohol, coffee, caffeine products, energy drinks and cigarettes. Time spent with exercise was similar in both groups. Handiness revealed. There was no signi cant difference of any outcome between groups at baseline (Table 1).
Time x group analysis of the d2 test showed a trend (p < 0.1) towards an improved cognition. Posthoc ttests for inner-subject effects sustained this trend for all d2 items in the exercise group, i.e. attention (p < 0.001), as well as concentration performance (p < 0.001), and performance speed (p < 0.001). The attention score improved within the exercise group by 17.  Table 2).
There were inner-subject effects on sleepiness (p = 0.019), with an increase of 0.59 ± 4.30 cm VAS in the intervention and 2.97 ± 3.54 cm VAS in the control group, with no signi cance between groups (posthoc unpaired t-test p = 0.171). The magnitude corresponds to a strong effect (d = 0.60, r = 0.29).
Covariate analysis including gender did not reveal differences in the time x group analysis. Female gender led to decreased errors in the balance test (∆-8.25 ± 5.44 versus ∆+2.71 ± 7.27, p = 0.029) in the intervention group. Pressure pain threshold was signi cantly reduced in female participants in the exercise group (p = 0.038), and grip strength in both female groups, exercise (p = 0.004) and control (p = 0.011). Gender had no effects on attention and false memories.
No harms have been observed within this study.

Discussion
Our study indicates strong effects of a single exercise intervention on the d2 cognitive and attention scores, as well as sleepiness following one night of sleep deprivation. The power of these assumptions is restricted due to the pilot character of the study, and as the global statistical test did only indicate trends but not a statistical difference based on an α-error < 0.05. The study was designed in line with the sleep loss experiment described by Diekelmann [22], outcome measures were all established and validated, and we controlled for important in uencers such as the incidence of sleeping disorders and the standardisation of meals.

Generalisability of effects
The experimental induction of sleep deprivation was successful, as demonstrated by an increase on the sleepiness scale in the control group by 3.0 points. This in line with other experimental set ups [22,29]. Perceived sleepiness in the exercise intervention group did not relevantly increase (0.59 points). Although this corresponds to a strong effect size (Cohen's d = 0.6) favouring the exercise group, we could not detect statistical signi cance. From a clinical perspective our results are comparable to a study, showing exercise to signi cantly decrease feelings of lethargy, a sensation that might be similar to sleepiness [30].
Exercise showed small effects on memory, however without being signi cant. It is to doubt that this effect is of clinical importance. Sleep is the best remedy for memory consolidation [31].
In contrast, the effects on cognitive attention in our paradigm indicate strong in uence by exercise and are of clinical relevance (mean improvement 15% in attention). One night on sleep deprivation is suggested to rather impact implicit but not explicit learning of serial-response time tasks [32]. This may be corroborated by our ndings in the d2 test, suggesting a higher performance speed. However, reduced errors as re ected in the d2 global score as well as an increased concentration performance could indicate that acute exercise-induced effects on memory and cognition are not only limited to this improved automatization. This is in accordance with a recent study showing acute exercise to improve executive cognitive functions in young adults [33].
Grip strength was assessed according to current principles, and the baseline force of the dominant hand of women (31.0 ± 4.0 kg) and man (48.2 ± 6.9 kg) is in agreement with German normative reference values [34]. Our data suggest an increase (2 kg) of maximal hand-grip strength. Even though grip strength measures have been validated to be robust in regard to test-retest settings, we cannot exclude effects of habituation. Still, an increase in grip strength could only be observed with in the exercised participants, in the control group, grip strength was reduced. Today, no link between sleep loss and grip strength has been reported. Consequently, effects of exercise on grip strength following sleep loss might be possible.
We could show that sleep deprivation mediates a decrease in pressure pain threshold in both groups, which is an indicator of facilitated pain perception. This is in line with previous ndings of sleep disorders and pain sensitivity [35]. Training studies have shown that acute resistance exercise [36] as well as 12 weeks of aerobic exercise [37] are able to increase the pressure pain thresholds in general, effects we could not con rm following sleep loss.
We did not detect differences in mood as assessed by the MDBF. Our intervention phase may have been too short to cause signi cant effects. There is one study reporting a single session of 40 min aerobic exercise (70% VO 2max ) to cause improvements in morning subjective mood states [30]. A study investigating a multimodal 8month program in caregivers including exercise sessions could observe positive effects on stress and negative affection [38].

Implications of the exercise program
The participants in our study performed a single bout of exercise. Studies dealing with the effects of acute exercise on cognition have usually not been performed adding sleep-loss as an additional factor.
As described above, Lluch et al. found acute exercise to improve mood and motivation [30]. It has been reported, that an acute bout of exercise improves the performance on an executive function task in individuals with schizophrenia [39]. High-intensity resistance exercise is more likely to reduce interference and improve reaction time in the evaluation of executive functions immediately after exercise, whereas low (40%) and moderate (70%) intensity improves performance on plus-minus tasks three hours after exercise [40]. The plus-minus task is a test for switching attention and is slightly comparable to the d2. As we did not assess the 3-hours follow-up period we are not aware if the effects of a single bout of vigorous exercise were even more pronounced at this time. An extended follow-up measurement should therefore be considered for future studies. A meta-analysis suggests acute bouts of exercise to reduce state anxiety [36], an outcome we did not assess in our study.
Finally, the timing of the exercise intervention possibly had an impact on the observed motor skills. It might be possible that exercise elicits larger effects when followed shortly after sleep, and not after sleep deprivation [31].

Limitations
We are aware that the lack of signi cances between groups limits the generalisability of ndings. Effectsizes are a well-established method to detect meaningful group differences, and effects found appear quite impressing. In addition with the small sample size and the character of the study design -i.e. a pilot study-, one should be aware that ndings are only hints and will need con rmation.
As detailed above, this study did not cover the complete range of psychological cognition tests. It was designed to get a rst insight into a complex understanding. Our study only observed one sleepless night in healthy and physically t participants, but we did not extend sleep loss to several nights to display more realistic scenarios of daily life.
In addition, we did not control for cardiovascular tness, as all of our participants were young academics at the sports campus, and their hand grip -correlating with oxygen uptake-was in the expected agerelated group. However, tness may moderate the interaction of acute exercise and cognition [41], and should be assessed on an individual level.

Conclusions
Acute circuit training following sleep loss seems to exert effects on attention and consciousness. These effects are small on the basis of a single bout of exercise, and it remains unclear whether they are meaningful for e.g. night-shift worker, or not. On the basis of this pilot study, it will be possible to perceive larger and longer-lasting prospective studies to evaluate the health impact of continuous exercise on physical and mental outcomes.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request (in compliance with data privacy).

Competing interests
All authors declare no con ict of interest.

Funding
There was no funding.

Authors' Contributions
JF: conception and design, data collection, analysis and interpretation of data, manuscript writing and nal approval of the manuscript. SG: data collection, manuscript writing and nal approval of the manuscript. WB: conception and design, nancial support, manuscript writing, nal approval of manuscript. All authors read and approved the nal manuscript.  Figure 1 Study Design. A single-centre randomised controlled study investigating the effects of a 20-minutes moderate-to-vigorous circuit training (see icons) on parameters of cognitive performance when compared to a passive control. Participants checked in at 18:00 the day of the experiment, starting with a collective standardised Mediterranean dinner. Baseline parameter were assessed at 22:00 and thereafter the list of Desse, Roediger and McDermott (DRM) words was presented.24 In the following night, sleep loss of the participants was assured by the study team. At 6:30 participants were randomised into two groups with the "exercise group" leaving to a nearby gym where they participated in an exercise program, whereas the "control group" remained passively seated in the common area. At 7:30 all outcomes were re-assessed, the DRM word were recalled, and the study ended thereafter.

Figure 2
Selective and sustained attention (D2-Score). The circles display the changes to baseline for exercise (solid black) and control (open grey circles). The d2 test of attention consists of a general attention score and subindices including concentration performance and performance speed.25 It consists of a general attention score (total errors per total task, in %), with lower scores indicating higher performance (left scale). Subindices (right scales) include performance speed (total number processed, max. 299), and concentration performance (speed minus error count), with higher scores indicating improved performance and speed. There is a statistical trend (time x group analysis, p < 0.1) that exercise (solid circles) improved attention (less errors), speed (more numbers processed), and performance (count of processed numbers minus errors) then control (open circles). p indicates the level of signi cance, d represents the effect size according to Cohen, suggesting that a d = 0.2 represents a 'small' effect size, d = 0.5 represents a 'medium' effect size and d = 0.8 represents a 'large' effect size.