Diurnal variation in variables related to cognitive performance: a systematic review

This systematic review adheres to the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) [27]. The corresponding checklist for PRISMA 2020 is provided in Appendix 1, indicating the page references for the information included in the present review.

The criteria for study inclusion were derived from the Cochrane guidelines for conducting systematic reviews [28]. These inclusion and exclusion criteria were established and unanimously agreed upon by all nine authors. After the initial screening of studies, three authors (AR, MM, and TB) independently evaluated the eligibility of each manuscript by examining the titles and abstracts in a standardized manner, ensuring blinding during the assessment process. To be deemed eligible, the manuscript had to meet the specified inclusion criteria:

A systematic search for English-language literature in the grey literature was performed at Liverpool John Moores University electronic library, Manipal Academy of Higher Education electronic library, and electronic databases (PubMed, Scopus, and Web of Science) from August 2021 to May 2022, concluding on May 21, 2022. The search aimed to find pertinent content concerning cognitive performance variables and their variation throughout the day, utilizing specific search syntax with Boolean operators in titles, abstracts, and keywords of indexed documents: (“time of day” OR “time-of-day” OR “daily rhythm” OR “daily variation” OR “daily fluctuation” OR “diurnal rhythm” OR “diurnal variation” OR “diurnal fluctuation” OR “circadian rhythm” OR “circadian variation” OR “circadian fluctuation”) AND (“cogni*” OR “cognitive performance” OR “attent*” OR “attention control” OR “sustained attention control” OR “selective attention” OR “accuracy” OR “alert*” OR “decision-making” OR “decision making” OR “reaction time”) was conducted (Appendix 2). The study employed supplementary advanced search methods, including the incorporation of wildcards, truncation, and proximity searching. As part of the secondary search (conducted by MM & TB), the reference lists of all included papers were manually screened to identify any additional relevant publications. Additionally, forward reference searching was conducted by exploring citations and authors to identify potential follow-up studies. To minimize potential selection bias, one author (SP) independently carried out the search for study selection. The PRISMA 2020 flow diagram [27] was used to illustrate the flow of papers, encompassing searches of databases, registers, and other sources, throughout the study selection process.

The article was included if the data from male participants could independently be identified in the case of the study population being both male and female. In instances where the abstract and/or the title did not provide enough information to indicate whether the article met inclusion criteria, the article was obtained and read by a third reviewer (SP), who determined the relevance of the manuscript for the review. For articles where the primary objective was not a TOD investigation, with a minimum of two timepoints (morning and evening), the manuscript was excluded. All conference abstracts, literature reviews, and letters to the editor were not included as such studies are not critically appraisable and/or methodologically-quality-assessable.

Various factors pertinent to research design and chronobiological studies were quantified, which included room temperature control, sleep patterns, food intake, light intensity, fitness levels, and the use of randomization and counterbalancing techniques [3‐5]. Each factor was recorded with a binary response of “yes” or “no,” while fitness levels were further categorized as trained or untrained. In cases where an article did not mention or refer to a specific factor, a negative response (no) was noted.

To evaluate the risk of bias in the study, two distinct tools were employed, following the Cochrane Scientific Committee’s quality assessment recommendations. The assessment of randomized studies utilized the risk of bias (ROB) 2.0 tool, while the ROBINS-I tool was applied to evaluate non-randomized studies. Although there were some similarities in features between both tools, they were primarily focused on specific outcomes. The evaluation involved fixed sets of bias domains, enabling an overall risk of bias judgment, with scores categorized from “low” to “critical.” Manuscript quality was independently assessed by two reviewers (AR and TB), who identified discrepancies in agreement across four domains of risk of bias among the 11 studies included in the review (5.6% of cases). To resolve these discrepancies, a third reviewer (SP) was consulted. For a clear visual representation of the results, Figs. 2 and 3 display a “traffic light” plot for each domain.

We initially identified 523 articles from primary database searches. Additionally, 1868 articles were found through organization searches (university databases) and citation searches. Figure 1 provides a breakdown of the number of articles found in each electronic database and other search methods, along with a comprehensive flow chart detailing the steps taken during the literature search. After eliminating duplicates, 444 titles from the databases were saved in the reference manager (Mendeley, Elsevier, Amsterdam, Netherlands). Subsequently, we thoroughly examined the titles, abstracts, and keywords of these manuscripts, resulting in 63 reports chosen for full-text analysis. Among these reports, 8 met the inclusion criteria and were included in the systematic review. Moreover, through organization searches and citation searching, we identified 45 additional reports that were evaluated for eligibility. Among these, 3 reports fulfilled the inclusion criteria and were deemed eligible, raising the total number of accepted studies to 11. Detailed explanations for exclusion can be found in Fig. 1.

Table 1 presents detailed characteristics of the participants across 11 studies, including a total of 151 male participants (with an average of 14 participants per study). The number of participants in each study ranged from 8 to 25. Among these studies, 63.6% focused on assessing circadian chronotype, with different questionnaires used, such as the morningness-eveningness questionnaire (Horne and Ostberg, 1976), the Composite Scale of Morningness [29], and a subjective amplitude scale. The results revealed that 77.1% of the participants were classified as having an intermediate chronotype, 11.0% as morning chronotype, and 11.9% as an evening chronotype. However, three studies did not report any information regarding the chronotype of their participants.

Morning sessions took place between 06:00 and 11:30 h, while evening sessions ranged from 16:00 to 21:10 h. In addition to these time-points, ten studies used extra timepoints for assessing diurnal variation. The number of time points assessed varied, meeting the inclusion criteria of at least two time-points. Among the cognitive aspects studied, reaction time was evaluated in 8 studies (72.7%), attention in 4 studies (36.3%), accuracy in 3 studies (27.2%), consistency in 2 studies (18.1%), vigilance in 2 studies (18.1%), and alertness in 1 study (9%). Various cognitive tests were utilized, including simple reaction time tasks, letter or sign cancellation tasks, signal detection tasks, badminton serves, dart throws, p300 tests, and selective and constant attention tasks.

In ten studies, performance variables exhibited TOD effects, with significant differences between morning and evening values. Four studies reported significantly better reaction times in the evening (ranging from 9.0 to 13.4%), while two studies found better reaction times in the morning (up to 34.2%). Two other studies found no differences in reaction times across different times of the day. Attention levels were found to be lowest in the morning in two studies (7.8% amplitude) and highest in the morning in two other studies (40.3% amplitude). Accuracy levels showed some variation, with one study reporting the highest values in the afternoon (14:00 h), another in the evening, and one observing no differences. The study that assessed consistency found better values in the evening, while the other study found no differences. Alertness also displayed diurnal variation, with the highest values observed in the late evening (20:00 h) compared to morning, afternoon, or early evening values by 7.3%.

Due to significant methodological and clinical heterogeneity among the studies, a meta-analysis was not feasible. Factors such as missing data, population differences, metrics, outcomes, and study designs made it impractical to pool the data for a meta-analysis. Moreover, the relatively low number of studies (11) with small average sample sizes and high heterogeneity would likely lead to underpowered results and challenges in detecting significant effects. Consequently, the study presented unweighted results and did not pursue a meta-analysis, considering the potential for compounded errors and inappropriate summaries.

Table 2 presents comprehensive information concerning various aspects of research design such as randomization, counterbalancing, light intensity recording, meal control, room temperature control, and sleep and fitness regulation. These factors are particularly crucial in conducting chronobiological investigations.

It was observed that none of the studies fulfilled all the essential criteria for chronobiological research. Among the included studies, 5 of them implemented counterbalancing to minimize learning effects, while 8 studies conducted TOD sessions in a randomized order. Notably, 4 studies incorporated both counterbalancing and randomization in their protocols, while only one study lacked both. Regarding specific controls, the majority of the studies (N = 8) regulated meals and sleep, whereas less than half (N = 5) provided details about room temperature control. Remarkably, only 3 studies effectively controlled all three aspects. All 11 studies did, however, furnish information about the “fitness” levels of their participants, who were either healthy males or sports players.

Three non-randomized studies utilized the ROBINS-I tool (refer to Fig. 2), and the detailed findings are available in the same figure. All three studies exhibited a low risk of bias in the classification of interventions (domain 3) and deviations from intended interventions (domain 4). Additionally, they had a low risk of bias due to missing data (domain 5) and in the selection of reported results (domain 7). The level of bias associated with participant selection ranged from low to moderate (domain 2), while bias arising from confounding (domain 1) and bias in outcome measurement showed moderate risk (domain 6). In conclusion, two of the studies received a low overall risk of bias judgment, while one study obtained a moderate overall risk of bias judgment.

A total of eight studies employed the risk of bias (ROB) 2.0 tool (see Fig. 3). Across all studies, there was a low risk of bias due to missing outcome data (domain 3) and the selection of reported results (domain 5). Regarding deviations from intended interventions (domain 2), the risk of bias ranged from low to moderate, while the randomization process (domain 1) and the measurement of outcomes (domain 4) showed some concerns regarding bias. In summary, all eight studies exhibited some concerns regarding the risk of bias across all domains.

Four studies have investigated the TOD effects on attention (Table 1) [20, 22, 30, 31]. Two studies reported better selective attention and constant attention in the morning (08:00 h) using a selective attention test, with values declining as the day progressed potentially due to the training experience of the players recruited in the various studies [21, 22]. Interactions around daytime sleepiness, time awake, and sleep build-up influence TOD aspects related to cognitive function [32, 33], thus suggesting that observations around attention are highly affected by sleep homeostasis. However, a study performed by Higuchi et al. (2000) reported reduced attention levels in the morning (08:00 h) compared to the late morning (11:00 h) which was sustained until late evening (20:00 h) using a P300 test. Another study performed by Souissi et al. (2019) reported reduced attention levels during the early morning (07:00 h), morning (09:00 h), afternoon (13:00 h), and late afternoon (15:00 h) compared to the afternoon (11:00 h) and evening (17:00 h) when using a number cancellation test. Overall, cognitive performance related to attention displayed contradictory findings; however, these variations can be attributed to the fact that different tests were used to assess attention, such as a P300 test, a selective attention test, and a number cancellation test, thus making it difficult to compare findings between different journals. It is well established that sleep inertia is affected by the circadian phase, and when subjective ratings of fatigue values are higher, visual and/or selective attention performance is negatively affected in the morning. In addition, in the post-lunch dip, sleepiness has been found to increase and attention has been found to decrease, with reduced alertness, subjective sleepiness, fatigue, and negative mood states increased [31, 34].

Two studies investigated TOD effects on vigilance, which varied with the outcome assessment type. When using an adapted sign cancellation test, vigilance is reported to be better in the late morning (10:00 h) and evening (18:00 h) than in the early morning (06:00 h) and afternoon (14:00 h) [35]. The observed variations in vigilance might be due to the improvement in visuomotor coordination [36] and core temperature [37]. The increase in motor contractile properties [38] during the day might be responsible for the increase in motor coordination and an increase in nerve conduction velocity [39] which in turn leads to better visuomotor coordination. When the letter cancellation test (LCT) is used, the LCT of 2 letters demonstrates several variables to display significant variation over TOD, with peaks occurring at different times of the day based on the performance variable examined. Similarly, an LCT of 3 letters demonstrated findings in-line with the observations present in the LCT of 2 letters [40].

Eight studies investigated TOD effects on reaction time (Table 1). Four studies demonstrated a faster reaction time in the evening which ranged from 9 to 13.4% when compared to other times of the day [17, 18, 20, 25]. In fact, late morning (10:00 h) values also showed better reaction times than morning, and afternoon values [20, 37]. Two studies reported faster reaction times in the morning than other times of the day when using a simple reaction time test, finding reaction times to reduce as the day progressed, with an amplitude of 34.1% [21, 22]. There was a decline in the reaction time performance post midday in comparison to morning, which might be due to the accumulation of tiredness after midday [21, 22]. These results are in line with a study that reported a fall in performance in the afternoon due to tiredness resulting from time awake [25]. The discrepancies between studies in the literature could be due to the training experience of the participants and due to differences in the population (trained vs untrained) recruited in the studies [22]. It has previously been established that amplitude of TOD differences between morning and evening is higher in trained compared to untrained individuals [41]. Furthermore, the level of training can also affect the reaction time performances, as suggested by a previous study which found that people who exercised regularly had faster reaction times compared to sedentary people [42]. Interestingly, two studies did not show any significant diurnal variation in reaction time [23, 43], due to the complexity of the tasks [23]. Nevertheless, time of the day and duration of time awake play a major role in response time observed [23] as reported by an earlier study that showed an increase in wakeful time and adverse circadian phases resulted in a delayed reaction time [44].

One study investigated the TOD effect on alertness and observed that alertness peaks in the late evening (20:00 h) by up to 7.3 % and was lowest in the morning (08:00 h). It is believed that the increase in body temperature might lead to physiological arousal that enhances cognitive performance as it modulates neurobehavioral performance [45]. Three studies investigated the effect of TOD on accuracy with discrepancies in the results present in all three studies. The study performed by Edwards et al. (2005) found better badminton serve accuracy values in the afternoon (14:00 h) compared to morning (08:00 h) and evening (20:00 h) in both short and long serves. These findings are like the findings observed in tennis serves, where both accuracy tests displayed high levels of test-retest reliability [24]. Another study found dart-throwing accuracy to significantly improve as the day progressed with the best accuracy observed in the evening (19:00 h) compared to the afternoon (15:00 h) and better than the morning (07:00 h) [25] in long-distance throws only. Similar observations were reported in the consistency of dart throws with the highest consistency present in the evening (19:00 h), compared to values observed in the early morning (07:00 h). As dart throwing requires a combination of hand-eye coordination and muscle contraction, when performing longer dart throws there is a larger emphasis placed on muscle contraction (strength), thus, findings established have observed TOD variations in line with core body temperature [26, 41]. In shorter throws, the emphasis is placed more on control mechanisms and factors related to fatigue, hence little to no variation of TOD established [15, 26]. However, one study did not show any significant difference throughout the day in accuracy for hits, false alarms, correct rejections and misses [18]. However, chronotype of the individual was found to affect accuracy, with evening types being more accurate than morning types in both morning and evening sessions. Accuracy is not parallel with the circadian patterns of body temperature with lower levels of accuracy present when temperature was at its highest [25]. These results depend on the skill level of the players recruited in this study [46], the fatigue levels due to time awake [15], changes in the “basal arousal.” Other reasons for the conflicting results may be related to the variation and lack of control in factors deemed important for TOD research (Table 2).

As far as we are aware, only three systematic reviews have looked into issues around chronobiology study design [3‐5]. In agreement with these previous reviews, an apparent lack of control in the research studies selected was also established within this review. It is well known that the periodicity of the body clock in human beings is influenced by external environmental rhythmic cues which impact the constant adjustment of the body clock (zeitgebers). In TOD studies, rhythmic cues such as activity, feeding/fasting, and light-dark cycles are the main factors that require additional control [47]. Light intensity was only reported in one study (9%), with no other studies reporting any information around light or dark exposure. The regulation of alertness and mood in human studies is highly affected by light exposure [47, 48]. Studies have observed that light exposure influences several cognitive processes related to attention, memory, and arousal [49‐52], with short-wavelength light negatively affecting reaction times [53]. Although there is a lack of clarity regarding whether or not light exposure results in increased cognitive performance for cognitive tasks which require sustained attention, light exposure is believed to improve such performance [47, 48, 54]. Therefore, there is a great importance to control light and/or dark exposure in cognitive studies. Three studies (27 %) failed to provide information around the control of meals, a factor which plays a vital role in cognitive performance. It has been established that “meals” potentially improve cognition and alertness [55], with the timing of meals, the characteristics of the meal, and the timing of meals affecting cognitive performance [56]. The size and macronutrient content of the meal influence mental performance, while the TOD at which a meal is consumed will affect cognitive performance [56]. In addition, alterations in meal timings have been shown to improve cognition [55]. The lack of standardization makes the comparison of results challenging. All studies reported information related to participant “fitness levels,” although it must be noted that personal characteristics of individuals can influence cognitive performance, such as age and level of training (sedentary vs. non-sedentary), which play a major role. It has been found that the level of training is closely associated with a better brain structure and brain functioning and, thus, results in better cognitive performance in trained individuals [57, 58].

When looking at sleep, 3 studies (27 %) failed to provide any information related to maintaining similar sleeping habits to “normal life.” No information was provided to participants to ensure habitual rising and waking times were maintained, that they should not stay up late, or whether any of the individuals had a prevalence of sleep insomnia or were sleep deprived. Sleep plays a significant role in cognitive performance and lack of sleep has shown to have negative effects on an extensive variety of performance variables related to decision-making, memory and attention [59‐61]. Reaction times and focused attention are worsened with one or two nights without sleep [59]. The presence of TOD in cognitive performance is highly associated with sleep homeostasis, time awake and previous sleep drive, and circadian rhythmicity, but how these processes interrelate is not well known. Diurnal variation in cognitive performance will differ in accordance with the “sleep status” of the individual. Individuals who are sleep deprived perform better around midday in tasks requiring episodic memory, while well-rested individuals showed more stable performance [60]. The time-since-last sleep is closely related to increased levels of fatigue, and as the amount of time awake increases, a negative effect is observed on cognitive performance, on restorative influences of sleep, and on arousal [62]. In addition, chronotype of individuals has also shown to play a role in simple and complex measures of cognitive performance [63]. It is well known that evening types have a significantly higher daytime sleepiness, thus resulting in worse cognitive performance in the morning when compared to morning types [16]. A total of 3 studies (27%) failed to assess chronotype scores in their participants.

Finally, other important factors to report are the mean ± SD of familiarization sessions. Not providing detailed and accurate statistics and information around this displays random and systemic bias. When counterbalancing and randomizing sessions, internal validity is guaranteed through the control of potential confounders. These are created by the effects of sequence and order and remove selection bias, balance, and both known and unknown confounding factors. In this systematic review, seven of the studies (64%) randomized their sessions, and only three (27%) counterbalanced their sessions. Significant methodological differences were observed across the accepted research manuscripts with the amount and type of familiarization varying across studies. Appropriate familiarization will ensure cognitive performance prior to conducting experimental sessions, which demonstrates a plateau effect [3].

As previously suggested in TOD studies, the importance of establishing laboratory-based protocols which are more rigorous is essential. There is a need of the methodological control and quality to improve, such as the appropriate timing of morning and evening sessions when assessing cognitive performance. The timing of morning and evening sessions assessing cognitive performance varied from 06:00 to 11:30 h and between 16:00 and 21:10 h, which is not within the appropriate timeframe needed to establish diurnal variation with timings closer to the nadir and peak of body temperature deemed more suitable. The lack of standardization of methodologies and factors that affect cognitive performance might explain why findings observed conflicting differences in several performance variables. The willingness of individuals to undertake early morning sessions and the laboratory opening times within research “buildings” affect this.

One of the major strengths of this systematic review is that it is the first review providing an in-depth overview related to cognitive performance and TOD. The review was performed in a structured manner following the PRISMA 2020 guidelines [27]. In addition, only four other reviews have provided detailed information around factors affecting cognitive performance in chronobiological studies [3‐5]. Further, the diversity and range of databases used within this review’s search strategy and the specific search terms utilized further strength. Finally, the inclusion criteria were strongly adhered to, and only studies which assessed diurnal variation and cognitive performance were included.

A limitation of the present systematic review was the large differences in methodology and cognitive performance tests used in the 11 included studies. Hence we were unable to conduct a meta-analysis and pool the datasets observed to further assess evidence associated to cognitive performance and TOD. This was mainly due to differences present between the 11 studies related to methodological and clinical heterogeneity [64]. Study design across studies displayed irregularities when assessing the methodological design. There was also disagreement as to whether cognitive performance displays TOD or diurnal variation and the timing of when this was observed.