Introduction

Regular exercise is known to confer numerous health benefits and can be an effective tool in both preventing and treating many chronic medical conditions, including obesity and type 2 diabetes [1,2]. The global increase in chronic diseases is associated with the so called ‘Western Lifestyle’, which includes high calorie diets and sedentary behavior [3]. Moreover, disruptions in natural circadian rhythms caused by irregular work/sleep schedules, exposure to artificial light, and around-the-clock dietary intake are increasingly part of everyday life and are emerging independent risk factors for chronic diseases [4], and premature aging [2].

The specific timing of exercise has been shown to significantly modulate the circadian clocks and affect the body’s core clock and natural rhythms [3,5,6]. For example, exercising in the morning can advance the circadian rhythm, while evening exercise can delay it [3,7]. Thus, consistently scheduled exercise at certain times of the day may contribute to resetting the body’s circadian clock, and help to restore physiologic circadian rhythms [3], potentially aiding to prevent metabolic and other chronic diseases [3,8]. Afternoon exercise has gained some traction as being superior in improving metabolic derangements for those living with prediabetes or type 2 diabetes [1], or in those with excess body weight [9], and also in general populations [3,4,10]. Even though exercising at any time is widely believed to be better than not exercising at all, intentionally planning the timing of exercise may maximize the physiologic benefits derived from physical activity [4,8].

While consistently timed afternoon/evening exercise can be achieved in the setting of short-term clinical studies, it is unclear if this is sustainable over the long-term outside the experimental setting. Specifically timed exercise may impose burdens on individuals’ quality of life, making it essential to understand patient experiences and adherence outside research settings, as these aspects remain understudied and underreported.

It is well established that interventions which demand a great deal from patients, and lead to less satisfaction are unlikely to be taken up or maintained in daily life outside of research settings [11]. Health interventions and therapies that are clinically suitable and physiologically sound may not always be sustainable over the long-term, because patient perspectives of such interventions are not universally taken into account. Considering patients’ unique circumstances, beliefs, preferences and facilitators/barriers is important for increasing intervention acceptability and effectiveness.

Patient-reported outcome measures (PROMs) can be utilized to describe an individual’s personal, subjective evaluation of a given intervention based on its impact on their quality of life, symptoms, functioning, and physical, mental, and social wellbeing. PROMs can provide clinicians and researchers with unique information that cannot be obtained from purely biomedical measures. Understanding PROMs related to a certain treatment can help improve treatment protocols and facilitate mobilization of interventions into clinical practice and real-life situations and thus, achieve higher likelihood of adoption and adherence to the studied intervention [12]. Some studies incorporate pre- and post-intervention PROMs as pre-specified outcomes [13]. While there are many different PROMs, most are in the form of validated scales, some examples include measures of: health-related quality of life, depression, anxiety, pain, hunger, etc [13–16]. The aim of our systematic review was to identify and synthesize the findings from all PROMs that have been reported in interventional studies on timed exercise interventions.

Methods

We conducted a systematic review and adhered to the Cochrane handbook guidance document [17] and the established reporting methods specified by the PRISMA (Preferred Reporting Items for Systematic Reviews) statement [18]. This review was part of a systematic review with a larger scope, with a protocol registered in the Prospective Register of Systematic Reviews (PROSPERO, registration number CRD42022322646; 3 May 2022. crd.york.ac.uk/prospero/display_record.php?ID=CRD42022322646). The broader review included the impact of a variety of different timed health behavior interventions (diet, exercise, sleep) on PROMs. Initially, the search aimed to identify the effect of all timed lifestyle interventions on PROMs. However, the large volume of relevant articles led to organizing studies into distinct categories of timed dietary interventions, timed exercise interventions, and sleep-related interventions. This approach allowed for separate systematic reviews of each intervention, enabling a more focused and detailed analysis of their effects on PROMs.

1. Search methods for identification of studies

• Electronic searches: We conducted a systematic literature search from inception until August 2023. Databases searched included: MEDLINE, SCOPUS, Embase, APA PsycInfo, CINAHL, and Web of Science. No date limits were applied. The search was conducted exclusively in English. Searches combined keywords and subject headings from three concepts: 1) patient perspectives (e.g., patient engagement OR patient perspective OR patient reported outcome measures OR patient reported experiences OR surveys OR qualitative research OR interviews OR focus groups); 2) diet, exercise or sleep (e.g., health promotion OR exercise/ physical activity OR diet/nutrition OR sleep OR behavior); and 3) timed interventions. The search strategy was formulated through comprehensive discussions between the primary authors and the health sciences librarian (DL), seeking to incorporate all variations of relevant keywords in the published literature into the search. The detailed search strategy can be found in the supplementary material (S1 Appendix).
• Reference list searches: The main literature search was supplemented by manually searching the reference lists from studies that met our inclusion criteria and systematic reviews on exercise.

1. Study inclusion and exclusion criteria

• Types of studies: We included all original interventional studies of any size that assessed timed exercise interventions, irrespective of comparator studied or randomization. Studies using retrospective and secondary data were excluded. Studies were excluded if they were dissertations, posters, conference abstracts, commentaries, reviews, letters to editors, or questionnaire development/validation programs.
• Types of participants: The review included studies in which human adults, or their surrogates reported a PROM related to timed exercise interventions. Studies conducted on children/adolescents or non-humans were excluded.
• Types of interventions: The review included all forms of timed exercise interventions, both acute exercise bouts and long-term training studies. Timed exercise interventions were defined as physical activity that was specified to be undertaken at a particular time of day. Studies were included if they included a comparison of any different timing of exercise [i.e., morning vs afternoon; morning vs evening; afternoon vs evening]. No restrictions were placed on the type, intensity and mode of exercise.
• Types of outcome measures: The review only included studies reporting one or more PROM – using either validated measures or bespoke tools.

1. Screening and selection of literature

• Title and abstract screening: Results of the search were imported into Endnote software for processing, including manual removal of duplicates and screening based on eligibility criteria. The cleaned list of references was imported into the ASReviews machine learning software [19] for title and abstract screening. ASReviews uses machine learning to rank the titles and abstracts on relevance, based on the patterns being established by the reviewer, and pre-specified articles of interest. A single reviewer (MK) conducted the initial screening using the software. The process for training and evaluating the AI tool was carefully conducted. Initially, we added 10 relevant and 10 irrelevant articles, clearly identifying them as such. After this initial training, the entire database of articles was input into the tool. From there, the reviewer (MK) continually marked each title and abstract as relevant or irrelevant. This interactive process allowed the tool to learn and refine its judgment, progressively surfacing the most relevant articles for review. This approach ensures that the tool becomes more accurate over time, although it does rely heavily on human input during the early training stages to establish a solid foundation for relevance assessment. The developers of the application recommend stopping the manual screening when 300 consecutive non-relevant articles have been reviewed and discarded. For precautionary reasons, we increased our margin to 500.
• Full-text review and selection: The full-texts from all articles identified as potentially relevant to the study at hand were reviewed in depth by two reviewers (MK and CF). Full-text articles were compared against predefined study eligibility criteria, to determine which studies were included and which were excluded. Agreement between the two reviewers was sought and disagreements were resolved through consensus or with the senior investigator (DJTC) as needed.

1. Data management and analysis

• Data extraction: Relevant data was extracted by two reviewers (MK and CF) in November and December 2023, using a pre-designed data extraction worksheet. Data included study details (author, year, design, duration), number and sex of participants, intervention description, and study outcomes related to PROMs. Details describing each exercise intervention followed the TIDieR method [20], to the extent permitted by the reported information.
• Missing data: We planned to do no imputation of missing data elements and planned to simply report where elements of methodology or reporting were lacking from each included study.
• Assessment of risk of bias in included studies: For Randomized Controlled Trials, we used Cochrane’s Risk of Bias 2 (RoB 2) tool to assess study quality [21]. The RoB 2 tool also has an extension for Crossover Trials [22], which we used for these specific study designs. Finally, for non-randomized studies of interventions, we used the ROBINS-I tool [23]. Risk-of-Bias assessments were conducted by two independent reviewers (MK and RS) with disagreements resolved through consensus discussions.
• Unit of analysis issues: Conversions were used to standardize units of measurement for homogeneity and comparability across studies.

1. Data analysis and synthesis

• Data synthesis: Results were reported following the guidelines of the Synthesis without meta-analysis (SWiM) in systematic reviews [24]. Assessing the certainty of the findings was not feasible given the varied nature of the PROMs reported by the different studies.
• Measures of treatment effect: Given the heterogeneity of PROM outcomes, a vote-counting method was applied [25]. Studies were classified based on whether they showed a reduction, no effect, or an increase in the outcome measure. Details of the treatment effect magnitude was also extracted.
• Study grouping: Because relevant PROMs may vary if the exercise is a single bout, or part of a longer-term program of physical activity, we felt that it was important to group the studies along these lines. We classified studies into acute exercise (assessing response to a single bout of exercise – which are most likely to impact immediate PROMs like exertion or fatigue) and long-term training (assessing incorporation of timed exercise over a sustained period of at least 1 month – most likely to impact PROMs like mental health, energy, etc…).

Results

Study selection and characteristics

A total of 37,198 articles were located through database searches. Of these studies, 10,358 were removed when filtering for duplicates. ASReviews was used to screen 6042 titles/abstracts, after which we had reached 500 consecutive non-relevant articles and therefore stopped screening further studies using the software making the decision that all potentially relevant studies had been screened. Eight additional articles were identified through manual searches of reference lists. This resulted in 98 unique articles, of which 51 were included in the analysis after full text review. Of these, 26 were dietary interventions, 9 were sleep interventions and 16 on exercise interventions – only the latter were included in this review (Fig 1, S2 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram.

https://doi.org/10.1371/journal.pone.0321526.g001

Study design and risk of bias

Of the 16 timed exercise studies that are the focus of this review, we identified 10 cross-over RCTs, 5 parallel group RCTs and 1 non-randomized interventional study (Table 1). Details of included studies are reported in the supplementary material (S3 Appendix) and a summary of the main outcomes are shown in Table 2. Across all studies there was a total of 403 participants. The majority (n = 11) of manuscripts described acute exercise studies, with fewer studies (n = 5) examining the impact of sustained timed exercise (duration: 8–24 weeks) on PROMs. Most included studies were found to have moderate to serious risk of bias (Table 3). Details of the risk of bias assessment can be found in the supplemental material (S4 Appendix). One of the RCTs was evaluated to have serious risk of bias due to missing outcome data and 10 of the studies were noted to have some level of risk of bias where none of the studies reported a clear method of randomization nor the reasoning behind the reported results, which made them unlikely to be representative of the broader population.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study Characteristics and Sample description.

https://doi.org/10.1371/journal.pone.0321526.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Results of outcomes measures.

https://doi.org/10.1371/journal.pone.0321526.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Risk of Bias for RCTs, cross over design and non-RCTs using the Cochrane tools.

https://doi.org/10.1371/journal.pone.0321526.t003

Discussion

The aim of this systematic review was to synthesize the evidence on the effect of exercise timing on PROMs. There is a relative paucity of research on this topic, as only 16 studies with a total of 403 participants were included in the review. Studies primarily investigated the effect of exercise timing on sleep quality (n = 8), with other PROMs explored less frequently. More than half of the reported PROMS were not significantly affected by the timing of exercise. While it is difficult to make any conclusive statements about the effects of these interventions due to the heterogeneity in the interventions and outcomes, in general, morning exercise seemed to have more positive associations with PROMs [28,29,33]. However, in two instances, certain PROMs were positively affected by evening exercise such as decreased stiffness [41], and improved vitality [38].

Studies included in our review found that exercise timing did not impact appetite [26] and hunger [27]. These findings are consistent with a review on the impact of exercise on appetite regulating hormones, which revealed that exercise suppresses appetite through stimulating anorexic hormones and this effect is not influenced by exercise timing [42].

Included studies were also inconclusive [27,28,32] for a clear effect of timing of exercise on perceived exertion, with 2 [30,31] out of 5 studies reporting significant increases in perceived exertion in evening exercisers, particularly notable in a study involving “early riser” cyclists whose evening exercise coincided with their habitual sleep time [30].

Literature has revealed that exercise is beneficial for mood and mental health outcomes [43]. Despite inconsistent effects of exercise timing on mood [33–35], one study found that morning exercise was associated with positive emotions when compared to afternoon or evening exercise [33]. Furthermore, the overall QoL score did not change due to exercise timing [38,39], but less bodily pain was reported with morning exercise in acute [35] and long-term [41] studies. Interestingly, evening exercise was found to reduce joint stiffness in adults with rheumatoid arthritis [41], potentially explained by the physiological changes controlled by the circadian rhythm, such as inflammatory cytokines and body heat that gradually improve mobility throughout the day [44]. Finally, the relationship between exercise timing and sleep parameters was mixed in our review. There is a robust body of evidence establishing a positive relationship between exercise and sleep in general [45] which often discourages exercising close to bedtime [46], for better sleep hygiene. However, in the included studies, exercise timing did not have significant impact on subjective sleep outcomes [26,29,33,35–37,39], apart from one study in older adults, who experienced better sleep satisfaction with evening exercise [40]. These findings align with a recent systematic review which indicated that there is no detrimental effect of evening exercise on sleep [10].

Overall, the studies included in this review were evaluated to have significant risk of bias, based on the Cochrane assessment tools. One common limitation for several of the non-randomized studies was the fact that the samples were small and conveniently chosen. Measures were often not taken to ensure population representativeness [36,38]. Furthermore, the PROMs reported in this review were largely secondary outcomes in studies that were powered to detect differences in biological parameters. Therefore, it is likely that the results of this review are underpowered to comment conclusively on the impact of exercise timing on PROMs. For enhanced understanding of the impact of timed behaviors on PROMs, future research should consider using PROMs as primary or co-primary outcomes and studies should be powered accordingly.

The articles included in this review analyzed the effects of exercise within controlled research settings, these designs are unable to fully consider the impact of timed exercise protocols in real-life situations, inclusive of social duties and work commitments. A drawback of the current body of research is that while consistent afternoon/evening exercise can be achieved in short-term clinical studies without significant detrimental effect on PROMs, it remains unclear if this would be viable for most people in the long run in the context of their everyday lives outside controlled experimental conditions. Moreover, the significant heterogeneity in exercise protocols, participant demographics, exercise duration, timing, and intervention lengths across studies, makes drawing firm conclusions challenging. The lack of standardization in these parameters highlights the need for more consistent research methodologies for reporting on PROMs in this field. Furthermore, the PROMs that have been collected to date do not explicitly explore participants’ acceptability of the intervention. To gain a comprehensive understanding of the practicality and long-term effectiveness of exercise interventions, future research should explore how timed exercise can be sustained in everyday life, taking into account individuals’ social duties and professuional commitments in addition to the kinds of PROMs that have been assessed to date. By addressing the challenges of real-life scenarios and adopting more standardized research methodologies, we may be able to draw more reliable conclusions about the sustainability of exercise timing interventions on long-term health and well-being.

The participants in the various studies covered a wide range of ages, and people with varying physical fitness backgrounds, and the included studies used different types of exercise interventions. This is a strength as it shows relative consistency in the findings, regardless of differences in these factors. Conversely, this may also be viewed as a limitation given that it makes direct comparison between the studies difficult. One limitation to the studies included is that approximately half of the studies included only male participants. Therefore, it is possible that the outcomes reported are not applicable to females and women, and future research is needed to understand potential sex and/or gender differences. More high quality RCTs (with lower risk of bias) investigating the effect of exercise timing on PROMs are needed. Other limitations of this review include the fact that only studies published in the English language were included, and that we were unable to complete a meta-analysis of the included studies given the marked heterogeneity in study design, intervention type, and outcome measures.

Conclusion

In conclusion, this systematic review suggests that timed exercise interventions do not seem to have a consistent effect one way or the other on PROMs that have been assessed to date. Some studies suggested that evening exercise may have a positive effect on perception of sleep quality, when compared to morning exercise. On the other hand, other studies suggest that morning exercise may be beneficial in several domains such as reducing hunger and improving anxiety. Overall, the findings of this review showed mixed results on multiple patient-reported outcomes and thus results remain inconclusive. That said, however, there does not seem to be a clear detrimental effect of afternoon/evening exercise compared to morning exercise with respect to most PROMs – which means that if ongoing and future clinical studies prove its metabolic superiority, evening exercise may be acceptable to individuals without clear detrimental effect on PROMs.

Supporting information