Impact of Seasonality on Physical Activity: A Systematic Review

Background: The purpose of this study was to collect and analyze the available scientific evidence of the impact of seasonality on physical activity (PA). PA refers to walking, biking, sports and/or active recreation. Methods: The search was performed in the following databases: PubMed, PEDro, Cochrane and Embase. All publications from January 2015 to September 2020 assessing seasonal variations on physical activity development in adults were selected. Results: A total of 1159 articles were identified, of which 26 fulfilled the selection criteria involving 9300 participants from 18 different countries. The results obtained suggest that seasonality affects PA independently of the countries, pathologies of the participants and the tool to collect PA information. Conclusions: PA level varies across the seasons, with higher PA level in summer compared with other seasons, especially in winter. Sedentary behavior follows the opposite trend. Impact of seasonality variations should be considered in clinical research involving PA as a primary outcome as well as in interventions on PA promotion.


Introduction
The World Health Organization (WHO) defines physical activity (PA) as "any bodily movement produced by skeletal muscles that require an expenditure of energy. Physical activity refers to all movements, including during leisure time, for transportation to and from places, or as part of a person's work. It considers sports that can be practiced at any level as: walking, biking, active recreation, and different games" [1].
The WHO recommends 150 min of moderate-intensity physical activity (PA) per week in adulthood and old age [2]. However, the percentage of the world's population that does not reach the minimum levels of PA is still high [3]. Precisely, according to the WHO, 23.3% of the global population in 2010, and 27.5% in 2016 [3,4]. One out of four adults is not active enough. In terms of the geographical zone, the US (32%) and Eastern Mediterranean regions (31%) [5] exceed the world average. As well, in the European Union (EU)-two thirds of the population does not reach the minimum recommendations for adults [6,7]-and the Arabic region the inactivity rates are over 60% [8].
Seasonality seems to impact PA levels and the exacerbations of some diseases and mortality [29,34]. Temperate climates have revealed higher mortality rates in winter than summer [40,41]. Seasonality is also known to affect more the aging people [29,34]. Moreover, in populations with pathologies, the risk of exacerbations increases during winter [15]. For example, preoperative lung cancer patients are much less physically active in the winter season, affecting their functional capacity. Thus, they could not be considered suitable for some surgical interventions during winter months [42]. In addition, adherence decreases in diseases such as COPD and HF during summer months. [15,43]. Additionally, due to the lack of adapted indoor facilities, wheelchair users are affected by seasonal variations and unfavorable weather conditions [44].
In recent years, PA promotion programs [2,3,27] have increased worldwide. However, despite the considerable heterogeneity of environmental conditions, there is little research on their influence on PA. Moreover, current guidelines and consensus do not adapt to the different periods of the year and the challenges they pose for PA implementation [45]. In recent years, the concern about climate change and global warnings has increased, including the analysis of its impact on health [46]. Accordingly, there has been increasing attention toward consideration of such change as a barrier to physical activity, including interest in other variables that could be modified by it, such as seasonal variations [47].
As humans cannot modify meteorological and seasonal conditions at their own will, the most intelligent response is to understand better how they affect PA to adapt and reduce or stop the adverse impact on the PA levels populations [42]. Being aware of how weather conditions affect physical activity can help policymakers and healthcare providers to adopt recommendations to mitigate its effects [47]. Thus, collecting data on PA and seasonality is crucial because it provides information on what strategies and interventions need to be modified during the different seasons of the year to avoid physical inactivity [39].
The aim of this systematic review is to compile and evaluate current available evidence about the impact of seasonality on PA and to describe the different strategies and tools used to collect variables related to PA.

Materials and Methods
This study follows the guidelines of the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) [48]. The Supplementary Table S1 provide the details of domain-specific score.

Research Strategies
The systematic search was executed using a structured electronic search in PubMed, PEDro, Cochrane and Embase databases in the October-December 2020 period. PA and seasonality have been the main two elements of search. Both on MeSH Terms (motor activity, exercise, training and seasons) and free terms keywords (physical activity, season), (Supplementary Table S2). Truncation has also been used for the keyword "season". A manual search which included the references and related articles, has been also carried out.

Selection Criteria
Articles published in any language between January 2015 and September 2020 (both included) assessing the influence of seasonality on PA were included. Concerning exclusion criteria, systematic reviews, meta-analysis, case studies and those studies conducted with participants under the age of 18 were excluded. Similarly, studies evaluating the influence of weather instead of seasonality were also excluded. Lastly, we excluded studies that do not differentiate between established standard seasons (e.g., referring to rainy/non-rainy seasons, school year vs. summer holidays) or those that measuring PA only within a season.

Assessment of Methodological Quality
For the quality assessment, the "standardized instruments from the Joanna Briggs Institute System for the Unified Management, Assessment and Review of Information" (JBI SUMARI) checklist was used to report and critically appraise the methodological aspects of included studies [49]. These instruments included the JBI Critical Appraisal Checklist for Comparable Cohort, the JBI Critical Appraisal Checklist for Cross-sectional Studies, the JBI Critical Appraisal Checklist for Randomized Control Trial [50] and the JBI Critical Appraisal Checklist for Quasi-Experimental Studies, and were chosen accordingly to the study design [49][50][51].

Data Extraction
Climate information was collected according to Köppen climate classification, first published in 1936. This instrument classifies climate into five main classes: tropical zone (A), arid zone (B), temperate zone (C), snow zone (D), and polar zone (E). The five main climate classes are further subdivided into 30 climate subtypes. Each subtype is defined by two or three letters code: the first letter is referred to the main class of climate, the second letter indicates the seasonal precipitation type, while the third letter indicates the level of heat [52].
The variables included in Tables 1 and 2 were gathered, such as: country of implementation, climate related information, objective of the study and year of publication, characteristics of the participants (age, sex, chronic diseases, and other relevant information about the population), seasons, tools for collecting information on PA, measurement time of each outcome and PA results for each included study. Two different reviewers selected studies, rated methodological quality, and extracted data independently. If there were any disagreements between both investigators, a third independent researcher determined inclusion/exclusion. To describe the seasonal differences in physical activity Cross-sectional study

Search Results
As shown in the PRISMA flow diagram (Figure 1), after the initial search and eliminating duplicates, 1159 articles were identified, of which 1126 studies were eliminated after reading the title and summary. Of the 33 remaining, after critical reading of the complete text and including one reference from manual search, 26 studies were finally selected for the systematic review.

Search Results
As shown in the PRISMA flow diagram (Figure 1), after the initial search and eliminating duplicates, 1159 articles were identified, of which 1126 studies were eliminated after reading the title and summary. Of the 33 remaining, after critical reading of the complete text and including one reference from manual search, 26 studies were finally selected for the systematic review. Concerning the quality appraisal, cohorts studies were over 63% of the items achieved, except for one study [55]. Items related to missing data and lost follow-up were not achieved in any of the studies. Regarding cross-sectional studies, all studies were over 75% and three of them accomplished 100% of the criteria. Lastly, RCT were over 60% of the items achieved. None of them blinded participants to treatment assignment. The details of domain-specific score are provided in Supplementary Tables S3-S5.
Regarding types of climates, according to the Köppen climate classification, 9 types of climates were represented in the selected articles. Most of the studies were developed Concerning the quality appraisal, cohorts studies were over 63% of the items achieved, except for one study [55]. Items related to missing data and lost follow-up were not achieved in any of the studies. Regarding cross-sectional studies, all studies were over 75% and three of them accomplished 100% of the criteria. Lastly, RCT were over 60% of the items achieved. None of them blinded participants to treatment assignment. The details of domain-specific score are provided in Supplementary Tables S3-S5.
A total of 9300 people took part in the set of studies collected by this review. There was a great variability between the sample sizes. Four studies included more than 1000 participants and nine less than 100 participants. Regarding the age of the people included, the mean age ranged from 30 years [44] to 80.3 years [37]. Twelve out of the 26 publications included participants with a specific pathology: chronic obstructive pulmonary disease (COPD) [15,19,20,59,60], heart failure (HF) [23,24,45], lung cancer [42], coronary heart disease [56], spinal cord injury [44] and type II diabetes and/or hypertension [54].

Physical Activity and Seasonality
Most of the studies (22 out of 26) found significant variations in PA among different seasons. In contrast, three [9,15,45], noticed no significant differences in PA related to seasonality. The three that found no significant differences were included in countries with low winter temperatures (Canada, Norway, Denmark, and Sweden) with average minimum temperatures in winter from −15 to −37 • C. However, the study conducted by Sayegh et al. [5], in a subtropical desert climate with very high summer temperatures and humidity, showed a significant decrease in PA in summer.  Overall, studies showed a higher level of PA in summer compared to winter and compared to all other seasons. Some studies compared seasons more favorable for PA, spring/summer vs. autumn/winter, finding statistically significant differences. Others works discovered that spring is the season of the year with the highest level of physical activity. Finally, a few studies noticed significant decreases in PA in winter compared to the rest of the seasons.
Some of the included studies showed results disaggregated by age group. Cepeda et al. [34] observed that middle-aged participants (50-64 years) and young elderly (65-74 years) were more physically active in summer than in winter. Meanwhile, the elderly (≥75 years) displayed no seasonal variations. Finally, Wesolowska et al. [17] remarked a higher number of steps in summer and spring compared to winter in all age groups.
Regarding the level of activity of the participants, Arnardottir et al. [37] noticed more physical activity in summer and, according to the stratified results, this summer-winter difference was significantly more elevated in the high activity level group than the low activity group. In the same line, Shoemaker et al. [24] also found a greater impact on seasonality in participants with fewer comorbidities and with physical activity longer than 2.2 h per day. Lapointe et al. [56] discovered that seasonal variations influence physical activity, but only in the active group (lower physical activity in autumn and winter than in spring and summer). For the low-activity group, no significant differences between seasons were observed. Nevertheless, in the study by Kim et al. [27] carried out on women, members of the active group were more likely to maintain the increase in the number of steps achieved in spring at the arrival of autumn-winter, contrary to women in the other two groups (low active and somewhat active), that showed a significant decrease.
Concerning sedentary behavior (SB): only three out of five studies assessing SB, found statistically significant differences: more sedentary time in winter compared to summer [37,58] and in autumn-winter compared to spring-summer [54].

Seasonal Variations on Physical Activity
This systematic review identified 26 of 1159 articles that fulfilled the selection criteria to determine the impact of seasonality on PA. Overall, the results of the review showed that PA increases significantly in the summer-spring months compared to winter. It occurs independently of the countries' climate, the characteristics and previous pathologies of the participants. Regarding the geographical areas, results were similar despite the climate of the region except for the sub-desertic climate where PA level decreased in summer.
Seasonality variations on PA appeared in the included studies in this systematic review independently of whether they are performed in healthy participants or in subjects with specific pathologies such as COPD, heart failure, lung cancer, coronary heart disease, spinal cord injury, type II diabetes and/or hypertension.
According to variations within the states, some countries like the United States, Canada and Australia cover immense land areas. It means that they certainly have a wide range of climates and seasonal variations. On the contrary, Scotland, Netherlands and France do not have as much land extension; however, there are discrepancies due to urban-suburban, mountain-coast or north-south disparities such as in the Scandinavian countries where there are many distinctions concerning the daily sunshine hours in the same territory [39]. Consequently, these countries may also require specific studies in different areas, as the geographical location is as important as the size of the nation.
Exploring the three selected studies that do not find statistically significant differences, two out of the three do it with clinically significant differences in practice. For instance, the study developed by Hoaas et al. [15] on COPD patients from Norway, Denmark and Australia, shows an increased number of steps at all locations during summer compared to all other seasons. This increased number of footsteps, furthermore, exceeds the minimum clinically important difference established for COPD patients (between 600 and 1100 footsteps/day). The magnitude of this PA growth is related to a reduced risk of hospital admissions [61]. Nevertheless, this difference is not statistically significant probably due to modest sample size in each group. In addition, the study conducted by Klompstra et al. [45] in participants with heart failure from Sweden, stablishes a change of 250 METs/week as clinically significant and some of the patients increased their PA in summer over this amount. Unexpectedly, this study observes that a large percentage of participants increased their PA in winter compared to summer. Authors concluded that his variation may be explained because the differences between summer and winter temperatures during the study were not as marked as expected and also the low response of the participants in the winter period (58%) may have caused a response bias. On the contrary, Kim et al. found statistically significant -but small in magnitude-differences in PA among seasons with a foreseeable little clinical impact as they do not exceed 300 steps/day.
According to the results of our review, the level of physical activity of the participants may be relevant, in seasonal variations of PA showing a greater impact of seasonality in participants with a higher level of PA [24,27,37,56]. In those population, the increase of PA activity in spring-summer is bigger than in low PA level participants. In this point of view, the results imply that participants who are physically inactive may not change their level of PA even in more favorable season.
Regarding different levels of intensity of PA, some studies collected them separately and their results could be relevant. For example, Cepeda et al. [32] found that the greatest seasonal variation was in light PA levels. Likewise, in the work of Furlanetto et al. [18] in patients with COPD, they found no differences in activities from moderate to vigorous intensity (above 3 METs). Nevertheless, they found differences in the time spent in activities with an intensity above 2 METs and suggest that this could be a more appropriate measure for subjects with a low activity profile.
Results of this work are consistent with a previous systematic review published in 2007 including a total of 37 studies between 1980 and 2006 [37]. The findings showed the level of PA also varies with seasonality, being higher in spring and summer and lower in winter. 27 out of the 37 studies included in the review found statistically significant differences in PA. Nevertheless, this previous review only included studies developed in 8 different countries showing a less range of diversity of territories. In addition, it analyzed together the impact of both, weather and seasonality, on PA. Additionally, a more detailed analysis is required in order to take into account stratification about age groups or level of PA.

Tools for Physical Activity Assessment
Studies included in the systematic review use different tools for PA information collection. In general, the results remain the same despite the tools used to collect PA both direct (objectives) and subjective instruments. However, some publications that applied both objective and subjective methods assessed if there was a correlation between different methodologies. Both Cooke et al. [54] and Wesolowska et al. [17] found that there was no correlation between objective and subjective methods. In fact, Wesolowska et al., in the absence of a statistically significant correlation between pedometer values (objective) and IPAQ scores (subjective), suggested that the participants may not correctly assess their own level of physical activity using subjective tools [17].
The use of objective methods seems to display numerous advantages as they measure the changes in physical activity and sedentary behavior more accurately than questionnaires [62][63][64]. They offer objective feedback [17,34], encouraging the interruption of long periods of sedentary behavior. Even in real time, they facilitate compliance with physical activity goals. They contribute as well, to design interventions in physical activity promotion and sedentary behavior [34]. They bring to the patients the opportunity to work out when they have contraindications to strenuous exercise. Additionally, they help positively in those who are less motivated to move and procure reproducible results [17].
Nevertheless, some objective methods (i.e., accelerometer) are not very accurate in distinguishing activities that fluctuate according to seasonal patterns such as cycling or swimming, with higher practice in summer. Thus, studies that rely purely on the use of accelerometers may then underestimate seasonal differences [65]. Objectives tools are also restricted estimating upper body movements during transport and heavy lifting activities [37]. More importantly, it may bias the assessment of the level of PA due to its effect on the promotion of physical activity. Nevertheless, most of the studies that incorporated objective methods used 1-week measurement periods which give a reproducible and practical dimension of physical activity and sedentary time [66]. Limiting the use of pedometers or accelerometers to specific weeks would help to reduce their effect/impact on the results. This strategy may be beneficial for some research, but also counterproductive or limiting, as it reduces the ability to identify variations in PA throughout the year due to the influence of external factors such as seasonality [39,67,68].
Subjective methods are commonly used to collect PA information because of their low cost, easier and faster administration, compared to direct methods and their possibility to measure different types of activities [32]. Furthermore, compared to other methods such as pedometers or accelerometers, they have a low level of influence on the results [67]. However, regarding the limitations of subjective methods, they might lead to an erroneous estimation of the activity performed [17,32] and may not detect seasonal variation while objective methods have done so, as in the study by Cooke et al. [54].

Study Designs for Seasonality Assessment
Due to the one-year periodic variations, the optimal way to analyze seasonality would be through a longitudinal study using the same individuals-if possible, for more than 1 year, to reveal parallel and divergent trends between years with more or less adverse climatic conditions [44]. Drawing conclusions about seasonality based on non-longitudinal designs, which compare different groups of people from different seasons with short sampling periods, may not be the most appropriate design [24].
Cross-sectional collection methods do not provide information on trajectories at the individual level [16], neither, do they observe the changes in PA of the same group of patients in different seasons. If in addition to being a cross-sectional model, there is a deficit in the number of patients recruited in one or more seasons, and the results regarding seasonal variations in physical activity could be influenced. As in the work performed by Hoaas et al. [15], where the low number of physical activity data collected during summer could have influenced the results; or in the study by Kong et al. [42] and the study conducted by Klompstra et al. [45], where the number of patients observed in winter is limited.

Limitations and Strengths of the Review
The present review was conducted following the PRISMA checklist. One of the limitations was the heterogeneity of the tools for the PA assessment, sample size, characteristics of the population and measurements periods. At the same time, the consistency of the results even with this heterogeneity represents the magnitude of the impact of seasonality on PA levels. The studies have been carried out in 18 different countries in the two hemispheres and on four continents.
Additionally, there are other constraints: (a) the small sample sizes of some of the studies [13,19,23,[55][56][57]; (b) the low number of articles carried out in two or more countries and studies showing results disaggregated by age and gender population groups; (c) some of the included studies are conference abstracts [53,54,56,[58][59][60], although there was some sufficient information for the synthesis of results; (d) limited date range for study publication; (e) regarding the study design, most of the publications are observational studies both cohorts (15 studies) or cross-sectional (seven studies) with only four studies with a quasi-experimental [58] or experimental design [16,20,23]. Given the characteristics of the research question, it was expected that most of the designs were observational.
Future studies on the effects of seasonality on PA are, in general, required with the widest possible diversity of locations, pathologies and population groups and with the application of the most appropriate methodologies to capture and quantify seasonal variation.

Implications of the Results for Clinical Research and PA Promotion Interventions
As the results of this systematic review showed that the influence of seasonal variations on physical activity is relevant for the general population-and certain groups-it should always be considered as one factor that may influence PA outcomes. In both, PA promotion interventions and clinical research, seasonality should be perceived as one of barriers for users to join in physical activities.
The results of this review may be helpful to identify the better time to set up or change the physical activities for people. It may also be useful, to implement PA maintenance strategies in seasons with a tendency to reduction (autumn-winter). This includes initiatives on the environment and facilities that allow opportunities to perform PA despite the characteristics of the different seasons. On the other hand, establishing light PA strategies might replace the increased sedentary time during winter and autumn, which suffers the greatest seasonal variation [34]. It can also be productive to incorporate specific and individualized education according to the season. Moreover, depending on the environmental context, encourage PA regardless of seasonal changes.
The results may also be useful for the interpretation of PA assessments. This is particularly relevant, where there may be differences in seasonal variations either because the research has been carried out in several countries or in a single one with significant variations across the country. In research studies it will be essential to take into account the effect of seasonality in both the initial and subsequent measurements during the follow-up period. It will be important to be careful when extrapolating PA results to locations with different seasonality conditions.

Conclusions
PA level follows seasonality variations finding higher PA level on summer compared with other seasons, especially on winter. Sedentary behavior follows the opposed trend as PA level regarding seasonality. Results are consistent in different countries and populations with chronic diseases, but future studies are required to get more detailed about its impact on gender and different age ranges or previous intensity of PA level of the individuals. Impact of seasonality variations should be considered in clinical research involving PA as a primary outcome and necessarily for interventions on PA promotion. Public health interventions could be implemented in order to analyze the potential impact of seasonality as a barrier for PA development in each specific context.