Effect of Physical Activity, Smoking, and Sleep on Telomere Length: A Systematic Review of Observational and Intervention Studies

Aging is a risk factor for several pathologies, restricting one’s health span, and promoting chronic diseases (e.g., cardiovascular and neurodegenerative diseases), as well as cancer. Telomeres are regions of repetitive DNA located at chromosomal ends. Telomere length has been inversely associated with chronological age and has been considered, for a long time, a good biomarker of aging. Several lifestyle factors have been linked with telomere shortening or maintenance. However, the consistency of results is hampered by some methodological issues, including study design, sample size, measurement approaches, and population characteristics, among others. Therefore, we aimed to systematically review the current literature on the effects of three relevant lifestyle factors on telomere length in human adults: physical activity, smoking, and sleep. We conducted a qualitative systematic review of observational and intervention studies using the Preferred Reporting Item for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. The systematic literature search covered articles published in MEDLINE and EMBASE databases (from 2010 to 2020). A total of 1400 studies were identified; 83 were included after quality control. Although fewer sedentary activities, optimal sleep habits, and non- or ex-smoker status have been associated with less telomere shortening, several methodological issues were detected, including the need for more targeted interventions and standardized protocols to better understand how physical activity and sleep can impact telomere length and aging. We discuss the main findings and current limitations to gain more insights into the influence of these lifestyle factors on the healthy aging process.

Aging is defined as a diminishing of biological functions over time, in which there is a progressive loss of the function and structure of tissues and systems [6,18]. The underlying mechanisms of the this process have been studied in depth and it seems they are related with genetic and epigenetic alterations, proteostasis loss, telomere shortening, autophagy deterioration, and mitochondrial dysfunction, among others [1,4,18,19]. However, the specific role of each mechanism remains unknown. Furthermore, these mechanisms may be modulated by other risk factors, such as diet, tobacco smoking, physical activity (PA), stress, or sleep, which, in turn, have been associated with certain diseases related to aging [18]. Therefore, a better knowledge of effective strategies for slowing the aging process is an important priority for researchers and policy makers.
Currently, several biomarkers of biological aging have been developed. Telomeres are regions of repetitive sequences (TTAGGG) n and associated protective proteins located at the ends of the cap-shaped linear chromosomes that preserve the integrity and stability of chromosomes during DNA replication. They are essential to chromosome stability and protect genomic DNA through various mechanisms [20][21][22][23][24]. Usually, a relative quantification protocol by calculating the telomere/single copy gene ratio is undertaken.
One of the factors potentially contributing to the heterogeneous results is the measurement approach used to determine TL. This is an important issue given the fact that TL differs in leukocyte subtypes isolated from the same individual. Although blood leukocytes are the most commonly used samples to determine TL, it is known that TL may substantially vary across various tissue types. The terminal restriction fragment (TRF) approach is the oldest method for analyzing TL and remains the "gold standard" even in current studies. It uses the southern blot or in-gel hybridization with a specific proof for telomere DNA [17,18]. Other methods include the fluorescent in situ hybridization (FISH) by using digital microscopy (quantitative FISH as Q-FISH) and flow cytometry (Flow-FISH). Another method, telomere shortest length assay (TeSLA), measures the distribution of shortest telomeres in heterogeneous situations. It provides a better measurement of the shortest length of telomeres. However, the quantitative polymerase chain reaction (PCR) is most widely used for its high throughput. Despite the fast, highly sensitive nature of PCR approach, which requires low amounts of DNA, the distribution of long and short telomeres cannot be obtained, and it is not possible to evaluate TL of individual cells or chromosomes [17,18]. This method is often chosen for large cohort studies, using DNA from leukocytes. Additionally, telomerase activity, gene expression, or epigenetic regulation of telomere-related genes are analyzed to delve deeper on the mechanisms [20,21,24].
TL is determined by the interaction between endogenous and external factors that explain greater or lesser susceptibility to accelerated aging [22]. Therefore, the exposome could well play a fundamental role through various biological mechanisms in the maintenance of the telomeres during aging [41,42]. Several studies have analyzed the association between lifestyle and TL, both as a protective factor [16,43,44], and as inducer of their shortening [36,45,46]. Diet has been one of the most commonly studied lifestyle factors with regard to TL. Several studies have shown the benefit of consuming foods rich in antioxidants or the benefits of various diets, such as Mediterranean type diets, on telomere attrition, as witnessed in a number of the already published reviews [47][48][49][50][51][52]. However, other relevant lifestyle factors related to aging and chronic diseases [53][54][55][56][57][58][59], such as PA, smoking or sleep have not been examined in such depth. Therefore, the aim of this work is to conduct a systematic review of the current literature (including observational and intervention studies) on the effects of PA, smoking, and sleep, on TL in relatively healthy human adults.

Search Strategy
We conducted a literature search for studies investigating the relationship between PA, smoking and sleep on TL in adults. In line with established guidelines (PRISMA) (http://www.prisma-statement.org/; accessed on 10 November 2021), a systematic review was carried out in which studies from the MEDLINE and EMBASE databases have been included. This was supplemented by a manual search of articles published from 2010 to March 2020. The search strategy for each lifestyle factor was made independently. The search terms employed were as follows: ("Physical Activity" OR "Exercise" OR "Smoking" OR "Tobacco" OR "Sleep") AND ("Telomere"). The protocol for the systematic review is available at PROSPERO, the NIHR International prospective register of systematic reviews (https://www.crd.york.ac.uk/PROSPERO/; accessed on 10 November 2021). Registration number: CRD42020192881.

Inclusion and Exclusion Criteria
Bibliographical inclusion was restricted to observational (cross-sectional, longitudinal studies, case-control studies and cohorts) and experimental studies (interventions) from original articles and carried out on healthy adults of both sexes, different ethnic backgrounds or with overweight, obesity or diabetes as these conditions (mainly overweight and obesity) are prevalent in studies focusing on the general population. Excluded were studies in adolescents, children, pregnant women, individuals with infectious diseases, and other chronic pathologies, or individuals consuming addictive substances, or with respiratory disorders or chronic sleep problems. Reviews, meta-analyses, conference summaries, unpublished studies, and studies published in languages other than English were also excluded from this systematic review.

Quality Control Analysis
The selection and assessment of the quality of the articles included herein were performed independently by three reviewers (R.B., C.O.-A., and J.V.S.). The quality of each study was determined by the tools provided by the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health (NIH) for evaluating the quality of observational and experimental studies. Once a score was obtained, the studies were grouped by quality, as good, acceptable, or low quality. Following this appraisal, 78% of the studies included in this review were classified as good quality.

Association between PA and TL
The impact of PA as a relevant factor in preventing diseases associated with aging was reported in several studies [53][54][55][56][57][58][59]. However, the underlying mechanism remains to be clarified. One of the possible mechanisms could be lower telomere attrition. To date, the optimum amount of PA and the type of exercise for maintaining TL remains under debate. Of the 83 articles reviewed, 49 analyzed PA and TL [45,46,. Cross-sectional studies (n = 23) are presented in Table 1 [45,46,. Table 2 [27,[81][82][83][84][85][86][87][88][89][90][91][92][93] shows observational studies (n = 14) using case-control and longitudinal approaches (n = 3); and Table 3 [44,[94][95][96][97][98][99][100][101][102][103][104] shows intervention studies. In these tables, the details for each study including author, reference, year, design, participants and characteristics, TL analysis methods, tissue for TL analysis, exposure assessment methods, and main results, are presented.     Most studies on the effect of PA on TL (37/49) were analyzed at the observational level in most cross-sectional studies (Table 1); PA was associated with longer TL. However, this association was mainly reported in subgroup analyses and not for the whole population included in the study. This subgroup analysis involved participant characteristics (men or women, older or young, among others) or different levels of PA (active versus inactive, moderate versus vigorous, regular, or not regular, etc.) Thus, as multiple comparisons were carried out in the same study, some of these comparisons could be statistically significant by chance, if no correction for multiple comparisons was considered. Therefore, the effect of PA on TL in these cross-sectional studies carrying out multiple comparisons is probably overestimated in the published papers. Among the cross-sectional studies, 3 [60,62,78] did not observe any association between PA and TL. The others reported mixed results (no association for some subgroups, but associations in at least one subgroup). One study [72], reported a U-shaped association, with only moderate PA being associated with longer TL. In these cross-sectional studies [45,46,, heterogeneity was also detected in the methodology used. PA was assessed by questionnaires in the most studies: some also included objective assessment (accelerometers, pedometers, or several PA tests) in the whole population or in subsamples. Likewise, variability in the methods used to assess TL was detected (17/23 used quantitative PCR, 5/23 used southern blot and 1/23 used Flow-FISH). Blood was the tissue selected for analysis in cross-sectional studies.
To note, several publications (7/24) were carried out on the same epidemiological study [in the United States population of the "National Health and Nutrition Examination Survey" (NHANES)] analyzing different levels of PA or participant characteristics [45,46,61,65,68,69]. Although the results differed depending on the subgroup and exposure analyzed, in general, in NHANES, undertaking any PA is associated with longer TL [45,46,[61][62][63][64][65][66][67][68][69]. Thus, individuals that undertook vigorous PA presented greater TL than those with moderate activity or those who did not undertake any PA. No differences were observed between the two latter groups [45]. However, in a subgroup analysis in NHANES, the authors reported that the presence of overweight or obesity could counter the protective effect of PA on TL [61]. Moreover, the differences between the degree of PA and TL were only observed among individuals aged 40 to 64 years [69].
In NHANES, individuals reporting vigorous activity had longer TL (reported in bp increase) (than sedentary individuals and those reporting low or moderate PA. Other studies in different populations, also estimated the effect of PA in terms of bp. Thus, Shadyab et al. [73], in a group of post-menopausal women in the Women's Health Initiative (WHI) study, observed a significant difference in bp between those who reported greater PA compared to those reporting the least PA. Moreover, a greater TL was observed in women whose walking speed was faster [73]. Similarly, in women from the Nurses' Health Study (NHS), it was observed that those who undertook vigorous or moderate activity for 2-4 h a week had a greater TL than those less active or sedentary women. Additionally, in the Berlin Aging Study II (BASII), it was estimated that regular PA over the previous 10 years was required to achieve a sustained effect on TL [71].
Furthermore, the role of type of PA, such as easy or fast walking, cycling, or undertaking aerobic or callisthenic exercises, on TL was assessed. Studies note an increase in TL in those performing aerobic or callisthenic exercise only [63]. Nevertheless, these results are contradictory to those found in this same study in a population of lower sample size in previous years [60]. In NHANES, only running was associated with greater TL, this relationship perhaps explaining the benefits of aerobic activity on certain pathologies or on total mortality [70]. Regarding more moderate activities-participants who accumulated more steps or who were in the highest quartile had longer TL than those with fewer steps per day [64]. In longitudinal studies, an increase in exercise-induced caloric expenditure was associated with less shortening of the telomeres over time [73]. Regarding only sedentary individuals in a women's accelerometer study, a difference of 170 bp was observed between the women in the highest sedentary quartile and those of the lowest level. However, when the self-reported sedentary lifestyle of the participants was studied, this was not associated with TL, possible due to the imprecise nature of self-reported questionnaires [75]. With regard to a sedentary lifestyle in a young adult population, it was observed that adults who spent more time watching the television presented shorter TL than those who watched it less [68]. For each hourly increase of television, the risk of having short telomeres increased by 7% [68]. On classifying the subjects into age groups, this association was only significant in adults from 20 to 40 years old [80]. Table 2 [27,[81][82][83][84][85][86][87][88][89][90][91][92][93] shows case-control (n = 11) and longitudinal studies (n = 3) analyzing the association between PA and TL. The results were mixed. In a majority of studies (n = 8/11), TL was longer in physically active cases than in controls. However, in three studies, no differences between TL in cases and control were found [85,88,89]. In general, sample size in the case-control studies was smaller than in the cross-sectional studies and participants characteristics were maximized in terms of PA (for example cases being elite sprinters in comparison with untrained controls [90], or cases being ultramarathon runners in comparison with controls [82]. The methods to assess TL were diverse, including PCR (7/11), Southern blot (3/11) or FISH (1/11). Regarding the type of tissue analyzed, although in general leukocytes were isolated from blood, two studies analyzed muscle [87,89]. Moreover, in one study, gene expression analysis of the Telomerase Reverse Transcriptase gene (TERT) gene was also examined [27]. Specifically, Denham et al. [27] compared several factors associated with cellular aging in endurance athletes with their corresponding active controls and found a greater TL and greater expression of the TERT and TPP1 genes in athletes.
The three prospective studies [91][92][93] also presented mixed results. Weisher et al. [93] found no association between changes in PA and changes in TL over 10 years of followup in a large sample size of participants in the Copenhagen City Heart Study (CCHS). However, Soares-Miranda et al. [91] found an association between changes in PA and in TL in subjects participating in the Cardiovascular Health Study (CHS). Subgroup associations were found in the third study [92]. Table 3 [44, [94][95][96][97][98][99][100][101][102][103][104] shows intervention studies (n = 12) analyzing the association between PA and TL. Several trials (9/11) found that individuals in active PA interventions had an increase in TL relative to controls, but others did not observe such differences (3/11) or reported subgroup results. Again, mixed results were found, mainly due to the great heterogeneity of the interventions, study designs (non-controlled or controlled, and different control groups among the randomized trials), sample size, as well as the participant's characteristics. Thus, seven studies measured TL using only PCR; in one study, PCR was used combined with gene expression or Flow-FISH; one study used FISH, and in two, gene expression (combined with microRNA analysis in one of them), was used. Likewise, some heterogeneity was detected in the tissue analyzed (mainly in blood, but two in muscle and one in saliva). All of this contributed to the possible heterogeneity.
Three intervention trials did not report associations between PA and TL [99][100][101], but even these trials were very heterogeneous. Friedenreich et al. [99], analyzing 212 healthy women aged 50-74 years and comparing 1 year intervention (in two groups: a group with aerobic exercise, 45 min/d, 5 d/week versus an inactivity group) did not find significant association between aerobic exercise intervention and TL. Laye et al. [100] did not observe changes in TL by subjecting eight trained individuals to seven marathons in seven consecutive days. However, there was greater mRNA regulation of the shelterin complex, involved in TL protection [100]. A trial of 439 elderly post-menopausal women [101] compared TL after 12 months of PA and/or dietary intervention, plus a control group, and did not find any differences between the different groups. However, after 12 months, in a subgroup analysis, the women with the shortest baseline telomeres had a greater mean increase in TL than those with the longest baseline TL.
An important consideration for intervention trials is the length of the intervention. Brandao et al. [95] observed that an 8-w intervention combining aerobic and strength training in sedentary women with obesity increased TL. However, it has been reported that different training modalities may exert different effects. Thus, Werner et al. [44] in a complex intervention showed that endurance and interval training increased TL, but not endurance training. In this study, telomerase activity was also measured and was positively correlated with TL. Similarly, Zietzer et al. [104] in an intervention study carried out in youth and elderly participants showed that in the youngest population, telomerase activity increased significantly in a single session of aerobic exercise [104]. When endurance athletes were compared with individuals with an average level of activity, a greater TL was observed compared with the control group. However, on stratifying by age, this difference was not observed in the youngest group (22-27 years) but was much more noticeable in the elderly group (66-77 years) [104].
With regard to the different types of PA, the results have been mixed, mainly due to the heterogeneity of the intervention, tissue, and assessment method and to the small sample sizes. Bicycle endurance exercise has been associated with increased gene expression levels of "Telomeric repeat-containing RNA" (TERRA) that is transcribed from the telomeres and is associated with longer TL, suggesting benefits of aerobic exercise [97]. Likewise, an upregulation of various telomeric genes, such as the TERT or SIRT6 (Sirtuin 6) genes, as well as a differential regulation in microRNA that codifies genes involved in the homeostasis of telomeres was found after 30 min of intense cardio-respiratory exercise in young males [96].

Association between Telomere Length and Tobacco Smoking
Of the 83 articles reviewed, 27 analyzed cigarette smoking and TL, all of which were observational studies (ethical reasons are crucial here explaining the lack of intervention studies). We grouped these observational studies in cross-sectional (Table 4) [124,127], and 5 longitudinal studies [121][122][123]125,126]. Despite some heterogeneity in the results, population characteristics, and methodology used, most studies show an association between tobacco smoking and reduced TL.
Only 2 of the 21 cross-sectional studies report no association between tobacco smoking and TL [60,105]; in one additional study, the initial association between tobacco smoking and TL was non-significant after adjusting for BMI [120]. In another report [108], although smoking status was not associated with TL, there was an inverse association between the number of cigarettes smoked and TL. The initial investigation by the NHS, analyzing the association between smoking and TL in 2284 women, found no relationship between both variables [60]. Two years later, in 5862 women, that same study observed a slight association between TL and smoking, specifically 8% greater TL in subjects who had never smoked as opposed to smokers [117] The authors stated that the combined effect of various risk factors (diet, PA, weight, alcohol, and tobacco) was much greater than the individual additive effect [117]. Various authors have also found greater telomere shortening in smokers as opposed to non-smokers [119,[121][122][123]. Tobacco possibly accounts for 0.64-1.23% of TL variation [116]. More specifically, there could exist 50 bp of variation between smokers and non-smokers [119], 73 bp [125], or 190 bp according to others [112].  Ex-smokers have a lower loss of TL compared to current smokers [119]. Similarly, a significant inverse relationship was found between the number of cigarettes consumed and TL [108,114].
Regarding longitudinal studies (Table 5) [121][122][123]125,126], more discrepancies were found. Although in these studies, significant associations between tobacco smoking and TL were found at baseline, some of these studies, such as the Epidemiologische Studie zu Chan-cen der Verhütung, Früherkennung und optimierten Therapie chronischer Erkrankungen in der älteren Bevölkerung (ESTHER) cohort [125], did not detect longitudinal associations. However, other longitudinal studies [122,123] reported significant associations in the follow-up. In order to explain the potential discrepancies in cross-sectional and longitudinal associations, Zhang et al. [121] analyzed in 5624 participants from an American population. They analyzed differences in TL between smokers and non-smokers following 16 years of follow-up, with concurrent and past smoking status reported biennially, and taking into account the sex influence. The authors observed that smoking was associated with reduction in TL over time, but this association was attenuated in men. The reason for this attenuation was a higher rate of smoking cessation among men with shorter TL, probably because of the poor health of these individual with shorter TL. In conclusion, this study suggested that time-varying and sex-specific associations should be measured in the studies analyzing the association between tobacco smoking and TL.
As with PA, several methodologies were used in studying the association between tobacco smoking and TL. The most frequent method was by PCR, followed by southern blot. However, other techniques were also used. It is interesting to mention the assessment of DNA methylation in several studies [106,109,110], taking into account the strong association between tobacco smoking and DNA methylation.  In addition, some protein biomarkers and gene expression assays have been assessed in various studies (Tables 4 and 5). For example, Song et al. identifying a higher level of different protein biomarkers of DNA damage and dysfunction of the telomeres in smokers compared to non-smokers [77]. Patel et al. could not find any differences on the levels of candidate gene expression between smokers and non-smokers [114]. On the other hand, contradictory findings were reported in a methylation and gene expression study carried out on twins. The authors observed greater telomerase activity and greater TL in smokers than in non-smokers [124].
Lastly, the effect of second-hand smoke or "passive smoking" and TL has been analyzed and it was suggested that high degree of exposure to second-hand smoke could accelerate the shortening of the telomeres that naturally occurs with aging [111]. However, another publication did not find any association between passive smoking and TL [119]. More research is needed to evaluate the effects of second-hand smoke on TL that takes into account various confounders such as air pollution, environmental work pollutants, and other related factors.

Association between Telomere Length and Sleep
In this systematic review, we analyzed 11 articles on the association between TL and sleep (Table 6) [43,[128][129][130][131][132][133][134][135][136][137]. In comparison with the other considered lifestyle factors in this review (PA and tobacco smoking), very few studies focused on sleep. The influence of sleep quality and sleep disorders on aging and on physical and mental health has been recognized more recently [138,139]. In general, aging is associated with poor sleep. Sleep quantity and efficiency vary across the lifespan, sleep duration being shorter, and sleep efficiency being lower at older ages; these difficulties worsen after 60 years, and with some differences between men and women [139][140][141]. Sex-specific differences are complex and vary by outcome [140]. A recent review including 1.1 million people from the Netherlands, United Kingdom and United States, noted that women aged 41 years and older, despite self-reporting shorter sleep duration and lower sleep efficiency than men, had longer and more efficient sleep objectively measured by actigraphy [142], suggesting that the studies analyzing the association between sleep and TL should consider this heterogeneity. Currently there is a great scarcity of intervention trials on this issue.
All but one study identified for this review were observational (Table 6). Moreover, data from the intervention study only analyzed associations between sleep and TL crosssectionally at baseline [132]. Only one study [135] had longitudinal data. Ten articles measured TL using PCR and 1/11 used southern blot [129]. Similarly, most studies have relied on self-reported sleep characteristics, with only two studies assessing objective data (polysomnography, actigraphy, etc.) [131,134], contributing to a low evidence level. Furthermore, even when self-reported sleep characteristics were assessed by questionnaires, high heterogeneity was observed. In several studies sleep quality and duration was assessed using the Pittsburgh Sleep Quality Index (PSQI). The PSQI [143], assesses sleep habits with a scale including 19 items used to derive seven component scores (sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbance, sleep medication and daytime dysfunction. These scores are summed to yield a "global sleep quality score". Some used the global sleep quality score [132] to analyze the association between sleep quality and TL, while others [128] used the sleep quality sub-scale score. More heterogeneity was observed in other studies analyzing the association between the self-reported sleep quality and TL in which the PSQI was not administered [43,[129][130][131][135][136][137]. Another variable commonly investigated was sleep duration. In general, this sleep characteristic was obtained by asking participants to state their average nightly sleep duration. Studies differed in their inclusion of weekend nights and in the response categories used. Some studies also analyzed the influence of insomnia or sleep disturbances on TL [129,135,136] using various questionnaires. Despite this heterogeneity, several studies found a significant association between poorer sleep quality (or sleep duration) and shorter TL [43,[128][129][130]135,137]. However, several of these studies observed subgroup-specific effects. For example, Jackowska et al. [130], found an association between sleep duration and TL in men but not in women. Grieshober et al. [129] noted that the association found between longer sleep duration and longer TL in 3145 postmenopausal women from the WHIS cohort was stronger in African American than White women, and did not reach the statistical significance in White individuals when the analysis was stratified by race. Likewise, Liang et al. [43], observed a significant positive association between sleep duration and TL in the NHS. This study further uncovered a modifying effect of age, such that sleep duration was only associated with TL in women younger than 50. Differences by age were also observed by Cribbet et al., [128], but in the opposite direction.   In this population [128], sleep duration was associated with TL only in older, but not in middle-aged, adults. In addition, mixed results were reported on the effect of long sleep on TL (U-shaped or linear effect). For example, Zhao et al. [137], observed longer TL in Mexican Americans who reported sleeping at least 9 h/night, whereas Tempaku et al. [134] in Brazil, observed associations between longer sleep and shorter TL. Cribbet et al. [128] also described a U-shaped association between sleep duration and TL, with the extreme of sleep duration having the shortest TL. Interestingly, in one study examining TL in different types of leukocytes [132], poorer global sleep quality was associated with shorter TL in lymphocytes, but not in granulocytes. This is the only study analyzing the sleep influence in different cell types and additional studies are needed to better understand this relationship. Furthermore, Nguyen et al. [131], using objective measures of sleep (actigraphy), did not find any association between sleep duration, and most other sleep metrics, and TL.
Only one study was identified that assessed the relation between chronotype and TL [135]. Individuals with an evening chronotype had a shorter TL. However, when longitudinal analyses were performed, chronotype did not predict changes in TL. Overall, despite many observational studies reported associations between self-reported sleep duration and sleep quality and TL, these results are mainly cross-sectional, limiting the ability to draw causal inference. More longitudinal and intervention studies are needed to increase our understanding on how poor sleep characteristics affects TL.

Discussion
This systematic review evaluated the current evidence on the influence of three relevant lifestyle factors (PA, tobacco smoking and sleep characteristics) on TL in adults. In the selected studies on TL, PA was the most studied lifestyle factor, followed by smoking and sleep. In general, being physical active, not smoking and having good sleep quality are associated with a reduced risk of morbimortality for the main chronic diseases [8][9][10][11][144][145][146][147]. There is a challenge in searching for the best biomarker of aging and despite these limitations and the advantage of measuring a combination of biomarkers instead of only one [148,149], TL is the most extensively studied biomarker of age-related diseases [7,. In this regard, many results have been obtained focusing on the association between TL, aging, cardiovascular diseases, diabetes, obesity and mortality [35][36][37][38][39][40]148,149]. Therefore, the demonstration of an effect of these lifestyle factors on TL could provide important information on the mechanisms by which these factors have protective health effects, as well as providing evidence to validate TL as a good biomarker of aging with a good capacity to capture healthy lifestyle interventions based on these factors. However, we have detected that the evidence level of the results obtained on the association among these lifestyle factors and TL is low, these being mainly based on crosssectional studies. Thus, in this review, no intervention studies were detected analyzing the effect of sleep characteristics on TL. Likewise, due to ethical limitations, no experimental publications have analyzed the effects of tobacco smoking on TL on humans. Regarding PA, there are several intervention studies (12/49) examining the effect of different PA intervention on TL. However, the complexity of the PA interventions hinders comparison between publications and precludes definitive conclusions. Moreover, there is a scarcity of longitudinal studies for all three lifestyle factors. In addition, most of the articles have relied on the self-reported measurement of PA, smoking or sleep parameters, instead of using objective determinations (accelerometers, plasma/urine cotinine, actigraphy, respectively, among others). Moreover, substantial heterogeneity existed across the review studies for several study parameters including sample size and statistical power, participant characteristics (sex, age-group, ethnicity, etc.); biological sample and method used to analyze TL (although in general TL was measured by PCR in leukocytes isolated from blood); and statistical control for potential confounders (adjustment for anthropometric and other lifestyle factors). All of this may contribute to the variations in observed results.
Despite the low-level of scientific evidence provided by the cross-sectional studies analyzed for the three lifestyle factors, and the fact that we cannot discuss the associations in terms of causality, in general, most studies [43,44,46, have reported (for the whole population or at the subgroup level) that physically active individuals have longer TL; that tobacco smoking is associated with shorter TL; and that a poor sleep quality and duration were related to shorter TL. Some studies have postulated several mechanisms by which PA may have an effect on TL. It is speculated that exercise can induce different signaling mechanisms, either epigenetic or transcriptomic [69,150]. Chilton et al. [151], in a comprehensive review on the relationship between exercise, aging and telomeres in human and animal models discussed the lack of good evidence at the mechanistic level to support this association, but suggested reductions in oxidative stress and inflammation as potential mediators. As noted in this systematic review, the optimum amount and the type of PA for maintaining TL remains under debate. Lack of agreement among studies could result from different factors, such as: time-need for interventions longer than 12 months to see the long-term effect of PA on telomeres; variability in trials-major random error and consequent lower precision in estimating the effect and low statistical power in the intervention studies due to the complexity of the intervention; or specific characteristics of the participant that could hinder the protective effect of exercise on telomeres [99]. In addition, a U-shaped relationship is thought to exist between endurance exercise and the regenerative capacity of skeletal muscle [72,89]. The possible negative effect on TL of high intensity exercise could be due to an inducing oxidative stress through an increase in reactive oxygen metabolites and free radicals [150].
Tobacco is one of the main risk factors for cardiovascular events and mortality. In addition, those who start smoking at an early age and continue during adulthood lose a decade of life expectancy compared to never smokers [152]. Tobacco smoking is a source of free radicals and oxidants that increase inflammatory response and oxidative stress [111,119,153]. Both processes could increase telomere shortening, contributing to the acceleration of aging and possibly an increased risk of age-related diseases [115,119,120,125]. Although several cross-sectional studies have reported an inverse association between tobacco smoking and TL, evidence for the causality of this association is low and several confounding factors may play a role.
Finally, interest in analyzing sleep and health status has increased in recent years. The relation between short and long sleep with numerous cardiometabolic pathologies has been studied, and a U-shaped association has been demonstrated [154,155]. However, some inconclusive results exist due to uncertainties in self-reported sleep characteristics, different methodological designs and the different populations analyzed [154][155][156]. Furthermore, the sleep process is highly affected during aging, and elderly individuals suffer the greatest disturbances in their sleep patterns [157]. Several authors claim that aging or the presence of other comorbidities could be the cause of the association found between sleep and TL [131,136]. Our systematic review suggests that there is a possible association between shorter TL and various sleep-related parameters, but without establishing a causal relationship between the two factors. Therefore, more longitudinal and intervention studies on sleep and TL are needed. The methodological differences in the evaluation of sleep and TL increase the inconsistency in the results obtained.
In general, a great number of studies analyzed for PA, smoking and sleep reported statistically significant results in subgroup analysis. These associations may be false positive associations due to multiple comparisons or real associations based on heterogeneous effects such as the previously described differences per sex [158] or per age groups [139]. Furthermore, the beneficial effect of PA, non-smoking, or quality sleep on TL may be due to the interaction between various lifestyle factors, such as those studied in this systematic review, or even with others (diet, educational level, etc.), and more studies are needed to assess the role of healthy lifestyle factor combinations in diverse populations.

Conclusions
The results obtained in this systematic review show that the influence of PA, smoking, and sleep on TL of adults is mixed, and depends on the type of study, population analyzed, and methodology (study design, exposure and TL assessment, subgroup analysis, etc.). The evidence level of the associations is very low due to the cross-sectional nature of the studies. Future research using prospective and well-powered intervention trials with standardized protocols and objective measures is needed. Despite these limitations, current studies suggest a low-evidence level association between moderate, regular, and daily non-sedentary activities and longer TL. Likewise, not smoking and having optimal sleep duration (7 h) and quality, are associated with longer TL. Further research is needed to improve design and reduce heterogeneity and to consider the diversity by sex, age, ethnic groups, and other lifestyle characteristics. Funding: This study was partially funded by the Conselleria de Innovación, Universidades, Ciencia y Sociedad Digital de la Generalitat Valenciana (grants PROMETEO 17/2017, PROMETEO 21/2021 and APOSTD/2019/136); and the Spanish Ministry of Health (Instituto de Salud Carlos III) and the Ministerio de Economía y Competitividad-Fondo Europeo de Desarrollo Regional (FEDER) (grants PI19/00781 and SAF2016-80532-R). This study is also supported by the grant PID2019-108858RB-I00 funded by AEI 10.13039/501100011033 and by "ERDF A way of making Europe"; and the University Jaume I (grant UJI-B2018-69). The APC was funded by the University of Valencia.