A systematic review was conducted and the recommendations from the Preferred Reporting Itens for Systematic Review and Meta-Analyses (PRISMA) were considered (Liberati et al., 2009). The search strategies were reported to ensure the integrity of the results and allow the updating using the same methods to bring emerging evidence into the review. The Boolean and proximity operators were used and the search strategy was correctly adapted for each database used (Table 1) (Sampson et al., 2008, 2009). Studies were identified by searching the following electronic databases: PubMED/MEDLINE (via National Library of Medicine) (2000–2017) and Science Direct (Elsevier) (2000–2017). The last search was conducted in February 2017.

Once the abstracts were reviewed, the complete versions of the papers that met the criteria were obtained. In addition, the reference lists of the papers that fulfilled the inclusion criteria were analyzed for identification of additional studies. The exclusion of studies with irrelevant content and duplicates was carried out after the title, abstract and full-text were read.

An “athlete” was defined according to the Medical Subject Headings (MeSH) and was considered to be an individual who has developed skills, physical fitness and strength, or who has participated in sports running (MeSH, 2015). We have considered the definition proposed by Stirling and Kerr (2006) that defines a “recreational athlete” as being an individual who plays on a sports team at an amateur level, works out 1–4 times a week, does not train and compete nationally or internationally, and does not train for the same activity for more than 8 h per week. Novice runners were those individuals who had not been running on a regular basis in the previous 12 months (10 km total in all training sessions in the previous 12 months), and recreational runners were considered as individuals running a mean of 24.94 km/week (Videbaek et al., 2015).

The following thesaurus terms registered in the database from MeSH were also used: “running,” “aerobic exercise,” “inflammation,” and “cytokines.” These terms were associated with the free terms “recreational runners,” “novice runners,” “marathon runners,” and “ultramarathon.”

The inclusion criteria were as follows: randomized controlled trials (RCTs) and non-randomized controlled trials (NRCTs); studies investigating the acute and chronic effects of running on markers of inflammation in marathon runners, recreational runners and novice runners; the terms runners, marathon runners, recreational runners and novice runners should be cited in the paper; only healthy participants; only full-text article citations with no restriction on languages; with individuals aged over 19 as the World Health Organisation (WHO) defines adolescence as the period in human growth and development that occurs between childhood and adulthood, from ages 10–19 (WHO, 2015)¹. Meeting abstracts, unpublished data, observational studies, review articles, studies using walking and jogging as independent variables, and studies on the effects of any kind of supplements on running, diet restrictions, use of devices (e.g., equipment, compression garments), comparisons between running and other sports, and effects of environmental conditions (ex. dry and hot) were excluded.

The outcome measures assessed for acute and chronic effects of marathon and recreational running were interleukin (IL): IL-6, IL-10, IL-8, IL-1ra, IL-1β, IL-2 and IL-12, TNF-α, C reactive-protein (CRP), interferon-gamma (IFN-γ), soluble receptors, and transformation growth factor-beta (TGF-β). These mediators were chosen after an initial analysis and review of the literature. They were identified as the main outcomes in studies published with marathon runners, recreational runners and novice runners (Nieman et al., 2005; Santos et al., 2007; Scott et al., 2011; Abbasi et al., 2013; Jee et al., 2013; Shin and Lee, 2013).

The quality and assessment of all eligible articles was evaluated using a modified version of the Downs and Black checklist (Downs and Black, 1998). Disagreements between authors were discussed and subsequently solved. This modified version consists of 27 objective questions (Downs and Black, 1998).

Results of the research strategy are presented in Figure 1. Initially, 60 studies were selected, with 51 studies being finally included according to the inclusion/exclusion criteria. Nine studies were excluded as follows: one study was excluded due to the use of heat stress, three because the subjects were adolescents, one following the reading of the full paper, two because of comparison with other sporting activities, and one because of medication use; in addition, one paper was not available in full-text version (Saravia et al., 2010). A total of 49 studies verified the acute effect of running on inflammation and two studies focused on the chronic effects.

Quality assessment of the studies according to the modified Downs and Black scale is summarized in Table 2. One important finding was that characteristics of the patients' included were not cleary described in 32 studies. Important adverse events and description of patients' characteristics lost to follow-up was not reported in 34 and 32 studies, respectively. None of the studies were randomized controlled trials and power was provided in 4 studies (Table 2).

An overview of the studies' characteristics is provided in Table 3 with sample size, age, sex, and exercise protocols. A summary of outcome measures in selected studies is presented on Table 4.

The 51 studies included resulted in a total of 1,421 subjects, of whom 163 were female and 1,234 males; 24 subjects were not identified by sex in one study (Neidhart et al., 2000). All trials provided age ranges for the subjects and the mean age was 39.16 years.

The common protocols adopted included marathon in 17 studies, ultra-marathon in 22, half-marathon in three studies, different distance protocols (42.195, 21.1, 12, 10 km and treadmill) in seven studies, and chronic training only in two studies (see Table 3).

The timing in data collection could be considered an important limiting factor for adequate comparisons, given the heterogeneity observed across the reported acute (i.e., 5–20 min) (Fallon et al., 2001; Kim et al., 2007; Vaisberg et al., 2013; Grabs et al., 2015) and delayed responses (24 h to 8 days) (Uchakin et al., 2003; Siegel et al., 2007; Hewing et al., 2015). This is not a simple issue given the different kinetics and biological availability of the molecules considered as IL-6 and CRP (Kasprowicz et al., 2013; Reihmane et al., 2013). Of note, some inflammatory markers (e.g., monocyte chemoattractant protein-1, granulocyte colony-stimulating factor) (Suzuki et al., 2003) have not been included in the current review. Further studies should elaborate appropriate study designs that consider both the appropriateness of the inflammatory markers selected as well as their kinetics.

Another important confounding factor is the different experience of runners. Thus, 26 studies (4 studies with male and female participants, and 15 studies with only male participants) reported a range of 4–17.5 years of experience with competitions (e.g., marathon, and ultra-marathon experience) (Schobersberger et al., 2000; Fallon et al., 2001; Nieman et al., 2003, 2016; Suzuki et al., 2003; Tomaszewski et al., 2003; Kim et al., 2007, 2015; Millet et al., 2011; Jee and Jin, 2012; Vaisberg et al., 2012, 2013; Waśkiewicz et al., 2012; Bernecker et al., 2013; Jee et al., 2013; Karstoft et al., 2013; Kasprowicz et al., 2013; Kłapcinska et al., 2013; Saugy et al., 2013; Shin and Lee, 2013; Wilhelm et al., 2014; Grabs et al., 2015; Hewing et al., 2015; Mrakic-Sposta et al., 2015; Krzeminski et al., 2016; Mohamed et al., 2016). However, 25 studies did not report this information (Fehrenbach et al., 2000; Mattusch et al., 2000; Neidhart et al., 2000; Niess et al., 2000; Suzuki et al., 2000; Bonsignore et al., 2002; Uchakin et al., 2003; Peters et al., 2004; Siegel et al., 2007; Ng et al., 2008; Papassotiriou et al., 2008; Chimenti et al., 2009, 2010; Auersperger et al., 2012; Nickel et al., 2012; Rehm et al., 2013; Santos et al., 2013; Bachi et al., 2015; Gill et al., 2015b; Arakawa et al., 2016; Nielsen et al., 2016; Stuempfle et al., 2016; Vernillo et al., 2017). This is not a trivial issue given that training experience of runners could have a potentially additive effect to the influence of runners' age on inflammation that warrants further research. Besides, only 34 articles especified the training preparation (km/week that ranged between 21.4 and 161 and time that ranged between 3.9 and 10 h/week) of runners before races (Fehrenbach et al., 2000; Mattusch et al., 2000; Niess et al., 2000; Suzuki et al., 2000, 2003; Fallon et al., 2001; Bonsignore et al., 2002; Nieman et al., 2003, 2016; Tomaszewski et al., 2003; Uchakin et al., 2003; Peters et al., 2004; Chimenti et al., 2009; Millet et al., 2011; Auersperger et al., 2012; Jee and Jin, 2012; Nickel et al., 2012; Vaisberg et al., 2012, 2013; Waśkiewicz et al., 2012; Bernecker et al., 2013; Karstoft et al., 2013; Kłapcinska et al., 2013; Rehm et al., 2013; Reihmane et al., 2013; Santos et al., 2013; Shin and Lee, 2013; Wilhelm et al., 2014; Bachi et al., 2015; Grabs et al., 2015; Hewing et al., 2015; Mrakic-Sposta et al., 2015; Arakawa et al., 2016; Krzeminski et al., 2016). Furthermore, only two studies cited the control of intensity during training (Mattusch et al., 2000; Auersperger et al., 2012). This aspect would be important for a better understanding of the relationship between running and inflammation from a dose-response perspective. For instance, higher values for IL-6 after a limited-time test were observed in sedentary individuals when compared to long- and middle-distance runners, but with no differences between groups for TNF-α (Mohamed et al., 2016). Given the growing popularity of races that last various days, further studies are warranted to elucidate if chronic exposure to high volumes of endurance running are detrimental for health. From an evolutionary perspective, this is an interesting topic given that daily running volumes of modern hunter-gatherers are far below (e.g., ~10–15 km) (O'Keefe et al., 2010; Boullosa et al., 2013) the training and competitive volumes of runners competing in ultra-endurance events. In addition, another new aspect that must be raised in further studies is the imbalance between training phases and recovery not reported in the studies included in this systematic review. Thus, since functional overreaching might be related to inflammation (Steinacker et al., 2004), further studies should explore these relationship in conjunction with other biological markers of overreaching.

Another important characteristic for a better characterization of runners is their fitness level. For instance, maximum oxygen consumption (VO₂max), which is the gold standard for aerobic evaluation, has been reported only in 14 articles (3 studies with participants of both sexes (Nieman et al., 2003, 2016; Ng et al., 2008; Chimenti et al., 2009; Jee et al., 2013), and 11 studies with males (Millet et al., 2011; Jee and Jin, 2012; Waśkiewicz et al., 2012; Kłapcinska et al., 2013; Shin and Lee, 2013; Wilhelm et al., 2014; Kim et al., 2015; Krzeminski et al., 2016; Mohamed et al., 2016). The participants of these studies could be considered recreational runners when classified by their actual VO₂max (44–51 and 35–41 mL.kg-1.min-1, for males and females, respectively) (Martin and Coe, 2007). In contrast, no study included elite runners when classified by their actual VO₂max (70–85 mL.kg-1.min-1 and 61–73 mL.kg-1.min-1 for males and females respectively) (Martin and Coe, 2007). Moreover, runners' classification in the current review has been challenging when using the selected criteria (see Table 3) (Stirling and Kerr, 2006; MeSH, 2015). Thus, further studies should provide all these informations for a better characterization of runners. As we did not perceive an influence of aerobic fitness on inflammatory markers, further studies should elaborate on this relationship while controlling other runners' characteristics as training experience. Additionally, the influence of other fitness components as muscle strength capacity should be assessed in further studies for verifying the potentially protective effect for muscle damage and therefore on inflammation.

Another important limitation for generalization of the results refers to the heterogeneity of running exercises (e.g., distance, intensity) used for evaluation of acute inflammatory responses. Furthermore, ambient characteristics (e.g., altitude, temperature) and race profile (e.g., uphill and downhill running) which have been suggested to influence muscle contraction and physiological responses (Vernillo et al., 2017), have not been always reported (Millet et al., 2011; Saugy et al., 2013). These aspects should be controlled in further studies for isolating the relative effect of every specific factor on inflammation.

Given the relationship between adipose tissue and inflammation (Pedersen and Febbraio, 2012), attention should be paid to overweight and obese runners. For instance, a higher level of CRP 24 h following a marathon has been observed in obese non-elite runners when compared to lean elite runners (Nickel et al., 2012). Furthermore, obese non-elite runners when compared to lean elite and lean non-elite runners demonstrated a higher level of IL-6 and a lower level of IL-10 serum levels at baseline (Nickel et al., 2012). However, all groups presented an increase for serum IL-10 and TNF-α, and a decrease for serum IL-6 levels, immediately post-marathon (Nickel et al., 2012). It must be considered that increments in IL-10 induced by exercise may be responsible for the elevation in IL1-ra which exerts an anti-inflammatory action by antagonizing IL-1 and IL-1β (Dinarello, 2000; Moldoveanu et al., 2001; Petersen and Pedersen, 2005; Pedersen, 2011). Nevertheless, it could be suggested that, in overweighted individuals, a higher pro-inflammatory status at baseline and post-marathon could be observed, with unknow consequences for health in the long term.

One relevant issue refers to the link between inflammation and cardiovascular health. Interestingly, the exercise-induced increase of IL-6 after the marathon in 20 lean male runners was associated with a lower prevalence of arrhythmias during and after the marathon race (Grabs et al., 2015). When produced by muscle contraction, IL-6 stimulates the synthesis of other anti-inflammatory cytokines such as IL-1ra and IL-10, thus providing an inhibitory effect on pro-inflammatory cytokines such as IL-1β and TNF-α (Pedersen and Febbraio, 2012; Pedersen, 2013). However, CRP, a strong predictor of cardiovascular events, is an acute phase protein synthesized in the liver by the stimulation of IL-6 (Ridker et al., 2002). Chronic endurance training may decrease CRP values, especially when accompanied by a loss in fat mass, therefore promoting further reduction of risk for cardiovascular events (Fallon et al., 2001; Tomaszewski et al., 2003; Walsh et al., 2011; Grabs et al., 2015; Kim et al., 2015). Of note, CRP may be more susceptible to chronically decrease in individuals presenting higher baseline levels (Barnett et al., 2005). Therefore, caution should be taken when evaluating the anti- and pro-inflammatory effects of running in individuals with different characteristics regarding cardiovascular risk factors (e.g., body composition) (Moldoveanu et al., 2000; Petersen and Pedersen, 2005; Walsh et al., 2011).

Most studies included in this systematic review were acute interventions (49 studies). However acute changes in inflammatory markers might not be related with anti- and pro-inflammatory outcomes during chronic aerobic training interventions. For instance, there were divergent responses for CRP changes in chronic interventions (Mattusch et al., 2000; Auersperger et al., 2012). Thus, while Mattusch et al. (2000) observed a reduction in CRP levels, Auersperger et al. (2012) did not observe any change. Therefore, further studies must consider this important limitation, while providing training load charactersitics as volume, intensity, and frequency of training sessions. An important question to be answered refers to the minimal training load required for runners of different levels when preparing different competitive distances, while analyzing the impact of these factors on inflammatory markers. Additionally, there is a prevalence of male runners on literature therefore more studies with female runners are needed.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.