Marathon running is predominantly determined by maximal oxygen uptake (V[O.sub.2] max), lactate threshold, running economy and oxygen uptake kinetics (Joyner and Coyle, 2008). It has been acknowledged that to reach elite-level performance a synergy of physiological and psychological traits, combined with an optimal environment are required (Tucker et al., 2013). Physiological parameters of heritability such as V[O.sub.2] max for endurance performance indicate significant genetic components. For example, Bouchard et al. (1999) demonstrated that the heritability of V[O.sub.2]max falls in the range of 40-50%, indicating such parameters present a considerable genetic influence. In addition, elite marathon runners of Kenyan or Ethiopian descent account for 90 of the top 100 marathon running performances of all time and hold the six previous world records (International Association of Athletics Federations, 2018). Based on this geographical concentration of unparalleled success, assertions of a genetic endowment for marathon running performance have been further amplified (Tuker et al., 2013; Vancini et al., 2014). Therefore, a major focus of sport genomics has been to identify specific genes and their polymorphisms associated with elite-level performance to facilitate in the identification of future athletic talent (Roth, 2012; Popovski et al., 2016). Specifically, one gene that has been extensively studied due to its associated role in endurance performance is angiotensin I-converting enzyme (ACE) (Gayagay et al., 1998; Montgomery et al., 1998). The ACE gene functional polymorphism is based on either the presence (insertion [I]) or absence (deletion [d]) of an intron sequence. The insertion (I) allele has been positively associated with elite endurance running performance in Caucasians due to lower circulating and tissue ACE activity (Myserson et al., 1999). No significant differences in the ACE I/deletion (D) genotype were found between Kenyan elite marathon runners and their respective general population (Scott et al., 2005). This ethnic disparity led to the contention that the East African running phenomenon is not a genetically mediated one (Wilber & Pitsiladis, 2012). On the contrary, the existing research to date, may be interpreted as evidence for the polygenic nature of complex endurance-related traits, due to the basis of researching the broader context of endurance performance, but also highlighting the limitations of a single candidate gene approach (Ahmetov et al., 2016). Furthermore, the use of case-control studies has given rise to inconsistent findings (Drozdovska et al., 2013; Tural et al., 2014), which may be attributed to small sample sizes resulting in insufficient statistical power to demonstrate significant effects of numerous genes each with small contributions (Wang et al., 2013). Where previous studies have been conducted on general endurance performance rather than specific to elite marathon running, ensuring sufficient statistical power is a particular challenge in elite-level performance since elite athletes are by definition, limited to a small number of superior individuals. Therefore, it has been suggested that the impact of these issues can be reduced by pooling single studies into a meta-analysis (Lopez-Leon et al., 2016). To date, meta-analyses have confirmed the associations of ACE and peroxisome proliferator-activated receptor alpha (PPAR[alpha]) genes with endurance performance (Ma et al., 2013; Lopez-Leon et al., 2016). However, it has been suggested that at least 93 genetic markers are associated with elite endurance athletic status (Ahmetov et al., 2016). Yet, to date, no review has identified the genetic markers associated specifically to elite marathon running per se.
Therefore, the objectives of the current systematic review were to critically examine the evidence concerning genes that may play a role in the performance of elite-level international marathon athletes, and to provide insight into the predictive utility of genetic testing in identifying future marathon success. This investigation of multiple genes, and their polymorphisms, is advantageous as it allows a more comprehensive picture of genotype-phenotype relationships to emerge, thus improving the future application of genomics as a practical tool in marathon talent identification.
A published literature search that investigated the association between genes and marathon performance was conducted up to May 2018 and obtained according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009). A predetermined search strategy was conducted on systematic databases for publications indexed in PubMed, SPORTDiscus and Google Scholar using combinations of the following search terms: "genes" OR "genetics" OR "genomics" AND "marathon" OR "endurance" OR "elite performance" in the titles of papers. Additional publications were also considered by cross-referencing. A hand search of the reference lists of articles included in the final analysis that were identified via the database search was conducted, as were the first 20 "related articles" of those included database search articles on PubMed.
All publications retrieved were screened by title and any duplicates or those irrelevant to the research question were removed. Abstracts of the remaining studies were then similarly screened and 24 studies were selected for full-text assessment against the predetermined inclusion and exclusion criteria outlined below.
Inclusion and exclusion criteria
The present review included case-control, cohort and genome-wide association studies (GWAS). To be included, studies were required to provide data on the genotypes associated with elite international marathon athlete status and/or marathon running performance published in peer-review journals. Those studies identified as 'elite' were based on participants having been international competitors and/or national representatives in the marathon distance as determined by entry standards from the International Association of Athletics Federations (IAAF, 2018). There were no restrictions applied regarding the age, gender or ethnicity of the participants. Studies were excluded if they were: (i) review articles, congress abstracts, editorials or other non-original articles; (ii) reported in a language other than English. Overall, 14 studies were included for qualitative synthesis. The study selection process and reasons for exclusion are illustrated in Figure 1.
Studies were assessed for inclusion by two independent reviewers, RK and DF, with disagreements resolved by discussion, and arbitration (HJM) if necessary. The reviewers, RK and DF were not blinded to authors, institutions or journals of publication. If a decision on whether to include or exclude a paper could not be made from the title and abstract, full text was obtained and checked.
Data extraction and quality assessment
For all selected studies, the following data were extracted: (i) first author name; (ii) publication date; (iii) participant characteristics; (iv) study design; (v) genetic markers measured; (vi) allelic frequencies; and (vii) bias. These outcomes were extracted for the narrative review.
Risk of bias assessment
The risk of bias of individual studies was assessed using the Cochrane Collaboration's risk of bias tool (RoB 2.0; Higgins et al., 2016). Studies were given an overall risk of bias grade of either "high", "some" or "low" calculated from the following five domains: a) sequence generation, b) allocation concealment, c) blinding, d) missing outcome data, e) selective reporting of results. If details for a particular domain were insufficient, the risk of bias was assessed as "unclear". Assessments were performed independently by two authors (RK and DF) with disagreements resolved by discussion, and arbitration (HJM) if necessary. Components were assessed independently, as per PRISMA (Moher et al., 2009) and Cochrane collaboration (Higgins et al., 2016) recommendations.
Qualitative analysis was carried out on the studies selected focusing on the association between genetic markers, elite marathon athlete status and running performance. A narrative review was provided in text and tables to summarise study characteristics. If deemed appropriate, a meta-analysis was planned for single nucleotide polymorphisms (SNPs) that had been investigated in at least three studies. However, a meta-analysis was not performed due to the heterogeneity of the study characteristics and SNPs measured with very few replication studies. Where data was not available in the full-text, original authors were contacted.
Summary of studies retrieved
The search strategy identified 241 studies, of which, following cross-referencing and exclusions, 14 full-text articles were identified and included in the qualitative synthesis (see Figure 1).
The characteristics of the 14 studies are summarised in Table 1. A total of 160 different polymorphisms in 27 genes were studied in 10,442 participants, of which 2,984 were marathon distance runners, 6,109 were non-athlete controls or 310 power athletes. Of the 160 polymorphisms, 159 were collectively investigated in two studies (Tsianos et al., 2010; He et al., 2015), where a total of 139 polymorphisms were studied in one paper alone by He et al., (2015). Where, the study investigated 84 polymorphisms in two genes, of which there were no reported associations (He et al., 2015). Only the ACE SNP I/D rs4646994 was investigated in three studies (Amir et al., 2007; Ash et al., 2011; Tobina et al., 2010). Thirteen of the studies were case-control or cohort designs with only one applying the new GWAS approach (Ahmetov et al., 2015), although GWAS was not applied to the comparison of the athlete groups with the control groups. While 11 studies...