Methods of Monitoring Internal and External Loads and Their Relationships with Physical Qualities, Injury, or Illness in Adolescent Athletes: A Systematic Review and Best-Evidence Synthesis

Background With the increasing professionalisation of youth sports, training load monitoring is increasingly common in adolescent athletes. However, the research examining the relationship between training load and changes in physical qualities, injury, or illness in adolescent athletes is yet to be synthesised in a systematic review. Objective The aim of this review was to systematically examine the research assessing internal and external methods of monitoring training load and physical qualities, injury, or illness in adolescent athletes. Methods Systematic searches of SPORTDiscus, Web of Science, CINAHL and SCOPUS were undertaken from the earliest possible records to March 2022. Search terms included synonyms relevant to adolescents, athletes, physical qualities, injury, or illness. To be eligible for inclusion, articles were required to (1) be original research articles; (2) be published in a peer-reviewed journal; (3) include participants aged between 10 and 19 years and participating in competitive sport; (4) report a statistical relationship between a measure of internal and/or external load and physical qualities, injury or illness. Articles were screened and assessed for methodological quality. A best-evidence synthesis was conducted to identify trends in the relationships reported. Results The electronic search yielded 4125 articles. Following screening and a review of references, 59 articles were included. The most commonly reported load monitoring tools were session ratings of perceived exertion (n = 29) and training duration (n = 22). Results of the best-evidence synthesis identified moderate evidence of positive relationships between resistance training volume load and improvement in strength, and between throw count and injury. However, evidence for other relationships between training load and change in physical qualities, injury, or illness were limited or inconsistent. Conclusions Practitioners should consider monitoring resistance training volume load for strength training. Additionally, where appropriate, monitoring throw counts may be useful in identifying injury risk. However, given the lack of clear relationships between singular measures of training load with physical qualities, injury, or illness, researchers should consider multivariate methods of analysing training load, as well as factors that may mediate the load–response relationship, such as maturation. Supplementary Information The online version contains supplementary material available at 10.1007/s40279-023-01844-x.


Introduction
Training and physical activity are integral for physical development [1]. When an athlete completes a training session, there is an acute increase in fatigue, which, with recovery, is then typically followed by a supercompensatory response [2]. Improving physical qualities has previously been shown to improve physical performance [3,4], decrease injury risk [5], improve recovery [6], and influence selection [7] in team sports, and therefore forms a significant focus of the training process. However, without adequate recovery following training, the athlete may suffer decreased performance and potentially injury or illness [8,9]. This relationship was originally referred to as the general adaptation syndrome [2], and despite this model having undergone refinement [10], the principle of providing a sequentially greater training stimulus, followed by adequate rest and recovery, remains the premise on which most modern training programmes are based. Colloquially, this balance between fitness and fatigue has been termed the 'Goldilocks effect' and highlights the need to understand both the positive and negative responses to training load [11].
To ensure appropriate prescription of training and rest, load monitoring programmes are often implemented, particularly in elite sport [12]. However, with the increasing professionalisation of youth sports, greater emphasis is being placed on quantifying the training loads of adolescent athletes [13][14][15]. There are both internal and external methods of monitoring training loads. External methods of monitoring load measure the work performed by an athlete, including resistance training volume load (sets × reps × load) and running metrics through global navigation satellite systems (GNSS) [16]. Alternatively, internal load monitoring methods capture the physiological (e.g., heart rate; HR) and psychophysiological (e.g., session rating of perceived exertion; sRPE) responses to the external load [16]. In comparison to external load, internal load is a more accurate measurement of the individualised response to training stress [17]. However, it is challenging to prescribe training based on internal load, as this is influenced by numerous factors, for example, hydration status [18]. Therefore, it is often more practical to prescribe training based on external loads. Given the limitations of internal and external load metrics, both internal and external loads will often be integrated in a load monitoring regimen.
Throughout adolescence, an athlete's response to training load will change due to factors such as maturation and training exposure [19], and therefore they are likely to have fluctuating responses to training load. For example, changes in sex hormones throughout maturation facilitate greater strength and hypertrophy adaptations [19,20]. Given the unique environment of adolescent athletic development, multiple attempts at developing training models to optimise adolescent athletic development have been proposed, such as the long-term athlete development model [21] and the youth physical development model [22]. These models propose that the development of certain physical qualities should be emphasised at different points throughout maturation. This highlights the need for a systematic review of the literature to understand current evidence about the complex nature of the load-response relationship in adolescent athletes.
Given the increased focus on training load monitoring in adolescent athletes, a systematic review of the literature is appropriate to guide practitioners and researchers on the relationship between methods of monitoring training load and physical qualities, injury, or illness. Subsequently, the aim of this systematic review was to detail the methods of reporting internal and external loads in adolescent athletes and describe their relationship with changes in physical qualities, injury, or illness.

Design and Search Strategy
This review was registered via PROSPERO (CRD42021245503). An electronic search was conducted of the CINAHL, SPORTDiscus, Web of Science, and SCO-PUS databases. Search terms and strategy are reported in Table 1. Search terms were crafted by reviewing known original research and reviews relevant to the topic [23]. No searches were mapped to medical subject heading terms. The search strings were initially searched independently and then combined with AND. Strings were adjusted based on database-specific truncation, wildcard, and proximity operators. The search was restricted to studies published in English. Articles were retrieved from the earliest possible date until March 2022.

Inclusion and Exclusion Criteria
The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines were followed to screen articles [24]. Article screening was performed by CD and JW; a third reviewer (RJ) was used to resolve any conflicts. Inclusion criteria were original research investigations, full-text articles written in English, published in a peer-reviewed academic journal, with participants aged 10-19 years old who participated in competitive sport [25]. Competitive sport was defined as any game or activity that involves physical exertion and skill, played against other teams or individuals [26]. Additionally, all studies were required to report a statistical relationship between a measure of internal or external training load and physical quality, injury or illness. Manuscripts were excluded if they were commentaries, letters, editorials, conference proceedings, case reports, conference abstracts or non-peer-reviewed articles and studies with < 1 week of load monitoring or alterations to load such as 'shock periods' [27].
Both observational and intervention-based studies were included, provided there was an indication of the relationship between load and change in physical quality, injury, or illness. Load was defined as "the cumulative amount of stress placed on an individual from multiple training sessions (structured or unstructured) over a period of time." [28]. Physical quality was defined as any test of an element of fitness, such as strength, power, endurance, or speed. Illness was defined as any non-musculoskeletal medical reporting event. Additionally, injury was defined as a medical reporting event, whether or not it resulted in time loss [29]. Due to various methods of reporting injury and illness data, the definitions were deliberately kept broad. Finally, studies were included if they reported either the incidence or burden of injury (hours or sessions of training lost).

Assessment of Study Quality
A modified Downs and Black [30] checklist was used to assess methodological quality by a single reviewer (CD) (Supplementary Material 1, see electronic supplementary material [ESM]); if clarification was required for any of the studies, a second reviewer was consulted (JW). This checklist has previously been used in sport science systematic reviews that similarly included a variety of study designs [31]. Items were scored as 1 (yes) or 0 (no or unable to determine), with a maximum score of 12.

Data Extraction and Analysis
Data were extracted by CD from included studies into a custom Google spreadsheet (Alphabet, Mountain View, Table 1 Search terms used   Variable Search strings Adolescent Adolescen* OR teen* OR Pubescent OR junior OR "School athlet*" OR youth* OR "Under#11" OR "Under#12" OR "Under#13" OR "Under#14" OR "Under#15" OR "Under#16" OR "Under#17" "Under#18" OR "Under#19" Athletes archer* OR athlete* OR baseballer* OR basketballer* OR batsm?n OR boarder* OR bobsledder* OR bowler* OR boxer* OR canoeist* OR cricketer* OR cyclist* OR dancer* OR footballer* OR golfer* OR gymnast* OR handballer* OR hurdler* OR jockey* OR kayaker* OR marathoner* OR netballer* OR orienteer* OR racewalker* OR rower* OR Rugby OR sailor* OR skater* OR skier* OR softballer* OR sportsm?n OR sportspeople OR sportsperson* OR sportswom?n OR sprinter* OR swimmer* OR volleyballer* OR weightlifter* OR wrestler* OR "badminton player*" OR "baseball player*" OR "basketball player*" OR "football player*" OR "handball player*" OR "hockey player*" OR "lacrosse player*" OR "martial artist*" OR "netball player*" OR "race walker*" OR "soccer player*" OR "softball player*" OR "squash player*" OR "tennis player*" OR "volleyball player*" OR "water polo player*" OR "weight lifter*" OR *rider* OR *runner* Load monitoring "Training load*" OR "Physical load*" OR "work load*" OR load* OR "Training practice*" OR "Global workload index" OR "NASA-TLX" OR "*RPE" OR "Perceived Exertion" OR trimp OR GPS OR "Training volume" OR "Training frequency" Physical qualities perform* OR fitness OR strength OR power OR cognitive OR aerobic OR skills OR physiolog* OR Jump OR physical N5 (Measure* OR assess* OR test* OR utility OR instrument* OR checklist* OR questionnaire* OR capacity OR perform* OR qualities) Injuries and illness injur* OR Illness OR "Upper respiratory tract infection" OR URTI NOT "systematic review" OR "Rat" CA, USA). Extracted data included participant characteristics such as age, stature, body mass, maturation level (if reported), sport, and playing level. The study results extracted were the method of monitoring the training load, and the measurement of change in physical quality, injury, or illness. Statistical interpretations of the results were only provided if reported in the original research. Contributing findings included in the best-evidence synthesis were any reported statistical relationship from included studies. Unclear or erroneous data, such as data with multiple decimal places or implausible values, were reported, but not included in the best-evidence synthesis. Assessments of physical qualities were grouped into relevant categories, being strength, power, aerobic fitness, repeated sprint ability, flexibility, muscular endurance, and change of direction. Studies included in this systematic review included a number of different study types (i.e., intervention and observational) and different statistical methods (i.e., correlation, hypothesis testing, effect sizes). As such, the heterogeneity of the results precluded meta-analysis, and data were therefore synthesised according to the following criteria [23,32]: Strong evidence: Consistent findings across two or more studies, and at least 75% of all contributing findings.
Moderate evidence: Consistent findings across two or more studies, and at least 50% of all contributing findings.
Limited evidence: Consistent findings identified in one study, and at least 50% of all contributing findings.
Inconsistent evidence: Conflicting findings across multiple studies, or less than 50% of contributing findings.
No evidence: No changes reported.

Search Findings and Study Selection
The search results are highlighted in Fig. 1. A total of 85 full-text articles were screened, with 59 studies included in the final review.

Research Reporting Quality
Methodological scores ranged from 6 to 11 with a mean of 8

Included
Reports sought for retrieval (n = 7) Reports not retrieved (n = 0) Fig. 1 Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) flow diagram of search strategy water polo (n = 1), basketball (n = 2), Irish dancing (n = 1), rugby league (n = 1), tennis (n = 7), weightlifting (n = 2), track & field (n = 4), baseball (n = 3), Australian football (n = 1), Gaelic football (n = 1), The most commonly reported internal load monitoring tools were sRPE (n = 29) and heart rate (n = 7). The most commonly reported external load monitoring tools were training duration (n = 22) and GNSS (n = 5). Physical qualities investigated included strength (n = 5), aerobic fitness (n = 19), speed (n = 12), power (n = 3), change of direction (n = 7), flexibility (n = 1), muscular endurance (n = 1) and repeated sprint ability (n = 3). Additionally, 34 studies investigated injury, and six studies investigated illness. Table 3 presents the results of the best-evidence synthesis. There was moderate evidence of a relationship between resistance training volume load and strength. Additionally, there was moderate evidence of a relationship between throw count and training duration, and injury. Evidence for all other relationships was either limited or inconsistent. Table 4 presents the relationships between external training loads and physical qualities. Nineteen studies investigated the relationship between external training loads and physical qualities [34, 39, 44, 47, 48, 51, 54-56, 64, 76, 91, 93, 94]; only one reported no significant relationships [44]. There was inconsistent or limited evidence of a relationship between GNSS metrics with change in physical qualities. Significant results were found for positive [51] and negative [94] relationships between high-speed running and changes in aerobic fitness, and a positive relationship for acceleration/deceleration and total distance with changes in sprint speed [51].

Relationship Between External Training Loads and Physical Qualities
Training duration showed inconsistent evidence of a relationship with changes in physical qualities. Results for training duration were non-significant [54], negative [76], and positive [39] with aerobic fitness; non-significant [91] and negative [76] for power; non-significant [91] and negative [54] for speed; inconsistent for change of direction [76]; and non-significant [91] and positive [48] for strength.
Resistance training metrics showed inconsistent evidence of a relationship to changes in speed, but there was moderate evidence of relationship to changes in strength. Relationships between resistance training metrics and speed were non-significant [44,91], or irregular [64]. Relationships with strength were positive between chin up 3 repetition maximum (RM) and upper body exercises, upper body volume (sets × reps × mass [kg]), and total (upper and lower body) volume [91], positive between bench press 3RM and upper body exercises and upper body volume [91], positive for snatch 1RM and total volume between medium and low volume groups [55], and positive for snatch and squat 1RM and number of lifts performed at 100% 1RM [56]. Relationships with power were observed to be non-significant for plyometrics volume measured via number of contacts [44], and positive for lower body exercises, lower body volume, and total volume [91]. Additionally, one study found upper-body resistance training volume to be related to 800-m time [61].
Of the studies that found significant results, one found that greater training load decreased the risk of injury in at least one variable [35]. The remaining 21 studies found greater training load, in at least one variable, was associated with increased injury risk [37-39, 57, 62, 66, 68, 78, 79, 81, 82, 87-89, 94]. However, when pooled, < 50% of contributing findings were significant. For GNSS and injury risk, positive relationships with high and very high accelerations [38], and both positive [38] and negative [37] relationships with total distance were reported.

External Training Loads and Illness
The only study investigating the relationship between external training load and illness found the total duration of training and matches over a week was related to increased risk of illness that caused the withdrawal of an athlete from either training or competition (OR 1.12 [95% CI 1.00-1.26]; p < 0.05) [40]. Table 6 presents the relationships between internal training loads and physical qualities. Sixteen studies investigated the relationship between internal training loads and change in physical qualities [34, 41, 46, 47, 49-51, 54, 70, 71, 73, 74, 76, 84, 94, 95]. Of these studies, six found no significant relationships [34,47,51,74,84]. Heart rate metrics had inconsistent or limited evidence of a relationship to changes in physical qualities. Positive relationships with aerobic fitness were observed for individualised training impulse (iTRIMP) [47,94], while Banister's training impulse (bTRIMP), Lucia's training impulse (LuTRIMP), and Edwards' training impulse (eTRIMP) all had both non-significant and positive relationships observed [34,47,49,51,94]. Maximal sprint speed was also found to have a positive relationship with eTRIMP [94], although the strength of the evidence was limited.
Studies investigating differential ratings of perceived exertion (dRPE) were limited, with various methods of quantifying load and inconsistent results. A positive relationship was seen between dRPE and aerobic fitness, but there were non-significant findings for speed and power [54]. Relationships between aerobic conditioning training load and physical qualities were negative for speed [46], and non-significant for power, change of direction, or aerobic fitness [46]. Tactical or skill-based training load showed both non-significant [46] and negative [70,71] relationships with aerobic fitness and negative relationships with repeated sprint ability [70]. A positive relationship was observed between strength and conditioning load, determined by the sRPE from all off-court training including resistance and metabolic conditioning, and repeated sprint ability, but there were non-significant results for speed, change of direction, aerobic fitness, and power [70]. Resistance training load showed positive relationships with speed, change of direction, and power [46]. Table 7 presents the relationships between internal training loads and injury. Ten studies found significant relationships between internal training load and injury [33,40,45,63,69,72,75,77,83,90], whilst one found no relationship [58]. Studies used a number of different definitions of injury, including reporting of a physical complaint or medical attention [33,40,69], time-loss injuries [45,63,72,75,83,90], and time loss > 3 weeks [58]. However, when pooling all the contributing findings from included studies, only 25% of contributing findings showed a relationship between internal training loads and injury.

Internal Training Loads and Injury
The evidence of a relationship between sRPE and injury risk was limited. There were positive [40,69,75,90], nonsignificant [45,83], and variable [63,77] relationships between 1-week sRPE and injury risk. Two-week training load and injury had positive [33], and non-significant [33,63,75] results. No significant relationship was seen for 3and 4-week training load, annual high-intensity training load, or annual training load and injury risk [45,63,75]. Daily training load [90], prior day's training load [90], and Table 3 Best-evidence synthesis of relationship between monitoring tools and change in physical qualities, injury or illness ↑↑↑ Strong positive relationship, ↑↑ moderate positive relationship, ↑ limited positive relationship, ↓↓↓ strong negative relationship, ↓↓ moderate negative relationship, ↓ limited negative relationship, ? inconsistent significant relationships, − no significant relationship reported, bTRIMP Banister's training impulse, dRPE differential ratings of perceived exertion, eTRIMP Edwards' training impulse, GNSS global navigation satellite system, iTRIMP individualised training impulse, luTRIMP Lucia's training impulse, sRPE session rating of perceived exertion, TeamTRIMP team training impulse     [76] individual sessional load [63] were all found to be positively related to injury risk. Some studies investigated the change in training loads using statistical methods such as the acute to chronic work ratio (ACWR), monotony, and strain. These alternative methods of analysing internal training loads had inconsistent relationships with injury risk. Results were non-significant [33,45,63,72,75,83,90] and positive [59,60,63,72] for ACWR; and non-significant [40,63,75] and positive [40,75] for strain and monotony. Table 8 presents the relationships between internal training loads and illness. Seven studies investigated the relationship between internal training load and illness [40-42, 53, 90, 95]. Both non-significant [40-42, 53, 95] and positive [90] relationships were reported for sRPE. The only study that investigated the relationship between HR and injury risk found a positive relationship [81].

Discussion
The aim of this review was to detail the methods of reporting internal and external loads in adolescent athletes and use best-evidence synthesis to report their relationship with changes in physical qualities, injury, or illness. Common internal methods of monitoring load included sRPE, dRPE, HR, and novel scales of perceived intensity, while common external methods of monitoring load included GNSS, resistance training volume, training duration, throw count, and accelerometry. Findings showed there was moderate evidence of a relationship between resistance training volume load and strength, and between training duration and throw count and injury. However, all other relationships between training load and physical qualities, injury, or illness were limited or inconsistent. An indirect finding of this review was the common use of univariate statistical techniques to establish the load-response relationship in adolescent athletes. Whilst the findings of this review indicate limited evidence for most relationships between training load and changes in physical qualities, injury and/ or illness, this may be due to highly complex interactions, as opposed to relationships not existing. For example, a number of factors outside of training load, such as sleep, stress, and maturation, will influence these relationships, but were not quantified. Based on the findings and interpretation of this review, it is recommended that researchers and practitioners should consider (1) accounting for resistance training volume load when monitoring strength training; (2) monitoring training duration, and throws, if appropriate, for potential increases in injury risk; (3) assessing factors, such as maturation, that may influence how adolescent athletes respond to load; and (4) the appropriateness of the statistical methodology used to establish a load-response relationship.

Methods of Monitoring Training Loads
A variety of internal and external load monitoring tools were used, with the distribution between the use of internal (n = 32) and external (n = 35) methods of monitoring load close to even. The most commonly reported internal load      [43] monitoring tools were sRPE and heart rate, whilst the most commonly reported external tools were training duration and GNSS. The prevalence of these methods throughout the literature likely reflects the accessibility and relative ease with which they can be used. For example, sRPE gives an overview of the load of an entire training session and is commonly used to accumulate the load across multiple forms of training (e.g., field-based training and resistance training) [96]. Alternatively, heart rate and GNSS are becoming increasingly accessible for practitioners and help provide greater information regarding the distribution of intensity across a training session [97]. It should be acknowledged, though, that the use of heart rate and GNSS is associated with added expense due to the equipment involved, which may limit its accessibility in adolescent sport. Furthermore, it does require additional expertise to collect and analyse the data appropriately [12]. Additionally, practitioners in adolescent settings are often constrained by both time and financial resources. Therefore, the methods of monitoring training load that are used throughout the adolescent literature may be an outcome of accessibility and relative ease of use rather than their relationship with changes in physical qualities, injury, or illness. Consequently, practitioners and researchers should carefully consider what the monitoring methods that are being used will add to a training environment and also whether the budget and expertise are available to help interpret the subsequent information.

Training Loads and Physical Qualities
There was moderate evidence of a relationship between resistance training volume and strength, with three studies and 53% of the results indicating a positive relationship and no results indicating a negative relationship. Resistance training volume is a commonly used monitoring tool for strength training and represents the product of the number of repetitions performed multiplied by the external load lifted [98]. Developing strength is recommended throughout all stages of adolescent development [22], as strength can be protective against injuries [5], facilitate performance [4], and underpins the development of other physical qualities, such as power [99]. Despite its importance, limited research (n = 4) has reported the relationship between training loads and strength. Additionally, all the studies were observational, limiting the ability for causal inference to be drawn.
One of the studies found that a medium-volume group had greater improvements in their snatch 1RM as compared with a low-volume group, but not compared with the high-volume group [55]. These results indicate that there may be an upper limit to the load-response relationship, but this has not been explored in detail in adolescent athletes. Nonetheless, volume load appears to demonstrate the strongest evidence for a relationship with changes in strength in adolescent athletes, and therefore warrants consideration by practitioners. Increases in strength occur as a result of a combination of neural and muscular factors [100]. In pre-peak height velocity (PHV) athletes, most strength-based adaptations occur as a consequence of increased coordination [22,101]. Strength increases seen from resistance training volume may be due to greater opportunities to practice. Post-PHV alterations in sex hormones enhance capacity for muscular adaptations, such as hypertrophy, to resistance training [22,100]. Therefore, although the mechanisms are likely to differ, resistance training volume load should be a focus throughout all stages of adolescent athletic development. This may have practical implications in the programming and periodisation of resistance training in adolescent athletes. However, there is no evidence on how much resistance training volume should be prescribed, and future research should investigate the minimal effective doses.
There were no consistent relationships between training monitoring tools and aerobic fitness across 11 studies. The most commonly reported monitoring tools were sRPE (n = 8), GNSS devices (n = 5), and heart rate monitors (n = 6). Interestingly, a relationship between upper-body resistance training load and 800-m time was found in one       [46] study [93], but this relationship is likely to be spurious. Measures of gross volume or load, such as total distance and TRIMPs, may not accurately represent the work performed, as they provide no information as to the distribution of volume or intensity. Some studies provided more informative measures of training load, such as distance and time between speed thresholds, but this did not improve any relationship [47,51,86]. The lack of consistent findings may be due to factors that mediate the response to aerobic training, such as maturation [102,103], changes in body mass [104], and variety in the monitoring tools and testing methods used to assess aerobic fitness [19]. Previously, it has been shown that adolescent athletes may have altered responses to aerobic training throughout maturation [19]. However, no studies investigating the relationship between training loads and aerobic fitness reported the maturation level of the participants.   Additionally, numerous training methods can enhance aerobic capacity, such as cross-training modalities (e.g., cycle or rowing ergometers), which may influence the effectiveness of some monitoring tools in accurately assessing overall training load (e.g., GNSS devices). Therefore, practitioners should consider external factors (e.g., maturation and body mass) that may influence aerobic capacity, and all forms of training that are being completed by the athlete. The 'Goldilocks' effect of the load-response relationship was evident in this review, with several studies finding that greater training loads were related to the decreased expression of physical qualities [46,70,71,76,94]. Given that the athletes in all studies were training throughout the period of investigation, it is unlikely de-training occurred. An alternative explanation for the decreased expression of physical qualities may be that excessive training loads and inadequate recovery caused substantial fatigue within the tested athletes [105], with studies reporting daily training loads as high as 1400 AU, equivalent to > 4.5 h of 'hard' training (i.e., > 8 RPE on a CR10 scale) [70,96]. Interestingly, two studies that found a negative relationship between sRPE and physical qualities were conducted with tennis players on international tours [70,71]. Travel can influence performance and recovery through factors such as compromised sleep and nutrition [106,107]. Therefore, although speculative, altered ability to recover may have played a mediating role in the results observed. Practitioners should also be cautious in interpreting a negative relationship between training load and physical qualities as advocating for a decrease in load, as this may hamper long-term athletic development. To state that more training results in decreased expression of physical capacity without offering solutions for reducing this risk, outside of simply reducing load, is unproductive. Instead, an increased focus should be placed on increasing or maintaining training loads whilst protecting athletes from injuries and fatigue by manipulating or accounting for factors that may mediate the load-response relationship.

Training Loads and Injury
There was moderate evidence of a relationship between training duration and throw count, and injury. However, there are limited applications of this finding as the relationship is likely due to increased exposure to risk. There were no other clear relationships between either internal or external monitoring tools and injury. Different metrics to assess distribution of training load were used, including the ACWR [33,35,38,45,63,72,83,90], monotony [63,75], and strain [63,75]. Analysis of included studies was also affected by inconsistent definitions of injury. For example, methods of reporting injury included reporting of a physical complaint or medical attention [33,40,69], time-loss injuries or illness [45,63,72,75,83,90], and time loss > 3 weeks [58]. Therefore, the inconsistent collection and analysis of methods used across different studies may unintentionally impede practitioners and researchers from drawing consensus across investigations into training load and injury.
The ACWR was used across seven of the 12 studies that investigated the relationship between internal load measures, such as sRPE, and injury risk [33,38,45,63,72,83,90]. The ACWR is a monitoring method that quantifies the acute changes in training load (e.g., most recent 7 days) relative to chronic training load (e.g., most recent 28 days) [108]. However, there are inconsistent approaches to calculating the ACWR, including variable time frames and different statistical approaches, such as exponentially weighted moving averages or rolling averages [109], and coupled or uncoupled chronic workloads [110]. The different statistical methods used to calculate ACWR can substantially alter the outcome, with one study demonstrating that quadratic calculation of the relationship between ACWR and injury was   [90] statistically significant, whereas linear was not [63]. Additionally, methodological pitfalls associated with the ACWR have been highlighted in studies that show that actual training loads confer no greater predictive value for injury risk than random chronic training loads [111]. Therefore, there is limited evidence for the use of ACWR as a metric to guide decisions around injury risk in an adolescent load monitoring programme. The monitoring tool with the strongest relationship between training load and injury was training duration, with 15 of 17 studies investigating this and 56% of contributing findings indicating a positive relationship. However, the use of various methods of reporting training duration makes it difficult to draw conclusions. For example, some studies examined training duration in the previous week [57], fortnight [39], over a season [66], weekly change in training duration [68], or duration relative to age [79,82]. Whilst there were inconsistencies in the reporting mechanism, there remains moderate evidence that increased training duration in preceding periods increases injury risk. Superficially, this finding may have practical applications as training duration is simple to collect and easy to analyse [12]. However, this relationship is likely due to athletes having greater risk of injury simply due to increased exposure. It should be noted that despite the potential for a greater number of injuries, training is necessary to develop physical qualities, tactical knowledge, and technical skills. Finding a balance between training exposure and athletic development is needed. Whilst this may be the focus of future research, it may be difficult to generalise research-based results to specific populations, due to the multi-factorial nature of injury.
Overall, there was limited evidence of a relationship between training loads and injury risk in adolescent athletes. Furthermore, training load, when administered appropriately, may also be protective against injury, highlighting the 'Goldilocks' effect [11]. Therefore, practitioners should exercise caution when using singular training loads to assess injury risk in adolescent athletes in isolation from mediating factors. Other factors that should be considered when assessing injury risk may include sleep, stress, nutrition, biomechanics, and injury history [112]. However, this list is non-exhaustive, and the highly complex nature of injuries means that identifying and accounting for all risk factors in an applied setting is difficult.

Training Loads and Illness
The evidence of a relationship between training loads and illness was limited or inconsistent with only six studies investigating these outcomes and only 4.6% of contributing findings indicating a relationship between training load and illness. The body interprets exercise as a stressor, similar to other psychological and physiological stressors [113].
Short-term periods of stress are thought to be immunoprotective, whereas prolonged exposure to stress is immunosuppressive [113]. Interestingly, the two studies that found a significant relationship between training load and illness had the longest observational period of any included studies (20 weeks and two seasons) [40,90]. Given the delayed relationship between prolonged periods of high stress and illness, studies of insufficient length may have confounded the results of the best-evidence synthesis. However, it is not known what amount of exposure to excessive stress increases the risk of illness. Additionally, given the general nature of stress, other stressors that adolescent athletes face, such as academic, social, and performance pressure, will likely contribute to this relationship, and should be accounted for [114].

Limitations and Future Directions
The results of this review provide important considerations for researchers and practitioners investigating and monitoring the training loads of adolescent athletes. However, there are limitations within this review that should be considered before implementing the findings. A limitation of the bestevidence synthesis methodology was the use of 'vote-counting' criteria, with no weightings applied to the magnitude of the stimulus or strength of the relationship [32]. Vote counting was used due to the lack of a validated method of quantifying stimulus magnitude and strength of relationships across different load monitoring tools and heterogeneous statistical methodologies. While standardisation of reporting training load metrics may assist in facilitating future meta-analyses, it is unlikely that a consistent framework will be universally adopted, due to barriers such as variation in the appropriateness of different metrics between sports, advances in technology, practitioner preferences, and the ever-increasing number of methods used to quantify training load. Additionally, a key consideration for training adolescent athletes is the effect of maturation on the response to training [19]. However, only four studies reported the maturation levels of their participants, limiting the ability to draw conclusions on the response to training load at different stages of adolescence. Previously, it has been shown that using chronological age as a surrogate for maturation is flawed as adolescents mature at different rates [115]. Given that maturity status can be assessed with relative ease (e.g., peak height velocity [116]), researchers may wish to consider reporting these data when investigating adolescent populations. This information would help inform future research on the role of maturation in the load-response relationship.
The lack of consistent findings in this review may be due to the multi-factorial nature of the load-response relationship. The individual response to training load is both positively and negatively influenced by factors such as physical qualities [6,117,118], stress [119], sleep [120], nutrition [121], and academic stress [122]. For example, one study found that self-esteem, sleep, and nutrition altered the injury rates in adolescent athletes in a multi-sport cohort [89]. It has also been demonstrated that increased stress levels correlate to a reduced adaptation to aerobic training [119]. The heterogeneity of the included studies and the complex nature of any latent relationship may have caused further noise in attempting to establish relationships with training load. The ability to adequately recover from a training dose is inextricably linked to non-training-related factors. Therefore, the 'Goldilocks' effect should not be viewed as solely being related to load. However, it is not feasible to accurately measure all of the factors that may influence the response to training load. Instead, practitioners may be best served to understand that rapid increases in stress, or prolonged periods of excessive stress, are likely to have negative outcomes and proactively modify loads accordingly.
To address the complex nature of the load-response relationship, it has recently been proposed that advanced statistical methods may be appropriate [123]. Most studies included in this review used logistic and linear regression methodologies, which are bound by fairly stringent assumptions (e.g., normality of residuals, homogeneity of variance) and are susceptible to issues such as multicollinearity [124]. These limitations may be accounted for by using alternative statistical techniques such as dimension reduction or feature selection algorithms. Compared with univariate correlation analysis, statistical methodologies that use dimension reduction (such as principle component analysis) or feature selection algorithms (such as elastic net regressions) may be more appropriate to establish a load-response relationship. By accounting for multi-collinearity, these techniques may be less likely to report spurious correlations. These techniques have previously been used to establish the relationship between training load and changes in aerobic fitness in adult athletes [124], as well as for talent identification [125]. Consequently, it is recommended that researchers consider the appropriateness of the statistical technique used when attempting to establish a dose-response relationship.

Conclusion
This systematic review is the first to investigate and detail the relationships between internal and external methods of monitoring training load and their relationship with changes in physical qualities, injury, or illness in adolescent athletes. The most commonly reported monitoring tools were sRPE and training duration. There was moderate evidence of a relationship between resistance training volume load and strength, and between throw count, training duration, and injury. However, all other relationships were either limited or inconsistent. The lack of consistent or strong relationships with load monitoring tools is likely due to the complex, individualised response to training load. Furthermore, whilst there was a general trend that greater training duration increased injury risk, inconsistencies in the reporting of training duration, and injury definitions, makes drawing conclusions difficult, and there is limited practical application of this finding. This systematic review's lack of clear trends is potentially due to the univariate nature of the data provided, which fails to account for the complex nature of any relationship between load and training outcomes where numerous mediating factors likely influence the load-response relationship. Therefore, researchers may wish to assess the interactions between multiple training loads through advanced statistical methods and their outcomes and consider mediating factors, such as maturation, that may influence this relationship.
Based on the current evidence, resistance training volume appears to be the best load monitoring tool for improving strength in adolescent athletes. Collecting resistance training volume is highly practical, requires relatively few resources to collect, and is simple to analyse. Throw count and training duration may also be valuable to assess injury risk in sports where they are applicable. Whilst the development of strength should be a key focus of adolescent development [22], this measure is only relevant to resistance training and likely only captures a small portion of the adolescent monitoring puzzle. As such, other methods are needed to quantify training and non-training stressors that are likely to influence training outcomes.

Declarations
Funding Open Access funding enabled and organized by CAUL and its Member Institutions. Availability of data and material All data and material reported in this systematic review are from peer-reviewed publications.

Conflicts of interest/Competing interests
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.