The Effects of Resistance Training on Architecture and Volume of the Upper Extremity Muscles: A Systematic Review of Randomised Controlled Trials and Meta-Analyses

: To systematically review the effects of exercise on fascicle geometry and muscle size parameters of the upper extremity muscles, the CENTRAL, CINAHL, PubMed and OpenGrey databases were searched on 31 July 2021. Finally, 17 randomised controlled trials (RCTs) were included in this systematic review. High-intensity bench press training ( g = 1.03) and 12 RM bench press exercises ( g = 1.21) showed a large effect size on increasing pectoralis major muscle size. In the elbow extensors, large effects were reported for an increase in muscle size with isometric maximal voluntary co-contraction training ( g = 1.97), lying triceps extension exercise ( g = 1.25), and nonlinear periodised resistance training ( g = 2.07). In addition, further large effects were achieved in the elbow ﬂexors via traditional elbow ﬂexion exercises ( g = 0.93), concentric low-load forearm ﬂexion-extension training ( g = 0.94, g = 1), isometric maximal voluntary co-contraction training ( g = 1.01), concentric low-load forearm ﬂexion-extension training with blood ﬂow restriction ( g = 1.02, g = 1.07), and nonlinear periodised resistance training ( g = 1.13, g = 1.34). Regarding the forearm muscles, isometric ulnar deviation training showed a large effect ( g = 2.22) on increasing the ﬂexor carpi ulnaris and radialis muscle size. Results show that these training modalities are suitable for gaining hypertrophy in the relevant muscles with at least four weeks of training duration. Future RCTs should investigate the effects of exercise modalities on the triceps brachii fascicle geometry, the infraspinatus muscle thickness (MT) and the subscapular MT due to their associations with sports performance. ‘Cross-sectional Area’, ‘Muscle Volume’, ‘Muscle Structure’, and ‘Muscle Length’ without any time, language, type


Introduction
Training-induced muscle adaptations are one of the core elements in training strategies for players, coaches, sports teams, sports federations or non-athletes. The number of studies focusing on muscle architecture has increased due to increasing access to technology for non-invasive muscle visualisation methods, e.g., magnetic resonance imaging (MRI) and ultrasound measurements. For example, investigating relationships between muscle architectural parameters and sports performance, muscle strength or sports injuries, and adaptations resulting from training, detraining, bed rest, or micro-gravity has received attention from researchers. Approximately 65% of PubMed database records containing the term "muscle architecture" have been published in the last decade (Supplementary Table S1).
Training-induced muscle architectural changes may depend on the exercise's contraction type. Eccentric (lengthening) and concentric (shortening), and isometric training can lead to comparable hypertrophic responses in skeletal muscles [46,47]. Kawakami et al. [48] noted muscle size increments are accompanied by pennation angle increases in hypertrophied muscles. By comparison, Franchi, Reeves, and Narici [46] highlighted that the underlying myogenic and molecular responses may be different in eccentric and concentric muscle actions because the eccentric training is considered to favour increases in fascicle length, and concentric training to favour higher increments of pennation angle [46]. A recent study by Pincheira et al. [49] showed that eccentric training can increase fascicle length by increasing sarcomere lengths. Another study stated that concentric, eccentric and isometric exercises can lead to similar increases in total DNA and RNA quantities, which are representative of muscle hypertrophy; however, concentric and isometric training increases muscle insulin-like growth factor 1 mRNA levels, whereas eccentric training does not increase these levels [46]. In short, there may be different underlying myogenic and molecular mechanisms of different training-induced muscle adaptations depending on the contraction type.
In consideration of the importance of the architectural parameters of upper extremity muscles for strength, power, rate of force development and sports performance, screening training-induced adaptations in the architecture of the upper extremity muscles may be a reference point for future training and conditioning directions for both athletes and non-athletes who target the upper extremities. Therefore, this systematic review with meta-analyses aimed to screen and reveal the effects of exercise on all available upper extremity muscles' volumes and architectural parameters that include fascicle geometry and muscle size variables.

Materials and Methods
This review followed the guidance of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [50]. The PRISMA 2020 checklist includes a 27 -item checklist for all sections of a systematic review. The PRISMA 2020 checklist is shown in Supplementary File S1. Before this systematic review, a review protocol was registered on INSPLASY (INPLASY202050074) [51].

Eligibility Criteria and Study Selection Process
Firstly, the duplicate records were automatically removed via the EndNote X 9 computer [52] program by the first author. Then, the remaining citations were imported to the Rayyan web application [53], which was designed for screening eligible studies for systematic reviews. The first and second authors independently screened the citations for eligibility, and they were blinded to decisions until the end of the screening process. Any conflicts that arose about the inclusion of the studies was firstly solved by discussion between the first and second authors. The third and fourth authors were considered referees if there were unresolved discussions. This conflict -solving mechanism was also applied during the risk of bias assessment and data extraction processes. Bangor University libraries retrieved non-available full-texts.
The following inclusion criteria were considered (1) being a randomised controlled trial, (2) being a full-text journal article in the English language, (3) exercise interventions lasting at least four weeks in healthy adults between 18 and 50 years old, (4) solely investigating exercise interventions, (5) using a non-invasive imaging technique (i.e., magnetic resonance imaging (MRI), ultrasonography) to assess muscle architectural parameters of a defined muscle or muscle groups of the upper extremities; and (6) presenting outcomes related to at least one muscle architectural parameter.

Outcome Measures
Changes in architectural parameters involving cross-sectional areas, fascicle length, muscle thickness, muscle volume and pennation angle of upper extremity muscles due to an exercise intervention were the outcome measures of this systematic review and meta-analysis.

Risk of Bias Assessments of Eligible Studies
For assessing the risk of bias of the included studies, the Cochrane Collaboration's tool for assessing the risk of bias in randomised trials [54] was employed. The first and second authors independently assessed the risk of bias for each eligible study regarding random sequence generation (selection bias), allocation concealment (selection bias), blinding participants and personnel (performance bias), blinding outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias) and other biases. Risk of bias categories were marked as "high risk of bias", "unclear risk of bias", and "low risk of bias".

Data Extraction
Information for groups, age, gender, number and physical activity levels of participants, type of exercises allocated for groups, the materials used during the exercise, exercising procedure, duration, number of sessions, sets and repetitions, targeted muscle or muscle groups, measurement device and region, type of muscle architectural parameters, pre-test and post-test values, statistical analyses and results were independently extracted from eligible studies by the first and second authors.

Meta-Analyses
Meta-analyses of this review were conducted via the Review Manager computer program (RevMan 5.4.1) [55]. A non-training control of the placebo group was considered the comparator for this systematic review. RevMan automatically calculates Hedges' (adjusted) g effects size (standardised mean difference (SMD)) using the mean difference Appl. Sci. 2022, 12, 1593 4 of 21 (MD) from baseline and the standard deviation (SD) of these mean differences for exercise and control groups [56]. The difference between the Hedges' g and Cohen's d is the adjustments of Hedges' g effect size calculations for a small sample having fewer than 20 participants [57]. Effect size interpretation was considered the commonly used interpretation for both Hedge's g [58] and Cohen's d [59] that small (0.2), medium (0.5) or large (0.8) [60].
The standard deviations of the mean changes from baseline is defined as a common missing outcome data [61] and difficulties for running a meta-analysis without missing SDs explained by previous systematic reviews [62,63]. For calculating missing SDs, a formula was defined as [64,65]: SDchange means the SD of the mean changes from baseline, SDbaseline represents the SD of the pre-test, SDfinal corresponds to the SD of the post-test, and the r symbolise the correlations between the baseline and final measurements; this correlation value is not generally presented in the studies. For instance, among the studies eligible for this systematic review, none demonstrate this r-value. Based on this, this systematic review employed the following process for obtaining the missing outcome data: Firstly, given additional data, e.g., confidence intervals, p-values, t-values, F-values, and standard errors were controlled and missing SDchanges from baseline were estimated using RevMan [55]. However, the first step could not be applied due to the lack of information in the included studies. As a second step, corresponding authors of the included studies were contacted to request their data -sets or the mean and SDchanges from baseline values, as previously recommended [61][62][63]. Thirdly, if the corresponding authors did not share their data with this systematic review, and the SDbaseline and SDfinal values were known, the SDchange value was calculated by assigning a value of 0.7 to the r in the formula [64,65], to provide a conservative estimate [66] as undertaken by previous systematic reviews [67][68][69][70]. Finally, if there were still missing outcome data, the study was not included in the meta-analysis and is mentioned separately in the -text.
The heterogeneity of a meta-analysis was measured by the chi-squared (χ 2 or Chi 2 ) statistic and the level of heterogeneity was estimated by the I 2 statistic, which indicates the percentage ratio of the variability in effect estimates caused by heterogeneity rather than chance [71]. I 2 results were interpreted as low (25%), moderate (50%) and high (75%) [72]. When statistical heterogeneity was absent (p > 0.05 in the Chi 2 statistics), a meta-analysis was performed for a continuous data, inverse variance, fixed-effect model [73] and a 95% confidence interval (95% CI). However, when statistical heterogeneity was observed, a metaanalysis was performed using a more conservative random effect model for continuous data, inverse variance and a 95% CI [73].

Level of Evidence of the Meta-Analyses
Each meta-analysis result in RevMan was exported to GRADEpro GDT software [74], and the level of evidence (LoE) of meta-analyses was graded by applying the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) method as described in the GRADE handbook [75], and recommended by the Cochrane Collaboration's tool for assessing the risk of bias in randomised trials [54], and the Cochrane Handbook for Systematic Reviews of Interventions [61]. The GRADE approach categorises the LoE of each meta-analysis as high, moderate, low and very low [75]. The GRADEpro GDT software measures LoE of meta-analyses based on the study design, risk of bias, inconsistency, indirectness, imprecision and publication bias features.

Risk of Bias Assessments
The low risk of bias scores range from two [92] to five [77] of seven sections of Cochrane Collaboration's tool for assessing the risk of bias in randomised trials [ among the included RCTs [76][77][78][79][80][81][82][83][84][85][86][87][88][89][90][91][92]. A risk of bias summary figure, which the shows revi authors' judgements about each risk of bias item for each included study (Figure 2), a a risk of bias graph, which shows the review authors' conclusions about each risk of b item presented as percentages across all included studies ( Figure 3) were created RevMan for further use in determining the level of evidence of meta-analyses us GRADEpro GDT software [74].

Risk of Bias Assessments
The low risk of bias scores range from two [92] to five [77] of seven sections of the Cochrane Collaboration's tool for assessing the risk of bias in randomised trials [54] among the included RCTs [76][77][78][79][80][81][82][83][84][85][86][87][88][89][90][91][92]. A risk of bias summary figure, which the shows review authors' judgements about each risk of bias item for each included study (Figure 2), and a risk of bias graph, which shows the review authors' conclusions about each risk of bias item presented as percentages across all included studies ( Figure 3) were created via RevMan for further use in determining the level of evidence of meta-analyses using GRADEpro GDT software [74].

Participants' and Intervention Characteristics of the Included Studies
Participants' characteristics in the eligible studies, including age, gender, sample size and physical activity level of participants, are presented in Supplementary Table S1. Additionally, intervention characteristics of the included studies that involve the type of exercises allocated for groups, exercise material, exercising procedure, total weeks, sessions, sets and repetitions, targeted muscle or muscle groups, measurement device and region, type of muscle architectural parameters, pre-test and post-test values, statistical analyses and results are shown in Supplementary Table S2.

Participants' and Intervention Characteristics of the Included Studies
Participants' characteristics in the eligible studies, including age, gender, sample size and physical activity level of participants, are presented in Supplementary Table S1. Additionally, intervention characteristics of the included studies that involve the type of exercises allocated for groups, exercise material, exercising procedure, total weeks, sessions, sets and repetitions, targeted muscle or muscle groups, measurement device and region, type of muscle architectural parameters, pre-test and post-test values, statistical analyses and results are shown in Supplementary Table S2.

The Posterior Arm
Six weeks of lying triceps extension exercise [76], which was performed via a dumbbell adjusted at 80% of 1 repetition maximum (RM), showed a large effect size (g = 1.

The Posterior Arm
Six weeks of lying triceps extension exercise [76], which was performed via a dumbbell adjusted at 80% of 1 repetition maximum (RM), showed a large effect size (g = 1.

The Posterior Arm
Six weeks of lying triceps extension exercise [76], which was performed via a dumbbell adjusted at 80% of 1 repetition maximum (RM), showed a large effect size (g = 1.

The Posterior Arm
Six weeks of lying triceps extension exercise [76], which was performed via a dumbbell adjusted at 80% of 1 repetition maximum (RM), showed a large effect size (g = 1.

Level of Evidence of the Meta-Analyses
The LoE of each meta-analysis was determined using the GRADEpro GDT software [74] according to the GRADE handbook [75] as described in the methodology section of this systematic review. The LoE values of meta-analyses ranged from very low to high. The LoE of each meta-analysis is presented in Supplementary File S3.

Discussion
To the authors' knowledge, this systematic review with meta-analyses is the first screening of RCTs for the effects of all types of exercises on the architecture of upper extremity muscles. This systematic review with meta-analyses aimed to overview the effects of exercise interventions on improving the architecture of upper extremity muscles. The meta-analyses of this systematic review revealed that most exercise interventions with at least 4 -weeks of exercise duration showed large effect sizes for increasing the size of individual upper extremity muscle or muscle groups (Figures 25-28). In summary, the following exercises showed large effects on increasing the size of the targeted muscles: highintensity concentrically-biased bench press training for the pectoralis major; lying concentrically-biased triceps extension, isometric maximal voluntary co-contraction training, and nonlinear periodised resistance training for the triceps brachii; traditional concentric elbow flexion exercise, low-load concentric forearm flexion-extension training without and with blood-flow restriction, isometric maximal voluntary co-contraction training, and nonlinear periodised resistance training for the biceps brachii; and isometric ulnar deviation training for the flexor carpi ulnaris and radialis.

Level of Evidence of the Meta-Analyses
The LoE of each meta-analysis was determined using the GRADEpro GDT software [74] according to the GRADE handbook [75] as described in the methodology section of this systematic review. The LoE values of meta-analyses ranged from very low to high. The LoE of each meta-analysis is presented in Supplementary File S3.

Discussion
To the authors' knowledge, this systematic review with meta-analyses is the first screening of RCTs for the effects of all types of exercises on the architecture of upper extremity muscles. This systematic review with meta-analyses aimed to overview the effects of exercise interventions on improving the architecture of upper extremity muscles. The meta-analyses of this systematic review revealed that most exercise interventions with at least 4-weeks of exercise duration showed large effect sizes for increasing the size of individual upper extremity muscle or muscle groups (Figures 25-28). In summary, the following exercises showed large effects on increasing the size of the targeted muscles: high-intensity concentrically-biased bench press training for the pectoralis major; lying concentrically-biased triceps extension, isometric maximal voluntary co-contraction training, and nonlinear periodised resistance training for the triceps brachii; traditional concentric elbow flexion exercise, low-load concentric forearm flexion-extension training without and with blood-flow restriction, isometric maximal voluntary co-contraction training, and nonlinear periodised resistance training for the biceps brachii; and isometric ulnar deviation training for the flexor carpi ulnaris and radialis.       In addition to the training modalities included in the meta-analyses, six RCTs [75,76,[80][81][82]87] were not included in the meta-analyses due to missing outcome data. The findings and intervention characteristics of the RCTs are presented in Supplementary Table S3. Among these RCTs, Matta et al. [85] investigated the effects of a nonlinear periodised strength training program on biceps brachii and triceps brachii MT and the triceps brachii long head PA and reported significant alterations in the outcome measures depending on the arm sites. The study of Matta et al. [85] was the only RCT that measured the PA of a muscle, which is a fascicle geometry component, among the eligible RCTs. The triceps brachii long head PA was significantly correlated with the strength parameters of the elbow extensors [45]. By comparison, the triceps brachii long head FL was one of the best predictors of a better swimming performance [36] and significantly correlated with lifting performance parameters [17]. However, there was no RCT that investigated the effects of an exercise intervention on the FL of triceps brachii long head. Although it did not meet the inclusion criteria of this systematic review, a recent uncontrolled trial [93]   In addition to the training modalities included in the meta-analyses, six RCTs [75,76,[80][81][82]87] were not included in the meta-analyses due to missing outcome data. The findings and intervention characteristics of the RCTs are presented in Supplementary Table S3. Among these RCTs, Matta et al. [85] investigated the effects of a nonlinear periodised strength training program on biceps brachii and triceps brachii MT and the triceps brachii long head PA and reported significant alterations in the outcome measures depending on the arm sites. The study of Matta et al. [85] was the only RCT that measured the PA of a muscle, which is a fascicle geometry component, among the eligible RCTs. The triceps brachii long head PA was significantly correlated with the strength parameters of the elbow extensors [45]. By comparison, the triceps brachii long head FL was one of the best predictors of a better swimming performance [36] and significantly correlated with lifting performance parameters [17]. However, there was no RCT that investigated the effects of an exercise intervention on the FL of triceps brachii long head. Although it did not meet the inclusion criteria of this systematic review, a recent uncontrolled trial [93]  In addition to the training modalities included in the meta-analyses, six RCTs [75,76,[80][81][82]87] were not included in the meta-analyses due to missing outcome data. The findings and intervention characteristics of the RCTs are presented in Supplementary Table S3. Among these RCTs, Matta et al. [85] investigated the effects of a nonlinear periodised strength training program on biceps brachii and triceps brachii MT and the triceps brachii long head PA and reported significant alterations in the outcome measures depending on the arm sites. The study of Matta et al. [85] was the only RCT that measured the PA of a muscle, which is a fascicle geometry component, among the eligible RCTs. The triceps brachii long head PA was significantly correlated with the strength parameters of the elbow extensors [45]. By comparison, the triceps brachii long head FL was one of the best predictors of a better swimming performance [36] and significantly correlated with lifting performance parameters [17]. However, there was no RCT that investigated the effects of an exercise intervention on the FL of triceps brachii long head. Although it did not meet the inclusion criteria of this systematic review, a recent uncontrolled trial [93] compared the effects of concentrically-biased cable push-down and cable overhead extension exercises, and Stasinaki and colleagues [93] did not report significant alterations in the FL of the triceps brachii long head even when the concentric elbow extension starts from a fascicle lengthened position. This may be due to the effects of concentric training. A future RCT should examine the impacts of eccentric training on the FL of triceps brachii long head.
In terms of the muscle size parameters, the triceps brachii MT has been found to be strongly correlated with elbow extension strength [41]. Additionally, the triceps brachii MT was stated as being significantly correlated with better swimming performance (r = −0.56) [36]. Moreover, elbow extensors' and flexors' muscle size parameters (ACSA, PCSA and MV) showed significant strong correlations with elbow joint torque (r = 0.705-0.945) [39]. Furthermore, the elbow extensors' cross-sectional muscle area (CSA) was correlated with rowing performance, and was the significant best predictor of arm pull during the rowing activity in rowers (r 2 = 0.195) [37]. Elbow flexors CSA showed a strong correlation with elbow flexion maximal power (r = 0.81) [40]. Arm muscles CSA was significantly correlated with shot put performance (r = 0.68) [38]. The pectoralis major muscle CSA was strongly correlated with bench press strength (r = 0.866), and muscle volume was strongly correlated with bench throw peak power (r = 0.821) [43]. Either concentric, isometric, eccentric or blood-flow restricted resistance training modalities led to significant muscle hypertrophies. Based on these findings, athletes, healthy individuals aiming to increase their related performance or muscle strength parameters, astronauts after a space mission [94] and patients experiencing muscle atrophies after bedrest [95,96], which were mentioned above, may refer to the training regimens that showed large effects sizes on increasing the pectoralis major, arm and forearm muscles' size parameters. However, exercise selection should cautiously be made due to the small numbers of studies included in each meta-analysis.
Additionally, the infraspinatus MT was significantly correlated with shoulder external rotation strength in professional baseball pitchers (r = 0.287) [44]. The subscapular MT was the best single predictor for powerlifting performance in professional powerlifters [17]. However, this systematic review did not detect any RCTs focusing on exercise-induced alterations in these muscle architectural parameters. Future RCTs may be conducted to investigate exercise-induced alteration in these muscle architectural parameters in the relevant samples, such as exercise-induced alterations in the infraspinatus MT in baseball pitchers, in the subscapular MT in powerlifters, and in the fascicle geometry of the triceps brachii in swimmers.
Regarding the effect size calculations of the RCTs, initially, none of the RCTs reported the required SDs of the mean changes from baseline for exercise and control groups. The difficulties associated with conducting a meta-analysis without this variable are well described in the literature [61][62][63]. Therefore, this systematic review strongly suggests that future RCTs should share their raw data, or mean changes from baseline and SDs of the mean changes from baseline, with their publications for more comparable future studies and meta-analyses. Additionally, for the effect sizes reported in individual RCTs, the calculations were generally in respect of the baseline and post-test scores of the intervention groups of post-test scores of the intervention and control groups. Both approaches may lead to wrong interpretations and fewer comparisons between the RCTs. Therefore, this systematic review strongly suggests that future RCTs should calculate the effect sizes based on the mean changes from baseline and their SD in an intervention group comparing the same parameters with a control group, as calculated in this systematic review. Finally, random allocation, and blinding of participants and assessors, were the most common risks of bias among the RCTs. Thus, following the CONSORT statement [97] for parallel-group randomised trials may reduce the risk of biases caused by the methodology, and this can be recommended for future RCTs.
A limitation of this study may be the small numbers of the RCTs included in each metaanalysis. A further limitation of our review is the inclusion of only English language articles, which may have led to the omittance of some data in the analysis. Similar to previous relevant meta-analytic studies that included both genders in the meta-analyses [62,63,98], this systematic review did not address the question of the influence of sex for a differential response to training in the meta-analyses, which adds a limitation to the outcomes provided. More RCTs may have led to stronger conclusions in this systematic review. Another limitation is not being able to perform assessments of meta-regression or publication bias, which are not suitable for performing each meta-analysis due to the few RCTs [61] included in the meta-analyses of this systematic review. An additional confounding factor is the difference between training interventions, which can lead to uncountable variability in the results of the meta-analyses.

Conclusions
Regarding the pectoralis major muscle size, 6-weeks of high-intensity bench press training [91] and 10 weeks of 12 RM bench press exercises [82] can be applied for hypertrophy in this muscle. To achieve hypertrophy in elbow extensors, 6-weeks of lying triceps extension exercise [76], isometric maximal voluntary co-contraction training [83,84], and 12-weeks of nonlinear periodised resistance training [89] may be a suitable intervention. From the perspectives of elbow flexors, 6-weeks of traditional elbow flexion exercises [77], 4-weeks of concentric low-load forearm flexion-extension training [79], isometric maximal voluntary co-contraction training [83,84], 4-weeks of concentric low-load forearm flexion-extension training with vBFR [79], or 12-weeks of nonlinear periodised resistance training [88,90] can be applied to gain hypertrophies in the elbow extensors. Finally, 6-weeks of isometric ulnar deviation training can be used to increase the flexor carpi ulnaris and radialis muscle size [78].
However, these results should be cautiously interpreted due to the small numbers of the RCTs included in each meta-analysis. More RCTs are needed to provide more precise and more robust conclusions about the effects of exercise on the architecture of the upper extremity muscles. Additionally, all the eligible studies of this systematic review were restricted to muscle size measurements, and not did not expand towards the fascicle geometry such as the FL of the triceps brachii long head. Future RCTs can examine the effects of exercise on the triceps brachii FL and PA, the infraspinatus MT and the subscapular MT, due to their associations with sports performance.

Acknowledgments:
This systematic review will also be published as a chapter of Gokhan Yagiz's Ph.D. Thesis. The authors wish to thank those authors who shared the requested missing outcome data with this systematic review.
Conflicts of Interest: Gokhan Yagiz, Esedullah Akaras, Hans-Peter Kubis and Julian Andrew Owen declare that they do not have any conflicting interests in this systematic review's content.