This meta-analysis was performed using the recommendations and criteria defined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Liberati et al., 2009). Computer-aided searches were conducted using the following databases: MEDLINE (PubMed), SWETSWISE, EMBASE, and SPORTDiscus^TM. The period of search history assessed was inclusive to August 2018. An extensive manual search and cross-referencing of journals, reference lists, was also performed with citations and abstracts from studies published in foreign language journals and scientific conferences were excluded. Descriptive terms and keywords that were used to retrieve studies included: “resistance training and muscular strength,” “resistance training and training volume,” “single vs. multiple-sets,” and “one vs. multiple-sets and muscular strength.” Boolean operators, including AND, OR, and NOT, were used to focus literature searches with literature searches reduced to studies involving humans only.

As a result of systematic computerized database searches, journals were retrieved from 1960 to August 2014 in where S vs. M3 were examined, from different population demographics (trained, untrained, male, and female subjects). After preliminary literature searching, reference lists of articles were screened for additional studies of relevance on muscular strength development. During the first selection round, appropriate study titles were screened for relevance with the inclusion of either resistance training or training volume. In the second selection round, GR, LK, and DB read the abstracts and then selected the article if resistance training for muscle strength was evaluated before and after a minimum RT intervention period of 4-weeks. This minimum time course was chosen due to reports of muscular adaptations in response to RT (Stock et al., 2016). In the third selection round, full articles were read.

Studies were deemed eligible in this review if they met the following conditions; (a) human subjects free from chronic disease, muscular, or orthopedic injuries, or physical limitations; (b) trained and untrained adult male or female subjects between 18 and 45 years; (c) subject's descriptive characteristics included in the report (height, weight, training status, and training experience); (d) subjects training at least one primary muscle group-pectoralis major, deltoids (anterior, lateral, posterior); bicep brachii, or tricep brachii; latissimus dorsi; quadriceps (vastus medialis, vastus intermedius, vastus lateralis, rectus femoris); hamstrings (bicep femoris, semitendinosus, semimembranosus; (e) at least one performed pre-to-post measure of muscular strength; (f) studies that compared S vs. M3 performing resistance exercise only (active control group); (g) training protocols lasting a minimum of 4-weeks; (h) and appropriate information to calculate training ES. This meta-analysis included both randomized trials (RAN) and randomized control trials (RCTs) that observed the intervention treatments using stratified resistance exercises with S vs. M3. RAN allocation ensures no systematic variances between the intervention groups; however, no control group may influence the assessment of outcomes (Schünemann et al., 2013). RCTs are a more specific method for defining a cause-effect relationship between treatments and outcomes.

Titles and abstracts of retrieved journal articles were independently evaluated for content relevance by three reviewers (GR, LK, and DB). Abstracts that contained the necessary information regarding the pre-set inclusion and exclusion criteria were retrieved and independently evaluated for full-text eligibility. Potential studies that did not have descriptive data tables but presented pre- to post-primary strength data in the form of figures resulted in extraction using WebPlot-Digitiser (Web Plot Digitiser V.3.11. Texas, USA: Ankit Rohatgi, 2017). Where differences between reviewers (GR, LK, and DB) occurred then additional dialogue and agreements were made by consensus. Ten randomly selected studies underwent post-hoc reassessment with the extracted results compared. For each reviewer coder drift was set at <10% in all cases, and inter-rater (GR and DB) reliability was >95%. Studies were read and individually coded for the following variables; (1) subject's descriptive characteristics, including age, training experience, and sample size; (2) programme characteristics including training frequency, number of sets performed per exercise, the number of reps performed per exercise; (3) measurement of pre-post-strength outcome(s) and; (4) treatment effects of mean (M) and SD values of changes in pre- and post-strength outcomes for RT intervention and control groups.

Internal validity of retrieved studies was evaluated using the Physiotherapy Evidence Database (PEDro) scale. The PEDro scale (Verhagen, 1998; Maher et al., 2003) has 11 measures, with a maximum score of ten. However, a maximum score from the PEDro scale, in this case, was eight, as the therapists, assessors and technicians conducting the interventions cannot be blinded. Studies were included in this analysis if they had a PEDro score of ≥ four, as this was considered as having acceptable internal validity. Methodological quality was independently assessed by reviewers (GR, LK, and DB). Variances of judgement concerning the scoring of the journal articles were agreed between reviewers through consensus.

Descriptive statistics were calculated to describe and summarize the results of the systematic review process. Data of individual study characteristics were entered into a spreadsheet (Microsoft, Redmond, WA, USA) to compare pre-post-strength outcomes of each study for coding, review and data reference. Descriptive statistics containing sample size (n), mean (M) and SD were extracted from each study. This provided data for the mean differences in pre- to post-intervention between groups (e.g., S and M3) on several strength outcomes. Muscular strength was deemed a continuous data variable; therefore, the standardized mean difference (SMD) with 95% confidence intervals (95% CI) were used to establish the ES measures. For each strength outcome variable, a SD score was calculated by using Cohen's d index of a single ES (di = [M1–M2]/SDpi) (Cohen, 1998), where d = ES, i = individual study, M1 = pre-intervention mean, M2 = post-intervention mean, and SDp = pooled standard deviation. The SD was calculated by summing the extracted pre-intervention and post-intervention SDs and dividing by two. If the standard error of measurement (SEM) of the mean was specified, the SD was calculated using the formula (SD = SEM^*square root of N) (Howell, 2012). Separate ES was weighted to account for individual sample sizes. If a study reported, exact P-values for a change of strength, the SD of change was calculated. Studies that did not report, exact P-values, the SD of change was calculated using the pre- and post-intervention SDs. Due to diverse population demographics and methods with the included studies, a random-effects inverse variance (IV) using the DerSimonian-Laird method (DerSimonian and Laird, 1986) was applied with the effects measure of SMD. If a study had numerous time-periods, only the pre- to post-intervention strength outcomes were extracted and entered for analysis. The data was then used to compute ES estimates and CI. For each strength measure, an ES was calculated as the pre- to post-intervention change, divided by the pre-intervention SD (Morris and DeShon, 2002).

Meta-Essentials (Suurmond et al., 2017) was initially used to input pre-post-strength outcome data with each row denoted as an individual ES for a treatment group. If treatment groups had multiple ES, then each ES was coded in a separate row. This aided with the computation of ES, SEM, and study size to allocate appropriate weight to each study, and estimate a study effect. To determine the significance of the ES, the chi-square (Chi²) test was performed in each model used. For the statistical analyses, Review Manager (RevMan) version 5.3.5 was used to calculate the difference in SD of post-intervention strength outcomes and the generation of forest plots. Data needed were either; (1) means and SDs (pre- and post-strength change); (2) CI data for pre- to post-strength change for each treatment group (3) P-values for pre- to post-strength difference for each treatment group, or if only the level of significance was available, and; (4) default P-values (e.g., P ≤ 0.05 becomes P ≤ 0.49, P ≤ 0.01 becomes P ≤ 0.0099, and P ≥ not significant becomes P ≥ 0.05).

The random-effects model was implemented to allow for variability between the studies due to high heterogeneity. A random-effects model conceptualized the existing series of studies under investigation to be a random sample selected from a larger population of studies. In the random-effects model meta-analysis, there are two sources of variability; (1) variability of the effect parameters, and; (2) sampling variability of experimental units (i.e., subjects) into studies. If individual parameter estimates of each study lead to high levels of heterogeneity, the random-effects analysis considers the “true variance” (or the remaining unmeasured random-effects between studies) in addition to the modeled between-study variances and sampling error typically assumed in fixed-effects models. It should be highlighted that the random-effects model typically gives less specific estimates and larger CIs.

To evaluate heterogeneity between studies, the I-squared (I²) index test and Cochran Q (Q) heterogeneity statistic were applied. The I² test was used to assess the degree of heterogeneity for each outcome, with an I² > 50% applied to indicate heterogeneity. Non-significance signifies that the results of the different studies were similar (P ≥ 0.05) and P < 0.05 denotes a statistically significant effect. The Q statistic uses the sum of squared deviations of each estimate resulting from the pooled estimate and weights the contribution of each study. The Q heterogeneity test was applied to evaluate heterogeneity prior to estimating tau-square (Tau²), then I² and Tau² statistics were calculated. Comparing the Q statistic with an X² distribution with k⁻¹ degrees of freedom (where k denotes the number of included studies) allowed P-values to be attained. All analyses were conducted at the 95% confidence level. The ES of ≤ 0.2, ≤ 0.5, ≤ 0.8, and ≥0.8 were considered trivial, small, moderate and large, respectively (Cochran, 1954).

For the assessment and evaluation of publication bias, the use of funnel plot assessments with Duval and Tweedie's (2000) trim and fill correction was applied. The purpose of the “trim and fill” was to identify and correct for funnel plot asymmetry ascending from publication bias. This method is to; (1) remove the smaller studies causing funnel plot asymmetry; (2) apply the trimmed funnel plot to estimate the true “centre” of the funnel, then; (3) replace the removed or omitted studies around the center. Forest plots were produced to display the study-specific ES and the corresponding CI. All forest plots generated were visually examined against its standard error (SE) to account for publication bias also known as the “file drawer problem.” This refers to the influence of the results of a study that introduces bias into the scientific literature by selective publication, primarily by the propensity to publish positive results but not to publish negative results (Scargle, 2000).

Separate subgroup analysis on ES was performed with the resulting moderators, including; (1) single-joint or multi-joint resistance exercise on 1RM strength gains (trained only subjects); (2) single-joint or multi-joint resistance exercise on 1RM strength gains (untrained only subjects). In the subgroup analysis, mean differences in ES were computed for each study to produce a study-level ES for the difference between S and M3 allowing for the generation of forest plots. Sensitivity analysis was performed, by identifying any studies that were highly influential which may bias the analysis. This was achieved for each model by eliminating one study at a time and then inspecting the set-volume predictor. Influential studies were removed if they caused a significant change in the magnitude of the coefficient or change from significant (P ≤ 0.10) to non-significant (P ≥ 0.10) or vice versa.

The initial examination generated 8,418 related abstracts and citations. Thirty-seven full-text articles were initially deemed to meet the inclusion criteria. A total of 19 potentially relevant journal articles met the pre-set inclusion and exclusion criteria (Table 2) and were further assessed for content applicability. Six studies (Kraemer, 1997; Borst et al., 2001; McBride et al., 2003; Galvão and Taaffe, 2005; Munn et al., 2005; Rønnestad et al., 2007) were rejected prior to data extraction, with Galbraith plot identifying one further article (Starkey et al., 1996) as an outlier and was omitted. Descriptions for the seven studies that were excluded are detailed in Table 3.

Following appraisal and sensitivity measures 12 full-text articles (Reid et al., 1987; Kraemer, 1997; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Kelly et al., 2007; Bottaro et al., 2009; Baker et al., 2013; Sooneste et al., 2013; Radaelli et al., 2014) met pre-set inclusion criteria (Table 2). Journal articles included in this analysis had dates ranging from 1987 to 2014. In total, 12 studies provided data on 393 subjects (Table 4) with the both randomized control groups (RCT [n = 4]) and random assignment of treatment conditions (RAN [n = 8]) experimental designs included.

The mean age of the subjects was 25.2 (± 5.4 years). The training status of subjects included in the 12 studies was untrained (n = 6) and trained (n = 6). Assigned cohorts consisted of male (n = 8 [67%]), female only groups (n = 1 [12%]), and mixed-sex studies (n = 3 [25%]) which were included in the analysis. The RT period ranged from 6 to 26 weeks (mean = 11.2 [±5.4] weeks), weekly training frequency ranged from 2 to 3 days per week (2.7 [±0.49] per week), and the repetitions used ranged from 3 to 18 repetitions (9.0 [±1.7]) per week. The total number of sets per week ranged from two-to-three-sets (2.7 [±1.7]) for S and six-to-nine-sets for M3 (7.5 [±2.1]) per exercise. Also, training loads ranged from 63 to 90% 1RM (76.7 [±4.1]) with the subject's resistance training characteristics and weekly training volume specified in Table 5.

The PEDro scale was based on the Delphi list (Verhagen, 1998) with column 1a not used in the calculation of the scores. Only criterion 2–11 are scored giving a total out of ten. Each column number corresponds to the following criteria on the PEDro scale (Table 6): 1^a = eligibility criteria (1a = eligibility criteria specified [1 = yes/0 = no]); 2 = subjects randomly allocated; 3 = allocation was concealed; 4 = groups similar at baseline; 5 = blinded subjects; 6 = therapists blinded; 7 = assessors blinded; 8 = follow-up measures obtained for >85% of subjects; 9 = intention to treat analysis; 10 = between groups statistical comparison; 11 = point measures and measures of variability. The included studies had PEDro scores that ranged from five through to six (Table 6). Though the maximum PEDro score is 11, it is problematic and unrealistic to achieve this total. Realistically the maximum score on PEDro was eight as it is problematic to blind both participants and researchers to an exercise intervention. Consequently, included journal articles had a common area of bias as subjects, therapists or researchers were not blinded. Galbraith plots were used to investigate for study heterogeneity and identification of potential outliers. Examination of Galbraith plots exposed no outliers (Figure 2). To assess for publication bias trim and fill funnel plot were also performed in all comparison models. This was to safeguard for overestimations of the ES of set-volume and strength outcomes in the included studies. The shape of the funnel plot did not expose any evidence of apparent asymmetry.

Pre- to post-strength outcomes were assessed via a meta-analytic procedure for all included studies. Subgroup analysis was then performed with multi-joint and single-joint exercises combined into separate subdivision analysis. A random-effects model was incorporated into each strength measure due to the potential of pooled study data generating significant heterogeneity with I² used to evaluate heterogeneity.

The pooled mean ES estimates (untrained and trained) of multi-joint and single-joint data (Tables 7, 8) comprised of 70 treatment groups from 12 studies (Reid et al., 1987; Kramer et al., 1997; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Kelly et al., 2007; Bottaro et al., 2009; Baker et al., 2013; Sooneste et al., 2013; Radaelli et al., 2014). The random-effects model exposed a considerable amount of variability between studies. Heterogeneity prior to taking Tau² into consideration (Q heterogeneity test) was: Chi² = 131.56, d.f. = 11, P < 0.00001. The heterogeneity statistic I² (%) = 92 [interpreted as high, (Higgins et al., 2003)], and the tau² test (between-trials variance) = 0.74. When a random effect analysis was implemented, a large effect was detected for combined multi-joint and single-joint exercises on RT set training volume [mean effect size (ES) 0.93; 95% CI 0.41–1.45]. Pre- to post-intervention strength change was greater with M3 compared to S (ES difference 0.25) with the effect statistically significant (P = 0.0005). The mean for S was 0.64 (95% CI 0.44–0.84). The mean ES for M3 was 0.89 (95% CI 0.66–1.12).

Examination of the effects of pre- vs. post-training strength (untrained subjects only) categorized as either S or M3 from six studies (Reid et al., 1987; Paulsen et al., 2003; Humburg et al., 2007; Bottaro et al., 2009; Sooneste et al., 2013; Radaelli et al., 2014). The random-effects model exposed a considerable amount of variability between studies. Q heterogeneity test was: Chi² = 78.30, d.f. = 5, P < 0.00001. The heterogeneity statistic I² (%) = 94 [interpreted as high, (Higgins et al., 2003)], and the tau² test (between-trials variance) = 0.94. A large effect was observed on untrained only subjects for multi-joint and single-joint exercises combined (ES 1.20; 95% CI 0.39–2.01). Pre- to post-intervention strength change was greater when M3 was compared to S (ES difference 0.22) with statistical significance (P = 0.004). The mean for S was 0.60 (95% CI 0.34–0.85). The mean ES for M3 was 0.82 (95% CI 0.52–1.11).

Separate subgroup examination on the effects of pre- vs. post-training strength (trained subjects only) categorized as either S or M3 from six studies (Kramer et al., 1997; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Kelly et al., 2007; Baker et al., 2013). The random-effects model exposed a considerable amount of variability between studies. Q heterogeneity test was: Chi² = 29.21, d.f. = 5, P < 0.0001. The heterogeneity statistic I² (%) = 83 (interpreted as high; Higgins et al., 2003), and the tau² test (between-trials variance) = 0.42. A moderate effect was observed on trained only subjects for multi-joint and single-joint exercises combined (ES 0.63; 95% CI 0.05–1.22). Pre- to post-intervention strength gain was greater when M3 was compared to S (ES difference 0.28) with statistical significance (P = 0.03). The mean for S was 0.68 (95% CI 0.35–1.01). The mean ES for M3 was 0.96 (95% CI 0.61–1.31).

Examination of upper body exercises (multi-joint and single-joint exercises combined) comprised 10 studies (Reid et al., 1987; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Bottaro et al., 2009; Baker et al., 2013; Sooneste et al., 2013; Radaelli et al., 2014) are displayed in the forest plot (Figure 3). The random-effects model exposed a large amount of variability between studies (I² = 93%). Removal of Humburg et al. (2007) and Radaelli et al. (2014) data, resulted in moderate heterogeneity (Q heterogeneity test was: Chi² = 21.71, d.f. = 7, P = 0.003). The heterogeneity statistic I² (%) = 68 (interpreted as moderate heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.27, and a small effect was observed (ES 0.37; 95% CI −0.09 to 0.82). Pre- to post-intervention strength change was greater when M3 was compared with S (ES difference 0.19) with no statistical significance (P = 0.11). The mean ES for S was 0.68 (95% CI 0.42–0.94). The mean ES for M3 was 0.87 (95% CI 0.61–1.14).

Examination of upper body exercises (multi-joint and single-joint exercises combined) on untrained subjects comprised of six studies (Reid et al., 1987; Paulsen et al., 2003; Humburg et al., 2007; Bottaro et al., 2009; Sooneste et al., 2013; Radaelli et al., 2014). The random-effects model exposed a large amount of variability between studies (I² = 94%). Removal of Humburg et al. (2007) and Radaelli et al. (2014) data, resulted in large heterogeneity (Q heterogeneity test was: Chi² = 11.45, d.f. = 3, P = 0.010). The heterogeneity statistic I² (%) = 74 (interpreted as high heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.54, and a small effect was observed (ES 0.35; 95% CI −0.49 to 1.19). Pre- to post-intervention strength change was greater when M3 was compared with S (ES difference 0.23) with no statistical significance (P = 0.42). The mean ES for S was 0.81 (95% CI 0.67–0.95). The mean ES for M3 was 1.04 (95% CI 0.66–1.41). Subgroup examination of S vs. M3 pre- to post-intervention strength differences on trained only and untrained only subjects was not viable due to inadequate study data.

Examination of upper body exercises (multi-joint and single-joint exercises combined) on trained subjects comprised of four studies (Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Baker et al., 2013). Q heterogeneity test was: Chi² = 3.09, d.f. = 3, P = 0.38. The random-effects model exposed a low amount of variability between studies. The heterogeneity statistic I² (%) = 3 (interpreted as low, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00. When a random effect analysis was implemented, a moderate effect was detected on trained only subjects for upper body exercise resistance exercises (ES 0.63; 95% CI 0.34–0.92). Pre- to post-intervention strength change was greater when M3 was compared to S (ES difference 0.33) with statistical significance (P < 0.0001). The mean for S was 0.59 (95% CI 0.16–1.02). The mean ES for M3 was 0.92 (95% CI 0.53–1.31).

Examination of lower body exercises (multi-joint and single-joint exercises combined) on untrained and trained subjects comprised of 10 studies (Reid et al., 1987; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Kelly et al., 2007; Bottaro et al., 2009; Radaelli et al., 2014) are displayed in the forest plot (Figure 4). Q heterogeneity test was: Chi² = 54.00, d.f. = 9, P < 0.00001. The random-effects model exposed a high amount of variability between studies I² (%) = 83. Removal of Hass et al. (2000) and Humburg et al. (2007) data, resulted in no heterogeneity (Q heterogeneity test was: Chi² = 4.05, d.f. = 7, P = 0.77). The heterogeneity statistic I² (%) = 0 (interpreted as no heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00, and a small effect was observed (ES 0.35; 95% CI 0.10–0.60). Pre- to post-intervention strength change was greater when M3 was compared with S (ES difference 0.27) with statistical significance (P = 0.006). The mean ES for S was 0.60 (95% CI 0.36–0.84). The mean ES for M3 was 0.87 (95% CI 0.57–1.17).

Examination of lower body exercises (multi-joint and single-joint exercises combined) on untrained comprised of five studies (Reid et al., 1987; Paulsen et al., 2003; Humburg et al., 2007; Bottaro et al., 2009; Radaelli et al., 2014). Q heterogeneity test was: Chi² = 34.99, d.f. = 4, P = < 0.00001). The random-effects model exposed a high amount of variability between studies I² (%) = 89. Removal of Humburg et al. (2007) data, resulted in no heterogeneity (Q heterogeneity test was: Chi² = 2.36, d.f. = 3, P = 0.50). The heterogeneity statistic I² (%) = 0 (interpreted as no heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00, and a moderate effect was observed (ES 0.49; 95% CI 0.14–0.83). Pre- to post-intervention strength change was greater when M3 was compared with S (ES difference 0.28) with statistical significance (P = 0.005). The mean ES for S was 0.75 (95% CI 0.30–1.2). The mean ES for M3 was 1.03 (95% CI 0.51–1.55).

Examination of lower body exercises (multi-joint and single-joint exercises combined) on trained subjects comprised of five studies (Kramer et al., 1997; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Kelly et al., 2007). Q heterogeneity test was: Chi² = 15.24, d.f. = 4, P = < 0.004. The random-effects model exposed a high amount of variability between studies I² (%) = 74. Removal of Hass et al. (2000) data, resulted in no heterogeneity (Q heterogeneity test was: Chi² = 0.34, d.f. = 3, P-value = 0.95). The heterogeneity statistic I² (%) = 0 (interpreted as no heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00, and a trivial effect was observed (ES 0.18; 95% CI −0.23 to 0.58). Pre- to post-intervention strength change was greater when M3 was compared with S (ES difference 0.32) with no statistical significance (P = 0.39). The mean ES for S was 0.63 (95% CI 0.18–1.08). The mean ES for M3 was 0.95 (95% CI 0.43–1.47).

Outcomes for S vs. M3 on multi-joint exercise (combined trained and untrained subjects) classified as S or M3 are displayed in the forest plot (Figure 5). The forest plot includes the mean ES and CIs for strength change separated for interventions featuring S and M3 and the overall effect test and heterogeneity analysis. The pooled mean ES estimates of S vs. M3 on multi-joint exercise data comprised of 34 treatment groups from eight studies (Kramer et al., 1997; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Baker et al., 2013; Radaelli et al., 2014). The random-effects model exposed a considerable amount of variability between studies. Q heterogeneity test was: Chi² = 69.07, d.f. = 7, P c< 0.00001. The heterogeneity statistic I² (%) = 90 (interpreted as high, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.86. A moderate effect was observed for multi-joint exercise and M3 (ES 0.83; 95% CI 0.14–1.51). Pre- to post-intervention strength change was greater when M3 was compared to S (ES difference 0.22) with statistical significance (P = 0.02). The mean for S was 0.61 (95% CI 0.34–0.88). The mean ES for M3 was 0.83 (95% CI 0.53–1.13).

Separate subgroup examination on the effects of pre- vs. post-training strength (trained subjects only) categorized as either S or M3 from five studies (Kramer et al., 1997; Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Baker et al., 2013). The random-effects model exposed a low degree of variability between studies. Q heterogeneity test was: Chi² = 7.73, d.f. = 4, P = 0.10. The heterogeneity statistic I² (%) = 48 (interpreted as low-moderate, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.07. A moderate effect was observed on trained only subjects for multi-joint exercises combined (ES 0.52; 95% CI 0.10–0.94). Pre- to post-intervention strength change was greater when M3 was compared to S (ES difference 0.25) with statistical significance (P = 0.02). The mean for S was 0.67 (95% CI 0.30 to 1.04). The mean ES for M3 was 0.92 (95% CI 0.56–1.28). Subgroup analysis of S vs. M3 pre-to post-intervention strength differences on untrained subjects was not feasible due to inadequate study data.

Subgroup analysis on the effects of pre- vs. post-training strength multi-joint upper body exercises (trained and untrained subjects) categorized as either S or M3 from seven studies (Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Baker et al., 2013; Radaelli et al., 2014). The random-effects model exposed a significant amount of variability between studies. Q heterogeneity test was: Chi² = 70.88, d.f. = 6, P < 0.00001. The heterogeneity statistic I² (%) = 92 [interpreted as high, (Higgins et al., 2003)] and the tau² test (between-trials variance) = 1.57. A large effect was observed (ES 1.15; 95% CI 0.17–2.12). Pre- to post-intervention strength change was greater when M3 was compared to S (ES difference 0.24) with statistical significance between RT set volumes (P = 0.02). The mean for S was 0.55 (95% CI 0.23–0.86). The mean ES for M3 was 0.79 (95% CI 0.44–1.14).

Subgroup examination of S vs. M3 pre- vs. post-training strength differences on trained subjects multi-joint upper body exercises comprised of four studies (Hass et al., 2000; Schlumberger et al., 2001; Rhea et al., 2002a; Baker et al., 2013). The random-effects model exposed a low to moderate amount of variability between studies. Q heterogeneity test was: Chi² = 4.56, d.f. = 3, P = 0.21. The heterogeneity statistic I² (%) = 34 (interpreted as low-moderate, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.07. A moderate effect was observed (ES 0.52; 95% CI 0.08–0.96). Pre- to post-intervention strength gain was greater when M3 was compared to S (ES difference 0.32) with statistical significance (P = 0.02). The mean for S was 0.59 (95% CI 0.09–1.08). The mean ES for M3 was 0.91 (95% CI 0.49–1.32). Subgroup examination of S vs. M3 pre- to post-intervention strength differences on untrained subjects was not feasible due to limited study data.

Analysis of S vs. M3 pre- to post-intervention strength differences on subject's multi-joint lower body exercises comprised of five studies (Kramer et al., 1997; Rhea et al., 2002a; Paulsen et al., 2003; Humburg et al., 2007; Radaelli et al., 2014). The random-effects model showed a significant amount of variability between studies. Removal of (Humburg et al., 2007) data resulted in no heterogeneity (Q heterogeneity test was: Chi² = 0.13, d.f. = 3, P = 0.99). The heterogeneity statistic I² (%) = 0 (interpreted as none, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00 and trivial effect observed (ES 0.09; 95% CI −0.32 to 0.51). Pre- to post-intervention strength change was greater with M3 compared with S (ES difference 0.17) with no statistical significance (P = 0.66). The mean ES for S was 0.81 (95% CI 0.41–1.21). The mean ES for M3 was 0.98 (95% CI 0.45–1.51). Subgroup examination of S vs. M3 pre- to post-intervention strength differences on trained only and untrained only subjects was not feasible due to inadequate study data.

Outcomes for 1-vs.-3 sets categorized as S or M3 for single-joint resistance exercises are displayed in the forest plot (Figure 6). The pooled mean ES estimates of single-joint resistance exercises comprised of 36 treatment groups from nine studies (Reid et al., 1987; Hass et al., 2000; Schlumberger et al., 2001; Paulsen et al., 2003; Humburg et al., 2007; Kelly et al., 2007; Bottaro et al., 2009; Baker et al., 2013; Sooneste et al., 2013). The random-effects model exposed a low to moderate amount of variability between studies. Q heterogeneity test was: Chi² = 10.61, d.f. = 8, P = 0.22. The heterogeneity statistic I² (%) = 25 (interpreted as low-moderate, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.03 and a small effect was observed (ES 0.49; 95% CI 0.26–0.72). Pre- to post-intervention strength change was marginally greater with M3 compared with S (ES difference 0.19) with statistical significance (P < 0.0001). The mean ES for S was 0.57 (95% CI 0.26–0.87). The mean ES for M3 was 0.76 (95% CI 0.54–0.98).

Subgroup examination on the effects of pre- vs. post-training strength (trained subjects only) categorized as either S or M3 from four studies (Hass et al., 2000; Schlumberger et al., 2001; Kelly et al., 2007; Baker et al., 2013). Q heterogeneity test was: Chi² = 3.91, d.f. = 3, P = 0.27. The random-effects model exposed a low amount of variability between studies. The heterogeneity statistic I² (%) = 23 (interpreted as low, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.04. When a random effect analysis was implemented, a small effect was detected on trained only subjects for single-joint resistance exercises (ES 0.37; 95% CI −0.01 to 0.75). Pre- to post-intervention strength change was greater when M3 was compared to S (ES difference 0.09) with no statistical significance (P-value = 0.06). The mean for S was 0.75 (95% CI 0.12–1.28). The mean ES for M3 was 0.84 (95% CI 0.39–1.28).

Examination of the effects of pre- vs. post-training strength (untrained subjects only) categorized as either S or M3 from five studies (Reid et al., 1987; Paulsen et al., 2003; Humburg et al., 2007; Bottaro et al., 2009; Sooneste et al., 2013). The random-effects model showed a low to moderate amount of variability between studies. Q heterogeneity test was: Chi² = 6.76, d.f. = 4, P = 0.15. The heterogeneity statistic I² (%) = 41 (interpreted as low-moderate, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.06, and a moderate effect was observed on untrained only subjects for single-joint exercises (ES 0.56; 95% CI 0.21–0.91). Pre- to post-intervention strength change was marginally greater when M3 was compared to S (ES difference 0.23) with statistical significance (P = 0.002). The mean for S was 0.51 (95% CI 0.24–0.77). The mean ES for M3 was 0.74 (95% CI 0.47–1.02).

Subgroup examination of upper body single-joint exercises are presented in the forest plot. The pooled mean ES estimates comprised of six studies (Reid et al., 1987; Hass et al., 2000; Humburg et al., 2007; Bottaro et al., 2009; Baker et al., 2013; Sooneste et al., 2013). The random-effects model exposed a low to moderate amount of variability between studies. Q heterogeneity test was: Chi² = 2.74, d.f. = 5, P = 0.74. The heterogeneity statistic I² (%) = 0 (interpreted as no heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00, and a trivial effect was detected (ES 0.20; 95% CI −0.07 to 0.47). Pre- to post-intervention strength change was comparable when M3 was compared with S (ES difference 0.07) with no statistically significance (P = 0.16). The mean ES for S was 0.69 (95% CI 0.28–1.11). The mean ES for M3 was 0.76 (95% CI 0.42–1.09). Subgroup examination of S vs. M3 pre- to post-intervention strength differences on trained only and untrained only subjects was not possible due to inadequate study data.

Examination on lower body single-joint exercises comprised of six studies (Reid et al., 1987; Hass et al., 2000; Schlumberger et al., 2001; Paulsen et al., 2003; Kelly et al., 2007; Bottaro et al., 2009). The random-effects model exposed a moderate amount of variability between studies (I² = 63%). Removal of Hass et al. (2000) data, resulted in no heterogeneity (Q heterogeneity test was: Chi² = 2.24, d.f. = 4, P = 0.69. The heterogeneity statistic I² (%) = 0 (interpreted as no heterogeneity, Higgins et al., 2003), and the tau² test (between-trials variance) = 0.00, and a small effect was observed (ES 0.41; 95% CI 0.08–0.74). Pre- to post-intervention strength change was greater when M3 was compared with S (ES difference 0.32) with statistical significance (P = 0.02). The mean ES for S was 0.40 (95% CI 0.17–0.63). The mean ES for M3 was 0.72 (95% CI 0.51–0.93). Subgroup examination of S vs. M3 pre- to post-intervention strength differences on trained only and untrained only subjects was not viable due to inadequate study data.

The scientific literature on daily set-volume has been heavily contested with some suggesting that S produce similar adaptations to MS (Silvester et al., 1982; Messier and Dill, 1985; Reid et al., 1987; Pollock et al., 1993; Starkey et al., 1996; Hass et al., 2000). However, others indicate that MS produces greater strength, hypertrophy, and power adaptations (Kraemer, 1997; Hass et al., 2000; Borst et al., 2001; Marx et al., 2001; McBride et al., 2003; Paulsen et al., 2003). In the context of pre-flight conditioning, the initial physical training status of each crew member must be fully considered. Specifically, as the inclusion of smaller doses of daily RT (e.g., S per exercise) may be appropriate to develop muscular strength in less conditioned (untrained) crew members, whereas more substantial doses of RT (e.g., ≥M3 per exercise) may be essential to attain further strength.

Currently, the prescription of RT for pre-flight conditioning is recommended from ground-based models and experience gained during previous missions. Present recommendations for ST consist of one-to-three-sets per exercises for untrained individuals and MS used with variations in RT volume and intensity over a period for trained individuals [American College of Sports Medicine Position Stand (ACSM), 2009]. These recommendations on terrestrial RT set-volume have been subject to strong criticism due to reported methodological constraints within the included meta-analyses (Smith and Bruce-Low, 2004; Winett, 2004; Otto and Carpinelli, 2006; Carpinelli, 2012; Fisher, 2012). For example, the meta-analysis by Rhea et al. (2003) presented data that may have nullified the mean ES and spuriously affected the results. This is due to increasing the heterogeneity of the meta-analysis, therefore, erroneously favoring MS programming. Rhea and colleagues reported that untrained subjects should perform four-sets for strength increases (ES = 2.28 ± 1.96 SD), compared with S (ES = 1.16 ± 1.59 SD), two-sets (ES = 1.75 ± 3.23 SD), and M3 (ES = 1.94 ± 3.23 SD). However, the ES standard deviation for M3 was significantly larger in comparison to the other treatment groups, with no explanation provided. Otto and Carpinelli (2006) stated that several critical errors could invalidate the results including bias in the selection process of studies, incorrect classification of subjects training status, and variances of RT loading.

A meta-analysis by Wolfe et al. (2004) suggested that untrained subjects in the initial stages of training (6–15 weeks) perform with an S programme. Wolfe et al. (2004) also reported that untrained subjects had comparable pre-post-strength gains as those of trained individuals when performing MS. When subjects performed, RT exercises to physical failure vs. subjects' perceived end, multiple-set programmes generated more substantial increases in strength compared to S (P ≤ 0.002). A meta-analysis by Krieger (2009) examined the effects of S vs. multiple-sets of exercises on strength and suggested that two-to-three-sets were associated with a more significant ES than S. Krieger (2009) concluded that two-to-three-sets per resistance exercise were associated with 46% greater strength gains than S in both trained and untrained subjects. Finally, a meta-analysis by Fröhlich et al. (2010) of 72 studies found S training regimes to be the equivalent of multiple-set training for short intervention phases but multi-set training to be superior overextended RT intervention periods.

This current meta-analysis provides evidence that supports the contention that M3 RT leads to greater strength gains than S programming for untrained individuals or those astronauts that may be on the early phase of pre-flight conditioning. However, this analysis examined the differences in pre-post-strength gain between body segments and joint types rather than only pooling RT exercises together to generate ES. For upper body exercise, untrained subjects demonstrated a larger pooled mean ES estimate for M3 0.82 (95% CI 0.52–1.11) compared with S (0.60; 95% CI 0.34–0.85), however these results were not statistically significant (P = 0.42). Also, a significant degree of heterogeneity was present when exercises were “pooled” together. This intriguingly may explain why other previous meta-analyses reported varying results with regards to specific set-volume and encountered criticism from others as the included studies combined RT exercises to aggregate ES. Subgroup examination on untrained subjects was only feasible for upper and lower body and single-joint only exercises due to limited available data for multi-joint exercise. Untrained subjects pre- to post-strength gains on upper body exercise was greater when M3 was compared with S (ES difference 0.23) but was not statistically significant (P = 0.42). In contrast, lower body exercises reported significantly greater strength changes (P = 0.005) with M3 compared with S (ES difference 0.28). Strength gains on single-joint exercises exposed a larger pooled mean ES estimate for M3 (0.74; 95% CI 0.47–1.02) compared with S (0.51; 95% CI 0.24–0.77). When S was used as a control group and M3 used as the experimental set-volume, a moderate ES of 0.56 (95% CI 0.21–0.91; P = 0.002) suggested that M3 was effective in generating larger strength increases.

Even though evidence from this analysis supports increase set-volume for strength gain, caution is warranted concerning the interpretation of the findings as exercises performed in the included studies may not transfer toward improved operational task functioning. Furthermore, even though superior increases in strength may be produced with M3, the necessary time and physical effort required to attain such strength developments may not be essential for preparation toward spaceflight. The time to perform this additional conditioning may be better served in completing other necessary operational tasks.

Current recommendations from NASA for pre-flight conditioning suggest that trained astronauts who are in the final stages of pre-spaceflight preparation perform MS with variations in RT volume and intensity (American College of Sports Medicine Position Stand (ACSM), 2009; Garber et al., 2011). Several ground-based meta-analyses (Rhea et al., 2003; Peterson et al., 2004; Wolfe et al., 2004; Krieger, 2009; Fröhlich et al., 2010) that are included within the American College of Sports Medicine Position Stand (ACSM) (2009) recommendations have reported superior results when performing MS for trained subjects. Rhea et al. (2003) reported strength increases in the bench press of 20% for S, compared with a 33% increase with M3. Pre- to post leg strength increased by 25.4% for S compared with an increase of 52.1% with M3. However, the inclusion of the Rhea et al. (2002a) study may have generated errors or bias when comparing the means of the one-set and three-set groups (bench press ES = 2.3 and leg press ES 6.5). Also, the post-training standard deviation bench press findings were two-to-three times greater to the pre-training standard deviation in both groups, and the authors did not provide confidence intervals for ES. The meta-analysis by Peterson et al. (2004) stated that experienced athletes should complete eight-sets per muscle group to increase muscular strength. However, due to the low number of ES in the eight-set group, these conclusions may be unreliable when evaluating the specific percentage of 1RM training. An additional meta-analysis performed by Wolfe et al. (2004) reported that MS generated more significant increases in strength (P ≤ 0.001) for trained subjects.

This current meta-analysis cautiously provides evidence concerning increased sets and strength gain with support toward the use of additional sets (up to M3) compared to S for trained astronauts pre-flight conditioning programme. In this analysis, we reveal that multi-joint only exercises on trained subjects demonstrated a larger pooled mean ES estimate for M3 (0.99; 95% CI 0.60–1.38) compared with S (0.68; 95% CI 0.27–1.10) suggesting that M3 was more effective in producing larger strength gains. Further examination of upper body multi-joint only exercises further supports the use of M3. The pooled mean ES estimate for S was 0.59 (95% CI 0.09–1.08) compared with M3 (0.91; 95% CI 0.49–1.32) suggesting that M3 was more effective in producing strength gains. For single-joint exercises on trained subjects, there was a larger pooled mean ES estimate for M3 (0.84; 95% CI 0.39–1.28) compared with S (0.75; 95% CI 0.12–1.28). However, this was not statistically significant in producing more substantial strength gains.

Exercise prescription currently implemented by space agencies include the prescription of greater RT set-volume, but crew members do not perform the volume in a linear manner. It could be hypothesized, therefore, that by increasing set-volume prior to the commencement of spaceflight may develop astronaut's pre-flight levels of upper and lower body strength. Similarly, Krieger (2010) found that this range of sets and reps was superior in producing muscular hypertrophy. This suggests that the atrophy effects of long duration spaceflight could be mitigated, but not prevented, by a more rigorous deployment of multi-set RT methods. Astronauts that are appropriately conditioned and have progressed in their individualized RT toward a trained state should perform MS per exercise to maximize muscular strength, as it could safeguard against decreases in strength. However, caution should be given with the prescription of daily RT set-volume as exposing astronaut's to chronically high training volume may expose crew members to overtraining syndrome before spaceflight. Also, the inclusion of increased set-volume may also come at the determinant of the application toward training intensity and loading (e.g., 85% 1RM) and may feasibly inhibit strength development (Stone et al., 2006). Collectively, when accounting for other programme variables (RT loading and training frequency), M3 appears to a certain degree to be more advantageous for strength development in individuals that are trained. However, careful consideration must be given toward the astronauts initial pre-training status because of variance between individuals. Furthermore, comprehensive monitoring should be considered to prevent the potential of excess fatigue and to overtrain (Day et al., 2004; Halson, 2014).

Historically, exercise has been prescribed as a method to counteract the undesired physiological effects of long duration exposure to weightlessness and to safeguard astronaut's health (both acute and chronic). Space agencies have recently acknowledged that the pre-flight physical fitness of astronauts must be higher than age centered normative measurements before spaceflight. The preparation of crew members has advanced from methods that traditionally only improved physiological tolerance toward spaceflight to practices that prepare individuals to live and work in space (Garshnek, 1989). These physical training protocols are employed as a method of countering detrimental physiological adaptations that result from μG. The physical conditioning and preparation of astronauts for spaceflight are essential; however, space agencies approach mission preparation differently. It is evident that most agencies foster a graded dose-response continuum with the initial phase of training emphasizing lower set-volume progressing to MS at the concluding stages of flight preparation.

Based on evidence generated from this analysis, it would be appropriate to suggest that astronauts perform RT via a graded dose-response continuum in preparation for in space transit. However, published research in this area is limited with evidence centered on ground-based studies from older adult populations and bed rest studies that may not fully replicate the physiological pre-spaceflight muscle conditioning status of astronauts. More evidence is required concerning the effects of RT bouts and the dose-response relationship on the training status, age and sex differences for protecting the muscular systems. A greater understanding of the mechanisms for physiological conditioning would, therefore, help to develop appropriate individual centered countermeasures that mitigate deconditioning and must be an essential area of focus for future space physiology research. Ensuring each astronaut's pre-spaceflight strength is at the desired level may help to counteract in-flight and post-flight physiological stress which is paramount for the development of a viable human space exploration programme that goes beyond Earth's orbit.

There are numerous strengths of this meta-analysis that separate it from previous investigations that compare S vs. M3 configurations. Considerations toward evidence derived from previous critical reviews that identify limitations regarding the quality and selection of included studies that may bias outcomes have been applied [in part]. These review articles debated the findings from previous meta-analytical studies, identifying confounding factors and imprecisions that may have influenced outcome reliability. Previous review articles in this area, have reported issues concerning robustness and rigor that control confounding variables when comparing strength outcomes. Subsequently, this meta-analysis has been founded on critical consensus, current studies, and previous evidence gathered from meta-analyses presenting a re-examination of the evidence. Greater emphasis was placed on strict inclusion and exclusion criteria, specifying why each study was excluded to allow for greater transparency. The identification and the inclusion criteria of studies were restricted to reduce heterogeneity between studies making it easier to apply the results to specific population groups (trained and untrained). This enabled direct assessments to be drawn from studies that compared S and M3 while endeavoring to control for other variables that influence strength outcomes.

This meta-analysis endeavored to limit where possible publication bias which is the most significant source of type I error (the increase of false positive results). Graphical representation via funnel plots evaluated the presence of publication bias which was not present in other meta-analytical studies (Rhea et al., 2002b, 2003; Peterson et al., 2004; Wolfe et al., 2004; Krieger, 2010). Where no publication bias was present, the ES of each included study is distributed symmetrically around the underlying true ES, with a more random variation of this value in smaller studies. Asymmetry in the plot is suggestive of bias, most often due to smaller studies which are non-significant or have an effect in the opposite direction from that assumed (Greco et al., 2013). Within our design, we were conscious of potential heterogeneity and implemented a multi-level model as a method for assessing heterogeneity across studies. Also, visual inspection of forests plots further verified this and indicated the need for further subgroupings of studies which reduced the degree of heterogeneity.

Moreover, subgroup analysis was performed to allow for the assessment of training status (trained and untrained) separately and also the analysis of multi-joint and single-joint exercises. This helped to reduce heterogeneity between groups and allowed for more consistent comparisons to be made. Lastly, efforts were made to restrict the subject pool to population groups (trained and untrained) that were physically active and within a comparable age range that mirrored that of an astronaut's pre-flight fitness status. This is due to the disparity existing with astronaut's fundamental physiological characteristics, with physical fitness levels ranging from sedentary individuals who performed little or no physical activity to athletic standard.

Limitations of this meta-analysis include the use of small sections of the published information that is often derived from an insufficient range of methodological designs that has been determined by the inadequacy of available primary data. As with all meta-analyses, limitations and restrictions exist that include the assessment of aggregate outcomes that inevitably do not assess the same construct (i.e., “comparing apples and oranges”). This is the process of search and retrieval of available journal articles, and the potential effect of publication bias (Rosenthal, 1991). Cooper and Hedges (1994) stated that “retrieval bias” may exist even when robust inclusion criteria for gathering journals articles is applied. For example, in this analysis, retrieval bias may have occurred due to the selection and retrieval of journal articles published only in English. Although we strived to include journal articles from high-quality sources, the number of relevant studies was restricted, and there remained variances in experimental design and control among included studies. For example, four of the 12 included journal articles used RCTs. The other eight used a RAN that did not include a control group; instead, a repeated measures design was used with a baseline measure implemented as the control. However, baseline measures were not uniformly applied across those studies.

In this meta-analysis, the strength increases may be due to other physiological factors that have been acknowledged to affect strength changes including sex, age, physical activity levels, prior training history, genetics and endocrine status. This analysis was unable to subcategories subjects into either male or female population groups; instead, it combined population groups. Several studies have reported that sex influences muscle morphology and functioning (Chorney and Bourgeois, 1999; Sale, 1999). Häkkinen et al. (1996) reported that males have greater muscle strength and size than women due to higher levels of anabolic hormones and greater body mass. Also, it has been inferred that due to lower blood androgen levels in women that the response to RT would induce less muscle hypertrophy compared to men. Equally, some studies have reported no differences between sexes with similar improvements in strength adaptations (Colliander and Tesch, 1991; O'Hagan et al., 1995; Roth et al., 2001).

For this analysis, every effort was made to assess strength outcomes of comparable training designs and methodological construct. However, attempting to include studies that had a standardized RT programme was problematic. Even when attempting to control for variance within studies, confounding issues that may have influenced strength development were present. The RT loading had various ranges (63–93% 1RM) that could not be controlled for which could affect strength gains. As training loading is one of the most critical parameters of ST and the volume of RT training makes it challenging to provide a clear dose-response relationship that supports M3 compared to S programming for strength gain. The previous analyses found that loading of 60% of subjects 1RM (Rhea et al., 2003; Peterson et al., 2004) was sufficient to increase strength for untrained subjects, with trained subjects recommended to perform between 80 and 85% of their 1RM [American College of Sports Medicine Position Stand (ACSM), 2009]. The disparity in the training programme type and the order of resistance exercise between groups were not comparable in all included studies which feasibly could impact upon the set number and muscular strength. Scientific literature is currently equivocal regarding the implementation of heavy or light RT loads for muscular adaptations (Fisher et al., 2017). It must be acknowledged that the different repetition ranges included within this meta-analysis may also be a possible limitation. Several studies have reported comparable strength gains when training at different repetition ranges (Morton et al., 2011; Fisher et al., 2017) whereas others have shown improved strength when individuals trained with higher RT load and a low number of repetitions (Campos et al., 2002). This could be explained by the specificity of the strength test, as training with a low number of repetitions is closer to the 1RM test (Buckner et al., 2017; Gentil et al., 2017).

Finally, strength adaptations may have resulted from the performance of repeated 1RM testing as the resistance exercise loading specificity of the 1RM-tested exercises may have indirectly affected the subject's strength. For example, a subject's pre-to-post leg extension results may have increased, due to the performance of the leg press but not to the same amount as a leg press itself. This is supported by Dankel et al. (2016) who performed 1RM and maximum voluntary contraction testing on elbow flexion. The results intimated that the increase in the 1RM result might not be solely correlated to set-volume, but due to the “learning effect” of the specific resistance exercise. This could explain the variation in untrained subjects' strength due to the principle of training specificity, as neurological learning effects could, therefore, explain the broad range of heterogeneity between the inexperienced group.

PJ is employed by KBR. The remaining 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.