Effects of Resistance Training Performed to Failure or Not to Failure on Muscle Strength, Hypertrophy, and Power Output: A Systematic Review With Meta-Analysis

Abstract

Vieira, AF, Umpierre, D, Teodoro, JL, Lisboa, SC, Baroni, BM, Izquierdo, M, and Cadore, EL. Effects of resistance training performed to failure or not to failure on muscle strength, hypertrophy, and power output: A systematic review with meta-analysis. J Strength Cond Res 35(4): 1165–1175, 2021—The aim of this review was to summarize the evidence from longitudinal studies assessing the effects induced by resistance training (RT) performed to failure (RTF) vs. not to failure (RTNF) on muscle strength, hypertrophy, and power output in adults. Three electronic databases were searched using terms related to RTF and RTNF. Studies were eligible if they met the following criteria: randomized and nonrandomized studies comparing the effects of RTF vs. RTNF on muscle hypertrophy, maximal strength, and muscle power in adults, and RT intervention ≥6 weeks. Results were presented as standardized mean differences (SMDs) between treatments with 95% confidence intervals, and calculations were performed using random effects models. Significance was accepted when p < 0.05. Thirteen studies were included in this review. No difference was found between RTF and RTNF on maximal strength in overall analysis (SMD: −0.08; p = 0.642), but greater strength increase was observed in RTNF considering nonequalized volumes (SMD: −0.34; p = 0.048). Resistance training performed to failure showed a greater increase in muscle hypertrophy than RTNF (SMD: 0.75; p = 0.005), whereas no difference was observed considering equalized RT volumes. No difference was found between RTF and RTNF on muscle power considering overall analysis (SMD: −0.20; p = 0.239), whereas greater improvement was observed in RTNF considering nonequalized RT volumes (SMD: −0.61; p = 0.025). Resistance training not to failure may induce comparable or even greater improvements in maximal dynamic strength and power output, whereas no difference between RTF vs. RTNF is observed on muscle hypertrophy, considering equalized RT volumes.

Introduction

It has been widely shown that resistance training (RT) improves maximal and explosive strength, and it is beneficial for sports performance in athletes (^25,26) and for functional capacity in nonathletes (¹⁴). One of the most popular approaches to prescribing RT is establishing the load correspondent to a range of maximal repetitions (RMs) (^3,28,29), in which transient concentric failure is achieved at the end of performed sets (i.e., 4–6RMs or 10–12RMs). It has been argued that performing RT sets to failure (RTF) would maximize motor unit (MU) recruitment and would, consequently, optimize neuromuscular adaptations to RT (^59,60). This theory is based on the size principle, suggesting that as consecutive repetitions are performed, the lower-threshold MUs (i.e., those composed of type I muscle fibers) are fatigued and, consequently, high-threshold MU (i.e., those composed of type II muscle fibers) recruitment will be maximized (^6,59,60). However, there are some controversial findings in studies investigating this issue using electromyography signals (EMGs) because it has been demonstrated that a higher EMG signal is achieved and stabilized before the repetition corresponding to muscular failure (^10,55). In addition, over several sets of an exercise, such accumulated fatigue can result in a reduction in total overload (i.e., volume and load), compared with nonfailure sets (^17,36). Based on the association between total mechanical overload (i.e., sets × repetitions × load) and muscle hypertrophy (^27,51), a reduction in overload could impair the hypertrophic responses.

In healthy people, RTF seems to result in marked neuromuscular gains (^1,24,28,44), although evidence that training leading to concentric failure is superior to RT not leading to failure (RTNF). Indeed, several studies have shown that RTF did not induce additional muscle strength gains compared with RTNF in young trained and untrained populations (^{12,13,25,26,35,48}), whereas a smaller number of studies observed greater strength increase after RTF (^8,15,47). In addition, it seems that RTF does not induce further muscle size enhancements in young (^35,40,48) and older individuals (⁵³), although this outcome is less investigated. Regarding muscle power output, it has been shown that RTF may compromise muscle power improvements in highly trained athletes (^25,26), whereas another study has shown similar muscle power adaptations between RTF and RTNF in older adults (²). Therefore, there are controversial results regarding the influence of repetitions to failure in the RT adaptations. In this regard, a meta-analysis study could be useful to provide more solid evidence, due to the controversial findings.

In a previous meta-analysis by Davies et al. (⁶), who subsequently published an article erratum (⁷), the authors demonstrated that there are no differences in strength gains after RTF and RTNF. Notwithstanding, this study only addressed maximal strength, and analyses on muscle hypertrophy and muscle power output adaptations were not performed. In addition, since the publication of this work (⁶), several other articles on this topic have been published, providing more data on the comparison between failure and not to failure approaches during RT (^{2,12,35,40,53}).

Although sets using repetitions to failure promote mechanical and muscle function adaptations, it is also notable that a certain degree of fatigue may induce high levels of discomfort and physical exertion, preventing the correct execution of the movement, particularly in nonexperienced practitioners (^56,59). In this sense, it is believed that the performance of RTF in the long term could increase the risk of overuse injuries (⁶). Moreover, repetition to failure also implies a longer time under tension, leading to greater increases in blood pressure, heart rate, and rate-pressure product (^16,34,39), which may increase the risk of cardiovascular complications in some populations. In addition, RTF induces a greater metabolic impact at the cellular level (^17,18), which may result in the need of a greater time of recovery between exercise sessions.

Therefore, it seems relevant to determine whether there are additional benefits of performing RTF or if its adaptations are comparable with those observed when submaximal repetitions are performed, taking into account the influence of total volume on these adaptations (i.e., compensating or not compensating the number of repetitions with additional sets). Thus, the purpose of this study was to systematically review randomized and nonrandomized longitudinal studies on the effects of RT performed to concentric failure or not to failure on muscle strength, hypertrophy, and maximal power output in healthy young subjects and in older adults. In addition, we also assessed if there was an influence of RT volume (equalized or not equalized) in this comparison. Our hypothesis was that RTF would not provide additional benefits on muscle strength and power output. Moreover, we also hypothesized that no difference between RTF and RTNF would be observed on muscle hypertrophy, considering equalized RT volumes.

Methods

Experimental Approach to the Problem

To test the authors' hypothesis, we performed a systematic review with a meta-analysis of longitudinal studies investigating the effects of RT performed to failure vs. not to failure on muscle strength, hypertrophy, and power output. This review was registered at http://www.crd.york.ac.uk/prospero as CRD42020155608. The study has been reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (^32,52).

Subjects

Three hundred eighty-four subjects from 13 studies were included in the analysis. Ninety-two studies included only men and only women, respectively; and 2 studies included a mixed sample. Three hundred thirty-two subjects were young adults (mean age range: 21.6–26.7 years old), while 52 were older adults (mean age range: 65.6–66.7 years old). In addition, four studies were performed with athletes (30.8%), eight with untrained individuals (61.5%) and one with recreationally resistance-trained individuals (7.7%) (Table 1). All studies mentioned that written informed consent documents were obtained for all subjects. This study was approved by the Institutional Review Board of Universidade Federal do Rio Grande do Sul (Federal University of Rio Grande do Sul).

Procedures

Eligibility Criteria

In this review, we considered longitudinal studies (randomized or nonrandomized) that compared the effects of RTF vs. RTNF for a period ≥6 weeks on muscle hypertrophy, muscle strength, and maximal power output in young and older trained and untrained adults. Studies were included if they assessed at least one of the following outcomes: (a) muscle hypertrophy using ultrasound, magnetic resonance images, or computed tomography measurements; (b) dynamic muscle strength through isoinertial tests (i.e., 1RM, 3, or 6RMs) in at least one exercise; or (c) maximal power output using vertical jumps, free weights, or plated load RT exercises. In the case of studies with different publications related to the same outcome (i.e., short-term and long-term adaptations), only one study was included. In this case, we chose to include studies with training periods similar to those of the other studies included in this review.

Search Strategy

The search was performed in October 2019 using the electronic databases MEDLINE (PubMed), Web of Knowledge (Web of Science), and Cochrane. In addition, manual searches from the references of the included studies were performed. After starting the literature search, we excluded 2 databases previously described during the registration. These exclusions were due to the high cost associated with the search in these databases. In addition, we performed a search for gray literature at http://www.opengrey.eu/. There was no limitation on the language or year of the study.

The search comprised the following terms and MeSH terms (and their respective related terms): “humans,” “adult,” “aged,” “resistance training,” “maximal repetitions,” “muscular failure,” “muscular exhaustion,” “muscle fatigue,” “failure,” “repetition failure,” “repetition exhaustion,” and “repetition maximum.” To optimize the capture of relevant references, such terms were combined by boolean operators (OR and AND). The full search strategy performed in the PubMed database is available as Supplemental Digital Content 1 (see Supplementary Material, https://links.lww.com/JSCR/A251).

Selection of Studies

The selection of studies was based on the eligibility criteria previously adopted and performed independently and in duplicate. First, 2 pairs of researchers independently evaluated the titles and abstracts of all studies found in the search (A.F.V. and E.L.C., and J.L.T. and S.C.L.; each pair evaluated half of the studies). Articles whose abstracts did not provide sufficient information as per the inclusion and exclusion criteria were assessed separately in full. Subsequently, each study selected in the previous phase was fully evaluated and selected by the reviewers independently. Disagreements were resolved by consensus, and in cases of persistence, a third investigator resolved the disagreement between the other pair of researchers (E.L.C. and J.L.T.). To avoid inclusion of duplicate studies (different publications using identical study groups), we screened the period and place of recruitment, and the authors were contacted for clarification when necessary.

Data Extraction

Standardized forms were adopted for data extraction, and the data extraction was performed independently by 2 pairs of reviewers (A.F.V. and E.L.C., and J.L.T. and S.C.L.; each pair extracted data from half of the studies). Eventual disagreements were resolved by consensus or by a third investigator from another pair (E.L.C. and J.L.T.). In this phase, the main characteristics of the selected studies, such as sample size and sample characterization, variables related to interventions, and results of the outcomes of interest, were detailed. Missing data were requested from the researcher of the study in question. In case of no answer, denying provision, or data loss, the article or outcome was excluded. For data presented only graphically, the results were extracted using DigitizeIt software.

The extracted outcomes were the absolute deltas of the values referring to muscle hypertrophy, maximal strength, and maximal power output. When not available, the delta was calculated from the values obtained before and after the intervention, and the delta SD was imputed by the equation proposed by Higgins and Green (²⁰).

In studies comparing RTF with 2 or more RTNF groups using different training approaches, we chose the most similar comparator group in terms of training variables (i.e., same number of exercises and more similar contraction speed in the concentric and eccentric phases). However, studies in which the RTF group was compared with RTNF groups using equalized and nonequalized volumes (^2,35,53), both comparisons were included in the analysis. In these cases, the sample of the RTF group was divided in half in each comparison to avoid overestimating the study weight and the sample size in the analyses (¹⁹).

Risk of Bias Assessment

The assessment of risk of bias in the individual studies included adequate random sequence generation, allocation concealment, blinding of subjects and personnel, blinding of outcome assessment, description of losses and exclusions, and intention-to-treat analysis, as proposed by the Cochrane Collaboration (²⁰). When these characteristics were described in the published document, it was considered that the criteria were met and they were classified as “low risk” or “high risk.” Studies that did not describe these data were classified as “unclear risk.” This evaluation was performed independently by 2 pairs of reviewers (A.F.V. and E.L.C., and J.L.T. and S.C.L.).

Statistical Analyses

Statistical analyses was performed through a meta-analysis comprising the comparison of RTF with RTNF on muscle hypertrophy, maximal dynamic strength, and maximal power output. Subgroup analyses included comparisons between RTF and RTNF with and without equalized volumes, as well as overall analysis (i.e., considering both cases within the same comparison). In addition, sensitivity analyses were performed for lower-body and upper-body maximal dynamic strength, muscle power output, and muscle hypertrophy. Moreover, sensitivity analyses on maximal strength and maximal power were also performed for sports athletes and nonathletes, as well as single-joint and multijoint exercises for the upper and lower body. Furthermore, in each analysis, we removed studies individually, one after the other. This strategy (known as “leave-one-out”) is recommended for further exploring between-study heterogeneity (beyond the I-squared calculation) because it allows for the identification of any substantial change in the direction of results whenever a single study is removed from the analysis. This exploratory procedure is a recommended practice, particularly for meta-analyses with a limited number of studies (⁵⁸). In all comparisons, we considered only analyses including at least 3 studies. Thus, to include at least 3 studies in each sensitivity or subgroup analysis, we pooled data from exercises composed of similar muscle groups. For example, for the analysis of lower-body muscles, we pooled data regarding the leg press and squat exercises, whereas for upper-body analysis, we pooled data from bench row and biceps curl exercises.

The results are presented as standardized mean differences (SMDs) between treatments with 95% confidence intervals (CIs). The calculations were performed using random effects models. The statistical heterogeneity of treatment effects between studies was assessed using the I² inconsistency test, considering that values higher than 50% indicated high heterogeneity (²¹). Values of α ≤ 0.05 were considered statistically significant.

Publication bias was verified through visual observation of the funnel graph of the analyzed variable. Asymmetry was tested using the Begg and Egger test and was considered to be significant when p < 0.10. In case of publication bias, the trim-and-fill test was used to estimate the publication bias effects on interpreting the results. All analyses were performed using Stata version 15.1.

Results

Study Selection

The search of MEDLINE (PubMed), Web of Knowledge (Web of Science), and Cochrane databases provided a total of 4,150 citations. In addition, 3 studies were identified through manual searches from the references of the included studies, and 271 studies were found through the search for gray literature. After duplicates were removed, 4,096 studies remained. Of these, 4,080 studies were discarded after reviewing the titles and abstracts. Thus, the full texts of the remaining 16 citations were examined in more detail (none of the gray literature). After integral reading, 13 studies met the inclusion criteria and were included in the quantitative analysis (Figure 1).

Figure 1.: Flowchart of studies included. RM = repetition maximum.

Description of the Studies

The 13 studies included in this review contributed to the analysis and comprised 384 subjects. Of these, 171 individuals participated in RTF interventions, whereas 213 participated in RTNF. Regarding the sexes in study samples, 9 and 2 studies included only men and only women, respectively, and 2 studies included a mixed sample. Most subjects included were young adults (86.5%), whereas only 52 (13.5%) were older adults. Furthermore, 4 studies were performed with sport athletes (30.8%), 8 with untrained individuals (61.5%), and 1 with recreationally resistance-trained individuals (7.7%) (Table 1).

Table 1 - Characteristics of subjects.^*

Studies	Groups	Subjects (n)	Sex (M/F)	Age (y)†	Body mass (kg)†	BMI (kg·m−2)†	Training status
Cadore et al. (2),‡,§	Failure‡,§	17	17/0	66.1 ± 5.0	79.4 ± 10.6	27.1 ± 3.2	UT
	Nonfailure‡	20	20/0	66.7 ± 6.1	80.3 ± 10.6	27.7 ± 2.8	UT
	Nonfailure§	15	15/0	65.6 ± 3.4	87.9 ± 0.1	30.9 ± 4.8	UT
da Silva et al. (53),‡,§	Failure‡,§	17	17/0	66.1 ± 5.0	79.4 ± 10.6	27.1 ± 3.2	UT
	Nonfailure‡	20	20/0	66.7 ± 6.1	80.3 ± 10.6	27.7 ± 2.8	UT
	Nonfailure§	15	15/0	65.6 ± 3.4	87.9 ± 0.1	30.9 ± 4.8	UT
Drinkwater et al. (8)	Failure	15	15/0	NR	NR	NR	SA
	Nonfailure	11	11/0	NR	NR	NR	SA
Drinkwater et al. (9)	Failure	7	7/0	NR	NR	NR	SA
	Nonfailure	7	7/0	NR	NR	NR	SA
Folland et al. (13)	Failure	12	8/4	22.0 ± 2.0	70.0 ± 3.0	NR	UT
	Nonfailure	11	7/4	20.0 ± 1.0	68.0 ± 7.0	NR	UT
Izquierdo et al. (25)	Failure	14	14/0	24.8 ± 2.9	81.1 ± 4.2	24.9 ± 2.5	SA
	Nonfailure	15	15/0	23.9 ± 1.9	80.5 ± 7.4	24.6 ± 1.9	SA
Izquierdo-Gabarren et al. (26)	Failure	14	14/0	25.4 ± 4.2	79.8 ± 5.3	NR	SA
	Nonfailure	15	15/0	26.7 ± 5.7	83.2 ± 6.3	NR	SA
Kramer et al. (30)	Failure	16	16/0	NR	78.4 ± 8.4	NR	T
	Nonfailure	14	14/0	NR	76.8 ± 10.1	NR	T
Martorelli et al. (35),‡,§	Failure‡,§	30	0/30	22.3 ± 3.8	63.7 ± 22.5	NR	UT
	Nonfailure‡	27	0/27	21.6 ± 3.3	62.5 ± 14.1	NR	UT
	Nonfailure§	32	0/32	21.7 ± 2.8	60.2 ± 13.5	NR	UT
Nóbrega et al. (40)	Failure	14	14/0	NR	NR	NR	UT
	Nonfailure	14	14/0	NR	NR	NR	UT
Rooney et al. (47)	Failure	13	NR	NR	NR	NR	UT
	Nonfailure	14	NR	NR	NR	NR	UT
Sampson and Groeller, (48)	Failure	10	10/0	23.4 ± 6.6	76.9 ± 0.2	NR	UT
	Nonfailure	10	10/0	23.7 ± 6.2	85.0 ± 13.7	NR	UT
Sanborn et al. (49)	Failure	9	0/9	NR	62.8 ± 9.2	NR	UT
	Nonfailure	8	0/8	NR	70.9 ± 12.1	NR	UT

^*M = males; F = females; BMI = body mass index; T = trained; UT = untrained; SA = sports athletes; NR = not reported.

^†Mean ± SD.

The interventions consisted of RT performed with 1–40 sets of 1–22 repetitions at intensities ranging between 65% 1RM and 92% 1RM. The weekly frequency of RT protocols ranged from 2 (46.2%) to 3 times (53.8%), and RT periods ranged from 6 to 14 weeks of follow-up. Of the 13 studies included, 5 studies used comparator groups with a nonequalized RT volume, 5 with an equalized RT volume, and 3 with both comparisons. In addition, 4 studies assessed muscle hypertrophy, 12 assessed maximal strength, and 5 assessed maximal power output (Table 2).

Table 2 - Training characteristics of studies.^*

Studies	Groups	Sets	Repetitions	Intensity	Rest† (s)	Volume‡	Frequency (days per week)	Duration (wk)	Hypertrophy test	Strength test	Power output test
Cadore et al. (2),§,‖	Failure§,‖	2 to 3	∼7.7–∼22.2	65–75% 1RM	120	Nonequalized	2	12	NE	NE	SJ and CMJ
	Nonfailure§	2 to 3	4–10	65–75% 1RM	120	Equalized
	Nonfailure‖	4 to 6	4–10	65–75% 1RM	120
da Silva et al. (53),§,‖	Failure§,‖	2 to 3	∼7.7–∼22.2	65–75% 1RM	120	Nonequalized	2	12	Ultrasound QF	1RM LP and KE	NE
	Nonfailure§	2 to 3	4–10	65–75% 1RM	120	Equalized
	Nonfailure‖	4 to 6	4–10	65–75% 1RM	120
Drinkwater et al. (8)	Failure	4	6	80–105% 6RM	230	Equalized	3	6	NE	6RM BP	40 kg BP
	Nonfailure	8	3	80–105% 6RM	100
Drinkwater et al. (9)	Failure	4	6	90–100% 6RM	165	Equalized	3	6	NE	3RM and 6RM BP	40 kg BP
	Nonfailure	8	3	90–100% 6RM	73
Folland et al. (13)	Failure	4	10	75% 1RM	30	Equalized	3	9	NE	1RM KE	NE
	Nonfailure	40	1	75% 1RM	30
Izquierdo et al. (25)	Failure	3	5–10	80–100% 6-10RM	120	Equalized	2	11	NE	1RM BP and SQ	60% 1RM BP and SQ
	Nonfailure	6	3–5	80–100% 6-10RM	120
Izquierdo-Gabarren et al. (26)	Failure	3 to 4	4–10	75–92% 1RM	NR	Nonequalized	2	8	NE	1RM BR	BR at different loads
	Nonfailure	3 to 4	2–5	75–92% 1RM	NR
Kramer et al. (30)	Failure	1	8–12	75–81% 1RM	120 to 180	Nonequalized	3	14	NE	1RM SQ	NE
	Nonfailure	3	10	66–75% 1RM	120 to 180
Martorelli et al. (35),§,‖	Failure§,‖	3	NR	70% 1RM	120	Nonequalized	2	10	Ultrasound EF	1RM BC	NE
	Nonfailure§	3	7	70% 1RM	120	Equalized
	Nonfailure‖	4	7	70% 1RM	120
Nóbrega et al. (40)	Failure	3	Not clear	80% 1RM	120	Nonequalized	2	12	Ultrasound VL	1RM KE	NE
	Nonfailure	3	Not clear	80% 1RM	120
Rooney et al. (47)	Failure	1	6	6RM	No rest	Equalized	3	6	NE	1RM BC	NE
	Nonfailure	6	1	6RM	30
Sampson and Groeller (48)	Failure	4	∼6.1	85% 1RM	180	Nonequalized	3	12	Magnetic resonance EF	1RM BC	NE
	Nonfailure	4	∼4.2	85% 1RM	180
Sanborn et al. (49)	Failure	1	8–12	8 to 12RM	NR	Nonequalized	3	8	NE	1RM SQ	NE
	Nonfailure	3	2–10	2 to 10RM	NR

^*RM = repetition maximum; QF = quadriceps femoris; EF = elbow flexors; VL = vastus lateralis; LP = leg press; KE = knee extension; BP = bench press; SQ = squat; BR = bench row; BC = biceps curl; SJ = squat jump; CMJ = countermovement jump; NR = not reported; NE = not evaluated.

^†Between sets.

^‡Between intervention and comparator groups.

Risk of Bias

Among the studies included, 15.4% (2/13) clearly reported random sequence generation, 15.4% (2/13) reported allocation concealment, none (0/13) implemented blinding or masking procedures to subjects or personnel, 15.4% (2/13) blinded the assessors to the outcomes, 38.5% (5/13) described sample losses and exclusions, and none (0/13) performed intention-to-treat analyses (see Table S1, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251).

Effects of Interventions

Maximal Dynamic Strength: Meta-Analysis and Subgroup Analysis

Data on maximal strength were obtained from 12 studies (^{8,9,13,25,26,30,35,40,47–49,53}) comprising a total of 384 individuals. In the overall analysis, no difference was found between changes induced by RTF and RTNF (SMD: −0.08; 95% CI: −0.42, 0.26; p = 0.642; I²: 67.3%) (Figure 2). We also did not observe publication bias in this analysis (Egger's regression, p = 0.786) (see Figure S1, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251).

Figure 2.: Forest plot of the maximal strength promoted by resistance training (RT) performed to concentric failure vs. RT performed not to failure. The estimation for each subgroup (1: nonequalized volume and 2: equalized volume) and the combined effect (overall) are detailed. The squares and error bars signify the standardized mean differences (SMDs) and 95% confidence interval (95% CI) values; the diamonds represent the pooled estimates of random effects meta-analyses.

The results of the subgroup analysis based on RT volumes (i.e., equalized or not equalized) showed an advantage in maximal strength gains in favor of the nonfailure approach, considering nonequalized RT volumes (SMD: −0.34; 95% CI: −0.67, −0.003; p = 0.048; I²: 31.5%) (Figure 2). On the contrary, no significant difference was found between RTF and RTNF when considering equalized RT volumes (SMD: 0.16; 95% CI: −0.38, 0.69; p = 0.566; I²: 75.3%) (Figure 2).

Each study included in this analysis was removed individually. Therefore, when we removed 5 of the 12 studies included (^{26,30,40,48,49}), this difference in favor of RTNF when considering nonequalized volumes disappeared.

Sensitivity Analyses

Owing to the heterogeneity between the studies, sensitivity analyses were performed in the interventions for the following (see Figures S2–S7, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251): lower-body exercises in which a significant difference was observed in favor of RTNF (SMD: −0.38; 95% CI: −0.69, −0.07; p = 0.015; I²: 23.1%); lower-body exercises (comparator, nonequalized volume) (SMD: −0.37; 95% CI: −0.91, 0.17; p = 0.179; I²: 54.6%); lower-body exercises (comparator, equalized volume) in which there was a difference in favor of RTNF (SMD: −0.41; 95% CI: −0.82, −0.00; p = 0.049; I²: 0.0%); multijoint lower-body exercises (SMD: −0.43; 95% CI: −0.93, 0.07; p = 0.089; I²: 45.3%); single-joint lower-body exercises (SMD: −0.34; 95% CI: −0.74, 0.07; p = 0.105; I²: 0.7%); upper-body exercises (SMD: 0.25; 95% CI: −0.30, 0.81; p = 0.367; I²: 76.0%); upper-body exercises (comparator, nonequalized volume) (SMD: −0.30; 95% CI: −0.72, 0.13; p = 0.168; I²: 0.0%); upper-body exercises (comparator, equalized volume) (SMD: 0.57; 95% CI: −0.20, 1.33; p = 0.146; I²: 79.1%); multijoint upper-body exercises (SMD: 0.34; 95% CI: −0.43, 1.10; p = 0.387; I²: 64.1%); single-joint upper-body exercises (SMD: 0.21; 95% CI: −0.61, 1.03; p = 0.615; I²: 83.1%); sport athletes (SMD: 0.08; 95% CI: −0.50, 0.66; p = 0.783; I²: 64.5%); and nonathletes (SMD: −0.16; 95% CI: −0.58, 0.26; p = 0.465; I²: 70.0%).

Muscle Hypertrophy: Meta-Analysis and Subgroup Analysis

Data on muscle hypertrophy were obtained from 4 studies (^35,40,48,53) comprising a total of 189 individuals. Resistance training performed to failure was associated with a greater increase in muscle size compared with RTNF (SMD: 0.75; 95% CI: 0.22, 1.28; p = 0.005; I²: 63.8%) (Figure 3). We did not observe publication bias in this analysis (Egger's regression, p = 0.457) (see Figure S8, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251).

Figure 3.: Forest plot of the muscle hypertrophy promoted by resistance training (RT) performed to concentric failure vs. RT performed not to failure. The estimation for each subgroup (1: nonequalized volume and 2: equalized volume) and the combined effect (overall) are detailed. The squares and error bars signify the standardized mean differences (SMDs) and 95% confidence interval (95% CI) values; the diamonds represent the pooled estimates of random effects meta-analyses.

The results of the subgroup analysis on the influence of RT volume (equalized or not equalized) showed an advantage in favor of the failure approach when considering nonequalized RT volumes (SMD: 0.82; 95% CI: 0.09, 1.56; p = 0.028; I²: 70.7%) (Figure 3). On the contrary, no significant difference was found between RTF and RTNF when considering equalized RT volumes (SMD: 0.59; 95% CI: −0.39, 1.58; p = 0.239; I²: 70.4%) (Figure 3).

Each study included in this analysis was removed individually. Therefore, when we removed 3 of the 4 studies included (^35,48,53), there were changes in these results.

Sensitivity Analyses

Owing to the heterogeneity found between the studies, sensitivity analyses were performed in the interventions for the following (see Figures S9 and S10, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251): lower-body muscles (SMD: 0.34; 95% CI: −0.31, 0.99; p = 0.302; I²: 48.6%) and upper-body muscles (SMD: 1.15; 95% CI: 0.60, 1.70; p = 0.000; I²: 38.3%). Because of the low number of studies (i.e., <3), we did not perform sensitivity analysis considering the subgroups (i.e., equalized and nonequalized RT volumes).

Muscle Power Output: Meta-Analysis and Subgroup Analysis

Data on maximal muscle power output were obtained from 5 studies (^2,8,9,25,26) comprising a total of 150 individuals. No significant effect was found between RTF and RTNF (SMD: −0.20; 95% CI: −0.53, 0.13; p = 0.239; I²: 32.7%) (Figure 4). There was no evidence of publication bias in this analysis (Egger's regression, p = 0.945) (see Figure S11, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251).

Figure 4.: Forest plot of the maximal power output promoted by resistance training (RT) performed to concentric failure vs. RT performed not to failure. The estimation for each subgroup (1: nonequalized volume and 2: equalized volume) and the combined effect (overall) are detailed. The squares and error bars signify the standardized mean differences (SMDs) and 95% confidence interval (95% CI) values; the diamonds represent the pooled estimates of random effects meta-analyses.

The results of the subgroup analysis on the influence of RT volume (equalized or not equalized) showed an advantage in favor of the RTNF approach when considering nonequalized RT volumes (SMD: −0.61; 95% CI: −1.15, −0.08; p = 0.025; I²: 27.5%) (Figure 4). On the contrary, no significant difference was found between failure and nonfailure training when considering equalized RT volumes (SMD: 0.03; 95% CI: −0.31, 0.36; p = 0.881; I²: 0.0%) (Figure 4).

Each study included in this analysis was removed individually. Therefore, when we removed 2 of the 5 studies included (^8,26), there were changes in these results.

Sensitivity Analyses

Owing to the heterogeneity found between studies, sensitivity analyses were performed in the interventions for the following (see Figures S12–S17, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251): lower-body exercises (SMD: −0.23; 95% CI: −0.59, 0.13; p = 0.211; I²: 0.0%); lower-body exercises (comparator, nonequalized volume) (SMD: −0.36; 95% CI: −0.92, 0.20; p = 0.207; I²: 0.0%); lower-body exercises (comparator, equalized volume) (SMD: −0.14; 95% CI: −0.61, 0.33; p = 0.563; I²: 0.0%); upper-body exercises (SMD: −0.16; 95% CI: −0.92, 0.61; p = 0.690; I²: 70.7%); upper-body exercises (comparator, equalized volume) (SMD: 0.19; 95% CI: −0.35, 0.74; p = 0.486; I²: 21.0%); jump power output (SMD: −0.28; 95% CI: −0.69, 0.14; p = 0.188; I²: 0.0%); bench press power output (SMD: 0.19; 95% CI: −0.35, 0.74; p = 0.486; I²: 21.0%); sports athletes (SMD: −0.14; 95% CI: −0.72, 0.44; p = 0.636; I²: 61.0%); and nonathletes (SMD: −0.28; 95% CI: −0.69, 0.14; p = 0.188; I²: 0.0%).

Discussion

This systematic review with meta-analyses summarized the evidence induced by RTF or RTNF on maximal strength, muscle hypertrophy, and maximal power output in adults. Overall analyses indicated no differences between RTF and RTNF with regard to training-induced gains in maximal strength and power output outcomes. When considering studies that consisted of nonequalized volume in RT, we observed an association between RTNF with greater maximal strength and muscle power increase, in comparison with RTF. Furthermore, we observed an association between RTF with greater muscle hypertrophy when compared with RTNF, considering nonequalized RT volumes; however, this difference was not found in the analyses when taking into account only studies that equalized the total training volume.

Some authors have argued that RTF could optimize neuromuscular improvement compared with RTNF based on size principle of motor unit recruitment (^11,54,59). Notwithstanding, only few studies have confirmed the superiority of RTF in isoinertial strength outcomes (^8,47), whereas most studies have not found differences between the RTF and RTNF approaches on strength increase (^{9,13,25,35,53}). In addition, a meta-analysis by Davies et al. (⁶), who subsequently published erratum (⁷), did not confirm differences between RTF and RTNF on strength enhancement.

Our meta-analysis agrees with most of the aforementioned studies because no differences were observed in the strength increase on overall analysis. However, when we examined the subgroup analyses splitting studies that either equalized or did not equalize the total volume, an advantage in favor of RTNF when considering nonequalized RT volumes was observed. One possible explanation could be that RTF induced greater load decrease across the sets due to transient muscle fatigue (^17,18), and a RT load is determinant of strength improvements (^27,51). Among the studies that did not equalize RT volumes, some of them used the same number of sets and lower numbers of repetitions in the nonfailure RT (i.e., a lower RT volume in the nonfailure groups) (^26,35,40,53), whereas 2 of them used 3 sets not to failure compared with one set to failure (i.e., a greater RT volume in the nonfailure groups) (^30,49). Therefore, although RT intensity seems to be a stronger determinant of strength increase than RT volume (^27,51), the influence of RT volume on this outcome cannot be ruled out. In fact, some previous studies have found greater strength improvement after RT regimes composed of more sets (^31,38), although this is controversial (⁴⁵).

A sensitivity analysis provided insights into this comparison (i.e., RTF vs. RTNF): when considering only lower-body exercises, there was a statistically significant greater strength increase in favor of the nonfailure approach in equalized RT volume studies (see Figures S7 and S8, Supplemental Digital Content 1, https://links.lww.com/JSCR/A251). By contrast, no differences were observed between the failure and nonfailure approaches in upper-body strength. Although some physiological responses in different muscle groups could affect adaptation induced by the failure or nonfailure strategies, our synthesis was limited and could not explore such factors in depth. Based on our findings, it seems that there is an influence of the muscle group assessed, and our data suggest that RTNF may be more beneficial to lower-body muscles, whereas no differences were observed for upper-body muscles. It should be mentioned, however, that sensitivity analyses on lower- and upper-body muscles combined different exercises (i.e., squat and leg press or bench row and biceps curl) to achieve at least 3 studies per analysis. Therefore, caution is needed when considering these results because, in some cases, different exercises may provide different stimuli.

Few studies have compared the effects of RTF with RTNF on muscle hypertrophy (^40,48), particularly after controlling for equalization of the RT volume (^35,53). Our synthesis has shown greater muscle size increase after repetitions to failure both in the nonequalized RT volume studies (i.e., a greater RT volume in the RTF groups), as well as in the overall analysis (considering equalized and nonequalized RT volumes). Nevertheless, considering only equalized volume comparisons, no differences were observed between RTF and RTNF interventions. Likewise, Martorelli et al. (³⁵) and Da Silva et al. (⁵³) observed greater muscle thickness increases in RTF compared with RTNF performed with 50% of the possible maximal repetitions, but these differences did not occur when equalizing RT volumes between groups. Indeed, the literature has shown that muscle hypertrophy is highly associated with the total RT volume, which determines the total mechanical overload applied on muscles (^27,51). Therefore, it seems that the advantage of performing repetitions to failure in the muscle hypertrophy is due to a greater training volume rather than a fatigue stimulus. In this way, it does not seem possible to compare the effects of the RTF and RTNF approaches on muscle hypertrophy with no volume equalization.

Our sensitivity analysis indicated that RTF and RTNF induced comparable increases in the lower-body muscles size, whereas RTF has shown superiority in upper-body muscle size gains. Because of the small number of comparisons, we did not perform a sensitivity analysis for lower-body and upper-body muscle splitting for equalized and nonequalized RT volumes in muscle hypertrophy. Thus, based on the influence of the total RT volume in muscle size gains, it is important to carefully consider the influence of repetitions to failure or not to failure in different muscle groups because of the small number of studies assessed in this article.

Muscle power output is associated with high performance in athletes (^22,23,33) and with functional capacity in nonathletes across their life span (^4,5,46). The present meta-analytic results do not show differences in muscle power increase after RTF and RTNF, either with equalized volume or in the overall analysis (considering equalized and nonequalized RT volumes). However, when considering the comparisons for RTF vs. RTNF with no volume compensation (i.e., a lower RT volume in the RTNF groups), there was a difference in favor of RTNF. These results may be explained by the current evidence that shows marked decreases in maximal muscle action velocity and, consequently, muscle power after RTF sets (^17,50). These characteristics (i.e., velocity and power) are determinants of muscle power increases after RT intervention (⁴²). Moreover, it has been shown that RT composed of shorter sets, higher mean velocity, and less fatigue induces greater improvements in muscle power output (⁴³).

However, there is one aspect that may limit the failure vs. nonfailure comparison in the muscle power outcome in the present meta-analysis, i.e., not all interventions were designed as power-type RT and used at the intended maximal velocity across the sets. Therefore, it is possible that even greater differences in favor of RTNF would be observed if the included studies had used maximal velocity in their training groups, as observed in the study by Izquierdo-Gabárren et al. (²⁶) investigating elite rowers. Different to maximal strength, sensitivity analyses did not show distinct patterns of differences between RTF and RTNF in upper-body and lower-body limbs regarding muscle power gains, including in the analysis of only jump power output.

When considering separately the studies that assessed athletes and nonathletes, no differences between RTF and RTNF on maximal strength and power output gains were observed. Among the 4 trials investigating sports athletes, 2 did not show differences between RTF and RTNF in power output increase (^9,25), and one showed greater gains after RTNF (²⁶). These findings might be relevant to highly trained athletes, considering the efficiency of the nonfailure approach, because RTF induces a greater metabolic impact at the cellular level (^17,18), which may result in longer recovery periods between exercise sessions. In addition, it has been suggested that the use of repetitions to failure for long-term training could increase the risk of overuse injuries and overtraining potential (⁶), although no direct evidence has been provided. Because athletes need to combine the RT with their specific sports training, it seems rational to use a low volume of repetitions during RT (i.e., training not to failure), considering that this approach will promote similar or even greater muscle strength and power improvement than training to failure. Analyses of maximal strength and power output when considering nonequalized RT volumes, as well as sensitivity analysis of lower-body strength, indicated that RTNF could be particularly beneficial for athletes who experience high demands of specific training and competition (⁴¹).

If we consider improvement in neuromuscular function and promotion of health, the same premise is useful for nonathletes because our findings suggest that RTNF can optimize strength enhancements and, at the same time, reduce high levels of discomfort and physical exertion, preventing the incorrect execution of the movement (⁵⁹). Moreover, RTF also implies a longer time under tension, leading to greater increases in blood pressure, heart rate, and rate-pressure product (^16,34,39), which may increase the risk of complications in populations at higher cardiovascular risk.

There are a number of limitations that should be acknowledged. The risk of biases of the included studies was high or unclear for assessed aspects in most of the studies. In addition, we observed a high heterogeneity in the maximal strength and muscle hypertrophy outcomes, and this heterogeneity remained moderate after sensitivity analyses for muscle hypertrophy and was low to high for maximal strength analyses. Moreover, although the RTF approach can be similar among studies, the same is not observed among RTNF approaches, thereby leading to possible influence in comparability between trials. For example, in the study by Nóbrega et al. (⁴⁰), the RTNF group performed a volitional number of repetitions (before failure), whereas in other studies, the number of repetitions was controlled and ranged from 50 to 70% of the maximal repetitions possible (^{2,25,26,35,53}), and these differences can confound the results. Likewise, to control the influence of RT volume on neuromuscular adaptations, some trials created a nonfailure group after equalizing the RT volume doubling the sets. Although it addresses an interesting aspect (i.e., RT volume influence), it may also generate others biases. One of them is creating RT groups with an unrealistic number of sets and training session durations. Another problem is adding a third training group in the comparison, thereby possibly increasing the probability of type II error. Furthermore, interventions ranged from 6 to 12 weeks; thus, the present findings cannot be extrapolated to long-term adaptations. In the study by Izquierdo et al. (²⁵), the total duration of the intervention was 16 weeks, but the last 4 weeks of the intervention were similar between RTF and RTNF (i.e., tapering time). Therefore, we included only the results concerning the first 11 weeks. Interestingly, differences in muscle power output in favor of the RTNF approach arose after the tapering period. The longer intervention study (i.e., 20 weeks) found in our literature search (⁵⁷) was excluded because the data were from the same research as other included study (⁵³). In this study, no differences in muscle strength and muscle size between the failure and nonfailure interventions were observed after 20 weeks, regardless of the RT volume (⁵⁷). The low number of included studies in the muscle hypertrophy outcome is another limitation that should be mentioned. Another characteristic of this study that is worthy of mentioning is the training status of the study subjects. Except for studies assessing athletes, most of the studies addressed untrained subjects, and only one study assessed recreationally resistance-trained individuals. For muscle hypertrophy, e.g., all included studies enrolled untrained subjects. Therefore, care should be taken when extrapolating our findings to well–resistance-trained individuals. Finally, some included studies assessed muscle power output through vertical jump tests, and there is a substantial problem with assessing power through vertical jump because simply changing the jump strategy may induce changes in muscle power (³⁷). It is important to note that none of the aforementioned limitations could be controlled by the authors of the current study.

This study also has its strengths. We implemented a rigorous process to adhere to recommended practices in systematic reviews, which included duplicate and independent procedures in all review stages (except meta-analyses) and taking extra care to avoid duplicating data generated by common study samples (a unit-of-analysis error). Besides an update in the maximal dynamic strength outcomes, this is the first meta-analysis comparing the effects of RTF with RTNF on muscle hypertrophy and muscle power output. In addition, the consideration of RT volume (i.e., equalized and nonequalized) allowed us to summarize the influence of RT volume on the adaptations because it is not feasible to determine possible differences between RTF and RTNF without considering the influence of RT volume. Moreover, many sensitivity analyses allowed us to assess some potential confounders, such as training status (i.e., athletes and nonathletes) and different muscle groups (i.e., upper-body and lower-body muscles).

In summary, RT performed until concentric failure did not provide additional gains in maximal dynamic strength and maximal power output, regardless of the training volume (i.e., equalized and nonequalized). In addition, there were significant greater muscle strength and muscle power improvements in the RTNF approach considering only nonequalized volume studies. Moreover, when considering studies that equalized the training volume, no difference between RTF and RTNF was observed on muscle hypertrophy. Furthermore, sensitivity analyses indicated that RTF and RTNF induce similar strength and power gains in athletes and nonathletes individuals, regardless of the RT volume. In addition, our findings suggest that different muscle groups (i.e., upper-body and lower-body muscles) respond differently to RTF and RTNF, and this should be taken into consideration when prescribing RT.

Practical Applications

Strength and conditioning coaches should be aware that RTF and RTNF induce comparable improvements in maximal strength and muscle power output in athletes and nonathletes, and even a small advantage of RTNF can be observed in strength increase, particularly when considering lower-body muscles. Regarding muscle hypertrophy, it seems that RTF promotes greater muscle size gains, but these gains are related to a greater RT volume rather than any influence of transient induced fatigue across the sets, as can be observed in the absence of differences between RTF and RTNF when considering equalized RT volumes. Therefore, although the RTF approach is widely used, its use should consider the risks. In athletes who combine RT with their specific sports trainings, it seems rational to use RTNF, considering that this approach will promote similar, or even greater (i.e., considering overall analyses in nonequalized RT volumes), muscle strength and power gains compared with RTF. Along with a possible advantage in power and strength development, RTNF may allow for a faster recovery after RT sessions and even prevent overuse injuries and overtraining potential in the long term. Considering nonathletes individuals, the approach should consider cardiovascular and injuries risks, time to train, and personal preference (i.e., some people may be more susceptible to discomfort associated with muscle fatigue and can even worsen their exercise execution). Finally, RTF and RTNF adaptations may be different after upper-body and lower-body exercises, and this should also be considered when prescribing RT.