REF-1623 · Ralston, Kilgore, Wyatt & Baker, 2017 · Sports Medicine · Paper · PubMed · Free full text

More weekly sets per exercise gave more strength, but the extra was modest: pooling 61 training groups from 9 studies, high weekly set counts beat low ones by an effect size of 0.18 (95% CI 0.06-0.30) and medium beat low by 0.15 (0.01-0.30), while low-set training on its own still raised strength by an effect size of 0.82 (0.47-1.17); for exercise-specific 1RM the high-versus-low gap was 0.14 and not significant (p = 0.06), and the abstract does not say how many sets low, medium and high mean

Main caveat: Abstract only; no risk-of-bias or quality assessment is reported in it, and heterogeneity values are not given.
moderate evidenceMeasured on non-athletes to trained athletesRead from the abstract only

Ralston and colleagues pooled training studies that compared different numbers of weekly sets per exercise and asked how much strength each dose bought. From 6962 records found by September 2016, nine studies met their deliberately strict criteria, giving 61 treatment groups.

Weekly sets were grouped as low, medium or high, and outcomes were analysed for all exercises together, for multi-joint and isolation exercises separately, and for 1RM on the trained exercise. High weekly sets came out ahead of low in every analysis, by effect sizes of 0.14 to 0.23 (significant in all but the 1RM analysis, p = 0.06), and medium beat low by 0.15.

Low weekly sets still improved strength, with pooled pre-to-post effect sizes of 0.80 to 0.95. The authors read this as further evidence that strength rises step by step with weekly sets: medium and high doses outperformed low, and low gave the smallest change. They suggest that for well-trained people either medium or high weekly sets may be a suitable dose.

Participants

61 treatment groups from 9 studies. Men by the PubMed indexing (MeSH: Male; age headings Adolescent, Young Adult, Adult and Middle Aged); the conclusion refers to novice, intermediate and well-trained male trainees. Not in the abstract: the total number of participants, their ages, and how many were trained or untrained.

Protocol

Search: five databases with terms for strength training, resistance training, single sets, multiple sets and volume, to September 2016; 6962 potentially relevant studies, 9 eligible under pre-set criteria.
Exposure: low (LWS), medium (MWS) or high (HWS) weekly sets per exercise. The set ranges for each category: not in the abstract.
Outcomes: pre-to-post strength per exercise; multi-joint and isolation exercises separately; exercise-specific 1RM; trained versus untrained.
Analysis: random-effects pooling; Cochran Q and I² for heterogeneity; sensitivity analyses by subgroup.
Not in the abstract: risk-of-bias or quality assessment, heterogeneity results, study durations, loads, weekly frequency, and any comparison of high with medium weekly sets.
Funding and interests: no funding; the authors declare no relevant conflicts.

Key Results

Comparison (outcome)Difference, effect size (95% CI); pPre-to-post effect size in each group (95% CI)
High vs low (all exercises)0.18 (0.06 to 0.30); p = 0.003low 0.82 (0.47-1.17); high 1.01 (0.70-1.32)
Medium vs low (all exercises)0.15 (0.01 to 0.30); p = 0.04low 0.83 (0.53-1.13); medium 0.98 (0.62-1.34)
High vs low (multi-joint)0.18 (0.01 to 0.34); p = 0.04low 0.81 (0.65-0.97); high 1.00 (0.77-1.23)
High vs low (isolation)0.23 (0.06 to 0.40); p = 0.008low 0.95 (0.30-1.60); high 1.10 (0.26-1.94)
High vs low (exercise-specific 1RM)0.14 (−0.01 to 0.29); p = 0.06low 0.80 (0.47-1.13); high 0.97 (0.68-1.26)
Low sets worked: every low-set estimate is well above zero (0.80 to 0.95). The advantage of more sets is about one-sixth to one-quarter of the gain low-set training produced on its own (our arithmetic, 0.14/0.80 to 0.23/0.95, a rough comparison of standardised effects).
1RM: for the lift that was trained, the high-versus-low interval crosses zero; the authors call the gain marginally greater.
Isolation exercises: the group intervals are wide (high 0.26 to 1.94).
High versus medium: not reported in the abstract, so it cannot say whether the highest dose beats the middle one.
The authors' wording: for novice and intermediate men they write that "LWSs do not lead to strength gains compared with MWS or HWS". Our reading: the low-set groups did gain (effect size about 0.8); the phrase means smaller gains than with more sets, not no gain.

Practical Application

A low weekly set count per exercise already builds strength; more sets add a modest extra. Across 9 studies, low weekly sets raised strength by an effect size of about 0.8 and high weekly sets by about 1.0; the gap (0.18) was real but small, and on the 1RM of the trained lift it was not significant (0.14, p = 0.06).
Moving off the lowest dose is supported; chasing the highest is not shown. Medium already beat low (0.15), and the abstract gives no high-versus-medium comparison. A later model in the base finds diminishing returns for strength as sets rise (REF-1477).
Inference, not tested here: for an endurance athlete short of time, a low set count is a valid floor, not a failure. Low-set groups reached about four-fifths of the high-set pre-to-post effect (0.82 vs 1.01, our arithmetic, a rough comparison of standardised effects); adding sets is the lever to pull when strength becomes the priority and recovery allows it.
The categories are per exercise and their limits are not in the abstract. Not from this paper: the base's cycling guide describes the trials' shape as 3-4 sets of low reps, 1-3 times a week (SYNTH-43), which is 3-12 weekly sets per lift (our arithmetic); where that sits among this paper's categories cannot be told from the abstract.ACSM's 2026 position stand, read in full, gives 2-3 sets per session for strength and at least 10 weekly sets for muscle size (REF-1476).
Elsewhere in the base. One hard set 1-3 times a week raised 1RM in trained men (REF-586), while in an earlier meta-regression 2-3 sets per exercise gave 46% greater strength gains than one (REF-1626).

Limitations

Abstract only; no risk-of-bias or quality assessment is reported in it, and heterogeneity values are not given.
Nine studies, mostly or all men by the PubMed indexing; women are not represented in the indexing.
The set ranges behind low, medium and high are not in the abstract, so no weekly set number can be taken from this entry.
No high-versus-medium comparison in the abstract.
Pre-to-post effect sizes within groups are not controlled effects; only the between-category differences compare doses.
The search ended in September 2016; the "restrictive" inclusion criteria are not described in the abstract.
How this entry was checked and changed
Why this level

risk of bias not read (abstract only): a meta-analysis of 9 studies whose abstract reports no risk-of-bias or quality assessment; heterogeneity was tested but its values are not given, and neither are the set counts behind the low, medium and high categories

Content correction 2026-11-06
2026-11-06: added from the abstract and the PubMed record; level B+ (meta-analysis of 9 studies; risk of bias not read, abstract only).