REF-1181 · Martins et al., 2026 · PLOS One · Paper · PubMed · Free full text

Will slow bodyweight strength work make a runner faster?

In 34 soldier-runners, 4 wk of 84-s sets (7-s reps) raised knee torque and jump; 3000 m gains (−2.8% vs −0.8% in controls) were not significant. Short, underpowered; no economy data.

B+moderate evidenceMeasured on recreational to trained athletes

Martins, Souza and colleagues matched 36 soldier-runners into triplets by age and weekly mileage and randomised each triplet to predominantly dynamic strength training (PDST), predominantly isometric strength training (PIST) or no strength training (CON); 34 finished.

Both training groups did 8 sessions over 4 weeks of bodyweight squats, unilateral lunges, supine hip lifts and plantar flexions, 4 sets of 12 repetitions of 7 s each, so every set lasted 84 s. PDST used 3 s up, 1 s hold, 3 s down; PIST used 1 s down, 5 s hold, 1 s up.

Knee-extension peak torque rose 13.3% (PDST) and 14.2% (PIST) vs 2.5% in CON; only PIST differed significantly from CON (P = 0.034). Countermovement jump rose 5.4% and 4.1% vs 0.7%. The 3000 m time trial improved in every group, including controls (time effect P = 0.002), with no group difference (interaction P = 0.421; −2.8%, −2.6% and −0.8%).

Ground contact time and vertical oscillation, which the authors use as running-economy proxies, did not differ between groups. The authors conclude that high-TUT training improved neuromuscular characteristics without significant changes in running performance or running economy over 4 weeks.

How this bears on the Russian static-dynamic method (Seluyanov): the closest tested form in endurance athletes, but still indirect.

The method's claims (practice descriptions, not this paper) are that slow, continuous-tension sets of roughly 30-60 s at a light load, taken to a burning sensation with short rests, grow type I fibres and their mitochondria and so improve endurance. PDST shares the slow tempo and long sets (84 s) with light (bodyweight) resistance, in runners.

It differs in rest (90 s between sets), in that failure, burning and the prevention of relaxation at the end of each repetition were not specified, in having no load progression, and in lasting only 4 weeks. No fibre type, mitochondrial or capillary measure was taken, and running economy itself was not measured.

The endurance outcome it did test, the 3000 m time trial, did not improve significantly more than in controls (−2.8% vs −0.8%; the trial was underpowered for it). It neither confirms nor refutes the method's endurance claim; it shows that 4 weeks of this kind of work did not produce a detectable performance change.

Protocol

Participants: 40 recruited, 36 randomised, 34 analysed (PDST 11, PIST 12, CON 11); men aged about 19; running 12-22 km a week, at least 3 runs a week, 3000 m pace 4:00-5:30 min/km, at least 6 months of distance running; one weekly session of military cross-training including strength work; "not highly trained in strength".
Allocation: triplets matched on age and weekly mileage by Euclidean-distance clustering, then randomised within triplets by script; participants blinded to group, trainers and assessors not.
Exercises: bodyweight squat, unilateral lunge (2 sets per leg instead of 4 because of fatigue), supine hip lift, plantar flexion.
Load: bodyweight only; no progression.
PDST tempo: 3 s concentric, 1 s isometric, 3 s eccentric, 0 s (3/1/3/0); 12 repetitions = 84 s per set. Relaxation between repetitions not addressed; range of motion not stated as partial.
PIST tempo: 1 s eccentric, 5 s isometric, 1 s concentric (1/5/1/0), squats and lunges held at 100-110° knee flexion; 12 repetitions = 84 s.
Volume and rest: 4 sets per exercise; 90 s between sets, 120 s between exercises; 2 sessions a week, 48 h apart, for 4 weeks (8 sessions); metronome, supervised group sessions.
Failure: not specified in the protocol (the Discussion refers to "execution to near failure" without measuring it).
Running: usual running continued, self-reported; CON did no strength training.
Outcomes: 3000 m track time trial (primary for sample size), session RPE × time (internal running load), knee-extensor isometric peak torque at 70° knee flexion (load cell), countermovement jump, jump loss after the time trial (neuromuscular fatigue), ground contact time and vertical oscillation (Garmin HRM-Run); post-tests 72-96 h after the last session.

Key Results

OutcomePDST (n = 11)PIST (n = 12)CON (n = 11)Time × group
3000 m time, pre → post808.0 → 786.0 s832.6 → 810.8 s818.0 → 810.3 sP = 0.421, η²p 0.05
3000 m change−2.8 ± 4.3%−2.6 ± 3.8%−0.8 ± 1.8%vs CON P = 0.291 and 0.355
Peak torque, pre → post222.0 → 245.3 Nm225.9 → 255.6 Nm232.4 → 237.3 NmP = 0.007, η²p 0.28
Peak torque change+13.3 ± 12.4%+14.2 ± 10.3%+2.5 ± 7.9%vs CON P = 0.063 (d 1.04) and 0.034 (d 1.12)
Countermovement jump, pre → post34.4 → 36.2 cm34.1 → 35.4 cm33.8 → 34.0 cmP = 0.001, η²p 0.36
Countermovement jump change+5.4 ± 3.4%+4.1 ± 3.1%+0.7 ± 2.5%vs CON P = 0.003 (d 1.54) and 0.034 (abstract 0.030; d 1.13)
Ground contact time239.2 → 233.2 ms244.5 → 236.9 ms241.3 → 238.8 msP = 0.215
Vertical oscillation10.6 → 10.1 cm10.5 → 9.8 cm10.7 → 10.4 cmP = 0.429
Internal running load (RPE × min)115.3 → 108.7120.3 → 112.7119.0 → 115.6P = 0.592
Correlation, peak-torque change vs 3000 m change (training groups, n = 23)r = −0.42 (95% CI −0.78 to −0.05), P = 0.049, adjusted for mileage
Weekly mileage had a strong effect in the time-trial model (ANCOVA F(1,30) = 21.19, P < 0.001, η² 0.39); the authors say it "may have influenced the results", so self-reported running volume is a major confounder of the time-trial comparison.
The paper's text says post hoc differences were "between the PDST and PIST groups"; the percentage-change analysis that follows shows differences of each training group vs CON. This looks like a reporting slip.

Mechanisms

Measured: isometric knee torque and jump height rose; running time, ground contact time, vertical oscillation and perceived effort did not change differently from controls.
Authors' interpretation: early gains are neural (recruitment of higher-threshold motor units under metabolic fatigue, coordination); the more specific gain in isometric torque in PIST reflects the match between holds and the isometric test.
Authors' interpretation: 4 weeks was chosen to minimise peripheral aerobic adaptations such as angiogenesis and mitochondrial biogenesis, and was likely too short for running performance to change.
Not measured: fibre type, mitochondria, capillaries, lactate, oxygen cost of running (running economy), tendon stiffness.
Inference: the moderate correlation between torque gain and time-trial gain is within-group and observational; it does not show the training made anyone faster.

Practical Application

Will slow bodyweight strength work make a runner faster? In 34 soldier-runners, 4 wk of 84-s sets (7-s reps) raised knee torque and jump; 3000 m gains (−2.8% vs −0.8% in controls) were not significant. Short, underpowered; no economy data.
As an entry block for runners new to strength work, slow bodyweight sets are a low-equipment way to gain some strength and jump height; the authors suggest it as preparation for heavier or plyometric work, which is where the evidence for running economy and time-trial gains lies (SYNTH-44).
Static-dynamic method: the closest tested form in endurance athletes, and still indirect. Slow, light, long sets in runners improved strength and jump, but 3000 m time did not improve significantly more than in controls in 4 weeks; nothing about type I fibres, mitochondria or capillaries was measured, and failure, burning and continuous tension were not specified. It does not support adding this work to an endurance plan for an aerobic benefit.
For a runner's strength slot, the base's default is still heavy, few repetitions, short of failure, twice a week for at least eight weeks ([SYNTH-44]); the 4-week length and the bodyweight-only load here are why this trial cannot challenge that.

Limitations

Four weeks (8 sessions); bodyweight only, no progression; the authors call the short duration an important limitation.
Sample size was planned on η²p 0.47 for a running outcome from another study; the observed interaction was 0.05, so the trial was underpowered for the time trial.
Running economy was not measured; ground contact time and vertical oscillation were used as indirect proxies, and vertical oscillation was not pre-registered.
Trainers and assessors were not blinded; running volume during the trial was self-reported, and mileage strongly affected the time-trial change.
Men only, recreational level, little strength-training background; the isometric peak torque test (70°) favours the isometric group, and strength was measured in one muscle group only.
Reporting slips: the abstract gives P = 0.030 for the isometric group's jump change, the results P = 0.034; the post hoc wording names PDST vs PIST where the data show each group vs CON.
Funding: GPEPS study group, FAPITEC/SE (Sergipe state research foundation), Prince Sultan University, and Princess Nourah bint Abdulrahman University (PNURSP2026R424). Conflicts of interest: none declared (PubMed record).
How this entry was checked and changed
Why this level

parallel RCT with n=34 (11-12 per arm), participants blinded but trainers and assessors not; only 4 weeks, sample size planned on an optimistic effect (η²p 0.47), and running economy was not measured despite the abstract's wording; read from the full text

Full text (checked 2026-10-16)
Forty physically active young male runners, Brazilian Army soldiers, weekly running routine of 12–22 km; final sample PDST (n = 11), PIST (n = 12), CON (n = 11) (full text).
PDST 3/1/3/0 cadence; PIST 1/5/1/0 cadence; 12 repetitions, 7 seconds per repetition, 84 seconds of TUT per set (full text).
Four sets per exercise; 90-second rest between sets, 120 seconds between exercises; twice a week, 48-hour interval, eight sessions (full text).
Progressive overload was not included (full text).
3000 m time PDST 808.0 ± 76.9 to 786.0 ± 73.5 s; PIST 832.6 ± 64.9 to 810.8 ± 73.7 s; CON 818.0 ± 47.1 to 810.3 ± 58.0 s (full text).
Time × group interaction: TT p 0.421, η2p 0.05; PT p 0.007, η2p 0.28; CMJ p 0.001, η2p 0.36 (full text).
ΔTT PDST: −2.8 ± 4.3%, PIST: −2.6 ± 3.8%, CON: −0.8 ± 1.8% (p = 0.291; p = 0.355) (full text).
ΔPT 13.3 ± 12.4%, 14.2 ± 10.3%, 2.5 ± 7.9%; p = 0.063; d = 1.04 and p = 0.034; d = 1.12 (full text).
ΔCMJ 5.4 ± 3.4%, 4.1 ± 3.1%, 0.7 ± 2.5%; p = 0.003; d = 1.54 and p = 0.034; d = 1.13 (full text).
Weekly running mileage effect on ΔTT: F(1,30) = 21.19; p < 0.001; η² = 0.39 (full text).
ΔPT and ΔTT r = −0.42 (95% CI = −0.78; −0.05); p = 0.049 (full text).
Sample size from η²p of 0.47 (Lum et al.); 36 participants for power 0.80 (full text).
Funding: FAPITEC/SE, Prince Sultan University, PNURSP2026R424 (full text).