REF-1182 · Kordi et al., 2020 · Scandinavian Journal of Medicine & Science in Sports · Paper · PubMed
Do on-bike isometric pushes at crank angles raise sprint power?
In 24 elite track sprinters, 6 weeks of 3-s max pushes at 3 angles instead of the heavy lift: W/kg +4.3% vs −0.5%; watts +2.9%, not above controls. Not randomised.
Kordi et al. asked whether elite sprint cyclists gain peak power when their main heavy gym lift is replaced by maximal isometric efforts on the bike.
Twenty-four track sprinters (17 men, 7 women; including tandem pilots and visually impaired stokers) were split by their head coach, not at random, into a "best practice" control group (CON, n = 11, 9 men and 2 women, mostly podium-level) and an experimental group (EXP, n = 13, 8 men and 5 women).
For 6 weeks of an off-season maximal-strength phase both groups kept the same track and road sessions. CON did its usual heavy gym work (3-5 sets of 3-5 reps at 85-95% 1RM of a squat or deadlift, then a second multi-joint lift and unilateral exercises).
EXP instead did 3-s maximal pushes on an ergometer whose flywheel was clamped still, at crank angles of 45, 90 and 135 degrees from top dead centre with each leg leading, building from 1 × 3 to 4 × 4 per angle and lead leg; afterwards they did 3 × 5 back squats at about 70-75% of predicted 1RM.
Relative peak power rose 0.8 ± 0.7 W/kg (18.7 → 19.5 W/kg, +4.3%) in EXP and did not change in CON (−0.1 ± 1.0 W/kg), a group × time interaction of p = 0.02. Absolute peak power rose 46 ± 62 W (+2.9%, p = 0.05) in EXP, but the interaction was not significant (p = 0.14).
The cycling-specific isometric torque rose 12.5% in EXP and not in CON (interaction p = 0.002); knee-extensor strength, voluntary activation, lean mass and muscle architecture showed no between-group differences.
Our reading: a plausible first signal that maximal on-bike isometric work can stand in for the heaviest lift in a sprinter's strength phase without losing power and may add a little, but the groups were chosen by the coach, absolute power did not beat the controls, and part of the relative gain reflects a 1-kg drop in body mass.
Protocol
Participants: 24 track sprint cyclists (17 men, 7 women), 23 ± 3 y, 80.8 ± 10.9 kg, 172.0 ± 9.4 cm. Solo men (n = 13) 200 m best 9.6-10.8 s, solo women (n = 5) 10.9-12.0 s; two men's tandems and one women's tandem (n = 6), stokers visually impaired; medallists at Olympic, Paralympic and World level; all with at least 2 years of resistance training.
Allocation: "Full random allocation of participants was not possible"; the head coach assigned riders on availability, training schedule, location (the ergometer was at one central venue) and perceived willingness to try the method.CON n = 11 (9 men, 2 women), "largely comprised of podium-level" sprinters; EXP n = 13 (8 men, 5 women), under-23 programme, national and podium-level riders. Baseline peak power did not differ (1537 ± 307 vs 1541 ± 389 W; 18.7 ± 2.5 vs 19.0 ± 3.5 W/kg). No blinding; menstrual phase not controlled; no sample-size calculation or registration reported.
Timing: out of the competitive season, in a designated "maximum strength" phase, after 1-2 weeks off and a 2-week return to full training. Tests 36 h after the last session, same time of day (± 1 h) and weekday before and after.
Training common to both groups, per week: 2 track sessions (a "high torque" day of stationary or slow-moving maximal efforts, 3-12 reps of 6-20 s; a "high power" day of 3-5 reps of 10-35 s from higher cadences), 2-3 gym sessions, 1-2 road rides of 60-90 min at RPE 2-4 of 10. Training was identical except the gym sessions.
CON gym sessions (supervised by the riders' strength coach and the lead researcher): a bilateral compound lift (back squat, front squat or deadlift by preference), 3-5 sets × 3-5 reps at 85-95% 1RM, 2-3-s lowering, maximal concentric intent, 3-5 min between sets, not to failure; then 3 sets of a second multi-joint exercise at similar loads (cleans, barbell jumps, dumbbell lunges, or single- or double-leg press); then unilateral knee extensions, hamstring curls and calf raises (6-12 reps, about 70-90% of predicted 1RM); then trunk work.Exact loading was individualised within these ranges.
EXP gym sessions (supervised by the lead researcher): maximal cycling-specific isometric efforts on a custom ergometer set to each rider's track position, with a car-jack clamp and rubber stopper pressed against the flywheel so the cranks could not turn, instrumented cranks and real-time torque feedback.
Crank angles: 45, 90 and 135 degrees from top dead centre, chosen as the angles of highest torque in the pedal stroke; each angle with the right leg leading and with the left leg leading (for example, all sets right-leg-leading at 45 degrees, then all sets left-leg-leading at 45 degrees); order of angles randomised each visit.
Each effort maximal, 3 s long; 60 s passive rest between reps, 2 min between sets and between crank positions.
Sets × reps at each angle: week 1 1 × 3, week 2 2 × 4, week 3 3 × 3, week 4 4 × 3, week 5 4 × 3 and 4 × 4, week 6 4 × 4. The paper's example (3 sets with the right leg leading at 45 degrees, then 3 sets with the left leg leading at 45 degrees) implies the count applies to each lead leg at each angle.
Then 3 × 5 back squats at about 70-75% of predicted 1RM, maximal concentric intent, 3 min between sets, agreed with the coaches so the riders would not regress if the new method failed; then the same trunk exercises as CON.As described in the Methods, EXP sessions did not include CON's second multi-joint lift or the unilateral exercises; the abstract ("training content was identical except participants in the EXP group replaced their usual compound lower body resistance training exercise") and the Design say only the main lift differed, so this is unclear.
What "replaced" means here: EXP gave up the heavy (85-95% 1RM) compound lift and kept a lighter, fast squat; it was a replacement of the heaviest loading, not an add-on to the full CON session and not a removal of all lifting.
Outcomes, in order: DXA body composition; ultrasound of vastus lateralis (pennation angle, thickness; both legs, 6 images); isometric knee-extensor maximal voluntary force (hip about 125 degrees, knee about 115 degrees), voluntary activation by femoral-nerve stimulation and rate of torque development (torque at 50-200 ms); cycling-specific isometric torque (ISO-CYC: cranks fixed at 90 degrees, seated, 3 maximal efforts of 3-5 s per side); isovelocity sprint test (seated, 4-s maximal efforts at 60, 115, 125, 135 and 180 rpm, 3 min apart), with peak power from a quadratic power-cadence fit and maximal torque and cadence extrapolated from a linear torque-cadence fit.
Reliability (the lab's earlier data): peak power CV 2.7% (ICC 0.96), extrapolated maximal torque CV 3.6%, ISO-CYC CV 3.9%, knee-extensor force CV 4.1%.
Analysis: 2 × 2 mixed ANOVA (group × time) with Bonferroni-adjusted follow-ups, Cohen's d, Pearson correlations of percentage changes across all 24 riders; significance p ≤ 0.05.
Funding: a collaborative studentship between Northumbria University and the English Institute of Sport (Acknowledgements). No conflict-of-interest statement appears in the published PDF.
Key Results
| Outcome | EXP (n = 13) | CON (n = 11) | Group × time (full text) |
|---|---|---|---|
| Relative peak power (W/kg) | 18.7 ± 2.5 → 19.5 ± 2.3 (+0.8 ± 0.7 W/kg, +4.3% in Table 1, p = 0.004) | 19.0 ± 3.5 → 18.9 ± 3.1 (−0.1 ± 1.0 W/kg) | p = 0.02 |
| Absolute peak power (W) | 1537 ± 307 → 1581 ± 287 (+46 ± 62 W, 3 ± 4%, d = 0.17, p = 0.05; the abstract says P < .05) | 1541 ± 389 → 1536 ± 366 (−5 ± 98 W, p = 0.844) | p = 0.14 (not significant) |
| Individual relative change | 10 of 13 above +2%; mean 4 ± 5%, range −1% to 16%, d = 0.33 (small) | 2 riders above +2%; mean 0 ± 5%, range −8% to 7%, d = −0.05 | — |
| Extrapolated maximal torque (Nm) | 207 ± 32 → 214 ± 32 (+7.1 ± 6.5 Nm, p = 0.007) | 194 ± 34 → 196 ± 34 (+2.4 ± 9.7 Nm) | p = 0.14 (not significant) |
| Extrapolated maximal cadence; optimal cadence | 276 → 277 rpm (p = 0.70); 138 → 138 rpm (p = 0.99) | 289 → 284 rpm (p = 0.36); 144 → 142 rpm (p = 0.27) | no significant interactions |
| Cycling-specific isometric torque, ISO-CYC (Nm) | 400 ± 78 → 450 ± 113 (+12.5%, p = 0.001) | 383 ± 98 → 367 ± 82 (−4.2%, p = 0.27) | p = 0.002 |
| Knee-extensor isometric force, MVF (Nm) | 309 ± 75 → 323 ± 65 (+4.5%) | 296 ± 44 → 296 ± 44 (0%) | no difference reported in either group |
| Voluntary activation (%) | 97.1 → 97.3 | 97.6 → 97.5 | no difference |
| Knee-extensor torque at 50 and 100 ms | unchanged | unchanged | — |
| Knee-extensor torque at 150 and 200 ms (Nm) | 227 → 240 (p = 0.03); 241 → 260 (p = 0.003) | 226 → 227; 242 → 244 | p = 0.18 and p = 0.054 (not significant) |
| Vastus lateralis pennation angle (degrees) | 17.1 → 18.0 (+5.2%) | 17.1 → 18.2 (+6.4%) | time effect p < 0.001 in both; interaction p = 0.06 |
| Vastus lateralis thickness (mm) | 22.4 → 23.0 (+2.6%) | 21.9 → 23.3 (+6.3%) | time effect p = 0.034 in both; interaction p = 0.66 |
| Body mass (kg) | 82.1 ± 13.1 → 81.1 ± 12.0 (−1.2%) | 80.2 ± 8.3 → 80.6 ± 8.0 (+0.5%) | no pre-post difference in either group |
| Lean mass, lower-body lean mass, body fat | no change | no change | — |
| Correlation with % change in peak power (all 24) | torque at 150 ms r = .42 (p = .04), at 200 ms r = .47 (p = .02); ISO-CYC r = .22 (p = .30); knee-extensor MVF r = .36 (p = .08) |
Main result: relative peak power rose 4.3% with on-bike isometrics vs −0.5% with usual heavy lifting (interaction p = 0.02); absolute peak power rose 2.9% vs −0.3% but did not differ between groups (p = 0.14); the trained task (ISO-CYC) rose 12.5% vs −4.2% (p = 0.002).
Only two interactions were significant: relative peak power (p = 0.02) and ISO-CYC (p = 0.002). Absolute peak power and extrapolated maximal torque (both p = 0.14), late rate of torque development (p = 0.18, p = 0.054), knee-extensor force, voluntary activation, pennation angle (p = 0.06), muscle thickness (p = 0.66) and body composition showed no between-group differences.The abstract's "There were no other between-group differences" is accurate; the Discussion's "improved cycling peak power output" and "mediated by" maximal torque go further than these statistics.
Body mass carries part of the relative result (our calculation from Table 1). Absolute peak power rose 2.9% and relative 4.3% in EXP because body mass fell about 1.0 kg (−1.2%) while CON gained about 0.4 kg; neither change was significant within its group, but the two together widen the gap in W/kg.
Size of the change against test noise (our reading). The mean absolute gain (3 ± 4%) is about the size of the lab's peak-power CV (2.7%), so for an individual rider a 2-3% change after one test is hard to separate from noise.
Mechanisms
Measured: the power gain came with a rise in extrapolated maximal torque within EXP (+7.1 Nm, p = 0.007) and no change in maximal or optimal cadence, so within EXP the torque-cadence line moved up at the low-cadence, high-force end; the torque change did not differ from CON (p = 0.14).
Measured: strength rose in the trained task (ISO-CYC +12.5%, interaction p = 0.002) but not significantly in a different task, the seated knee-extensor test (+4.5% vs 0%), and voluntary activation did not change. This is task and position specificity, in line with REF-1159.
Measured: within EXP, knee-extensor torque at 150 and 200 ms rose (p = 0.03, p = 0.003) but torque at 50 and 100 ms did not; the interactions were not significant. Across all 24 riders, the late-torque change was the only measure that correlated with the change in peak power (r = .42 and .47, 18-22% of the variance); the ISO-CYC change did not (r = .22, p = .30).
Measured: pennation angle and muscle thickness rose in both groups with no between-group difference; lean mass did not change. No structural change explains the group difference.
Authors' interpretation: maximal isometric efforts on the bike let a sprinter repeat forces in race positions that are unreachable or unsustainable while pedalling, an overload that the authors say raised peak power "by improving the maximum torque". No mediation analysis was run and the torque interaction was not significant, so "mediated" is the authors' reading.They also read CON's thickness gain as a stronger hypertrophy stimulus from heavy lifting; the thickness interaction was p = 0.66.
Inference: a rise in late but not early rate of torque development, with unchanged activation, is what sustained maximal contractions would be expected to produce. This agrees with REF-1161, where sustained 3-s holds raised maximal strength but only explosive contractions raised early force.
Inference: EXP had less to lose and more to gain than the "podium-level" CON (more under-23 and national riders, more women); a lower starting level and coach choice of who tried the method can produce part of a between-group difference on their own.
Practical Application
Do on-bike isometric pushes at crank angles raise sprint power? In 24 elite track sprinters, 6 weeks of 3-s max pushes at 3 angles instead of the heavy lift: W/kg +4.3% vs −0.5%; watts +2.9%, not above controls. Not randomised.
For a track sprinter in a maximal-strength phase: the tested dose is 3-s maximal efforts at 45, 90 and 135 degrees from top dead centre, each leg leading, 60 s between efforts and 2 min between sets, building from 1 × 3 to 4 × 4 per angle and lead leg over 6 weeks, 2-3 gym days a week, followed by 3 × 5 fast squats at about 70-75% 1RM.It held or slightly raised peak power where the heaviest lift was dropped; it did not beat heavy lifting on absolute power (B, not randomised).
Do not read it as "drop the gym". EXP kept a lighter fast squat and the shared track "high torque" sessions; the trial is a 6-week swap inside an off-season block, not a season without lifting. Heavy lifting stays the cyclist's default in the base (SYNTH-43); isometric work stays a tool for a named goal (SYNTH-88).
TRANSFER — road and endurance riders (inference, untested): nothing here measured sustained power, efficiency or a road sprint after hours of riding. For endurance cyclists the tested strength routes are heavy lifting (REF-146, REF-147, REF-075) and all-out pedalling against very high resistance (REF-158, B+; REF-428, A-); on-bike isometrics for them would be the base's inference from a sprint trial.
TRANSFER — off-bike holds (inference, untested): a single-leg leg-press or split-squat hold at a "pedal angle" is not what was tested. The gain was task-specific (ISO-CYC +12.5% while the seated knee-extensor test did not differ between groups), and isometric gains spread mainly to neighbouring angles (REF-1159, B+), so expect less transfer from a gym hold than from pushing the cranks in the riding position.
Equipment (practice, not the paper's): the trial used a custom ergometer with a clamped flywheel, instrumented cranks and torque feedback, under supervision. A home trainer or bike is not built to be locked against maximal seated pushes; an improvised lock is untested and is not a recommendation.
Judging the result in one rider (the base's reading): with a peak-power CV of about 2.7% in this lab, a change of 2-3% after a single before-and-after test is within noise; repeat the sprint test before calling a response.
Limitations
Not randomised: the head coach allocated riders on availability, location and willingness; the authors state they "cannot discount a possible bias inherent in the coaches approach". CON was mostly podium-level riders and EXP mixed under-23, national and podium-level riders, with 5 of 13 women vs 2 of 11.
Small (13 vs 11); no sample-size calculation, registration or blinding is reported; many outcomes tested without correction across outcomes.
The headline between-group effect is in W/kg only; absolute peak power did not differ from CON (p = 0.14), and part of the W/kg gap comes from a non-significant 1-kg drop in body mass in EXP (our calculation).
The groups may not have differed only in the isometric work: the Methods describe EXP sessions as the isometric work, a lighter squat and trunk exercises, without CON's second multi-joint lift and unilateral exercises, while the abstract and the Design say only the usual compound lift was replaced; CON's loading was individualised within ranges and not precisely controlled.
Six weeks, one off-season phase that followed 1-2 weeks off; no follow-up, no track or race performance, no longer-duration sprint (sprint power was 4-s isovelocity efforts in the lab).
Elite track sprinters only (Tier 4-5): nothing on road or endurance cyclists, other events or less-trained riders.
Bespoke equipment (custom ergometer with a clamped flywheel and instrumented cranks); the set × rep count applies to each lead leg only by the paper's example, not by an explicit statement.
The abstract gives absolute peak power as "P < .05" while the results give p = 0.05; Table 1's 4.2% for maximal torque is larger than the change in group means (207 → 214 Nm, our calculation about 3.4%), presumably a mean of individual changes.
Funding: a Northumbria University and English Institute of Sport studentship; no conflict-of-interest statement in the published PDF. Folland, a co-author here, is also an author of REF-1159, one of the three joint-angle-specificity papers the Introduction cites for its rationale (ref. 14, with Thepaut-Mathieu 1988 and Young 1985), and of REF-1161, which this paper does not cite.
How this entry was checked and changed
Why this level
parallel controlled trial, n = 24 (13 + 11), read from the full text (published version, open access). The rubric's B+ needs a randomised trial of 20-39; here allocation was explicitly not random (made by the head coach on availability, location and willingness), the control group was mostly podium-level riders while the experimental group mixed under-23, national and podium-level riders, the sexes were unevenly split (5 of 13 vs 2 of 11 women), no blinding, sample-size calculation or registration is reported, and the only between-group difference in sprint power was in power per kilogram
Full text (checked 2026-10-20)
24 track sprint cyclists (17 males, 7 females, age, 23 ± 3 year; mass 80.8 ± 10.9 kg); 200 m personal best for the males (n = 13) between 9.6 and 10.8 seconds, females (n = 5) 10.9 and 12.0 seconds (full text).
"Full random allocation of participants was not possible"; CON n = 11, 9 male, 2 female, largely podium-level; EXP n = 13, 8 male, 5 female; minimum of 2-year history of resistance training (full text).
CON: 3 to 5 sets of 3 to 5 repetitions at 85%-95% of 1 RM, 2-3 seconds descent, 3-5 minutes between sets; unilateral exercises 6-12 repetitions at ~70%-90% of predicted 1 RM (full text).
EXP: 45°, 90°, and 135 from top dead center; efforts maximal and 3 seconds; 1 set of 3 repetitions in week 1, 2 × 4 in week 2, 3 × 3 in week 3, 4 × 3 in week 4, 4 × 3 and 4 × 4 in week 5, 4 × 4 in week 6; 60 seconds between reps, 2 minutes between sets (full text).
EXP also performed 3 sets of 5 reps of the back squat at ~70%-75% of predicted 1 RM with 3-minute rest, after the isometric training protocol (full text).
Weekly: track cycling sessions (n = 2), gym sessions (n = 2-3), road rides (n = 1-2) of 60-90 minutes; high torque day 3-12 reps of 6-20 seconds; high power day 3-5 reps of 10-35 seconds (full text).
Absolute PPO in EXP 46 ± 62 W, 3 ± 4%, d = 0.17, P = .05; CON −5 ± 98 W, P = .844; group × time interaction P = .14 (full text).
Relative PPO group × time interaction P = .02; 0.8 ± 0.7 W/kg (P = .004) in EXP and − 0.1 ± 1.0 W/kg in CON; Table 1 18.7 ± 2.5 to 19.5 ± 2.3 vs 19.0 ± 3.5 to 18.9 ± 3.1 (full text).
Ten out of 13 participants in EXP increased relative PPO by >2%; average 4 ± 5%, range −1% to 16%, d = 0.33; CON 0 ± 5%, −8% to 7%, d = −0.05 (full text).
Extrapolated maximal torque 7.1 ± 6.5 Nm, P = .007 in EXP vs 2.4 ± 9.7 Nm in CON, group × time P = .14; maximal cadence P = .70 and P = .36 (full text).
ISO-CYC increased in EXP by 12.5% (P = .001), group × time interaction P = .002; 400 ± 78 to 450 ± 113 vs 383 ± 98 to 367 ± 82 (full text).
Torque at 150 and 200 ms increased in EXP (P = .03 and .003), interaction P = .18 and 0.054; pennation angle P < .001 over time, CON 6.4% vs EXP 5.2%, group × time P = .06; muscle thickness P = .034, group × time P = .66 (full text).
MVF 309 ± 75 to 323 ± 65 in EXP vs 296 ± 44 to 296 ± 44 in CON; body mass 82.1 ± 13.1 to 81.1 ± 12.0 vs 80.2 ± 8.3 to 80.6 ± 8.0 (full text).
Table 3 (r, P) against the relative change in PPO, n = 24: Torque150 .42, .04; Torque200 .47, .02; ISO-CYC .22, .30; MVF .36, .08; variance explained (18%-22%) by changes in RTD (full text).
Reliability: PPO CV = 2.7%, ICC = 0.96; maximal torque CV = 3.6%; ISO-CYC CV = 3.9% (full text).