REF-1620 · Cusson-Fradet, Domingue & Lajoie, 2026 · Journal of Sports Sciences, published online October 2026, pp. 1-17 · Paper · PubMed

An 8-minute cycling test paced by perceived effort alone (SCT-RPE: three 2-min stages at "Moderate", "Strong" and "Very strong", then recovery and a 6-s sprint) gave a power-to-heart-rate index that was strongly associated with peak work rate across 21 cyclists (R² = 0.77, 13.4 W per index unit) with between-day typical errors of 3.29-7.97% for its endurance variables; whether it detects change within one rider was not tested

Main caveat: One study, 21 cyclists; the abstract does not give sex, age or training level.
limited evidenceMeasured on a mixed groupRead from the abstract only

Cusson-Fradet, Domingue and Lajoie tested a short self-paced cycling test in which the rider, not a power target or a heart-rate clamp, sets each stage by perceived effort. Twenty-one cyclists rode three 2-minute stages at the verbal anchors Moderate, Strong and Very strong, recovered, then sprinted for 6 seconds.

The authors computed the ratio of power to heart rate (PO/HR) and a weighted version of it (wPO/HR) for each stage and for the whole test, and compared them with a graded exercise test and with sprint performance; the test was repeated to measure reliability.

They kept the overall weighted index, wPO/HRoverall, as the candidate endurance-fitness index, because it was closely associated with peak work rate on the graded test, reasonably reliable, and weighted towards power, which suits a test regulated by effort.

Endurance variables showed moderate to very high day-to-day reliability; the sprint peak power was reliable and correlated almost perfectly with separate sprint performance (r = 0.93; a correlation, not agreement in watts).

The authors conclude that the test shows cross-sectional convergent validity and acceptable reliability, and that its responsiveness over time is still to be established.

Participants

n = 21 cyclists (sex, age and training status not given in the abstract).

Protocol

Test: three 2-min stages at RPE Moderate, Strong and Very strong (the abstract does not name the scale; Borg CR100 per the first author's forum post), a recovery period, then a 6-s sprint (Ppeak); about 8 min in all.
Indices: PO/HR and weighted wPO/HR for each stage and overall.
Validity: against graded exercise test outcomes (peak work rate, Wpeak) and sprint performance.
Reliability: repeated trials on different days (ICC3,1 and the coefficient of variation of the typical error, CVTE).
From the first author's forum post (4 Nov 2026), not the abstract and not checked against the paper: wPO/HR = PO/√HR; power and heart rate averaged over the last 30 s of each stage; 60 s of passive recovery and 54 s of easy pedalling before the sprint; Ppeak = the highest 1-s power; three trials over two laboratory visits 2-14 days apart. Neither the abstract nor the post describes the warm-up.

Key Results

MeasureResult
wPO/HRoverall against WpeakR² = 0.77, p < 0.001, slope 13.4 W per index unit
Endurance-related variables, between daysICC3,1 0.72-0.94; CVTE 3.29%-7.97%
Ppeak (6-s sprint), between daysICC3,1 0.86; CVTE 8.96%
Ppeak against separate sprint performancer = 0.93
The 13.4 W per unit slope is a comparison between cyclists, not a change within one cyclist.
No result on whether the index moves with training, detraining or fatigue: the authors state that longitudinal responsiveness remains to be established.

Practical Application

A short submaximal check can be paced by effort alone. In 21 cyclists, power relative to heart rate in three 2-min effort-paced stages ranked riders' peak work rate well (R² = 0.77), and, because the stages are set by effort rather than by a power or heart-rate target, it needs no prior maximal test to set up (by design; the first author's forum post says the same).
Do not read a single change as fitness or fatigue yet. The endurance variables' between-day errors were 3.29-7.97%; for the retained index itself the first author's forum post gives 4.3% (typical error 0.94 units, ICC 0.92) and says a change of about 1 unit is within the test's error (not in the abstract).Inference, not from this paper: that is about the size of change a training block may bring in a trained rider. The test's ability to follow one rider over time was not studied. Read it as a pattern over several tests, beside RPE and how training is going (SYNTH-41, SYNTH-53).
Inference, not tested here: fatigue can hide in a power-to-heart-rate ratio. Two days after an ultramarathon, runners had a lower heart rate and a higher RPE at the same treadmill speeds (REF-757; the overreaching studies with the same pattern are gathered in SYNTH-41).In an effort-paced test that would mean less power and a lower heart rate at the same effort, which move a power-to-heart-rate ratio in opposite directions, so the index may change less than power does when the rider is tired. The abstract calls the weighting power-dominant (PO/√HR in the first author's forum post), which softens this but does not remove it.This is why the base reads its own submaximal check as a set (power or output, heart-rate recovery and RPE together, SYNTH-41).
Inference: 2-min stages are short for heart rate. Heart rate is still rising at the end of a 2-min stage, so how hard the rider starts each stage may shift the reading; keep the start of each stage the same from test to test.
The sprint is a neuromuscular check, not a fine one. Its day-to-day error of about 9% (CVTE 8.96%) is larger than most changes worth acting on (inference, not from this paper).
Not from this paper: the heart-rate-clamped reference test (LSCT) and its reliability (typical errors of 1.3-4.4% across the measures of its reliable stages) are in REF-752; how the base reads its own check is in SYNTH-41.

Limitations

One study, 21 cyclists; the abstract does not give sex, age or training level.
Cross-sectional validity only: the association with peak work rate is between riders; no test of responsiveness to training or fatigue within a rider.
Validity was judged against a graded exercise test and sprint tests, not against a time trial or race performance.
The weighting behind wPO/HR (PO/√HR) and the 30-s sampling window come from the first author's forum post, not from the abstract; the full text was not checked for them.
Written from the abstract, the record metadata and the first author's intervals.icu forum post (4 Nov 2026); full text not read. A correction to the article was published on 3 Nov 2026 (doi 10.1080/02640414.2026.2753281); its content was not read, so whether it changes any number here is unknown.
How this entry was checked and changed
Why this level

a test validation study is B by design (rubric); the marker step (rule 3) is not applied because the index is the subject of the study; not B- although read from the abstract only, because the abstract gives the design, comparators, sample and every statistic the entry quotes (the index formula and sampling window are not in it, and no number here rests on them); not B+, which needs a cross-sectional measurement study of ≥100 (n = 21)