REF-1277 · Ross et al., 2015 · Mayo Clinic Proceedings · Paper · PubMed

My fitness stopped improving — train more or harder?

In 121 sedentary obese adults, harder sessions at equal energy left 0% non-responders vs 17.6%; double the amount cut 38.5% to 17.6%. Not athletes (B+, from the abstract).

Main caveat: Read from the abstract only. The non-response definition, the mean changes and their SDs, the number randomised and any control arm could not be checked.
moderate evidenceMeasured on non-athletes

Ross, de Lannoy and Stotz tested whether the intensity and the amount of exercise separately change how many people fail to improve their cardiorespiratory fitness (CRF).

Sedentary, abdominally obese, middle-aged adults were randomly assigned to five sessions a week for 24 weeks of low-amount, low-intensity exercise (LALI: 180 kcal per session for women and 300 for men, at 50% of VO2peak), high-amount, low-intensity exercise (HALI: 360 and 600 kcal at 50%) or high-amount, high-intensity exercise (HAHI: 360 and 600 kcal at 75%).

The analysis included the 121 participants who completed at least 90% of the prescribed sessions (75 women; 53.2 ± 7.5 y): 39 LALI, 51 HALI and 31 HAHI. CRF was measured with a treadmill test at 4, 8, 16 and 24 weeks; the study ran from 1 September 2009 to 31 May 2013. CRF rose within all three groups by 24 weeks (P < .001).

At 24 weeks 38.5% (15 of 39) of LALI, 17.6% (9 of 51) of HALI and 0% (0 of 31) of HAHI participants were CRF non-responders. At a fixed intensity, more exercise cut the non-response rate by 50% (P = .02); at a fixed amount, higher intensity eliminated it (P = .001).

Over time, non-responders fell between 4 and 8 weeks in LALI and by 16 weeks in HALI and then plateaued, while in HAHI they kept falling through 24 weeks. Our reading: the non-responder counts fall with amount and above all with intensity; the abstract gives no group mean changes, but a larger mean gain alone would produce this pattern.

With a fixed threshold, a group whose mean gain is larger has fewer people below it even if the spread is the same (our inference), and the abstract gives neither the threshold nor a control arm's SD of change, so it cannot show that the low-intensity "non-responders" differed from the others in anything but noise.

Protocol

Participants: n = 121 completers (75 women, 62%); sedentary, middle-aged (53.2 ± 7.5 y), abdominally obese; completed at least 90% of five weekly sessions over 24 weeks. The number randomised and dropouts by arm are not in the abstract.
Arms (randomised): LALI — 180 kcal (women) or 300 kcal (men) per session at 50% of VO2peak, n = 39; HALI — 360 or 600 kcal at 50%, n = 51; HAHI — 360 or 600 kcal at 75%, n = 31. Five sessions a week for 24 weeks.
Outcome: CRF (VO2peak) by treadmill test at 4, 8, 16 and 24 weeks; the share of CRF non-responders in each arm over time. The definition of non-response is not in the abstract.
Dates: 1 September 2009 to 31 May 2013.
Not in the abstract: the training mode, mean CRF changes and their SDs, the non-response threshold, a non-exercise control arm, trial registration.
Funding and interests: Canadian Institutes of Health Research (grant OHN-63277 in the PubMed record; CIHR is also the funder listed on Crossref). Conflicts of interest are not stated in the abstract or the PubMed record.

Key Results

Arm (energy per session, women/men; intensity)CompletersCRF non-responders at 24 weeksNon-responders over time
LALI: 180/300 kcal at 50% of VO2peak3938.5% (15 of 39)fell between 4 and 8 weeks, then plateaued
HALI: 360/600 kcal at 50%5117.6% (9 of 51)fell by 16 weeks, then plateaued
HAHI: 360/600 kcal at 75%310% (0 of 31)fell continually over 24 weeks
CRF rose within all three groups at 24 weeks (P < .001); the size of the gains is not in the abstract.
Amount at a fixed intensity (LALI vs HALI): the non-response rate fell by 50% (P = .02).
Intensity at a fixed amount (HALI vs HAHI): non-response was eliminated (P = .001).
Spread: neither the SD of change in each arm nor a non-exercise control arm's SD of change is in the abstract, so the share of non-responders cannot be checked against measurement noise.

Mechanisms

Measured: treadmill CRF at four time points; counts of participants classed as non-responders.
Authors' interpretation: at a fixed amount, raising intensity to the level of consensus recommendations eliminated non-response; low-intensity exercise may not be enough to improve CRF in a substantial share of sedentary obese adults.
Inference: a count of non-responders depends on the mean change, the spread and the threshold. The HAHI arm's 0% fits a larger mean gain; it does not show that every individual responds to intensity, nor that the LALI non-responders were a distinct kind of person (REF-430: a responder label must be shown to exceed noise; REF-1274: in controlled trials most of the spread in VO2max change was measurement error).
Inference: the plateau of the low-intensity arms after 8-16 weeks is the group's count levelling off, which fits a fixed load ceasing to add once the body has adapted to it (REF-1218, a fixed load stopped raising VO2max after about three weeks in nine people); the high-intensity arm kept improving to 24 weeks.

Practical Application

My fitness stopped improving — train more or harder? In 121 sedentary obese adults, harder sessions at equal energy left 0% non-responders vs 17.6%; double the amount cut 38.5% to 17.6%. Not athletes (B+, from the abstract).
The direction for a stall, in untrained adults: at a fixed amount, raise the intensity; at a fixed intensity, add amount. The same two exits appear in REF-995 (higher intensity at the same energy) and REF-996 (more sessions a week), in adults who were untrained (REF-995) or not described as trained (REF-996); for trained athletes this is transfer (SYNTH-42, "Done and flat").
Low-intensity-only plans level off within months in sedentary people. At 50% of VO2peak the number of non-improvers stopped falling after 8-16 weeks; at 75% it kept falling to 24 weeks. For a trained athlete 75% of VO2peak is not "high intensity", so the threshold does not carry over (inference).
"0% non-responders" is not "everyone responds". Without the definition and a control arm's SD of change the counts cannot separate real non-response from noise; never tell an athlete they are a non-responder from one block (SYNTH-13) (inference from REF-430 and REF-1274).
Rule out the rest before changing the lever. Not from this paper; the base's own practice (SYNTH-42): check that the block was done, and rule out illness and under-fuelling, before calling it flat.

Limitations

Read from the abstract only. The non-response definition, the mean changes and their SDs, the number randomised and any control arm could not be checked.
Per protocol. Only completers of at least 90% of sessions were analysed; HAHI had the fewest completers (31, against 39 and 51), and if completion differed by arm the comparison may be biased (inference).
No control-arm spread in the abstract. Individual-response claims need the SD of change in a non-exercising arm; without it the non-responder shares cannot be told apart from test noise and within-person variation.
Arithmetic. 38.5% to 17.6% is a 54% relative fall, reported as "by 50%" (our arithmetic; consistent in direction).
Who. Sedentary, abdominally obese adults around 53, 62% women; the exercise mode is not given in the abstract.
Funding and interests: Canadian Institutes of Health Research; no device maker, federation or data platform named. Conflicts of interest not stated in the abstract or the PubMed record.
How this entry was checked and changed
Why this level

randomised, but analysed per protocol in completers only (how many were randomised is not in the abstract), and the outcome is a count of "non-responders" whose definition, and the control-arm SD of change that would calibrate it, are not in the abstract; the rubric's A- for a well-powered RCT (40 or more) is held one step lower for these reasons. Read from the abstract (closed access; no copy in Europe PMC or OpenAlex, and the publisher page sits behind a bot-check page)