The short version
- A meta-analysis of 15 studies found no significant difference between training to failure and stopping short for muscular strength (ES -0.09) or hypertrophy (ES 0.22).
- In the trials within that analysis that did not equate training volume, non-failure training was significantly better for strength (ES -0.32, 95% CI -0.57 to -0.07).
- A separate meta-analysis of 15 studies found only a trivial hypertrophy advantage for set failure (ES 0.19), and no advantage at all for momentary muscular failure specifically (ES 0.12, p = 0.343).
- In resistance-trained participants, training to failure showed a small but significant hypertrophy benefit of 0.15 (95% CI 0.03 to 0.26).
- In 19 trained adults over five weeks with volume equated, training at 0 to 1 reps in reserve produced no greater strength or muscle cross-sectional area gains than training at 4 to 6 reps in reserve.
There is a specific kind of gym advice that survives because it sounds like honesty. Push every set until you physically cannot move the bar, the reasoning goes, because the last rep is the one that counts and everything before it is a warm-up. It is a satisfying story about effort. Two meta-analyses have now tested it and the story does not hold.
The strength answer is clear, and it is not the one you expect
Grgic and colleagues meta-analysed 15 studies comparing training to muscular failure against stopping short of it. For muscular strength the pooled effect was -0.09 (95% CI -0.22 to 0.05) — no significant difference, with the point estimate leaning very slightly toward not going to failure. For hypertrophy the pooled effect was 0.22 (95% CI -0.11 to 0.55), also not significant.
Two subgroup findings sharpen it. In studies that did not equate training volume between groups — which is what happens in real life, where a set taken to failure costs you reps later — the analysis showed a significant advantage for non-failure training on strength (-0.32, 95% CI -0.57 to -0.07). And in resistance-trained participants specifically, failure showed a significant but small effect for hypertrophy (0.15, 95% CI 0.03 to 0.26). Their conclusion: training to failure does not seem to be required for gains in strength and muscle size, and does not seem to be detrimental either.
The hypertrophy answer, from a second team
Refalo and colleagues approached the same question through proximity-to-failure and analysed 15 studies. Using any definition of set failure, they found a trivial advantage for failure over non-failure for hypertrophy: effect size 0.19 (95% CI 0.00 to 0.37, p = 0.045), with no moderating effect of volume load or relative load.
Then the subgroup analyses took most of it back. For momentary muscular failure specifically — genuinely unable to complete another repetition — versus non-failure, there was no advantage at all (0.12, 95% CI -0.13 to 0.37, p = 0.343). And training to a high velocity-loss threshold above 25% was no better than a moderate 20 to 25% (0.08, p = 0.529). Their read is that the relationship between proximity-to-failure and hypertrophy is potentially non-linear: getting close matters, arriving does not.
What changes inside the muscle when you go closer
A 2023 trial is useful here because it deliberately equated the thing the meta-analyses could not. Nineteen resistance-trained adults trained squat, bench and deadlift twice weekly for five weeks, one group ending sets with 0 to 1 reps in reserve and the other maintaining 4 to 6. Total training volume did not differ significantly between the groups.
Both groups improved their one-rep maxes, with no significant group-by-time interactions for any of the three lifts, and none for vastus lateralis cross-sectional area either. What did differ was neural: the near-failure group's motor unit firing rate versus recruitment threshold relationship changed, with the authors interpreting the shift as increased firing rates in lower-threshold motor units. Interesting, real, and not something you can see in a mirror after five weeks.
The part the meta-analyses cannot measure
Every one of these studies runs for five to twelve weeks in mostly young participants, and the outcomes are strength and size. What none of them capture well is the running cost: the sets you drop later in the session, the sessions you cut short later in the week, the joints that stop enjoying it. Grgic's finding that non-failure came out ahead for strength precisely in the trials where volume was allowed to fall naturally is the closest thing in this literature to a measurement of that cost.
This is also where a lot of supplement money quietly goes. A caffeine dose makes a grinding set feel more survivable, which is a genuine effect and a real reason people buy a pre-workout supplement. It does not make the grinding set necessary.
What we'd actually tell you
Take your compound lifts to roughly one to three reps in reserve. That is close enough to capture almost everything the evidence says proximity buys you, and it leaves you able to complete the rest of the session and the rest of the week — which is where progressive overload actually happens.
Save true failure for the last set of an isolation exercise, where the systemic cost is low and the feedback on your rep judgement is useful. And if you are new to lifting, know that the meta-analytic answer is that you can build strength and muscle without ever meeting the bar you cannot move. The intensity story is compelling. It is just not a requirement.
Good questions
Do I have to train to failure to build muscle?
No. Two meta-analyses covering 30 studies between them found failure training no better than stopping short for strength, and at best trivially better for hypertrophy. One found no advantage at all for true momentary failure. Getting close to failure appears to matter; arriving there does not, and it costs you work you could have done in the sets and sessions that follow.
How many reps in reserve should I leave?
One to three on compound lifts is a defensible default. In a five-week trial that equated total volume, a group stopping at 4 to 6 reps in reserve gained as much strength and muscle size as a group stopping at 0 to 1. The evidence supports being in the neighbourhood of failure rather than at it, so leave enough in the tank to complete the session as planned.
Is training to failure bad for you?
Not inherently. The meta-analysis was explicit that failure training does not seem to have detrimental effects on strength or hypertrophy either. The catch is what it does to everything after it: in trials where volume was allowed to fall naturally rather than being equated, non-failure training came out significantly ahead for strength. Failure is not dangerous, it is expensive.
Should I go to failure on squats and deadlifts?
We would not. The systemic and technical cost is highest on heavy compound lifts, the evidence says you gain nothing for strength by going there, and form degrades exactly when the load is heaviest. Keep one to three reps in reserve on the big lifts and put your failure sets on isolation work where the downside is a sore biceps rather than a rounded back.
How do I know how close to failure I actually am?
Most people are further away than they think, especially early on. A practical calibration is to occasionally take a light isolation set to genuine failure and note how the last few reps felt — bar speed slowing sharply is the most reliable cue. Rep-speed loss is the same signal the velocity-loss research uses, and the trials found stopping at a moderate 20 to 25% velocity loss worked as well as pushing past 25%.
Sources
- Grgic J, Schoenfeld BJ, Orazem J, Sabol F. Effects of resistance training performed to repetition failure or non-failure on muscular strength and hypertrophy: A systematic review and meta-analysis. J Sport Health Sci, 2022. View study
- Refalo MC, Helms ER, Trexler ET, Hamilton DL, Fyfe JJ. Influence of Resistance Training Proximity-to-Failure on Skeletal Muscle Hypertrophy: A Systematic Review with Meta-analysis. Sports Med, 2023. View study
- Ruple BA, Plotkin DL, Smith MA, et al. The effects of resistance training to near failure on strength, hypertrophy, and motor unit adaptations in previously trained adults. Physiol Rep, 2023. View study
These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.
