How close to failure does a set need to be?

It does NOT say "train to failure." Five overclaims a reader would reach for that this paper does not support. (1) It does not identify an optimal RIR or say hypertrophy is maximised at 0 RIR — the abstract reports the direction of a slope, never a peak, and the authors write that "the exact relationship between RIR and muscle hypertrophy and strength remains unclear." (2) It does not say training far from failure produces no strength gain. "Negligible relationship" means RIR barely moved the outcome, so strength was gained across the range — the opposite of the reading most people take. (3) It measures nothing about fatigue, injury risk, joint stress, or adherence, so it cannot support either "failure is safe" or "failure is unsustainable." (4) It is not an experiment on RIR. The exposure was assigned retrospectively by the reviewers: "the RIR associated with each effect in the analysis was estimated on the basis of the available descriptions of the training interventions in each study." The authors call the models exploratory and warn that "caution is warranted when interpreting the present analysis due to its exploratory nature." Never present this as a controlled dose-response trial. (5) It does not report how many studies or participants were pooled, or over what durations — the abstract gives no such figures, so any k, n, effect size, or confidence interval attached to this card has been imported from a different paper.