Reliability, Device Agreement and Validity of Load-Velocity Profiles: A Systematic Review with Meta-analysis
Sports medicine - open · 2026
Velocity sensors and their 1RM predictions look reliable on paper, but missing error data and big heterogeneity mean treat the numbers as estimates, not a replacement for testing.
The paper
Systematic review with meta-analysis, 63 studies on velocity sensor validity/reliability (Part I) and 38 studies on velocity-based 1RM prediction models (Part II), covering commercial LPT and IMU devices in resistance training populations.
What they found
Pooled ICCs of 0.91-0.92 (95% CI 0.83-0.97) for sensor validity/device agreement and 0.90-0.91 (CI 0.85-0.95) for intra/inter-day reliability; velocity-based 1RM prediction showed ICC 0.90 (CI 0.83-0.94) for reliability and 0.91 (CI 0.72-0.98) for validity, with substantial heterogeneity across moderators.
Study design
Systematic review with meta-analysis, PROSPERO-registered, pooling k up to 608 individual effect sizes from over 100 primary studies of varying design (mostly cross-sectional validity/reliability studies) in trained and recreationally active lifters.
Methodology
Search covered PubMed/MEDLINE, Web of Science and Scopus with adapted COSMIN quality appraisal, which is appropriate for measurement-property reviews, but the abstract reports no formal GRADE certainty rating and flags 'substantial heterogeneity' without abstract-level I2 figures or clear moderator breakdown. Critically, the authors admit a 'dearth of measurement error and agreement analyses,' meaning the headline ICCs describe relative consistency, not the absolute error (SEM, limits of agreement) a clinician would need to trust an individual predicted 1RM. No mention of publication bias assessment or funding/COI in the abstract, and pooling across very different sensor brands, exercises and load ranges risks mixing apples and oranges under a single ICC.
The appraisal
ICCs above 0.90 look statistically reassuring but relative reliability statistics can stay high even when absolute agreement (the number that matters for setting a training load) is poor, and the authors themselves say this data is largely missing. The wide confidence intervals (down to 0.72 for the validity pooled ICC) and large heterogeneity, especially in lower body lifts, mean the 'good-to-excellent' summary masks real device- and exercise-specific unreliability rather than confirming it. Worth noting too that the 1RM prediction validity estimate is built on only 9 studies, so that 0.72-0.98 range isn't a stable, well-populated finding, it's a wide interval from a thin evidence base.
The gap
Head-to-head measurement error and agreement data (Bland-Altman/SEM) by device and exercise, so a coach can know the actual kilogram error margin on a predicted squat or bench 1RM rather than a pooled correlation coefficient.
Landmark context
This sits in the velocity-based training literature that grew out of early load-velocity profiling work (Jovanović and Flanagan-era propositions and the single-device validation studies that followed), and is one of the larger attempts to quantitatively pool validity and reliability data across commercial devices and prediction models rather than rely on narrative synthesis or single-device papers.
What to do Monday
Keep using velocity-based 1RM estimates for tracking within-athlete trends on a fixed device and exercise, but don't swap devices mid-programme or treat the predicted number as equivalent to a directly tested 1RM, particularly for squat and other lower body lifts where heterogeneity was worst.
In practice
For rehab caseloads where you're using a velocity sensor to autoregulate late-stage strength work (e.g. post-ACLR or tendon rehab where you track velocity loss to manage fatigue), this data supports trusting the device for within-session and within-athlete trend tracking, but not for calling a predicted 1RM the number you write into a return-to-sport strength benchmark; get an actual tested max or a supervised near-max lift at those decision points. On the S&C floor, use the predicted 1RM to drive daily load autoregulation and long-term trend monitoring on a fixed device and fixed exercise, but recalibrate against a real tested 1RM every training block, especially for squat and other lower-body lifts where the heterogeneity here was worst, and never compare kg numbers across two different sensor brands or between an IMU and an LPT as if they were interchangeable. The caveat that applies to both settings: if you change hardware, restart your baseline, don't assume continuity of the number.
The five papers that matter — every week, free.
We read the whole sports-medicine, rehab & performance literature and appraise the papers worth your time.
Get the free weekly issueRelated appraisals
Read in your language: English · Français · Deutsch · Español · Italiano · Nederlands · Dansk · Svenska