Accuracy of wearables for determining the maximal oxygen uptake and lactate threshold: a qualitative systematic review.

Introduction: The limitations of costly laboratory methods could be addressed by advances in wearable sensors and associated machine learning algorithms (13). Key advantages of wearable technologies (WT) include non-invasive and continuous monitoring. However, it is necessary to control and ensure their accuracy. The aim of this review was to analyze and systematically summarize recent studies on the validity of maximal oxygen uptake (V̇O2max) and lactate threshold (LT) as estimated by consumer wearables. Methods: The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed in preparing this review. The literature was systematically searched in the Web of Science, SPORTDiscus, and Medline databases. The risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies, version 2, tool. Results: Of the 252 records initially identified, 13 studies met the defined inclusion criteria. Most studies investigated Garmin smartwatches with compatible chest belt heart rate (HR) sensors and the Firstbeat Technologies algorithm. V̇O2max was measured in all cases in the laboratory with a treadmill graded exercise test (GXT), while estimations were generated from submaximal outdoor runs. Studies with healthy untrained adults, recreational athletes and highly trained or professional athletes were equally represented. In seven studies wearable devices proved to be valid or acceptable for V̇O2max estimation compared to the gold standard measurement. Three studies demonstrated validity of LT estimation. Discussion: Our review demonstrates valid estimations of V̇O2max using WT in populations of healthy untrained adults, recreational athletes and team sports professional athletes. Usefulness in elite endurance sports is questionable and may depend on artificial intelligence (AI). Two or more submaximal runs as an index test could improve validity, but further examination is needed. An important finding is that valid estimations were calculated from submaximal tests. These make consumer wearables a user-friendly alternative to laboratory GXT, which has some limitations. To provide insight into stability of algorithms, there is a need for longitudinal studies that would monitor the accuracy of WT over months or even years. Ongoing research into the latest models of smartwatches and synthesis of data is critical to understanding their suitability and usefulness.