JMIR Form Res. 2026 Aug 12;10:e100764. doi: 10.2196/100764.
ABSTRACT
BACKGROUND: Consumer wearable activity monitors are increasingly being used as end points in exercise-oncology trials and in clinical decision-making, yet their accuracy is unvalidated in lung cancer, where slow, fragmented gait may challenge step-counting algorithms.
OBJECTIVE: This study assessed the criterion validity of the Fitbit Charge 6 device against video-recorded direct observation in adults with lung cancer under controlled laboratory conditions. We aimed to describe step-count agreement across walking bout durations and gait speeds, and to compare its accuracy in classifying active vs sedentary minutes and detecting spurious steps across nonwalking activities.
METHODS: Fourteen adults diagnosed with stage I-IV lung cancer completed a cross-sectional, in-laboratory validation study at The Ohio State Wexner Medical Center. Participants wore the Fitbit Charge 6 device on their nondominant wrist while completing variable-duration walking trials (5, 15, and 30 seconds); self-selected gait speed trials across 8 progressively faster speeds; and standing, sitting, lying, and fidgeting tasks. All activities were video recorded and coded at a 1-second resolution. Step count agreement was evaluated using repeated-measures Bland-Altman analysis (mean bias and 95% limits of agreement), supported by mean absolute percentage error (MAPE) and intraclass correlation coefficients (ICCs). Minute-level activity intensity classification was assessed via a pooled confusion matrix using a majority-rule active-minute threshold of ≥30 seconds. Spurious step detection was descriptively analyzed across nonwalking minutes, stratified by fidgeting status.
RESULTS: Across 126 walking trials, the Fitbit device undercounted steps by only a small absolute margin, which was consistent across bout durations (bias of 1-3 steps), but relative agreement was poor and strongly duration dependent (MAPE 50.6% at 5 seconds vs 16.3%-18.7% at 15-30 seconds). The ICC was low (overall ICC[A,1]=0.21). Across 111 gait speed trials, undercounting was the greatest at gait speeds below 0.6 m/s (bias of approximately 13 steps; MAPE 56.6%) and was minimized near 1.0-1.2 m/s, with a quadratic mixed-effects model confirming a nonlinear speed-error relationship (P<.001). For activity intensity classification, sensitivity was high (0.91), but specificity was modest (0.63), and the positive predictive value was low (0.31), reflecting frequent misclassification of sedentary minutes as active. Among 267 nonwalking minutes, 33 (12.4%) contained at least one spurious step, with higher false-positive rates during fidgeting (24/167, 14.4%) than nonfidgeting (9/100, 9.0%) periods.
CONCLUSIONS: The Fitbit Charge 6 device provides improved step counts during sustained, moderate-speed walking but introduces a clinically meaningful error during short bouts and at slower gait speeds, which are frequently noted in adults with lung cancer. High sensitivity but low specificity for activity intensity classification suggests systematic overestimation of active minutes. These findings have implications for the design and interpretation of exercise-oncology interventions relying on consumer wearable-derived end points in this population.
PMID:42585577 | DOI:10.2196/100764
