Personalized Reference Values for the Two-Minute Walk Test: An Analysis of Cross-Sectional Data From the National Institutes of Health Toolbox Study

Abstract

Objectives: To develop reference values for the Two-Minute Walk Test (TMWT) via 2 previously untested methods: (1) smooth age-based statistical models and (2) a neighbors-based approach accounting for age, sex, and height. Design: Cross-sectional observational study. Setting: National Institutes of Health Toolbox study sites across the United States. Participants: A total of 1385 healthy, community dwelling adult participants (age 18-85 years) in the National Institutes of Health Toolbox study were included in this analysis. Intervention: None. Main Outcome Measures: Reference values for TMWT were generated using 2 approaches: (1) Generalized Additive Models for Location Scale and Shape, wherein TMWT values were modeled as a smooth function of age, and (2) a semiparametric neighbors-based approach. The performance of references values was then adjudicated by examining precision (ie, the average interquartile or interdecile range of reference values), and coverage (ie, the proportion of realized values included within a given inter-percentile interval). Agreement between methods was examined by intraclass correlation coefficient. Results: Neighbors-based reference values demonstrated a smaller average interquartile range (149 ft; 95% confidence interval [CI], 146-152 ft), compared with age-based reference values (158 ft; 95% CI, 155-162 ft), but similar average interdecile range (neighbors-based: 369 ft; 95% CI, 360-377 ft; age-based: 374 ft; 95% CI, 366-383 ft). Coverage appeared accurate via both approaches. Agreement between approaches was high (intraclass correlation coefficient=0.96), although differences were apparent on a case-by-case basis. Conclusions: Both age-based and neighbors-based reference values offer viable options for interpreting a person's TMWT performance. In this analysis, the neighbors-based approach (adjusting for height) yielded potentially clinically relevant differences in reference values for persons at extremes of height.