Zackary S. Cicone1, Michael V. Fedewa2, Clifton J. Holmes3, Michael R. Esco, FACSM2, Hayley V. MacDonald2. 1Shenandoah University, Winchester, VA. 2The University of Alabama, Tuscaloosa, AL. 3Washington University in St. Louis, St. Louis, MO.

BACKGROUND: Single frequency bioimpedance analysis (SFBIA) is a simple alternative to isotope dilution techniques for assessing total body water (TBW). How characteristics related to the sample (e.g., age, health status) and SFBIA methodology (e.g., criterion technique, device, frequency, index) influence SFBIA accuracy has yet to be comprehensively examined. The aims of this systematic review and meta-analysis were to 1) quantify the accuracy of SFBIA for predicting TBW and 2) determine the potential impact of study-level effect modifiers. METHODS: Five electronic databases were searched for studies that compared isotope dilution TBW values to SFBIA. Standardized mean difference (SMD) effect sizes were calculated using the Gibbons method for each comparison and an overall estimate was generated using a three-level random-effects model. Within- and between-study level variance was evaluated using one-sided log-likelihood-ratio tests. When appropriate, subgroup analyses were performed to identify potential study-level moderators. RESULTS: Aggregate-level data from 51 studies (255 individual effects) were included in the final analysis. Study samples included predominantly healthy participants with large ranges in mean age (0 to 82 y) and body mass index (14.1 to 50.2 kg/m2). The overall SMD indicated a negligible difference between SFBIA and criterion dilution methods (SMD=-0.04, p=0.67), but lacked homogeneity at both the within- (σ2=0.45) and between-study (σ2=0.26) levels (all p<0.001). Moderator analysis revealed that the interaction between frequency and index (p<0.01) influenced the observed error between SFBIA and criterion methods. Resistance index (Ht2/R) produced less error than impedance index (Ht2/Z) across all frequencies (all p>0.10), with Ht2/R at 50 kHz producing the most accurate estimate of TBW (β=0.06, p>0.05). Additionally, there was a small yet significant effect for sample sex (% women, β=-0.003, p<0.05), suggesting that SFBIA may underestimate TBW in samples that are predominantly women. No main effects were observed for other study-level factors (e.g., sample characteristics or BIA methodology). CONCLUSION: Overall, Ht2/R produced less error in TBW estimation than Ht2/Z, with Ht2/R at 50 kHz providing the smallest mean difference in TBW when compared to isotope dilution. These results suggest that SFBIA may provide acceptable estimates of TBW across a range of diverse samples.

This document is currently not available here.