Validity Of A-mode Ultrasound Estimating Percentage Of Body Fat Comparing To Lab Methods: A Meta-analysis

Amplitude mode ultrasound (A-US) measures the strength of reflected sound waves throughtissue and is used to measure the thickness of subcutaneous fat tissue to estimate body fat (%BF).Validation studies have tested A-US agreement by using established methods such as hydrodensitometry, dual energy absorptiometry, air displacement plethysmography and multi-component models. PURPOSE: This meta-analysis investigated the validation studies of A-US to estimate %BF among healthy populations of all ages. METHODS: Using the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines, databases were searched for A-US validation studies in humans written in English. The Pearson Correlation Coefficient (r) was extracted from each study, and results were synthesized by JAMOVI software programs. The sensitivity analysis was performed to minimize heterogeneity by removing studies falling out of the 95% CI. RESULTS: A total of 19 studies met the inclusion criteria, representing 1350 participants (male 831; female 519). Various laboratory methods were used, including DEXA (n = 10), ADP (n = 8), and 3-component model (2C + water)(n = 1). Results were presented for combined samples of males and females (n = 7) and with results separated by gender (n = 12). Combined sample results were only analyzed if gender separated results were not reported. A strong positive correlation was identified between A-US and lab methods for studies reporting both genders together (r = 0.89 [0.86 ~ 0.92] p < 0.05); for males (r = 0.91 [0.88 ~ 0.95] p < 0.05); and for female (r = 0.85 [0.77 ~ 0.93], p < 0.05) based on a random effects model. High heterogeneity was detected (I2 range 93.2 ~ 96.8%). The funnel plot showed an asymmetry indicating potential publication biaswith studies falling outside the 95% CI. After removing these studies, the r’s maintained a strong relationship (r = 0.81 ~ 0.92, p < 0.05) and heterogeneity was low (I2: 0 ~ 30.9%, p > 0.05). The studies removed were from non-white populations or used either uncommon A-US devices or sites. CONCLUSIONS: This present study showed a strong correlation between A-US and lab methods to estimate %BF of healthy population, but there were concerns of heterogeneity due to varying samples and methods. Before recommending for use, the prediction error needs to be quantified.