Abstract
BACKGROUND: In Ethiopia, use of advanced body composition measurement methods may not be feasible due cost and unavailability of the facilities. This study developed and validated body fat percent prediction equation for adults using locally appropriate data. METHODS: The study was conducted from February to April 2015 among 704 randomly selected adult employees of Jimma University. The total sample was spilt and randomly assigned to a training (n = 352) sample used for developing Ethiopian body fat percent (BF%) prediction equation and a testing (validation) sample (n = 352) used for determining the validity of the equation. A multivariable linear regression model was used to develop BF% prediction equation on the training sample using Air displacement Plethysmography (ADP) measured BF% as dependent variable and age, sex and body mass index as predictor variables. For the testing (validation) sample, BF% measured using ADP and the one predicted using the newly developed Ethiopian and Caucasian BF% prediction equations were compared using validity measures, Kappa statistics and agreement between the two measures was determined using Bland Altman plot. RESULTS: A multivariable linear regression model run on the testing population showed that age, sex and BMI were significant predictors of ADP measured BF%. Accordingly, the BF% prediction equation of Ethiopian adults was generated as follows: BF% = -8.601 + BMI (1.521) + Age (0.243) + Sex (-10.568), where sex = 1 for males and 0 for females. Comparison of measured and predicted BF% showed that there was no significant (P = 0.932) difference between ADP measured BF% and BF% predicted using Ethiopian equation with a mean (±SD) difference of 0.03 (±5.44). Conversely, there was a significant difference (<0.0001) between ADP measured BF% and the Caucasian Equation estimated BF% with a mean (±SD) of 6.83 (±5.57). In both males and females, the Ethiopian equation demonstrated a very good to excellent sensitivity, specificity, positive predictive value and negative predictive values. Conversely, the Caucasian equation had poor sensitivity and negative predictive values, while it demonstrated an excellent specificity and positive predictive value. Likewise, there was a substantial Kappa agreement for males (K = 0.741) and for females (K = 0.720) between Ethiopian equation and ADP in diagnosing obesity among males based on BF%, while there was a slight Kappa agreement for males (K = 0.156) and a fair Kappa agreement for females (K = 0.365) between Caucasian equation and ADP (P < 0.001). Bland Altman plot showed a good agreement between ADP measured BF% for the Ethiopian Equation and not for the Caucasian equation. It was observed that the Ethiopian equation has a better prediction of BF% when compared to the measured one, but the Caucasian equation consistently underestimated BF% for all samples with different levels BF%. CONCLUSION: The new Ethiopian BF% prediction equation performed very well in predicting BF% in the testing population in terms of validity measures, Kappa agreement and Bland Altman plot; while the Caucasian equation significantly underestimated body fat percent among Ethiopian adults. The results imply that the new Ethiopian equation can be used as a cost effective and user friendly screening method for early detection of obesity for the prevention of associated morbidity and mortality in Ethiopian adults.