以重水稀釋法驗證不同生物電阻抗分析儀測量台灣人之體脂率

Using Deuterium Oxide Dilution Method to Validate Percent Body Fat Measured by Bioelectrical Impedance Analysis in Taiwanese

背景：在二組分身體組成模式中，重水稀釋法是測量全身水含量以推算體脂率的黃金準則。臨床上則以生物電阻抗分析儀最常被用來測量體脂率，此類生物電阻抗分析儀多為國外製造，但目前台灣尚未有本土文獻以驗證兩種方式測得體脂率之一致性。
方法：此橫斷性研究，採立意取樣，由至成大醫院家庭醫學部體重管理門診就診的台灣成年受試者取358位（男/女：137/221），平均年齡34.1±9.9歲，平均身體質量指數30.9±9.6 kg/m2。經隔夜空腹並排空尿液，以重水稀釋法和台灣地區常用的四種生物電阻分析儀來測定受試者全身水含量並推算體脂肪量及體脂率。使用SPSS 17版套裝軟體進行資料整理與統計分析。
結果： 3 5 8 位受試者中， 以重水稀釋法測得之平均體脂率為33.5±10.6%，與四種不同生物電阻抗分析儀測得之體脂率比較呈顯著差異。以Bland -Altman一致性分析發現四種生物電阻抗分析儀皆與重水測得之體脂率具一致性，然Tanita TBF-410機型於受檢者平均體脂率為50%以上時會略為明顯高估。以線性迴歸分析，依變數為重水稀釋法測得之體脂率，自變數為不同生物電阻抗分析儀測得之體脂率、身體質量指數（倒數值）、性別、年齡及腰圍時，所得到迴歸式之調整後決定係數為0.430到0.755，若於線性迴歸式中加入臀圍或腰臀圍比皆無法提高調整後決定係數值。
結論：本研究中所使用的四種生物電阻抗分析儀所測得之體脂率數值與以重水稀釋法測得之體脂率雖具有一致性但仍有系統性誤差，宜於線性迴歸模式中校正之。

Background: Deuterium oxide (D2O) dilution method is the golden standard of measuring the total body water (TBW) to deduce the percent body fat (PBF) in the 2-compartment model of body composition. Clinically, BIA (bioelectrical impedance analysis) is the most common way for measuring PBF. However, most BIA devices come from foreign, especially Western, manufacturers, and there has been no study examining the consistency between the PBF obtained by a BIA device and the one by D2O dilution in Taiwanese.
Methods: A total of 358 Taiwanese outpatients (male/female=137/221, mean age=34.1±9.9 years old, and mean body mass index=30.9±9.6 kg/m2) visiting National Cheng Kung University Hospital were enrolled. After overnight fasting and bladder emptying, the subjects had their PBF values measured by D2O dilution and four frequently used BIA devices. Statistical and data analyses were conducted using SPSS 17th version.
Results: For the 358 subjects, the mean PBF by D2O dilution (D2O-PBF) was 33.5±10.6%, significantly different from the one measured by the four BIA devices (BIA-PBF). The results of Bland-Altman analysis indicated an overall acceptable consistency between BIA-PBF and D2O-PBF, with the only exception occurring in the case of Tanita TBF-410 that reported a slight overestimation when the average PBF of the subjects exceeded 50%. In terms of multiple linear regression, when D2O-PBF was set as the dependent variable and BAI-PBE, 1/BMI, gender, age, and waist circumference as the independent variables, the adjusted coefficient of determination (R2) fell in the range of 0.430-0.755 and remained consistent even after hip circumference or waist-to-hip ratio was added into the regression model.
Conclusion: While a high degree of consistency between D2O-PBF and BIA-PBF is verified in the study, there are evident systematic errors that should be corrected by the linear regression model appropriately.