Department of Pediatrics, University of Miami Miller School of Medicine, FL 33101, USA.
Physiol Meas. 2010 Apr;31(4):497-512. doi: 10.1088/0967-3334/31/4/003. Epub 2010 Feb 24.
The hydration of fat-free mass, or hydration fraction (HF), is often defined as a constant body composition parameter in a two-compartment model and then estimated from in vivo measurements. We showed that the widely used estimator for the HF parameter in this model, the mean of the ratios of measured total body water (TBW) to fat-free mass (FFM) in individual subjects, can be inaccurate in the presence of additive technical errors. We then proposed a new instrumental variables estimator that accurately estimates the HF parameter in the presence of such errors. In Monte Carlo simulations, the mean of the ratios of TBW to FFM was an inaccurate estimator of the HF parameter, and inferences based on it had actual type I error rates more than 13 times the nominal 0.05 level under certain conditions. The instrumental variables estimator was accurate and maintained an actual type I error rate close to the nominal level in all simulations. When estimating and performing inference on the HF parameter, the proposed instrumental variables estimator should yield accurate estimates and correct inferences in the presence of additive technical errors, but the mean of the ratios of TBW to FFM in individual subjects may not.
去脂体重的水合作用,或水合分数(HF),通常在双 compartment 模型中被定义为一个恒定的身体成分参数,然后根据体内测量来估计。我们表明,该模型中广泛使用的 HF 参数估计器,即个体中测量的总身体水分(TBW)与去脂体重(FFM)比值的平均值,在存在附加技术误差的情况下可能不准确。然后,我们提出了一种新的工具变量估计器,即使存在此类误差,它也可以准确地估计 HF 参数。在蒙特卡罗模拟中,TBW 与 FFM 的比值的平均值是 HF 参数的不准确估计器,并且在某些条件下,基于它的推断的实际第一类错误率超过了名义 0.05 水平的 13 倍以上。工具变量估计器在所有模拟中都是准确的,并且保持了接近名义水平的实际第一类错误率。在估计和进行 HF 参数推断时,在存在附加技术误差的情况下,建议的工具变量估计器应该会产生准确的估计值和正确的推断,但是个体中 TBW 与 FFM 的比值的平均值可能无法做到这一点。