Jaydeep Sinha, Hesham Al-Sallami, Stephen Duffull
Otago Pharmacometrics Group, School of Pharmacy, University of Otago, Dunedin, New Zealand
Objectives:
The FFM model developed by Janmahasatian et al. (FFMJan) [1] has been extensively used in clinical pharmacology studies. The FFMJan model was developed based on a linear relationship (ZJan model) between bioimpedance (Z) and body mass index (BMI) in a population that descended from European ancestry. Bioimpedance (Z) essentially reflects the body composition, which is known to vary widely between different ethnicities [2,3]. For example, at a fixed BMI, the estimated body fat percentage of Asian Indians was reported to be 7.6% (95% CI: 5.0% – 10.2%) higher on average than Europeans [2]. This indicates that that FFMJan that follows the European body composition pattern would be anticipated to over-estimate FFM in Asian Indians. A recent validation study by Srigiripura et al. [4] has reported that the FFMJan model over-predicted FFM in the Indian population when compared to the measured FFM data. Therefore, there is a need to extend the FFMJan model in order to account for inter-ethnic difference in the body composition. An extended model would incorporate ethnic specific parameters.
Objective-1: To derive an extended version of the FFM model (FFMExt) that would accommodate estimable ethnicity specific parameter(s).
Objective-2: To estimate the ethnic specific parameters for an Indian population (FFMExt(Ind)).
Methods:
Objective-1: In the proposed extended FFM model (FFMExt), the relationship between Z (ZExt) and BMI was relaxed to accommodate a non-linear relationship by incorporating a set of ethnic-specific composition parameters Ψ {ψ1, ψ2, ψ3} to the coefficients of ZJan model. An extended equation for bioimpedance, ZExt, was developed and combined with the existing FFMJan model. By rearranging the equation, the final equation of FFMExt was obtained which contains body composition parameters Ψ {ψ1, ψ2, ψ3}.
Objective-2: The Ψ parameters of the FFMExt model were estimated against predictions from a reference model that was developed specifically for the Indian population by Kulkarni et al. [5]. FFM (FFMKul) was simulated based on demographic data (age, sex, height, and weight) of 100 adult Indian medical patients. The Ψ parameters were estimated using NONMEM (version 7.3). Model selection was based on the likelihood ratio test (LRT). A visual plot of the reference FFMKul vs. the predicted (FFMExt(Ind)) data was used to evaluate the (FFMExt(Ind)) model.
Results:
Objective 1, the final equation of the FFMExt model was derived and is shown in Eq 1 and 2. When all values of Ψ = 1 the FFMExt equation simplifies to the existing FFMJan model.
For males:
FFMExt = 9270*WT / [ψ1 *216*BMI + ψ2*6680*BMI (1- ψ3)] (1)
For females:
FFMExt = 9270*WT / [ψ1 *244*BMI + ψ2*8780*BMI (1- ψ3)] (2)
In Objective 2, an initial comparison between the two models revealed that the FFMJan model over-predicted the reference model (FFMKul). The mean differences in prediction (FFMKul – FFMJan) [95% CI] were -1.9 [(-2.6) – (-1.2)] kg and 0.3 [(-0.2) – 0.9] kg in male and female subjects respectively, and the root mean squared errors (RMSE) were 3 kg and 1.6 kg in male and female respectively. The final FFMExt(Ind) had two estimable ethnic-specific parameters (ψ1 was estimated to be 0). Sex was identified as a significant covariate on ψ2, which was the proportionality term, but not on ψ3, the exponent. The estimates (%RSE) of ψ2 were 0.77 (3.2%) and 0.70 (3.3%) for males for females respectively, and ψ3 was 0.72 (1.3%). Note, for the Indian population the first term of the denominator drops as ψ1 = 0. The final models of FFMExt(Ind) have been shown in Eq 3 and 4.
For males: FFMExt(Ind) = 9270*WT / [0.77*6680*BMI 0.28] (3)
For females: FFMExt(Ind) = 9270*WT / [0.70*8780*BMI 0.28] (4)
Conclusions:
This work extends the Janmahasatian’s FFM model to prediction of FFM in other populations by incorporating ethnicity specific correction factor(s). Importantly, the correction factor(s) can be directly estimated from bioimpedance (Z) measurements, and there is no need to measure FFM in the respective population, which has significant benefits for future clinical pharmacology studies.
References:
[1] Janmahasatian S et al. Clinical Pharmacokinetics. 2005. 44(10): p1051-65.
[2] Rush EC et al. The New Zealand Medical Journal. 2004. 117(1207): p1-9
[3] Rush EC et al. British Journal of Nutrition. 2009. 102(4): p632-41
[4] Srigiripura et al. Int J Nutr Pharmacol Neurol Dis. 2017. 7(4): p94-100.
[5] Kulkarni et al. J Appl Physiol. 2013. 115(8): p1156-1162.
Reference: PAGE 28 (2019) Abstr 9020 [www.page-meeting.org/?abstract=9020]
Poster: Methodology - Other topics