Gregory Haber1, Joshua Sampson1, Katherine M Flegal2, Barry Graubard1. 1. Biostatistics Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, MD, USA. 2. Prevention Research Center, Stanford University Medical School, Stanford, CA, USA.
Abstract
BACKGROUND: Several studies have assessed the relation of body composition to health outcomes by using values of fat and lean mass that were not measured but instead were predicted from anthropometric variables such as weight and height. Little research has been done on how substituting predicted values for measured covariates might affect analytic results. OBJECTIVES: We aimed to explore statistical issues causing bias in analytical studies that use predicted rather than measured values of body composition. METHODS: We used data from 8014 adults ≥40 y old included in the 1999-2006 US NHANES. We evaluated the relations of predicted total body fat (TF) and predicted total body lean mass (TLM) with all-cause mortality. We then repeated the evaluation using measured body composition variables from DXA. Quintiles and restricted cubic splines allowed flexible modeling of the HRs in unadjusted and multivariable-adjusted Cox regression models. RESULTS: The patterns of associations between body composition and all-cause mortality depended on whether body composition was defined using predicted values or DXA measurements. The largest differences were observed in multivariable-adjusted models which mutually adjusted for both TF and TLM. For instance, compared with analyses based on DXA measurements, analyses using predicted values for males overestimated the HRs for TF in splines and in quintiles [HRs (95% CIs) for fourth and fifth quintiles compared with first quintile, DXA: 1.22 (0.88, 1.70) and 1.46 (0.99, 2.14); predicted: 1.86 (1.29, 2.67) and 3.24 (2.02, 5.21)]. CONCLUSIONS: It is important for researchers to be aware of the potential pitfalls and limitations inherent in the substitution of predicted values for measured covariates in order to draw proper conclusions from such studies. Published by Oxford University Press on behalf of the American Society for Nutrition 2021.
BACKGROUND: Several studies have assessed the relation of body composition to health outcomes by using values of fat and lean mass that were not measured but instead were predicted from anthropometric variables such as weight and height. Little research has been done on how substituting predicted values for measured covariates might affect analytic results. OBJECTIVES: We aimed to explore statistical issues causing bias in analytical studies that use predicted rather than measured values of body composition. METHODS: We used data from 8014 adults ≥40 y old included in the 1999-2006 US NHANES. We evaluated the relations of predicted total body fat (TF) and predicted total body lean mass (TLM) with all-cause mortality. We then repeated the evaluation using measured body composition variables from DXA. Quintiles and restricted cubic splines allowed flexible modeling of the HRs in unadjusted and multivariable-adjusted Cox regression models. RESULTS: The patterns of associations between body composition and all-cause mortality depended on whether body composition was defined using predicted values or DXA measurements. The largest differences were observed in multivariable-adjusted models which mutually adjusted for both TF and TLM. For instance, compared with analyses based on DXA measurements, analyses using predicted values for males overestimated the HRs for TF in splines and in quintiles [HRs (95% CIs) for fourth and fifth quintiles compared with first quintile, DXA: 1.22 (0.88, 1.70) and 1.46 (0.99, 2.14); predicted: 1.86 (1.29, 2.67) and 3.24 (2.02, 5.21)]. CONCLUSIONS: It is important for researchers to be aware of the potential pitfalls and limitations inherent in the substitution of predicted values for measured covariates in order to draw proper conclusions from such studies. Published by Oxford University Press on behalf of the American Society for Nutrition 2021.
Entities:
Keywords:
Berkson error; all-cause mortality; bias; body composition; epidemiology; prediction equations; research methods
Authors: Nathaniel Schenker; Lori G Borrud; Vicki L Burt; Lester R Curtin; Katherine M Flegal; Jeffery Hughes; Clifford L Johnson; Anne C Looker; Lisa Mirel Journal: Stat Med Date: 2010-11-30 Impact factor: 2.373
Authors: Dong Hoon Lee; NaNa Keum; Frank B Hu; E John Orav; Eric B Rimm; Walter C Willett; Edward L Giovannucci Journal: Eur J Epidemiol Date: 2018-08-16 Impact factor: 8.082
Authors: Dong Hoon Lee; Leandro Fórnias Machado de Rezende; José Eluf-Neto; Kana Wu; Fred K Tabung; Edward L Giovannucci Journal: Int J Cancer Date: 2019-01-24 Impact factor: 7.396
Authors: Hee-Kyung Joh; Edward L Giovannucci; Kimberly A Bertrand; Soo Lim; Eunyoung Cho Journal: J Natl Cancer Inst Date: 2013-04-08 Impact factor: 11.816
Authors: Shinyoung Jun; Alexandra E Cowan; Johanna T Dwyer; Wayne W Campbell; Anna E Thalacker-Mercer; Jaime J Gahche; Regan L Bailey Journal: J Nutr Date: 2021-12-03 Impact factor: 4.687
Authors: Victoria Rosberg; Julie Kk Vishram-Nielsen; Anna M Dyrvig Kristensen; Manan Pareek; Thomas S G Sehested; Peter M Nilsson; Allan Linneberg; Luigi Palmieri; Simona Giampaoli; Chiara Donfrancesco; Frank Kee; Giuseppe Mancia; Giancarlo Cesana; Giovanni Veronesi; Guido Grassi; Kari Kuulasmaa; Veikko Salomaa; Tarja Palosaari; Susana Sans; Jean Ferrieres; Jean Dallongeville; Stefan Söderberg; Marie Moitry; Wojciech Drygas; Abdonas Tamosiunas; Annette Peters; Hermann Brenner; Ben Schöttker; Sameline Grimsgaard; Tor Biering-Sørensen; Michael H Olsen Journal: Prev Med Rep Date: 2022-01-27