Premium
Kidney function changes with aging in adults: comparison between cross‐sectional and longitudinal data analyses in renal function assessment
Author(s) -
Chung Sang M.,
Lee David J.,
Hand Austin,
Young Philip,
Vaidyanathan Jayabharathi,
Sahajwalla Chandrahas
Publication year - 2015
Publication title -
biopharmaceutics and drug disposition
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.419
H-Index - 58
eISSN - 1099-081X
pISSN - 0142-2782
DOI - 10.1002/bdd.1988
Subject(s) - renal function , linear regression , population , random effects model , regression analysis , statistics , creatinine , sample size determination , analysis of variance , regression , medicine , longitudinal study , repeated measures design , demography , mathematics , meta analysis , environmental health , sociology
The study evaluated whether the renal function decline rate per year with age in adults varies based on two primary statistical analyses: cross‐section (CS), using one observation per subject, and longitudinal (LT), using multiple observations per subject over time. A total of 16628 records (3946 subjects; age range 30–92 years) of creatinine clearance and relevant demographic data were used. On average, four samples per subject were collected for up to 2364 days (mean: 793 days). A simple linear regression and random coefficient models were selected for CS and LT analyses, respectively. The renal function decline rates per year were 1.33 and 0.95 ml/min/year for CS and LT analyses, respectively, and were slower when the repeated individual measurements were considered. The study confirms that rates are different based on statistical analyses, and that a statistically robust longitudinal model with a proper sampling design provides reliable individual as well as population estimates of the renal function decline rates per year with age in adults. In conclusion, our findings indicated that one should be cautious in interpreting the renal function decline rate with aging information because its estimation was highly dependent on the statistical analyses. From our analyses, a population longitudinal analysis (e.g. random coefficient model) is recommended if individualization is critical, such as a dose adjustment based on renal function during a chronic therapy. Copyright © 2015 John Wiley & Sons, Ltd.