Centile Charts II: Alternative Nonparametric Approach for Establishing Time-Specific Reference Centiles and Assessment of the Sample Size Required

Abstract

Background: Reference intervals, and more generally centile estimates, are used to characterize a reference population for the purposes of interpreting an individual patient's clinical measurement. We describe methods of calculating reference intervals where these centiles vary with a covariate, usually age or time. Methods: The US Food and Drug Administration and the IFCC have made recommendations on two approaches: the parametric approach, which models the structural characteristics of the data set with a theoretical distribution, and the nonparametric approach, which makes no particular assumption about this structure. In this report we propose a nonparametric procedure that relies on the principles of regression and show how sample size determination can be assessed. We also show how the sample size calculation is influenced by the distribution of the times measured. Results: We illustrated our method on three data sets and compared the results for our proposed nonparametric method with parametric estimates. We showed that the bias is reduced and that the nonparametric method is less likely to produce fluctuating profiles. Conclusions: To achieve adequate precision the sample size needs to be larger than 120, as has often been recommended. If there is doubt about the parametric model, then threshold sample sizes may need to be as high as 500.

Laboratory Coagulation Parameters in Extremely Premature Infants Born Earlier than 27 Gestational Weeks upon Admission to a Neonatal Intensive Care Unit

“…Reference ranges were defined as the 2.5th-97.5th centile of the population sample [21]. Using these reference intervals, PT, APTT and fibrinogen reference ranges were determined to be 14.4-36.7 and 40.5-158.5 s and 0.7-4.8 g/l, respectively.…”

A Non‐parametric Conditional Bivariate Reference Region with an Application to Height/Weight Measurements on Normal Girls

“…If the age dependence of the distribution is relevant, which is clearly the case when constructing a weight reference, a conditional reference curve (or reference chart) is needed. Much of the literature on reference charts is on such conditional references -for example (Koenker and Bassett, 1978;Cole, 1988;Altman, 1993;Altman and Chitty, 1994;Wright and Royston, 1997;Morrell et al, 1997;Yu and Jones, 1998;Heagerty and Pepe, 1999;Griffiths et al, 2004).…”

Time-Related Centile Ranges for Quality of Life Outcomes in Renal Transplantation

“…The range of normal for a given variable is typically defined as the range of values for that variable in the population of interest that fall between the 2.5 th and the 97.5 th percentile (Griffiths, Iles, Koduah, & Nix, 2004). A basic assumption must be that the population from which the reference range is derived is representative of "typical" individuals and therefore the values for that variable in that population are considered to be the normal range.…”

“…Reference ranges, often referred to as normal ranges for measurements, have long been used in health care to evaluate the health status of patients. The range of normal for a particular variable is typically defined as the area between the 2.5 th and the 97.5 th percentiles (Griffiths, Iles, Koduah, & Nix, 2004). The assumption is made that the reference population in which the variable is measured is a population representative of "normal" or "healthy" individuals, thus the term normal range.…”

“…The range of normal for a given variable is typically defined as the range of values for that variable that fall between the 2.5 th and the 97.5 th percentile in the population of interest (Griffiths, Iles, Koduah, & Nix, 2004). The assumption is made that the reference population in which the variable is measured is a healthy population representative of "normal" or "healthy" individuals, thus the term normal range.…”