Establishing age/sex related serum creatinine reference intervals from hospital laboratory data based on different statistical methods

Abstract

Background

This is a retrospective study on a large hospital database to establish age- and sex-related mean values and reference ranges for serum creatinine (Scr), obtained with an IDMS-traceable, enzymatic method, in a Caucasian population.

Methods

The database was filtered for unique entries to reduce the presence of correlated and pathological data. Three different statistical methods, a non-parametric method, the Bhattacharya procedure and a non-linear fit of the cumulative Gaussian distribution were used to estimate the serum creatinine–age dependency for men and women, from birth till 100 years of age.

Results

Scr increases with age, equal for boys and girls, up to 14 years and with a much steeper slope for boys than for girls between 14 and 20 years. We show that the Scr–age pattern is constant between 20 and 70 years with a mean of 0.90 mg/dL [0.63–1.16 mg/dL] for men and 0.70 mg/dL [0.48–0.93 mg/dL] for women. Above 70, Scr starts to slowly increase again.

Conclusions

Indirect methods confirm the available reference intervals from healthy-volunteer studies and add information on age-periods not covered by these studies. As such, indirect methods can be used complementary to healthy-volunteer studies.

Introduction

Although the use of estimated glomerular filtration rate (eGFR) from statistical models, such as the Modification of Diet in Renal Disease (MDRD) Study equation [1], [2], is getting more and more common to assess renal function, reference values for serum creatinine (Scr) may still be necessary to compare patients' results. The reason for this is that statistical models are based on assumptions and have their own limitations. For instance, the MDRD Study equation, which is recommended for diagnosing Chronic Kidney Disease (CKD) by US, UK, New Zealand, Australia and guidelines of other countries, is only valid in adults between 20 and 65–75 years old, with decreased renal function (eGFR  < 60 ml/min/1.73 m²) [3], [4], [5]. This means that children, the elderly and healthy individuals are not included in the model. Scr reference values may be helpful in these age/sex subgroups to spot abnormalities.

In the past few years, many publications have addressed the topic of the diversity of Scr assays [6], [7], each with their specific problems. Most clinical labs are using an alkaline picrate assay (so-called Jaffe method). One major issue was that the alkaline picrate method has shown a considerable lack of specificity. Attempts to correct for this nonspecificity resulted in the use of many different assays among which a compensated Jaffe method, where a correction for a constant bias as compared to the Isotope Dilution Mass Spectrometry (IDMS) reference method has been introduced. Although many attempts to solve this specificity problem, the current tendency amongst clinical labs is the increasing use of enzymatic methods to determine Scr. These methods have the advantage of being IDMS-traceable, that is, mathematical correction factors are no longer necessary. Although not current common practice, the scientific community is clear in its recommendation to use enzymatic assays for the determination of Scr [6]. As a consequence, the number of publications reporting reference values for Scr, determined by enzymatic assays, is very limited up till now, and can only date back to 2002, the time when sufficient commercial enzymatic assays became available. This also demonstrates one of the drawbacks of so-called healthy-volunteers' studies, which are recommended by the IFCC to determine reference values [8], [9]: if the biochemical method improves or changes, the study has to be repeated. With the continuous evolution of measurement procedures, this process has to be repeated too often. In such studies, the “a priori” selection of reference individuals for obtaining reference values is required. This selection is difficult, costly and time consuming: physical examinations, laboratory tests and a medical interview based on a questionnaire should guarantee the health status of the selected individuals. There is no doubt that this reduces the risk of including abnormal values in the data, but it can never be 100% guaranteed. Clinical laboratories, seeking accreditation for compliance with ISO 15189:2003 need to demonstrate that the physiological reference intervals are appropriate for the patient population served and for their measurement systems as well. Giving the cost, time and difficulty of healthy-volunteer studies, few labs do so. Most of them use the reference values from scientific or commercial literature. Literature for reference values for Scr measured by an enzymatic method and obtained according to the IFCC guidelines is quite limited, as shown by Ceriotti et al. [10]. In their paper they referred to only one paper with pediatric data [11] and 5 reports with adult data, of which three of them were on the same group of subjects and had to be considered together. From the remaining three reports on adults, only two were obtained in a European population [12], [13], the other in an Australian white population [14]. The results were comparable, giving reference limits of 0.67–1.19 mg/dL for men and 0.51–1.02 mg/dL for women (smallest and largest values of three reference papers are presented here). These data were not able to cover the complete age-range, so clear gaps are observed: newborns (although average values for neonates between 0 and 21 days are given), the adolescent period (15–20 years) and the elderly (above 65–75 years old). For these age groups, there are no recent IDMS-traceable Scr reference values available.

Indirect methods have been described to produce reference intervals [15], [16], [17]. These methods use mathematical and statistical procedures, sometimes in combination with exclusion criteria, on retrospective data of patients obtained daily in the laboratory. These methods ‘try’ to differentiate between pathological and non-pathological results. Although, we do not see these methods as a complete alternative for IFCC recommendations to produce reference intervals, they can be complementary, filling the gaps where healthy-volunteer study results are lacking. On the other hand, healthy-volunteer results may be used to confirm assumptions on which indirect methods are based; as such, these healthy-volunteer results can be used to validate the indirect method. Indirect methods have the advantage that they are cheap, fast and can be repeated at any time, acting as self-validating. When the measurement method evolves, indirect methods can be used to generate new reference intervals, as soon as enough data from the new measurement method is available. Moreover, data mining should not only be seen as a means to obtain reference intervals. Median or means of the data are much more robust than the tailing reference values, especially if the majority of the age/sex subgroup distribution is coming from healthy people. Then, trends may be observed which may help to understand the evolution of Scr with age or sex, in newborns, young adolescents or the elderly.

In this study we present age/sex reference intervals for Scr, obtained from an enzymatic assay, in a Caucasian population, spanning the complete lifetime.