Background: Reference intervals for serum creatinine remain relevant despite the current emphasis on the use of the estimated glomerular filtration rate for assessing renal function. Many studies on creatinine reference values have been published in the last 20 years. Using criteria derived from published IFCC documents, we sought to identify universally applicable reference intervals for creatinine via a systematic review of the literature.
Methods: Studies were selected for inclusion in the systematic review only if the following criteria were met: (a) reference individuals were selected using an “a priori” selection scheme, (b) preanalytical conditions were adequately described; (c) traceability of the produced results to the isotope dilution–mass spectrometry (IDMS) reference method was demonstrated experimentally, and (d) the collected data received adequate statistical treatment.
Results: Of 37 reports dealing specifically with serum creatinine reference values, only 1 report with pediatric data and 5 reports with adult data met these criteria. The primary reason for exclusion of most papers was an inadequate demonstration of measurement traceability. Based on the data of the selected studies, we have collated recommended reference intervals for white adults and children.
Conclusion: Laboratories using methods producing traceable results to IDMS can apply the selected reference intervals for serum creatinine in evaluating white individuals.
The usefulness of plasma or serum creatinine measurements for the identification of renal insufficiency is hampered by its covariation with sex, age, race, diet, and muscle mass of each individual in whom it is being measured (1)(2). Moreover, the analytical quality of the test is often suboptimal due to the nonspecificity of the Jaffe method used in most routine laboratories (3).
The recent campaign of the National Kidney Disease Education Program, initiated in 2000 by the US National Institutes of Health (4), has led to increased focus on the measurement of creatinine in serum (5). Concurrently, there has been an international effort to improve creatinine standardization by assuring traceability to the reference system (6). These efforts highlighted limitations of commercially available methods for creatinine measurement as well as the differences in results among methods resulting from a lack of assay standardization and the nonspecificity of the alkaline picrate method used in the majority of clinical laboratories (3)(7)(8). A further difficulty associated with these efforts is the need for development of scientifically sound reference intervals for creatinine assays once they have been aligned with high-order reference standards. In fact, the Modification of Diet in Renal Disease (MDRD) formula is validated only for adults (>18 years) and those with renal disease (9), and thus there is still the need for reference intervals. The variation of creatinine concentration with age and sex makes the task of defining scientifically sound reference intervals very demanding.
To determine whether data obtained using standardized assays already exist in the peer-reviewed literature, which could thus be endorsed for global application, the IFCC Committee on Reference Intervals and Decision Limits (C-RIDL) undertook a systematic review of the literature. The presence of clear demonstration of the traceability of the results to higher-order reference methods (10)(11) or reference materials was the main criterion that guided our work.
Materials and Methods
We performed a literature search on the Medline database for the period January 1987 to January 2007 because reference measurement procedures were not described until the mid-1980s (10)(11). Separate searches were performed for the terms “reference interval(s),” “reference values,” “reference range(s),” and “normal values” with “title” as tag term. All these searches were crossed (operator AND) with the search for the term “creatinine OR creatininium” (field limit “humans”). A further check was performed crossing (operator AND) “creatinine OR creatininium” (field limit “humans”) with “reference intervals” (without any field limit). Other references were added by manual search. The report of Schlebusch et al. (12), which was not indexed in Medline, was provided by one of the authors (G.K.).
The selection of studies to be included in this systematic review was guided by the following 4 criteria derived from the IFCC recommendations on reference intervals (13)(14)(15)(16)(17) and the Clinical Laboratory Standards Institute (CLSI) C28-A2 document (18).
Criterion 1: selection of reference individuals.
Criterion 2: study design.
Criterion 3: analytical correctness.
Accurate description of the measurement method used including quality control procedures and data on analytical performance. Experimental demonstration of traceability of the results produced to the isotope dilution–mass spectrometry (IDMS) reference methods (10)(11).
Criterion 4: statistical treatment of the collected data.
Selection of an adequate number of subjects for determination of reference intervals by nonparametric statistical methods, with appropriate partitioning.
evaluation of studies
All retrieved reports (n = 304) were first screened by 1 of the authors (F.C.), who selected for further evaluation an initial set of 37 studies specifically dealing with creatinine reference intervals in serum. These reports were reviewed again and confirmed for inclusion in the analysis by 3 other authors (J.B., J.Q., and J.H.) according to the above criteria.
From the initially selected 37 studies (see Supplemental Data), 8 were discarded because they used data mining (indirect) methods to derive reference intervals. Sixteen studies did not adequately demonstrate traceability of creatinine results to the reference measurement system, 2 had an insufficient number of subjects, 2 dealt only with neonates, and 3 could not be accessed (not available in the libraries of the authors or from the Internet). In this evaluation, there were no conflicting opinions among the reviewers.
Six reports contained information demonstrating traceability to the reference method (12)(19)(20)(21)(22)(23). Only 1 of these dealt with the pediatric population (12). Three were on the same group of subjects and can be considered together (21)(22)(23). All these reports were found to comply with all the criteria and obtained very similar results (Table 1⇓ ), but we selected the report of Junge et al. (19) for final consideration because it gave a better description of the experiments performed to demonstrate traceability of the results. All original data and some relevant missing information were requested directly from the authors.
Although no study of children was found to be in full compliance with either the defined analytical or statistical criteria, the paper of Schlebusch et al. (12) was further considered, and the authors were contacted to obtain more detailed information on subject selection, method performance, and statistical methods. We received all original data, and the age-specific reference intervals were derived following the approach suggested by Royston and Wright (24) with a nonparametric method (25).
selection of the reference individuals
Junge et al. (19) a priori selected 252 volunteers (age 18–74 years) in whom health status was checked by a brief physical examination, laboratory tests (standard clinical chemistry profile and urinalysis), and a medical interview based on a questionnaire. Exclusion criteria for the chemistry profile were C reactive protein >10 mg/L; γ-glutamyl transferase >132 U/L for males, >78 for females; and alanine aminotransferase >100 U/L for males, >70 U/L for females. All the other evaluated tests (cholinesterase, hemoglobin, leukocytes, platelets) were within the established reference intervals. No special dietary recommendations were given, and individuals taking drugs were excluded.
In the original paper of Schlebusch et al. (12), only a partial description of the selected population of 521 children was provided, and no information was furnished on cord blood donors and neonates; however, the authors sent us the following additional details. Children included in the study were carefully selected over a 2-year period from those presenting to the allergy ward. Subjects presenting with massive eczema were excluded. Other general clinical criteria for exclusion included the known presence of chronic renal insufficiency, acute renal failure, glomerulonephritis, interstitial nephritis, cystic kidneys, renovascular disorders, diabetes, myopathies, iron deficiency anemia, iron overload, acute and chronic infection, and pancreatic disorders.
In the group of preterm neonates, exclusion criteria included infusion therapy or treatment with indomethacin for the occlusion of ductus arteriosus. Most preterm neonates were treated with antibiotics beginning the third day.
Junge et al. (19) described sample collection and handling. Analyses were performed on fresh serum samples. The subjects were not fasting (G.K., personal communication), but fasting status does not seem to influence creatinine results (26).
Schlebusch et al. (12) gave no detailed information regarding sample collection. Sera were frozen at −20 °C and analyses were performed up to 2 years later. Creatinine is a very stable analyte in serum (27). Moreover, the use of an enzymatic method eliminates problems related to possible interfering substances generated during storage (28).
Junge et al. (19) used an enzymatic method (Creatinine Plus, Roche Diagnostics) on an automatic analyzer (Roche Hitachi 717). This method is based on the same reaction sequence as the Creatinine-PAP (phenol aminophenazone) method (29) consisting of 4 consecutive enzymatic steps via creatininase, creatinase, sarcosine oxidase, and peroxidase, but using a different chromophore (2,4,6-triiodo-3-hydroxybenzoic acid) and a modified detergent composition. The method was standardized by use of 6 human serum pools with values assigned by the Institute for Clinical Biochemistry, University of Bonn, Germany, using an IDMS reference method (10) (G.K., personal communication). This method is included in List I of the Joint Committee on Traceability in Laboratory Medicine (JCTLM) related to the currently available reference methods (30). Indicated quality control data showed intra- and interassay imprecision (CV) of 1.4% and 1.9%, respectively, at a creatinine concentration of 88 μmol/L.
Schlebusch et al. (12) used the same analytical method (Creatinine Plus) on a different analyzer (Roche Hitachi 911). The instrument was calibrated with the manufacturer’s calibrators, but no experiment was performed to demonstrate traceability to the reference procedure. However, Miller et al. have demonstrated the accuracy of this assay on the Roche Hitachi 911 in an external quality assessment study (bias 0.09 μmol/L at a concentration of 80 μmol/L) (7). A recent report on the recalculation of the Modification of Diet in Renal Disease (MDRD) study equation has confirmed the excellent correlation of this method with IDMS (IDMS = 1.0 × Roche – 3.01 μmol/L) (31). Further evidence confirming the accuracy of this assay has been generated in a study involving 172 laboratories from 6 European countries, performed under the auspices of the European Community Confederation of Clinical Chemistry and Laboratory Medicine (EC4) (32). Data from the Italian branch of this study, involving 3 laboratories using the Roche enzymatic method, showed a mean bias of +2.6 μmol/L at a concentration of 76 μmol/L when results were compared with the IDMS value (8). No information related to the assay reproducibility was found in the original report, but the data provided by the authors indicated a within-run CV (21 replicates) of 1.7% at 48 μmol/L (using a serum pool) and of 1.0% at 88 μmol/L (using Roche control material Precinorm U). Note that in the study, the analyses were performed in duplicate.
statistical treatment of the collected data
Schlebusch et al. (12) reported the 2.5th and 97.5th percentile limits of the value distribution. The publication did not report what statistical approach was used in the calculation of these population centiles or results regarding the normality of the value distributions in the different age groups; the authors communicated to us that “STATEX,” an internal Roche computer software that employs a nonparametric approach, was used for calculation. Thus, using the original data, we recalculated the 2.5th and 97.5th percentiles using a nonparametric approach and SAS software (25). The resulting numbers, shown in Table 3⇑ , were identical in most cases, but not always. Moreover, age-specific reference intervals were derived following the approach suggested by Royston and Wright (24). In brief, each age was incremented by 0.01 years to avoid taking the logarithm of zero. The 471 derived ages and corresponding reference values (cord blood data were not used for this calculation) were fitted using fractional polynomials to derive a function for age-adjusted mean creatinine. In a second step, the residual deviations of each creatinine value from the fitted age-adjusted mean creatinine were examined for normality. Once approximate normality was confirmed, the absolute values of these residual deviations were fitted using fractional polynomials to derive a function for age-adjusted mean absolute residual. Fractional polynomials were calculated using the SAS programs developed by Sauerbrei et al. (33). Multiplying by the square root of π/2, the mean absolute residuals were converted to standard deviations following the approach suggested by Altman (34). Addition and subtraction of 1.96 times the age-adjusted standard deviation to the age-adjusted mean creatinine provided age-adjusted upper and lower limits for the reference interval.
Creatinine concentrations expressed in mg/dL and μmol/L were fitted separately.
Age-adjusted mean creatinine was best fitted by the following second-order fractional polynomial functions.
Mean creatinine (μmol/L)age = −2.37330 − 12.91367 ·loge(age) + 23.93581 · (age)0.5
Mean creatinine (mg/dL)age = −0.02324 − 0.14545 ·loge(age) + 0.26964 · (age)0.5
The cumulative distribution of the residuals appeared nearly linear on normal probability paper, confirming an approximately normal distribution. The age-adjusted standard deviation was derived using the following best-fit second-order fractional polynomial functions.
SD (μmol/L)age = √(π/2) · 4.20393 − 2.44027 ·loge(age) + 0.59763 · age0.5
SD (mg/dL)age = √(π/2) · 0.00649 − 0.03621 · loge(age) + 0.04124 · age
Age-adjusted reference intervals were calculated using mean creatinineage ± 1.96 · SDage.
Although age exerts a highly significant effect on creatinine (P <0.0001), sex and its interaction term with age do not exert any detectable effects (P = 0.9523 and P = 0.485, respectively).
comparison with other reference intervals reported in the literature
Table 1⇑ compares results from published studies evaluating reference intervals for serum creatinine concentrations in adults obtained with methods traceable to the reference system (19)(20)(21)(22). The data obtained by Mazzachi et al. (20) and Rustad and colleagues (21)(22) are quite similar to those of Junge et al. (19). In the populations studied by these investigators (all white individuals), the reference intervals appear to be very similar from Northern Europe to Australia. A multicenter study recently performed in Spain on 248 women and 220 men gave similar results for men (64–106 μmol/L), but slightly lower results for women (52–85 μmol/L) (35). That study, however, used a compensated alkaline picrate method. No reliable data are available for other racial/ethnic groups.
The suggested reference intervals for children, although hampered by some weaknesses, appear to us acceptable, because the data on cord blood are very similar to those of the adult female population obtained in other studies, and the trend of the values with age is quite similar to that reported by other authors using assays that have not been standardized (36)(37)(38).
The literature search on the subject of creatinine reference intervals demonstrated that, despite the relevance of the subject, very few papers fulfill the criteria necessary to endorse the obtained reference intervals for global application. Nevertheless, we think that the reference intervals reported in Tables 2⇑ and 3⇑ may be adopted by any laboratory serving a similar population and using a method of comparable specificity that is traceable to the creatinine reference system. The comparability of the results obtained in individuals coming from different countries (Germany, Scandinavia, Australia) reinforces the conclusion that the proposed reference intervals are universally valid for whites. Due to the multiethnicity of the population, especially in larger cities, a validation of the applicability of the proposed interval to the population served using the approach proposed by the CLSI C28 document (18) is highly recommended. Although it is known that serum creatinine concentrations are higher in those of African descent (9), no literature data on reference intervals fulfilling our criteria were available in either African or Asian populations.
We elected to consider only data obtained with enzymatic assays because of the higher specificity of this analytical approach. It is well demonstrated that the subtraction of 18–25 μmol/L (19)(23) to eliminate protein-related unspecific interference on alkaline picrate assays significantly improves the correlation of these assays with enzymatic ones. In this situation, the obtained reference intervals are very similar to those of the enzymatic methods (19)(20)(21)(23)(35) as indicated in the note for Table 1⇑ . However, on individual samples, especially at the low creatinine concentrations found in children, large differences can be seen. In agreement with the IFCC Working Group on Standardization of Glomerular Filtration Rate Assessment, “the use of assays that are more specific for serum creatinine, such as those based on some enzymatic procedures, may provide more reliable estimated GFR values” (6). Only assays using enzymatic principles have the analytical specificity to guarantee traceability of each individual result to the reference measurement system for creatinine measurement. Care must be exercised when using enzymatic methods with icteric samples, however, because a negative bias can occur (39).
Grant/funding Support: This work was performed as part of the activities of the IFCC Committee on Reference Intervals and Decision Limits, with the support of IFCC Scientific Division.
Financial Disclosures: Many of the authors served as consultants to several diagnostics manufacturers, but none of these activities represent actual conflicts of interest with respect to the current report.
Acknowledgments: We thank Paul Horn for the calculation of the nonparametric reference intervals for children.
- © 2008 The American Association for Clinical Chemistry