Volume 42, Issue 7

This manuscript reviews the introduction of the concept of reference values, the corresponding philosophy, and subsequent recommendations reflecting the different subtopics of the field. The generally unrecognised phenomenon that laboratory results of the population tended to be lognormal instead of Gaussian attracted the attention of the author, who became sceptical of the concept of normal values because of its ambiguity. Together with N.-E. Saris a new concept of reference values was launched at a congress in 1969. Briefly, clinical measurements should be interpreted against values from proper control subjects. Subsequently international, regional and national societies established expert panels which produced recommendations covering the general principles and terminology of reference values, the concept of health, standardised specimen collection and preanalytical factors, statistical treatment of collected values, stratification of data, relating observed (patient) values to reference values, etc. The importance of using correct terminology, taking into consideration the aim of ordering the laboratory test and some neglected procedures (e.g., survival values) are emphasised. New developments such as reference changes are mentioned. The field has evolved largely as the result of constructive international teamwork.

The concept of reference values is widely accepted, but their application has been quite lax over the years. This is due primarily to the difficulty of properly selecting and documenting samples from a reference population. In the absence of a clear description of reference individuals, reference values lose their meaning, are ambiguous at best, and are often confused with decision limits. The clinical medicine perspective of reference values is to rule out diseases and to define health, while that of preventive medicine is to appreciate the state of health. Defining reference limits and normality in this context is a great challenge. Advances in the fields of genomics and proteomics and the rapid pace of technological advances help highlight the biological diversity among individuals. However, there is a great need for reference values that are representative of healthy humans and presented in a manner that they can be utilized by all laboratories. In addition, as secure information technology becomes available, the goal of using an individual as their own reference during a lifetime is now possible, provided that consistency of databases is ensured.

The selection of individuals to represent a group for any purpose can commit the effort to success or failure from the very outset. The description of this group must reflect what is, in reality, the purpose for which the group is being studied. Considering that reference intervals are designed to assist clinicians during diagnostic triage, criteria must reflect the nature of the subjects entering the system for care. The majority of people seeking care have non-life-threatening, yet troublesome complaints. In other words, persons who are in a reasonable state of health and who are not suffering from significant illness. Therefore the relevant cohort should fit this description, not individuals who lack any clinical or laboratory issues. Our focus is to employ reference intervals to exclude disease and not to define normality. Normality in the eyes of statisticians is quite different from the word as viewed by clinical staff, yet the term is used interchangeably without concern for its ultimate use. As a result medicine today stumbles over this issue at every turn. Scientists and artists do not speak the same language. Compounding this awkward situation has been the conviction that by increasing laboratory precision and accuracy, the utility of laboratory data, reference intervals in particular, will improve clinical understanding. This has, after many years, not proven to be the case. What has become very evident, however, is that unsustainably large sums of money are being devoted to improving laboratory performance. To what end, when we are unable to grasp the significance of what is available now including published reference intervals for common test values? A realistic perspective on the value of prevailing laboratory performance, as illustrated by reference intervals, must be adopted. The obligation of laboratories, its regulators and professional groups is to create a sustainable mechanism to provide clinically usable information, in a timely and affordable manner, to our population and to those legions of less fortunate humans.

The International Federation of Clinical Chemistry (IFCC) published between 1987–91 a series of six recommendations on reference values in laboratory medicine. This paper reviews the history and scope of the fifth part of these recommendations. This fifth recommendation deals with statistical methods used for analysis of reference values and estimation of reference intervals. The RefVal program, which implements the recommended method, is also described.

A well-known transformation from the bell-shaped Gaussian (normal) curve to a straight line in the rankit plot is investigated, and a tool for evaluation of the distribution of reference groups is presented. It is based on the confidence intervals for percentiles of the calculated Gaussian distribution and the percentage of cumulative points exceeding these limits. The process is to rank the reference values and plot the cumulative frequency points in a rankit plot with a logarithmic (ln=log e ) transformed abscissa. If the distribution is close to ln-Gaussian the cumulative frequency points will fit to the straight line describing the calculated ln-Gaussian distribution. The quality of the fit is evaluated by adding confidence intervals (CI) to each point on the line and calculating the percentage of points outside the hyperbola-like CI-curves. The assumption was that the 95% confidence curves for percentiles would show 5% of points outside these limits. However, computer simulations disclosed that approximate 10% of the series would have 5% or more points outside the limits. This is a conservative validation, which is more demanding than the Kolmogorov-Smirnov test. The graphical presentation, however, makes it easy to disclose deviations from ln-Gaussianity, and to make other interpretations of the distributions, e.g., comparison to non-Gaussian distributions in the same plot, where the cumulative frequency percentage can be read from the ordinate. A long list of examples of ln-Gaussian distributions of subgroups of reference values from healthy individuals is presented. In addition, distributions of values from well-defined diseased individuals may showup as ln-Gaussian. It is evident from the examples that the rankit transformation and simple graphical evaluation for non-Gaussianity is a useful tool for the description of sub-groups.

Four existing methods for partitioning biochemical reference data into subgroups are compared. Two of these, the method of Sinton et al. and that of Ichihara and Kawai, are based on a quotient of a difference between the subgroups and the reference interval for the combined distribution. The criterion of Sinton et al. appears rather stringent and could lead to recommendations to apply a common reference interval in many cases where establishment of group-specific reference intervals would be more useful. The method of Ichihara and Kawai is similar to that of Sinton et al., but their criterion, based on a quantity derived from between-group and within-group variances, seems to lead to inconsistent results when applied to some model cases. These two methods have the common weakness of using gross differences between subgroup distributions as an indicator of differences between their reference limits, while distributions with different means can actually have equal reference limits and those with equal means can have different reference limits. The idea of Harris and Boyd to require that the proportions of the subgroup distributions outside the common reference limits be kept reasonably close to the ideal value of 2.5% as a prerequisite for using common reference limits seems to have been a major improvement. The other two methods considered, that of Harris and Boyd and the “new method” follow this idea. The partitioning criteria of Harris and Boyd have previously been shown to provide a poor correlation to those proportions, however, and the weaknesses of their method are summarized in a list of five drawbacks. Different versions of the new method offer improvements to these drawbacks.

Age-specific reference limits are required for many clinical laboratory measurements. Statistical assessment of calculated intervals must be performed to obtain reliable reference limits. When parametric, covariate-dependent limits are derived, normal distribution theory usually is applied due to its mathematical simplicity and relative ease of fitting. However, it is not always possible to transform data and achieve a normal distribution. Therefore, models other than those based on normal distribution theory are needed. Generalized linear model theory offers one such alternative. Regardless of the statistical model used, the assumptions behind the model should always be examined.

Despite being proposed over 30 years ago, the multivariate reference region (MRR) remains an interpretive tool with uncertain value in clinical practice. This paper reviews the mathematical theory of the MRR, presents topics in the clinical application of MRRs, and examines the advantages and disadvantages of MRRs. Finally, some thoughts are offered regarding future directions for work in this field.

The reference interval is probably the most widely used decision-making tool in clinical practice, with a modern use aiming at identifying wellness during health check and screening. Its use as a diagnostic tool is much less recognised and may be obsolete. The present study investigates the consequences of the newpractice for the interpretation of prospective value, negative vs. positive, the probability of confirming wellness, and number of false results based on selected strategy for reference interval establishment. Calculations assumed normalised Gaussian-distributed reference intervals with analytical variation set to zero and absolute accuracy. Also assumed is the independency of tests. Probability for no values outside reference intervals in healthy subjects was calculated from the formula p no outside =(1−p single ) and according to the formula for repeated testing: p one outside =n×p single (1−p single ) n−1 etc. Here n is the number of tests performed and p single is the probability of one result outside reference limits with the general formula p i outside n−i =k×p single i (1−p single ) n−i , with k being the binominal coefficient and i the number outside the reference intervals. Use of the 99.9 centile for health checks will increase the probability for no false from 60% to 99% for 10 tests, and from 46% to 98% for 15 tests. The probability for one false-positive result in 10 tests in a panel can be reduced from 32% to 1% if the 99.9% centile is substituted for the 95% centile. For two in 10 tests, the probability can be reduced from 8% to below 0.1%. In both cases, selection of the 99.9% centile improves the diagnostic accuracy. Reference intervals are needed as a “true” negative reference for absence of disease, and should cover the 99.9% centile of the reference distribution of an analyte to avoid false positives. For this new use, it is critical that reference persons are absolutely normal without clinical, genetic and biochemical signs of the condition being investigated. However, reference intervals cannot substitute clinical decision limits for diagnosis and medical intervention.

Reference intervals for healthy subjects and diseased populations are important benchmarks for the clinical interpretation of laboratory test values. It is important that the testing conditions used to collect the reference data be closely matched with the testing conditions used for patient data. Interlaboratory differences and intralaboratory changes, especially changes in analytic set-points, can markedly affect the clinical interpretation of tests. If laboratory testing methods could be harmonized, laboratories could potentially share reference data to make these data more reliable. This chapter illustrates the use of reference intervals for healthy subjects and reference data from diseased populations for medical decisions. One example illustrates the effects of age differences on test result interpretation. Other examples illustrate use of thyroid-stimulating hormone (TSH) and combinations of TSH with free thyroxine measurements for diagnosing thyroid diseases. Both univariate and multivariate reference intervals are discussed. Model systems for using disease prevalence and cost utility functions are provided to illustrate optimization of medical decisions.

There is a need for revisiting theoretical concepts and practical applications of conventional population-based reference values to make for better clinical use of laboratory data. Knowledge of the underlying biological variation of quantities examined in medical laboratories is vital to understanding the proper generation and application of traditional population-based reference values. Appreciation of the biological changes that occur over the span of life is a necessary prerequisite to deciding whether stratification of reference values according to age is likely to be necessary. Knowledge of the detail of predictable biological cyclical rhythms is required for correct clinical interpretation of laboratory data and appropriate collection of specimens at times relevant to the clinical purpose. Quantitative data on inherent within- and between-subject biological components of variation have shown the marked individuality of most quantities of interest in laboratory medicine. This individuality casts light on why examinations are not generally very successfully applied in population screening or case-finding. Consideration of individuality demonstrates why stratification of reference values is often very advantageous. Individuality provides an indisputable argument for better use of individual specific reference values.

When a biological quantity examination exhibits a high degree of individuality, developing a strategy for interpreting these values in an individual context can be a useful alternative. Time-series analysis is the appropriate statistical framework to build a model for explanation of the behaviour of laboratory information and to forecast future values. The key concepts in this approach are autocorrelation and withinperson variance. Unfortunately, the powerful tools provided by time-series analysis require many observations, a requisite difficult to meet in every day practice. However, introducing some restrictions in the autocorrelation parameter of the most reliable model, the first order autocorrelation model, and using the average within-person variance from a selected population, it is possible to build predictive reference intervals for an individual, based on only few observations. The most common case is the minimum time series: when there are just two observations. The statistical significance of the change from a previous observation is a problem that arises from both quality control (delta checks) and the interpretative diagnostic fields (reference change limit). Applying the same restrictive criteria, it is possible to develop specific limits for a difference between consecutive observations based on a within-person variance selected from the distribution of variances found in a sample of similar individuals.

This article is intended as a guide for the production of biological reference values of healthy people (physiological reference values) by several clinical laboratories using the same measurement system. This guide is a proposal from the Catalan Association of Clinical Laboratory Sciences to be applied worldwide at a regional level. This guide makes it possible for all clinical laboratories in a region using the same measurement system to adopt the same physiological reference limits. The model presented here is based on the assumption that the production of physiological reference values is a professional task that should be shared by both clinical laboratories and the in vitro diagnostics industry.

A suggestion for a standard procedure to establish biological reference intervals for biochemical quantities by a multicenter approach is presented. This procedure was developed for and used in the Nordic Reference Interval Project 2000 (NORIP). This project established biological reference intervals for 25 frequently requested biochemical quantities through cooperation of 102 Nordic laboratories. Each laboratory performed collection of reference samples and measurement using their routine methods. The bias of each routine method was eliminated by use of common reference materials measured in each of the participating laboratories.

The aim of this study was to investigate similarities and differences in the distribution of serum concentrations of nine proteins in two racial groups (Caucasian and Asian Indian) of adult males living in the same geographical area (Leeds, Bradford, UK) for at least two generations. This is part of a larger study to determine the need for separating reference intervals for racial and ethnic groups worldwide. The distributions of concentrations for all proteins evaluated in the Indians fit ln-Gaussian distributions, indicating probable homogeneity. However, for the Caucasians, the distributions for α 1 -antitrypsin and possibly haptoglobin were not ln-Gaussian. In the former case, this is undoubtedly due to the number of Caucasians with lower-concentration phenotypes (Pi MS and MZ). Although haptoglobin differences may be due to genetic variants as well, this is not a complete explanation. In addition, the Indians have lower serum concentrations of orosomucoid (α 1 -acid glycoprotein), as has been reported by others. It is apparent that for some proteins, including α 1 -antitrypsin, orosomucoid, and possibly haptoglobin, the populations show differences that require the use of separate reference intervals. In addition to genetic influences, environmental differences cannot be ruled out as partial causes for some of the differences noted.

The IFCC Committee on Plasma Proteins has been investigating regional differences for commonly assayed plasma proteins to determine whether universal reference intervals can be applied. As a part of this study, we launched an Asian project analyzing the concentrations of 13 serum proteins whose values are standardized to CRM470, and five newer analytes: retinol-binding protein (RBP), cystatin C (CysC), lightchain-κ (L-κ), and light-chain-λ (L-λ). In Tokyo, Seoul, Kuala Lumpur, Hong Kong, Taipei and Shanghai, serum samples were collected from 146 to 415 apparently healthy individuals with nearly equal gender ratios. All assays were performed in Tokyo on a Behring Nephelometer II (BN II). Seven chemical analytes (aspartate aminotransferase (AST), alanine aminotransferase (ALT), γ-glutamyltransferase (γGT), creatinine, total cholesterol (TC), triglycerides (TG) and high-density lipoprotein cholesterol (HDL-C)) were also measured. These results were used for excluding individuals with possible latent clinical disorders. Positive acute phase reactants were consistently lower, and negative ones were higher, in Tokyo than those in other cities. The most conspicuous difference was observed in C-reactive protein (CRP). There were no regional differences in transferrin, albumin, or CysC. Creatinine was much lower in Tokyo despite comparable CysC levels. ALT and γGT were higher in Shanghai, Taipei and Seoul; γGT and TG were higher in Shanghai; and HDL-C was higher in Tokyo. Gender-related differences in reference intervals were observed for immunoglobulin (Ig)M, haptoglobin, RBP, transferrin, α 2 -macroglobulin (A2M), transthyretin, α 1 -acid glycoprotein, CysC, and C4 in all cities. Slight age-related differences were observed, irrespective ofthe region, in IgA and ceruloplasmin (increase) and A2M (decrease). Environmental factors and lifestyle seem to have a great influence on many commonly measured analytes.

Laboratory tests were at the origin of the modern approach to the evaluation of the health status in many Preventive Medicine Centers. Guided by the theory of reference values and its applications, years of efforts have been devoted to defining the health status of people, and understanding the biological variation of the population. At the Nancy Center for Preventive Medicine, 40,000 people are invited every year for health examination; many coming with their families. French legislation authorized every citizen to have full health examination every five years. Annually, 600,000 people are seen at 70 centers. The visit includes answering questionnaires, physiological and biological testing modified for age, risks, and social criteria. The process concludes with consultation with a general practitioner. Data accumulated over 30 years have led to the identification of primary factors influencing biological variation of common laboratory tests. For example, body mass index (BMI) has to be considered, in addition to age and gender, for measurement of γ-glutamyltransferase, alanine aminotransferase, and other analytes. More recently, the familial Stanislas cohort has shown that apolipoprotein E (apoE) genetic polymorphism has to be considered as a factor for variation in some individuals. For example, there was greater than 100% difference in the mean concentration of apoE between ε 2/ ε 2 and ε 4/ ε 4 individuals. This example nicely illustrates the relationship between genotype and phenotype. Demonstrating the effects of multiple gene polymorphisms on phenotypes will be a valid approach to predictive laboratory medicine.

Reference intervals are recommended for naturally occurring quantities and required in the evaluation of new components in order to provide clinically useful information. The aim of the present study is to present a method for selecting reference individuals for the determination of fasting venous plasma glucose (f-vPG) reference intervals and ways to determine if disease groups can share reference intervals with an ideal reference population. Reference subjects were randomly selected, eligibility was judged according to predetermined inclusion and exclusion criteria. Using the literature we selected risk indicators for diabetes mellitus (DM) and used these indicators to rule out high-risk individuals in order to obtain a reference distribution of f-vPG determined using individuals with low risk of DM. The distributions of f-vPG in the high-risk individuals was compared with that determined for the low-risk group. We then estimated the ability of the high-risk individuals to share the reference interval of the low risk individuals, and calculated the fraction that was outside this interval. Distributions were also investigated for linearity in the cumulated frequency rankit distribution of ln-values. The allowable difference between two reference limits could not exceed 0.375 times the population biological variation. Most risk indicators were powerful predictors of high f-vPG values. Subgroups with these risk indicators should not be included in the homogeneous ln-normally distributed reference distribution. Distributions of f-vPG concentrations in individuals with risk factors were not homogeneous and varying percentages of individuals were outside the reference distribution, having f-vPG greater than 7.0 mmol/l. We conclude that randomisation is only useful to recruit candidate reference subjects. To rule out subjects according to clinical risk factors for diabetes, it is necessary to identify a reference population with low risk of exhibiting increased f-vPG concentrations. This method may be used to validate a reference interval for a particular analyte with respect to an investigated disease, and to stratify risk factors of importance.

It has previously been shown that thyroid antibodies affect thyroid stimulating hormone (TSH) concentrations in men and women and that TSH levels are predictive of future thyroid disease. We investigated the validity of the National Academy of Clinical Biochemistry (NACB) guidelines regarding the TSH reference interval by studying 1512 individuals. Two hundred and fifty had at least one thyroid antibody, 121 were taking medications other than estrogens and occasional analgesics, and 105 reported a family history of thyroid disease. Serum TSH, thyroid peroxidase antibodies (TPOab) and thyroglobulin antibodies (Tgab) were determined on AutoDELFIA and TSHRab by a radioreceptor assay (RRA) from Brahms Diagnostica. For individuals without thyroid antibodies and other risk factors, no effect of age and gender was seen for serum TSH. Neither medication nor the presence of Tgab alone had any influence on serum TSH. TPOab alone or in combination with Tgab were associated with an increased serum TSH level. The ‘cumulative percentage distributions’ of subgroups, as well as the combined population, was ln-Gaussian distributed. The central 95% of the population was within the 95% CI in rankit-plots. Consequently, a common reference interval for serum TSH of 0.58–4.07 mIU/l for all adults between 17 and 66 years of age was established. This reference interval is much higher than expected from the NACB-guidelines.

Biochemical measurements in “special fluids” are complicated with the problem of reference intervals. Reference intervals are difficult to establish for these types of samples since they are usually only collected in patients with clinical suspicion of disease. Determination of neurotransmitter metabolites in cerebrospinal fluid illustrates this difficulty. This paper will review the factors and circumstances that have been identified or are suspected to modifythe concentration of 5-hydroxyindoleacetic acid (5-HIAA) and homovanillic acid (HVA) in cerebrospinal fluid. In addition to obvious parameters such as age-related variation that can affect the concentration of 5-HIAA and HVA in cerebrospinal fluid, a varietyof other factors can explain the wide range of “control” group sizes reported in the literature. Reference intervals must take into account the purpose of cerebrospinal fluid examinations, whether they be prospective studies to explore physio-pathologic relationships or for diagnostic purposes. In the latter case, certain neurological disorders cannot be excluded if a single measured value is within the reference interval.

This manuscript explains the establishment andvalidation of metrological traceability of calibration for routine measurement procedures using common medical decision-making criteria. Metrological traceability is considered the basis for achieving comparability of measurement results in laboratory medicine. This concept is supported by European legislation, which demands that manufacturers provide assurance andd emonstrate metrological traceability of in vitro Diagnostic Medical Devices. The guidance to comply with these legislative requirements is available in different CEN/ISO standards and is used as a basis of this manuscript. The goals andaccomplishments in metrological traceability of SI- and non-SI analytes is considered. Specific problems, such as non-availability of primary reference materials and measurement procedures, lack of official endorsement, andnon-commutability of certain reference materials are discussed. With respect to non-commutability, the use of split-sample measurements is advocated. Also, the expression of measurement uncertainty associatedwith the application of the metrological traceability chain is discussed. In addition, the needfor post-market vigilance assessment of traceable performance is considered. Finally, laboratory medicine scientific and professional societies, diagnostics manufacturers, and clinicians are urged to share responsibilities for understanding the implications of metrological traceability of routine measurements.

This is a brief compilation of criteria on analytical quality that are prerequisites for the establishment and use of reference intervals in laboratory medicine. In this context, analytical quality is characterized by the precision, accuracy and long-term stability of laboratory methods performed with diagnostic kits, including their application on various instrument platforms. Systematically, analytical quality can be differentiated into quality of the method and its performance, and quality of long-term use. Factors influencing these types of quality in the development of diagnostic kits are discussed. The responsibility for maintaining analytical quality is shared by the two parties involved in the provision and use of reference intervals: The manufacturers of diagnostic kits are committed to providing complete description and traceability of their methods, whereas users must consider the effects of physiological, biometrical, preanalytical and analytical influences on the reference interval.

Interpretation oflaboratory test results requires comparison to some type of reference value or reference interval. These comparisons can be cross-sectional (population-based reference interval and cut-off values) or longitudinal (reference change value). Quality specifications for cross-sectional comparison have been established by determining the influence of analytical bias and imprecision on the percentage ofthe healthy population falling outside the reference limits, when sharing population-based reference intervals in a Gaussian distribution ofresults. Quality specifications for longitudinal comparisons are equally important and are often overlooked, since less work has been done in this area. Some criteria suggest that a difference between consecutive results designates a true change in a patient health status when the difference is higher than the within-subject biological variation plus the within-laboratory analytical variation. In this chapter we discuss the clinical considerations and laboratory-related factors that must be considered when quality specifications are applied to sharing reference comparisons. Real life experience shows that different analytical methods can produce comparable results when common quality goals are established, and quality can be achieved through a willingness to work together. Within the existing organization, the current specifications for analytical quality and a dedication to quality health care makes it possible to achieve transferability between laboratories within a geographic area.

Internal quality control should assure that the desired quality goals are achieved during reference value studies. Quality goals are often stated in the form of allowable limits of error, such as an allowable total error or an allowable bias. For reference value studies, it may be more appropriate to utilize a goal for allowable bias. In either case, it is possible to calculate a metric in the form of the critical systematic error that can be used to guide selection or design of the internal quality control procedure. A graphical tool, called the critical-error graph, facilitates the selection by superimposing the calculated critical systematic error on the power curves of different control rules and numbers of control measurements. Examples are provided to illustrate the calculation of the critical systematic error from both an allowable total error goal and an allowable bias goal, using figures from an extensive tabulation of available total error and bias goals.

In Lundberg's brain-to-brain turnaround time loop, the postanalytical phase is undoubtedly one of the most difficult phases. Of course, error prolongation can cause serious problems when quality is not sufficiently assessed and assured in the preanalytical and analytical phases. However, in the postanalytical phase following reporting, interpretation and action are at stake; and with this the wellbeing of the patient. Here exists a grey area of responsibility between the laboratory and physician. The laboratory should correctly report the measurement as data. But to place this in the correct context requires additional information on (specific) reference values, specificity, sensitivity and possible interpretation of the data. Here is where the ultimate interaction between the laboratory and the physician (representing the patient) occurs. What happens when errors occur in this phase? Do errors occurring in the previous phases effect this phase? What types of errors occur in the postanalytical phase? Postanalytical quality is the ultimate check on the coherence of the preanalytical, analytical and postanalytical quality; it is a check on the overall quality. It ties together the quality of the question to be answered, the analytical quality that is achieved and the usefulness of the answer obtained. In addition, it also checks quality in the context of the patient and physician interaction. Can errors be avoided and can a predictable quality level be assessed and be assured? A model, called NEXUS vision is described.