Recommendation for the review of biological reference intervals in medical laboratories

Introduction

Support for the interpretation of laboratory tests is a major concern for medical laboratories particularly in the interpretation of results. Properly validated reference intervals for each quantitative result as given in the report is one of the main criteria for medical decision taken using biological examination.

The concept of reference values was designed in the 1970s by a Scandinavian group, then, it was developed and completed by numerous works of national societies (French and Spanish) as well as at the international level, particularly within the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) and National Committee for Clinical Laboratory Standards (NCCLS) (now CLSI – USA) in the 1980s. Documents from public or standardization bodies have institutionalized the recommendations of the national scientific societies.

Thus, EN ISO 15189 standard [1] for Medical Laboratories and Directive 98/79/EC [2] for in vitro diagnostic devices (IVD) recommend the report of results to include, as applicable, reference interval limits provided by Medical Laboratories and the package inserts of IVD as well.

Numerous articles published in recent decades are partly outdated and does not always correspond to the needs expressed by professionals [3], [4], [5], [6], [7], [8], [9]. A joint working group from the CLSI and the IFCC has revised documents published previously. A joint paper was published under the title “Defining, Establishing and Verifying Reference Intervals in the Clinical laboratory: Approved Guideline-Third Edition – EP28-A3c” [10]. The French Society (Société Française de Biologie Clinique – SFBC) provided recommendations for determination and review of reference intervals [11].

Terminology

The following definitions have been approved by the IFCC, the International Council for Standardization in Hematology (ICSH) and the World Health Organization (WHO), then by the CLSI. They have been included in full in the latest document published jointly by the IFCC and CLSI [10].

• Observed value: value of an analyte obtained by observation or measurement of a test subject, which should be compared with reference values, a reference distribution, reference limit or reference interval

• Reference distribution: the distribution of reference values

• Reference individual: a person selected on the basis of well-defined criteria

• Reference population: a group consisting of all reference individuals

• Reference interval: the interval between two reference limits (these included) e.g.: 95% of apparently healthy men from 18 to 65 years

• Reference limits: a value derived from the reference distribution and used for descriptive purposes

• Reference values: the value obtained by observation or measurement of a defined quantity on a reference individual

Figure 1 shows the relationship between the various terms defined above.

Figure 1:

Relationship between the terms used for the “Reference Values Concept” [4], [10].

The reference interval is the interval specified in the distribution of values obtained from populations of healthy subjects. This is generally defined as an interval corresponding to 95% of the population, centered on the median. It can vary depending on the type of primary population sample and the analytical method. In some cases, only one reference limit may be used, usually an upper limit. The determination of the reference interval is based on statistical calculations. It is purely descriptive of a given population.

The reference limits [12] (defining a reference range) are associated with a well-defined reference population, generally consisting of healthy individuals. They are used to compare an observed value (a result from the patient) to reference data obtained from this group of well-defined subjects. They are one of the keys for medical decision making which should take into account the specificities of each patient. Reference limits are descriptive of a given health state and can sometimes be used, in well-defined cases, as decision limits.

The medical decision limits [12] are used by the clinician as a threshold below or above which a medical action is recommended. While reference limits are generally two (upper and lower limits), the number of decision limits is variable according to concerned laboratory test and clinical setting. They are based on a clinical assessment and are set either by statistical methods (e.g. Bayesian approach) or from epidemiological studies.

In some cases, for some analytes, reference ranges are replaced by decision limits set by national or international consensus (e.g. total cholesterol, glycated hemoglobin...). For these analytes it is unnecessary to determine reference limits or to validate data from the literature.

Finally, it should be borne in mind that the comparison of a laboratory result to a reference or decision limits is not the only way of interpretation of laboratory tests: as recalled again Dalton about the measurement of serum creatinine? This also applies to many other analytes: low variation of an observed value (even within the reference range) may have a pathophysiological significance (for the diagnosis, monitoring of treatment or prevention) [13].

Protocol for determining reference limits

For a new examination or a new method, if there is no reliable data in the literature, the laboratory will use the basic protocol to determine the reference interval limits. When information is available, it may be preferable to validate published reference intervals.

The basic protocol includes a series of successive steps fully described in documents formerly published by the IFCC [3], [4], [5], [6], [7], [8], [9] and recent recommendations of the IFCC and CLSI [12]. Here is a simplified summary:

1. List the factors of biological and analytical variations (from literature data)

2. Determine the exclusion and partition criteria on the basis of adapted questionnaire

3. Make a written consent form and have it signed by the selected individuals

4. Categorize potential reference individuals based on questionnaire data and other assessment of health status modes

5. Exclude individuals from the reference sample based on predetermined criteria

6. Set the appropriate number of individuals reference

7. Prepare selected individuals for sample collection according to the procedures normally used for patients in the laboratory

8. Collect and process samples

9. Collect reference values: analyze specimens following well-defined methods

10. Check the reference values. Establish a histogram to evaluate the distribution of data

11. Identify possible errors and/or outliers

12. Analyze the reference values: select a statistical method and calculate the limits of reference and the reference interval

13. Document all steps and procedures followed.

The following describes some points.

Transferability

The IFCC protocol is considered the “gold standard”, but it is unsuitable for routine practice of clinical laboratories because it is too heavy and too complex to implement. Also it seems unrealistic that each laboratory determines its own reference ranges for each new test method or analytical system introduced in the laboratory. There is no simple, universal method. The main difficulty is the selection of the population and the definition of a “healthy” individual. If recruitment is not properly done, the calculated reference limits may be skewed regardless of the statistical method used. In practice, the most common factors of exclusion are: the presence of an acute or chronic illness, long-term medication, overweight and consumption of tobacco and/or alcohol. Inclusion and exclusion criteria are to be documented according to the test and disease associated.

Review of reference intervals

In an attempt to overcome this difficulty it is proposed that only a “review or verification” of the published reference limits shall be made. The transfer of data produced by other laboratories or IVD manufacturers, combined with a simple validation process, could be a great help.

However, certain conditions must be fulfilled so that the transfer process is acceptable, including the selection process of the population and that measurement methodologies (preanalytical and analytical) are similar. For this purpose, the revised recommendations IFCC/CLSI [10] propose several solutions based on different scenarios:

Case 2: Comparison of populations

When a laboratory wishes to transfer the reference intervals established by another laboratory (article, laboratory with the same analytical system, users group...) or by a manufacturer, the transfer of the reference interval exceeds the strict comparison of analytical systems framework, it becomes a matter of comparing reference populations.

Several approaches are proposed in the latest recommendations of the IFCC and CLSI [10].

1. Subjective method

   To verify that the essential elements of the original study are consistent with the working conditions and the population of the laboratory.

   Key elements to consider are:

   • the geographic and demographic criteria

   • the preanalytical procedures

   • the analytical performance

   • the description of the reference population and the protocol used

   • the statistical method for determining the reference range

   If this is the case, the reference range of the original study can be transferred without verification. Providing all the necessary information is still the main limitation.

2. Verification of the reference interval from a sample of apparently healthy subjects (Figure 2)

   If the subjective method is not applicable, the laboratory verifies the reference interval published (IVD manufacturer’s application sheets, another laboratory data, scientific articles, etc. …).

   Protocol:

   • Selection of 20 apparently healthy individuals, taking into accounts the required exclusion and partition criteria (gender, age, absence of disease, medication...).

   • Determination of reference intervals with the method to be tested. The homogeneity of the set of data will be checked to ensure that the whole range is covered by data. If some data are discrepant as compared to the whole, they could be eliminated using a so-called method for rejection of outliers. Preanalytical and analytical conditions of the tested method and the original one will be consistent.

   Interpretation:

   • reference limits to be checked are accepted if the number of results outside the limits is ≤2.

   • New selection of 20 biological samples is analyzed if the number of results outside of the proposed limits is equal to 3: the same procedure as above is applied. Under these conditions, the reference limits to be checked are accepted if the number of results of the new selection out of range is ≤2.

   • If four or more results are outside of the limits, it is advisable to review the analytical procedure, to consider the possible presence of biological and/or demographic differences and to determine the reference limits of the new method following the original protocol. It would ne helpful to examine possibility if given reference interval data could be matched with some other refrenec interval source prior to start detailed RI examination with a large number of samples.

Figure 2:

Validation of reference interval [10].

Multicenter studies to determine reference intervals

Given the difficulty in selecting reference individuals, pooling of data produced by different laboratories can simplify some tasks to achieve the desired goal [10]. Indeed, among the many factors of variation, two are particularly important:

• The reference population: the documentation of ethnic and racial differences is relatively scarce, but it cannot be overlooked.

• The influence of the analytical method is true. However, standardization efforts undertaken in recent years allow reducing the effects from laboratory to laboratory. However, the inter methods differences are often very important, especially in immunochemistry.

  It is therefore possible to produce common reference intervals through multicentre studies, subject to comparable analytical systems.

To do this the following prerequisites must be met:

• The selection of reference individuals will be consistent with the basic protocol

• The implementation of the pre analytical phase will be the same on each site and well described

• Traceability of results and the inter-laboratory standardization will be effective

• A common program of quality control will be implemented

• Finally it is recommended that each laboratory validates established reference ranges in its own environment.

Traceability

Traceability of all actions will be provided. All steps for determining Reference Intervals will be documented:

• Preanalytical conditions

• Analytical method

• Selection of the reference samples

• Statistical method

Other procedures for transferring and validating reference intervals

Tate et al. [20] published recently a critical review. Some examples are given, including the Canadian study from Adeli et al. and those from Koerbin et al. in Australia. Adeli et al. [21], [22], [23] conducted the study of the Canadian Laboratory Initiative on Pediatric Reference Intervals (CALIPER) and the Canadian Health Measures Survey (CHMS) which was carried on well-defined populations according to a pre-analytical and analytical protocol tightly controlled. This multicenter study compared the references intervals between the key analytical systems. Data for children, adults and elderly Canadians are presented.

Koerbin et al. [24] determined the bias between eight major analytical platforms using a similar method of comparison.

Tate [20] also offers as an alternative approach to mining data from its own laboratory. This approach is acceptable only if the laboratorian is able to identify healthy individuals not affected by a disease. This method is not recommended by the CLSI-IFCC.

Ozarda et al. [25] aims to derive reliable country specific Reference intervals through multicenter studies at a global level and a protocol was developed. It could be applied to markers traceable to a reference system and non-standardized biochemical markers if there are no regional and/or ethnic groups’ differences.