Reference intervals – ever met a normal person?

This issue, like all issues of the journal, features articles in which results of measurements in patients or patient groups are interpreted against a reference population or cohort.^1–5 Proper establishment of reference values is a tedious and often underestimated task. The outcome of laboratory measurements, and hence the reference intervals to which they have to be compared, is dependent on a plethora of factors related to both nature and nurture. Thus, reference intervals depend on factors including race, age, gender, diet, body weight, smoking habits, medication, diurnal and seasonal rhythms, reproductive status (stage of menstrual cycle, oral contraceptive use, pregnancy, menopause), body position, manner of sample collection, specimen handling and storage and the method of measurement. Also, the definition of the reference interval, i.e. either the 95% or 99% distribution of the concentration of the analyte measured in healthy persons, affects the interpretation of an individual patient’s laboratory results. An additional issue is the uncertainty of the reference cohort’s health status: definitions of ‘apparently healthy’ or ‘self declared healthy’ leave room for wider interpretation than possibly desirable. Although, in many cases, the health of the reference group is further substantiated by demonstrating that the results of other laboratory tests, e.g. those that reflect thyroid, liver or kidney disease are normal, ³ the health of a person assigned to the control group can often not be established with certainty.

Reference intervals are helpful in both diagnosis and management of patients. Successful treatment may be accompanied by aberrant laboratory results returning to within their reference interval. This is the case with electrolyte concentrations, blood cell counts and activities of liver enzymes for instance. A prerequisite here is that laboratory results in patients behave in a similar way to those in reference populations or cohorts. That this assumption is not always correct is illustrated by Corte et al. ¹ who show that long-term biological variation of troponin concentrations appears to be considerably lower in healthy volunteers than in patients with kidney failure. This finding may have implications for manufacturers of reagents as well as for interpretation of results by clinicians and establishment of guidelines for decision making.

For some categories of results, referring to a reference interval is meaningless. Whilst a decrease of preoperatively elevated tumour marker concentrations to within the reference interval may indicate a successful surgical intervention, especially if the contribution of other tissues to the blood marker concentration can be neglected, subsequent changes in marker concentration within the reference interval may be significant. A successive increase in serum prostate-specific antigen, e.g. from 0.1 just after prostatectomy via 0.5 to 1.0 and 1.8 µg/L in a 55-year-old man would indicate disease progression, although all values lie within the reference interval. It should be borne in mind too that the reference values apply to men with their prostate in situ and not to patients after prostatectomy. Similar reasoning applies to patients with ovarian and other malignancies.

Children are usually underrepresented in general populations used to establish reference intervals. Application of adult-derived reference intervals to children may not be appropriate as reported by Dodd et al., ² who found that widely used reference intervals for sodium may be too wide for children, too narrow for potassium and strongly age dependent for creatinine. Obviously, studies are recommended to define more appropriate reference intervals for children. At the other end of the age spectrum, multi-comorbidity amongst older people may influence reference intervals derived from individuals considered to belong to the unaffected or healthy group. Larsson et al. ⁴ addressed this when setting reference intervals for parathyroid hormone (PTH). Notwithstanding the extreme narrow age range in their control group, i.e. 70 years only, the findings indicate that exclusion of people with a variety of comorbidities and abnormal laboratory results has limited influence on the reference interval. It seems safe to assume that the findings are valid for people with a wider age range too. Unfortunately, the authors did not include vitamin D measurements, as it has recently been shown that vitamin D deficiency should be taken into account when establishing PTH reference values. ⁶

Introduction of new methods requires both analytical and clinical validation including a critical appraisal of reference intervals. Immunoassays by their nature appear to be especially noteworthy here. Woloszynek et al. ⁵ took this to heart and established reference intervals for anti-Mullerian hormone (AMH) following their analytical validation of a recently released AMH assay. Whitehead et al. ³ recognized that the occurrence of analytes in multiple forms may affect reference values and they illustrated this with monomeric and larger forms of prolactin. As long as assays do not properly discriminate between the various forms of the analyte prevailing in the circulation and the various forms thus contribute to the signal obtained, the results and hence the reference interval are dependent on the proportion of the various species of the analyte. It is possible that such proportions may differ between patient and reference groups, to further complicate the issue.

The establishment of proper reference intervals requires continuous attention and focus. Ideally, reference intervals should refer to the analyte and the proper patient group only and not to the method for assessment of the analyte. For most analytes, however, this is still far from reality, especially when internationally recognized commutable standards are lacking. In such cases, harmonization or the introduction of a method-related normalizing factor (e.g. as has previously been used for GFR estimating equations) may provide a temporary solution. In that sense, reference intervals, six years after the statement by Barth, ⁷ still require further clarity.