Abstract
Barth, 1 in an Editorial concerning an excellent comprehensive review on reference intervals, 2 reminded us that the results of laboratory tests are used in many different clinical settings including screening, diagnosis and monitoring. Moreover, the concept of using population-based 0.99 interfractile intervals in screening and diagnosis and 0.95 interfractile intervals for disease monitoring 3 was said to have merit.
However, as briefly pointed out by Ceriotti et al., 2 the within-subject biological variation (CVI) is much smaller than the between-subject biological variation (CVG) for nearly all quantities measured in laboratory medicine. 4 The major consequence of this marked individuality is that conventional population-based reference intervals, irrespective of whether generated by an individual laboratory or common or harmonized, are of very limited utility in evaluating the results of an individual in screening or diagnosis, since many individuals will have values which are highly unusual for them but still lie within reference intervals.
However, most laboratory tests are used in monitoring, either in acute settings in hospitals or over longer terms in chronic disease. More importantly, therefore, conventional reference intervals are of extremely limited value in this clinical setting. Individuals can have significant changes in results when all lie within the reference interval: such changes will usually be ignored by both laboratories and clinicians. In addition, results can change from inside the interval to outside (and vice versa) without significance: laboratories would conventionally ‘flag’ the results outside the reference limits, probably stimulating some unnecessary clinical activity, if only repetition of the test.
A much better way of monitoring individuals, rather simpler than the methods briefly mentioned by Ceriotti et al., 2 is to use reference change values (RCV), the generation and application of which have been documented in detail. 5
Changes in serial results from an individual may be due to the individual improving or deteriorating, but are also due to three inherent sources of variation, namely preanalytical variation (CVP), analytical imprecision (CVA) and CVI. For a change to be significant, the difference in results must be greater than this inherent variation, termed the RCV, and calculated as: RCV = 21/2 Z (CVP 2 + CVA 2 + CVI 2)1/2, where Z is the number of standard deviations appropriate to the desired probability (for example, 1.96 for P < 0.05 and 2.58 for P < 0.01). When preparation of the individual for sample collection, and sample collection, handling and storage prior to analysis, are both optimized, as they should be by good training of staff and adherence to standard operating procedures, CVP becomes minimal and the formula reduces to: RCV = 21/2 Z (CVA 2 + CVI 2)1/2. Thus, RCV are very simple to calculate since all laboratories know the analytical imprecision of each of their methods in detail from internal quality control techniques and within-subject biological variation data are available for many quantities. 4 Estimates of within-subject biological variation are constant over time, geography and methodology, and in health and chronic stable disease: 6 they are therefore ubiquitously applicable, so that, unlike population-based reference intervals, laboratories do not have to generate their own data. A further approach, as yet seemingly unapplied in everyday practice, is that the RCV formula can be rearranged to make Z the unknown: Z = change/[21/2 (CVA 2 + CVI 2)1/2]: thus, using the change in serial results found in an individual and the known analytical imprecision and within-subject biological variation, the probability that any change was significant could easily be determined.
RCV can be used to highlight significant changes on individual laboratory reports as has been done for many years in NHS Tayside, Scotland. 5 Moreover, RCV can be used to set objective criteria for use in delta-checking quality control techniques. They can also be used in auto-verification and auto-validation strategies. 7 Rather than spending considerable resources in defining reference intervals, laboratories are urged to apply well-established methodology to calculate RCV and to use these in everyday practice, providing considerable advantage in the monitoring of changes in serial results from individuals.
DECLARATIONS