| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters |
1
Wadsworth Center for Laboratories and Research, New York State Department of Health, Albany, NY 12201
2
Department of Biomedical Sciences, School of Public Health, State University of New York, Albany, NY 12201
aAuthor for correspondence. Fax 518-474-9185; e-mail clinchem{at}wadsworth org
To the Editor:
A recent opinion article (1) in this journal presented sound metrologic principles and a plan of action to improve analytical accuracy in medical laboratories. In addition to the use of Certified Reference Materials outlined by Müller (1), use of proficiency testing and external quality-control programs can have a broader impact on interlaboratory comparability. Ideally, target values should be derived from reference methods and reference materials, but both are limited in their availability, and the mean (or other indicator of central tendency) of participant data is frequently used. This has been shown to provide a basis of comparison often comparable to reference methods (2).
The means of subsets of methods (peer-group means) are often reported as target values, a practice usually ascribed to method-dependent behavior of proficiency test specimens and so-called matrix effects (3)(4); this use of multiple "true values" in proficiency testing has been criticized (5)(6).
In the US, the CLIA’88 serves as the basis for any accredited external quality-control program. The current regulations (7) state that only when "a specific method’s results demonstrate bias that is not observed with actual patient specimens, as determined by a defensible scientific protocol, a comparative method or a method group (‘peer’ group) may be used".
Nonetheless, peer group evaluation is commonplace, perhaps reflecting potential for sanctions associated with failure in proficiency testing. Here we share two examples where this practice would have masked true analytical bias.
Case 1.
In a recent proficiency test, the results for thyroxine from a participant laboratory demonstrated a high bias relative to both the robust mean (8) of all participants and that of the instrument peer group (Fig. 1
). The degree of the high bias was less noticeable within the peer group; had the results been evaluated against the peer group, a CLIA score of 100% would have been obtained. The results were not peer graded, and three of the five results fell out of the allowable range and produced an unsatisfactory score of 40% (Fig. 1A
, specimens 1–3). In spite of the bias demonstrated by this method, other laboratories using it attained a CLIA score of 100% when evaluated against the overall range or against the peer group-specific range, had the latter been used.
|
Examination of this failure indicated that the bias was attributable to the precipitation, over time, of magnetic particles in the reagent solution onto the side of the reagent container, producing a shift in reported values. Precipitation would not have occurred had the reagent been consumed within 24 h. The failing laboratory had a low test volume and stored the unused reagent at 4 °C between analyses. The identified source of the bias was verified by reproducing the high-bias phenomenon with the stored reagent and by eliminating the high bias using the stored reagent but mixing it just before use (Fig. 1
). After this investigation, the manufacturer issued a service notice in February 2001 and initiated activities to improve the reagent quality.
Case 2.
Calcium is a non-peer-graded analyte in our proficiency test program with an allowable limit of ±0.25 mmol/L (1.0 mg/dL) from the all-method mean value (7). Methods agree well, and the failure rate in the New York State program is low [in the year 2000, 43 results of 6185 (0.7%) were unsatisfactory]. Over the course of several test events, a negative bias was consistently present for a particular instrument model with the o-cresolphthalein complexone method. The reported calcium values were either at or below the lower acceptable limit. The results within this instrument group were similar, and had peer group-specific target values been used, laboratories using this analyzer would have been awarded passing scores.
The manufacturer indicated that they accommodated the known bias by a commensurate downward adjustment of the reference interval [1.85–2.30 mmol/L (7.4–9.2 mg/dL)]. This range was lower than that typically used by clinical laboratories [2.10–2.55 mmol/L (8.4–10.2 mg/dL)] or found by reference methods (9).
Further investigation by the manufacturer led to changes in the calcium calibrator set point values. The set point adjustment (
9%) produced an increase in the recommended reference interval; proficiency test results for the analyzer now appear consistent with all-participant mean values.
These two examples demonstrate positive outcomes that may occur by avoiding peer-group evaluation in external quality assessment programs. Problems were identified that affected the quality of patient testing, resolutions were found, and comparability among methods was advanced. This supports the view of Büttner (5) that in "proficiency testing, so called ‘peer group mean values’ ... do not lead to any improvement of the trueness or therefore of the comparability".
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  W. G. Miller, G. L. Myers, and R. Rej Why commutability matters. Clin. Chem., April 1, 2006; 52(4): 553 - 554. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and peer-group mean (
) without (A) and with (B) reagent mixing. Target concentrations for specimens 1–5 were 113, 142, 221, 76, and 31 nmol/L (8.8, 11.0, 17.2, 5.9, and 2.4 µg/dL), respectively. Peer-group target concentrations were 121, 160, 263, 80, and 36 nmol/L (9.4, 12.4, 20.4, 6.2, and 2.8 µg/dL), respectively.