| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Department of Biochemical Medicine, NHS Tayside, Ninewells Hospital and Medical School, Dundee DD1 9SY, Scotland
aE-mail callum.fraser{at}tuht.scot.nhs.uk
To the Editor:
Apple et al. (1), on behalf of the Committee on Standardization of Markers of Cardiac Damage of the IFCC, recently reviewed and abstracted the scientific literature to provide recommendations pertaining to the quality specifications for B-type natriuretic peptide (BNP) and N-terminal proBNP (NT-proBNP) assays (together abbreviated here as NPs). They stated that a decision concerning what is acceptable imprecision is needed and concurred with the goals derived using the model proposed by Cotlove et al. (2) based on the concept that analytical imprecision (CVA) should not significantly affect clinical use. This model suggests that this desired negligible clinical impact can be obtained when CVA is lower than or equal to one-half the intraindividual biological variation (CVI). The mathematical explanation of this and the many other models available for the setting of quality specifications for analytical performance characteristics have been described in detail (3). The authors state that, because of high biological variation for NPs (CVI, 30%–50%), very low CVA may be unnecessary: however, for monitoring of therapy with serial BNP measurements in clinical cases, it may be desirable to minimize CVA (1).
Furthermore, Apple et al. (1) state that a desirable CVA of <15% at NP concentrations within the reference interval is recommended. If an eventual goal is to rely on monitoring of marker trends over time, then an optimal imprecision of <10% is advocated.
It is interesting that those who produce allegedly evidence-based guidelines, recommendations, and scientific statements always seem to end up with "round numbers" such as 15%, 10%, and 5% as goals for CVA (3). Using readily available data on CVA and CVI, it is easy to work out whether the recommendations are cogent, particularly the latter, because the model of Cotlove et al. (2) is actually concerned with the clinical setting of monitoring individuals.
Consider that CVI is 40%, as suggested by Apple et al. (1) and recently studied in detail (4). A significant change in serial results occurs only if the reference change value (RCV), sometimes called the critical difference, is exceeded. RCVs are easily calculated as 21/2 x Z x (CVA2 + CVI2)1/2 (3), where Z is the number of standard deviations appropriate to the probability selected. If CVA was 15%, then the RCV for P <0.05 would be 118%, and if CVA was 10%, the RCV would be 114%. Thus, the statement made by Apple et al. (1) that an optimal CVA of 10% would be advocated for monitoring individuals is not evidence based. If CVI is much larger than CVA, then it is simply not worthwhile reducing CVA to less than one-half of CVI, even in this clinical use of results.
Whether the RCV can be reduced to make NPs more useful for monitoring individuals over time is a question not addressed by Apple et al. (1) but is discussed briefly by Bruins et al. (4). In particular, do replicate assays or replicate samples, as advocated for cholesterol and high-sensitivity C-reactive protein (5), help in this regard to improve the utility of a test in monitoring individuals? This again can be calculated easily because:
 |
It is also easy to calculate the number of samples required to obtain an estimate within a certain percentage of the true individual homeostatic set point of the individual from the formula based on a simple standard error of the mean estimate (3):
 |
NPs are very like high-sensitivity C-reactive protein when the relative magnitudes of CVI and CVA are considered objectively, and it has been suggested that professional guidelines on the latter are less than objective (5). It is not difficult (6) to calculate the effects of CVA on RCV, on the results of analyzing a sample more than once and/or taking more than 1 sample, and on the number of samples needed to obtain an estimate of an individual’s homeostatic set point within a stated closeness at a predetermined probability.
The above does assume that the variation in NP concentrations can be described as random variation around a homeostatic set point, and this widely used model (3) has been implicitly and explicitly used by Apple et al. (1). However, it may be that this model is not totally appropriate for NPs. It could be argued that any change in BNP concentration in an individual should be considered as potentially and clinically relevant, even when smaller than the RCV calculated from the CVI derived using the homeostatic model, because such a change would reflect an alteration in activation of the neuroendocrine system as a result of specific pathophysiologic mechanisms. In this alternative model, any change in NP concentrations in an individual should be interpreted by taking into account clinical history and examination, including the response to specific treatments, as well as laboratory findings. Perhaps clinical criteria should be used to evaluate the pathophysiologic relevance of a "significant" variation in NP concentrations (which can be mathematically defined as (21/2 x Z x CVA) in an individual patient. If this alternative model is considered appropriate, then the recommendation that it may be desirable to minimize CVA (1) has considerable merit, but not just to CVA of 10% as recommended. The lower the CVA, the higher Z will be for any specific change, and the more significant will be the observed changes in NP. This consideration is not discussed by Apple et al. (1).
I again strongly suggest that those who produce allegedly evidence-based guidelines, recommendations, and scientific statements be urged to do all the calculations outlined here and think on their ramifications on clinical utility before disseminating their work.
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  M. R. Fokkema, Z. Herrmann, F. A.J. Muskiet, and J. Moecks Reference change values for brain natriuretic peptides revisited. Clin. Chem., August 1, 2006; 52(8): 1602 - 1603. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
