Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Representatives of Bayer Diagnostics respond

Representatives of Bayer Diagnostics respond To the Editor The data provided by Carmel et al. (their Table 1 )suggest that they might be missing cobalamin-deficient patients because they failed to choose an appropriate medical decision point for their patient population with the Bayer ADVIA Centaur VB12 assay. Although the data suggest a “systematic” bias between the RIDA and the ADVIA Centaur, the dispersion (the standard deviation of the difference between ADVIA Centaur and RIDA results) is equivalent to that for Quantaphase vs RIDA. The ADVIA Centaur-vs-RIDA dispersion is 77 pmol/L, and the Quantaphase-vs-RIDA dispersion is 75 pmol/L; these are essentially the same (P = 0.926 by the F-test). In fact, after correcting for systematic bias, ADVIA Centaur and Quantaphase agree better with each other (dispersion of 65 pmol/L) than does either with RIDA. (We assume that Carmel et al. believe that the Quantaphase performs acceptably because they chose it as the referee assay.) Contrary to the assertion of Carmel et al., their data suggest that they may improve detection of cobalamin deficiency in their laboratory by adjusting their medical decision point (“shifting the reference range”, as they put it). Two authoritative sources state that the delineation between normal and cobalamin-deficient serum concentrations is poorly defined (1)(2). The ADVIA Centaur VB12 Assay Method Manual corroborates this by providing ranges for both normal and cobalamin-deficient subjects, and these ranges overlap. Choosing the optimal medical decision point is a substantial endeavor. The predictive value of a test depends on disease prevalence in the population being screened—which may vary considerably among laboratories—as well as on the inherent clinical sensitivity and specificity of the test (3). For these reasons, it is impossible for a manufacturer to recommend a medical decision point suitable for all laboratories. Thus, the ADVIA Centaur Assay Method Manual encourages individual laboratories to establish their own reference ranges, but Carmel et al. fail to provide any information that they have done so. They apparently have chosen as their medical decision point the 2.5th percentile of the normal population from the ADVIA Centaur VB12 Assay Method Manual. This decision point optimizes the clinical specificity of the assay at the expense of clinical sensitivity, which at least partially explains why Carmel et al. might be missing cobalamin-deficient patients. We believe the best way to compare clinical performance of tests is through ROC curve analysis (4). Such a comparison automatically compensates for systematic analytical biases among the methods and the prevalence of disease in the study population. Such an analysis requires a symmetric study design and that each test be treated equally. A study design that uses the RIDA (even in combination with other clinical and laboratory information) to identify cobalamin-deficient patients is biased against any other cobalamin assay. The study by Carmel et al. also falls short of identifying clinical false negatives with the RIDA. Furthermore, we advise extreme caution before drawing conclusions from samples stored at −20 °C for up to 10 years (Dr. Carmel has told us that this is the approximate storage interval) when an authoritative source (1) indicates that samples stored at −20 °C are stable for 8 weeks. Finally, 17 banked diagnosed cobalamin-deficient samples used to determine the deficient range of the ACS:180 assay in 1996 and stored since then at −80 °C were run concurrently on the ACS:180, ADVIA Centaur, and where possible, the Bio-Rad RIA. Cutoffs were taken from the manufacturers’ package inserts (211 ng/L = 156 pmol/L for ACS:180 and ADVIA Centaur; 200 ng/L = 148 pmol/L for Bio-Rad RIA). The results in Table 1 show that the recent determinations compare favorably with historical data and show similar diagnostic classifications across methods. An additional 49 low-cobalamin samples (range, 144–389 ng/L; mean, 222 ng/L) were obtained and tested on the ADVIA Centaur and Bio-Rad RIA, and clinical concordance between the two methods was 92%. On the basis of the data presented in the two studies above, we conclude that the ADVIA Centaur VB12 assay provides accurate cobalamin values consistent with the clinical interpretation. Table 1. Comparison of cobalamin results. Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L 1 L, low; N, normal. 2 Not run, sample not run because of insufficient volume. 3 Samples 12–17 contain antibodies to intrinsic factor. Open in new tab Table 1. Comparison of cobalamin results. Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L 1 L, low; N, normal. 2 Not run, sample not run because of insufficient volume. 3 Samples 12–17 contain antibodies to intrinsic factor. Open in new tab References 1 Tietz NW. Clinical guide to laboratory tests 1995 : 636 -639 WB Saunders Philadelphia. . 2 Jacobs DS, Kasten BL, Demott WR, Tilzer I. Laboratory test handbook 1996 : 350 -352 Lexi-Comp, Inc. Hudson, OH. . 3 Galen SG, Gambino SR. Beyond normality: the predictive value and efficiency of medical diagnoses 1975 : 10 -14 John Wiley & Sons New York. . 4 National Committee for Clinical Laboratory Standards. Assessment of the clinical accuracy of laboratory testing using receiver operating characteristic (ROC) plots; approved guideline. NCCLS Document GP10-A. Villanova, PA: NCCLS, December 1995.. © 2000 The American Association for Clinical Chemistry This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Clinical Chemistry Oxford University Press

Representatives of Bayer Diagnostics respond

Download PDF
2 pages

Loading next page...
 
/lp/oxford-university-press/representatives-of-bayer-diagnostics-respond-GR2vmLsIsG

References (1)

Publisher
Oxford University Press
Copyright
© 2000 The American Association for Clinical Chemistry
ISSN
0009-9147
eISSN
1530-8561
DOI
10.1093/clinchem/46.12.2018
Publisher site
See Article on Publisher Site

Abstract

To the Editor The data provided by Carmel et al. (their Table 1 )suggest that they might be missing cobalamin-deficient patients because they failed to choose an appropriate medical decision point for their patient population with the Bayer ADVIA Centaur VB12 assay. Although the data suggest a “systematic” bias between the RIDA and the ADVIA Centaur, the dispersion (the standard deviation of the difference between ADVIA Centaur and RIDA results) is equivalent to that for Quantaphase vs RIDA. The ADVIA Centaur-vs-RIDA dispersion is 77 pmol/L, and the Quantaphase-vs-RIDA dispersion is 75 pmol/L; these are essentially the same (P = 0.926 by the F-test). In fact, after correcting for systematic bias, ADVIA Centaur and Quantaphase agree better with each other (dispersion of 65 pmol/L) than does either with RIDA. (We assume that Carmel et al. believe that the Quantaphase performs acceptably because they chose it as the referee assay.) Contrary to the assertion of Carmel et al., their data suggest that they may improve detection of cobalamin deficiency in their laboratory by adjusting their medical decision point (“shifting the reference range”, as they put it). Two authoritative sources state that the delineation between normal and cobalamin-deficient serum concentrations is poorly defined (1)(2). The ADVIA Centaur VB12 Assay Method Manual corroborates this by providing ranges for both normal and cobalamin-deficient subjects, and these ranges overlap. Choosing the optimal medical decision point is a substantial endeavor. The predictive value of a test depends on disease prevalence in the population being screened—which may vary considerably among laboratories—as well as on the inherent clinical sensitivity and specificity of the test (3). For these reasons, it is impossible for a manufacturer to recommend a medical decision point suitable for all laboratories. Thus, the ADVIA Centaur Assay Method Manual encourages individual laboratories to establish their own reference ranges, but Carmel et al. fail to provide any information that they have done so. They apparently have chosen as their medical decision point the 2.5th percentile of the normal population from the ADVIA Centaur VB12 Assay Method Manual. This decision point optimizes the clinical specificity of the assay at the expense of clinical sensitivity, which at least partially explains why Carmel et al. might be missing cobalamin-deficient patients. We believe the best way to compare clinical performance of tests is through ROC curve analysis (4). Such a comparison automatically compensates for systematic analytical biases among the methods and the prevalence of disease in the study population. Such an analysis requires a symmetric study design and that each test be treated equally. A study design that uses the RIDA (even in combination with other clinical and laboratory information) to identify cobalamin-deficient patients is biased against any other cobalamin assay. The study by Carmel et al. also falls short of identifying clinical false negatives with the RIDA. Furthermore, we advise extreme caution before drawing conclusions from samples stored at −20 °C for up to 10 years (Dr. Carmel has told us that this is the approximate storage interval) when an authoritative source (1) indicates that samples stored at −20 °C are stable for 8 weeks. Finally, 17 banked diagnosed cobalamin-deficient samples used to determine the deficient range of the ACS:180 assay in 1996 and stored since then at −80 °C were run concurrently on the ACS:180, ADVIA Centaur, and where possible, the Bio-Rad RIA. Cutoffs were taken from the manufacturers’ package inserts (211 ng/L = 156 pmol/L for ACS:180 and ADVIA Centaur; 200 ng/L = 148 pmol/L for Bio-Rad RIA). The results in Table 1 show that the recent determinations compare favorably with historical data and show similar diagnostic classifications across methods. An additional 49 low-cobalamin samples (range, 144–389 ng/L; mean, 222 ng/L) were obtained and tested on the ADVIA Centaur and Bio-Rad RIA, and clinical concordance between the two methods was 92%. On the basis of the data presented in the two studies above, we conclude that the ADVIA Centaur VB12 assay provides accurate cobalamin values consistent with the clinical interpretation. Table 1. Comparison of cobalamin results. Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L 1 L, low; N, normal. 2 Not run, sample not run because of insufficient volume. 3 Samples 12–17 contain antibodies to intrinsic factor. Open in new tab Table 1. Comparison of cobalamin results. Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L Sample . Bio-Rad RIA . . ACS:180 . . ADVIA Centaur . . 1996 value, ng/L . 2000 value, ng/L . 1996 value, ng/L . 2000 value, ng/L . 2000 value, ng/L . 1 156 L1 Not run2 176 L 176 L 195 L 2 174 L Not run 160 L 182 L 186 L 3 164 L Not run 152 L 171 L 155 L 4 202 N Not run 201 L 196 L 199 L 5 148 L Not run 170 L 159 L 165 L 6 141 L Not run 204 L 172 L 181 L 7 186 L 176 L 215 N Not run Not run 8 236 N 206 N 210 L 188 L Not run 9 109 L Not run 151 L 150 L 162 L 10 182 L Not run 161 L 148 L Not run 11 254 N 250 N 246 N Not run Not run 123 140 L 154 L 170 L 161 L Not run 13 92 L Not run 187 L 159 L 181 L 14 55 L Not run 157 L 148 L 164 L 15 60 L Not run 156 L 168 L 158 L 16 65 L Not run 155 L 152 L 154 L 17 77 L Not run 119 L 115 L 141 L 1 L, low; N, normal. 2 Not run, sample not run because of insufficient volume. 3 Samples 12–17 contain antibodies to intrinsic factor. Open in new tab References 1 Tietz NW. Clinical guide to laboratory tests 1995 : 636 -639 WB Saunders Philadelphia. . 2 Jacobs DS, Kasten BL, Demott WR, Tilzer I. Laboratory test handbook 1996 : 350 -352 Lexi-Comp, Inc. Hudson, OH. . 3 Galen SG, Gambino SR. Beyond normality: the predictive value and efficiency of medical diagnoses 1975 : 10 -14 John Wiley & Sons New York. . 4 National Committee for Clinical Laboratory Standards. Assessment of the clinical accuracy of laboratory testing using receiver operating characteristic (ROC) plots; approved guideline. NCCLS Document GP10-A. Villanova, PA: NCCLS, December 1995.. © 2000 The American Association for Clinical Chemistry This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Clinical Chemistry – Oxford University Press

Published: Dec 1, 2000

There are no references for this article.