Clinical Chemistry 46, No.

10, 2000 1719

in-house ELISAs for antineutrophil cytoplasmic step between the addition of sample range 0.31-17.90 g/L, mean differ-
antibodies directed against proteinase 3 and
myeloperoxidase. Pathology 1999;31:38 – 43. and labeled antibody (7 ) and the use ence 12.6%, SD 8.6%; and IgM, range
of neural network classifier systems 0.27-5.96 g/L, mean difference
Anthony A.M. Ermens1* that analyze reaction kinetics (13 ). 13.2%, SD 8.2%. The small SD indi-
Angelique J.M. Bayens1 Serum immunoglobulins can be cated that none of the samples exhib-
Adriënne Crooymans1 markedly increased in patients pre- ited the prozone effect. A percentage
Anita A.M. Broekman-van Hout2 senting with large myeloma tumor difference less than the mean plus 2
Hans L.P. van Duijnhoven2 burdens and may lead to falsely low SD was considered acceptable and
results in nephelometric assays (11 ). was determined to be 15% for IgG,
1
Klinisch Laboratorium We combine 50-␮L aliquots from 30% for IgA, and 30% for IgM. Large
Diaconessenhuis each of 10 samples to dilute each differences were considered sugges-
Postbus 90052 sample 10-fold and eliminate any tive of a prozone effect.
5600 PD Eindhoven, The Netherlands prozone effect. The concentrations of The ability of this approach to
IgG, IgA, and IgM in the pool are identify samples exhibiting the pro-
2
Algemeen Klinisch Laboratorium measured using a nephelometer zone effect during routine analysis
Elkerliek Ziekenhuis (BNII; Dade Behring, Inc.) and com- was evaluated during a 6-month pe-
5800 AB Helmond, The Netherlands pared with the mean values when all riod. Approximately 750 samples/
samples in the pool are analyzed month were received, and 460 pools
(calculated value). When the two val- were analyzed. Ten samples from
*Author for correspondence. Fax 31-40- ues for an immunoglobulin differ by five different myeloma patients were
2335595; e-mail [email protected].
a specified quantity, all samples in identified as being falsely low be-
the pool are reanalyzed after a 10- cause of the prozone effect (Table 1).
fold dilution. Four samples were from patients
Criteria for detecting the prozone with IgA myeloma, and one was
effect are based on data obtained from a patient with IgG myeloma.
Dilution Protocols for Detection of from routine samples during a 10- The discrepancy between the mea-
Hook Effects/Prozone Phenomenon day period. Measured immunoglob- sured and calculated pool was 62-
ulin concentrations for 27 pools (10 88% (initial difference; Table 1).
To the Editor: samples per pool) were compared When the sample generating the er-
The prozone or (high-dose) hook ef- with the mean values of samples in roneous value was identified and the
fect, documented to cause false-neg- the pools. The range of values for the “correct” result (obtained after dilu-
ative assay results ⬎50 years ago (1 ), measured serum pools and the dif- tion) was used in the calculation, the
still remains a problem in one-step ferences between the measured pool difference between the measured
immunometric assays (2–9 ), immu- value and the value derived from the and calculated pool was within the
noturbidimetric assays (10 ), and im- mean of individually measured sam- established limits of 30% for IgA and
munonephelometric assays (11 ) for ples in the pool (calculated value) for 15% for IgG (corrected difference;
immunoglobulins. To detect the pro- each immunoglobulin were as fol- Table 1). The falsely low values dif-
zone effect, samples are often tested lows: IgG, range 10.20-32.50 g/L, fered from the actual results by as
undiluted and after dilution (9 ). If mean difference 4.6%, SD 4.1%; IgA, much as 11-fold for IgA and 40-fold
the result on dilution is higher than
for the undiluted sample, then the
undiluted sample most likely exhib- Table 1. Detection of the prozone effect in nephelometric assays for
ited the prozone effect. Unfortu- immunoglobulin (Ig) by monitoring the percentage of difference between
nately, this approach increases labor measured and calculated pool values.
and reagent costs for assays that may Difference between
only rarely encounter extremely high measured and calculated
pool values,c %
analyte concentrations. An alterna-
tive approach involves pooling pa- Myeloma Original Ig Diluted Ig Initial Corrected
tient samples and measuring the Patient class result,a g/L result,b g/L difference difference
pool and a 10-fold dilution of the 1 IgA 8.26 59.00 64 7
pool (12 ). If one or more of the sam- 2 IgA 3.91 43.10 77 16
ples in the pool is falsely low because 3 IgA 8.46 47.60 75 5
of the prozone effect, then the results 4 IgA 7.21 77.90 88 21
from the undiluted and diluted pools 5 IgG 3.08 126.00 62 2
(after correcting for the 10-fold dilu- a
Result was obtained for either IgA or IgG depending on the myeloma class.
tion) will differ significantly (12 ). b
Samples were diluted 10-fold before re-analysis.
Other approaches to eliminate the c
An aliquot (50 ␮L) from 10 samples was combined and assayed. The measured value for the pool was
prozone effect include using two- compared with the sum of the 10 individual measurements divided by 10. The percentage of difference
between the two values is shown.
step immunoassays that have a wash
1720 Letters

for IgG (Table 1). The prozone effect Clayman RV, Nahm MH. Extremely high values We report on the effectiveness of a
of prostate-specific antigen in patients with
is not restricted to IgA and IgG be- adenocarcinoma of the prostate; demonstra- simple, inexpensive method reported
cause we identified samples exhibit- tion of the “hook effect”. Clin Chem 1988;34: by Cole et al. (3 ) that uses a pooled
ing this phenomenon when measur- 2175-7. sample to detect “hook effect”.
8. Pesce MA. “High-dose hook effect” with the
ing IgM (data not shown). Centocar CA 125 assay. Clin Chem 1993;39:
Cole et al. (3 ) recommend batching
A 2% incidence (1 of 46 pools) for 1347. patient samples in groups of 10,
the prozone effect when measuring 9. Saryan JA, Garrett PE, Kurtz SR. Failure to forming a pooled sample with each
detect extremely high levels of serum IgE with
immunoglobulins may be higher an immunoradiometric assay. Ann Allergy
sample diluted 10-fold by the other
than at institutions not specializing 1989;63:322-4. samples in the batch. In addition, the
in the treatment of multiple my- 10. Jury DR, Mikkelsen DJ, Dunn PJ. Prozone effect pooled sample is diluted 10-fold,
and the turbidimetric measurement of albumin
eloma. However, the incidence of in urine. Clin Chem 1990;36:1518-9.
producing a 100-fold final dilution.
multiple myeloma over the age of 25 11. Van Lente F. Light scattering immunoassays. “Hook” samples produce a higher
is 30 per 100 000 (14 ), and most lab- In: Rose NR, de Macario EC, Folds JD, Lane result for the 100-fold dilution pool
oratories will eventually encounter a HC, Nakamura RM, eds. Manual of clinical
laboratory immunology, 5th ed. Washington,
than the 10-fold dilution pool. Each
sample exhibiting the prozone effect DC: ASM Press, 1997:13-9. of the 10 samples must then be rean-
when measuring immunoglobulins 12. Cole TG, Johnson D, Eveland BJ, Nahm MH. alyzed at a higher dilution to detect
by nephelometry. Reporting of an Cost-effective method for detection of “hook
effect” in tumor marker immuometric assays.
the out-of-range result. Cole et al.
erroneous result can have serious Clin Chem 1993;39:695-6. also describe a modification in which
medical implications, and sample 13. Papik K, Molnar B, Fedorcsak P, Schaefer R, up to 30 patient samples are pooled.
pooling is a simple method for de- Lang F, Sreter L, et al. Automated prozone
effect detection in ferritin homogeneous immu-
For the past 6 years, we have used
tecting falsely low concentrations at- noassays using neural network classifiers. Clin the modified protocol described by
tributable to the prozone effect. Al- Chem Lab Med 1999;37:471-6. Cole et al. (3 ) in combination with
though this screening approach 14. Cooper MD, Lawton AR. Disorders of the im- predilution of samples from patients
mune system. In: Braunwald E, Isselbacher
increases reagent costs by 10% and KJ, Petersdorf RG, Wilson JD, Martin JB, known to have extremely high re-
involves additional labor to prepare Fauci AS, eds. Harrison’s principles of inter- sults. We use 50 ␮L of sample from
and analyze pools, it is considerably nal medicine. New York: McGraw-Hill, 1987:
1396-403. each patient sample in an analytical
more cost-effective than analyzing all run to create a pooled sample. Each
samples undiluted and after dilution, run usually contains 20 –30 patient
which doubles reagent costs. Further- Anthony W. Butch samples, so the final pool dilutes
more, this simple prozone detection each patient sample 20- to 30-fold.
method can be adapted to other neph- University of Arkansas The answer for the pooled result
elometric assays with the potential for for Medical Sciences should not exceed the highest patient
erroneous results from antigen excess. Department of Pathology result in the analytical run. If the
4301 West Markham pooled sample is higher than the rest
Little Rock, AR 72205 of the patients, all samples are re-
References peated after dilution.
1. Landsteiner K. The specificity of serological With this protocol, we have found
reactions. Cambridge, MA: Harvard University
Press, 1946;240-52. two samples, one for prostate-spe-
2. Brensing AK, Dahlmann N, Entzian W, Bidling- Address correspondence to this author cific antigen (PSA) and one for CA-
maier F, Klingmuler D. Underestimation of LH at: UCLA Medical Center, Department of 125, with falsely low values attribut-
and FSH hormone concentrations in a patient Pathology and Laboratory Medicine,
with a gonadotropin secreting tumor: the high 10833 Le Conte Ave., Mailroom A2-179 able to antigen excess. The most
dose “hook effect” as a methodological and CHS, Los Angeles, CA 90095-1713. Fax recent was a CA-125 value that gave
clinical problem. Horm Metab Res 1989;21:
697-8.
310-794-4864; e-mail abutch@mednet. a result of 375 units/mL when ana-
ucla.edu. lyzed undiluted. The pooled sample
3. Haller BL, Fuller KA, Brown WS, Koenig JW,
Evelend BJ, Scott MG. Two automated prolactin from that analytical run had a value
immunoassays evaluated with demonstration
of a high-dose “hook effect” in one. Clin Chem
of ⬎500 units/mL (reportable range,
1992;38:437-8. 15–500 units/mL). The final patient
4. Petakov MS, Damjanovic SS, Nikolic-Durovic result was 23 000 units/mL. The pa-
MM, Dragojlovic ZL, Obradovic S, Gilgorovic To the Editor:
MS, et al. Pituitary adenomas secreting large
tient had no previous laboratory re-
amounts of prolactin may give false low Hook effect is an infrequent event sults at our institution, so the error
values in immunoradiometric assays. The that is notoriously difficult to detect would not have been detected by
hook effect. J Endocrinol Invest 1998;21:
184-8. in the clinical laboratory (1 ). One of delta checking and may not have
5. Flam F, Hambraeus-Jonzon K, Hansson LO, the best methods of detection is to been apparent to the ordering physi-
Kjaeldgaard A. Hydatidiform mole with non- run samples both undiluted and di- cian. Although the manufacturer
metastatic pulmonary complications and a
false low level of hCG. Eur J Obstet Gynecol luted (2 ). Any sample that does not claims that this assay will not hook
Reprod Biol 1998;77:235-7. dilute properly may have antigen back into the normal range until con-
6. Zweig MH, Csako G. High-dose hook effect in a excess. This method prevents the re- centrations exceed 100 000 units/mL
two site IRMA for measuring thyrotropin. Ann
Clin Biochem 1990;27:494-5. porting of falsely low results but (CA-125 II product insert; Centocor
7. Vaidya HC, Wolf BA, Garrett N, Catalona WJ, incurs substantial time and expense. Diagnostics Division), erroneous re-