Interlaboratory
Bilirubin
Variability
Richard L. Schreiner, MD, and Melvin R. Glick, PhD
From the Departments of Pediatrics and Pathology, Indiana University School of Medicine and James Whitcomb Riley Hospital for Children, Indianapolis
ABSTRACT. Stabilized liquid quality control sera, some with elevated biirubin concentrations, were distributed to laboratories for total and conjugated (direct) bilirubin
analyses. Interlaboratory variability was high; for
exam-ple, results ranged from 10.9 to 24.0 mg/100 ml for a
serum with a mean biirubin concentration of 18.1 mgI 100 ml. Coefficients of variation were typically 10% to 12% for total biirubin and approximately 24% for conju-gated bilirubin analyses. It is believed that these stabi-lized sera, when used as quality control materials, can help reduce the unacceptably large interlaboratory van-ability found in this study. Pediatrics 69:277-281, 1982; bilirubin, jaundice, laboratory variability.
Pediatricians and other clinicians rely on single
or serial bilirubin determinations for jaundiced
neo-nates. They assume that reliable results emanate
from the clinical laboratories, and that these results
are reproducible and comparable from one
labora-tory to another. In 1960, Mather’ described the
then-prevailing state of affairs:
Although bilirubin determinations are perhaps the most notoriously unreliable of any in clinical chemistry, in
situations where the pediatrician has a fair degree of
confidence in the laboratory results, it is my impression that he often goes overboard in demanding precision, riding the infant’s biirubin “curve” as he would a bron-cho-every small buck is to be combatted with spurs, a hard bit and a platelet pack; each small subsidence is a victory.
The College of American Pathologists’
comprehen-sive chemistry survey in 1968 certainly supported
this pessimistic view of the accuracy of biirubin
measurements in the clinical chemistry
laborato-ries. At 1,500 participating laboratories, total
bili-Received for publication June 3, 1981; accepted July 27, 1981. Presented in part at the 33rd National Meeting of the American
Association for Clinical Chemistry, Kansas City, MO, July 1981.
Reprint requests to (R.L.S.) Department of Pediatrics, Indiana
University School of Medicine, 1100 Michigan St, Indianapolis,
IN 46223.
PEDIATRICS (ISSN 0031 4005). Copyright © 1982 by the
American Academy of Pediatrics.
rubin concentrations for lyophilized quality control
sera were determined. When the mean
concentra-tion was 13.5 mg/100 ml, the coefficient of variation
(CV) was 13.8%. Good performance limits (+1.5
SD) were judged to be 10.7 to 16.3 mg/100 ml and
acceptable performance (±2 SD), 9.8 to 17.4 mg/
100 In 1970, a similar survey showed a CV of
14.6% for a mean bilirubin concentration of 15.8
mg/100 ml with good performance (±1 SD) 13.5 to
18.1 mg/100 ml and acceptable performance (±2
SD) 11.2 to 20.4 mg/100 2 There are no reasons,
nor any data, to suggest that the situation has
improved since then, although advances in
instru-mentation may have affected the analytic precision
which could be obtained.3
One major problem in the quality control and
interlaboratory comparison of biirubin
measure-ments is the lack of reliable standards and external control materials with high bilirubin concentrations
in the range expected from jaundiced newborn
in-fants.4 The purpose of this study was to evaluate
the current accuracy and precision of bilirubin
mea-surements in hospital laboratories by using a new
stabilized biirubin solution.5
MATERIALS AND METHODS
Preparation and Storage of Material
Four different concentrations (levels 1 to 4,
Ta-ble) of Ultimate-D (Beckman Instruments Inc,
Brea, CA), a serum-based biirubin solution
stabi-lized by the addition of ethylene glycol, were used.
Plasma samples with elevated biirubin
concentra-tions were also pooled from the preexchange blood
of selected infants undergoing exchange
transfu-sions. Aliquots of these plasma samples were placed
in test tubes and kept frozen at -20 C until the
survey.
Transportation and Analysis
After communication with laboratory personnel
Se-TABLE. Intenlaboratory Survey of Biirubin Determinations
278 BILIRUBIN VARIABILITY
Survey Specimen Identification
Total Biirubin (mg /100 ml) Conjuga ted Biirubin (mg/i 00 ml)
Mean Range SD CVf Mean Range SD CV
(%) (%)
Level 1 1.3
n=107
0.0-3.2 0.4 29.7 0.7
n=96
0.0-1.1 0.2 32.6
Level 2 5.9
n=140
2.3-8.0 0.7 12.1 3.8
n=95
1.2-6.0 0.9 22.1
Level 3 9.1
n=139
6.8-12.1 0.9 10.5 5.8
n=95
1.8-9.2 1.4 23.6
Level 4 18.1
n=135
10.9-24.0 1.9 10.4 11.9
n=95
5.2-22.5 2.8 23.5
Preexchange 9.1
n=93
0.7-12.9 1.7 19.1 0.7
n=61
0.0-2.5 0.6
-4’Includes all results submitted. t CV, Coefficient of variation.
:1:n, No. of determinations.
§Non-Gaussian distribution.
rum specimens were transported directly to the
hospital laboratory. All biirubin solutions were
pro-tected
from light and were kept at 0 C or colderwhile in transit to the laboratories of survey partic-ipants. With each group of serum samples delivered,
a messenger made a circuit through the geographic
area to deliver the specimens. An extra aliquot of
each level of quality control serum was transported
with the test samples; this extra aliquot was
re-turned with the messenger for analysis of biirubin
in our laboratory. In this way the integrity and
stability of each shipment of control sera were
monitored. Participants were asked to determine
the total and conjugated (direct-reactmg) bilirubin
concentration of each quality control serum
pro-vided. A questionnaire was included for reporting
the method, instrument, and biirubin standard
so-lution used for calibration of the instrument. The
laboratory technologists were not aware of the
ex-pected bilirubin values for the sera.
Determination of the mean, range, SD, and CV
for the histograms was generated by standard com-puter programs.
During the two months of this survey, 141 sets of
results were reported to us from 67 laboratories.
Among the larger groups of identified analytical
systems there were: 21 acas (automated clinical
analyzers) (DuPont Instruments, Wilmington, DE);
11 continuous-flow instruments (Technicon
Cor-poration, Tarrytown, NY); six centrifugal analyzers;
and 29 single-beam manual spectrophotometers,
mainly of various models manufactured by either
Gilford Instruments Laboratories, Inc (Oberlin,
OH),
Coleman Instruments Division, Perkin-ElmerCorp (Norwalk, CT), or Turner Associates (Palo
Alto, CA). Dual wavelength direct spectroscopic
analysis (not diazo reaction) was the method from
21 laboratories; of these, nine used the neonatal
method available on the aca, five used model
BR-II
bilirubinometers (Advanced Instruments, Inc,Needham Heights, MA), and five used A/O
biiru-binometers (American Optical, Buffalo).
When not specified by the constraints of the
instrument, most of the diazo methods selected
were modifications of the Jendrassik and Grof
pro-cedure.6 Three laboratories used dimethylsulfoxide
as a catalyst for nonconjugated biirubin reaction.
Calibration materials generally were those of the
manufacturer’s specification; otherwise, products
from American Monitor Corp (Indianapolis, IN)
and General Diagnostics (Morris Plains, NJ)
pre-dominated.
RESULTS
The mean, range, SD, and CV of reported
bii-rubin concentrations are shown in the Table. The
wide ranges of results reported for quality control
specimens with biirubin concentrations in the two
higher ranges are further highlighted by the
histo-grams given in Figs 1 and 2. For comparison, Fig 3
shows the range of values obtained from our
labo-ratory with a modified Nosslin method during the
two-month survey period.7 Thirteen different
tech-nologists each performed at least one determination
during this time. During this same time the
intra-laboratory CV was 3.4% (mean biirubin
concentra-tion 19.6 mg/100 ml) for a “known” quality control
serum (General Diagnostics).
Stability of the new liquid quality control
mate-rial is confirmed by the results for the specimens
transported back to us by the messenger. The range
of results in our laboratory was within ±0.3 mg/100
ml for all mean biirubin concentrations below 10
mg/100 ml; and ±0.5 mg/100 ml for the higher
108 113 118 123 128 133 138 143 148 153 158 163 168 173 178 183 188193 198203208 213 218 223228233238
Total bilirubin (mg/dl±O.2)
U C a 3
g
>.
U
a, a.
La-Fig 1. Biirubin results from all participating laboratories (open bars) with use of level 4 stabilized serum. Shaded bars represent results obtained from direct spectroscopic (no
chemical reaction) estimations of biirubin.
15 18 81 84 81 90 93 96 99 102 105 108 11.1 114 111 120
Total bilirubin (mg/dl t 0.1)
Fig 2. Bilirubin analyses on level 3 stabilized serum reported by participating laboratories in which either diazo methods (open bars) or direct spectroscopic (shaded bars) methods were used.
DISCUSSION
The results of this study demonstrate the lack of
accuracy of serum bilirubin measurements,
espe-cially at the high concentrations seen in some
new-born infants. These results probably will not be
surprising to clinicians who care for a large number
of sick newborns, especially when these infants are
transferred from many different hospitals. The
re-suits reported here are consistent with the
sugges-tion that little or no improvement in intenlaboratory
biirubin results has occurred in the past 20 years, at least for elevated bilirubin concentrations.
It is also worth noting that while the results of
this bilirubin survey are consistent with those
re-ported earlier, the usual reporting of statistical
re-suits alone does not allow a full appreciation of the
whole story. Indeed, although we have chosen to
include all reported values in our calculations, most
survey reports exclude these “outliers” for valid and
well accepted (with statistically derived criteria)
reasons before the final calculation. For example,
the total biirubin results obtained for the level 4
serum with a mean biirubin concentration of 18.1
mg/100 ml and an SD of 1.9 mg/100 ml imply that
approximately 95% of the laboratories would report
biirubin concentrations between 14.3 and 21.9 mgI
100 ml. While such a range seems undesirable, those
results outside this too-wide range are even more
disturbing. The results greater than ±2 SD were
10.9, 11.6, 22.0, and 24.0 mg/100 ml. While results
such as these would warrant dramatically different
treatment of the newborn, the evidence suggests
that actual newborn serum bilirubin results
fre-quently show wider variation than the data
re-ported here. Our data probably represent better
than usual results.
No one wifi dispute the importance of accuracy
and precision in the determination of serum
biliru-bin concentration in the newborn infant. In older
children and adults this accuracy and precision are
not nearly so important; however, in the newborn,
damage due to the deposition of bilirubin in the
brain (kernicterus) results in devastating neurologic
damage. Prevention of kernicterus consists of
screening and monitoring serum biirubin
2
108 113 118 123 128 133 138 143148 153 158 163 168 173 178 183 188193 198203208 213 218 223228233238
. Total bilirubin (mg/dl±O.2)
U C a,
a.
a,
U-Fig 3. Intralaboratony bilirubin results (shaded bars) from level 4 sera determined by
our laboratory. During two months covered by this survey, 28 determinations by 13
technologists were performed. Open bars show all other results reported from 67 partici-pating laboratories.
decide the proper time to implement therapy.
How-ever, in order to be able to decide reasonably when
to implement therapy one must have a reliable
measurement of the biirubin concentration. Other
laboratory tests, such as column chromatography,
and most recently, hematofluorometry,8 have been
advocated to evaluate the ability of the patient’s
serum proteins to bind more biirubin, thus
provid-ing assistance in deciding upon the appropriate time
for exchange transfusion. However, serum bilirubin
concentration and clinical evaluation of the patient
remain the major parameters considered in such a
decision.9”#{176}
Numerous studies have shown a lack of a
cmi-cally useful correlation of serum bilirubin with the
risk of kernicterus, especially in very
low-birth-weight infants.”2 All or nearly all of these studies
have been retrospective in design, in which the
serum bilirubin measurements were determined by
routine methods in the clinical chemistry
labora-tory. One must wonder whether this lack of a
posi-tive correlation is, at least in some part, due to the lack of accuracy or precision of the serum bilirubin
measurement.
We believe one of the major problems in the lack
of accuracy and precision in biirubin measurement
today in the clinical chemistry laboratory is the
relative lack of emphasis on quality control for
bilirubin concentrations greater than 8 mg/100 ml.
In addition, some laboratory instruments are
call-brated at lower concentrations than are appropriate
if serum from a newborn is to be used. The recently
introduced liquid biirubin solutions for quality
con-trol and calibration allow us to evaluate this range
of values. We suggest that the use of these new
materials for calibration and routine quality control
may help eliminate much of the interlaboratory
variability demonstrated in this survey.
ACKNOWLEDGMENT
The four concentrations of Ultimate-D used in this
investigation were supplied by Beckman Instruments, Inc, Brea, CA.
REFERENCES
1. Mather A: Reliability of biirubin determinations in icterus of the newborn infant. Pediatrics 26:350, 1960
2. Doumas BT, Perry BW, Sasse EA, et al: Standardization in
bilirubin assays: Evaluation ofselected methods and stability
of biirubin solutions. Clin Chem 19:984, 1973
3. Ross JW, Fraser MD, Moore TD: Analytic clinical
labora-tory precision-State of the art for thirty-one analytes. Am J Chin Pathol 74:521, 1980
4. Lucey JF, Phillips CL, Utterback JG, et al: A difference in the incidence of hyperbilirubinemia among premature in-fants in two hospitals. Pediatrics 30:3, 1962
5. Louderback A, Jendrzejczak B, Doumas BT, et al: A new
approach in stabilization of a biirubin standard. Fresenius
ZAnal Chem 301:145, 1980
6. Gambino SR: Bilirubin (modified Jendrassik and
Grof)-Provisional, in Meites S (ed): Standard Methods of Clinical Chemistry. New York, Academic Press, 1965, vol 5, p 55
7. Nosslin B: The direct reaction of bile pigments in serum-Experimental and clinical studies. Scand J Chin Lab Invest 12(Suppl 49):1, 1960
assay of biirubin and biirubin binding capacity in blood of 11. Kim MH, Yoon JJ, Sher J, et al: Lack of predictive indices jaundiced neonates: Comparisons with other methods. Pe- in kernicterus: A comparison of clinical and pathologic
fac-diatrics 66:411, 1980 tors in infants with or without kernicterus. Pediatrics 66:
9. Gitzelmann-Cumarasamy N, Kuenzle CC: Biirubin binding 852, 1980
tests: Living up to expectations? Pediatrics 64:375, 1979 12. Turkel SB, Guttenberg ME, Moynes DR. et al: Lack of
10. Levine RL: Bilirubin: Worked out years ago? Pediatrics 64: identifiable risk factors for kemicterus. Pediatrics 66:502,
380, 1979 1980
THE FIRST AMERICAN PLEA FOR A PERMANENT HOSPITAL FOR SICK
CHILDREN (1852)
In 1852, James Stewart (1799-1864), under the pen name “Phiopedos,”
originated a plan for the establishment of a children’s hospital in New York
City. When the institution was opened on March 1, 1854, under the name of the
New York Nursery and Child’s Hospital, it was the only hospital on this
continent devoted to children. (There had been a hospital devoted exclusively
to sick children in Boston as early as 1846, but financial difficulties forced it to
close after a few years.)
“Phiopedos” wrote’:
It must be evident to all who will reflect upon the large amount of sickness there is among the children of the poor in our city [New York], that hospital accommodations for them are among its most urgent wants. In the dwellings of the very poor there is almost always ..absence of everything necessary for the ordinary relief of the sick, and especially of the unremitting attention that is needed by them. The necessity of constant occupation to obtain the means of existence, precludes the possibility ...of devoting any time to the requirements of the sick; and it is from this want of attendance, next to want of pure air, that children suffer most. For those who have the necessary comforts for the sick, or who
have the time that they may bestow upon their families, when they most require it,
dispensary attendance is sufficient for their wants in sickness; but when it is known that many children are absolutely destitute of all these-indespensable as they are-the necessity of proving well-ventilated accomodations is evident; a place where all the wants
of the sick [child] may be supplied, and especially when personal care must form an
essential pant of the arrangement;-a need only to be supplied by the establishment of a well-organized hospital. ...
There are not hospitals enough for the ordinary wants of our city, now containing more than half a million inhabitants. Among so large a number of people, many more hospitals than now exist could be filled with distinct classes, either of people or of
diseases. Where sickness among children exceeds to a great degree sickness among
adults ...there will always be a sufficient number of applicants to fill any number of hospitals that will be established for their exclusive use.
The need, also, of a special hospital for children is evident when it is considered that a large number is to be provided for, and ...[with] attendants adapted exclusively to them. ...
After a careful consideration of the truths here presented, surely no one can hesitate to do what is in his power to assist in the establishment and support of a Child’s Hospital in New York.
Noted by T.E.C., Jr, MD
REFERENCE
