Clinical Chemistry 46: 399-403, 2000;
(Clinical Chemistry. 2000;46:399-403.)
© 2000 American Association for Clinical Chemistry, Inc.
Cerebrospinal Fluid Protein Concentrations in Children: Age-related Values in Patients without Disorders of the Central Nervous System
Daniel Bioua,
Jean-François Benoist,
Claire Nguyen-Thi,
Xuan Huong,
Philippe Morel and
Martine Marchand
a Author for correspondence. Fax: 33-1-40-03-47-90; e-mail daniel.adrbp{at}wanadoo.fr
 |
Abstract
|
|---|
Background: The published reference values for cerebrospinal fluid
(CSF) total protein concentrations in children suffer from two major
drawbacks: (a) the age-related range often is too broad
when applied to the steeply falling concentrations in early infancy;
and (b) no values have been published for widely used
dry chemistry methods.
Methods: We conducted a 2-year retrospective survey of CSF
results obtained in a children’s hospital with a dry chemistry-based
method set up on the Vitros 700 analyzer.
Results: The data related to ambulatory children up to 16 years
of age and term neonates with no clinical or biological signs of brain
disease (n = 1074). Seven age groups with significantly different
CSF protein values were identified, and their age-related percentiles
(5th, 50th, and 95th) were determined. On the basis of the upper 95th
percentile, from age 0 to 6 months the CSF protein concentrations fell
rapidly from 1.08 to 0.40 g/L. A plateau (0.32 g/L) was reached from
age 6 months to 10 years, followed by a slight increase (0.41 g/L) in
the 10–16 years age range.
Conclusions: These results imply that CSF total protein
concentrations in the pediatric setting, particularly in infants, must
always be interpreted with regard to narrow age-related reference
values to avoid false-positive results.
|
Introduction
|
|---|
Measurement of total protein in cerebrospinal fluid (CSF) is used
mainly to detect various central nervous system (CNS) diseases
associated with either increased permeability of the blood-CSF barrier
or increased intrathecal immunoglobulin synthesis. The largest
increases are seen in bacterial meningitis, with smaller increases in
other inflammatory, traumatic, or tumoral disorders and after
intracerebral hemorrhage (1).
In the pediatric setting, CSF testing must be done rapidly and with a
small sample volume. The CSF protein assay on the Vitros 700 (formerly
the Kodak Ektachem; now available from Ortho-Clinical
Diagnostics) meets these conditions. However, the central 95%
percentile reference interval given by the manufacturer (120–600 mg/L)
and determined by Lott and Warren (2) apparently is based on
a small number of samples from adults only. The authors claimed there
were no statistically significant age-related differences. However,
because of the immaturity of the blood-CSF barrier in fetuses and
neonates, it is well known that CSF protein concentrations in newborn
infants, and especially premature newborns, are higher than in children
and adults (3).
The published reference values for CSF total protein concentrations
suffer from two major drawbacks: (a) the values are method
dependent (type of dye, precipitating agent, and calibrator); and
(b) the age-related range often is too broad when applied to
the steeply falling CSF protein concentrations in newborn infants
(Table 1
) (3)(4)(5).
Curiously, since the first report by Lott and Warren (2), no
other age-related reference intervals have been published for CSF
protein concentrations determined by dry chemistry methods in children.
We therefore conducted a 2-year retrospective survey of CSF results
obtained in a children’s hospital.
|
Patients and Methods
|
|---|
patients
Patients used for reference data.
For obvious ethical reasons,
CSF was not sampled in healthy children. We therefore used data on CSF
specimens obtained between May 1997 and May 1999 by lumbar puncture
from term infants (gestational age, 38–41 weeks) and children (up to
16 years) consulting Robert Debré Children’s Hospital for
suspected CNS disease (n = 2182). Also included in this study were
CSF samples from term newborns hospitalized in the maternity or
neonatology departments (n = 512). All specimens from children
hospitalized in other wards were discarded to exclude the many CNS
disorders and iatrogenic complications that can increase CSF protein
concentrations (1).
All specimens with the following features were excluded: pigmented,
turbid, purulent, or bloody aspect; presence of an erythrocyte pellet
after centrifugation; abnormal erythrocyte or leukocyte count; positive
Gram staining; positive microbiological culture; positive bacterial
antigens; antibiotic therapy before CSF sampling; and samples from
patients hospitalized for >1 day. The remaining 1074 specimens (1013
from ambulatory children and 61 from hospitalized newborns) were used
to define the reference population.
Patients used for method comparison.
All unselected
consecutive CSF specimens obtained during a 2-month period (n =
337) from hospitalized or consulting patients were included in the
study to obtain CSF protein values covering a very wide range (0.1–2
g/L, the upper limit of linearity). CSF samples large enough for
duplicate assays were centrifuged and assayed immediately with two
analyzers (see below).
assays methods
Vitros 700.
The Vitros 700 CSF protein determination is based
on biuret colorimetry, but with decreasing reflectance density (670 nm)
of the Cu2+-azoic dye complex when
Cu2+ is complexed by the peptide bond
(2). The instrument is calibrated with a mixture of a
polymer and purified human protein. An endpoint measurement is made 5
min after the addition of 10 µL of the CSF specimen to the slide. The
respective between-run imprecision for two controls (Liquid Performance
Verifier 1 and 2; Vitros) was as follows: n = 60;
x = 0.69 g/L; CV = 5.3%; and n = 59;
x = 1.46 g/L; CV = 3.4%.
Hitachi 911 (Roche Diagnostics).
In this method, CSF
protein reacts with a pyrogallol red-molybdate complex to form a
blue-purple complex that absorbs at 600 nm (6). In the
protocol provided by the manufacturer (Biotrol Diagnostic), an endpoint
measurement is made 10 min after the addition of 20 µL of the CSF
specimen. The calibrator is bovine albumin. The between-run imprecision
for two serum controls (Lyphochek; Bio-Rad) was as follows:
n = 63; x = 0.32 g/L; CV = 6.2%; and n
= 63; x = 0.70 g/L; CV = 2.9%.
statistical analysis
Because we could not predict the distribution of the
population in each age group, nonparametric tests were used to estimate
the statistical significance of the reference ranges. The
Kruskal–Wallis test was used to test for statistical significance of
differences in CSF protein values between the all age groups. When
significance was found, pairwise comparisons between successive age
groups were made with the Kolmogorov–Smirnov test. When there was no
statistically significant difference (P >0.05), the
successive age groups were pooled. The StatView program (SAS
Institute), Ver. 5.0, was used for statistical analysis.
|
Results
|
|---|
reference intervals
After birth, the distribution of CSF total protein concentrations
fell rapidly (Fig. 1
A). This fall was maximal during the first 100 days of life, and
was much less pronounced thereafter. Over an average 6 months to 1 year
age range, a plateau was reached. We divided the data into 11 age
groups: 1–8 days, 8–30 days, 1–2 months, 2–3 months, 3–4 months,
4–5 months, 5–6 months, 6 months to 1 year, 1–2 years, 2–10 years,
and 10–16 years. Applying the nonparametric Kruskal–Wallis test to
the age groups defined above, we found a highly significant age effect
on the distribution of CSF protein values. The nonparametric
Kolmogorov–Smirnov test was then applied to identify an age effect
between successive age groups. The difference was not statistically
significant for the three age groups between 3 and 6 months and the
three between 6 months and 10 years (results not shown). In contrast, a
significant difference was found between all of the other successive
paired age groups.
View larger version (32K):
[in this window]
[in a new window]
|
Figure 1. Age-related changes in CSF total protein values in
children.
(A), correlation in children up to 16 years of age.
(Inset), correlation in infants <1 year of age.
(B), relationship between the 5th, 50th, and 95th
percentiles of the CSF protein concentration and the mid point of each
age group. Median (x), upper and lower 95th percentiles ( ), 97.5th
and 2.5th percentiles ( ). Each age group is represented by its mid
point age, i.e., 4, 19, 46, 76, 136, 1918, and 4749 days. The
relationship is best described by a smoothed curve and a logarithmic
scale for age up to 16 years. (Inset), linear scale in
infants up to 6 months of age.
|
|
On the basis of these findings, the reference population was again
distributed into the following seven age groups: 1–8 days, 8–30 days,
1–2 months, 2–3 months, 3–6 months, 6 months to 10 years, and 10–16
years. The relationship between the 5th, 50th, and 95th percentiles of
the CSF protein concentration and the mid point of each age group (4,
19, 46, 76, 136, 1918, and 4749 days) gave a statistical estimate of
the age dependence of the reference interval (Fig. 1B
) and
confirmed the rapid fall between birth and 6 months of age. However,
from 10 years onward, a trend toward higher values was found. These
successive downward and upward trends in CSF protein concentrations
were highly significant for each of the above age groups when tested
with the Kolmogorov–Smirnov test (Table 2
). Inside a given age group, a more precise estimate of the
upper limit of the reference interval could be obtained by
interpolation of the patient’s age on the 95th percentile plot.
View this table:
[in this window]
[in a new window]
|
Table 2. Age-related percentiles of CSF total protein (g/L) in infants and
children, as determined by the Vitros
method.
|
|
method comparison
The equation of the regression line of best fit was y
(Hitachi) = 0.802x (Vitros) + 0.00039 g/L (n =
337; SE of the slope, 0.008; SE of the intercept, 0.004;
r = 0.985; mean Hitachi = 0.360 g/L; mean
Vitros = 0.449 g/L). The values determined by the pyrogallol red
method (y; Hitachi) correlated well with those obtained by
the copper-binding method (x; Vitros). However, regression
analysis showed a significant proportional negative bias, at the 95%
confidence interval, of the pyrogallol method (y) compared
with the copper-binding method.
|
Discussion
|
|---|
The CSF protein concentration depends on the serum protein
concentration and on the permeability of the blood-CSF barrier. Because
of the higher permeability of the immature barrier of preterm and term
infants, their CSF has higher protein content than that of children and
adults. Maturation of the barrier gradually increases after birth,
reaching its maximum at the third month of life (3). This
was shown by the age distribution of the values in the reference
population, with a maximal decrease during the first 6 months of life.
The rate of the fall during the first trimester demonstrates the poor
predictive performance of the CSF protein test when appropriate narrow
age groups are not used in this period.
Given the smallest sample size of our age groups (n = 26) and the
fact that the theoretical minimal sample sizes required for the
respective 97.5th and 95th percentile estimates are 40 and 20 values
(1), we defined the upper limit of the present reference
intervals as the 95th percentile. On the basis of this upper limit, the
physiological maturation of the blood-brain barrier during early
infancy is illustrated by a corresponding rapid fall (from 1.08 to 0.60
g/L) between birth and 3 months of age. Thereafter, the decline in CSF
protein was less pronounced, reaching a minimal value (0.32 g/L) in the
6 months to 10 years age range. From 10–16 years, a slight increase
was observed, in keeping with the significant increase reported by
Tibbling et al. (7) in aging adults (from 0.37 ± 0.06
g/L to 0.53 ± 0.13 g/L between 17–30 years and 61–77 years,
respectively). In agreement with the above finding, the 95% central
reference interval (2.5–97.5) of our oldest age group (10–16 years)
was lower (0.10–0.43 g/L) than that of adults (0.12–0.60 g/L), as
determined with the same analyzer used by Lott and Warren
(2), thus confirming the slight but continuous increase in
CSF protein concentrations starting in early adolescence. Nevertheless,
given the continuous increase in aging adults, the adult reference
interval reported by Lott and Warren (2) seems to be an
approximation of the true situation.
In the literature, the reference values for CSF protein often are
expressed as gaussian statistics (mean ± SD) and are often based
on small samples with no details of value distribution. This approach
is at the very least questionable. More worrying is that the age range
is often too large [0.15–0.45 g/L above 1 month of age
(4)] and thus fails to reflect the rapid fall in
physiological CSF protein concentrations during the first 3 months of
life. To our knowledge, the most complete report covering the entire
period of development through infancy and childhood is from Statz and
Felgenhauer (3). Our sample is even larger, enabling us to
define narrower reference intervals, particularly for the 1–2 months
and 2–3 months age ranges, during which the most rapid decline in CSF
protein concentrations occurs. Our data may lead to better diagnostic
performance (specificity, sensitivity, positive and negative predictive
values, and efficiency) of neonatal CSF protein assays.
When necessary, CSF protein values should be checked with a different
chemical method to detect possible interference by drugs used to treat
CNS disorders (2). We found a good correlation between the
control and routine methods (Hitachi vs Vitros) with, however,
significantly lower values in the former. These findings are in
agreement with those of Lott and Warren (2), who reported
higher results (+26%) with the copper-binding methods (biuret and
Vitros) than with a turbidimetric method. In addition, these authors
found a good correlation between the Vitros method and the biuret
method, taking the latter as reference. Thus, the Hitachi values, which
were on average 21% lower than the Vitros values, must be corrected
when the present Vitros reference intervals are used to avoid
false-negative results.
In conclusion, the pediatric reference intervals reported here
complete previous studies in adults and children. Narrow and
age-related reference intervals are proposed for an analyzer that is
widely used in clinical chemistry and is based on dry chemistry. These
proposed intervals may improve the predictive performance of CSF
protein assays, particularly in the first 3 months of life.
|
Supplemental Information
|
|---|
Data from this study are available as a supplement from the
Clinical Chemistry Web site. The file can be accessed by a
link from the on-line Table
of Contents
(http://www.clinchem.org/content/vol46/issue3/).
|
Footnotes
|
|---|
Service de Biochimie-Hormonologie, Hôpital Pédiatrique Robert Debré, 48 Blvd Sérurier, 75019 Paris, France.
|
References
|
|---|
-
Silverman LM, Christenson RH, Grant GH. Amino acids and proteins. Tietz NW eds. Textbook of clinical chemistry 1986:607-615 WB Saunders Philadelphia. .
-
Lott JA, Warren P. Estimation of reference intervals for total protein in cerebrospinal fluid. Clin Chem 1989;35:1766-1770.
[Abstract/Free Full Text]
-
Statz A, Felgenhauer K. Development of the blood-CSF barrier. Develop Med Child Neurol 1983;25:152-161.
[ISI][Medline]
[Order article via Infotrieve]
-
Soldin JS, Brugnara C, KC Gunter, JM. Hicks, eds.
Pediatric reference ranges, 2nd ed. Washington: AACC Press, 1997:181pp..
-
Klein JO, Marcy SM. Bacterial sepsis and meningitis. Remington JS Klein JO eds. Infectious diseases of the fetus and newborn infant 4th ed. 1995:835-890 WB Saunders Philadelphia. .
-
Watanabe N, Kamei S, Ohkubo A, Yamanaka M, Ohsawa S, Makino K, Tokuda K. Urinary protein as measured with a pyrogallol red-molybdate complex, manually and in a Hitachi 726 automated analyzer. Clin Chem 1986;32:1551-1554.
[Abstract/Free Full Text]
-
Tibbling G, Link H, Ohman S. Principles of albumin and IgG analyses in neurological disorders. I. Establishment of reference values. Scand J Clin Lab Investig 1977;37:385-390.
[ISI][Medline]
[Order article via Infotrieve]
The following articles in journals at HighWire Press have cited this article:
|
|
|
|
 
P. J. Owen-Lynch, C. E. Draper, F. Mashayekhi, C. M. Bannister, and J. A. Miyan
Defective cell cycle control underlies abnormal cortical development in the hydrocephalic Texas rat
Brain,
March 1, 2003;
126(3):
623 - 631.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
Compiled by David E. Bruns, Editor (dbruns@aacc.org)
Clin. Chem.,
April 1, 2001;
47(4):
797 - 797.
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
