close

Вход

Забыли?

вход по аккаунту

?

The latex test revisited. Rheumatoid factor testing in 8287 rheumatic disease patients

код для вставкиСкачать
95 1
THE LATEX TEST REVISITED
Rheumatoid Factor Testing in 8,287 Rheumatic Disease Patients
FREDERICK WOLFE, MARY ANN CATHEY, and F. KAY ROBERTS
Rheumatoid factor (RF) testing by latex fixation
in 8,287 outpatients yielded a sensitivity of 81.6% and
78.0% at titers of 1:20 and 1:80, respectively, and a
specificity against noninflammatory rheumatic disorders
(NIRD) of 96.6% and 97.9% and against NIRD plus
inflammatory disorders of 95.2% and 96.8%, respectively. The predictive value of a positive test result at the
clinic prevalence rate for rheumatoid arthritis (RA)
(16.4%) was approximately 80%, and was 70% at 10%
prevalence and 10% at 1% prevalence. No associations
of RF with age or sex were found in non-RA patients.
RF titers increased minimally with age in RA patients
and were higher in men than in women. This study
suggests that latex testing is far more specific than has
been believed and that the titer is not spuriously increased with age.
Although rheumatoid factor (RF) is found in a
number of rheumatic disorders, such as systemic lupus
erythematosus (SLE) and Sjogren’s syndrome, as well
as in nonrheumatic conditions (1,2), its central role in
the clinic lies in its utility as an aid in the diagnosis of
From the Arthritis Center, the University of Kansas School
of Medicine, and the Arthritis, Rheumatism, and Aging Medical
Information System (ARAMIS).
Supported in part by grants from the National Institutes of
Arthritis, Diabetes, Digestive, and Kidney Diseases (AM-21393) to
the Arthritis, Rheumatism, and Aging Medical Information System
(ARAMIS).
Frederick Wolfe, MD: Clinical Professor of Medicine,
University of Kansas School of Medicine, and Director, Arthritis
Center: Mary Ann Cathey, MN: Research Associate, Arthritis
Center; F. Kay Roberts, BS: Outcome Assessor, Arthritis Center.
Address reprint requests to Frederick Wolfe, MD, Arthritis
Center, 1035 North Emporia, Wichita, KS 67214.
Submitted for publication August 21, 1990: accepted in
revised form February 28, 1991.
Arthritis and Rheumatism, Vol. 34, No. 8 (August 1991)
rheumatoid arthritis (RA). The proportion of RA patients positive for RF has ranged from 30% to more
than 90% in various studies (3) and depends to a major
extent on the definition of RA used in the clinic. But,
a contemporary view of RF prevalence in RA is
available from the American College of Rheumatology
(formerly, the American Rheumatism Association)
1987 study of the criteria for the classification of
rheumatoid arthritis (4). In that study, 80.4% of RA
patients were positive for RF, a proportion similar to
that found by other contemporary studies.
There is less agreement and less knowledge,
however, concerning the specificity or the predictive
value of RF testing, and almost all data have been
derived from population studies. A number of factors
have been identified that might influence specificity.
Specificity is linked to the presence or absence of
other RF-related disorders in the sample under study.
Thus, RF would have poor specificity in a clinic with a
large proportion of SLE patients. There is little knowledge of RF prevalence in the rheumatic disease clinic,
where its prevalence may be higher than in the community, even when patients with RF-related disorders
are excluded. Ehrlich, for example, noted an increased
prevalence of RF and RA in patients with erosive
osteoarthritis (OA) of the hands ( 5 ) . Such disorders
might have increased rheumatoid factor positivity by
virtue of the confusion with RA as well.
Data from population studies have suggested
that RF increases with age (&lo). Since patients with
disorders such as OA are generally in the older age
groups, RF prevalence might be expected to be increased among those with such chronic rheumatic
conditions. Thus, age, rheumatic disorders, a known
“false” positive rate in the community, and the pres-
952
WOLFE ET AL
ence of RF in certain nonrheumatic disorders have led
to the continuing reemphasis that RF positivity is
nonspecific and not diagnostic of RA.
Even though the view that RF positivity is
nonspecific is well known, RF testing in the community is common and may be inappropriately selected,
often being a part of laboratory “panels” and “profiles” that include disparate items such as C-reactive
protein, serum uric acid, anti-streptococcal antigens,
fluorescent antinuclear antibodies, and other immunologically based tests. When performed in settings in
which the prevalence of RA is low, the predictive
value of positive tests may be limited.
Although RF testing is extremely common, we
have noted that there is limited information concerning
RF in patients with rheumatic complaints who do not
have rheumatoid arthritis. In particular, RF specificity
and the predictive value of positive and negative tests
has not been studied. A knowledge of the predictive
value of these tests is important to the understanding
of the indications for their use. To further define the
role of RF testing in patients with rheumatic complaints, we have studied 8,287 patients seen consecutively in an outpatient rheumatic disease clinic. Of this
group, 1,362 patients had RA, 1,480 had other inflammatory disorders, and 5,445 had noninflammatory
rheumatic conditions.
PATIENTS AND METHODS
Patient population. Consecutive new patients (n =
8,287) attending an outpatient arthritis clinic in Wichita, KS,
between April 1976 and June 1990 had blood drawn for RF
assay. In addition, each underwent a comprehensive history
and physical examination using a standardized protocol,
with radiographs frequently being obtained. These data,
including diagnosis, were entered into the Arthritis Center
(Wichita, KS) computers as part of the ARAMIS (Arthritis,
Rheumatism, and Aging Medical Information system) data
bank (1 1).
Approximately 43% of the patients attending this
clinic were self referred. The clinic serves a catchment area
that includes the entire state of Kansas, with the exception
of the northeast part of the state. According to the 1990 US
census, 91.9% of the population of Kansas is white; 92.8% of
the patients attending this clinic are white. Of the adult
patients, 81.4% are high school graduates, and their mean
educational level is 12.4 (2.54 SD) years.
Diagnostic groups. Patients were considered to belong to 1 or more groups (subsets). Group 1 consisted of
those patients diagnosed as having RA according to either
the 1958 (12) or 1987 (4) criteria for the diagnosis or classification of RA. Group 2 consisted of patients with other
inflammatory rheumatic disorders (IRD): HLA-B27 spondylarthropathies, Reiter’s syndrome, psoriatic arthritis,
ankylosing spondylitis, monarthritis, gout, scleroderma, juvenile rheumatoid arthritis, pseudogout, undifferentiated
polyarthritis, polymyalgia rheumatica, cranial arteritis,
mixed connective tissue disease (MCTD), dermatomyositis,
SLE, other “collagen vascular” disorders, palindromic
rheumatism, Sjogren’s syndrome, and similar disorders.
Group 3 had noninflammatory rheumatic disorders (NIRD),
including osteoarthritis, low back pain syndromes, diffuse
idiopathic skeletal hyperostosis (DISH), cervical pain syndromes, fibromyalgia, and similar disorders. The 3 groups
were mutually exclusive and exhaustive, except that patients
with inflammatory disorders might also have additional
noninflammatory disease diagnoses. Group 4 consisted of
the combination of groups 2 and 3, or all patients who did not
have RA. Group 5 represents all patients. Generally, group 5
can be thought of as providing information about all rheumatic disease patients seen in an outpatient rheumatology
clinic; group 3 or group 4 as providing an approximate model
of what can be seen in community studies or general medical
clinics (omitting RA). Patients in these categories almost
always satisfied American College of Rheumatology diagnostic or classification criteria if criteria for the disorders
existed at the time of diagnosis. For other conditions, the
diagnosis was based on diagnostic recommendations found
in standard textbooks of rheumatology over the time period
of the study. In each instance, the diagnosis was based on
history, physical examination, laboratory, and radiographic
data; in some patients, it was based on findings at the
followup examination.
Rheumatoid factor was tested using a commercial
latex test kit (Rapi Tex R F Test; Behring Diagnostics,
Somerville, NJ). Tests were considered positive at a dilution
of 1:20 or above. Testing was performed on the day the
sample was drawn. A small number of patients had a second
test performed. The reasons for performing a second test
included 1) participation in an investigational drug study that
required a new sample or 2) clinical reasons relating to
uncertainty of diagnosis of the subject’s rheumatic disorder.
On initial testing 1,445, or 17.4%, of the 8,287 patients were
positive for R F at a dilution of 1:20 or above. An additional
72 patients were positive at repeat testing. Except as described in the text, this report deals with patients who were
R F positive at some time during their assessment in the
arthritis clinic.
Statistical analysis. Data were analyzed using version
6.04 of PC SAS (13) and the 1990 revision of PC BMDP (14).
Chi-square testing was employed for unordered categorical
data, and ridit analysis was used for ordered data. Ridit
analysis is a categorical data analysis technique that accounts for the inherent ordering of categories and is the
analytic method of choice for such data (15). In our analyses,
we used this method to compare males with females regarding the ordered categories of RF titer. Regression techniques
were used to study the effect of age on RF positivity and
titer. For the analysis of RF positivity at a single level (e.g.,
positive or negative at 1:80) binary logistic regression was
used; where the effect of age on the ordered categories (RF
titers) was studied, ordinal logistic regression was the analytic method (16). All tests were 2-tailed; t-tests were used to
analyze the continuous demographic variables. P values
953
RHEUMATOID FACTOR TESTING
Table 1.
Demographic characteristics of 8,287 rheumatic disease clinic patients*
Age, years
(SD)
Sex,
%
Caucasian,
%
Black,
%
HS graduate,
Category (n)
All RA patients (1,362)
Males
Females
All IRD patients (1,480)
Males
Females
All NIRD patients (5,445)
Males
Females
55.3 (14.29)
55.8 (13.02)
55.1 (14.78)
45.6 (19.10)
43.4 (18.44)
46.9 (19.36)
53.3 (16.42)
52.6 (15.64)
52.6 (16.71)
29.1
70.9
36.3
63.7
28.1
71.9
94.2
95.5
93.7
91.5
92.4
91.0
92.5
92.2
92.6
3.0
1.5
3.6
4.1
3.2
4.7
4.1
4.3
4.1
78.1
76.4
78.8
81.2
79.3
82.3
80.1
78.6
81.3
%
~
* HS = high school; RA = rheumatoid arthritis; IRD = inflammatory rheumatic disorders; NIRD
noninflammatory rheumatic disorders.
were not corrected for multiple comparisons. Statistical
significance was declared at the 0.05 level.
Calculation of posttest probabilities (for Figures 3-5)
relied on the following relationships. Positive likelihood ratio
= true positive rate to false positive rate. Negative likelihood ratio = false negative rate to true negative rate. Pretest
odds = (Pi1 - P). Posttest odds = likelihood ratio x pretest
odds. Posttest probability = (posttest odddl + posttest
odds).
RESULTS
Tables 1 through 3 display the demographic,
diagnostic, and rheumatoid factor characteristics of
the study subjects. More than two-thirds of the RA
and NIRD patients were women, and 63.7% of those
with IRD were women (Table 1). Overall, 16.4% had
RA, 17.9% other inflammatory rheumatic conditions,
and 65.7% noninflammatory disorders. Also, 16.6% of
Table 2. Rheumatoid factor positivity at dilutions of 1:20 to
1:5,120 for 1,362 male and female clinic patients with rheumatoid
arthritis*
Variable,
rheumatoid
factor titer
21:20
5 I :40
?1:80
rl:160
21:320
?1:640
2 1 : 1,280
2 1:2,560
2 15,120
Combined
group
(n = 1,362)
Females
(n = 966,
70.9%)
Males
(n = 396,
29.1%)
81.6
80.2
78.0
71.9
62.8
49.2
37.5
23.7
5.3
78.3
76.4
73.9
67.8
58.4
44.1
33.3
20.5
4.8
89.9
89.4
87.9
81.8
73.5
61.6
47.7
31.6
6.6
* Values are the percentage positive for rheumatoid factor. Females
and males differ in the proportion who are seropositive (titer 2 1:20)
(chi-square 28.6, P < 0.001). The distribution of positive latex
fixation test results at 1:20 through 15,120 differs by sex at the 0.001
level (ridit analysis).
=
women and 16.1% of men attending the clinic had
rheumatoid arthritis.
Effect of sex on rheumatoid factor positivity.
Seropositivity, defined as a positive R F titer of 1:20 or
above, was more common in men with RA than in
women with the disorder ( P < 0.001) (Table 2). At all
levels of RF positivity, the proportion of men exceeded that of women. Since this finding is related to
the increased proportion of seronegative women, we
next studied the distribution of positive RF titers by
ridit analysis. Among the seropositive of both sexes,
men had consistently higher RF titers (chi-square [ridit
scores] 14.9, P < 0.001).
Among those with IRD, 11.4% of women and
7.4% of men were RF positive at a titer of 1:20 ( P =
0.03). At 1530, the percentages were 8.6 and 5.0,
respectively. The IRD group is heterogeneous (as
noted in Figure I ) , and seropositivity may be explained by the relative proportion of certain disorders
Table 3. Relationship of age to rheumatoid factor positivity at a
dilution of 1:80*
Age group
(years)
520
21-30
31-40
41-50
51-60
61-70
71-80
>80
RA patients
(n = 1,362)
IRD patients,
(n = 1,480)
NIRD patients
(n = 5,445)
n
% positive
n
% positive
n
% positive
4
78
141
225
353
346
183
32
50.0
64.1
79.4
82.7
76.2
78.6
79.8
78.1
144
210
248
228
245
246
133
26
4.2
6.2
4.8
11.8
10.2
6.1
6.8
3.8
149
360
719
917
1,310
1,120
650
220
1.3
3.1
1.8
1.9
2.1
2.0
2.3
2.7
* The effect of age on rheumatoid factor positivity at 1:80 was not
significant in logistic regression models (P > 0.1) for any of the 3
study groups. RA = rheumatoid arthritis; IRD = other inflammatory
disorders; NIRD = noninflammatory disorders.
954
WOLFE ET AL
PSORUTlC MHRITIS
NECK PUN SYNDROMES
MICROSING
svmonms
U L PERIPHERAL J O l M OA
POLWYUGIA RHEUMATICA
YOS1T1S
LUPUS ERITHENATOSUS
COLUGtN-VhSCUUR
PUINDROUK: RHNUATISU
SJOCREWS SiNDROME
0
20
40
60
80
100
PERCENT POSITIVE
Figure 1. Rheumatoid factor positivity at 1:80 in 8,287 patients with rheumatic disorders.
Noninflammatory disorders include neck pain syndromes (n = 420), low back pain (n = 1,226),
osteoarthritis (OA) of the hips (n = 167), OA of the knees (n = 894). OA of the hands (n = 739,
all peripheral joint OA combined (n = 1,590), diffuse idiopathic skeletal hyperostosis (D.I.S.H.;
n = 20), and fibromyalgia (n = 726). Inflammatory disorders include B-27 spondyloarthropathies (n = 22), Reiter’s syndrome (n = 51), psoriatic arthritis (n = 118), ankylosing spondylitis
(n = 80), monoarthritis (n = 141). gout (n = 150), scleroderma (n = 44), juvenile rheumatoid
arthritis (J.R.A.; n = 92), pseudogout (n = 178), undifferentiated polyarthritis (n = 787),
polymyalgia rheumatica (n = 32), cranial arteritis (n = 23), mixed connective tissue disease
(M.C.T.D.; n = 17), dermatomyositis (n = 32), systemic lupus erythematosus (n = 124), other
“collagen-vascular” disorders (n = 38), palindromic rheumatism (n = 17), Sjogren’s syndrome
(n = 26), and rheumatoid arthritis (n = 1,362). Patients may have more than one diagnosis, but
patients classified as having noninflammatory disorders may not have a coexistent inflammatory
disorder.
in the study population. In addition, the multiple
testing performed in this study suggests the possibility
that the significant P value might be the result of
chance. Unlike the RA group, however, ridit analysis
in those with IRD showed no association between RF
titer and sex (chi-square [ridit scores] 1.6, P = 0.28) in
those who were seropositive.
In the 5,445 patients with NIRD, no statistically
significant association between sex and RF positivity
or titer (chi-square [ridit scores] 1.2, P = 0.27) was
noted. Positivity at 1:20 was noted in 3.7% of women
and 2.9% of men. At 1:80,2.2% of women and 1.8% of
men were R F positive.
Effect of age on rheumatoid factor positivity.
Table 3 describes the effect of age on RF positivity at
a titer of 1:80 or greater. No statistically significant
association with age was found for RF positivity
(defined either as positive at 1:20 or positive at 1230)
for either the NIRD group or the IRD group in binary
and/or ordinal logistic regression models. For patients
with RA, there was no significant relationship between
age and RF positivity (at 1:20 or 1:80 titers), but the
RF titer (from 1:20 through 15,120) increased with
age, in ordinal logistic regression analysis (Wald chisquare 14.4, P < 0.0001). This was also true when the
analysis was performed including the seronegative
patients (titers “negative” through 15,120) (Wald
chi-square 13.1, P < 0.001). The relationships, however, were not strong. Kendall’s tau for the association of predicted probabilities and observed responses
in the first logistic model was only 0.06, and the
directly measured association between age and RF
titer using the Spearman correlation coefficient was
0.11 overall. These data indicate that a positive RF test
955
RHEUMATOID FACTOR TESTING
loo
I
Rheumatoid Arthritis (RA) n = 1,362
Inflammatory Disorders (IRD) n = 1.480
Non-inflommotory
Disorders (NIRD) n = 5.445
.
I"
1:zo
cI
2
2
0
r
;
;
m
,
;
c
,
0
V RA
R""5
vs
" R NIRD
D
+,
IRD
0 RA vs IRD
60
I__
0
0
0
2
4
6
8
10
When RF is used as an aid to the diagnosis of
RA in population studies, community surveillance,
general practice clinics, and rheumatology clinics, the
utility of the test depends upon the characteristics of
the control population. NIRD, at one extreme, represents the easiest separation of RA patients from other
rheumatic disease patients. Since the false-positive
rate at 1:20 in NIRD patients is 3.4% or 2.1% at 1:SO (a
rate toward the lower limit of published population
data), NIRD patients may be considered a possible
surrogate for controls in population surveys. A more
conservative estimate would use NIRD plus IRD. IRD
is almost certainly overrepresented in a rheumatic
disease clinic compared with the community, such that
the combination of NIRD and IRD might be considered to form the lower limit. The probability is that the
correct curve for the community is somewhere between the 2 leftmost curves in Figure 2. The IRD curve
alone is representative of the diagnostic utility of RF
as a screening tool in a group of patients with inflammatory rheumatic disorders.
The data shown in the ROC curves allow the
calculation of positive and negative likelihood ratios.
Likelihood ratios measure the ability of a test to
discriminate between patients with and without disease, in this instance, between those with and without
RA. A positive likelihood ratio is the ratio of those
with RA who are RF positive at a titer of 1:SO
(true-positive rate) to those who are RF positive but do
not have RA (false-positive rate). A negative likelihood ratio expresses a similar relationship, the ratio of
the false-negative rate to the true-negative rate. The
data in Figure 3 indicate positive likelihood ratios of
37.6, 24.4, and 10.6 and negative likelihood ratios of
0.22, 0.23, and 0.24 for the NIRD, NIRD plus IRD,
and IRD, respectively. These ratios allow the construction of curves to describe prior (pretest) and
posterior (posttest) probabilities for the diagnosis of
RA in the 3 groups, as demonstrated in Figures 3 and
4. The prior or pretest probability, as shown on the
horizontal axis of Figures 3 and 4, represents the
prevalence of RA in the clinic, but can also be considered to be the clinicians' probability estimate after
examining the patient, for example. The posterior or
posttest probability is the probability of having RA
given the pretest probability and given that the RF test
is positive (or negative) at the 1:SO dilution. Data on
calculation of posttest probabilities from likelihood
ratios and pretest probabilities are described in Patients and Methods. As shown in Figure 4, the vertical
line at 16.4 indicates the clinic prevalence of RA.
12
FALSE POSITIVE RATE (%)
Figure 2. Receiver operating characteristic (ROC) curves for RA
versus NIRD, RA versus IRD, and RA versus the combination of
noninflammatory plus inflammatory disorders (NIRD plus IRD) for
latex fixation test positivity at a dilution of 1:80.
result in RA patients is not associated with age, but
that there is a slight, but statistically significant, association of increasing R F titers with age in RA patients.
No such associations were noted in patients without RA.
Rheumatoid factor positivity and the diagnosis of
rheumatoid arthritis. Figure 2 plots the receiver operating characteristic (ROC) curves indicating the truepositive and false-positive rates for the 3 diagnostic
subsets in which RA can be found: RA versus NIRD,
RA versus IRD, and RA versus NIRD plus IRD. A
ROC curve is a graphic representation of the truepositive rate (sensitivity or "positive in disease") and
the false-positive rate ("positive in health") at various
levels of test positivity (17). Test specificity in Figure
2 can be expressed as 100% minus the false-positive
rate. Although the best cut-off for RF positivity is
somewhat subjective (17), depending on whether sensitivity or specificity is more important, there is little
loss of sensitivity (3.6%) coupled with gains in specificity as the cut-off is increased from a titer of 1:20 to
a titer of 1:SO. Sensitivity, however, decreases sharply
as titer cut-off changes from 1:SO to 1:160 (6.1%) as
indicated in Table 2 and Figure 2. The "best" positive
titer for RF at a clinical level is 1:80 when specificity is
most important, and 1:20 when sensitivity is the important determinant. But, all titers between 1:20 and
1:80 are roughly equivalent in terms of the accuracy of
classification ([sensitivity + specificity]/2). Accuracy
decreases rapidly at titers of 1:160 and above, however.
956
WOLFE ET AL
0.20
0.60
E
0.00
1.oo
g
0.90
0.90
0.80
0.80
0.70
0.70
0.40
0.80
1.00
1 .oo
L
I-
a
0
0
2
1:
3
I
0.60
&
0.50
W
- -RA vs
. . . RA vs
a
2
m
3
0
a
a
I-
-RA
NlRO
NlRO
vs IRD
0.60
+
IRD
0.50
0.40
0.40
0.30
0.30
0.20
0.20
0.10
0.10
v,
W
c
I-
v,
g
0.00
0 0
0.20
0.40
0.60
0.80
0.00
1.00
PRE TEST PROBABlLlM OF RHEUMATOID ARTHRITIS
Figure 3. Pretest probability versus posttest probability for rheumatoid arthritis (RA; n = 1,362) versus noninflammatory disorders
(NIRD; n = 5,443, RA versus inflammatory disorders (IRD; n =
1,480), and RA versus the combination of NIRD plus IRD (n =
6,925) for latex fixation test positivity at a dilution of 130. The
pretest probability represents the probability of having RA in each
of the 3 group comparisons prior to rheumatoid factor testing; the
posttest probability represents the probability after testing. The 3
curves above represent a positive test, while the curves below
represent negative tests.
and for women, 37.5, 21.6, and 8.6. Negative likelihood ratios for men were 0.29, 0.12, and 0.13, and for
women 0.22, 0.27, and 0.28, respectively, for RA
versus NIRD, NIRD plus IRD, and IRD.
“False” positive rheumatoid factor tests and the
specificity of the latex test. Patients with noninflammatory disorders may be considered “controls,” approximating “normals” for the purpose of R F testing. We
found the latex test to be highly specific in this group,
with only 2.1% of patients having false-positive test
results at the 1:80 dilution and 3.4% at the 1:20
dilution. Although R F “false” positivity has been
associated with a number of other conditions, including viral infections, parasitic infections, neoplasms,
and other hyperglobulinemic states (1,2), fewer than 5
patients had R F positivity that could be thus explained.
Rheumatoid factor positivity is well known in
other rheumatic conditions, though the percent positive has not been determined in a large, prospective
series of unselected, consecutive clinic patients. Figure 1 indicates the prevalence of R F positivity in
vl
t
a
z 0.90
c
/
_. . .-, .-. .-.
0.90
- 0.80
When compared against all other patients (dotted line),
the probability of a patient with a positive latex test
result having RA is approximately 80%. In population
surveys where the prevalence of RA is s l % , a positive R F test result may be expected to be associated
with RA in 20% or fewer instances. Similarly, when
used as a screening tool in general practice, where the
prevalence of RA is low, most patients with positive
R F test results will not have RA.
Both true-positive rates and false-positive rates
(Table 1) differed for men and women. The effects of
these differences were to make positive likelihood
ratios larger and negative likelihood ratios smaller for
men. As shown in Figure 5 for RA versus NIRD plus
IRD, a negative R F test result at a titer of 1230 makes
RA far less likely in a man than in a woman. Sex
differences, however, are relatively unimportant when
the test result is positive. We calculated likelihood
ratios for all comparison groups. For RA versus
NIRD, NIRD plus IRD, and IRD, positive likelihood
ratios for men were 46.1, 33.0, and 17.8, respectively,
- 0.70
-
-RA vs NlRO
. . . RA vs NIRD
--RA
vs IRO
- 0.60
+
IRD
- 0.50
- 0.40
- 0.30
- 0.20
-- 0.10
0.00
0.05
0.10
0.15
0.20
0.25
0.00
0.30
PRE TEST PROBABILITY OF RHEUMATOID ARTHRITIS: PRNALENCE
Figure 4. Detail of Figure 3 for pretest (prevalence) probability
versus posttest probability for RA versus NIRD, RA versus IRD,
and RA versus NIRD plus IRD for latex fixation test positivity at a
dilution of 1:80. The pretest probability represents the probability of
having RA in each of the 3 group comparisons prior to rheumatoid
factor testing; the posttest probability represents the probability
after testing. The 3 curves above represent a positive test, while the
curves below represent negative tests. See Figure 2 for definitions
and numbers of patients per group.
957
RHEUMATOID FACTOR TESTING
0.00
vr
1 .oo
0.20
0.40
0.60
0.80
1.00
1 .oo
c
g
t-
0.90
0.90
a
0.80
0.80
0.70
0.70
0.60
0.60
0.50
0.50
0.40
0.40
2
0
0.30
0.30
a
0.20
0.20
0.10
0.10
K
D
I
3
w
I
lY
>-
=’
m
K
c
v,
w
tt-
v,
0
a
nnn
- - -000
0.00
020
0.40
0 60
0.80
100
PRE TEST PROBABILITY OF RHEUMATOID ARTHRITIS
Figure 5. Pretest probability versus posttest probability for RA
versus NIRD plus IRD for latex fixation test positivity at a dilution
of 1:80. The pretest probability represents the probability of having
RA prior to rheumatoid factor testing; the posttest probability
represents the probability after testing. The curves above represent
a positive test, while the curves below represent negative tests. See
Figure 2 for definitions and numbers of patients per group.
various rheumatic conditions. Almost 15-20% of those
with SLE, dermatornyositis, MCTD, and similar disorders were positive for RF at a titer of 1:80. More
patients than would have been expected with gout,
DISH, and pseudogout were positive for RF. When
patients with SLE, dermatomyositis, MCTD, palindromic rheumatism, and Sjogrens’s syndrome (where
RF is expected to occur more frequently) are excluded, unexplained latex positivity occurs uncornrnonly , confirming the high specificity and predictive
value of R F testing.
DISCUSSION
Erik Waaler first identified rheumatoid factor in
1937 (18). But its potential clinical utility was identified
in the report of Rose et a1 a decade later (19). In the 25
years that followed, clinical tests for rheumatoid factor
were developed that used modifications of the original
sheep cell test. Human and rabbit immunoglobulin
were coated to a variety of substances including
human cells, latex particles of various sizes, and
bentonite (20). Three major test systems emerged, the
SCAT (sensitized sheep cell agglutination test), the
bentonite flocculation test (BFT), and the latex fixation test (LFT). The SCAT was shown to be the most
specific in the diagnosis of RA but the least sensitive,
and the LFT the most sensitive but least specific
(6,20-23). The ease with which the LFT could be
performed has led to its overwhelming adoption at the
clinical level, and most reports in the rheumatology
literature refer to this test. The SCAT and BFT are
frequently used in population studies because of their
apparent high specificity.
Several factors have led to concern about specificity, particularly of the LFT. Rheumatoid factor has
been found in a number of other conditions, including
rheumatic disorders, but also viral infections, parasitic
infections, such chronic inflammatory disorders as
tuberculosis, syphilis, and subacute bacterial endocarditis, malignancies, and hyperglobulinemic states
(2). In addition, RF positivity has been described as
increasing in prevalence with age. Waller et a1 studied
5,461 normal subjects in 1964 (6). Of the 97 subjects
over age 60,27.8% were positive by the LFT (although
some of these individuals may have had RA); 4.3% of
subjects between the ages of 30 and 59, and 3.1%
younger than 30 were positive for RF by the LFT.
Using the SCAT, the prevalence of positivity was
2.1% of the older age group compared with <0.25% in
younger subjects (6). Aho et a1 studied 7,200 subjects
in a population survey in Finland using the WaalerRose test (SCAT) (21). The false-positive rates in the
fourth through the ninth decades of life were 1.2%,
1.7%, 1.1%, 1.9%, 3.4%, and 3.1%, respectively.
Using the BFT, a false-positive rate of 5.6% in 906
Alaskan Eskimos was noted, 3.2% of those younger
than 55 and 13.8% of those above that age (8). Mikkelsen et a1 examined 6,590 R F tests in a population
survey of Tecumseh, Michigan in 1959-1960 (9). Using
the LFT at a dilution of 1:20 they noted that 3.35% of
all subjects had positive results. The rates increased at
each decade of age for both sexes, from 2.0% at ages
20-29 to 11.4% at ages >70. Although recent studies
have suggested that false-positive RF tests may be
followed by the clinical development of RA in some
patients (22,24), the thrust of the above studies is to
suggest caution in the interpretation of RF testing
when the LFT is used, and particularly so in the
elderly.
Almost all RF testing, however, is done in the
958
clinic for the purposes of diagnosis. No studies have
examined the issue of sensitivity and specificity in
clinical populations. Given the data from population
studies, it might be assumed that the use of the LFT in
an elderly, clinically ill population would lead to
unacceptable levels of specificity, since age itself or
other medical diseases might be associated with RF
positivity, or patients with rheumatoid arthritis might
be incorrectly diagnosed as having a different rheumatic disorder (e.g., OA of the hands or knees). The
results of our study of 8,287 consecutive rheumatic
disease patients, 2,053 of whom were age 65 or older,
show that this is not so. First, we found no relationship
between rheumatoid factor positivity or rheumatoid
factor titer and age in patients without rheumatoid
arthritis (see Table 3 and Results section). We noted
such a relationship only in patients with RA. These
data have two important clinical implications. First,
they suggest that the inferences that can be drawn
regarding a positive test for R F in the clinic are true
across all ages. Second, they underscore differences
between clinic data and those obtained in population
surveys.
In considering the “false” positive rates for
LFT, it is clearly important to consider the other
diseases in the sample that might be associated with
R F positivity. As shown in Figure 1 , slight increases in
the R F positivity rates above what was noted in
noninflammatory disorders were seen in most inflammatory disorders, and increases above 10% were
noted in disorders such as MCTD, dermatomyositis,
SLE, and others that are known to be associated with
RF. Thus, the R F positivity rate in the IRD group is
7.3% at a dilution of 1530. A truer estimate of the
false-positive rate with the latex test methodology can
be obtained by examining the data of patients without
inflammatory disorders, since for most patients with
inflammatory disorders, a positive RF test result is not
“false.” At dilutions of 1:80, only 2.1% of NIRD
patient tests were “false” positive, and, as indicated
above, no effect of age was seen. We attempted to
estimate the false-positive rate (for a diagnosis of RA)
of the LFT among rheumatic disease patients in general by combining the IRD and NIRD patients. Among
this combined group, 3.2% were positive for RF at
1230. Since inflammatory diseases are overrepresented
in rheumatic disease clinics compared with general
populations and with patients with musculoskeletal
pain seen in general medical clinics, the ROC curve
WOLFE ET AL
(Figure 2) for community and general practice patients
may be closer to the NIRD curve than the NIRD plus
IRD curve.
It has been noted almost universally that high
titers of rheumatoid factor rarely represent falsepositive results. Our data support this contention
(Figure 2). At titers of 1:1,280 or above, the falsepositive rate for the NIRD plus IRD group was 0.5%
and for IRD alone 1.3%. Figure 2 suggests that the
“best” positive titer using the latex method may be
1:80, since there is little loss of sensitivity but a
substantial increase in specificity. But 1:20 also works
well, trading a small loss in specificity for a small
increase in sensitivity. Increases in test specificity
with increasing titers, however, result in lower sensitivity, being less than 40% at a dilution of 1:1,280. The
American College of Rheumatology criteria for the
classification of rheumatoid arthritis, published in 1988
(4), indicate that a test method be used that produces
of the normal populafalse-positive findings in -4%
tion. Our data suggest that any dilution of 1:20 or
above will fulfill that requirement, but that the latex
test is in fact more specific than previous population
studies have indicated. For clinical diagnosis of RA,
rather than for classification of the disorder, a titer of
1:80 may be more useful, particularly in equivocal
cases. Positivity at high titers increases the certainty
of the diagnosis of RA and, although not studied in the
American College of Rheumatology RA classification
analysis (4), should increase classification certainty
as well.
The determination of sensitivity and specificity
of RF testing in our population and its subsets allows
the construction of pretest-posttest probability curves
and, therefore, some estimation of how these tests
might perform in community studies and clinics. Assuming a prevalence of 1% for RA in the community,
Figure 4 suggests that the probability of RA, given a
positive LFT, is 10%. Since joint pain is common in
the community (9), as is morning stiffness and reported
joint swelling (9), reliance on R F testing when using
nonspecific criteria for RA diagnosis, such as the
“probable RA” of the 1958 American Rheumatism
Association criteria (12), might be one of several
mechanisms that led to the acknowledged earlier overestimation of RA prevalence.
In rheumatic disease clinics where the prevalence of RA is high (16.4% in our clinic), a positive RF
test result is highly predictive of RA ( S O % ) , and a
RHEUMATOID FACTOR TESTING
negative test result suggests the diagnosis is unlikely
(<5%). If all RF-associated disorders are considered
as “positive” testing, then the utility of this test in
rheumatic disease clinics increases. Although RF testing could be used for screening purposes in rheumatic
disease clinics, we are not suggesting that this be done.
It is also important to note that in certain rheumatic
disease clinics, patients with systemic connective tissue disorders may be seen in equal or greater proportion to those with RA. In such settings, the predictive
value of RF testing for RA will be low.
Laboratory screening is, however, frequently
used in the community, where diagnostic acumen may
not be high. Every rheumatologist has seen patients
with back pain or osteoarthritis or fibromyalgia and a
positive RF test result labeled as having RA and
treated with disease-modifying antirheumatic drugs.
Commercial laboratories promote panels that include
disparate items such as C-reactive protein, antistreptococcal tests, rheumatoid factor, and uric acid.
Therefore, physicians may have results of RF testing
when the information is not germane to the diagnosis.
Our data suggest, however, that if the physician’s
estimate of the probability of RA is, for example, as
high as lo%, then RF testing will be a substantial aid to
diagnosis, a positive result linking to a posttest probability above 70% and a negative result a probability of
<3%. But if the tests are ordered when there is little
likelihood of RA (e.g., back pain or fibromyalgia) and
where the pretest probability is at the population level
(I%), their predictive value will be very low.
ACKNOWLEDGMENTS
We thank Drs. Michael Luggen, Theodore Pincus,
and Donna Hawley for their helpful assistance with the
manuscript. We dedicate this article to Dr. Donald M.
Mitchell of Saskatoon, Saskatchewan, Canada, who also
made these observations but was prevented from writing
about them by his premature death.
REFERENCES
1. Bartfield H: Distribution of rheumatoid factor activity in
nonrheumatoid states. Ann N Y Acad Sci 168:30-40,
1969
2. Carson DA: Rheumatoid factor, Textbook of Rheumatology. Second edition. Edited by WN Kelley, ED
Harris Jr, S Ruddy, CB Sledge. Philadelphia, WB Saunders, 1989
959
3. Wolfe F: Rheumatoid arthritis, Prognosis in the Rheumatic Diseases. Edited by BN Dordrecht. Amsterdam,
Kluwer 1991
4. Arnett FC, Edworthy SM, Bloch DA, McShane DJ,
Fries JF, Cooper NS, Healey LA, Kaplan SR, Liang
MH, Luthra HS, Medsger TA Jr, Mitchell DM, Neustadt DH, Pinals RS, Schaller JG, Sharp JT, Wilder RL,
Hunder GG: The American Rheumatism Association
1987 revised criteria for the classification of rheumatoid
arthritis. Arthritis Rheum 31:315-324, 1988
5. E,hrlich GE: Inflammatory osteoarthritis. 11. The superimposition of rheumatoid arthritis. J Chronic Dis 25:635643, 1972
6. Waller M, Toone EC, Vaughan E: Study of rheumatoid
factors in a normal population. Arthritis Rheum 7513520, 1964
7. Ferri S, Favaro L, Amoresano C: Environmental and
social factors in rheumatoid factor epidemiology. Scand
J Rheumatol Suppl 17:219-222, 1988
8. Beasley RP, Retailliau H, Healey LA: Prevalence of
rheumatoid arthritis in Alaskan Eskimos. Arthritis
Rheum 16:737-742, 1973
9. Mikkelsen WM, Dodge HJ, Duff IF, Kato H: Estimates
of the prevalence of rheumatic diseases in the population
of Tecumseh, Michigan 1959-60. J Chronic Dis 20:351369, 1967
10. Cathcart ES, O’Sullivan JB: A longitudinal study of
rheumatoid factors in a New England town. Ann N Y
Acad Sci 168:41-51, 1969
11. Fries JF: The chronic disease data bank: first principles
to future directions. J Med Philos 9:161-180, 1984
12. Ropes MW, Bennett GA, Cobb S, Jacox R, Jessar RA:
1958 revision of diagnostic criteria for rheumatoid arthritis. Arthritis Rheum 2:16-20, 1959
13. SAS Institute: SAS Technical Report P-200 SAS/Stat
Software: Calk and Logistic Procedures. Release 6.04.
Cary, NC, SAS Institute, 1990
14. Dixon WJ, Brown MB, Engelman L , Jennrich RI:
BMDP Statistical Software. Berkeley, University of
California Press, 1990
15. Fleiss JL: The Design and Analysis of Clinical Experiments. New York, John Wiley & Son, 1986
16. Hosmer DW, Lemeshow S: Polytomous logistic regression, Applied Logistic Regression. New York, John
Wiley & Son, 1989
17. Sackett DL, Haynes RB, Tugwell P: Clinical Epidemiology: A Basic Science for Clinical Medicine. Boston,
Little, Brown & Co., 1985
18. Waaler E: On the occurrence of a factor in human serum
activating specific agglutination of sheep blood corpuscles. Acta Pathol Microbiol Immunol Scand [B]: 10: 172178, 1940
19. Rose HM, Ragan C, Pearce E, Lipman MO: Differential
WOLFE ET AL
agglutination of normal and sensitized sheep erythrocytes by sera of patients with rheumatoid arthritis. Proc
SOCExp Biol Med 68:1-6, 1948
20. Egeland T , Munthe E: Rheumatoid factors. Clin Rheum
Dis 9:135-160, 1983
21. Aho K, Tiinamaija T, Heliovaara M, Palosuo T: Rheumatoid factors and rheumatoid arthritis. Scand J Rheumatol Suppl 74:4144, 1988
22. Aho K, Palosuo T, Raunio V, Puska P, Aromaa A,
Salonen JT: When does rheumatoid disease start? Arthritis Rheum 28:485-489, 1985
23. Valkenburg HA, Ball J, Burch TA, Bennett PH,
Lawrence JS: Rheumatoid factors in a rural population.
Ann Rheum Dis 26:497-508, 1966
24. Del Puente A, Knowler WC, Pettitt DJ, Bennett PH:
The incidence of rheumatoid arthritis is predicted by
rheumatoid factor titer in a longitudinal population
study. Arthritis Rheum 31: 1239-1244, 1988
Документ
Категория
Без категории
Просмотров
7
Размер файла
878 Кб
Теги
test, testing, factors, patients, rheumatic, revisited, latex, disease, 8287, rheumatoid
1/--страниц
Пожаловаться на содержимое документа