close

Вход

Забыли?

вход по аккаунту

?

The diagnosis of polyarteritis nodosa.

код для вставкиСкачать
1117
THE DIAGNOSIS OF POLYARTERITIS NODOSA
I. A Literature-Based Decision Analysis Approach
DANIEL A. ALBERT, DAVID RIMON, and MARC D. SILVERSTEIN
We investigated diagnostic testing in polyarteritis
nodosa (PAN) by calculating, from published data, the
sensitivity and specificity of visceral angiography and
muscle, nerve, testicle, kidney, and liver biopsy. Test
sequence strategies were constructed by Bayesian inference using a computer program written for this purpose. Test sequences were compared with an aggressive
strategy consisting of repeated tests until there was a
positive finding or until the available tests were exhausted, and a conservative strategy consisting of 1
biopsy procedure plus angiography. The Bayesian analysis agreed most closely with the conservative approach
for most prior probabilities (degree of suspicion) that a
patient had PAN. The aggressive strategy had an overall
sensitivity of 90% and specificity of 91%, whereas the
conservative strategy was 85% sensitive and 96% specific. Furthermore, the aggressive strategy was more
costly ($2,986 versus $1,961) and had a higher rate of
morbidity (3.8 versus 2.7 days of hospitalization per
patient evaluated) than did the conservative strategy.
The mortality rates of both strategies were equivalent
(-0.05 deaths per hundred patients evaluated). The
per-case cost of diagnosis increased as prevalence decreased, and a t 10% prevalence, the aggressive strategy
cost more than $17,000 per case diagnosed. Sensitivity
From the Department of Medicine. University of Chicago
Medical Center, Chicago, Illinois.
Dr. Albert is an Arthritis Foundation Arthritis Investigator.
Dr. Silverstein is a Henry J. Kaiser Family Foundation Faculty
Scholar in General Internal Medicine.
Daniel A. Albert, MD; David Rimon. MD: Marc D. Silverstein, MD.
Address reprint requests to Daniel A. Albert, MD, University of Chicago Medical Center, Rheumatology Section, Hospital
Box 74, 5841 South Maryland Avenue, Chicago. IL 60637.
Submitted for publication May 15, 1987; accepted in revised
form March I . 1988.
Arthritis and Rheumatism, Vol. 31, No. 9 (September 1988)
analysis revealed that the strategies were moderately
affected by the test characteristics, within reasonable
assumptions, but that the differences in conservative
and aggressive approaches remained. Thus, our analysis
based on available data and the assumption of test
independence suggests that the preferred diagnostic
evaluation of patients with symptoms suggestive of PAN
consists, in most cases, of a single biopsy procedure,
with angiographic evaluation if necessary.
Polyarteritis nodosa (PAN) is a rare disorder
with potentially lethal consequences that are favorably
affected by therapy (1). It differs pathologically and
clinically from other vasculitides by its propensity for
causing necrotizing occlusions and aneurysms in
medium-sized arteries. The diagnosis of PAN and its
distinction from other vasculitides is clinically important because of its unique natural history and particular organ system involvement, although it is like other
forms of vasculitis in its responsiveness to treatment
with steroids and cytotoxic agents (I).
The diagnostic approach to patients with symptoms suggestive of PAN is a subject of controversy,
and a variety of diagnostic procedures have been
recommended by different observers. Angiography
and biopsy of muscle, nerve, testicle, kidney, and liver
have all been suggested as diagnostic procedures of
choice; several different strategies consisting of multiple tests have also been advocated ( I ) . A prospective
trial of all tests in patients with suspected PAN is not
likely, because the disease is rare and because different diagnostic and therapeutic approaches are followed at different institutions.
Thus, we undertook this study to quantitate the
sensitivity and specificity of the various diagnostic
ALBERTETAL
1118
Table 1. Characteristics of diagnostic tests for polyarteritis nodosa*
Test
Angiography
Muscle biopsy
Symptomatic
Blindt
Nerve biopsy
Symptomatic
Blindt
Testicular biopsy
Renal needle biopsy
Liver needle biopsy
No. of patients
studied
Sensitivity
(%)
Specificity
(96)
Cost
Morbidity
(days)
Mortality
($)
(7%)
Ref.
366
61 (40/66)
99
1,581
2
0.054
1,4-7,17-19
31
73
66 (19129)
29 (15152)
97
97
91 1
91 1
1
0
0
7
204
13
39
14
71 (517)
19 (391204)
69 (9113)
13 (5139)
7 (1114)
97$
97$
97$
97$
97$
1,111
1
1,111
494
1,048
456
I
I
I
2
2
0
0
0
0.10
0.016
4,8,9
8-10
3
1
9,11
4,9,10,12,13,21
4,9,14,22,23
* Numbers in parentheses are the number of patients with positive findings per the total number of patients with polyarteritis nodosa. Cost (per
patient) represents 1984 US dollars (University of Chicago Medical Center charges). Morbidity represents days of hospitalization for testing per
patient. Mortality represents deaths per 100 patients evaluated.
t Autopsy series.
4: Estimate.
procedures and to assess t h e morbidity, cost, and
mortality linked with these procedures, based on available d a t a from published sources. We also sought to
formulate testing sequences that would ensure a high
probability of true positive or true negative diagnoses
in patients with various degrees of likelihood of having
PAN a n d to evaluate these test sequences by comparing their cost, morbidity, and mortality (2).
W e chose to evaluate t h e diagnostic strategies
by t h e cost, morbidity, and mortality attributable t o
the diagnostic tests, rather than patient outcome, for
several reasons. 1) Clinicians are more comfortable
with diagnosis as an outcome than with the results of
therapy. 2) T h e specifics of therapy for PAN are
controversial. 3) The outcomes of therapy are very
uncertain, a n d judgment about management must be
individualized. 4) The costs of treating patients with
false-positive test results are difficult to quantify because they depend heavily on t h e nature of t h e diseases that were missed. The findings of our analyses
indicate the best diagnostic approach in patients with
suspected PAN and indicate the areas in which additional d a t a might be useful in confirming our conclusions.
METHODS
Studies surveyed. We conducted a survey of the
medical literature (North American and European reports),
including a MEDLINE computer search, to evaluate all
reports concerning the diagnosis of PAN. Reports of individual cases were excluded. Reports were included if they
dealt primarily with the diagnostic evaluation of PAN,
included 10 or more patients, and contained sufficient clinical detail about individual cases to assure us that the patients
had PAN and not another disorder, such as rheumatoid
vasculitis, giant cell arteritis, or essential mixed cryoglobulinemia. One exception (3) was made because of a paucity
of data on the sensitivity of nerve biopsy in symptomatic
patients. Reports concerning biopsies must have explicated
the criteria used to classify the biopsy findings as positive or
negative. Similarly, angiographic studies must have documented the arterial bed visualized and the findings that were
considered evidence for a positive result. Of more than 200
articles screened, only the reports cited herein met our
criteria and were used for further analysis.
Published reports containing clinical, pathologic, and
angiographic data on patients with PAN were selected and
pooled to determine the sensitivity and specificity of each
diagnostic test.
Definitions. Sensitivity is the probability of a positive, or abnormal, test result in a patient who has the disease
in question-in this case, PAN. Specificity is the probability
of a negative, or normal, test result in a patient who does not
have PAN. Prior, or pretest, probability is the probability
that a patient with the specific symptom complex has PAN
before a diagnostic test is performed. Posterior, or posttest,
probability is the revised probability that the patient has
PAN after the diagnostic test result is known. The posterior
probability is dependent on the prior probability of disease
and the sensitivity and specificity of the diagnostic test.
Predictive value positive is the probability that a patient with
a positive, or abnormal, test result has PAN. Predictive
value negative is the probability that a patient with a
negative, or normal, test result does not have PAN.
Studies selected. Sixty-six patients with PAN were
studied by angiography (4-7). Thirty-one patients with muscle pain and tenderness underwent open muscle biopsy
(4,8,9), and 73 patients without muscle involvement underwent open muscle biopsy (8-10). Open testicular biopsy was
performed in 13 patients with PAN (9,l I). Seven patients
with PAN and peripheral neuropathy had a nerve biopsy (3).
Needle biopsy of the kidney was performed in 39 patients
with PAN and renal involvement (4,9,10,12,13),and needle
DIAGNOSIS OF PAN BY A DECISION ANALYSIS MODEL
1119
Table 2. TESTSEQ results for a 25% prior probability of a diagnosis of polyarteritis nodosa
(PAN)
PAN confirmed
(posterior probability >95%)
No.
of tests
PAN excluded
(posterior probability < 10%)
One test
Angiograph y
Symptomatic nerve biopsy
Two tests
Symptomatic nerve biopsy plus
Symptomatic muscle biopsy or
Blind muscle biopsy or
Testicular biopsy or
Kidney biopsy
Angiography plus
Symptomatic muscle biopsy or
Blind muscle biopsy or
Blind nerve biopsy or
Testicular biopsy
Symptomatic muscle biopsy plus
Blind nerve biopsy or
Testicular biopsy or
Kidney biopsy
Symptomatic nerve biopsy plus
Blind nerve biopsy or
Testicular biopsy or
Kidney biopsy
Blind muscle biopsy plus
Blind nerve biopsy or
Testicular biopsy plus
Blind muscle biopsy or
Kidney biopsy or
Blind nerve biopsy
Testicular biopsy
Testicular biopsy plus
Blind nerve biopsy or
Kidney biopsy
biopsy of the liver was performed in 14 patients with PAN
and liver involvement (4,9,14).
Biopsy results were considered positive when necrotizing arteritis was found. Results of angiography were
considered positive only when microaneurysms were found.
Calculation of test characteristics. The sensitivity of
each test was calculated from the data described above,
except for blind nerve biopsy, which was estimated from an
autopsy series (1). The sensitivity of angiography partly
depended on which visceral bed was examined. The sensitivity of angiography of the hepatic, renal, and mesenteric
vessels varied from a low of 50% (9 of 18) for superior
mesenteric vessels to a high of 81% (13 of 16) for hepatic
vessels. We pooled the data from all vascular beds. The
specificities of angiography and muscle biopsy were calculated from published reports; it was assumed that all biopsies
had the same specificity. Bayes’ formula, as follows, was
used to calculate the posterior probability of PAN in patients
after a positive test result or after a negative test result, at
various prior probabilities (levels of diagnostic suspicion).
For a positive test result, Bayes’ formula is:
P(D/T+)
=
P(D/T-)
=
P(T +/D) * P(D)
P(T+/D) * P(D) + P(T+/IS) * P(D)
For a negative test result, Bayes’ formula is:
P(T-/D)
*
P(D)
P(T-/D) * P(D) + P(T-/D) . P(D)
In collaboration with Kay Sandacz from the University of Chicago Computation Center, we developed a computer program, TESTSEQ (documentation available upon
request), written in BASIC on a DEC-20 computer. The
program identifies test sequences that satisfy user-specified
criteria of diagnostic confirmation or elimination at userspecified prior probabilities. This program will accept up to
10 diagnostic tests and will force all possible combinations of
positive and negative results of up to 4 tests, using Bayes’
formula, assuming test independence. Combinations of 4 or
more tests are generated by a random test sequence procedure. The random test sequence will pursue successful
sequences for 1,000 possible combinations. If no successful
sequences are encountered, the program notifies the user
and asks if these default limits should be overridden, and if
so, what new limits the user wishes to set. Sample sequences
were tested for several different prior probabilities, and the
program was found to perform accurately.
Diagnostic strategies. Diagnostic strategies, consisting of sequences of tests generated by TESTSEQ, were
further evaluated by modeling each strategy as a decision
tree (15). The sensitivity and specificity of each test were
used as probabilities of outcomes. Tests were assumed to be
independent. University of Chicago Medical Center charges
for diagnostic tests in 1984 were used as utilities. The
morbidity, in excess days of hospitalization, was calculated
as the sum of the products of individual test-specific morbidity rates times the number of each test ordered in each
strategy. No other hospital charges were included. The
mortality rate was calculated similarly, as the sum of the
products of the individual test-specific mortality rates (16
23) times the number of each test ordered in each strategy.
Each strategy was represented as an entire decision
tree, consisting of sequences of chance nodes. Our goal was
to show clinicians how 2 different diagnostic strategies
(aggressive and conservative) performed to allow a choice of
approaches to patients with suspected PAN. The 2 strategies
were evaluated by determining their performance in a hypo-
ALBERT ET AL
1120
Table 3. TESTSEQ results for a 50% prior probability of a diagnosis of polyarteritis nodosa
(PAN)
No.
of tests
PAN confirmed
(posterior probability >95%)
PAN excluded
(posterior probability < 10%)
One test
Testicular biopsy or
Symptomatic muscle biopsy or
Symptomatic nerve biopsy or
Angiography
None
Two tests
Blind muscle biopsy plus
Blind nerve biopsy or
Kidney biopsy or
Liver biopsy
Symptomatic nerve biopsy plus
Symptomatic muscle biopsy or
Testicular biopsy
Blind nerve biopsy plus
Kidney biopsy
Three tests
Any three
Angiography plus
Symptomatic muscle biopsy plus
Testicular biopsy or
Blind muscle biopsy plus
Symptomatic nerve biopsy or
Testicular biopsy or
Symptomatic nerve biopsy plus
Kidney biopsy or
Testicular biopsy
Any three
Symptomatic muscle biopsy plus
Symptomatic nerve biopsy plus
Kidney biopsy or
Liver biopsy or
Blind nerve biopsy plus
Testicular biopsy or
Blind muscle biopsy
Any three
Symptomatic nerve biopsy plus
Testicular biopsy plus
Kidney biopsy or
Liver biopsy
thetical population, at different prevalence rates of PAN.
The overall sensitivity, specificity, morbidity, and costs of
the strategies were determined using a software program
developed by one of us (MDS) (24).
Sensitivity analysis. For each strategy, the decision
analysis model was analyzed by varying the assumptions
over a wide range, as suggested by clinical experience, and
by determining the effect of changing values on the overall
sensitivity, specificity, cost, morbidity, and mortality of the
strategy by a one-way sensitivity analysis. For example, the
sensitivity of angiography and testicular biopsy was varied
from 40% to 80%. The specificity of angiography alone was
varied from 90% to 100%. Disease prevalence was varied
from 25% to 75%, and the proportion of males in the
population was varied from 50% to 100%.
RESULTS
Literature review. Table 1 shows the sensitivity,
specificity, mortality, morbidity (days of hospitaliza-
tion), and dollar cost of the diagnostic procedures, as
well as the numbers of patients on which the sensitivity and specificity calculations were based. For example, data from angiograms on 366 patients revealed a
sensitivity of 61% and a specificity of 99%. Angiography costs an average of $1,581, can be expected to
require a mean of 2 additional days of hospitalization,
and has a mortality rate of 0.054 per 100 patients.
In general, directed biopsies (according to clinical involvement) are significantly more sensitive than
are blind biopsies (66% versus 29% for muscle biopsy
and 71% versus 19% for nerve biopsy). In theory, this
should also be true for directed angiography (based on
abnormal liver function test results, gastrointestinal
hemorrhage, etc.), but the available literature does not
address this point; we were therefore unable to make
this dichotomy.
DIAGNOSIS OF PAN BY A DECISION ANALYSIS MODEL
Table 4.
(PAN)
1121
TESTSEQ results for a 75% prior probability of a diagnosis of polyarteritis nodosa
No.
of tests
PAN confirmed
(posterior probability >95%)
PAN excluded
(posterior probability < 10%)
One test
Testicular biopsy or
Symptomatic muscle biopsy or
Blind muscle biopsy or
Symptomatic nerve biopsy or
Angiograph y
None
Two tests
Blind nerve biopsy plus
Kidney biopsy or
Liver biopsy
None
Kidney biopsy plus
Liver biopsy
None
Any five
Testicular biopsy plus
Symptomatic muscle biopsy plus
Liver biopsy plus
Renal biopsy plus
Symptomatic nerve biopsy
Five tests
(example)
TESTSEQ analysis. The results of the TESTSEQ analysis, in which we performed sequences of
tests that would confirm a diagnosis (>95% posterior
probability of disease) or exclude a diagnosis (<lo%
probability), are summarized in Tables 2, 3, and 4,
respectively, for patients with low (25%), medium
(50%), or high (75%) probability of having PAN. In
Table 2, the analysis for a low prior probability of the
disease (25%) shows that angiography would be sufficient as a single test to verify a diagnosis of PAN with
more than 95% certainty. The only single test that
would rule out PAN with a resulting disease probability of less than 10% would be biopsy of a symptomatic
nerve.
If, however, clinicians order 2 tests, then several combinations of tests could be selected to confirm
or exclude PAN. For example, a biopsy of a symptomatic nerve could be used with either a blind or a
symptomatic muscle biopsy, a testicular biopsy, or a
kidney biopsy to confirm PAN with a probability of
more than 95% (Table 2). Similarly, angiography could
be used with any of the 4 tests shown to exclude PAN
with a resulting probability of disease of less than 10%.
It is important to note that strategies designed to
address both possibilities (confirm and exclude) would
have to sacrifice maximal sensitivity or specificity, and
no single strategy was thus optimal. We chose to
compare 2 strategies that were compatible with TESTSEQ results and with current clinical practice.
Diagnostic strategies. The aggressive strategy
(Figure 1) consists of initial angiography, followed by
testicular biopsy (for symptomatic males) or blind
muscle and blind nerve biopsy for individuals with no
other localizing findings. If muscle or nerve is symptomatic, then the sequence is preceded by muscle or
nerve biopsy and ends with needle biopsy of the
kidney. This strategy is aggressive in that it seeks to
confirm a suspected diagnosis; that is, it emphasizes
sensitivity over specificity. It characterizes an approach that is focused on the severity of the disease
(high mortality), the difficulty in confirming the diagnosis (because of nonspecific clinical features and the
sparsity of lesions), and the necessity of treatment
with cytotoxic agents (which favorably affects outcome, yet causes high morbidity). In short, this view
focuses on the difficulty of making the diagnosis and
the necessity of confirming it.
We subjected this strategy to sensitivity analysis in which we varied crucial assumptions of the
model (such as ma1e:female ratio). Substituting kidney
biopsy for nerve biopsy when there was no symptomatic site had little effect on the overall characteristics
of the strategy. The overall characteristics and the
sensitivity analysis of the strategy are shown in Tables
5 and 6, respectively.
We compared this extensive diagnostic evaluation with a more truncated investigation that was
limited to 2 diagnostic procedures (Figure 2). This
ALBERTETAL
1122
Figure 1. Tree diagram of the "aggressive" strategy for the diagnosis of polyarteritis nodosa (PAN). A series of diagnostic
tests begins with categorization of patients into those with symptomatic muscle, those with symptomatic nerve, and those
with no symptomatic site. Symptomatic muscle or nerve is biopsied (Bx),and if the results are negative, angiography is
then performed; in those with no symptomatic site, the angiography is performed first. Thereafter, men with testicular
symptoms will undergo testicular biopsy, followed by biopsy of nerve. Substitution of kidney biopsy for nerve biopsy in
the no-symptomatic-site branch makes little difference in the outcome. The overall strategy sensitivity is 91% and
specificity is 90%. 0 = procedure performed; @ = positive findings; 0 = negative findings.
conservative strategy emphasizes specificity over
sensitivity and is based on the following logic: A
false-positive diagnosis is potentially as serious as a
false-negative diagnosis because treatment with cytotoxic drugs can cause death or serious disease; the
diagnostic procedures result in significant morbidity,
even mortality; the costs of aggressive diagnostic
evaluation are prohibitive when applied to large populations with low prevalence of disease; and the patients who have occult disease that is missed by this
diagnostic strategy may have less severe disease than
patients who have clinically apparent disease. Occlusive visceral lesions would be found by angiography ,
and patients with smaller vessel disease or nonocclusive disease may not need to take cytotoxic drugs. The
characteristics of this strategy are shown in Table 5.
DISCUSSION
Each test that we analyzed has been recommended as a diagnostic procedure €or patients with
suspected polyarteritis nodosa. However, these recommendations are often based on small numbers of
patients from biased sample populations. Furthermore, there has been no formal attempt to compare
tests against each other and to develop a rational
approach for the diagnostic evaluation of PAN. Problems abound in the analysis of patients for the presence of PAN, including the heterogeneity of clinical
manifestations, the rarity of the disease, its mimicking
of other disorders, the poor sensitivity of any single
diagnostic test, the occurrence of skip lesions, and the
frequent problem that patients are not symptomatic in
DIAGNOSIS OF PAN BY A DECISION ANALYSIS MODEL
1123
Table 5. Decision analysis of 2 test strategies for the diagnosis of polyarteritis nodosa*
Aggressive strategy
Symptomatic nerve
cost ($)
Morbidity (days)
Mortality (n)
Sensitivity (%)
Specificity (%)
Predictive value positive (%)
25% prevalence
50% prevalence
75% prevalence
Predictive value negative (%)
25% prevalence
50% prevalence
75% prevalence
Symptomatic muscle
Conservative strategy,
no symptomatic site
No symptomatic site
2,585
3.70
0.081
97
88
2,642
3.77
0.083
96
88
2,953
3.71
0.054
90
91
1,965
2.69
0.054
85
96
73
89
96
73
89
96
76
91
97
83
96
99
99
97
91
99
96
88
97
90
75
95
86
68
* Cost (per patient) represents 1984 US dollars (University of Chicago Medical Center charges). Morbidity represents days of hospitalization
for testing per patient. Mortality represents deaths per 100 patients evaluated. Numbers represent a prevalence of 5O%, and a ma1e:female ratio
of 9:l.
sites that are easily tested by biopsy. Blind biopsies
are much less sensitive than biopsies of symptomatic
sites. We analyzed diagnostic tests for PAN by surveying the entire literature and culling those studies
which were large, less biased, and sufficiently detailed
to assure us that the patients had the disease in
question. We were able to pool these data to arrive at
test characteristics. Pooling data lessens, but does not
eliminate, center-to-center variability in test sensitivity and specificity.
A problem that often arises in the assessment of
diagnostic tests is the uncertainty in estimates of test
specificity. Test specificities are often determined in
healthy adults or other patient groups in whom the
disease under study is clearly not present. When the
test is applied to groups of patients who have symptoms of the disease under study, the actual specificity
is lower because of “false-positives” (25). Thus, prospective verification of specificity estimates would be
useful, but such verifications are rarely performed.
To achieve an overall strategy that has a sensitivity of more than 90%, several tests must be utilized,
Table 6.
such that patients who have a negative result on one
test will undergo a second or third test, and so on.
Only by combining the findings of 4 or more sequential
tests can an overall strategy sensitivity of 90% be
achieved, even when one assumes that the data provided by each test are independent. This is the most
generous assumption concerning the information content of tests used in sequence. However, the calculation is based on a conservative estimate of
angiographic sensitivity because we pooled data from
all vascular beds, and not every patient had every bed
examined, and because we used a very strict criterion
of positivity-the presence of microaneurysms. Clinicians can increase the sensitivity of this procedure by
accepting less stringent criteria for positivity, such as
thrombosis or stenosis, or by visualizing each vascular
bed sequentially until positive findings are noted.
To construct rational sequences of tests, we
analyzed all possible combinations of tests which,
when performed, would exceed any selected threshold
of probability in confirming or excluding a diagnosis of
PAN. We performed these calculations for a variety of
Sensitivity analysis of aggressive strategy
Variable
Range (%)
Sensitivity of strategy (%)
~
Angiogram sensitivity
Angiogram specificity
Testicular biopsy sensitivity
Prevalence of disease
% male
40-80
25-75
84-95
90 no change
85-93
90 no change
91 no change
82-92
91 no change
91 no change
100-50
92-85
92-90
40-80
90-100
Cost ($1
Specificity of strategy (%)
~~~~~
3,108-2,875
2,879-2.998
3,058-2,938
3,461-2,511 (no
symptomatic site)
2,946-2,996 (no
symptomatic site)
ALBERT ET AL
1124
Nerve
Bx @
PAN
I
Symptomatic Nerve
Nerve Bx
Angiography @
Nerve
Bx
0
PAN
Angiography
No PAN
Patients with
suspected
c
Symptomatic Muscle
Muscle Bx
@
Muscle Bx
0 Angiography
PAN
Muscle Bx
Angiography
PAN
@
PAN
Angiogra
No PAN
AngiographyO
PAN
I
No Symptomatic Site
Male
Testicular
AngiographyP
PAN
Bx
Testicular Bx
0
Muscle Bx
@
Angiographya
NO PAN
Female
Muscle
Bx
Muscle
Bx
’
PAN
No PAN
Figure 2. Tree diagram of the “conservative” strategy for the diagnosis of polyarteritis nodosa (PAN). A series of
diagnostic tests begins with categorization of patients into those with symptomatic muscle, those with symptomatic nerve,
and those with no symptomatic site. Like the aggressive approach, the symptomatic sites are biopsied (Bx) first, and if the
results are negative, angiography is done, whereas in patients with no symptomatic site, the angiography is performed first.
In men with testicular symptoms, a testicular biopsy is performed; in women and in men without testicular symptoms, a
blind muscle biopsy is done. Unlike the aggressive strategy, the diagnostic testing ceases after 2 procedures. The overall
strategy is 85% sensitive and 96% specific. 0 = procedure performed; @ = positive findings; @ = negative findings.
given prior probabilities. The prior probabilities correspond to the clinician’s estimation of the likelihood of
a patient having a particular disease prior to testing
(prior probability). It is best not to construe prior
probabiIity as prevalence, since PAN is extraordinarily rare in the general population (fewer than 1 in
100,000 individuals) (26). For these estimates, we
chose 25% as a representative low probability, 50% as
a representative medium probability, and 75% as a
representative high probability. A clinical scenario
that might suggest a 25% probability of PAN is a
60-year-old man with low-grade fever, weight loss, and
weakness for several months. A clinical scenario that
might suggest a 50% probability of PAN is a 40-yearold man with persistent abdominal pain and peripheral
neuropathy. Similarly, a 75% probability of PAN
might be suggested by constitutional symptoms, hypertension, and testicular pain in a drug addict. These
probabilities are “subjective,” however, and are the
product of the patient’s symptoms and signs and the
clinician’s experience. Thus, individual clinical manifestations and other data, such as serologic findings,
response to prior therapy, etc., are subsumed under
the umbrella of a prior probability.
The threshold we considered clinically reasonable to confirm the diagnosis of PAN was >95%, and
to exclude the diagnosis, <lo%. Under these conditions, our computer program, TESTSEQ, generated
combinations of tests that fulfilled these criteria. Other
criteria could be used; however, these representative
criteria demonstrated the important findings for our
analysis. One important conclusion is that it is easier
to verify than to reject the diagnosis at all levels of
prior probability. The number of tests that are necessary to verify a diagnosis diminishes as the prior
probability increases, and the converse is true for
rejecting a diagnosis. Strictly in terms of test characteristics, it appears that findings of angiography and
biopsy of symptomatic nerve and symptomatic muscle
are the most useful in both making and excluding a
diagnosis. Unfortunately, it was difficult to determine
whether angiography had increased sensitivity in
symptomatic patients, although one study suggested
that this was the case (4).
DIAGNOSIS OF PAN BY A DECISION ANALYSIS MODEL
Blind biopsy of muscle and nerve, as well as
kidney and liver biopsies, yield less useful information, and the latter 2 procedures are associated with
significant mortality. We presume these latter 2 procedures would have even lower sensitivity in patients
who have no clinical involvement of the liver or
kidney. Although we used the available data on testicular biopsy to estimate test characteristics, the information is sparse and precludes definitive recommendations regarding its use. Perhaps the most important
conclusion of this analysis is that there is no single
optimal strategy.
Since no single strategy was optimal, we compared 2 clinically acceptable, TESTSEQ-compatible
strategies that differed in their approach, using decision analysis. Strictly speaking, the Bayesian inference structure of TESTSEQ and the decision analysis
of the 2 strategies are different, even though they
utilize the same mathematical tools. The Bayesian
inference structure of TESTSEQ generates any test
sequence that will convert a given prior probability to
a posterior probability that exceeds a given threshold.
The decision analysis uses Bayesian inference, but the
structure of the tree is based on considerations other
than random test sequences, such as morbidity, mortality, and cost.
In the aggressive strategy, individuals who have
symptomatic muscles or nerves will undergo biopsy of
the symptomatic organ as a first procedure. Subsequently, angiography will be performed-the first test
for patients with no symptomatic site-followed by
testicular biopsy (if symptomatic) for males, and then
muscle or nerve and kidney biopsies for females and
asymptomatic males, or for those with negative findings on testicular biopsy. This strategy works well,
achieving approximately 90% sensitivity and 91 %
specificity overall, and it is somewhat sensitive to the
way in which angiograms are read. A more liberal
interpretation of angiographic abnormalities increases
the sensitivity of the strategy up to 95%. This, of
course, is achieved at a cost of diminishing the strategy specificity, although the exact trade-off for a given
increment of sensitivity in decreased specificity is not
known (Table 6).
As the proportion of males increases in the
population, the sensitivity of the strategy increases
because an extra test (testicular biopsy) can be done.
It should be noted, however, that the testicular biopsy
estimate is based on a very few cases, and it may
decrease if larger and newer studies are performed.
Furthermore, these data are based on open biopsy of
1125
the testes. We have no information on needle biopsy of
the testes. The portion of males in the population does
not change the strategy’s specificity, but does change
its sensitivity. That is, the entire diagnostic strategy
has an overall sensitivity and specificity for the diagnosis of PAN (just as an individual test does), and
these characteristics can be altered by the availability
of findings from testicular biopsy. Overall sensitivity is
not to be confused with predictive value, which may
also increase with increasing numbers of males in
the patient population (i.e., increased prevalence),
because PAN is a male-predominant disease.
Thus, our analysis focuses on 2 strategy characteristics independent of disease prevalence: sensitivity and specificity. We also examined strategy
characteristics that are dependent on prevalence: cost,
morbidity, mortality, and predictive value. For example, when we apply the aggressive diagnostic strategy
to populations with variable disease prevalence (from
25% to 75%), the strategy sensitivity and specificity
are unchanged, but cost, morbidity, and mortality
vary, as does predictive value, which ranges from 77%
to 97%. The most glaring problem with the aggressive
diagnostic strategy is its cost, both in patient mortality/
morbidity and in dollars. As many as 1 out of every
1,200 patients undergoing assessment with this strategy will die as a result of the diagnostic tests themselves. In addition, when the prevalence of the disease
or prior probability diminishes to as low as lo%, the
diagnostic cost per case diagnosed increases to more
than $17,000. Last, from a clinical standpoint, it is
unlikely that physicians will ask their patients to
undergo up to 5 invasive procedures to diagnose this
disease.
We therefore analyzed a more truncated version of the strategy. This version is limited to 2 tests
and therefore limits the risk. Patients with a symptomatic muscle or nerve would undergo biopsy of that
organ, and if the biopsy findings were negative, angiography would be performed. Patients with n o symptomatic site would undergo angiography. If in a male
patient with testicular symptoms, the findings from the
angiography were negative, a testicular biopsy would
be performed; if the negative angiographic findings
were in a female patient or an asymptomatic male
patient, a muscle biopsy would be performed.
Some physicians might opt for a sural nerve
biopsy instead of testicular or muscle biopsy, but the
estimate of 19% sensitivity for blind nerve biopsy is
based on an autopsy series, which should result in an
overestimation of the sensitivity. Thus, we would still
1126
recommend testicular biopsy (in a symptomatic male)
or blind muscle biopsy over blind nerve biopsy. Moreover, sural nerve biopsy can result in troublesome
anesthesia or dysesthesia. One study indicates that
positive findings from sural nerve biopsy are usually
accompanied by abnormal sensory nerve conduction
on electrophysiologic testing, even in the absence of
symptomatic neuropathy (27). This suggests that electrophysiologic testing might be part of the noninvasive
evaluation of a patient with suspected PAN, the results of which would influence the selection of the
invasive tests to confirm the diagnosis.
The truncated strategy works surprisingly well
and has an overall sensitivity of 85% and a specificity
of 96%. The morbidity rate is substantially lower than
that of the extended strategy, as is the cost. The
mortality rate, however, is not very different. This
lack of difference in mortality rates reflects the impact
of angiography on the rate of mortality from the
strategy and the absence of procedures that cause
mortality in the distal branches of the tree. Since
angiography is near the front of both the truncated and
the extended strategies, it is the test responsible for
most of the fatal events, even though liver and kidney
biopsy, on a test-by-test basis, result in higher rates of
mortality. The mortality rate for the no-symptomaticsite branch of the aggressive strategy is the same as
that for the no-symptomatic-site branch of the conservative strategy because there are no tests in the
distal branches of the aggressive strategy that cause
appreciable rates of mortality. If, however, one were
to replace blind nerve biopsy with kidney or liver
biopsy, the mortality rate from the aggressive strategy
no-s ymptomatic-site branch would be higher than its
counterpart in the conservative strategy.
All in all, the truncated strategy is more favorable than the extended strategy because the morbidity
and cost are reduced by one-third. The sensitivity and
specificity differences between the extended and truncated strategies appear to be an even trade-off, with
the truncated strategy giving up 5% in sensitivity in
return for a 5% increase in specificity. The mortality
rate for the truncated strategy appeared to be only
slightly lower than that of the more extended strategy.
This analysis is based on available data and,
thus, is limited by the published studies that were used
to formulate the database. Some of these studies were
of small numbers of patients and are therefore inherently less reliable, but the degree of confidence we
have in these numbers cannot be calculated. The 69%
sensitivity of testicular biopsy is based on 13 patients.
ALBERTETAL
In autopsy series, this varied from 29% (28) to 86%
(29). Although 69% sensitivity is plausible, further
study is needed. We performed a one-way sensitivity
analysis in which important test characteristics were
varied over wide ranges to determine the effect of
imprecision in the literature estimates on our analysis.
Over a wide range of plausible values, our conclusions
concerning the aggressive and conservative strategies
are unchanged. Since our major conclusion is that
there is little evidence to recommend the aggressive
strategy over the conservative strategy, there was no
reason to extend the sensitivity analysis to varying
more than one variable, i.e., two-way or three-way
sensitivity analysis.
We have made several assumptions, one of
which is that the information contained in the tests is
independent. This assumption maximizes the diagnostic power of the tests when placed in sequence;
however, it is reasonable, given the disparity in the
nature of the tests. Specifically, each test involves
either separate anatomy or different physiology. While
this is a strong assumption, it is not unreasonable.
First, the tests address different pathophysiologic aspects of this systemic disorder and utilize different
technologies. Second, the degree of dependence will
be the same for both strategies, and it will therefore
not affect the goal of the analysis-to compare the 2
approaches. Third, nonindependence will affect the
sensitivity, specificity, and predictive value of the
strategy but not the cost, morbidity, and mortality,
which are the major variables upon which the decision
rests. Unfortunately, there are no published data to
evaluate the nonindependence of these tests. In fact,
there is little available information on the nonindependence of any tests. McNeil et a1 (30) examined the
degree of independence of the intravenous pyelogram
and the renogram for the diagnosis of renovascular
hypertensive disease. Using this analysis as a worst
case (since these imaging techniques are fairly similar), we would presume that the degree of nonindependence is less than 10%. Thus, for all these reasons,
we believe the assumption 01 test independence is
warranted. Another assumption concerns the specificity of the biopsy procedures; these data are not
available, and we made the assumption that all biopsy
material has approximately the same specificity, that
is, 97%. In the following article in this issue of Arthritis
and Rheumatism (31), we report that we found no
false-positive test results in over a thousand of these
diagnostic procedures performed at the University of
Chicago.
DIAGNOSIS OF PAN BY A DECISION ANALYSIS MODEL
Despite these limitations, this analysis clearly
shows a rational approach to the patient suspected of
having polyarteritis nodosa, an approach that has been
based on the available data. Certain aspects of this
approach need scrutiny, and the overall strategy couId
be subjected to prospective testing, although the small
number of patients presenting to any given medical
center might necessitate a multicenter analysis. Because of the uncertainties regarding these recomrnendations, however, we undertook t h e analysis of patients seen at our medical center and report our
findings in t h e article that follows.
REFERENCES
I . Cupps TR, Fauci AS: Systemic necrotizing vasculitis of
the polyarteritis nodosa group, The Vasculitides. Philadelphia, WB Saunders, 1981
2. Pauker SG, Kassirer JP: The threshold approach to
clinical decision making. N Engl J Med 302: 1 109-1 117,
I980
3. Dyck PJ, Conn DL, Okazaki H: Necrotizing angiopathic
neuropathy. Mayo Clin Proc 47:461-475, 1972
4. Travers RL, Allison DJ, Brettle RP, Hughes GRV:
Polyarteritis nodosa: a clinical and angiographic analysis
of 17 cases. Semin Arthritis Rheum 8:184-199, 1979
5. D’Izarn JJ, Boulet CP, Convard JP, Bonnin A, LedouxLebard G: L’arteriographie dans la periarterite noueuse:
a propos de 15 cas. J Radiol Electrol57:505-509,1976
6. Chatel A, Gamier T, Bigot JM, Toueg C, HClenon C:
L’arteriographie dans la periarterite noueuse. J Radiol
60:113-120, 1979
7. Vasquez JJ, San Martin P, Brabado FJ, Guerra F,
Arnalich F, Garcia-Puig J, Sanchez Mejias E: Angiography findings in systemic necrotizing vasculitis. AngiolOgy 32:773-779, 1981
8. Maxeiner SR, McDonald JR, Kirklin JW: Muscle biopsy
in the diagnosis of periarteritis nodosa. Surg Clin North
Am 32:1225-1233, 1952
9. Sack M, Cassidy JT, Bole GG: Prognostic factors in
polyarteritis. J Rheumatol 2:411-420, 1975
10. Walker GL: Neurological features of polyarteritis nodosa. Clin Exp Neurol 15:237-247, 1978
11. Cohen RD, Conn DL, Ilstrup DM: Clinical features,
prognosis, and response to treatment in polyarteritis.
Mayo Clin Proc 55:146-155, 1980
12. Conn DL, McDuffie FC, Holley KE, Schroeter AL:
Immunologic mechanisms in systemic vasculitis. Mayo
Clin Proc 51:511-518, 1976
1127
13. Keeton GR: Diagnostic and therapeutic problems of
polyarteritis nodosa. S Afr Med J 56:634-637, 1979
14. Mowrey FM, Lundberg EA: The clinical manifestations
of essential polyangiitis (periarteritis nodosa), with emphasis on the hepatic manifestations. Ann Intern Med
40:1145-1164, 1954
15. Weinstein ML,Fineberg HV: Clinical Decision Analysis. Philadelphia, WB Saunders, 1980
16. Lang EK: A survey of the complications of percutaneous
retrograde arteriography. Radiology 81:257-263, 1963
17. Halpern M: Percutaneous transfemoral arteriography.
AJR 92~918-934, 1964
18. Seidenberg B, Hurwitt ES: Retrograde femoral (Seldinger) aortography. Ann Surg 163:221-226, 1966
19. Haut G, Amplatz K: Complication rates of transfemoral and transaortic catheterization. Surgery 63594596,
1968
20. Slotkin EA, Madsen PO: Complications of renal biopsy:
incidence in 5000 reported cases. J Urol87:13-15, 1962
21. Dodge WF, Daeschner CW, Brennan JC, Rosenberg
HS, Travis LB, Hopps HC: Percutaneous renal biopsy
in children. Pediatrics 30:287-296, 1952
22. Linder H: Das Risiko der perkutanen Leberbiopsie.
Med Klin 66:924929, 1971
23. Thaler H: Uber Vorteil und Risiko der Leberbiopsiemethode nach Menghini. Wien Klin Wochenschr 76533538, 1964
24. Silverstein MD:A clinical decision analysis program for
the Apple computer. Med Decis Making 3:29-37, 1983
25. Ransohoff DF, Feinstein AR: Problems of spectrum and
bias in evaluating the efficacy of diagnostic tests. N Engl
J Med 299:92&930, 1978
26. Kurland LT, Chuang T-Y, Hunder GG: The epidemiology of systemic arteritis, Epidemiology of Rheumatic
Diseases. Edited by RC Lawrence, L E Shulman. New
York, Gower, 1984
27. Wees SJ, Sunwoo IN, Oh SJ: Sural nerve biopsy in
systemic necrotizing vasculitis. Am J Med 71525-532,
1981
28. Nuzum JW Jr, Nuzum JW Sr: Polyarteritis nodosa.
Arch Intern Med 94:942-955, 1954
29. Dahl EV, Baggenstoss AH, DeWeerd JH: Testicular
lesions of periarteritis nodosa, with special reference to
diagnosis. Am J Med 28:222-228, 1960
30. McNeil BJ, Varady PD, Burrows BA, Adelstein SJ:
Measures of clinical efficacy. N Engl J Med 293:216-221,
1975
31. Albert DA, Silverstein MD, Paunicka K, Reddy G,
Chang RW, Derus C: The diagnosis of polyarteritis
nodosa. 11. Empirical verification of a decision analysis
model. Arthritis Rheum 31:1128-1134, 1988
Документ
Категория
Без категории
Просмотров
4
Размер файла
965 Кб
Теги
polyarteritis, nodosa, diagnosis
1/--страниц
Пожаловаться на содержимое документа