Download The Importance of Sputum Cytology in the

The Importance of Sputum Cytology in
the Diagnosis of Lung Cancer*
A Cost-effectiveness Analysis
Stephen S. Raab, MD; John Hornberger, MD, MS; and
Thomas Baffin, MD, FCCP
Study objective: To assess the potential health and cost effects of initial testing with sputum
cytology to diagnose lung cancer.
Design: Cost-effectiveness analysis.
Data sources: Surveillance Epidemiology and End Results (SEER) program; cost data from
Northern California Kaiser Permanente Hospitals and Universities of Stanford and Iowa;
National Center for Health Statistics; and a MEDLINE search.
Interventions: The use of sputum cytologies preceding other tests (ie, fine-needle aspiration,
bronchoscopy, thoracoscopy) in patients with suspected lung cancer.
Main outcome measures: Mortality associated with testing and initial surgical treatment (eg,
centrally located cancer), cost of testing
performance of thoracoscopy to remove a local-stage,
and initial treatment, life expectancy, lifetime cost of medical care, and cost-effectiveness.
Results: In central lesions, sputum cytology as the first test was the dominant strategy because it
both lowers medical-care costs ($2,516 per patient) and lowers the mortality risk (19 deaths in
100,000 patients) of the evaluation without adversely affecting long-term survival. In peripheral
lesions, sputum cytology costs less then $25,000 per year of life saved if the pretest probability of
cancer exceeds 50%. The estimated annual savings of adopting sputum cytology as the first test
for diagnosing lung cancer in the United States is at least $30 million.
Conclusions: Experience in regional centers indicates that sputum cytologic testing is infre¬
even in patients with
quently ordered before implementing invasive diagnostic techniques,
central lung masses. The study findings suggest that sputum cytology as the first test in suspected
lung cancer is likely to be cost saving without adversely affecting patient outcomes.
(CHEST 1997; 112:937-45)
Key words: cost-effectiveness; diagnostic testing; pulmonary lesions
Abbreviations: FNA=fine-needle aspiration
diagnosed as
having lung
year.1 In the process of diagnosing lung cancer, most
patients have chest radiographs that reveal pulmo¬
nary lesions suggesting the presence of a cancer. To
obtain a pathologic diagnosis, patients often undergo
additional testing, such as bronchoscopy, fine-needle
aspiration (FNA), thoracoscopy/thoracotomy, and,
less commonly, sputum cytology.2-5
\M ore than 170,000 patients
-^-*-
cancer
in the
are
United States each
*From the Department of Pathology (Dr. Raab), University of
Iowa Hospitals and Clinics, Iowa City, and the Departments of
Health Research and Policy (Dr. Hornberger) and Medicine
(Drs. Hornberger and Raffin), Stanford University School of
Medicine, Stanford, Calif.
received March 19, 1997; revision accepted April 23,
Manuscript
1997.
Reprint requests: Stephen S. Raab, MD, University
of Iowa
Hospitals and Clinics, Department of Pathology, 200 Hawkins Dr
6235 RCP, Iowa
City, IA 92242-1049
Patterns of testing for lung cancer in the United
States changed significantly in the past 30 to 40
Before the 1960s, sputum cytology was
considered the best initial test in the diagnostic
evaluation of patients with chest radiographs sug¬
gesting a cancer diagnosis.24 Though data on
usage rates of sputa cytologic tests are not gener¬
ally available, at the University of Iowa and Stan¬
ford University, approximately 15% of patients
with suspected lung cancer undergo sputum cyto¬
logic testing and <5% undergo multiple
sputa
FNA have re¬
cytologic tests. Bronchoscopy and
sputum cytologic testing in the initial eval¬
placed
uation of pulmonary lesions, in part, because of
these tests' higher probability of diagnosing cancer
when cancer is present.215 While other tests pro¬
vide, on average, greater diagnostic accuracy than
sputum cytology, they also have the disadvantages
years.4
CHEST 7112/4/ OCTOBER, 1997
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
937
of greater risks of morbidity and mortality and
more costly per test.2-4
being
To our knowledge, clinical outcomes and the cost
of alternative testing strategies in patients with sus¬
cancer have never been assessed in a
pected lungtrial.
In particular, there is no direct
controlled
of
comparison testing strategies that use sputum
as an initial inexpensive and safe.albeit
cytology
less accurate.procedure to testing strategies that do
not use sputum cytology. Additional evidence on the
and disadvantages of sputum
potential isadvantages
warranted.
Decision analysis has been
cytology
shown to be particularly well suited to summarizing
existing data for deciding whether to alter clinical
practice or if further examination of clinical practice,
eg, in a randomized trial, is warranted.
This article assesses the costs and clinical out¬
comes associated with sputum cytology as the initial
test in patients with pulmonary lesions seen on chest
nonsputa strategies per added year of life saved in
patients with suspected lung cancer.
Materials
and
Figure 1 shows the decision model of the evaluation of a
reference-case male patient age 50 years with suspected lung
cancer. The patient has no history of cancer. We consider two
scenarios characteristic of patients presenting with lung lesions:
(1) a patient with a smaller lesion (<2 cm) seen in the center of
the lung (central-lesion case), and (2) a patient with a moderately
sized lesion (>3 cm) seen at the periphery of the lung (periph¬
eral-lesion case). We assume there is no clinical or radiographic
evidence of extrapulmonary disease. Additional sensitivity analy¬
ses assess variations in the reference-case assumptions.
Diagnostic Strategies
In both scenarios, the
patient may have
an initial sputum
other
cytology
proceed directly
diagnostic tests. If a
cancer diagnosis is made on any test, no further testing for a
pathologic diagnosis will be performed and a pathologic diagnosis
of cancer is followed by a staging procedure (ie, CT scan or
mediastinoscopy). The model also considers two bronchoscopy
options, depending on whether transbronchial needle aspiration
is used as a staging procedure at the time of a diagnostic
If the transbronchial needle aspiration reveals
bronchoscopy.1619
cancer in the mediastinum, we assume that staging has been
completed, else we assume that another staging procedure, such
as CT scanning, will be done.20 We did not include consideration
of positron emission tomography as a staging procedure in this
test
radiographs (sputa strategies) compared to testing
strategies that do not use sputa (nonsputa strategies).
Specifically, we compared sputa to nonsputa strate¬
gies on the following: (1) mortality of testing and
initial surgical treatment (eg, performance of a tho¬
racoscopy to remove a local-stage centrally located
cancer); (2) cost of testing and initial surgical treat¬
life
ment; (3)
expectancy; and (4) lifetime cost of
medical care. We also estimate the cost of sputa and
to
or
Surgical
O
Diagnostic
test
Methods
Regional
disease
treatment
Non-surgical
treatment
Non-surgical
treatment
o
No
Go to next
diagnosis made
diagnostic test in sequence
Nontreatable lesion
No further testing
or
treatment
0
Treatable lesion (e.g.,
granulomatous disease)
Medical
therapy
Diagnostic tests options:
Sputa
aspiration
Bronchoscopy ± needle aspiration
Thoracostomy
Expectant management
Fine needle
Figure 1. Decision tree showing sequence
peripheral lesion seen on chest radiograph.
of
tests
and
treatments
for
patient with central
938
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
or
Clinical
Investigations
model because its sensitivity/specificity and safety are still under¬
in clinical studies.2122 The next test in the
sequence will be performed if (1) sputum cytologic testing
reveals no cancer or is indeterminant or (2) nonsputa test results
are indeterminate.
On the basis of the staging procedure, a patient may have local,
regional, or distant disease.23-24 Surgery is performed if the stage
of disease is local and is not performed if the stage of disease is
regional or distant.2528 If the pathologic diagnosis reveals an
illness other than cancer (eg, granulomatous disease), the treat¬
ment will be provided appropriate to that condition.
Other than sputum cytology, tests that may be used to obtain
a pathologic diagnosis include bronchoscopy, FNA, thoracoscopy,
and expectant management. Table 1 (top) shows the sequence of
testing considered for a central lesion and Table 1 (bottom) shows
the sequence of testing considered for a peripheral lesion. In
central lesions, bronchoscopy is assumed to precede FNA be¬
cause of the established greater diagnostic yield of bronchoscopy
in these cases.6 In contrast, in peripheral lesions, FNA is assumed
to precede bronchoscopy because of the established higher yield
of FNA in these cases.12
going evaluation
treatment (ie, thoracoscopy for local-stage, centrally located
cancers), cost of medical care of testing and initial treatment, life
expectancy, and lifetime cost of medical care. Mortality of testing
and initial treatment are estimated as the cumulative probability
of death associated with continuing to be tested until a pathologic
diagnosis is made and a definitive initial treatment is provided or
the patient has a fatal complication. The cost of testing and initial
treatment includes all expenses associated with the testing pro¬
cedures, test complications, staging, staging complications, initial
treatment, and treatment complications. Life expectancy is esti¬
mated for each stage (local, regional, and distant) based on
published reports.23 Total cost of medical care includes the cost
of testing plus the expenses associated with all testing, initial
treatment, and follow-up care. Besides estimating the mean of all
measures when input variables are assigned their reference-case
values, as described in the next section, we also estimate the
standard deviation of these measures as the input values were
varied at random over a uniform distribution across the range of
permissible reference-case values.
Assumptions
In the following sections, we describe the pretest probability of
Data Sources
for each scenario, estimated life expectancies as a function
of diagnosis and stage, test characteristics (sensitivity and speci¬
ficity), and cost of testing and treatment (Table 2).
Probability of Cancer, by Scenario and Staging: After patho¬
logic and staging tests, all cancers are classified according to site
of origin (primary or metastatic), stage (local, regional, distant),
and pathologic tumor type (small cell, non-small cell). Localstage cancers are removed by thoracoscopy or thoracotomy. The
probability of the stage of cancer following surgery reflects the
accuracy of the staging procedure in correctly classifying cancers
as local stage. Staging for most patients consists of a thoracic CT
cancer
Life expectancies were derived from published reports of the
Surveillance Epidemiology and End Results Program and the
National Center for Health Statistics.2329 The Surveillance Epi¬
demiology and End Results data form a population-based tumor
registry that is used to estimate mortality rates in this population.
Cost of medical care was obtained from the Northern California
Kaiser Permanente Hospitals. Test-associated costs were calcu¬
lated from costs obtained from the billing offices of Stanford
University and the University of Iowa. These costs include both
a procedural and a pathology cost. Test characteristics, such as
sensitivity, specificity, complication rates, and morbidity rates,
were based on a MEDLINE search for literature published
between 1980 and 1995.
Outcome Measures
For each
case
scenario and
testing strategy, we estimate the
mortality associated with testing, mortality associated with initial
Table 1.Possible Sequence of Tests Assessed in a
Central Lesion Case and a Peripheral Lesion Case
Strategy
Name*
First
Second
Third
Central lesion
T
BT
FT
BFT
BFE
Thoracoscopy
Bronchoscopy
FNA
Bronchoscopy
Bronchoscopy
Peripheral lesion
T Thoracoscopy
FT
FBT
FBE
FNA
FNA
FNA
Thoracoscopy
Thoracoscopy
FNA
FNA
Thorascopy
Expectant
management
Thoracoscopy
Bronchoscopy
Bronchoscopy
Thoracoscopy
Expectant
management
*T=thoracoscopy; BT=bronchoscopy plus thoracoscopy; FT=FNA
plus thoracoscopy; BFT=bronchoscopy plus FNA; BFE=bronchoscopy plus FNA plus expectant management; FBT=FNA plus
bronchoscopy plus thoracoscopy; FBE FNA plus bronchoscopy
plus expectant management.
=
scan.
Life Expectancies: Table 2 shows life expectancies based on
whether the patient has cancer, whether the cancer is primary or
metastatic to the lung, the pathologic type of cancer (small cell
and non-small cell), the cancer stage (local, regional, distant), and
treatment (surgery or no surgery).23 The stage-specific life ex¬
pectancies in Table 2 represent life expectancies for patients with
small cell carcinoma and non-small cell carcinoma, weighted by
the probability of these diagnoses.30 The life expectancy of
patients with local-stage small cell carcinoma is based on the
patient receiving adjuvant therapy in addition to surgery.
Test Characteristics: For each test in patients with a central
lesion, Table 3 (top) shows the assumptions of true-positive rate
(sensitivity),79'1113"15'31 true-negative rate (specificity),7"9111315'31
rate of testing and initial treatment, test-related nonfamortality
tal complication rate, cost of tests, and cost of treating nonfatal
complications. All nonfatal complications do not require treat¬
ment. In addition, there are several types of complications for
each test. The values in Table 3 represent averages of complica¬
tion types and costs of those treatable complications. Table 3
(bottom) shows assumptions for similar variables in patients with
peripheral lesions.
Sputa tests consist of three to five consecutive early morning
deep-cough samples.24 Ten percent of patients will be unable to
produce a sputum sample sufficient to make a diagnosis.2431
FNA procedures are performed using a percutaneous, trans¬
thoracic technique under CT scan or fluoroscopic guid12,29,30,32-40
Diagnostic accuracy of FNA generally depends
more on lesion size than on location;12 thus, the true-positive rate
of FNA is higher for the relatively larger peripheral lesion than
for the smaller central lesion with FNA. The rate of
rax
varies between 15% and
pneumotho¬
25%;15 thus, the nonfatal all-cause
complication rate was assumed equal to 15%.
CHEST/112/4/OCTOBER, 1997
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
939
Table 2.Variable Inputs to Model, Assigned RaseCase Values, and Sources
Variable
Pretest
Value
probabilities
Stage of disease
Local
Regional
Distant
Small cell carcinoma
Peripheral lesion
Central lesion
Primary lung cancer
0.30
0.35
0.35
17
17
17
0.10
0.25
0.95
17
17
17
25
27
6
2
1
17
17
17
17
17
17
17
Life expectancy, yr
No
Reference
cancer
By stage of disease and type
of treatment
Local-surgery
Regional-surgery
Distant-surgery
Local-no surgery
Regional-no surgery
Distant-no surgery
Metastatic cancer-no surgery
Costs
2.5
2
1
1
(1995 US dollars)
Treatment, by stage of disease
and type of care
Local-initial care (first 6 mo)
Local-continuing care
Local-terminal care (last 6 mo)
Regional-initial care
Regional-continuing care
Regional-terminal care
Distant-initial care
Distant-terminal care
Staging
Staging procedure
$26,335
$ 5,076
$18,356
$19,379
$ 7,052
$15,536
$14,776
$12,436
1,000
Nonfatal complication
600
Kaiser
Kaiser
Kaiser
Kaiser
Kaiser
Kaiser
Kaiser
Kaiser
Stanford, Iowa
Stanford, Iowa
biopsy and brush,
Because
central lesions are more easily visualized than peripheral lesions
with bronchoscopy, the sensitivity of bronchoscopy for central
lesions is higher than for peripheral lesions.6 Published mortality
rates for bronchoscopy vary between 0.01% and 0.5%;12 we
selected a base-case mortality rate of 0.1%. Nonfatal complica¬
tion rates vary in the literature between 0.8% and 25%;12 the
all-cause nonfatal complication was assumed to be 15%. The cost
of the expectant management includes the cost of chest radio¬
graphs performed at intervals of 3 to 6 weeks.
Besides making a pathologic diagnosis, thoracoscopy or thora¬
cotomy can be used to remove the lesion.3339 Assuming tissue is
obtained, the sensitivity and specificity of pathologic diagnosis
with these procedures are effectively 100%. In practice, however,
the diagnostic accuracy of these procedures is less accurate
because of inadequacies in tissue sampling.33'39-41 All local-stage
cancers are assumed to be excised. In a patient with a nonneo¬
plastic lesion, a pathologic diagnosis of "benign" is made and no
further testing is necessary. If a cancer diagnosis is made by a test
other than thoracoscopy/thoracotomy and the patient has localstage cancer or the test is a false-positive, we assume that
thoracoscopy/thoracotomy was used as definitive treatment.42
We included in the sequence of test strategies "expectant
management" to account for the approximately 5% of patients in
Bronchoscopy
costs
included those of
wash,4-6 and, if done, transbronchial needle aspiration.
whom none of the prior tests established a pathologic diagnosis
and the clinical probability of cancer with the lesion is deemed
insufficient to warrant more invasive testing.4345 Expectant man¬
agement consists of regularly scheduled chest radiographs (eg,
every 3 to 6 weeks) to monitor changes in the size and appearance
of the lesion that may indicate a cancer diagnosis.45 Some experts
may use longer intervals; however, the assumption of a shorter
interval testing period is conservative in that it predominantly
favors nonsputum strategies. Our analysis also considers the
tradeoff with expectant management of avoiding the risks and
costs of more invasive testing balanced against the risks and costs
of a delay in making a diagnosis of cancer, particularly in patients
with local-stage cancer.38-45
Costs of Medical Care: The cost of medical care is categorized
as initial care (first 6 months, not including costs related to
testing, or surgical complications), continuing care (yearly costs
after initial care costs), and terminal care (6 months prior to
death).40 These costs represent those derived from patient billing
records, and include both total hospital discharge and home
health-care costs. The cost of treating nonfatal complications is an
average reported in the Kaiser data taken across all the possible
complications, weighted by the probability of the complication.
The relatively higher cost of initial care for local disease results
from additional hospital costs of excising the tumor, whereas
patients with regional or distant disease have other costs associ¬
ated with outpatient care, including chemotherapy. Life expect¬
ancies and costs are discounted at a fixed annual discount rate of
3%46 and costs are reported in 1995 US dollars.
Results
We describe separately findings for the centraland peripheral-lesion cases.
Central Lesions
Table 4 (top) shows, for a patient with a central
lesion, mortality associated with testing and initial
treatment, cost of medical care of testing and initial
treatment, life expectancy, and lifetime cost of med¬
ical care. Mortality of testing and initial treatment at
its lowest is 3.6 deaths per 1,000 patients for sputa-
bronchoscopy-FNA-expectant
management and, at
its highest, is 6.1 deaths per 1,000 patients for
FNA-bronchoscopy-thoracoscopy. Sputa-bronchoscopy-FNA-expectant management results in 19
fewer deaths per 100,000 patients compared to the
nonsputa strategy of bronchoscopy-FNA-expectant
management. The cost of testing and initial treat¬
ment is $5,120 for sputa-bronchoscopy-FNA-expectant management and $11,217 for thoracoscopy. Sputa-bronchoscopy-FNA-expectant
management costs
lower than the
of bronchos¬
$330
nonsputa strategy
copy-FNA-expectant management.
is 9.88 years for
Life expectancy
sputa-bronchoscopy-FNA-expectant
management and 9.57 years for thoracoscopy, a
difference of 114 days. Lifetime cost of medical care
is $24,561 for sputa-bronchoscopy-FNA-expectant
management and $33,984 for thoracoscopy.
Cost Per Life-Year Saved: Table 5 (bottom) shows
940
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
Clinical
Investigations
Table 3.Test Characteristics in a Central Lesion Case and Peripheral Lesion Case*
Test
Probability of Nonfatal
Complication, %
Probability of
Procedural Death,
Cost of
True-Positive
Rate, %
True-Negative
40
80
75
100
95
95
95
0
200
0.05
15
0.1
100
0.5
15
15
1,200
1,800
10,000
20
85
30
100
95
95
95
100
0
0.05
0.1
Rate, %
Test,
$f
Cost of Treating Nonfatal
Complication, $f
Central lesion
Sputa
FNA
Bronchoscopy
Thoracoscopy
Peripheral lesion
Sputa
FNA
Bronchoscopy
Thoracoscopy
0
0.5
0
200
15
15
15
1,200
1,800
10,000
0
600
800
2,000
0
600
800
2,000
*See text and references 7-15 and 26-38.
f Costs estimated from Kaiser data.
Table 4.Central Lesion Case, Outcomes and Costs
Ry Test Strategy
Mortality of Testing and
Initial Treatment
(per 1,000 Patients)
Test
Strategy
Strategy*
1
2
3
4
5
6
7
8
9
10
SBFE
BFE
T
ST
SFT
FT
SBT
SFBT
BT
FBT
Cost of Testing and Initial
(SD)
3.6(1.1)
3.8(1.1)
5.0(1.0)
5.0 (1.0)
5.4 (1.0)
5.5 (1.0)
5.8 (1.0)
6.0(1.0)
6.0 (1.0)
6.1 (1.0)
Strategy
Mean
SBFE
BFE
SBFT
SFT
SBT
BFT
FT
ST
BT
T
(SD)
5,120 (37)
5,450 (38)
8,842 (50)
9,066 (47)
9.722 (48)
9,841 (49)
10,107(46)
10,178(43)
10,906 (46)
11,217 (43)
Mean
Between Test
Difference in Life
Expectancy, d
(SD)
2.2 (2.7)
0.0 (2.7)
0.4 (2.7)
0.4 (2.7)
107.3 (2.7)
Comparison
Mean
SBFE vs BFEf
SFT vs FT
SBFT vs BFT
SBT vs BT
STvsT
Lifetime Cost of Medical
Treatment, $
Life
Care, $
Expectancy, yr
Strategy
SBFE
BFE
SFT
SBFT
FT
SBT
ST
BFT
BT
T
(SD)
9.88 (1.9)
9.88 (1.9)
9.87(1.9)
9.87(1.9)
9.87(1.9)
9.86 (1.9)
9.86 (1.9)
9.86 (1.9)
9.87(1.9)
9.57 (2.0)
Mean
Strategy
SBFE
BFE
SFT
FT
SBFT
SBT
BFT
BT
ST
T
(SD)
24,561 (98)
24,712(111)
29,328(112)
29,948 (105)
29,970 (106)
30,105 (104)
30,378 (105)
31,067(107)
31,468 (109)
33,984(113)
Mean
Strategies11
Difference in Lifetime Costs
of Medical Care, $ Mean (SD)
-151(148)
-620 (154)
-408 (148)
-962 (150)
Cost-effectiveness, $
SBFE dominant*
SFT dominant
SBFT dominant
SBT dominant
ST dominant
-2,516 (157)
sputa, B=bronchoscopy, FNA, E=expectant management, T=thoracoscopy.
Sputa
strategy minus nonsputa strategy.
*
*S
=
F
=
f
Cost-effectiveness estimate not needed because sputa strategy prolongs life expectancy and is less costly.
^Those that differ in only whether sputa examination is performed as an initial test.
differences in life expectancies and lifetime cost of
medical care between strategies that vary in only
whether sputa examination is used as an initial test
before further testing or treatment. In all five com¬
parisons, initial sputa examination results in equiva¬
lent or longer life expectancy (range=0 to 107 days)
and lower lifetime medical costs (range $151 to
$2,516). The slightly longer life expectancy associ¬
ated with sputa strategies primarily reflects sputa
=
sparing
some
patients the additional testing and
mortality associated with more invasive testing.
Peripheral Lesions
Table 5 (top) shows, for patients with a peripheral
lesion, mortality associated with testing and initial
treatment with thoracoscopy, cost of medical care of
and initial
life
testing
treatment,
expectancy, and life-
CHEST / 112 / 4 / OCTOBER, 1997
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
941
Table 5.Peripheral Lesion Case, Outcomes and Costs*
Ry Test Strategy
Mortality of Testing and
Initial Treatment,
(per 1,000 patients)
Mean (SD)
Strategy
Mortality of Testing and
Initial Treatment
Test
Strategy
(per 1,000 Patients)
Strategy
SFBE*
FBE
T
ST
SFT
FT
SFBT
FBT
(SD)
3.4 (1.0)
3.5(1.0)
5.0(1.0)
5.0(1.0
5.5(1.0)
5.5(1.0)
6.0 (1.0)
6.1 (1.0)
Mean
5,047 (35)
5,064 (36)
9,470 (48)
9,810 (49)
9,916 (46)
10,286 (53)
10,803 (43)
11,236 (43)
SFBE
FBE
SFT
SFBT
FT
FBT
ST
T
Retween Test
Difference in Life
Comparison
Expectancy, d
(SD)
1.8 (2.7)
0.4 (2.7)
0.4 (2.7)
0.0 (2.8)
Mean
SFBE vs FBEf
SFT vs FT
SFBT vs FBT
STvsT
*See Table 4 for explanation of abbreviations.
f
Lifetime Cost of Medical
Life
Expectancy, yr
(SD)
9.783 (1.9)
9.780 (1.9)
9.780(1.9)
9.779(1.9)
9.779 (1.9)
9.779 (1.9)
9.779 (1.9)
9.778 (2.0)
Strategy
Mean
SFBE
SFT
SFBT
FT
FBT
ST
T
FBE
Care, $
Strategy
FBE
SFBE
SFT
FT
SFBT
FBT
ST
T
(SD)
24,322(109)
24,460 (109)
29,505 (104)
29,654 (105)
30,309 (103)
30,427 (103)
32,898 (110)
35,099 (112)
Mean
Strategies^
Difference in Lifetime Costs
of Medical Care, $ Mean (SD)
137 (154)
-149(148)
-118(146)
-2,201 (157)
Cost-effectiveness, $
$27,600
SFT dominant*
SFBT dominant
ST dominant
Sputa strategy minus nonsputa strategy.
* Cost-effectiveness estimate not
needed because sputa strategy prolongs life expectancy and is less costly.
^Those that differ in only whether sputa examination is performed as an initial test.
time cost of medical care. The mortality of testing and
treatment at its lowest is 3.4 deadis per 1,000
initial
patients for sputa-FNA-bronchoscopy-expectant man¬
agement compared to, at its highest, 6.1 deaths per
1,000 patients for FNA-bronchoscopy-thoracoscopy.
Sputa-FNA-bronchoscopy-expectant management re¬
sults in four fewer deaths per 100,000 patients com¬
of FNA-bronchoscopypared to the nonsputa strategy
expectant management. The cost of testing and initial
treatment is $5,047 for sputa-FNA-bronchoscopy-expectant management and $11,236 for thoracoscopy.
Life expectancy was 9.783 years for sputa-FNA-bronfor
choscopy-expectant management and 9.778 years
a
differ¬
management,
FNA-bronchoscopy-expectant
ence of 1.8 days. Lifetime cost of medical care is
$24,322 for sputa-FNA-bronchoscopy-expectant man¬
agement and $35,099 for thoracoscopy.
Cost Per Life-Year Saved: Table 5 (bottom) shows
differences in life expectancies and lifetime cost of
medical care between strategies that vary in only
whether sputa examination is used as an initial test
before further testing or treatment (eg, sputa-FNA-
bronchoscopy-expectant management vs just FNAEach of the
bronchoscopy-expectant amanagement).
life
sputa strategies provides
expectancy equivalent
to or longer than the comparable nonsputa strategies
days). In only the comparison of
sputa testing preceding FNA-bronchoscopy-expect¬
ant management is there an added lifetime cost to
the sputa strategy, equal to $138. In this least
favorable
(range=0 to
1.8
comparison to sputa on the basis of cost,
initial sputa examination costs $27,600 per year of
life saved. In the other three comparisons, initial
sputa examination results in lower medical cost and
equivalent or longer life expectancy.
Sensitivity Analysis
Table 6 shows the effect of varying the pretest
of cancer on the difference in life expect¬
probability
in lifetime cost of medical care, and
difference
ancy,
cost-effectiveness in the central- and peripheral-lesion
cases. The comparison is shown only between the
testing strategies that were least favorable to sputa. For
instance, in peripheral lesions, the comparison is be¬
tween using and not using sputa examination before
FNA-bronchoscopy-expectant management. In central
lesions, the comparison is between using and not using
sputa examination before bronchoscopy-FNA-expectant management. In both scenarios, sputa testing in¬
creases life expectancy as the probability of cancer
increases. Sputa testing costs less per life-year saved in
942
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
Clinical
Investigations
Table 6.As a Function of the Pretest Probability of Cancer, Differences in Lifetime Expectancy and Costs Between
Test
Straegies
Peripheral Lesions
(SFBE*
Probability
of Cancer
Difference
in Life
Expectancy, d
Minus
Central Lesions
FBE)
Difference in
Lifetime
Costs, $
(SBFE Minus BFE)
Cost per
Difference
in Life
Saved, $
463,270
69,856
23,471
2,891
Expectancy, d
Life-yr
0.208
0.1
264
0.165
1.045
200
1.014
0.3
2.166
138
0.5
2.112
0.7
2.651
21
3.521
2
0.9
4.695
155
4.748
*See Table 4 for explanation of abbreviations.
fCost-effectiveness estimate not needed because sputa prolongs life expectancy and is less
patients presenting with central lesions and is less cosdy
for selected patients with peripheral lesions; in partic¬
ular, in patients whose probability of cancer is >50%.
The conclusions regarding the lower cost and better
survival, especially in patients with central lesions,
remained stable
to variation in
other reference-case
assumptions.
DISCUSSION
This study suggests that sputum cytology could be
used to a greater extent in the evaluation of sus¬
pected lung cancer, particularly in patients with
central lesions, to reduce the cost of the initial
evaluation without adversely affecting test-related
mortality or life expectancy. Notably, in patients
presenting with a central lesion where the pretest
probability of cancer is >50%, sputa strategies lower
medical-care costs and may prolong survival by
reducing the needIn to proceed to riskier/costlier
testing strategies. patients presenting with a pe¬
ripheral islesion, where the pretest probability of
cancer
estimated to be >50%, the cost per
saved
with sputa strategies is <$25,000.
life-year
Adopting
sputum cytology as the first test for
with
a centrally located lung lesion seen on
patients
chest radiographs has the potential to provide sub¬
stantial medical-care savings in the United States. At
least 200,000 patients with suspected lung cancer
(^fraction of central lesions (60%) Xnumber of lung
cancers/year [170,000] -r-pretest probability of cancer
[50%]) would be eligible for sputum cytology as an
initial test. The potential savings in testing costs
alone with adoption of sputum cytology equals at
least $30 million each year (>200,000X$150 savings
per test per patient). Relatively few diagnostic strat¬
egies introduced into medicine during the 1980s and
1990s have been shown to offer such a potential
advantage to both prolonging survival and lowering
Difference in
Lifetime
Costs, $
227
49
-170
-304
-506
Cost per
Life-yr
Saved, $
502,150
17,640
SBFE dominant
SBFE dominant
SBFE dominant
costly.
medical-care costs. The savings may be underesti¬
mated because this study conservatively applied es¬
timates of the accuracy of sputum cytology, and risks
of morbidity and mortality of invasive tests, to the
lower ranges of those reported in the literature.
The decision to use sputum cytology may depend
on the patient's and clinician's beliefs about the
effects of testing strategy on quality of life. Besides
reducing the mortality risk of more invasive proce¬
dures, the sputa strategies also help patients avoid
the need for tests that are uncomfortable. In con¬
trast, the performance of up to six sputa cytologic
tests may delay establishing a diagnosis by a week or
more.47 Depending on the patient's attitudes.eg,
the patient's anxiety about knowing the diagnosis
outweighs the patient's fear of and discomfort with
invasive procedures.nonsputa strategies may result
in a higher "quality-adjusted" life expectancy even
though the sputa strategies result in longer life
expectancy. Thus, clinical decisions at the bedside
may need to be tailored to patients individual atti¬
tudes about the risks and benefits of invasive proce¬
dures to diagnosis lung cancer.
The assumptions incorporated into the model, of
necessity simplified the clinical decision, depicting
dominant practice options in our institutions. Prac¬
tices at other institutions may vary, only giving
greater salience to the need for a systematic evalua¬
tion of diagnostic options for patients with suspected
lung cancer. For example, incorporating FNA with
for both diagnosis and staging1619 may
bronchoscopy
be a cost-effective option when compared to surgical
mediastinal exploration.18 However, some investiga¬
tors have raised concerns about the yield of using
transbronchial needle aspiration alone as a staging
procedure,20 and have suggested that it should be
accompanied by CT scanning. Our findings suggest
that first doing bronchoscopy with or without FNA
results in higher costs than doing sputum cytologic
CHEST / 112 / 4 / OCTOBER, 1997
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
943
tests
first, followed by a staging procedure if cyto¬
the next
diagnosis
logies
if
tests
do
not
reveal a
diagnostic procedure cytologic
tests
cancer
reveal
a cancer
or
diagnosis.
Our findings indicate, at a minimum, there should
be a reconsideration of the appropriateness of spu¬
tum cytology in the diagnostic evaluation of patients
with suspected lung cancer. Decision-analytic meth¬
ods are particularly useful methods for assessing
whether the cost of further evaluation of a clinical
strategy.ie, in a randomized clinical trial.is justi¬
fied by the potential benefits.48 However, some
clinicians may view results based on a decision
analysis skeptically, particularly because the pub¬
lished literature on the safety and efficacy of tests to
the etiology of lung lesions may be out of
diagnose
date with today's procedures or reflect practices at
selected (eg, academic) medical centers instead of
practice in the community. Because the incidence of
test-related mortality is so low, a study to assess the
differential effect of testing strategies on this out¬
come would be prohibitively expensive. To allay
concerns, however, a study could assess the effect of
sputa vs nonsputa strategies on the number of more
tests (eg, bronchoscopy, FNA, thoracotomy)
costlywere
that
avoided, the differential cost of testing, and
the diagnostic accuracy of different strategies; in
did any patients undergoing sputa testing
particular,
have a significant delay in the diagnosis of a treatable
lung cancer?
Conclusions
Health provider organizations, clinicians, and pa¬
are seeking to lower the cost of medical care
while maintaining high-quality care. For the many
thousands of patients each year who undergo diag¬
nostic evaluation of suspected lung cancer.particu¬
with centrally located
larly those patients presenting
of
addition
lesions.the
sputa examination as an
lung
initial test shows significant promise in lowering the
costs of testing and initial treatment, lowering the
lifetime costs of medical care, lowering the risk of
death from testing and initial treatment, and thus
modestly improving life expectancy.
tients
ACKNOWLEDGMENTS: The authors with to thank Dr. Joe
Selby at Kaiser-Permanente in northern California for providing
data on the
cost
Johnston WW, Bossen EH. Ten years of respiratory cytopa¬
thology at Duke University Medical Center: II. The cyto¬
pathologic diagnosis of lung cancer during the years 19701974, noting the significance of specimen number and type.
Acta Cytol 1981; 25:103-07
4 Johnston WW, Elson C. Cytology of other body sites: respi¬
rator)' tract. In: Bibbo M, ed. Comprehensive cytopathology.
Philadelphia: Saunders, 1991; 320-98
5 Mehta AC, Marty JJ, Lee FWY. Sputum cytology. Lung
Cancer 1993; 14:36-85
6 Arroliga A, Matthay R. The role of bronchoscopy in lung
cancer. Lung Cancer 1993; 14:87-98
7 Farber S, Benioff M, Frost J, et al. Cytologic studies of
sputum and bronchial secretions in primary carcinoma of the
lung. Dis Chest 1948; 14:633-64
8 Koss L, Melamed M, Goodner J. Pulmonary cytology: a brief
survey of diagnostic results from July 1st, 1952 until Decem¬
ber 31st, 1960. Acta Cytol 1964; 8:104-13
9 McDonald J. Pulmonary cytology. Am J Surg 1955; 89:462-64
10 Ng ABP, Horack GC. Factors significant in the diagnostic
accuracy of lung cytology in bronchial washing and sputa
samples: II. Sputum samples. Acta Cytol 1983; 27:397-402
11 Pilotti S, Rilke F, Gribaudi G. Sputum cytology for the
diagnosis of carcinoma of the lung. Acta Cytol 1982; 26:
649-54
12 Salazar A, Westcott J. The role of thoracic needle biopsy for
the diagnosis and staging of lung cancer. Lung Cancer 1993;
14:99-110
13 Spjut H, Fier D, Ackerman L. Exfoliative cytology and
pulmonary cancer. J Thorac Surg 1955; 30:90-107
14 Tanaka T, Yamamoto M, Tamura T, et al. Cytologic and
histologic correlation in primary lung cancer: a study of 154
cases with resectable tumors. Acta Cytol 1985; 28:49-56
15 Umiker W. The role of exfoliative cytopathology in the
routine diagnosis of bronchogenic carcinoma: a five year
3
study of 152 consecutive, unrelated cases. Dis Chest 1961;
40:154-59
16 Garcia RF, Diaz Lobato S, Pino JM, et al. Value of CT-guided
fine needle aspiration in solitary pulmonary nodules with
negative fiberoptic bronchoscopy. Acta Radiol 1994; 35:
478-80
17 Harrow EM, Wang KP. The staging of lung cancer by
bronchoscopic transbronchial needle aspiration. Chest Surg
Clin North Am 1996; 6:223-35
18 Schenk DA, Chambers SL, Derdak S, et al. Comparison of
19
20
21
22
of treating lung cancer.
the Wang 19-gauge and 22-gauge needles in the mediastinal
staging of lung cancer. Am Rev Respir Dis 1993; 147:1251-58
Wang KP. Transbronchial needle aspiration and percutane¬
ous needle aspiration for staging and diagnosis of lung cancer.
Clin Chest Med 1995; 16:535-52
Castella J, Buj J, Puzo C, et al. Diagnosis and staging of
bronchogenic carcinoma by transtracheal and transbronchial
needle aspiration. Ann Oncol 1995; 6(suppl 3):S21-4
Chin R, Ward R, Keyes JW, et al. Mediastinal staging of
non-small-cell lung cancer with positron emission tomogra¬
phy. Am J Respir Crit Care Medet 1995;The152:2090-96 of FDG
usefulness
Sasaki M, Ichiya Y, Kuwabara Y, al.
positron emission tomography for the detection of mediasti¬
nal lymph node metastases in patients with non-small cell
lung cancer: a comparative study with x-ray computed tomog¬
23:741-77
raphy. Eur J Nucl Med 1996;
National Cancer Institute, Health. Lung cancer surveillance,
epidemiology and end results (SEER) program. Bethesda,
Md; National Institutes of Health Press, 1994
Mountain CF. A new international staging system for lung
cancer. Chest 1986; 89:225S-233S
Greenberg S, Fraire A, Kinner B, et al. Tumor cell type
'
REFERENCES
23
1 American Cancer Society. Cancer facts and figures. Atlanta:
American Cancer Society, 1995
2 Erozan Y, Frost J. Cytopathologic diagnosis of cancer in
pulmonary
Acta
material:
a
critical
Cytol 1970; 14:560-65
histopathologic
24
correlation.
25
944
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
Clinical
Investigations
staging in the prognosis of carcinoma of the lung.
Pathol Annu 1995; 2:398-405
26 Hoogstraten B. Lung tumors. Berlin: Springer-Verlag, 1988
27 Martini N. Operable lung cancer. CA Cancer J Clin 1993;
versus
43:201-14
28 Shields T. Surgical therapy for carcinoma of the lung. Lung
Cancer 1993; 14-47:121-47
29 National Center for Health Statistics, Public Health Service.
Vital statistics of the United States, life tables (vol 2, section
6). Washington, DC: US Government Printing Office, 1986
30 Ginsberg R. Surgeiy7 and small cell lung cancer: an overview.
Lung Cancer
31 DeMay RM. Respirator)7 cytology: the art and science of
cytopathology. Chicago: American Society of Clinical Pathol¬
1989;'5:232-36
ogy Press, 1996; 207-56
32 Austin J, Cohen M. Value of having a cytopathologist present
during percutaneous fine-needle aspiration biopsy of the
lung: report of 55 cancer patients and meta-analysis of the
literature. Am J Roentgenol 1993; 160:165-77
33 Coosemans W, Lerut T, Van Raemdonck D. Thoracoscopic
surgeiy: the Belgian experience. Ann Thorac Surg 1993; 56:
621-60
34
35
Cummings S, Lillington G, Richard R. Estimating the prob¬
ability of malignancy in solitary pulmonary nodules: a bayes¬
ian approach. Am Rev Respir Dis 1986; 134:449-52
Godwin JP, Speckman JM, Fram EK, et al. Distinguishing
benign from malignant pulmonary nodules by computed
tomography. Radiology 1982; 144:349-52
36 Harter LP^ Moss AA,' Goldberg HI, et al. CT-guided fineneedle aspirations for diagnosis of benign and malignant
disease. Am J Roentgenol 1983; 140:363-67
37 Khouri NF, Stitik FP, Erozan YS, et al. Transthoracic needle
aspiration of benign and malignant lung lesions. Am J Roent¬
38
39
genol 1985; 144:281-88
Lillington G, Cummings S. Decision analysis approaches in
solitary pulmonary nodules. Semin Respir Med 1989; 10:227-31
Miller J. The present role and future considerations of
video-assisted thorascopy in general thoracic surgery. Ann
Thorac
Surg 1993; 56:804-06
40 Warner K, Luce B. Cost-benefit and cost-effectiveness anal¬
ysis in health care: principles, practice and potential. Ann
Arbor, Mich: Health Administration Press, 1982
41 Cooper J. Perspectives on thoracoscopy in general thoracic
surgery. Ann Thorac Surg 1993; 56:697-700
42 Kaiser L, Bavaria J. Complications of thoracoscopy. Ann
Thorac
Surg 1993; 56:796-98
43 Garland L. The rate and growth of natural duration of
primary bronchial cancer. Am J Roentgenol Rad Ther Nucl
Med 1966; 96:604-11
44 Nathan M, Colloing V, Adams R. Differentiation of benign
and malignant pulmonary nodules by growth rate. Radiology
1962; 79:221-32
45 Nathan M. Management of solitary pulmonary nodules: an
organized approach based on growth rate and statistics.
46
47
48
JAMA 1974; 227:1141-44
Lipscomb J. Time preference for health in cost effectiveness
analysis. Med Care 1989; 27(suppl 3):S233-52
Fox CR, Tversky A. Ambiguity aversion and comparative
ignorance. Q J Econ 1995; 585-603
Hornberger J, Brown B, Halpern J. Designing a cost-effective
clinical trial. Stat Med 1995; 14:2249-59
CHEST/112/4/OCTOBER, 1997
Downloaded From: http://publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/21752/ on 05/12/2017
945