вход по аккаунту



код для вставкиСкачать
Image Analysis Combined with Visual Cytology in the
Early Detection of Recurrent Bladder Carcinoma
Allison M. Richman, M.P.H.1
Susan T. Mayne, Ph.D.1
James F. Jekel, M.D.1
Peter Albertsen, M.D., M.S.2
BACKGROUND. Early detection of recurrent transitional cell carcinoma of the bladder (TCC) is important to permit early treatment, which produces maximal preservation of the bladder and maximum survival.
METHODS. This retrospective cohort study attempted to determine the period of
time over which urinary DNA image analysis combined with visual cytology is
Department of Epidemiology and Public Health,
Yale University School of Medicine, New Haven,
Division of Urology, Department of Surgery,
University of Connecticut Health Center, Farmington, Connecticut.
useful in the early detection of recurrent TCC of the bladder. The authors believe
this study is unique in that it measured the effectiveness of this test (image analysis
plus visual cytology combined) at varying times before clinical diagnosis of recurrence was made. The cohort was comprised of 175 urologic patients from urologic
practices across the U.S. Data, collected between January 1991 and February 1994,
included cystoscopy, biopsy, DNA image analysis, and visual cytologic reports.
RESULTS. Sixty patients in the cohort were found to have active TCC whereas 115
patients had a history of, but no active, disease during the follow-up period. As
expected, the sensitivity and specificity of DNA image analysis in combination
with visual cytology, and DNA image analysis alone, were greatest when urinary
samples were obtained close to the time of diagnosis. In general, the longer the
interval from the combined tests to the time of diagnosis, the lower the sensitivity.
The combined tests had predictive value up to 3 months prior to clinical diagnosis
when any detectable cytologic abnormality was considered positive. At the optimal
cutoff points as determined from receiver operating characteristic curves, sensitivity increased when DNA image analysis was supplemented with visual cytology.
CONCLUSIONS. The combination of DNA image analysis and visual cytology provides a better method for the early detection of recurrent TCC than DNA image
analysis alone. This test potentially may be useful in providing information regarding bladder tumor recurrence up to 3 months prior to clinical evidence of disease.
Cancer 1998;82:1738–48. q 1998 American Cancer Society.
KEYWORDS: image analysis, image cytometry, aneuploidy, cytology, bladder neoplasms, sensitivity and specificity.
Supported by DIANON Systems, Inc, Stratford,
The authors thank DIANON for kindly providing
the laboratory data for this study and the urologists for kindly providing or allowing access to
patient record data.
Address for reprints: Allison M. Richman,
M.P.H., 1600 N. Rhodes St., Arlington, VA
Received October 1, 1997; accepted October 24,
uperficial tumors of the bladder recur locally in approximately
50% of transitional cell carcinoma (TCC) patients within 6 – 12
months of their removal.1 Therefore, patients with TCC are at high
risk for repeated tumor episodes. If left undetected and untreated,
such tumors may progress and require more radical and debilitating
procedures than if detected and treated early. The availability of an
effective and noninvasive method for monitoring patients with TCC
of the bladder would increase the lead time for treatment and improve
outcomes. In addition to predicting disease recurrence, biomarkers
also may be used as surrogate intermediate endpoints for monitoring
chemopreventive therapy.
DNA image analysis has received attention as a new, highly technical method for detecting and providing prognostic information re-
q 1998 American Cancer Society
/ 7bc2$$1622
04-09-98 15:07:34
W: Cancer
Bladder Carcinoma and Image Analysis/Richman et al.
garding TCC of the bladder. 2 – 18 A more traditional
way of measuring DNA content of bladder cells is flow
cytometry.19 – 26 Studies comparing the value of image
analysis with flow cytometry have shown varying results.27,28 Several recent studies on image analysis have
used bladder tissue specimens.2 – 12 Fewer studies have
attempted to evaluate the use of image analysis for
the detection and prognosis of bladder carcinoma using urine specimens.13 – 16
Cross-sectional studies have suggested that image
analysis is more effective in identifying tumors than
is visual cytology of urine specimens.13,14 In particular,
sensitivity is increased. Follow-up studies demonstrated through regression analysis that DNA image
analysis can predict TCC progression, recurrence free
survival, and overall survival.2 – 4 Gerber et al. found
that DNA image analysis produced an improved sensitivity (93% vs. 36%) but decreased specificity (41% vs.
95%) compared with visual cytology.15 However, when
both procedures were used together, both measures
attained the highest level. Amberson and Laino reported improved sensitivity for DNA image analysis
combined with visual cytology compared with visual
cytology alone (84% vs. 72%).16
Based on reports that sensitivity is improved when
DNA image analysis is combined with visual cytology
of urine specimens,15,16 the objective of the current
study was to attempt to replicate those findings within
a retrospective cohort design, to evaluate changes in
sensitivity and specificity over time using mostly noninvasively obtained urine specimens, and to determine
the length of time for which the combined tests have
clinical utility.
Subjects were selected from the data base of urologic
patients throughout the U.S. who had urine samples
analyzed by one diagnostic laboratory. The laboratory
data base was used to identify all patients for whom
DNA image analysis in combination with visual cytology had been performed.
Two mailings were done to identify patients eligible for inclusion in the study and to collect relevant
clinical data. The first mailing was sent to all urologists
in the laboratory’s data base meeting the following
criteria: 1) the urologist began using the laboratory
test (image analysis plus visual cytology) for urine
specimens prior to December 31, 1990, and 2) the urologist had at least one patient who had had at least
three independent urine samples tested. The mailing
inquired as to whether the patient had a history of
TCC and, if so, whether the TCC had recurred (i.e.,
showed a positive biopsy or positive clinical diagnosis)
during the study period. The responses were used to
/ 7bc2$$1622
04-09-98 15:07:34
exclude patients for whom the laboratory test was being used for reasons other than monitoring TCC and
to obtain preliminary estimates on the rate of recurrence in this population. The second mailing included
the data collection form used as a survey instrument
for the physician. It was sent to the 113 urologists who
reported having eligible patients in the first mailing.
All eligible patients were potential study subjects.
A total of 175 urologic patients, under the care of 50
private urologists, were studied. Recurrent patients
had at least one tumor episode diagnosed during the
follow-up period, whereas disease free patients had a
history of TCC but no active disease during this period,
according to the clinical data. The study period went
from January 1991 to February 1994, during which
time urine specimens were collected and clinical procedures were performed. All data were collected from
medical records; hence, the investigators had no direct
patient contact.
Clinical disease status was determined from the
results of biopsy and cystoscopy as reported in medical records. These clinical data were used as the ‘‘gold
standard’’ against which laboratory data were compared. Laboratory data were recorded for ten time periods prior to the end of the study period. The length
of these time periods, defined by the interval between
the laboratory test and clinical observation, generally
was chosen as a 3-month period to be consistent with
the expected length of time between follow-up visits
for a patient experiencing a recent disease episode.
Inclusion Criteria
The inclusion criteria were: 1) a positive biopsy within
the study period (case) or a history of TCC prior to the
study period but with no evidence of disease during
the study period (control); 2) at least three DNA image
analyses on urine specimens simultaneous with visual
cytology; 3) the patient’s urologist began using the
DNA image analysis before 1991; 4) the patient’s urologist responded to the first mailing; and 5) the patient
underwent at least one cystoscopy or biopsy during
the designated study period.
Exclusion Criteria
Exclusion criteria were: 1) bladder carcinoma of a histologic type other than TCC; 2) no cystoscopy or biopsy performed during the study period; and/or 3)
sufficiently poor data quality so that the disease status
could not be determined. Cystectomy was not a criterion for exclusion of patients, who were still at risk for
developing TCC in the ureters or the renal pelvis of
the kidneys. Patients with TCC in areas of the genitourinary tract other than the bladder were not excluded
because DNA image analysis measured TCC cells exfo-
W: Cancer
CANCER May 1, 1998 / Volume 82 / Number 9
liated in the urine originating from any site in the
urinary tract.
Definition of Endpoints
Biopsy results were considered the gold standard for
determining the disease status for bladder carcinoma,
except when no biopsy was performed. When no biopsy was performed, a negative cystoscopy was accepted as indicating that no tumor was present. A positive cystoscopy and subsequent treatment for tumor
recurrence were accepted as evidence of recurrent disease.
Study Period
The start of the study period began at the patient’s
‘‘reference date,’’ which was the date of the patient’s
first laboratory test after January 1, 1991. For participants with recurrent disease, the follow-up period extended until the first date of diagnosis of tumor recurrence after the reference date. For disease free participants, the endpoint was defined by the last followup visit to the physician’s office prior to February 28,
Clinical Data Collection Methods
Data were collected by two methods. First, for 60 patients in Chicago, New York, and Washington, DC, data
were abstracted by one of us (A.R.) from medical
charts in 5 doctors’ offices. The selection of these offices was based on location (proximity to the diagnostic laboratory or a large number of eligible patients in
one office) and willingness of the urologist to allow
data abstraction of his or her charts. Data on the other
115 patients, under the care of 45 urologists, were collected by means of the second mailing. Doctors were
asked to complete a data collection form on each of
their patients included in the study.
The data collection form was individualized for
each patient, based on the reference date. It requested
information regarding the dates and results of biopsies
and cystoscopies performed during the study period,
examined prospectively from the reference date. Pathology data included the date of the first biopsy on
or after the reference date, the histologic type, grade,
and stage of the tumor, and the pattern of growth (i.e.,
papillary, carcinoma in situ [CIS], or solid). Cystoscopic information included the date of each examination during the study period, whether or not a tumor
was visualized, the stage of the tumor, and the pattern
of growth. Telephone calls were made to the physicians’ offices for clarification when data were inconsistent internally. Those patients whose disease status
or clinical endpoints were unclear and could not be
validated were excluded from the study.
/ 7bc2$$1622
04-09-98 15:07:34
To maintain the confidentiality of patients and
physicians, no personal identifiers were entered into
the data set for analysis. Each patient was given a study
code number.
Clinical Procedures Performed during Follow-Up
During the follow-up period, every patient underwent
at least 1 clinical procedure; 36% underwent at least
1 biopsy. Information was collected on all 847 cystoscopies reported to have been performed during the
study period, 85 of which (10%) were positive. ‘‘Suspicious’’ cystoscopic findings were interpreted as negative. The number of cystoscopies per patient ranged
from 1 – 10 with an average of 4.8. Biopsies were recorded for 63 patients, 45 of which (71%) were positive.
Only the first biopsy performed during the study period was recorded for purposes of this analysis because
it served as an individual endpoint for follow-up of
the patient.
Laboratory Procedures
Laboratory results, including laboratory diagnosis and
aneuploidy score (AS), were obtained from the mainframe data base at the diagnostic laboratory. The term
‘‘laboratory diagnosis’’ or ‘‘laboratory test’’ is used to
refer to the combination of visual cytology and image
analysis results that were obtained as a single score
from the database. Data for image analysis alone are
presented; however, data for visual cytology alone are
For the DNA analysis, urine specimens (voided
urine, catheterized urine, or bladder washings) had
been collected, fixed, and prepared. Most urine specimens were voided urine but catheterized urine and
bladder washings were accepted when invasive procedures were performed for reasons other than this test.
The standard procedures of the Feulgen staining
method were used to stain the cells.29,30
Visual cytology
Visual cytologic evaluations initially were made by cytotechnologists, who recorded a preliminary diagnosis
using standard codes commonly used to report cytologic findings. The criteria used for assessment were
standard, accepted criteria, including nuclear to cytoplasmic ratio, intensity of staining, texture of chromatin, definition of boundaries, and degree of differentiation.5,10,31
Image cytometry
The Leitz Autoplan image analyzer was used (Leitz,
Wetzlar, Germany). The same slides used for cytology
were stained for DNA image analysis; the same group
of cells was examined using each method. A histogram
W: Cancer
Bladder Carcinoma and Image Analysis/Richman et al.
was produced showing the ploidy of the cells, and an
AS quantified the degree of abnormality of the DNA.
This score was calculated according to the method of
Bocking et al., in which the amount of deviation from
the normal 2c state was measured.32 This score then
was converted from a geometric scale to a logistic scale
to create an AS range from 0 – 100. A score of ¢ 25 was
considered abnormal.
Laboratory diagnosis
A laboratory diagnosis was given on a scale of one to
five that also was used for the visual cytologic evaluations: 1 Å negative, 2 Å reactive, 3 Å atypical, 4 Å
suspicious, and 5 Å positive. Initial diagnosis was
made by a cytotechnologist. A final diagnosis was
made by a pathologist who reviewed the 1) slides, 2)
cytotechnologist’s visual cytology report, 3) histogram,
and 4) AS. The quality of the specimen, such as the
number of cells and morphologic effects of treatment,
were noted to improve the validity of diagnosis.
Data Analysis
The goal of the analysis was to determine the ability
of the DNA image analysis, in combination with visual
cytology, to predict disease status at different points
in time prior to clinical evidence of disease. Data were
entered using EpiInfo Version 5 33 and were analyzed
using the Statistical Analysis System Version 6.34
The strategy of analysis involved classifying all patients as either recurrent or disease free, based on the
clinical evidence. The diagnostic tests then were classified into time intervals, based on the number of days
by which the laboratory test preceded the clinical endpoint. The taxonomy for the time intervals was: õ 2
days (simultaneous); 2 – 30 days; 31 – 60 days; 61 – 90
days; 91 – 80 days; 7 – 9 months; 10 – 12 months; 13 – 15
months; 16 – 18 months; and ú 18 months.
Sensitivity and specificity
Two-by-two tables of clinical disease status by diagnostic laboratory test results were generated for each
time interval. Sensitivity, specificity, false-positive and
false-negative error rates, and positive and negative
predictive values were determined for each table.35
Sensitivity and specificity are reported in the current
The true disease status was either positive (cases
with recurrent disease) or negative (disease free controls). Disease status was used as the standard against
which to compare two sets of laboratory results: the
laboratory diagnosis and the AS. The 2 1 2 tables were
repeated for different cutoff points, corresponding to
different levels of rigor in assigning a diagnosis. A dichotomous variable was created from the laboratory
/ 7bc2$$1622
04-09-98 15:07:34
diagnosis by cutting the ordinal variable at different
points: 1 (negative) versus 2 – 5; 1 – 2 versus 3 – 5; 1 – 3
versus 4 – 5; and 1 – 4 versus 5 (positive). The AS alone
was made into a dichotomous variable by considering
a score of 15/ positive, then 20/ Å positive, then 25/
Å positive, and finally 30/ Å positive.
Receiver operating characteristic curves
Receiver operating characteristic (ROC) curves were
used to assess the strength of the diagnostic test as
a predictor for bladder carcinoma recurrence in all
patients studied, and to determine the interval before
diagnosis for which the screening test was useful. ROC
curves also were used to compare the relative ability
of the laboratory diagnosis (including both DNA image
analysis and visual cytology) and the AS score, which
is determined from only the DNA image analysis, to
predict the true status.
Mailing Results
A total of 102 of 195 questionnaires from the first mailing were returned by the urologists (52%). Overall responses regarding 423 patients showed that 359 (85%)
had a history of TCC. Of 403 responses to the second
question, 198 patients were reported to have had active TCC (49%), defined as positive biopsy or positive
cystoscopy when no biopsy was performed.
Forty-eight of 113 second questionnaires were
completed and returned by mail (42%). These contained data on 147 patients, compared with the 418
patients requested (35%). Thirty-two patients were excluded (21.8%), 23 for lack of clinical follow-up and 9
due to inadequate data. The remaining 115 patients
whose data were collected by questionnaire were included in the study. Of the 60 patients whose data
were collected by chart abstraction, none were excluded. This left 175 patients with adequate data (37%
of the total 478 patients for whom data were sought).
Descriptive Statistics
Of 175 patients, 126 (72%) were male and 49 (28%)
were female. Their ages ranged from 32 – 94 years, with
a mean age of 69.8 years (standard deviation Å 11.3
years); there was no statistically significant difference
between the ages of men and women.
The study subjects were patients of 50 different
private physicians. One physician had 46 patients included in the study, and each of the other physicians
had between 1 and 6 patients included. The study patients lived in all parts of the U. S.: 34% were from the
Central region, 19% from the Mid-Atlantic states, 10%
W: Cancer
CANCER May 1, 1998 / Volume 82 / Number 9
from the Northeast, 18% from the Pacific region, 9%
from the Southeast, and 10% from the Southwest.
Disease characteristics
Of the 175 study subjects, 115 remained disease free,
whereas 60 patients had a positive diagnosis of TCC
during the study period. Of the 60 clinically diseased
patients, 45 (75%) were confirmed by biopsy, and 15
(25%) were diagnosed by cystoscopy alone. Six of the
15 patients diagnosed by cystoscopy alone had at least
1 negative biopsy during the follow-up time, prior to
the diagnosis. Of the 115 patients considered clinically
negative, 12 were biopsied with negative results, and
103 were diagnosed as negative by cystoscopy.
Among the 60 patients with recurrent disease, 50
(83%) had superficial tumors, 5 (8%) had smooth muscle invasion, and the pattern of invasion was unclear
for 5. Of the 50 superficial tumors, 30 were specified
further by tumor classification as determined by biopsy: 8 (27%) were Tis (CIS), 19 (63%) were Ta (no
invasion of the lamina propria), and 3 (10%) were T1
(invasion of the lamina propria but not the muscularis)
(TNM classification system).36
Of the 45 cases determined by biopsy, 13 (29%)
were Grade I (well differentiated), 15 (33%) were Grade
II (moderately well differentiated), 11 (24%) were
Grade III or IV (poorly differentiated), and the grade
was not reported for 6 (13%).
Among the 60 patients with recurrent disease, 37
(62%) exhibited papillary growth patterns alone, 2 (3%)
had solid tumors alone, 6 (10%) had CIS alone, 3 (5%)
had CIS and papillary growth patterns, 2 (3%) had solid
tumors and CIS, and 10 (17%) were not categorized
due to incomplete data. Of the tumors for the 15 cases
determined by cystoscopy alone, 8 were reported as
papillary, 1 was reported as CIS, and 6 were not specified.
Chart abstraction versus mailing
Of the 175 study subjects, data were collected by chart
abstraction for 60 and by mailing for 115. The proportion of positive cases in each group is equivalent: 35%
in the chart abstraction group compared with 34% in
the mailing group; however, biopsy confirmation of
disease was more likely in the chart review group (86%
vs. 69%). Both groups had a high proportion of superficial tumors; however, missing data make comparison
of percentages unreliable.
Effect of Time and Stringency of Criteria for a Positive
Test on Sensitivity and Specificity
The sensitivity and specificity of the laboratory diagnosis, resulting from the combined visual cytology and
image analysis test, are shown at various cutpoints as
/ 7bc2$$1622
04-09-98 15:07:34
FIGURE 1. Sensitivity of laboratory diagnosis for transitional cell carcinoma of the bladder at four cutoff points as a function of time. Laboratory
diagnosis (1 Å negative, 2 Å reactive, 3 Å atypical, 4 Å suspicious, and
5 Å positive) represents the combination of image analysis and visual
cytology. The time interval represents the time between the laboratory test
and clinical observation.
a function of time in Table 1 and Figures 1 and 2. Time
intervals, shown on the x axis, indicate the time from
the laboratory test date to the date of diagnosis. The
graphed lines show sensitivity and specificity over
time at four different levels of stringency for defining
a positive laboratory diagnosis. Sensitivity and specificity of the AS at various cutpoints are shown in Table
2 and Figures 3 and 4. As hypothesized, as the time
from the test to the time of clinical diagnosis increased
(up to approximately 9 months), the sensitivity of the
laboratory diagnosis decreased and, generally, specificity increased.
In addition, as generally was expected, sensitivity
decreased and specificity increased with increasing
stringency in the criteria for a positive laboratory diagnosis. Sensitivity and specificity also were affected by
changes in the cutoff point for the AS, although by
much less than by changes in the cutoff point for the
combined laboratory diagnosis.
When the laboratory diagnosis was made at essentially the same point in time as the clinical diagnosis
(interval 1), the striking effects of stringency of the
laboratory test criteria on sensitivity are evident in Table 1 and Figure 1. The sensitivity of the laboratory
diagnosis decreased from 80.4% (when the cutoff point
for the test Å 1 vs. 2 – 5) to 7.1% (when using a cutoff
W: Cancer
Bladder Carcinoma and Image Analysis/Richman et al.
Sensitivity and Specificity of Laboratory Diagnosis (DNA Image Analysis plus Visual Cytology) at Different Time Intervals and Cutoff Points for
all Patients (n Å 175)
Cutoff point
1 vs. 2/
1–2 vs. 3/
1–3 vs. 4/
1–4 vs. 5
0–1 day
2–30 days
31–60 days
61–90 days
91–180 days
7–9 mos
Sens: sensitivity; Spec: specificity.
diagnosis of recurrence was made. There was little effect of a longer time interval on sensitivity when the
test cutoff point was set at 1 – 4 versus 5, because the
sensitivity was very low to begin with.
FIGURE 2. Specificity of laboratory diagnosis for transitional cell carcinoma of the bladder at four cutoff points as a function of time. Laboratory
diagnosis (1 Å negative, 2 Å reactive, 3 Å atypical, 4 Å suspicious, and
5 Å positive) represents the combination of image analysis and visual
cytology. The time interval represents the time between the laboratory test
and clinical observation.
point of 1 – 4 vs. 5). Likewise, the specificity increased
from 41% at a cutoff point of 1 versus 2 – 5 to 97% at
a cutoff point of 1 – 4 versus 5.
In addition, keeping the cutoff point for the laboratory test at 1 versus 2 – 5, the sensitivity of the laboratory test fell from 80.4% when the test and clinical
diagnosis were made at essentially the same time to
25.9% when the test was done 7 – 9 months before the
/ 7bc2$$1622
04-09-98 15:07:34
ROC Curve
An ROC curve was developed for the combined laboratory diagnosis to determine which cutoff point for the
laboratory reports showed the best operating characteristics (Fig. 5). An ideal test would show an ROC
curve that rose along the left-hand y axis and touched
the upper left corner, at which point the sensitivity
would be 1.0 (100%) and the false-positive error rate
would be 0.0 (so that the specificity also would be
The combined test showed a disappointing ROC
curve. Its operating characteristics were not much better than the line of no benefit (a straight line from the
lower left corner to the upper right corner). The ROC
curve in Figure 5 plots the four possible cutoff points
at the simultaneous time interval. The 1 versus 2 – 5
cutoff point at the simultaneous time interval showed
a higher test sensitivity than did any other combination of characteristics. ROC curves were developed for
all the other cutoff points and time intervals, none of
which were as good as the one shown in Figure 5.
Image analysis in combination with visual cytology is
supported for use in monitoring patients for recurrence of bladder carcinoma and response to treatment. The time element in this study showed that DNA
image analysis plus visual cytology may be indicative
of malignant changes up to 3 months prior to discovery by clinical follow-up. The sensitivity of the test
W: Cancer
CANCER May 1, 1998 / Volume 82 / Number 9
Sensitivity and Specificity of Aneuploidy Score at Different Time Intervals and Cutoff Points for all Patients (n Å 175)
Cutoff point
õ 14 vs. 15/
õ 19 vs. 20/
õ 24 vs. 25/
õ 29 vs. 30/
0–1 days
2–30 days
31–60 days
61–90 days
91–180 days
7–9 mos
Sens: sensitivity; Spec: specificity.
The time interval represents the time between image analysis and clinical
FIGURE 4. Specificity of the aneuploidy score for transitional cell carcinoma at four cutoff points as a function of time. The aneuploidy score
represents a quantitative measure of image analysis ranging from 0–100.
The time interval represents the time between image analysis and clinical
remained ú 50% up to 3 months prior to the time of
clinical diagnosis. The laboratory test can be used with
noninvasively obtained urine samples as a screening
test for patients with a history of TCC to determine
when unplanned cystoscopies may be needed; it is
not intended as a substitute for adhering to clinically
accepted protocols for follow-up cystoscopies. It also
serves as a confirmation of cystoscopic observation
and may detect tumors not visible to the eye. The
addition of image analysis to cytology provides improved sensitivity over cytology alone. 15,16 Image alone
did not perform as well as the combination (sensitivity
was poorer) and therefore is not indicated for use
Sensitivity and specificity of the laboratory diagnosis
are comparable to that found in the literature. At the
simultaneous time interval and at the least stringent cutoff point, sensitivity was 80% whereas specificity was
41%. Other studies have reported values of sensitivity
from 76–93% and specificity from 41–94%. 13–16 The
poor specificity in this study may be related to the selection of patients with a history of TCC as the control
FIGURE 3. Sensitivity of the aneuploidy score for transitional cell carcinoma at four cutoff points as a function of time. The aneuploidy score
represents a quantitative measure of image analysis ranging from 0–100.
/ 7bc2$$1622
04-09-98 15:07:34
W: Cancer
Bladder Carcinoma and Image Analysis/Richman et al.
more conservatively. Treatment of invasive tumors, involving the removal of the bladder as well as surrounding organs including the prostate, drastically affects lifestyle and quality of life as well as survival rate.
FIGURE 5. Receiver operating characteristics (ROC) curve for the laboratory diagnosis (DX) for transitional cell carcinoma of the bladder (TCC) at
the simultaneous time period (0–1 day). Laboratory diagnosis represents
the combination of image analysis and visual cytology. The area under the
curve represents the strength of the test as a predictor of TCC recurrence.
Implications for Use of DNA Image Analysis with Visual
Positive DNA image analysis and visual cytology results must be confirmed by other diagnostic procedures. There may be a need to examine areas not accessible by a cystoscope, to investigate the possibility
of CIS not visible to the eye, or to consider the possibility of a false-positive test result. False-positive results
may lead to unnecessary and expensive investigational
procedures as well as undue anxiety. Although a falsepositive result is not very likely with high cutoff values
for a positive diagnosis (high cutoff values often are
used in a clinical setting for this reason), a false-negative result is comparatively more likely because sensitivity is not as high as one would like. Therefore, physicians should not be satisfied with following high risk
patients by this test alone.
The advantages for this kind of diagnostic test may
be substantial. Procedures involved in the laboratory
test require only a urine specimen, pose no risk to the
patient, are noninvasive, and are relatively inexpensive
compared with cystoscopy and other invasive procedures. Because bladder carcinoma patients are monitored for the rest of their lives, practicality is especially
important. If such a test leads to identification of recurrences up to 3 months before they would otherwise
be detected, then patients can be treated earlier and
/ 7bc2$$1622
04-09-98 15:07:34
Instrumentation and Measurement
Flow cytometry is an alternative method to image cytometry; however, image cytometry provides distinct
advantages over flow cytometry. Image cytometry 1)
detects abnormal DNA content occurring in extremely
low frequencies because individual cells are analyzed
rather than a population of cells, 2) requires fewer cells
and can be performed on smaller samples, 3) allows
visual control of the samples, which is helpful in eliminating artifacts, 4) provides a permanent record of the
cells, and 5) the same set of cells can be used for visual
cytology. 16,27,37 – 39
There are a variety of ways to measure and interpret DNA content such as those based on mean and
standard deviation of DNA content and measures expressing the shift from the normal diploid state.40 The
method used in this study was an index measuring the
deviation from the normal 2c state.32 The advantage
of using an index formulation is that a single number
with a continuous range expresses all DNA abnormalities. Difficulties may arise in discriminating aneuploidy from conditions leading to increased DNA replication. A less common alternative is to measure the
percentage of aneuploidy with ú 4c nuclei.
Methods of data collection for image cytometry
studies vary according to the type of instrumentation
and type of staining utilized.14,15,27,29,30,37,41,42 The Leitz
automated image analysis system used in this study
provides more reliable results but allows less control
than the operator-dependent models. Feulgen, used
in this study, and acridine orange are the two most
common stains for image analysis, each possessing
different benefits.14,29,30,42
Other biomarkers have been used to detect bladder carcinoma recurrence such as DNA 5CER (another
DNA ploidy marker), DD-23, M-344, Lewis X, epidermal growth factor receptor (EGFr), and F-actin.14,43 – 47
Another new biomarker based on monoclonal antibodies, the bladder tumor associated antigen (BTA)
test, is performed on urine specimens with a dipstick.48
Rapidly occurring technologic improvements and
identification of new markers are taking place.
Comments on the Data and Additional Areas for
This study provides a diverse geographic distribution
of 175 patients from across the U. S. and a wide range
of clinical practice protocols (50 urologists). Many
other studies of DNA image analysis have had patient
W: Cancer
CANCER May 1, 1998 / Volume 82 / Number 9
populations from a single geographic location and
usually were part of a hospital system.13,15,16
Unique to this study was the time element over
which the effectiveness of the test was evaluated. A
cross-sectional design commonly is used to investigate
sensitivity and specificity. Although some studies used
a follow-up design, few had repeated measures of DNA
image analysis at different time intervals. A retrospective cohort design with repeated measures allowed examination of the period of time in advance of clinical
evidence for which the test has predictive ability; it
indicates to clinicians how far in advance of clinical
symptoms a recurrent tumor may be detected.
Although the current study design provides useful
information, a comparison of DNA image analysis plus
visual cytology with visual cytology alone, rather than
with image alone, would have been more informative
regarding the benefits of adding the new procedure,
image analysis, to the old procedure, visual cytology.
The comparison performed showed that use of image
alone is not indicated.
Limitations were placed on the quality of the data
by response rates (52% for the first mailing) and data
collection methods. Based on the information in the
laboratory data base, there is no reason to believe that
patients of nonresponding physicians differed from
those included in the study. Chart review, performed
by the investigators for 34% of the patients, was not
feasible for the entire study due to the diversity of
geographic locations of the patients. Comparison of
patient characteristics collected by chart abstraction
and mailing (completed by physicians or their staff )
showed minimal differences between groups.
Biopsies were not available for all cases because
this was an observational study. Although it theoretically is possible that our gold standard may have been
incorrect in up to 25% of cases in which positive diagnoses were made without biopsies, this is not likely
because the physicians had cystoscopic findings, other
clinical evidence, and patient histories. Because the
gold standard is assumed to be correct, any variation
or error in the gold standard would only serve to decrease sensitivity and specificity; the actual performance of the test would have been underestimated.35
Other possible limitations in the data, such as inclusion of nonbladder TCC cases and treatment effects
on laboratory results, were investigated and determined to be inconsequential based on the small number of patients for which these issues were relevant.
Of the 60 patients whose data were collected by chart
review, there were only 4 patients (6.6%) who had TCC
of sites other than the bladder, 3 patients (5%) with
partial cystectomies, and none with radical cystectomies. This distribution is expected to be representative
/ 7bc2$$1622
04-09-98 15:07:34
of the whole sample. Concerns regarding the limited
ability of conventional urine cytology to reliably differentiate cell changes due to immunotherapy, chemotherapy, and radiation therapy and those due to carcinoma were addressed previously.36 Information regarding treatment within the last 30 days was collected
on test requisition forms. A random sample of 50 test
requisition forms from the larger laboratory database
showed that only 1 of 50 patients was treated with
such a therapy. Therefore, it was estimated that the
number of patients in the study group having treatment near and at the time of the test was very small.
There are few studies on the value of DNA image
analysis for early detection of TCC. Image analysis is a
relatively new technique. Of those studies performed,
none considered the variable of time in the ability of
the test to detect TCC. Larger and prospective studies
are needed to combine a time variable with stage,
grade, and pattern of growth to determine under what
conditions DNA image analysis is most informative.
Based on the results of the current study, we reached
the following conclusions:
1) The combined laboratory diagnosis performed better than the AS alone.
2) Sensitivity decreased as the time interval from the
laboratory tests to the clinical diagnosis increased.
When any detectable cytologic abnormality was
considered positive, the sensitivity of the combined
tests remained ú 50% up to 3 months prior to the
time of clinical diagnosis. Therefore, the combined
tests may be able to detect a recurrent bladder tumor up to 3 months in advance of clinical diagnosis.
3) A high sensitivity (80.4%) was achieved using the
combined tests under certain conditions (any detectable cytologic abnormality considered positive,
simultaneous test and diagnosis); however, specificity was low (41.1%) under the same conditions.
Keller JW, McCune CS, Sahasrabhe DM. Urologic and male
genital cancers. In: Rubin P. Clinical oncology: a multidisciplinary approach for physicians and students. 7th edition.
Philadelphia: W.B. Saunders Co, 1993:425–32.
Schapers RFM, Ploem-Zaaijer JJ, Pauwels RPE, Smeets AW,
van den Brandt PA, Tanke HJ, et al. Image cytometric DNA
analysis in transitional cell carcinoma of the bladder. Cancer
Lipponen PK, Eskelinen MJ, Jauhiainen K, Harju E, Terho
R, Haapasalo H. Prediction of superficial bladder cancer by
nuclear image analysis. Eur J Cancer 1992;29A(1):61–5.
Al-Abadi H, Nagel R. DNA content and survival rates of patients with TCC of the bladder. J Urol 1994;151:37–42.
W: Cancer
Bladder Carcinoma and Image Analysis/Richman et al.
van Velthoven R, Petein M, Oosterlinck WJ, Roels H, Pasteels
JL, Schulman C, et al. The use of digital image analysis of
chromatin texture in feulgen-stained nuclei to predict recurrence of low grade superficial transitional cell carcinoma of
the bladder. Cancer 1995;75:560–8.
Borchers H, Planz B, Jakse G, Bocking A. DNA aneuploidy
in G1-urothelial carcinomas of the urinary bladder. Urol Int
Yang Q, Xia Y, Wang Z, Wang G, Ding S, Shi S, et al. Morphometric diagnosis of bladder tumor. Oncology 1991;48:188–
de Meester U, Young IT, Lindeman J, van der Linden, HC.
Towards a quantitative grading of bladder tumors. Cytometry 1991;12:602–13.
Colombel M, De Laundit Y, Bellot J, Kiss R, Abbou C, Chopin
D. Prognostic evaluation of morphonuclear parameters in
superficial and invasive bladder cancer. Br J Urol 1995;
van Velthoven R, Petein M, Zlotta A, Oosterlinck W, van der
Meijden A, Zandona C, et al. Computer-assisted chromatin
texture characterization of Feulgen-stained nuclei in a series
of 331 transitional bladder cell carcinomas. J Pathol 1994;
Hemstreet GP III, Rollins S, Jones P, Rao JY, Hurst RE,
Bonner RB, et al. Identification of a high risk subgroup of
grade 1 transitional cell carcinoma using image analysis
based deoxyribonucleic acid ploidy analysis of tumor tissue.
J Urol 1991;146:1525–9.
van der Poel HG, van Caubergh RD, Boon ME, Debruyne
FMJ, Schalken JA. Karyometry in recurrent superficial transitional cell tumors of the bladder. Urol Res 1992;20:375–81.
Sherman AB, Koss LG, Wyschogrod D, Melder KH, Eppich
EM, Bales CE. Bladder cancer diagnosis by computer image
analysis of cells in the sediment of voided urine using a video
scanning system. Anal Quant Cytol Histol 1986;8(3):177–86.
Parry WL, Hemstreet GP III. Cancer detection by quantitative fluorescence image analysis. J Urol 1988;139:270–4.
Gerber WL, Lenahen PJ, Kendall AR, Mercer WE. Computerassisted microfluorometric detection of individual malignant bladder cells. Urology 1991;38(5):466–72.
Amberson JB, Laino JP. Image cytometric deoxyribonucleic
acid analysis of urine specimens as an adjunct to visual
cytology in the detection of urothelial cell carcinoma. J Urol
Koss LG, Sherman AB. Image analysis of cells in the sediment of voided urine. Monogr Clin Cytol 1984;9:148–62.
Koss LG, Eppich KM, Melder KH, Wersto R. DNA cytophotometry of voided urine sediment. Anal Quant Cytol Histol
Badalament RA, Gay H, Whitmore WF Jr., Herr HW, Fair WR,
Oettgen HF, et al. Monitoring intravesical bacillus CalmetteGuérin treatment of superficial bladder carcinoma by serial
flow cytometry. Cancer 1986;58:2751–7.
Bretton PR, Herr HW, Kimmel M, Fair WR, Whitmore WF
Jr., Melamed MR. Flow cytometry as a predictor of response
and progression in patients with superficial bladder cancer
treated with bacillus Calmette-Guérin. J Urol 1989;141:
DeVere White RW, Deitch AD, West B, Fitzpatrick JM. The
predictive value of flow cytometic information in the clinical
management of stage 0 (Ta) bladder cancer. J Urol 1988;139:
Giella JG, Ring K, Olsson CA, Karp FS, Benson MC. The
predictive value of flow cytometry and urinary cytology in
/ 7bc2$$1622
04-09-98 15:07:34
the follow-up of patients with transitional cell carcinoma of
the bladder. J Urol 1992;148:293–6.
Norming ULF, Nyman CR, Tribukait B. Comparative histopathology and deoxyribonucleic acid flow cytometry of random mucosal biopsies in untreated bladder carcinoma. J
Urol 1991;145:1164–8.
Norming ULF, Tribukait B, Nyman CR, Nilsson BO, Wang
N. Prognostic significance of mucosal aneuploidy in stage
Ta/T1 grade 3 carcinoma of the bladder. J Urol 1992;
Badalament RA, Hermansen DK, Kimmel M, Gay H, Herr
HW, Fair WR, et al. The sensitivity of bladder wash flow
cytometry, bladder wash cytology, and voided cytology in
the detection of bladder carcinoma. Cancer 1987;60:
Badalament RA, Kimmel M, Gay H, Cibas ES, Whitmore WF
Jr., Herr HW, et al. The sensitivity of flow cytometry compared with conventional cytology in the detection of superficial bladder carcinoma. Cancer 1987;59:2078–85.
Koss LG, Wersto RP, Simmons DA, Deitch D, Herz F, Freed
SZ. Predictive value of DNA measurement in bladder washings: comparison of flow cytometry, image cytophotometry,
and cytology in patients with past history of urothelial tumors. Cancer 1989;64:916–24.
Chabanas A, Rambeaud JJ, Seigneurin D, Fradet Y, Lawrence
JJ, Faure G. Flow and image cytometry for DNA analysis in
bladder washings: improved concordance by using internal
reference for flow. Cytometry 1993;14:943–50.
Kiss R, Salmon I, Camby I, Gras S, Pasteels JL. Characterization of factors in routine laboratory protocols that significantly influence the feulgen reaction. J Histochem Cytochem
Gurley AM, Hidvegi DF, Bacus JW, Bacus SS. Comparison
of the Papanicolaou and Feulgen staining methods for DNA
quantification by image analysis. Cytometry 1990;11:468–74.
Masson E, Herlin P, Bloyet D, Duigou F, Mandard AM. Development of an automatic program for the stereological
estimation of the volume weighted mean volume of cell nuclei. Pathol Res Pract 1992;188:438–42.
Bocking A, Chatelain R, Homge M, Daniel R, Gillissen A,
Wohltmann D. Representativity and reproducibility of DNA
malignancy grading in different carcinomas. Anal Quant Cytol Histol 1989;11:81–6.
Dean AG, Dean JA, Burton AH, Dicker RC. EpiInfo, Version
5: a word-processing, database, and statistics program for
epidemiology on microcomputers. Stone Mountain, GA:
USD, Incorporated, 1990.
SAS Institute, Inc. SAS/STAT. Version 6. Cary, NC: SAS Institute, Inc., 1990.
Greenberg RA, Jekel JF. Some problems in the determination
of the false positive and false negative rates of tuberculin
tests. Am Rev Respir Dis 1969;100:645–50.
Bane BL, Rao JY, Hemstreet GP. Pathology and staging of
bladder cancer. Semin Oncol 1996;23:546–70.
Ploem JS, van Driel-Kulker AM, Ploem-Zaaijer JJ. Automated
cell analysis for DNA studies of large cell populations using
the LEYTAS image cytometry system. Pathol Res Pract
Hemstreet GP, Bonner RB, Hurst RE, O’Dowd GA. Cytology
of bladder cancer. In: Vogelzang NJ, Scardino PT, Shipley
WU, Coffey DS, editors. Comprehensive textbook of genitourinary oncology. Baltimore: Williams & Wilkins, 1996:338–
W: Cancer
CANCER May 1, 1998 / Volume 82 / Number 9
39. National Research Council. New technologies. Biologic
markers in urinary toxicology. Washington, DC: National
Academy Press, 1995:200–2.
40. Montironi R, Scarpelli M, Sisti S, Ansuini G, Pisani E, Mariuzzi G. Prognostic value of computerized DNA analysis in
noninvasive papillary carcinomas of the urinary bladder. Tumori 1987;73:567–74.
41. Tanke HJ, Brussee JA, van Driel-Kulker AM, van der Burg MJ,
Schelvis-Knepfle CF, Jonas U, et al. Results of the automate
analysis of 328 bladder specimens using the Leyden Television Analysis System (LEYTAS). World J Urol 1983;1:77–81.
42. Hemstreet GP III, West SS, Weems WL, Echols CK, McFarland S, Lewin J, et al. Quantitative fluorescence measurements of AO-stained normal and malignant bladder cells.
Int J Cancer 1983;31:577–85.
43. Bonner RB, Liebert M, Hurst RE, Grossman HB, Bane BL,
Hemstreet GP III, Marker Network for Bladder Cancer. Characterization of the DD23 tumor-associated antigen for blad-
/ 7bc2$$1622
04-09-98 15:07:34
der cancer detection and recurrence monitoring. Cancer
Epidemiol Biomarkers Prev 1996;5:971–8.
Fradet Y, Islam N, Boucher L, Parent-Vaugeois C, Tardif M.
Polymorphic expression of a chuman superficial bladder tumor antigen defined by mouse monoclonal antibodies. Proc
Natl Acad Sci USA 1987;84:7227–31.
Sheinfeld J, Reuter VE, Melamed MR, Fair WR, Morse M,
Sogani PC, et al. Enhanced bladder cancer detection with the
Lewis X antigen as a marker of neoplastic transformation. J
Urol 1990;143:285–8.
Neal DE, Sharples L, Smith K, Fennelly J, Hall RR, Harris AL.
The epidermal growth factor receptor and the prognosis of
bladder cancer. Cancer 1990;65:1619–25.
Rao JY, Hemstreet GP III, Hurst RE, Bonner RB, Min KW,
Jones PL. Cellular F-actin levels as a marker for cellular
transformation: correlation with bladder cancer risk. Cancer
Res 1991;51:2762–7.
Bard Diagnostic Sciences. Available from URL: http:// [accessed August 12, 1997].
W: Cancer
Без категории
Размер файла
153 Кб
Пожаловаться на содержимое документа