Transitional cell cancer (TCC) of the bladder is known to be a frequently recurrent, partly progressing tumor type.1 Once diagnosed and treated by transurethral resection, bladder tumor patients have to be followed for many years.2 Frequent follow-up appointments negatively affect the quality of life of these patients.
Among the methods used in follow-up of bladder cancer, urinary cytology has been well established. Advantages and limits of the method in the detection of tumor recurrences have been described previously.3–5 The diagnostic utility of voided urine cytology was compared with that of bladder washing and remains controversial.4, 6 Yet, it was never taken into account that the time interval between delivery of voided urine and diagnostic result is an important factor in the acceptance of this method. Quick-staining cytology is a technique predominantly used in intrasurgical diagnosis. Its application to voided urine produced shortly before cystoscopy was thought to provide immediate information vital to urologists. Therefore, one objective of this study was to test and compare the accuracy of quick-staining cytology of voided urine specimens.
Morphometry of both urine and bladder washing specimens played a major role in previous attempts to surpass the limitations and to enhance the diagnostic efficiency of cytologic material.7–11 In general, morphometry added objectivity to reports otherwise highly dependent on diagnostic experience.3 Image cytometry on Feulgen-stained exfoliative material from the bladder introduced the biometric parameter ploidy. The possibility of selective nuclear measurements made static cytometry superior to flow cytometry, especially in scanty and heterogeneous material also containing nonepithelial cells.12 However, the lack of a well distinguishing threshold system hampers the wider application of DNA image cytometry in bladder cancer.13
Morphometric examination of Feulgen-stained specimens may include the determination of geometric and texture parameters. Depending on the material under investigation, previous studies gave some of these geometric and texture parameters a diagnostic value comparable to that of integrated optical density.14, 15 Taking part of these findings into consideration, the linked use of one parameter describing the DNA histogram and one parameter describing a nuclear geometric feature led to a system with an automated classification of the bladder washings under investigation.16 Consequently, the second objective of the study was to elucidate the diagnostic and clinical value of this special image analysis alone and in combination with quick-staining urinary cytology.
MATERIALS AND METHODS
- Top of page
- MATERIALS AND METHODS
The study consisted of 223 follow-up specimens collected from 124 patients between February 1996 and March 1998. All patients had a past history of transitional cell carcinoma of the bladder with at least one transurethral tumor resection before entrance into the study. Patients' ages ranged from 18 to 96 years, with a median of 73 years. Fourteen patients had received local chemotherapy, additionally. None of the patients received radiotherapy because of bladder cancer. Patients with already known metastatic disease were excluded from the study. Control interval of patients with more than one follow-up appointment ranged from 3 months to 1 year.
Follow-up specimens included voided urine, concurrent bladder washings, and, in cases of suspicious cystoscopy results, transurethral biopsies or tumor resections.
Voided urine produced shortly before cystoscopy was spun down, and further management followed a previously described protocol.17 Cytospin preparations were done after sedimentation. The work-up procedure resulted in cells on two glass slides. One glass slide was air dried by gentle warming and stained within 40 seconds with Hemacolor (Merck, Darmstadt, Germany). Cytologic examination followed immediately and specimens were classified in the following previously established diagnostic groups: negative for atypia and/or malignancy, atypia regardless of grade, suggestive of malignancy, and positive for malignancy.3, 13, 17 Specimens without any epithlial cells were considered insufficient for detection of bladder cancer. Signs of inflammation were graded separately. All cytologic results were available before cystoscopy. The second glass slide of each voided urine specimen was fixed by spray before drying and underwent Papanicoulaou staining. Cytologic evaluation of Papanicoulaou-stained specimens was done independently on the next day, reviewed by a second pathologist (SM), and followed the same classification scheme.
Bladder washing specimens were gathered during cystoscopy and processed within 2 hours. Two of the cytospin preparations were wet fixed, and one of these entered a Feulgen staining protocol.15, 16 The second glass slide entered routine cytology procedure. Cytologic description was available for all specimens but was not used in this study. Static image analysis was done on a dual parameter system with an integrated automatic threshold. Using a magnification of ×400, nuclei of 50 microscopic fields are imported via a camera, analyzed by the software (Gentian Scientific Software, the Netherlands), checked visually again, then finally computed; working procedure followed in principle a previously published protocol.16 The instrument is calibrated with lymphocytes or nuclei of squamous cells, intermediate cell type, as internal reference. By the calculation of configuration the software discharges the majority of overlapping nuclei and granulocytes. Internal reference cells are selected and additional nondiagnostic nuclei are dismissed by supplementary obligatory visual review. Reference cells' coefficient of variation is limited to 5%, automatically.18 In case of a coefficient of variation beyond that level, the software stops processing the sample, and the measurement or review has to be repeated. Primarily, histograms are classified by the DNA parameters 2c-deviation-index (2cDI) as defined by Böcking et al.19 and the 5c-exceeding-rate. Additionally, nuclei were classified by a shape factor based on smoothed differences in Freeman chain code and describing the degree of ellipse-shape (PASS16). The above-specified software gives a threshold based on the combination of the DNA parameter 2cDI and the morphometric shape parameter and results in the calculation of a low-, intermediate-, or high-risk factor for bladder cancer (Fig. 1). Each measurement aims at computing at least 200 cells. Within this study specimens contained 50 to 630 diagnostic nuclei of epithelial cells. Specimens with fewer than 100 epithelial nuclei counted were considered insufficient.
Figure 1. Threshold image analysis giving a risk factor for bladder cancer (graphic modified from the original report sheet). 5cER: 5c-exceeding-rate; PPR:.
Download figure to PowerPoint
For statistical evaluation, voided urine specimens with the cytologic report of suspicion or evidence of malignancy and bladder wash specimens with high-risk factors for the presence of bladder cancer were defined as positive. Evaluation of atypia in cytology and intermediate risk in image cytometry entered the negative group. Patients were classified as having recurrent disease if suspicious cystoscopy results or clinical judgments were followed by the histopathological confirmation of TCC. The histopathological report included tumor grading and staging according to the International Union Against Cancer (UICC) classification.20 Patients were considered free of recurrence if there was either a concurrent negative cystoscopy or, in cases of suspicion resulting from cystoscopy findings, histological exclusion of cancer and further negative clinical follow-up for at least 6 months. For estimation of the value of both methods, either alone or in combination, statistical analysis included calculation of sensitivity, specificity, and predictive value of a negative and a positive result.
- Top of page
- MATERIALS AND METHODS
During the given observation period, 223 follow-up appointments of 124 patients with a past history of TCC included the availability of voided urine produced shortly before cystoscopy and concurrent bladder wash. Eighty-three recurrences were found in those patients and proven by histology.
The distribution of grade and stage of recurrent tumors as well as the sensitivity of quick-staining urinary cytology and threshold image analysis of bladder washing specimens are given in Table 1. The majority of recurrent TCC was reported low grade by histopathology, 61 showing grade 1 or 2 TCC and low stage, showing stage pTa in 54 cases and pT1 in 19 cases. Within this study, carcinoma in situ (CIS) was found only supplementary to a papillary transitional cell carcinoma G2pTa. In this case cytology was positive and threshold image analysis reported high risk.
Table 1. Sensitivities of Quick-Staining Cytology of Voided Urine and Image Cytometry with Integrated Automatic Risk Classification of Bladder Washing Specimens and Their Correction by the Number of Specimens Insufficient for Bladder Cancer Detection
| || ||Voided urine cytology||Bladder wash risk factor image cytometry|
|Sensitivity||Corrected sensitivity (n insufficient specimens)||Sensitivity||Corrected sensitivity (n insufficient specimens)|
|TCC G1||(n = 22)||13.6||13.6 (0)||22.7||27.8 (4)|
|TCC G2||(n = 39)||46.2||51.4 (4)||58.9||69.7 (6)|
|TCC G3||(n = 22)||86.4||86.4 (0)||81.8||90.0 (2)|
|TCC pTa||(n = 54)||37.0||39.2 (3)||46.3||55.6 (9)|
|TCC pT1||(n = 19)||57.9||61.1 (1)||63.2||70.6 (2)|
|TCC pT ≥ 2||(n = 10)||90.0||90.0 (0)||90.0||100.0 (1)|
|Total||(n = 83)||48.1||50.6 (4)||55.4||64.8 (12)|
The overall sensitivity of cytology was 48.1% and reflects the high number of recurrent low grade transitional cell carcinomas under investigation. Depending on the grade of the underlying tumor, the sensitivity of quick-staining cytology was 86.4%, 46.2%, and 13.6% for grade 3 to grade 1 TCC, respectively. All reports stating negativity for atypia and/or malignancy by quick-staining and those stating positivity for malignancy were reproduced in the corresponding Papanicolaou-stained specimen. Three specimens primarily diagnosed as suggestive of malignancy were found to be positive for malignancy in the Papanicolaou-stained cytospin by both cytopathologists. For 11 voided urine specimens, atypia diagnosed in the quick-stained material could not be reproduced in Papanicolaou-stained material.
Image analytical work-up of bladder washing specimens revealed higher sensitivity for grade 1 and 2 tumors. Compared with the sensitivity of cytology, that of image analysis was lower in grade 3 TCC (Table 1). Considering the amount of material insufficient for cytologic diagnosis and/or image analysis sensitivities were recalculated and are listed also in Table 1. Among those voided urine specimens before bladder cancer, 4 of 83 were insufficient for diagnoses, either because of massive leukocyturia or hematuria. When nuclei of the 83 corresponding bladder washing specimens were computed, 12 failed to result in estimable histograms. Either diagnostic nuclei were covered by a high amount of leukocytes or epithelial nuclei were scanty, or both. Data demonstrate that also for 2 of 22 bladder washing specimens of consecutively diagnosed G3 tumors, image analysis was not possible and thus decreased the sensitivity.
Additionally, sensitivities were calculated on the basis of results complementary positive in voided urine cytology and threshold image analysis (Table 2). The data revealed an increase of sensitivity in bladder tumor detection for all tumor grades and stages. Compared with cytology alone the complementary use of both methods increased sensitivity depending on the grade of the underlying tumor by 18.2%, 20.5%, and 4.5%, respectively. Depending on the histopathological tumor stage, the sensitivity increased for 18.6% in pTa, 10.5% in pT1, and 10% in pT2 TCC. Thus, compared with voided urine cytology, the overall sensitivity improved for 15.8%. Paralleled to image analysis, the ancillary use of both methods revealed improvement, too. The increase of sensitivity ranged, depending on the grade and stage of the later-resected bladder tumor, from 5.2 to 10.0%; the overall increase was 8.5%.
Table 2. Value of Quick-Staining Cytology of Voided Urine in Combination with Image Cytometric Risk Analysis
| ||Percentage (%) of cases with either positive cytology or high risk in ICM||Percentage (%) of cases with both positive cytology and high risk in ICM|
|TCC G1 (n = 22)||31.8||4.5|
|TCC G2 (n = 39)||66.7||38.5|
|TCC G3 (n = 22)||90.9||77.2|
|TCC pTa (n = 54)||55.6||27.8|
|TCC pT1 (n = 19)||68.4||52.6|
|TCC pT ≥ 2 (n = 10)||100.0||80.0|
|Total (n = 83)||63.9||39.6|
The calculation of both percent negativity in the absence of TCC and the percentage of negative results that were true negatives was based on 140 follow-up cases. Primarily, a patient's follow-up was considered negative when either concurrent cystoscopy was inconspicuous or suspicous cystoscopy was followed by histological exclusion of cancer and further negative clinical follow-up for at least 6 months. A patient's follow-up was considered positive when concurrent cystoscopy revealed a tumor later histologically proven to be TCC. These estimations revealed 100% specificity, 76.65% predictive value of a negative result, and 100% predictive value of a positive result for quick-staining urinary cytology. Risk factor image analysis was described with 82.9% specificity, 75.8% predictive value of a negative result, and 65.7% predictive value of a positive result. Three of 24 follow-ups (3 of 22 patients) judged negative by cystoscopy and high risk by image analysis developed TCC within 6 months (1 patient) and 14 months (2 patients). Thus, based on tumor recurrence in later follow-up, specificity for image analysis amounted to 85.0%. The predictive value of a high-risk statement in image analysis was 70%. Specificity data, including calculations on time dependence, are summarized in Table 3. Among those patients with a high-risk estimation by image analysis and concurrent negative clinical follow-up were another five (six follow-ups) with grade 3 TCC and another four (5 follow-ups) with already recurrent TCC, grade 1 or 2, stage pTa or pT1, in their past history.
Table 3. Specificity and Predictive Value of a Positive Result of Risk Factor Image Analysis of Bladder Washing Specimens Dependent on Follow-up Time
| ||Bladder wash risk factor image analysis|
|Specificity correlated with concurrent positive cystoscopy (%)||82.9|
|Specificity, including further follow-up (%)||85.0|
|Predictive value of a positive result with concurrent positive cystoscopy (%)||65.7|
|Predictive value of a positive result, including further follow-up (%)||70.0|
Eight follow-ups were characterized by negative cytology, low risk by image analysis, and positive cystoscopy. Visible tumors were resected and histology staged all as TCC stage pTa, seven of them grade 1, and one grade 2. In 11 cases negative histology and negative clinical follow-up succeeded cystoscopy results suggestive of cancer. For one of these cases the classification “high-risk for bladder cancer” was reported by image analysis; none had positive cytology. In other follow-up cases with cystoscopically visible tumor and atypia in quick-staining voided urine cytology slides, image analysis subdivided cases. Exactly half of these cases, 2 of 4 grade 1 and 5 of 10 grade 2 TCC, were judged high risk by image analysis.
- Top of page
- MATERIALS AND METHODS
Many efforts have been made to overcome the unpredictability of the biological behavior of TCC.21, 22 Still, routine cytology, cystoscopical evaluation, and biopsy are the predominant diagnostic tools in the follow-up of bladder cancer patients.21
Reports on the sensitivity and specificity of Papanicolaou urinary cytology have documented the abilities and limitations of cytology in detecting primary and recurrent TCC.3, 5, 21 Discussions on the materials led to partly controversial results,4, 6) but the value of a comprehensive cytologic description on-site has not yet been studied. For detection of malignancy, the data demonstrate a positive agreement between quick-staining and Papanicolaou cytology for those cases diagnosed explicit negative and positive in the quick-staining. Still, there was some overestimation. The finding of atypia could not be reproduced in Papanicolaou-stained specimens for 11 cases. Within the specimen group under investigation there was minor interobserver variation, mainly restricted to grading of atypia. Any kind of therapy reduces sensitivity in routine cytology.15 The same has to be said for quick-staining cytology comparing the present findings with previously published data on the cytologic detection of mainly primary bladder cancer.17 In contrast, the specificity was not influenced by any preceding therapy. Thus, the high specificity of quick-staining cytology in cancer detection, the high predictive value of a positive result, and its high degree of agreement with traditional staining methods emphasize the effectiveness. Additional information on inflammatory processes demonstrates its partial superiority to immunological markers.23 With minimal patients inconvenience, a waiting time of 20 to 30 minutes, the detailed information on actual inflammatory changes is available right before planned cystoscopy and thus may influence further follow-up procedure.
Cytology provides information about the actual bladder status, but its reliability in predicting recurrence- or progression-free survival is low; and like all visual diagnostics the method is limited by a certain intraobserver and interobserver variation. Parameters determined by DNA image cytometry appear to be more valuable biomarkers and are mainly observer independent.11, 12 The predictive value of DNA measurements in bladder washings has been reported previously.10, 13 But the difficulties in comparison of results from different laboratories hamper its widespread application. To overcome interlaboratory variances, studies were performed on the influence of fixatives, different protocols of Feulgen staining, and image analysis systems differing widely in their set-up.24–26 Additionally, there is little agreement on the interpretation of DNA histograms. A generally accepted interpretation of the DNA histogram giving organ- or tumor-specific (or both) thresholds for low and high risk of recurrence or progression is still lacking.
A board of European pathologists summarized algorithms useful for diagnostic and prognostic interpretation.18 Among the algorithms suggested is the 2c deviation index (2cDI), also reported to be the most important independent predictor of survival in Ta, T1 TCC of the bladder.12 Including all nuclei measured into the calculation, the numerical classification of a DNA histogram by the 2cDI reflects also the need for special interpretation of epithelia with physiological polyploidization such as the transitional epithelium.27 The 2cDI and a shape descriptor are the basis of an image analytical software with an automated classification.16 Data evaluated in this study support and stress the previously published comparison with conventional bladder wash cytology.11 The tested system is user-friendly and appears to be very potent in selecting those patients with the need for more intensive care, even without a concurrent cystoscopically visible tumor. Patients with CIS or TCC grade 3, invasive beyond the lamina propria, or frequent recurrences of low-grade TCC in their past remain high risk by the tested image analysis system. These findings agree with previous reports on the aggressiveness of CIS4 and the relationship of multiple recurrences and prognosis.2
Additionally, image analysis was demonstrated to be helpful in subclassifying cases with suspicious cystoscopy results and negative cytology or cytologic findings of abnormal urothelial cells lacking decisive features of malignancy. Image analysis reported low risk for 14 of 83 TCCs. Still, the combinatory use of quick-staining cytology and image analysis with the presently integrated threshold missed 8 of 83 recurrent TCCs, all of them stage pTa, seven grade 1 and one grade 2. These findings might suggest minimal aggressiveness of the resected tumors and agree with the knowledge about the prognostic significance of stage and grade of the resected tumors.12 The comparison of data from this study with recently published data is limited by a different distribution of material and a slightly different grading system.13 Nevertheless, Mora et al.13 also reported 12 and 8 of 75 recurrent TCCs undetected by image analysis alone and cytology combined with image analysis, respectively. Findings in bladder washings of both studies differ from those in single-cell separations of biopsies, in which aneuploidy is found in up to 76.6% of separations when the Kolmogoroff-Smirnow test is used.7
In conclusion it can be stated that the combination of quick-staining cytology and image analysis with an automated risk classification lowers the limits of both methods and improves the morphological follow-up of patients with bladder cancer without increasing patient discomfort. This combination appears useful for possibly minimizing the number of follow-up appointments and thus may also improve the quality of life of a patient group increasing in number.