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

Background. Identifying a marker enabling prediction of recurrence in the group of superficial transitional cell carcinomas (sTCCs) of the bladder remains an important challenge today. This report quantitatively describes chromatin patterns with respect to such sTCC recurrence. Materials and Methods. Twenty-nine patients with sTCCs who did not exhibit tumor recurrence within a minimum of 24 months were compared with 21 patients with sTCCs who exhibited tumor recurrence two or three times in a 24-month period

As stated by Rubben et al.,' more than 60% of all bladder tumors are superficial and classified as Stages Ta-T1 and Grades 1-3 in accordance with the recommendation of the International Union Against Cancer.' These authors also report that even after the complete transurethral resection of all visible lesions, 50-70% of superficial bladder tumors recur, and up to 25% progress in stage and/or grade. Likewise, de Vere White et aL3 report that patients with superficial transitional cell carcinomas (sTCCs) of the bladder present two major clinical problems: (1) identifying those who will experience recurrent disease; and 2) recognizing those whose disease will progress. These authors3 add that the risk of progression is approximately 9% in patients with Stage 0 (Ta) lesions, whereas it increases to almost 43% for those with Grade 3 Stage A (Tl) lesions.
For nearly 30 years,4 the determination of nuclear DNA content (DNA ploidy level) has been used as a marker of bladder tumor aggressiveness. A large number of reports have shown that this DNA ploidy level is related significantly to histologic grade, and/or recur-rence, and/or prognosis in ~u p e r f i c i a l~,~-~~ and invatransitional cell carcinomas (TCCs) of the bladder. All these studies were performed by either flow or image cytometry, both of which provide similar results. lS Other research groups have focused their attention on the predictive value contributed by the quantitative description of morphometric and/or kilryometric nuclear features (cell nucleus size, shape, geometry, etc.). Blomjous et al.," Helander et al.,I7 Ooms et ah," and Lipponen et al.19 show that these parameters also offer significant information (in addition to that contributed by DNA ploidy level determination) for characterizing the aggressiveness of TCCs of the bladder and for predicting the recurrence of these sTCCs. In addition to DNA ploidy level determination and morphometric cell nucleus characterization, as emphasized by Van der Poel et a1.,20 a third category of parameters exists involving those parameters that quantitatively describe the cell nucleus chromatin patternz1 We already have demonstrated the usefulness of such computer-assisted microscopic assessments of chromatin patterns in Feulgen-stained nuclei with respect to characterizing bladder tumor aggressiveness.22 To our knowledge, this third type of parameter has not been used yet to predict tumor recurrence of bladder sTCCs.
The aim of the present work is a quantitative analysis of chromatin pattern and its relationship to recurrence of sTCCs of the bladder. For this purpose, 29 patients with sTCCs who did not exhibit tumor recurrence within a minimum of 24 months were compared with 21 patients with sTCCs who exhibited tumor recurrence two or three times during a 24-month period, for a total of 74 sTCCs. Quantitative chromatin pattern description was performed by digital cell image analysis of Feulgen-stained nuclei as previously described.22-2s Six morphonuclear parameters were described and used subsequently to determine a score, allowing the prediction of the biologic behavior of sTCCs, i.e., recurrence versus nonrecurrence, to occur in one calculation step. DNA ploidy also was determined in each sTCC in this study by DNA histogram sive12-14

Clinical Data and Histopathologic Diagnosis
We examined transurethral resection material from 50 patients (study period, 1987-1993). The histopathologic evaluation was performed on H & E-stained sections. All 50 patients were untreated, and of these, 29 had sTCCs that did not recur during a 24-67-month observation period; these patients belonged to the pri- We, thus, studied 74 sTCCs, the grade and the stage of which are depicted in Table 1.
Histologic grade and clinical stage were determined without knowing the results of the cell image analysis.
Histopathologic grading was performed according to the World Health Organization classification for urinary bladder tumors.28 Clinical staging was performed according to the TNM classification system.2

Cytologic Samples
The present study is prospective; however, because many centers are involved, each of our samples was buff ered-formalin fixed and paraffin embedded before digital cell image analysis was performed. This enabled us to work under standard conditions in fixing and processing the samples.25 Thus, the current series of 74 sTCCs of the bladder was obtained from archived material, i.e., formalin fixed paraffin embedded tissues. One to three paraffin blocks were available for each case according to tumor size, and five sections were cut from each block. The first, third, and fifth sections (5-pm thickness) underwent histopathologic grading and clinical staging (after H & E staining). The second and fourth sections (80-pm thickness) underwent a method recently described in detail." This method makes possible single-cell nuclei suspensions (after pronase digestion) that are centrifuged onto glass slides and then stained by the Feulgen reaction according to the procedure described p r e v i o~s l y .~~-~~

Quantitative Chromatin Texture Description
The quantitative description of the chromatin texture was performed by a SAMBA 200 microscope image processor (Alcatel-TITN, Grenoble, France) with a lOOX magnification lens (numerical aperture: 1.30). After a rapid low resolution analysis of the images to memorize the nuclear address at the motorized stage, each nucleus (200-300 nuclei selected per slide, i.e., 600-1800 nuclei selected per case) was analyzed in a high resolution mode." The SAMBA 200 cell image processor made enabled the quantitative description of 15 morphonuclear We describe only 6 of these 15 parameters below and in the Results section because the remaining 9 either were correlated statistically to the 6 illustrated here, or were not statistically significant with respect to the two histoclinical tumor groups (primary vs. recurrent) (data not shown). These nine parameters are detailed e l~e w h e r e .~~,~~ Because the morphonuclear parameters describe the topographic distribution of the pixels within the nucleus, the nuclear size must be determined first before the chromatin pattern is described quantitatively. The nuclear size was assessed by means of the nuclear area (NA), which corresponds to the calculation of the nuclear profile, i.e., the number of pixels occupied by a nucleus. The topographic distribution of the pixels (1 pixel = 0.16 pm2) then was computed on 256 levels of optical density (OD) with respect to each pixel. The nuclear images of the Feulgen-stained nuclei were digitized at a 540-nm wavelength.
The nuclear images were evaluated visually and accepted by the observers on the same basis as those previously d e~c r i b e d ,~~-~~ i.e., the nuclei corresponded to well preserved cells. Segmentation was satisfactory when the nuclear membrane was intact, and neither staining artifact nor an overlap between neighboring nuclei appeared in the samples selected.
In addition to the NA, the five parameters used for the quantitative description of the chromatin texture in terms of pixel distribution per nucleus (the mathematical formulae are those described by Gal10way~~ and Haralick et aL3') are listed below: the skewness index (SK) represents the asymmetry in the distribution of densitometric values in the nuclei. It assesses the proportion of pale and dense pixels in the nucleus the kurtosis index (K) represents the half-height width of the OD value distribution histogram in the nucleus. This parameter measures the homogeneity level of the optical density value distribution in the nucleus the variance of OD (VOD) is a parameter that also provides a quantitative description of the heterogeneity level of the optical density value distribution in the nucleus. The K values are high when a number of optical density value distribution peaks appear in one and the same nucleus. Conversely, the VOD values-even for a low K value-can become high if the distribution of the optical density values in the nucleus is totally anarchic The primary group includes the superficial transitional cell carcinomas (sTCCs), which did not recur over a minimum period of 24 months, and the recurrent group includes those recurring during this period.
t One urimarv and two recurrent tumors were uentauloid.
the short run length emphasis (SRL) is representative of the frequency of small dense chromatin clumps in the nucleus contrast (C) measures the number of boundaries between the nuclear regions of different extinction values; its value decreases more when the nucleus contains a small number of large, dense chromatin clumps.
A score was determined based on the six morphonuclear parameters described above. The rationale behind and the method for determining this score is explained below.

D N A Ploidy Level Determination
As indicated above, the nuclear integrated optical density (IOD) was computed for each pixel (0.16 m) on 256 densitometric levels. The nuclear DNA content of each nucleus thus was calculated from its IOD parameter value, which corresponded to the sum of the optical density (OD) values of each pixel of the nucleus.
The DNA histogram types were assessed from the IOD parameter, which relates to nuclear DNA content, and was calculated based on spectrophotometry and because the OD is defined as a function of the transmission values measuring the amount of absorbant material. Thus, the IOD assessment performed for each case enabled further determination of the DNA histogram types. The absolute IOD value assessed for each sample was normalized according to the procedure described previo~sly.~' This normalization enabled the 2C DNA content to equate to 2000 k 100 arbitrary units of IOD.31 The DNA histogram typing also was performed as was described for tumors of the nervous system.26r27 The different DNA histogram types defined in Table 2 were diploid (type I), hyperdiploid (type IV), triploid (type 11), Figure 1. Development of one geometric parameter, i.e., the nuclear area ( Fig. 1 A), three densitometric parameters, i.e., the Skewness index ( Fig. lB), the Variance of Optical Density (Fig. IC), and the Kurtosis index ( Fig. lD), one parameter issuing from the run length matrix, i.e., the frequency of the short-run length (Fig. lE), and one parameter issuing from the cooccurrence matrix, i.e., contrast (Fig. lF), in contrast to the clinical subgroups. The results are presented as mean values (black squares) f their standard errors (hatched rectangles) and k their standard deviations (open rectangles). There were 50 patients, of whom 29 had superficial transitional cell carcinomas (sTCCs) which did not recur over a minimum observation period of 24 months and a maximum observation period of 67 months (PRIM group). The remaining 21 patients had tumors which recurred once to three times over a period ranging from 3 to 24 months after their first diagnosis (REC group). There were 74 sTCCs in the study. CLINICAL SUB-GROUPS hypertriploid (type V), tetraploid (type 111), and polymorphic (type VI) patterns.

Determining a Score
Because this work attempted to identify those sTCCs that will and will not recur, a score was determined based on the six morphonuclear parameters described above and their development values assessed in the primary group. A subscore value of 1-5 was attributed to each of these parameters. Thus, the values of the score range from 6 to 30.
The subscore values were attributed in two different ways according to whether the mean parameter values increased or decreased from the primary to the recurrent group. Thus, as indicated on the right-hand side of Figures 1A (NA parameter), 1D (K parameter), and 1E (SRL parameter), subscore value 1 was attributed in the primary group to all the morphonuclear parameters whose values were less than the mean minus the standard error of the mean (SEM) value. Subscore value 2 was attributed in the primary group to all the morphonuclear parameters whose values ranged from the mean minus SEM to the mean plus SEM values. Subscore value 3 was attributed in the primary group to all the morphonuclear parameters whose values ranged from the mean plus SEM to the mean plus 5 X SEM values. Subscore value 4 was attributed in the primary group to all the morphonuclear parameters whose values ranged from the mean plus 5 X SEM and the mean plus 10SEM values. Lastly, subscore value 5 was attributed in the primary group to all the morphonuclear parameters whose values were greater than the mean plus IOSEM values, as is indicated in Figures lA, lD, and 1E. The same subscore values were attributed to the SK (Fig.  lB), VOD (Fig. lC), and C ( Fig. 1F) parameters, but in an inverse manner to that described in Figures lA, lD, and 1E.

Statistical Analyses
The results are given as the mean k SEM and k standard deviation. Statistical comparison of the data was performed using the Fisher's F test (one-way analysis of variance) after a check was performed on the equality of variances by the Bartlett's test, and on the normal distribution fit of the data by the chi-square test. The statistical levels of significance in the breakdown of the score values in contrast to various histopathoclinical tumor groups were assessed by means of the chi-square test.
All the statistical analyses were performed using the Statistica/Dos software package (Statsoft, Tulsa, OK).

DNA Ploidy Level Determination
In Table 2, the breakdown of the six DNA histogram types versus the two histoclinical tumor groups was similar. The level of statistical significance assessed for such a breakdown determined by the chi-square test was P > 0.05. This means, for example, that the primary group, which includes the sTCCs that did not recur over a minimum of 24 months, did not exhibit a significantly lower proportion of aneuploid (hyperdiploid + triploid + hypertriploid + polymorphic) cases than the recurrent group, which includes the sTCCs that recurred within this 24-month period.

Chromatin Texture Characterization
The data obtained for six morphonuclear parameters, i.e., one (NA) related to nuclear size and five (SK, VOD, K, SRL, C) to chromatin pattern, are described below. Two histoclinical groups, the primary and the recurrent groups, were studied. Figure 1A shows that the mean NA value was significantly lower (P < 0.001) in the primary group than in the recurrent one. The mean SK value was significantly higher (P < 0.001) in the primary than in the recurrent group (Fig. 1B). The same feature was observed with respect to the VOD ( P < 0.001, Fig. 1C) and C ( P < 0.001, Fig. 1F) parameters.
The mean K (P < 0,001, Fig. 1D) and SRL (P < 0.001, Fig. 1E) values were markedly lower in the primary than in the recurrent group. Figure 2 illustrates the development of the score values across the two tumor subgroups (primary versus recurrent). The iterative chi-square test revealed that a cutoff value separating the score values from those < 14 to those > 14 symbolized by the dark horizontal line in Figure 2, enabled discrimination (P < 0.001) between the sTCCs that had not recurred over 24 months (score < 14) and those that had recurred (score > 14). Of the 29 primary tumors, 21 had a score value < 14 and 8 > 14, whereas of the 45 recurrent tumors, 6 had a score value < 4 and 39 > 14.

Results Relating to the Score
In Figure 2, we linked each tumor to its recurrence with either a hatched or a black line. The hatched lines represent recurrent sTCCs with score values lower than (or equal to) the primary (or recurrent) tumors from which they stemmed. Likewise, the black lines in Figure  2 represent the recurrences with score values higher than those in the original sTCCs. We established this link to investigate whether the score value actually described biologic sTCC aggressiveness quantitatively. For this purpose, we studied the biologic behavior of three sTCCs (Group A in Fig. 2), each of which recurred twice, the first in Group B and next in Group C of Figure  2. Patient P1 had three tumors whose score values were maximal. This patient's tumor developed from a Grade 1 pathologically classified pT1 (Group A) through a Grade 1 pTa (Group B) to a Grade 2 pTa (Group C). Thus, this patient had a primary tumor with a high ini-tial score value, and this tumor had already recurred twice. As compared with Patient 2, it reasonably could be assured that this patient experience a new sTCC recurrence in the near future. Patient P2 already had had a recurrent sTCC when this investigation began (in Group A). At that time, this tumor was classified as Grade 1 pTa, and its score value was 24. This score value decreased to 16 in Group B, and the tumor remained Grade 1 pTa. In Group C, the score value decreased to 13, while the tumor remained Grade 1 pTa. In fact, the tumor as an sTCC could not have a better prognosis than Grade 1 pTa, which it retained throughout its biologic life, though this did not prevent it from recurring three times. As of this writing, the clinicians described Patient P2 as being in a state of clinical remission, reflected by a constant decrease in the score values of the successive recurrences. The developmental profile of patient P3 was the opposite of that of Patient P2. The tumor in Group A was Grade l pTa with a score value equal to 15; the score value increased to 18 in Group B and to 27 in Group C. In Group B, the tumor was Grade 2 pTa and in Group C, Grade 2 pT1. The development of the score values of Patient PD clearly reflects clinical progression in the same way as does the remission of Patient P2.
The biologic development of these three tumors into their subsequent two recurrences may suggest that histologic grade or clinical stage plays a dominant role in this type of biologic development and this may also depend either wholly or partly on DNA ploidy level (see Discussion). In other words, it may be thought that the information contributed by the score values is similar to that contributed by the combination of histologic grade, clinical stage, and DNA ploidy level, but this is not the case as is discussed below. Figure 3 represents a synthesis of this study's results. The predictive value (with respect to tumor recurrence) of DNA ploidy level and score value when determined individually and together is illustrated. In Figure  3, the empty areas represent the cases in which DNA ploidy level determination matched clinical reality (Lee, correctly predicted recurrence or nonrecurrence), whereas the hatched areas represent the cases in which DNA ploidy level determination did not match clinical reality. Likewise, the numbers preceded by an arrow represent the cases in which the score value determination did not match clinical reality, whereas the numbers in circles represent the cases in which the score value determination matched clinical reality. If a comparison is made with respect to potential recurrence between the low risk diploid pTa/pTl Grade 1 tumors (low-risk according to the criteria in the literature as mentioned in the Discussion) and those associated with a higher risk (i.e., the aneuploid pTa/pTl Grade 1 tumors), of PRIMARY RECURRENT TUMORS TUMORS Figure 3. A summary of the information contributed by the score value (as illustrated in Fig. 2) as compared with that contributed by DNA ploidy level determination, i.e., diploid versus aneuploid, in superficial Grade 1 pTa/pTl transitional cell carcinomas of the bladder. The empty areas represent the cases where DNA ploidy level determination matched clinical reality, i.e., correctly predicted recurrence or nonrecurrence, whereas the hatched areas represent the cases for which DNA ploidy level determination did not match clinical reality. Similarly, the numbers preceded by an arrow represent the cases where the score value determination did not match clinical reality, whereas the numbers in circles represent the cases for which the score value determination matched clinical reality.
the 32 patients represented in Figure 3, the score value made it possible to accounting for 16 (50%) who did not meet the traditional criteria given in the literature (the hatched areas in Fig. 3). The present score is redundant in relation to the traditional data given in the literature for 9/32 patients (28%). Consequently, 7/32 patients (22%) were not included in the individual predictive prognostic value of the score.

Discussion
Obtaining a marker predicting the recurrence rate in the group of superficial transitional cell carcinomas (sTCCs) of the bladder still remains an important challenge. The predictive value of numerous parameters has been investigated. These parameters can be classified into roughly five categories relating to: (1) histological grade (1, 2, and 3) and clinical stage (pTa, pT1); (2) immunohistochemical features; (3) the quantitative description of cell nucleus morphometry; (4) cell-cycle kinetics; and (5) DNA ploidy level. The aim of this study was to show that a sixth category of parameters, involving parameters quantitatively describing chromatin patterns, can be used to predict recurrence of sTCCs of the bladder.
Jordan et al.32show that the biologic potential (ag-gressive vs. nonaggressive) of sTCCs can be predicted from histologic grading and clinical staging. Consequently, the probability of recurrence is much higher in Grade 3 pT1 sTCCs than in those that are Grade 1 pTa, and this view is shared by other^.^,'^ Two examples of the parameters relating to histochemical features are reported below. The first concerns lectin histochemistry, for which Gabius et al.33 show that the chemical conjugation of carbohydrate derivatives is expressed differentially in normal bladder tissue versus the various sTCC grades and stages. Likewise, Yamazaki et al.,34 which describes squamous cell carcinoma-associated antigen, shows that the quantitative characterization of the squamous cell carcinoma-associated antigen can predict the clinical behavior of Grade 3 pT1 sTCC diseases. Indeed, their results show that the nonprogression rate of Grade 3 pT1 carcinomas with positive squamous cell carcinoma antigen expression in the cytoplasm was significantly lower than those with negative antigen e x p r e~s i o n .~~ Concerning the third category of parameters, a large number of studies have shown that the quantitative morphometric characterization of the cell nucleus can be helpful in determining clinical aggressiveness, i.e., recurrence, in sTCCs of the bladder.16-21,35 In particular, Lipponen et al.19 show that bladder sTCCs can be efficiently categorized into prognostic groups by means of nuclear image analyses, because tumors with high nuclear factor values should be considered for radical primary therapy and adjuvant therapy after transurethral resections.
Regarding the fourth category of parameters, which involve cell cycle kinetics, i.e., determining the percentage of cells in the S phase of the cell cycle or the growth fraction, several authors have shown that this type of determination, performed by either immunohistochem-iCa136,37 or flow ~y t o m e t r i c~,~, '~ methods, also can be helpful in determining the aggressiveness of sTCC of the bladder.
Lastly, of the five above-mentioned parameter categories, the fifth, relating to DNA ploidy level determination, has been investigated the most, for approximately 30 years.4 The purpose of this study is not to undertake an extensive review of the literature on the predictive value of DNA ploidy level determination with respect to tumor recurrence of sTCCs of the bladder but, rather, to show how the quantitative description of chromatin texture in Feulgen-stained nuclei can aid DNA ploidy level determination; we report a few examples below to illustrate our point.
In their study of 229 patients with Grade 1-11 and Stage pTa/pTl bladder tumors, Gustafsson et al.5 reported that no progressive cases were found among 175 patients with repeatedly diploid DNA patterns and that tumor progression was linked exclusively to an aneuploid DNA pattern. Similarly, in their study of sTCCs of the bladder, de Vere White et aL3 showed that all the patients with aneuploid histograms experienced recurrent disease as predicted. They conclude that for sTCC disease, DNA histograms seem to be able to provide prognostic information beyond that obtained from tumor grade and stage.
In fact, too rapid a systematization favoring the view that most, if not all, Grade 1 tumors of the bladder are DNA diploid and that most Grade 3 tumors are DNA aneuploid must be avoided.14 Wheeless et al.14 claimed that in patients with a history of superficial bladder cancer treated by transurethral resection, a clear, nontetraploid DNA aneuploid histogram is diagnostic for recurrent tumors. However, as our results show, this is far from being a universal truth. Our data clearly show that primary diploid sTCCs of the bladder can recur, and that some of the primary aneuploid bladder sTCCs under study did not recur over an observation period of 24-60 months. These results, therefore, conflict with the studies of Gustafson et al.5 and de Vere White et al.,3 but are in accordance with those of Fradet et al.,9 who state that ploidy is not an entirely reliable prognostic indicator because in their study, a significant proportion of nonprogressing pTa and pT1 tumors were aneuploid, whereas the samples were near diploid in 6/20 cases of cancer progression. Nevertheless, DNA ploidy level measurement represents a helpful marker in stratifying Grade 2 sTCCs of the bladder, which constitute a heterogeneous group with approximately equal proportions of DNA diploid and DNA aneuploid tum o r~.~~ The present data on the quantitative description of chromatin patterns in Feulgen-stained nuclei from sTCCs of the bladder show clearly that this description contributes significant additional information to that contributed by histologic grading, clinical staging, and DNA ploidy level measurement relating to tumor recurrence predictivity. All the monovariate statistical analyses (Fisher's F test) reported in Figures 1-4 illustrate the development of one individual morphonuclear parameter through the histoclinical subgroups under study. These analyses show clearly that the sTCCs in the primary group differed markedly from the recurrent ones in terms of morphonuclear characteristics. The tumors in the primary group, which included superficial bladder transitional cell carcinomas that did not recur over a minimum observation period of 24 months, had smaller nuclear sizes than the tumors that recurred once, twice, or even three times during this minimum 24-month observation period. These primary tumors also differed markedly from the recurrent ones based on their chromatin texture characteristics. The chromatin texture in the sTCC cell nuclei from the recurrent tumors contained larger pale areas, shown by decreasing SK parameter values, than in those of the sTCC cell nuclei in the primary group. However, this global decrease in chromatin condensation was accompanied by an increasing frequency of the small dense chromatin clumps (as assessed by the SRL parameter) and the large dense chromatin clumps (data not shown) in the recurrent group as compared with those in the primary one. This duality between the concomitant increase in the number of pale areas and of small and large chromatin clumps leads to an increasing heterogeneity in the pale chromatin area distribution in the nucleus (as revealed by the increase in the K parameter values) and to an overall decreasing chromatin heterogeneity distribution (as revealed by decreasing VOD parameter values) from the primary tumor group to the recurrent one. Furthermore, it should be emphasized that the frequency of the small (SRL parameter) and large (LRL parameter) dense chromatin clumps increased in such a marked manner in the recurrent as compared with the primary group that they joined together, resulting in a decrease in the absolute number of well delineated chromatin clumps in the nucleus and, thus, to a decrease in the C parameter values; as an inverse function, this decrease measures the number of boundaries between the nuclear regions of the different extinction values or, in other terms, the number, whether large or small, of dense chromatin clumps in the nucleus.
The additional information contributed by quantitative chromatin pattern description to that contributed by DNA ploidy level determination is clear when one compares the diploid Grade 1 pTa/pTl sTCCs, which are normally considered low risk, i.e., potentially nonrecurrent, to the aneuploid Grade 1 pTa/p'I'l sTCCs, which are considered high risk, i.e., potentially recurrent (see Fig. 3). Apparently in Figure 3 in accordance with the data in the literature involving histological grade, clinical stage and DNA ploidy level determination, only the biologic behavior (recurrence versus nonrecurrence) of 13/32 patients (41%) was predicted correctly, whereas this biologic behavior was predicted correctly in 25/32 patients (78%) when the score value alone was considered. In combining the data concerning histological grade, clinical stage, and DNA ploidy level determination with that concerning the score value, we were able to predict the biologic behavior of 29/32 patients (91%) correctly.
In conclusion, the present results show that the quantitative description of chromatin pattern by performing the digital cell image analyses of Feulgenstained nuclei can contribute helpful information in predicting tumor recurrence of low grade, sTCCs of the bladder. For this purpose, six morphonuclear parame-ters (relating to nuclear size, chromatin condensation level, and its heterogeneity distribution) were used to determined a score, with values ranging from 6 to 30. The information contributed by the score values was additional to that contributed by DNA ploidy level determination. Of 32 patients with Grade 1 pTa/pTl tumors, DNA ploidy level determination enabled the accurate prediction of tumor nonrecurrence or recurrence of 13 patients (41%), whereas score value determination enabled such a prediction for 25 patients (78%). Combining the information contributed by DNA ploidy level and the score values, it was possible to predict recurrence or nonrecurrence of 29/32 (91%) patients.
Since February 1992, we obtained data on 291 patients with sTCCs for the purpose of a prospective study. The 50 patients whose results we report here either had relapses within the 25-month study period (21 patients) or have not shown signs of a relapse as of this writing (29 patients). We are now in the process of completing the data illustrated in Figure 3 on the remaining 241 patients. If reference is made to the data in the literature, we will need approximately 3 more years to obtain all the data and to acquire a sufficient degree of clinical follow-up on each patient.