SEARCH

SEARCH BY CITATION

Keywords:

  • urine;
  • bladder;
  • kidney;
  • urothelial carcinoma;
  • genitourinary

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

BACKGROUND:

Urine cytology represents a major portion of testing volume in many cytopathology laboratories.

METHODS:

The authors previously reported a template designed to standardize urothelial diagnostic categories to enable clinicians to uniformly manage their patients. In this study, they examined the common cytomorphologic features observed in specimens diagnosed with atypical urothelial cells, cannot exclude high-grade urothelial carcinoma (AUC-H), which prove most predictive of high-grade urothelial carcinoma (HGUC).

RESULTS:

The most common morphologic features observed in the AUC-H specimens were hyperchromasia, irregular nuclear borders, increased nucleus-to-cytoplasm ratio, and anisonucleosis. Of the 58 patients who had specimens diagnosed with AUC-H, 95% ultimately were diagnosed with HGUC on follow-up biopsy over the study period. The small number of patients who had AUC-H with non-HGUC follow-up did not allow for a statistical comparison to determine the predictive ability of the selected criteria for HGUC. Next, the authors used the same features to examine a subset of urine samples that were diagnosed with atypical urothelial cells of unknown significance (AUC-US) in an attempt to improve the predictive value of this clinically frustrating category. A blind review was performed of 290 urine specimens from 217 patients. In contrast to the AUC-H specimen cohort, the majority of specimens with AUC-US did not contain atypical cells with the 4 common morphologic features. All 4 features significantly predicted HGUC in surveillance patients, but not in patients with hematuria.

CONCLUSIONS:

Hyperchromasia was the strongest predictor of HGUC by far in patients who were undergoing surveillance (odds ratio, 9.81). Hyperchromasia remained statistically significant in multivariate analysis, indicating its predictive strength even in the absence of other features. Cancer (Cancer Cytopathol) 2013;121:21–28 © 2012 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

Urine cytology represents a major portion of testing volume in many cytopathology laboratories.1-3 We previously reported a template (Table 1) that was designed to standardize our urinary tract diagnostic categories to enable our clinicians to uniformly manage their patients.4 After several years of using the template, we believed it would be prudent to examine its performance. Our diagnostic categories of high-grade urothelial carcinoma (HGUC) and atypical urothelial cells, cannot exclude HGUC (AUC-H) included almost equal percentages of patients (5% and 6%, respectively). The diagnostic category atypical urothelial cells of uncertain significance (AUC-US) included many more patients (26%) than we believed was acceptable for an indeterminate category.

Table 1. The Johns Hopkins Template for Reporting Urinary Cytology Specimens
Negative for urothelial atypia or malignancy (NUAM)
Urothelial carcinoma (UC)
 High grade (HGUC)
 Low grade (LGUC)
Atypical urothelial cells (AUC)
 Of undetermined significance (AUC-US)
 Cannot exclude HGUC (AUC-H)
Other (specify)

Our strategy was first to examine the common cytomorphologic features observed in the category AUC-H that proved most predictive of HGUC. These features were then used to examine a subset of AUC-US specimens as a training exercise in an attempt to improve the predictive value of this clinically frustrating category.5, 6 The objective of this study was to define those criteria that will not only detect patients with HGUC but also to define a separate category of cell changes that, although not normal, are not indicative of neoplasia and appropriately belong in an indeterminate, not suspicious category.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

Construction of the Database

The hospital laboratory information system was searched for cytology specimens that were diagnosed as AUC-H and AUC-US from July 1, 2007 to June 30, 2009. Of the patients who were diagnosed with atypical cells (AUC-US or AUC-H) on urine cytology between July 2007 and June 2009, 58 who were diagnosed with AUC-H by our faculty underwent at least 1 follow-up surgical biopsy. Eight patients had 2 urine specimens diagnosed as AUC-H. In total, 62 specimens were available for review.

In total, 607 patients with 1246 specimens were identified with a diagnosis of AUC-US. Of these specimens, 290 specimens from 217 patients were used in this study as a training set, and the remaining specimens were saved as a future test set. Specimens that did not have follow-up pathology (surgical biopsy or urine specimen) were not included. The specimens were classified by clinical indication (158, surveillance for neoplasia; 118, hematuria; 14, other indications), which was obtained by abstracting clinical histories.

Review of Specimens

A junior and senior pathologist (C.J.V. and D.L.R.), who were masked to clinical indication and outcome, separately evaluated each of the AUC-H specimens and determined 8 cytologic features that appeared to be predictive of HGUC. The presence of these features in each sample was scored separately and reconciled during a joint review. After unmasking of this data set, a similar masked review of AUC-US specimens was conducted using the 4 criteria that were identified as most predictive of HGUC in the AUC-H samples. The predictions were then matched with the follow-up biopsy or clinical outcomes, which were tracked over the 36 months after the July 2009 cutoff for inclusion in the study.

Statistical Methods

We constructed 2 × 2 tables to calculate sensitivity, specificity, positive predictive value, and negative predictive value; and chi-square tests were used to evaluate statistical significance. Univariate and multivariate analyses were performed to evaluate key morphologic features that predicted an outcome of HGUC on follow-up. For all statistical analyses, 2-sided tests were used, and P values ≤ .05 were considered statistically significant. The statistical software package SAS 9.2 for Windows (SAS Institute Inc., Cary, NC) was used for all statistical analyses.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

Patient Demographics

Of the 58 patients in the AUC-H cohort, 45 were men, and 13 were women (3:1 ratio) (Table 2). Patients ranged in age at first diagnosis from 48 to 90 years among men and from 39 to 83 years among women. Later decades predominated; however, the range among women included primarily the fourth and fifth decades, whereas men who had the diagnosis generally spanned the later decades. Of 217 patients in the AUC-US cohort, 160 were men, and 57 were women (2.83:1 ratio) (Table 3). Older age was associated with a higher risk of HGUC on follow-up, and the trend increased from 12% of patients in their fifth decade to 31% of patients in their eighth decade.

Table 2. Outcome According to Sex in 58 Patients Who Had a Diagnosis of Atypical Urothelial Cells of Undetermined Significance
 Outcome: No. of Patients (%)
SexHGUCNon-HGUC
  1. Abbreviations: HGUC, high-grade urothelial carcinoma.

Women13 (100)0 (0)
Men42 (93)3 (7)
Table 3. Outcome According to Age and Sex in 217 Patients Who Had a Diagnosis of Atypical Urothelial Cells, Cannot Exclude High-Grade Urothelial Carcinoma
 Outcome: No. of Patients (%)
Clinical CharacteristicHGUCNon-HGUC
  1. Abbreviations: HGUC, high-grade urothelial carcinoma.

Sex  
 Women11 (23)46 (77)
 Men25 (14)135 (86)
Age, y  
 20-290 (0)1 (100)
 30-390 (0)3 (100)
 40-490 (0)16 (100)
 50-593 (12)22 (88)
 60-699 (15)50 (85)
 70-7913 (18)59 (82)
 80-8911 (31)25 (69)
 90-990 (0)5 (100)

Morphologic Features Observed in Specimens With Atypical Urothelial Cells, Cannot Exclude High-Grade Urothelial Carcinoma

Of 725 patients who were diagnosed with atypical cells (AUC-US or AUC-H) on urine cytology between July 2007 and June 2009 who had a follow-up biopsy, 58 were diagnosed by our faculty as having had at least 1 sample with AUC-H, and 8 patients had 2 such samples. In total, 62 specimens were available for review. A junior pathologist and a senior pathologist were masked to patient history and outcomes and separately evaluated each specimen for individual cytologic criteria. Eight cytologic criteria were selected that were expected from experience to be the most predictive of HGUC (Table 4). Based only on the cytologic features, each reviewer predicted whether the sample represented a patient with HGUC. The predictions were then matched with the follow-up biopsy outcomes, which were tracked over the 36 months after the July 2009 cutoff for inclusion in the study.

Table 4. Cytologic Features Recorded for Each Sample Diagnosed With Atypical Urothelial Cells, Cannot Exclude High-Grade Urothelial Carcinoma According to Frequency Observed
Cytologic FeatureFrequency Observed: No. of Samples (%), n = 62
Individual abnormal cells46 (74)
Hyperchromatic nuclei44 (71)
Irregular nuclear borders41 (66)
Increased nuclear-to-cytoplasm ratio35 (56)
Anisonucleosis34 (55)
Elongated nuclei24 (39)
Cell clusters23 (37)
BK polyomavirus change11 (18)

The most common morphologic features identified in AUC-H specimens were individual atypical cells, hyperchromasia, and irregular nuclear borders followed by increased nucleus-to-cytoplasm ratio and anisonucleosis (Figure 1). BK polyomavirus changes were uncommon. No single criterion was observed in all samples. Hyperchromasia was marked, often obscuring nuclear detail. Irregular nuclear borders were observed in cells that had polygonal nuclei with flat, rather than round, edges. Anisonucleosis when present was striking, with variation between atypical cells often exceeding a 3:1 ratio. Although cells with atypical features often were observed as individual cells, larger fragments with the same atypical features could be identified in some samples.

thumbnail image

Figure 1. These photomicrographs of a urine specimen that was diagnosed as atypical urothelial cells, cannot exclude urothelial carcinoma (AUC-H) demonstrate the 4 most commonly observed features: hyperchromasia, irregular nuclear borders, increased nuclear-to-cytoplasm ratio, and anisonucleosis. Anisonucleosis is particularly striking in this specimen, in which many atypical cells vary at a greater than 3:1 ratio. Atypical cells are hyperchromatic, resulting in the dual population of nuclei commonly observed in specimens with AUC-H. A subsequent biopsy revealed high-grade urothelial carcinoma. The patient died 1 year later from metastatic disease (original magnification: top left, ×100; all others, ×400).

Download figure to PowerPoint

Outcome of Patients With a Diagnosis of Atypical Urothelial Cells, Cannot Exclude High-Grade Urothelial Carcinoma

This study was blinded to clinical history and outcome, and the outcomes were not revealed until a slide review was completed. Of the 58 patients who were diagnosed with AUC-H, 95% ultimately were diagnosed with HGUC on follow-up biopsy over the study period (Table 5). The small number (n = 3) of AUC-H patients without HGUC on follow-up did not allow for a statistical comparison of how predictive the selected criteria were for HGUC. These 3 patients did have low-grade papillary carcinoma; therefore, all 58 patients who were diagnosed with AUC-H had biopsy-proven urothelial cancer. Furthermore, among the women in this group, all had a high-grade lesion identified on biopsy. Most of the patients with AUC-H who were diagnosed subsequently with HGUC were in the surveillance group, whereas only 7 presented with hematuria. Two patients in the surveillance group had a tumor diagnosis of low-grade urothelial carcinoma after a cytologic diagnosis of AUC-H.

Table 5. Distribution of Patients With Atypical Urothelial Cells, Cannot Exclude High-Grade Urothelial Carcinoma According to Clinical Indication, Sex, and Outcome
 Outcome: No. of Patients (%)  
Clinical IndicationHGUCNon-HGUCSex: M:F RatioTotal No.
  1. Abbreviations: HGUC, high-grade urothelial carcinoma; M:F, male:female.

Surveillance47 (96)2 (4)3.949
Hematuria8 (89)1 (11)2.09
Total55 (95)3 (5)3.558

It is noteworthy that these data demonstrate that our AUC-H category has much higher predictability for HGUC than reported in our original template study.4 This current study includes an additional 18 months of follow-up (for a total of 3 to 5 years depending on sample collection), during which 16 additional patients were diagnosed with HGUC. 63% (39 of 62) of AUC-H patients had HGUC on follow-up in the original study compared with 95% (55 of 58) in this study. Unfortunately, this unexpected increase in HGUC found in patients with an AUC-H diagnosis does not allow for a statistical comparison between the 2 groups to determine how well the morphological features found in the AUC-H specimens predict HGUC. We therefore turned our attention to applying these morphologic findings to the AUC-US category.

Review and Analysis of Specimens With Atypical Urothelial Cells of Undetermined Significance

It is known that a finding of AUC-US is much less predictive of HGUC on follow-up4; therefore both outcome groups were likely to have enough patients for a proper statistical analysis. We performed a blind review of 290 urine specimens from 217 patients. In contrast to the AUC-H specimen cohort, the majority of AUC-US specimens did not contain atypical cells with the 4 common morphological features. BK virus changes were observed more commonly in AUC-US specimens than in AUC-H specimens (32% vs 18%, respectively).

After slide scoring was completed, patient outcomes were revealed. Overall, 22% of patients in the surveillance group had HGUC on follow-up compared with 13% of patients in the hematuria group (Table 6). Statistical analysis was performed to determine the association of each morphologic feature with patient outcome (Tables 7 and 8). All features were significantly capable of predicting HGUC in the surveillance group but not in the hematuria group (Figure 2).

thumbnail image

Figure 2. These are photomicrographs of a urine specimen that was diagnosed as atypical urothelial cells of unknown significance (AUC-US), although the specimen demonstrates the 4 most commonly observed features in specimens with atypical urothelial cells, cannot exclude high-grade urothelial carcinoma (AUC-H): hyperchromasia, irregular nuclear borders, increased nuclear-to-cytoplasm ratio, and anisonucleosis. Although these features are also observed often in individual atypical cells, a fragment of atypical cells is identified in this specimen. The atypical cells also possess the 4 most common features. This patient was diagnosed with invasive high-grade urothelial carcinoma 1 month later from a cytoscopic biopsy (original magnification: top left, ×100; all others, ×400).

Download figure to PowerPoint

Table 6. Distribution of Patients With Atypical Urothelial Cells of Undetermined Significance According to Clinical Indication, Sex, and Outcome
 Outcome: No. of Patients (%)  
Clinical IndicationHGUCNon-HGUCSex: M:F RatioTotal No.
  1. Abbreviations: HGUC, high-grade urothelial carcinoma; M:F, male:female.

Surveillance23 (22)81 (78)3.3104
Hematuria13 (13)88 (87)2.5101
Unknown0 (0)12 (100)2.012
Total36 (18)181 (82)2.8217
Table 7. Statistical Analysis of the Cytomorphologic Features Used to Predict High-Grade Urothelial Carcinoma in Specimens With Atypical Urothelial Cells of Undetermined Significance
IndicationFeatureSensitivity, %Specificity, %NPV, %PPV, %Pa
  • Abbreviations: N/C ratio, nuclear-to-cytoplasm ratio; NPV, negative predictive value; PPV, positive predictive value.

  • a

    P values were determined with the chi-square test.

HematuriaN/C ratio68.7540.0088.8915.49.5
 Hyperchromasia68.7550.5090.9118.03.16
 Irregular borders62.5048.0088.8916.13.43
 Anisonucleosis37.5072.0087.8017.65.43
 All features combined31.2581.1988.1720.83.38
SurveillanceN/C ratio86.6744.5393.4426.80.0016
 Hyperchromasia86.6760.1695.0633.77<.0001
 Irregular borders76.6761.7291.8631.94.0001
 Anisonucleosis63.3377.3490.0039.58<.0001
 All features combined50.0084.3887.8042.86<.0001
Table 8. Odds Ratios and P Values of the Cytomorphologic Features Used to Predict High-Grade Urothelial Carcinoma in Specimens With Atypical Urothelial Cells of Undetermined Significance
  Univariate AnalysisMultivariate Analysis
Clinical IndicationFeatureORPORP
  • Abbreviations: NA, not applicable; N/C ratio, nuclear-to-cytoplasm ratio; OR, odds ratio.

  • a

    P values < .05.

HematuriaN/C ratio1.47.511.10.88
 Hyperchromasia2.20.1711.59.12
 Irregular borders1.54.440.17.22
 Anisonucleosis1.54.440.87.83
 All features combined1.96.20NANA
SurveillanceN/C ratio5.22.004a3.02.074
 Hyperchromasia9.81<.0001a8.44.035a
 Irregular borders5.30.0004a0.54.457
 Anisonucleosis5.90<.0001a1.91.222
 All features combined5.40.0001NANA

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

The most common features observed in AUC-H specimens were individual abnormal cells with hyperchromasia, irregular nuclear borders, anisonucleosis, and increased nucleus-to-cytoplasm ratio, features that are shared with malignant cells from HGUC. However, in specimens that are diagnosed as HGUC, the cells often are larger and possess coarsely granular chromatin and/or large nucleoli.7-9 In AUC-H specimens, the atypical cells are often small, with marked hyperchromasia that obscures nuclear detail. Less common features are elongated nuclei and the presence of cell clusters. Because elongated nuclei and cell clusters are observed more often in samples from low-grade neoplasms, these features would not necessarily be expected to predict HGUC. Although the BK virus can mimic HGUC (decoy cells),7 BK viral changes were observed more commonly in AUC-US specimens than in AUC-H specimens.

Previous studies have demonstrated similar findings in atypical, yet not definitively malignant, urine specimens. Potts et al examined 153 urine specimens for 32 cytomorphologic criteria and observed that anisonucleosis and a high nuclear-to-cytoplasm ratio were important in separating HGUC from benign lesions.10 Deshpande and McKee observed that hyperchromasia and nuclear irregularities were associated with biopsy-proven low-grade urothelial carcinoma or HGUC; in their study, the nuclear-to-cytoplasm ratio was not associated significantly with outcome.11 In a descriptive study, Renshaw outlined several categories of atypical urine specimens and noted the presence of increased nuclear-to-cytoplasm ratio, hyperchromasia, and irregular nuclear membranes in these specimens.12 In another study, Mokhtar et al examined 72 voided urine specimens and observed nuclear membrane irregularities and hyperchromasia more frequently in patients with a (low-grade or high-grade) malignant follow-up biopsy.13 In the current study, we focused on the prediction of HGUC; the diagnosis of low-grade urothelial carcinoma in cytology specimens remains a challenge.14-17

The current study was performed with investigators masked to clinical indication and outcome. The follow-up period of the original template study was extended by 18 months, and HGUC was discovered in additional patients with AUC-H over that period. Although this false false-positive phenomenon in urine cytology is well known, the follow-up period required to prove a false-positive urine specimen is not well established. We did not expect such a high rate of biopsy-proven HGUC in patients who had AUC-H urine specimens. Because almost all patients had HGUC on follow-up, there was not enough statistical power to evaluate the specificity of the morphologic features that we identified in AUC-H specimens.

The most commonly identified morphologic criteria were applied to the larger cohort of patients with AUC-US to determine their utility in predicting HGUC. Our analysis indicates that the 4 features examined (hyperchromasia, irregular nuclear borders, anisonucleosis, and increased nuclear-to-cytoplasm ratio) were statistically significant in predicting HGUC on follow-up, but only for the patients who underwent surveillance (Tables 7 and 8). The data do not support the application of these criteria in patients with hematuria alone and may suggest that these features can be observed in stone atypia. It is noteworthy that this result differs from that reported by Deshpande et al, who observed that nuclear membrane irregularities and hyperchromasia retained a significant association with outcome when patients who had a previous history of HGUC were excluded.11 This may be because those authors considered both low-grade and high-grade urothelial carcinoma in their outcome analysis. In our study, hyperchromasia was by far the strongest predictor of HGUC for patients in the surveillance group, with an odds ratio of 9.81. Furthermore, hyperchromasia remained statistically significant in multivariate analysis, indicating its predictive strength even in the absence of other features (Table 8).

The statistical performance of the criteria is excellent given the relatively poor performance of the AUC-US category as a whole. For instance, in this subset of patients, the positive predictive value of AUC-US for HGUC is 19% and 13.8% for the surveillance and hematuria groups, respectively (Table 6). When all 4 morphologic criteria are used to predict HGUC in these specimens, the positive predictive value is 42.9% and 20.8% for the surveillance and hematuria groups, respectively (Table 7). Therefore, these criteria allow the classification of urine specimens into a higher risk category. A specimen that meets all 4 criteria is more likely to represent HGUC than a specimen that meets ≤3 criteria. However, requiring all 4 criteria sacrifices sensitivity at the expense of specificity (Table 7).

These results likely under represent the amount of true HGUC that exists in these patients. For instance, patients who are diagnosed with AUC-US on urine cytology are less likely to undergo follow-up biopsy; and HGUC, if present, may take longer to discover. This is partially evidenced by our finding that patients with AUC-H had an increased rate of HGUC when they were followed for an additional 18 months (Figure 3). Furthermore, specimen categorization into AUC-H often is biased by patient history, and there is a lower threshold to diagnose AUC-H in patients who have a known history of HGUC.

thumbnail image

Figure 3. This photomicrograph shows a urine specimen that was diagnosed with atypical urothelial cells, cannot exclude high-grade urothelial carcinoma (AUC-H) and demonstrates single cells with hyperchromatic nuclei, irregular nuclear borders, and an increased nuclear-to cytoplasmic ratio. The cell nuclei often assume a more polygonal (rather than round) shape. Although nuclear size does not differ dramatically among the abnormal cells, the nuclei are markedly hyperchromatic, and nuclear detail is observed. The follow-up bladder biopsies in 2009 were negative. Further workup that year revealed the presence of high-grade urothelial carcinoma in the right kidney, leading to a nephroureterectomy. Three years later, the patient had bladder dysplasia with high-grade urothelial carcinoma on urine cytology (original magnification, ×400).

Download figure to PowerPoint

Patients who have AUC-US specimens have a much lower risk of a subsequent HGUC diagnosis on follow-up (46 of 290 specimens; 16%) than patients who have AUC-H specimens (55 of 62 specimens; 89%). Further analysis is in process to determine which other factors were present that directed clinicians to biopsy patients in the AUC-US category or the negative for urothelial atypia or malignancy category. Our ultimate objective is to reduce the AUC-US category to a meaningful size for appropriate patient management.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

We acknowledge our urologic surgeons for their enthusiastic support of our efforts. In addition, we are grateful to our cytotechnologists and laboratory technicians for their excellent work in the preparation and screening of samples.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. FUNDING SOURCES
  9. REFERENCES