The accuracy of urinary cytology in daily practice

Authors

  • Sheldon Bastacky M.D.,

    Corresponding author
    1. Department of Pathology, University of Pittsburgh Medical Center (UPMC), Pittsburgh, Pennsylvania
    • Department of Pathology—PUH C622, University of Pittsburgh Medical Center, 200 Lothrop Street, Pittsburgh, PA 15213-2582
    Search for more papers by this author
  • Stacey Ibrahim M.D.,

    1. St. Joseph's Hospital (SJH), Asheville, North Carolina
    Search for more papers by this author
  • Sharon P. Wilczynski M.D.,

    1. Department of Pathology, City of Hope National Medical Center (COHNMC), Duarte, California
    Search for more papers by this author
  • William M. Murphy M.D.

    1. Department of Pathology and Laboratory Medicine, University of Florida College of Medicine (UFCM), Gainesville, Florida
    Search for more papers by this author

Abstract

BACKGROUND

Most studies of urinary cytology have been research analyses designed to test the method itself, and many claim that the high diagnostic yields in these studies cannot be achieved in daily practice. The authors examined the clinical and pathologic records in three hospital pathology practice settings—academic, community, and cancer referral settings—to determine the diagnostic yield of urinary cytology under daily clinical conditions.

METHODS

Records of consecutive urinary cytology specimens from 1672 patients reported from the years 1990–1994 were reviewed and correlated with histologic and clinical information. Initial analyses were based on the records themselves, without review of pathologic specimens. Subsequently, a subset of specimens was reviewed to determine reasons for noncorrelations.

RESULTS

Results confirmed that the diagnostic sensitivity and specificity of urinary cytology for high grade transitional cell neoplasms, as reported in the daily practice of pathology, are very high (79% and >95%, respectively). Disaggregated cells from low grade transitional cell neoplasms usually lack recognizable features of neoplasia, and attempts to classify such lesions cytologically result in low diagnostic yield, with an overall sensitivity of 26%. Of these 1672 patients, 707 had insufficient clinical information for analysis, despite diligent and persistent efforts to acquire it.

CONCLUSIONS

The diagnostic yield of consultations based on urinary cytology in the daily practice of pathology is high, regardless of whether the practice setting is referral-based or community-based. The available information indicates that in approximately 79% of patients with high grade transitional cell neoplasms, the neoplasms can be detected using urinary cytology. Conversely, a negative result is predictive of no cancer in more than 90% of cases. Sensitivity for detecting low grade urothelial lesions is low; however, most of these are easily detected cystoscopically. The authors' inability to acquire sufficient information to determine diagnostic yield in a large percentage of their cases was disturbing to them. Not only did this deficiency render their analyses incomplete, but lack of easily accessible follow-up and the apparent low priority for quality assurance activities among pathologists in all types of practice settings is likely to significantly reduce the feedback required for pathologists to acquire and maintain expertise in this very difficult area. Cancer (Cancer Cytopathol) 1999;87:118–28. © 1999 American Cancer Society.

Urinary cytology functions as the primary screening and surveillance modality for the detection of urothelial neoplasia at most medical centers. Although neoplastic urothelial cells were first recognized in the urine in 1864,1 it was not until 1945 that Papanicolaou and Marshall2 described the utility of urinary cytology in the diagnosis of urothelial malignancy. Subsequent to this, there have been many studies evaluating the diagnostic yield of this method,3–15 with cytohistologic correlation reported as high as 95% and sensitivity (positive/suspicious) for low grade urothelial neoplasms as high as 76%.13

Many critics of urinary cytology claim that the theoretic diagnostic yields achieved in these studies are not realistically attainable in daily practice due to a variety of factors, including suboptimal specimen quality and pathologists' lack of expertise with urinary cytology. In recent years, a variety of nonmorphologic techniques, including cell membrane blood group antigen expression, bladder tumor (basement membrane) antigen, tumor specific nuclear matrix proteins, DNA ploidy, molecular genetic markers, and others, have been developed with the aim of enhancing diagnostic yield relative to urinary cytology, challenging the latter's primary role in the evaluation of urothelial neoplasia.16–25 These techniques offer the potential advantage of increased sensitivity in the detection of incipient, low grade, and minimally sampled high grade urothelial neoplasms, but most of these are limited by their lower specificity relative to urinary cytology. A positive nonmorphologic test continues to require cystoscopic and pathologic correlation prior to definitive therapy. The current study was designed to determine the actual diagnostic yield of urinary cytology in daily practice in three different practice settings—academic, community, and cancer referral medical centers—based on a retrospective review of unaltered clinical and pathologic records of a series of patients undergoing urinary cytology.

MATERIALS AND METHODS

The clinical histories and pathology reports (cytology and histology) of 1672 patients, from 3 pathology practice settings (UPMC - 504, SJH - 782, COHNMC - 386) whose initial specimens were received prior to January 1995, were reviewed, ensuring a minimum of 2 years of follow-up. From each institution, the patients were selected consecutively based on either medical record or cytology accession numbers. The patients were classified into one of the following categories: histology-proven bladder carcinoma; history of bladder carcinoma on surveillance with no positive biopsies in the pathology department records; other pathologically documented, nonurothelial, genitourinary malignancies; no history of a genitourinary malignancy with screening for urothelial neoplasia, infection, etc.; and urinary cytology without patient history. Patients with urothelial malignancies limited to the upper urinary tracts were excluded from the study.

Histology reports at our institutions utilized terminology based on either the World Health Organization (WHO) or Third Series AFIP Fascicle schemes. For the purposes of this study, variations in reported histologic diagnoses for transitional cell neoplasms could be included under two basic categories: low grade and high grade. Low grade neoplasms included those reported as transitional cell papilloma; transitional cell carcinoma (TCC), low grade; and TCC, Grade 1 (Fig. 1). High grade neoplasms included those reported as TCC, Grade 2; TCC, Grade 3; or TCC, high grade (Fig. 2). A few nonurothelial primary and/or secondary neoplasms with positive urinary cytology were recorded as high grade. Similarly, variations in the diagnostic terminology of cytologic reports could be categorized as either positive/suspicious for low or high grade neoplasm (P/S) or negative. P/S included any reported diagnostic terms (e.g., transitional cell/urothelial carcinoma, malignant cells, dysplastic cells rule out neoplasm) that denoted the presence of a neoplasm. Negative included all other reported diagnostic terms (e.g., negative, atypical cells, dysplastic cells favor reactive). Pathologic interpretations of all genitourinary histologic specimens for each patient were reported with specimen type (biopsy, transurethral resection, cystectomy [partial and complete], cystoprostatectomy, ureterectomy, urethrectomy, and nephrectomy). Cytohistologic correlations were patient-based rather than specimen-based, such that one patient with multiple specimen correlations was tabulated only one time. A positive correlation satisfied one of two conditions: 1) a patient with at least one positive urinary cytology, concurrent with or preceding a positive histologic specimen within less than or equal to 1 year, or 2) a patient with consistently negative urinary cytology and histology specimens (or negative urologic evaluation in the absence of histology). A 1-year interval was arbitrarily selected for cytology-histology specimen correlations to avoid erroneous positive and negative correlations of 1) remote urothelial neoplasms undergoing spontaneous therapy-induced regression, or 2) more recently emerging neoplasms, which may not have been present at the time of a prior cytology examination, recognizing that this cutoff may exclude potential positive correlations. A discordant correlation occurred for patients with at least 1 positive diagnosis by either cytology or histology and consistently negative findings by the other modality within any 1 year following the positive diagnosis. The pathology slides of a subset of the discrepant cases were reviewed for specimen adequacy and cytologic findings. Specimens reported as suboptimal on review had 1 or more of the following deficiencies: less than 15 intermediate or basal urothelial cells, obscuring blood/inflammation, or poor cellular preservation. An explanation of discordant results was determined for each reviewed case.

Figure 1.

Low grade papillary urothelial carcinoma is represented in a urinary bladder washing (A, Papanicolaou stain, original magnification ×784) and a cystoscopic bladder biopsy (B, H & E, original magnification ×314).

Figure 2.

High grade urothelial carcinoma is represented in a urinary bladder washing (A, Papanicolaou stain, original magnification ×185) and a corresponding cystoscopic biopsy (B, H & E, original magnification ×93).

Review of the clinical information (charts or electronic records) of a sample of patients with negative cytology and no follow-up histology was performed directly to determine the disposition of these patients and indirectly to ascertain how urinary cytology screening was utilized at the three institutions. In addition, urologists at our institutions were surveyed in an effort to understand the role of urinary cytology in the practical clinical management of patients with suspected or documented urinary tract disorders (in particular, the role of cytology in the screening and surveillance of patients with urothelial malignancies).

RESULTS

Results are summarized in Figure 3 and Tables 1–4.

Figure 3.

A flow diagram of the results of urinary cytology-histology report correlation for 1672 cumulative patients from UPMC, SJH, and COHNMC (refer to “Results” section).

Table 1. Diagnostic Yield of Urinary Cytology
 OverallUPMCSJHCOHNMC
  1. UPMC: University of Pittsburgh Medical Center; SJH: St. Joseph's Hospital; COHNMC: City of Hope National Medical Center; UFCM: University of Florida College of Medicine; TCCa-1: transitional cell carcinoma, Grade 1.

Sensitivity64%47%85%66%
Specificity95%98%74%98%
PPV75%81%56%88%
NPV92%91%93%94%
Diagnostic accuracy89%90%77%96%
% TCCa-129%44%33%11%
Non-determinant45%11%82%4%
Table 2. Type, Quantity, and Quality of Specimens from Patients with Histologically Confirmed Bladder Neoplasms
Histologic diagnosisCytologic diagnosisInstrumented specimensMedian urine specimens per patient (range)Suboptimal specimensa
  • a

    “Suboptimal” is less than optimal and unsatisfactory (reported adequacy for each specimen).

TCCa 2/3 and(+)48%4 (1–29)3%
othern = 101 (79%)
n = 128 (71%)
(−)10%1 (1–3)9%
n = 27 (21%)
TCCa 1(+)37%3 (1–10)7%
n = 52 (29%)n = 14 (27%)
(−)42%2 (1–20)3%
n = 38 (73%)
Table 3. False-Negative Urinary Cytologies for 65 Patients
  • a

    No. (%) of patients with at least 1 suboptimal specimen.

Patients whose cytology was reviewed31 (48%)
Urine specimen adequacy
Suboptimala
 Reported 8 (26%)
 Review 18 (58%)
Review dxNo. of patientsNo. of patients with suboptimal cytology
 No tumor cells 19 (61%) 12 (63%)
 Tumor cells present 12 (39%) 6 (50%)
Table 4. Patients with Negative Urinary Cytology and No Histologic Follow-Up (n = 830)
• Patients with clinical history/follow-up (n = 220; 27%)
• Most common indications for urinary cytology: hematuria, irritative symptoms, history of a nonurothelial genitourinary malignancy, pyuria/surveillance for infection, history of calculi, abdominal mass, urinary retention, incontinence
• Demographic features of patient subset: no significant additional medical history, allograft recipients and other immunocompromised patients, diabetes mellitus, autoimmune disorders, other
• Diagnostic modalities: cystoscopy; intravenous/retrograde pyelogaphy; voiding cystourethrogram; upper urinary tract cytology; abdominal radiography (X-ray of kidneys, ureter, and bladder; computed tomography scan; ultrasound; renal perfusion scan); random renal biopsy; other
• Identified genitourinary abnormalities (50%): urinary tract infection, calculi, prostatic hypertrophy, prostatitis, cystitis, benign renal cysts, benign urothelial hemorrhage, renal hemorrhage, other

Demographic and Urinary Specimen Characteristics

Sources of urinary specimens included inpatients from general medicine, general surgery, urology, oncology, and psychiatry services and outpatients from urology, gynecology, oncology, and general medicine clinics; the majority of specimens were received from outpatient urology and general medicine clinics. Overall indications for urinary cytology for the determinant cases included 1) screening for urothelial neoplasms in patients with hematuria or irritative urinary symptoms (45%), 2) surveillance of patients with a history of urothelial neoplasia (29%), 3) screening for urinary tract involvement in patients with documented nonurothelial genitourinary malignant neoplasms or upper tract urothelial confined neoplasms (11%), 4) screening for infection (bacterial, fungal, or viral [cytomegalovirus, polyoma [BK]) and/or hemorrhagic cystitis in immunocompetent patients with suspected urinary tract infections or asymptomatic/symptomatic immunocompromised patients, many of the latter receiving chemotherapy (13%), and 5) miscellaneous (renal parenchymal disease, interstitial cystitis, urinary eosinophil evaluation, etc.) (2%). The exact percentages for each of the categories varied significantly among the three institutions: screening for urothelial malignancy was highest at SJH, intermediate at UPMC, and lowest at COHNMC, whereas surveillance of patients followed for treated urothelial neoplasms was highest at COHNMC, intermediate at UPMC, and lowest at SJH. COHNMC also used urinary cytology in many bone marrow transplant recipients to screen for opportunistic viral infections. The majority of urinary specimens were voided or not otherwise specified (72%) compared with a smaller number of instrumented specimens (catheterized, washing, barbotage, and brushing) (28%).

Most of the urine samples were submitted as a single specimen; however, a small percentage of patients contributed multiple concurrent specimens, including a few with paired voided and instrumented urinary specimens.

Cytohistologic Correlation Data

These data are shown in Figure 3. A total of 1672 patients were included in the study, with the following institutional breakdown: UPMC (504 pts), SJH (782 pts), and COHNMC (386 pts). Clinical history with variable follow-up data was available for 965 (determinant cases).

At least 1 urinary cytology was reported as positive/suspicious among 253 patients; 115 (45%) were confirmed histologically (91 high grade and 14 low grade urothelial carcinomas, and 10 nonurothelial malignancies), 39 (16%) were histologically negative, and 99 (39%) had no hospital record of a histologic evaluation (97 of these had no available clinical follow-up). Urinary cytology was reported as consistently negative/atypical cells in 1419 patients; 65 (5%) had at least 1 concurrent positive histologic evaluation (26 high grade and 38 low grade urothelial carcinomas and 1 lymphoma), 524 were histologically negative, and 830 had no hospital record of a concurrent histologic evaluation (220 had a documented urologic evaluation without biopsy and 610 had no or no readily accessible clinical history available).

Cytologic Diagnostic Yield

Cytologic diagnostic yield is shown in Table 1. The standard parameters assessing diagnostic yield of urinary cytology (relative to confirmatory histology) were calculated for the determinant patients, both overall and also individually for the three separate institutions. Composite sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy were 65%, 95%, 75%, 92%, and 89%, respectively. Histologically reported low grade urothelial neoplasms accounted for 29% of positive cases.

Characteristics of Urinary Specimens in Patients with Histologically Confirmed Bladder Neoplasms

These characteristics are shown in Table 2. Histologically confirmed high grade malignant neoplasms (TCCa 2, 3, and other) occurred in 128 patients (71%), of whom 101 (79%) had their neoplasms detected by urinary cytology. Low grade urothelial carcinoma involved 52 patients (29%) with histologically confirmed bladder neoplasms, for whom cytology was positive in only 14 cases (27%). The proportion of patients with an instrumented specimen, the number of urine specimens per patient, and the percentage of patients with consistently adequate specimens tended to be higher for patients with high grade malignant neoplasms detected by urinary cytology compared with patients who had high grade malignant neoplasms missed by urine cytology. For low grade urothelial carcinoma, these factors were quantitatively similar for patients detected and missed by urinary cytology. Approximately 80% of patients in a predominantly screening setting (UPMC) had an initially positive/suspicious urinary cytology; 91% were identified after 2 urinary examinations, and only 10% required between 3 and 6 cytologies (obtained on separate dates) to establish a diagnosis. Multiple positive urinary cytology-histology correlations were seen typically among patients undergoing surveillance for high grade urothelial carcinoma; in contrast, patients with multiple negative urinary cytologies and positive biopsies were more likely to have low grade urothelial neoplasms. Patients with high grade urothelial neoplasms and few urinary cytologies were followed in surveillance only by cytoscopy and biopsy; followed with urinary cytology, which was evaluated by an outside lab; followed at another location by a different physician; or lost to follow-up. Ninety-one percent of patients with both positive urinary cytology and positive biopsy had histologic confirmation within 3 months of the positive urinary cytology, with 79% of these patients having a correlation on the same day. Only 1 patient had histologic confirmation between 6 months and 1 year of the positive cytology (7 months).

Assessment of False-Negative Cases

False-negative cases are cited in Table 3 and Figures 4–6. Sixty-five of 180 (36%) histologically confirmed patients with bladder neoplasms had consistently reported negative urinary cytologies, including 27 patients (42%) with only low grade urothelial neoplasms. The cytology and corresponding histology for 31 patients (48%) were reviewed. (Some of the older false-negative cytology slides had been discarded after 5 years). At least 1 suboptimal urinary specimen was reported among 8 patients (26%), but on review (applying adequacy criteria described in “Materials and Methods”), 18 patients (58%) had at least 1 suboptimal specimen. On review, 19 patients (61%) had no recognizable tumor cells in any of their cytology specimens, and 12 of these patients (63%) had at least 1 suboptimal specimen. Twelve patients (39%) had rare suspicious or diagnostic malignant cells present, which were missed by the pathologist; 6 of these patients (50%) had at least 1 suboptimal specimen (Fig. 4). One patient had a low grade papillary urothelial carcinoma on biopsy, represented in the urinary cytology by rare fragments of low grade neoplastic cells, which were missed (Fig. 5); these fragments were inconspicuous at low magnification but diagnostic at high power. A patient correctly diagnosed as having a high grade urothelial carcinoma on the basis of an instrumented urinary specimen was reported as having atypical but degenerated urothelial cells in a concurrently obtained voided urine specimen; had only a voided specimen been submitted, the patient would have been falsely classified (Fig. 6). Most of the false-negative urine specimens (with missed tumor cells) from COHNMC were from patients with degenerated atypical urothelial cells who were undergoing therapy for urothelial carcinoma; these cells were interpreted as atypical benign urothelial cells with therapy effect.

Figure 4.

The urinary cytology of this high grade urothelial carcinoma was falsely reported as negative. Note the rare missed high grade urothelial cell in a background of acute inflammation and Candida yeasts (A, Papanicolaou stain, original magnification ×296). A transurethral bladder resection biopsy shows a high grade invasive urothelial carcinoma undermining benign urothelium with a small in situ component (B, Papanicolaou stain, original magnification ×93).

Figure 5.

A false-negative urinary cytology is shown for a patient with a low grade papillary urothelial carcinoma. The bladder washing contains small fragments and single reactive urothelial cells, and only rare fragments of the papillary neoplasm. Note the increased nuclear:cytoplasmic ratios, nuclear crowding, prominent contour, and mild chromatin derangements in the euchromatic tumor cell nuclei (A, Papanicolaou stain, original magnification ×925). Corresponding cystoscopic biopsy reveals a World Health Organization Grade 1 noninvasive papillary urothelial carcinoma (B, H & E, original magnification ×93).

Figure 6.

A comparison of concurrently obtained voided urinary specimen and a bladder washing is shown for a patient with a high grade urothelial carcinoma. The voided urine specimen (reported as atypical but degenerated) was hypocellular and the cells were poorly preserved (A, Papanicolaou stain, original magnification ×296) relative to the more cellular bladder washing (reported as positive for malignant cells) containing well-preserved high grade urothelial carcinoma cells (B, Papanicolaou stain, original magnification ×296).

Assessment of False-Positive Cases

False-positive cases are cited in Figures 7 and 8. A total of 39 patients had presumed false-positive urinary cytologies, i.e., a positive cytology and/or a negative urologic evaluation. Many of the patients included in this category had no readily documented urologic or histologic follow-up and are more accurately classified as indeterminant than as false-positive. Because the majority of these cases were not reviewed, we were unable to determine the distribution of causes resulting in false-positive diagnoses; rather, we are able to report causes in sporadic cases. One patient was incorrectly recorded in the chart as having a negative urinary cytology and was not evaluated further. Another patient did not return and was lost to follow-up. Two patients were biopsied elsewhere and were later confirmed to have urothelial carcinoma. A few discrepant cases contained atypical urothelial fragments with increased nuclear:cytoplasmic ratios falsely interpreted as low grade papillary urothelial neoplasms; the corresponding biopsies revealed reactive urothelium associated with instrumentation, polypoid cystitis, and urolithiasis effects (Fig. 7). Rare urine specimens falsely classified as high grade urothelial carcinoma were reported on biopsy to have either therapy effect (e.g., cyclophosphamide-induced cystitis) or mild flat dysplasia (Fig. 8). One patient (undergoing surveillance for urothelial carcinoma) with a urinary cytology reported as suspicious for a high grade urothelial neoplasm and a concurrent, superficial-epithelial denuded biopsy interpreted as low grade urothelial dysplasia was more likely to have an underreported biopsy interpretation than a false-positive cytology (Fig. 8).

Figure 7.

A false-positive urinary cytology is shown for a patient with a small bladder trigone lesion. The hypocellular bladder washing contains a single fragment of crowded, slightly atypical urothelial cells arranged in a papillary configuration (A, Papanicolaou stain, original magnification ×93). The nuclei, though overlapping, are euchromatic, and oval to round with smooth contours and evenly distributed chromatin (B, Papanicolaou stain, original magnification ×370). Cystoscopic biopsy reveals the focus of polypoid cystitis (C, H & E, original magnification ×60).

Figure 8.

“False-positive” urinary cytology is shown for a patient with a history of previously documented and treated urothelial carcinoma. The voided urinary specimen is hypocellular and contains rare, markedly atypical urothelial cells reported as suspicious for high grade urothelial carcinoma (A, Papanicolaou stain, original magnification ×925). One of five random bladder biopsies contains an isolated mucosal epithelial focus reported as low grade urothelial dysplasia (B, H & E, original magnification ×244).

Assessment of Patients with Negative Urinary Cytology and No Urothelial Histologic Evaluation (Table 4)

Of the 830 patients with negative urinary cytologies and no recorded histologic evaluation, clinical information was obtained on 270 patients (33%), 220 (81%) of whom had clinical history and follow-up information that was easily accessible (in computerized medical records). Indications for urinary cytology included isolated microscopic/macroscopic hematuria (31%); hematuria in combination with other signs or symptoms, including pyuria, irritative symptoms, urinary retention, incontinence, flank/abdominal pain, anticoagulation, and systemic cyclophosphamide therapy (31%); history of a nonurothelial genitourinary malignant neoplasm (18%); isolated pyuria/surveillance for infection (10%); and miscellaneous (irritative symptoms, urinary retention, anemia with unexplained weight loss, flank/abdominal pain, mass/abnormal radiographic finding, history of calculi, hypertension evaluation, azotemia, hydronephrosis, and history of pelvic irradiation) (10%). Additional demographic characteristics of this patient population included individuals with the following underlying disorders: allograft recipients (9%), diabetes mellitus (7%), nonbiopsied systemic autoimmune diseases (2%), documented glomerulonephritis (1%), chronic renal failure (1%), and HIV/AIDS (1%). Some type of diagnostic evaluation was documented in 242 patients (89%), including cystoscopy (60 patients), intravenous pyelography (58 patients), cystoscopy with retrograde pyelogram ± upper tract cytology ± voiding cystourethrogram (34 patients), abdominal (kidney, ureter, and bladder) radiography (16 patients), abdominal ultrasound (59 patients), abdominal computed tomography scan (23 patients), renal perfusion scan (3 patients), and random renal biopsy (5 patients). Genitourinary abnormalities were identified in approximately 50% of these evaluated patients, including urinary tract infection (32 patients); calculi (18 patients); BPH/bladder outlet obstruction (12 patients); benign renal cysts (12 patients); prostatitis/cystitis (10 patients), including 1 patient with interstitial cystitis; urothelial hemorrhage associated with coagulopathy, Foley catheter, mucosal hemangioma/arteriovenous malformation, radiation cystitis, graft-versus-host disease, or urothelial friability not otherwise specified (16 patients); renal hemorrhage associated with glomerulonephritis, allograft rejection, papillary necrosis, polycystic kidney disease, or mitomycin-induced hemolytic uremic syndrome (15 patients); miscellaneous (post–transurethral prostate resection, neurogenic bladder, urethral diverticula, renal calyceal blunting [? chronic pyelonephritis], double collecting system, tuberculosis, extrarenal pelvis and pelvic mass) (12 patients).

DISCUSSION

Many studies have been designed to evaluate the theoretic diagnostic yield of urinary cytology. In a review of many published retrospective cytohistologic correlation studies, sensitivity ranged from 44.7 to 97.3%, with a mean of 75.6%; specificity ranged from 88.1 to 99.5%, with a mean of 97.1%; and mean positive and negative predictive values were 89% and 98.2%, respectively. The mean false-positive rate was 1.5%. For papillomas (TCC 1, according to World Health Organization criteria), the sensitivity ranged from 1% to 20%, with a mean of 11.0%.9 Other studies have produced similar results,12–14 with consistently high diagnostic yields for high grade urothelial neoplasms and inconsistent diagnostic yields for low grade urothelial lesions. Wolinska et al. concluded that true papillomas could not be reliably diagnosed by urinary cytology because of the overlap with reactive conditions, but conceded that a specific constellation of findings (increased cellularity, hematuria, and atypically shaped cells) in conjunction with the appropriate clinical setting favored the likelihood of a papilloma being present.15

These studies suggest that there is a theoretic limit with respect to the cytologic detection of urinary bladder neoplasms. This limit is determined by the histologic classification used for correlation, cytomorphologic constraints, specific individual specimen quality, and interpretative experience and skill of the diagnostician. If the urinary specimen fails to sample neoplastic cells or cellular constituents, neither morphologic criteria nor ancillary techniques will detect a concurrent urothelial neoplasm. If the urinary specimen contains neoplastic cells that are either poorly preserved or morphologically indistinguishable, but genetically/antigenically distinctive from reactive cells, the cells will frequently be missed cytologically but may be detected by one or more ancillary techniques.16–25 Although many of these nonmorphologic techniques offer sensitivity advantages, they tend to be less specific than morphology. In spite of the limitations of urinary cytology, the various cytohistologic studies have validated the efficacy of the method when subjected to retrospective review by expert urinary cytopathologists. Our multicenter study has shown that under the conditions of daily practice, excellent diagnostic yields are achievable for high grade urothelial carcinomas, i.e., biologically significant neoplasms that are often cystoscopically difficult to detect. Overall, sensitivity for high grade neoplasms was 79%. Low grade urothelial lesions were detected in only 27% of affected patients, but fortunately they were easily detectable cystoscopically. Most importantly, the overall diagnostic accuracy was 89%.

Our study identified causes of false-negatives and false-positives commonly occurring in the daily practice of urinary cytology. Thirty-six percent of patients with histologically confirmed urothelial neoplasms were missed by urinary cytology; 42% of these false-negative patients had low grade neoplasms. Although a few of these neoplasms were cytologically distinguishable as malignancies, the majority lacked distinctive malignant features, overlapping with reactive conditions. Of the reviewed false-negative patients, many (58%) had at least 1 urinary specimen that was suboptimal due to low cellularity, poor preservation, or obscuring of cells. Furthermore, of the patients with missed high grade neoplasms, a disproportionate number had single voided urinary specimens; false-negative rates were decreased in high grade patients who had multiple, instrumented specimens. The advantage of multiple specimens may be related to nonuniform (inconstant) shedding of neoplastic urothelial cells, with multiple collections increasing the likelihood that the abnormal cells will be sampled. Instrumented samples offer the advantages of increased cellularity and better cellular preservation, whereas voided specimens allow for increased duration of sampling, sampling of the upper tracts, and ability to use tissue fragments as a soft criterion for low grade papillary lesions. Overall, instrumented samples produced a higher sensitivity of detection in patients with histologically confirmed urothelial carcinoma. The other significant cause of false-negative diagnosis was ascribing therapy-induced changes in neoplastic cells as a benign therapy effect.

The overall false-positive rate was 2.7%. The majority appear to have been erroneous overinterpretations of reactive conditions falsely classified as low grade urothelial neoplasms, and in patients who were lost to follow-up.

Urinary cytology is utilized for a variety of indications that vary significantly depending on the patient population, specific practice setting, and expectations of the clinicians requesting the examination. In community hospital practice, urinary cytology is used primarily to screen for urinary tract neoplasms presenting with persistent hematuria and irritative symptoms. Screening typically consists of 1–3 voided urinary cytologies ideally collected on separate days, with cystoscopy, upper tract radiographic contrast dye studies, and biopsy reserved for patients with 1) positive/suspicious urinary cytologies or 2) a high index of clinical suspicion. Follow-up is provided for patients with negative/atypical urine specimens and persistent symptoms. Although some of these patients are biopsied, the majority are not. The prevalence of urinary tract cancer in this patient population, though substantially increased over age-matched controls, is low compared with the various nonneoplastic disorders identified through the urologic investigation. Furthermore, a significant percentage of newly diagnosed urothelial neoplasms occurring in a screening population (33% and 44% at SJH and UPMC) are low grade, i.e., difficult to detect by urinary cytology. By contrast, urinary cytology in cancer referral practices is applied more frequently in the surveillance of tumor recurrence in patients with treated (commonly high grade) urothelial neoplasms. The cellular morphology is often altered by intercurrent cystitis and therapy effects.

It is apparent that the focus of urinary cytology applied to daily practice differs from that of methodologic retrospective cytohistologic correlation studies. The latter are analyses designed to determine maximal diagnostic yield based exclusively on morphologic factors, whereas the former serve as medical consultations influenced by a combination of clinical factors and morphology. As such, the practice setting invariably influences the diagnostic threshold of the interpreting pathologist in daily practice. In a practice setting oriented toward screening, the pathologist's threshold for making a positive diagnosis may be lower, because newly diagnosed urothelial neoplasms are typically confirmed histologically prior to the initiation of therapy. The perceived consequences of a false-positive diagnosis may not be clinically important, and the benefit of greater sensitivity may outweigh the cost of reduced specificity. In a surveillance mode, urinary cytology may be used diagnostically, equivalent to a histologic diagnosis. A false-positive diagnosis may result in an unnecessary (and often overly aggressive) therapy.

Diagnostic thresholds in all practice settings are optimized through the implementation of quality assurance reviews, particularly cytohistologic correlation and clinical follow-up. Cytohistologic correlation both refines and prioritizes individual pathologist's internalized diagnostic criteria, permitting more reliable distinction between cytologically overlapping entities, more precisely defining the limits of cytologic resolution. Examples include distinguishing reactive changes from low grade urothelial neoplasms at the low end of the spectrum and separating therapy and polyoma (BK) viral cytopathic effects from high grade urothelial neoplasms at the opposite end of the spectrum. Quality assurance provides the necessary feedback required to develop specimen adequacy criteria, demonstrating how and to what extent the number and types of samples, cellularity, and cellular preservation impact diagnostic yield. It is noteworthy that in our study the majority of reviewed patients with false-negative urinary cytology had at least one suboptimal urinary specimen, but that many of these specimens were actually reported as adequate. Clinicians need to refine collection methods to assure consistent adequacy of sampling, and pathologists must be equally vigilant in the implementation and reporting of adequacy standards to reduce false-negative rates. Finally, quality assurance provides the pathologist with feedback regarding how urinary cytology influences clinical management, which in turn modifies individual diagnostic thresholds. Without quality assurance, the opportunity to refine diagnostic skills is compromised, and diagnostic biases and systematic errors are perpetuated. This is particularly true for pathology practices in which the cytology and histology are usually examined by different pathologists. And yet, despite its enormous potential benefits, quality assurance is progressively underutilized and performed mainly to satisfy laboratory inspection requirements. It is an unfortunate consequence that, as pathologists' workloads increase in an expanding managed care environment, there is less available time and less incentive for continuing education, research, and quality assurance activities, none of which are reimbursable.

One result of this constellation of factors is the trend toward marginalizing the role of the pathologist from one of medical consultant to one of laboratory manager. In one of the participating hospitals in this study, for example, the collaborating pathologist was actually denied access to the medical records without the written consent of each of the involved patients, as if the medical consultation had been a laboratory test performed by an individual with no medical justification for the follow-up information. Perhaps in the future patient information will be more readily accessible if a national patient data base is developed, facilitating quality assurance reviews in all areas of medicine.

CONCLUSIONS

Points in summary:

  • Urinary cytology represents a medical consultation requested by one physician from another.

  • The diagnostic yield remains high, especially for patients harboring high grade carcinomas in their urinary bladder.

  • The diagnostic yield for low grade urothelial neoplasms is low, reflecting the benign nature of these tumors.

  • Diagnostic errors are unusual. The majority of false-negative results for high grade carcinomas reflect either the absence of tumor cells in the specimens or inadequate specimens containing few, obscured, or poorly preserved tumor cells. The majority of false-positive results occur in patients lost to follow-up, with a smaller number related to overinterpretation of reactive changes as a low grade urothelial neoplasm.

  • Both the expected yield and the clinical consequences influence a consultation based on urinary cytology in daily practice settings.

  • A significant minority of urinary specimens actually lack sufficient cells to be considered representative but have been interpreted as negative.

  • Trends in current medical practice provide little incentive for ongoing quality assurance, even in academic centers, and this impedes the feedback necessary for practitioners to maintain expertise.

  • These trends tend to marginalize practicing pathologist to a technical role, minimizing their medical contribution while overemphasizing the inherent imprecision of their interpretive skills.

Ancillary