- Top of page
- MATERIALS AND METHODS
- FUNDING SUPPORT
- CONFLICT OF INTEREST DISCLOSURES
Urothelial carcinoma (UC) accounts for > 70,000 new cancer cases per year and approximately 14,000 deaths annually in the United States. Approximately 70% of patients initially present with Ta/T1 tumors. Depending on risk stratification, up to 15% of these tumors progress to a muscle-invasive stage, which is associated with a limited 5-year survival rate. Early diagnosis is essential to prevent progression to a muscle-invasive stage, which is associated with a remarkable increase in mortality. Cystoscopy with subsequent transurethral resection remains the gold standard for the diagnosis of bladder cancer. In addition, current guidelines recommend performing urine cytology (CYT) as an adjunct tool in patients with suspected UC. Although it demonstrates a relatively high specificity, the diagnostic value of CYT is impaired by its low sensitivity. Therefore, several additional urine tests have been developed within the last 20 years to improve the detection rate of UC. These include fluorescence in situ hybridization (FISH) for the detection of chromosomal aberrations (loss of the 9p21 locus, site of the p16 tumor suppressor gene, has been identified as a common chromosomal aberration in patients with UC) and immunocytology, which in addition to normal urine CYT uses a cocktail of antibodies against UC-associated surface antigens such as carcinoembryonic antigen and mucin-like glycoprotein (the uCyt+ test). Furthermore, protein-based tests such as the quantitative nuclear matrix protein 22—enzyme-linked immunoadsorbent assay (NMP22-ELISA) have become widely available and are frequently used by specialized centers and practitioners. In the majority of previous studies, these tests all demonstrated improved sensitivity but decreased specificity compared with CYT. One limitation of some of these markers is a relatively high false-positive test rate because of endogenous or exogenous factors including mechanical manipulations, urinary tract infection (UTI), hematuria, and impaired renal excretory function.
Although several studies have been performed that directly compare the diagnostic performance of various urine tests, to our knowledge little is known regarding whether a combined application of these markers may improve their diagnostic performance in patients suspected of having UC.
Therefore, the objective of the current study was to investigate the diagnostic performance of combined applications of the 4 most broadly available urine markers (CYT, FISH, uCyt+ and NMP22-ELISA) in a large cohort of 808 patients suspected to have UC.
- Top of page
- MATERIALS AND METHODS
- FUNDING SUPPORT
- CONFLICT OF INTEREST DISCLOSURES
Cystoscopy is still the gold standard for the primary diagnosis and follow-up of patients with UC. However, the performance of cystoscopy every 3 to 6 months in the follow-up setting is associated with discomfort and stress for patients. Therefore, there has been an intense search for reliable noninvasive markers of UC within the last decade. The tests, which to our knowledge have shown the most promising results to date, include FISH for the detection of chromosomal abnormalities within malignant cells in the urine, uCyt+ for the detection of cancer-associated antigens, and a protein-based assay for the detection of NMP22. Several studies have been performed to compare these tests with conventional CYT and the majority revealed an increased sensitivity associated with a decreased specificity, with the result that these tests are not yet able to replace cystoscopy.[7, 8, 10] Because some markers are characterized by high specificity whereas others have the advantage of having higher sensitivity, the objective of the current study was to investigate whether the combined use of the 4 most broadly used urine markers (CYT, FISH, uCyt+, and NMP22-ELISA) could improve diagnostic performance in a large cohort of patients receiving diagnostic workup for symptoms believed to be suspicious for UC.
The overall performances of the applied tests were found to be similar to those of prior studies. The 2 tests that were found to perform best were CYT and FISH, which demonstrated AUCs of 0.78 and 0.79, respectively. Although CYT demonstrated a slightly higher specificity (87.5% vs 86.3 %), FISH had a better sensitivity (71.3% vs 67.8%). However, compared with previous studies, CYT achieved a considerably higher sensitivity. This most likely is the consequence of having a well-experienced cytologist involved, which has been demonstrated to be of major importance for the accuracy of CYT. NMP22-ELISA was the test with the highest overall sensitivity (84.4%) but lowest specificity (41.3%). When excluding patients with factors that influenced the performance of NMP22-ELISA, specificity was found to be clearly increased (75.3%) whereas sensitivity was far lower, which is in accordance with the findings of prior studies. Similar to other studies, uCyt+ outperformed CYT with regard to test sensitivity. However, the specificity of uCyt+ was found to be lower.
Whenever investigators perform combinations of biomarker tests for cancer, they are faced with the question of what are the criteria for a positive test combination. By considering a test combination to be positive if 1 of 2 tests is positive, a better sensitivity is expected with a parallel decrease in specificity. A combination requiring multiple positive tests to be considered positive is expected to be associated with an increase in specificity and a decrease in sensitivity. One reasonable method with which to compare various test combinations, which has also been performed by others, is ROC analysis of models containing various marker combinations. In this case, the AUC reflects the overall performance of the model. Moreover, the cutoff value with the optimal correlation of sensitivity and specificity can be determined.
The results of the current study indicate that particular test combinations demonstrate a clearly improved performance compared with single tests. Combining CYT and FISH clearly improved the detection of UC compared with the use of single tests. Using the optimal cutoff value of ROC analysis for this 2-test algorithm, only patients who are negative for both FISH and CYT are considered to have a negative test combination. Because one would hardly omit either a positive FISH result or a positive CYT result, this combination appears to be realistic for clinical practice. One reasonable approach would be a stepwise performance (CYT in the first setting and FISH in CYT-negative patients). The FISH method is associated with considerable costs and CYT has such a high specificity, that CYT-positive patients should undergo an invasive workup in any case. In the current study cohort, the performance of FISH in CYT-negative patients identified 12 additional tumors and reduced the percentage of missed grade 3/CIS tumors by 62%. However, 44 patients who were CYT negative were found to have false-positive FISH results. A 2-test combination of FISH and uCyt+ was found to have the highest AUC of all the 2-test combinations (AUC, 0.85). However, the implications of this model cannot be transferred directly into clinical practice. By dichotomizing test patterns into positive and negative according to the optimal cutoff value (based on ROC analysis), the result of the combination is still determined by FISH and an additional uCyt+ test does not alter the decision based on the applied model. This phenomenon also can be observed with other combinations. Therefore, the increased explanatory power of these combinations cannot be fully exploited by dichotomizing test combinations into positive and negative. Rather, these combinations may increase the accuracy of models predicting the probability of a patient having UC. One practical approach for calculating the risks of UC are nomograms that include demographic and clinical parameters. Nomograms including the results of urine markers have been recently implemented. The results of the current study indicate that urine marker combinations might increase the predictive ability of these nomograms, because the AUC of the base model containing factors frequently used for nomograms (age and hematuria status) was clearly improved by the addition of urine marker panels.
One issue that has been discussed frequently is whether urinary tumor markers might individualize the intervals between cystoscopies in the surveillance of patients with a history of bladder cancer. To our knowledge, because no single marker to date has demonstrated sufficient performance to replace cystoscopy in this setting, there is a need to examine whether combinations of urine markers may be suitable. Although the current study was performed in patients without a history of UC, the improved performance noted through the combination of urine markers may also be observed in patients under surveillance and models including multiple urine tests might be used for future studies addressing this issue.
The main limitation of performing multiple urine markers is the relatively high costs. However, regular cystoscopy is also associated with considerable costs. Therefore, it appears to be reasonable to address the cost-effectiveness of different approaches for UC surveillance in future studies. Moreover, cost-benefit analyses are required to analyze whether the performance of additional urine markers is cost-effective.
The current study has various limitations. First, a high percentage of urine samples were obtained by instrumented urinary sampling. Second, we did not assess the smoking status of the patients involved, which could have an effect on the base model applied in the current study.
In conclusion, to the best of our knowledge, the current study is the first to investigate whether the combined use of cytology, FISH, uCyt+ and NMP22-ELISA can improve diagnostic performance compared with single tests in patients who are suspected of having primary UC. The results demonstrate that the expedient combination of urine markers enhances their diagnostic accuracy. FISH appears to be a reasonable amendment to routine CYT and identifies a considerable percentage of tumors that are not detected by CYT alone with an acceptable rate of false-positive results. These results may form the basis of further studies investigating whether risk stratification based on the combined application of urine markers may individualize diagnostic algorithms and the surveillance of patients with symptoms that are suspicious for UC.