Urine Survivin As a Diagnostic Biomarker for Bladder Cancer: a Systematic Review BJUIBJU INTERNATIONAL Ja Hyeon Ku *† , Guilherme Godoy * , Gilad E

Home , Survivin

Urine survivin as a diagnostic biomarker for bladder cancer: a systematic review BJUIBJU INTERNATIONAL Ja Hyeon Ku *† , Guilherme Godoy * , Gilad E. Amiel *‡ and Seth P. Lerner * * Scott Department of Urology, Baylor College of Medicine, and ‡ Michael E. Debakey Veterans Affairs Medical Center, Houston, TX, USA, and † Department of Urology, Seoul National University College of Medicine, Seoul, Korea Accepted for publication 9 November 2011

To determine the clinical utility of urine What ’s known on the subject? and What does the study add? survivin as a bladder tumour marker we Although many tests for identifying patients with new or recurrent bladder cancer systematically reviewed the available have been used, a reliable method has yet to be established. Recently, increasing evidence. A comprehensive literature review attention has focused on the role of survivin in bladder cancer detection. was performed, from August 1997 to March 2011, using three search engines in Because urine survivin tests have better sensitivity than cytology, urine survivin could English including PubMed, Cochrane potentially replace routine cytology and might be used as an adjunct method for Library, and SCOPUS. Two reviewers cystoscopy. However, the clinical utility of urine survivin as a bladder tumour marker independently evaluated both trial identifi ed in the present study remains to be elucidated. eligibility and methodological quality and data extraction. We included studies that subsets of data. There were 2051 subjects, cytology, but did not match cytology for evaluated urine survivin, used cystoscopy including 1038 in the case group and 1013 specifi city. The clinical utility of urine and/or histopathology as the reference in the control group, and heterogeneity survivin as a bladder tumour marker standard, and allowed the construction of was present among diagnostic studies. The identifi ed in the present study remains to a 2 × 2 contingency table. Bivariate pooled sensitivity and specifi city for urine be elucidated. random effect meta-analyses were used to survivin tests were 0.772 (95% confi dence calculate the summary estimated of interval [ CI ] 0.745 – 0.797) and 0.918 (95% sensitivity and specifi city and to construct CI 0.899 –0.934), respectively. The area KEYWORDS a summary receiver-operating under the curve of urine survivin tests characteristics curve of urine survivin tests. was 0.9392. When a subgroup analysis bladder neoplasm , survivin , urine , In all, 14 studies were included in the with six studies was performed, urine diagnostic test , systematic review , present review; two studies had two survivin tests had better sensitivity than meta-analysis

INTRODUCTION compared with expensive and invasive tests [ 5 ] . Most of these tests have a better cystoscopy procedures. Although urine sensitivity for detecting bladder cancer than Bladder cancers are the second most cytology is highly sensitive in the urinary cytology, but the specifi city is lower. common urological malignancy in the recognition of high-grade urothelial It remains unclear whether or not these Western world, with an estimated 70 530 carcinomas, urine cytology has poor tests offer additional information that is new cases diagnosed each year in the USA sensitivity in low-grade bladder tumours useful for the detection and management of [ 1 ] . Most diagnosed patients (75 – 85%) [ 5 ] . In addition, urine cytology has poor non-muscle invasive bladder tumours [ 5 ] . present with non-muscle invasive disease, inter- and intra-observer reproducibility [ 6 ] . Therefore, a reliable diagnosis and which is characterised by a probability of Cystoscopy is the ‘ gold standard ’ method for surveillance method has yet to be recurrence and progression at 5 years of diagnosing bladder cancer and is also used established. 31 – 78% and 1 – 45%, respectively [ 2 ] . Early in the surveillance of patients with a history diagnosis and postoperative monitoring of of bladder cancer [ 3 ] . However, this method An ideal marker would be offi ce-based, rapid bladder cancer is benefi cial to improve is an invasive and costly procedure, may and inexpensive, and have high specifi city patient prognosis [ 3 ] because detecting miss small papillary tumours and carcinoma and sensitivity in the target population, bladder cancer earlier may decrease the rate in situ , and may be over used particularly in which might reduce the burden of frequent of progression [ 4 ] . patients with low-grade disease. cystoscopy and cost of care. Recently, increasing attention has focused on the role Screening, as well as follow-up of patients Many studies have searched for, developed, of survivin in bladder cancer detection. with bladder cancer, is usually performed and evaluated new tests for identifying Survivin is a structurally unique inhibitor of by urinary cytology as an adjunct to patients with new or recurrent bladder apoptosis protein (IAP) characterized by a cystoscopy. Urine cytology has advantages cancer that are more reliable than cytology highly conserved baculovirus IAP repeat

(BIR) domain. Survivin is a 12-amino acid search criteria. We then determined the fi nal version of the Quality Assessment of protein (16.5 kDa) encoded at chromosomal studies included in the meta-analysis after Diagnostic Accuracy Studies (QUADAS) tool region 17q25, and acts as an IAP by applying the same inclusion and exclusion [ 11 ] , which has 14 items. Every included inhibiting caspases 3, 7, and 9 [ 7 ] . Survivin criteria to the remaining content of the study was assessed by ‘ yes ’ , ‘ no ’ , and is almost exclusively expressed in malignant articles. To be enrolled in the present study, ‘ unclear ’ from the variations (items 1 and 2), epithelium and can be measured in urine at retrieved studies had to fulfi l the following the bias (items 3 – 7, 10 – 12, and 14) and the mRNA and protein level [ 8 ] . Survivin is inclusion criteria: provide a 2 × 2 the report quality (items 8, 9, and 13), expressed in human and rat embryogenesis contingency table; detect survivin in urine; respectively, and the causes of bias and tissues and most human tumours, but not in published in English; include a per patient variations were identifi ed. normal adult tissues (except the thymus analysis; and use cystoscopy and/or gland) [ 9 ] , which makes survivin especially histopathology as the reference standard. HOMOGENEITY TEST signifi cant compared with other IAP Exclusion criteria were as follows: non- members broadly distributed in normal adult human study; review articles; letters; Heterogeneity was explored using the tissues. Given the sharp differential editorial comments; case reports; articles chi-square test and likelihood ratio (LR) I2 . expression of survivin in cancerous vs that did not include raw data; and studies LR I2 measures the percentage of the total normal tissues, the detection of survivin on cancer not arising in the bladder. When variation across studies that are due to appears to be a suitable tool for cancer data or subsets of data were presented in heterogeneity rather than chance. LR I2 is diagnosis. As a member of IAPs, survivin is more than one article, only the largest series evaluated as follows: I2 = (Q – df)/Q × 100%, overexpressed in undifferentiated cells were included in the analysis. A cohort of where Q is Cochran ’ s heterogeneity statistic associated with a high potential of cell patients was not included more than once and df is the degrees of freedom. An I2 proliferation [ 10 ] . in the same analysis. Included studies met > 50% may be considered to represent the quality criteria for studies of diagnosis substantial heterogeneity [ 12 ] . For the We performed a systematic review of the of the Oxford Centre for Evidence-Based LR chi-square test, we judged that diagnostic accuracy of trials of urine Medicine. Briefl y, the studies had clearly heterogeneity was signifi cant for P < 0.05. survivin tests for bladder cancer. The present identifi ed different groups of patients (with When there was evidence of heterogeneity, study sought to evaluate and compare the and without disease, and positive and data were analysed using a random effects detection validity (sensitivity and specifi city) negative tests). All included studies had a meta-analysis to obtain a summary estimate of urine survivin tests in the diagnosis of true-negative group that was free of the for the test sensitivity with 95% CIs. Studies bladder cancer and to construct a summary disease according to the defi nition used in in which positive results were confi rmed receiver-operating characteristic (ROC) the present study. were conducted using a pooled specifi city curve. with 95% CIs. EXTRACTION OF DATA THRESHOLD EFFECT ANALYSIS MATERIALS AND METHODS One reviewer (J.H.K) screened the titles and abstracts identifi ed by the search strategy A different threshold may be used in SEARCH STRATEGY and then two reviewers (J.H.K. and G.G.) included studies to defi ne positive test independently assessed the full tests for results due to lack of standardisation. A A comprehensive computer literature search inclusion. The database was designed to differential threshold effect may be the for abstracts was performed to identify ensure the most relevant data with respect reason for detectable sensitivities and articles about the diagnostic performance of to author, year of publication, demographics specifi cities of test accuracy studies. urine survivin for the detection of bladder of the patients, tumour characteristics, Summary ROC curves were constructed to cancer. We did a systematic review of geographic location, period of recruitment, see the variation between studies and original articles published between August study design, adoption of the ‘ blind ’ method, summarise the results quantitatively. A 1997 and March 2011 that analysed the and reference standard. The data extracted threshold effect was suspected with the diagnostic role of urine survivin in patients from articles comprised the number of ROC space graph if the points aligned in a with bladder cancer. A literature search of true-positive, true-negative, false-positive, typical shoulder-like pattern. Computation PubMed, Cochrane Library, and SCOPUS and false-negative cases as confi rmed using of the Spearman correlation coeffi cient was databases using the keywords ‘ survivin ’ cystoscopy and/or histology. If not directly also assessed for a threshold effect. A high and ‘ bladder cancer ’ was performed presented, the statistical parameters were correlation between sensitivity and independently by two reviewers for this calculated from the sensitivity and specifi city indicated such a threshold effect. meta-analysis. The searches were restricted specifi city or predictive values and the to English language publications. results of the reference test as given. SUBGROUP ANALYSIS Disagreements were resolved by consensus SELECTION CRITERIA or arbitration by another reviewer. Subgroup analysis was used to investigate the differences in the performance of Articles were selected in two steps. First, QUALITY EVALUATION urine survivin tests and urine cytology. articles were excluded after inclusion and Comparisons of the diagnostic accuracy of exclusion criteria were applied to the titles Two reviewers independently evaluated the urine cytology and urine survivin tests and abstracts of the articles fulfi lling the quality of the included studies using the between subgroups were carried out by

comparing the diagnostic odds ratio (OR) of Identification: identified studies from FIG. 1. the subgroups. The diagnostic OR expresses the database (n = 343) Flow diagram of identiﬁ cation of relevant studies. how much greater the odds of having the Removed repeated sudies (n = 110) disease are for the people who have a positive test result than for the people who Screening: studies after excluded duplicates (n = 233) have a negative test result. Excluded studies because they are irrelevant (n = 197) To investigate the differences in the Eligibility: studies for full test review performance of a given method for different (n = 36) = subgroups, we performed another subgroup Excluded studies (n 22) other languages (n = 6) analysis. The pooled estimates were review of literature (n = 5) reappraised when only studies on urine- editorials (n = 1) based survivin mRNA test using reverse report of already reported date (n = 1) required study design not met (n = 5) transcriptase (RT)-PCR were included, and other procedure (n = 2) the reappraised results were compared with other end point (n = 2) Included: final studies included in the original results. this meta-analysis (n = 14)

STATISTICAL ANALYSIS Egypt [ 20,25 ] , and one was performed in (question 7) were avoided. The interpretation The homogeneity test, threshold effect Iran [ 22 ] . Of the 14 studies, three (21.4%) of the reference standard results generally analysis, pooled weighted sensitivity and enrolled patients prospectively [ 16,18,26 ] occurred without the interpreter ’ s specifi city, and summary ROC curve were and the manner in which data was collected knowledge of the index test results (71.4% performed using Meta-Disc (version 1.4) for the other 11 studies was not defi ned. of the studies had ‘ yes ’ responses to [ 13 ] . The results were regarded as signifi cant Patient recruitment was consecutive in two question 11). at a two-sided P value of 0.05; P < 0.05 was studies (14.3%) [ 18,24 ] and was not defi ned considered statistically signifi cant. in other studies. Five studies were conducted ASSESSMENT OF STUDY HETEROGENEITY by general RT-PCR [ 14,15,20,22,24 ] , fi ve were quantitative RT-PCR [ 18,19,23,25,26 ] , The forest plots of sensitivities and RESULTS one was nested RT-PCR [ 17 ] , one was ELISA specifi cities from 16 subsets of data from [ 21 ] , two were Bio-Dot [ 14,15 ] , and one was all 14 studies are shown in Fig. 2 . A LITERATURE RETRIEVAL Molecular beacon [ 27 ] . All studies used homogeneity test of sensitivity and conventional cystoscopy and/or specifi city showed the following: a Figure 1 shows the fl ow of studies through histopathology as a reference standard chi-square, 125.66 with 15 df (P < 0.001), the review. In all, 343 potentially relevant ( Table 1 ) [ 14 – 27 ] . The percentage of and I2 = 88.1%; and chi-square, 102.38 studies were identifi ed on the basis of the non-muscle invasive tumour cases with 15 df (P < 0.001), and I2 = 85.3%, above search items, and 110 repeated (carcinoma in situ , Ta, or T1) among all respectively. Therefore, notable documents were removed. After a reading of tumours (the three studies [ 14,20,25 ] for heterogeneities were detected. the titles and abstracts, 197 studies were which stages were not reported) was excluded because of irrelevance. Further between 26.7% and 90.0%, with an average ASSESSMENT OF THE THRESHOLD EFFECT assessment for more detailed information of 69.3% (514 among 742 tumours). identifi ed 22 ineligible studies. The reasons The representation in a ROC space showed for study exclusion were as follows: other QUALITY EVALUATION OF INCLUDED STUDIES that the pattern of the points in the plot languages (six); review of literature (fi ve) or was not a shoulder-arm shape ( Fig. 3 ). A editorial (one); report of previously reported Most studies had a suboptimal design for Spearman rank correlation was used as a data (one); required study design not met the reporting of selection criteria (21.4% for further test for the threshold effect. Analysis (fi ve); other procedures (two); and other ‘ yes ’ responses to question 2) and reporting of the diagnostic threshold showed that the endpoint (two). Finally, 14 studies were of uninterpretable and/or intermediate test Spearman correlation coeffi cients for urine included [ 14 – 27 ] . Two studies had two results (21.4% for ‘ yes ’ responses to survivin tests were 0.035 (P = 0.897) and a subsets of data [ 14,27 ] . There were 2051 question 13). Only three studies used the diagnostic threshold effect was therefore subjects, including 1038 in the case group ‘ blind ’ method (review bias, question 10) not observed. and 1013 in the control group. [ 16,24,28 ] and all the studies that did not execute the reference standard described SUMMARY ESTIMATES OF SENSITIVITY, STUDY CHARACTERISTICS insuffi cient detail to permit replication of SPECIFICITY, DIAGNOSTIC OR, AND the test (diagnostic review bias, question 9). SUMMARY ROC CURVE Three studies were conducted in the USA However, patients in all of the studies [ 14,16,23 ] , one was carried out in Japan received the same reference standard The pooled sensitivity and pooled specifi city [ 15 ] , fi ve were performed in China (differential verifi cation bias, question 6), values of urine survivin tests that were [ 17,19,21,24,27 ] , two were conducted in and in all studies, disease progression obtained from the random effects model are Germany [ 18,26 ] , two were carried out in (question 4) and incorporation bias shown in Fig. 2 . The results showed that the

pooled sensitivity for the urine survivin test was 0.772 (95% CI 0.745 – 0.797) and the pooled speciﬁ city was 0.918 (95% CI 0.899 – 0.934). The pooled diagnostic OR for urine survivin tests was 49.666 (95% CI 22.90 – 107.72). The pooled positive and negative LRs of urine survivin tests were 8.341 (95% CI 4.821 – 14.431) and 0.245 Gold standard Interpretation (95% CI 0.176 – 0.342), respectively. The summary ROC curves and the * Q index of urine survivin tests for bladder cancer detection are shown in Fig. 3 . Because of the heterogeneity and lack of a threshold effect, we chose the random-effects model to synthesize the ROC curves. The area under the curve (AUC) of urine survivin tests was 0.9392 and the * Q index estimates for these tests were 0.876.

Urine collection Detection method SUBGROUP ANALYSIS

Six of the 14 studies also included data from cytology tests [ 16,18,21,24,25,27 ] . The Study design results of the subgroup analysis that we performed for urine survivin tests and cytology are presented in Table 2 [ 14 – 27 ] . The sensitivity, diagnostic OR, and * Q index

Patient Patient enrolment values of the urine survivin tests were higher than those of cytology. By contrast, the speciﬁ city, positive LR and negative LR for cytology were higher than those of urine survivin tests. The pooled positive LRs were Patient age, Patient years (range) ≈ 12 for urine survivin tests and ≈ 25 for the cytology tests. The pooled negative LRs were

Male (%) 0.27 for urine survivin tests and 0.60 for the cytology tests, respectively. The AUC estimate was signiﬁ cantly larger for urine survivin tests than for cytology tests. Patient/ Patient/ control no. The results of another subgroup analysis performed for survivin mRNA in urine by RT-PCR are also presented in Table 2 . The Place of study pooled estimates were higher than those of the original results, although the corresponding 95% CIs predominantly overlapped with each other. Recruitment period DISCUSSION

Although the current standard of care at

Year of Year publication most institutions consists of urine cytology and cystoscopy, we have not found a speciﬁ c and sensitive non-invasive tool to detect primary and recurrent bladder cancer. . [ 26 ] 2010 na Germany 32/17 75.0 70.0 (na) na Prospective Voided Real-time RT-PCR Cystoscopy na al

Cystoscopy is currently considered the . [ 15 ] 2004 na Japan 40/9 82.5 69.0 (46 – 87) na na na RT-PCR Cystoscopy na . [ 20 ] 2006 1994 – 2000 Egypt 84/83 na na na na Catheter RT-PCR Cystoscopy na . [ 18 ] 2005 1999 – 2001 Germany 35/33 na 68.0 (na) Consecutive Prospective Voided Real-time RT-PCR Cystoscopy na . [ 23 ] 2007 na USA 24/94 na na na na Voided Real-time PCR Cystoscopy na . [ 16 ] 2004 na USA 117/92 73.5 73.5 (40 – 94) na Prospective Voided Bio-Dot Cystoscopy Blind et al . [ 14 ] dataset 1 . [ 14 ] dataset 2 2001 2001 na na USA USA 31/107 15/5 na na 69.7 (na) 69.7 (na) na na na na Catheter Catheter Bio-Dot, RT-PCR Cystoscopy na Cystoscopy na al . [ 17 ] 2004 na China 30/36 73.3 48.5 – 76) (41 na na na Nested RT-PCR Cystoscopy na al al

al . [ 22 ] 2006 na Iran 20/18 95.0 60.0 (17 – 82) na na na RT-PCR Cystoscopy na

. [ 27 ] dataset 1 . [ 27 ] dataset 2 2010 2010 2005 – 2006 na China 35/35 China 80.0 103/84 60.0 (42 – 82) 84.5 na na na na Voided na Molecular beacon, ELISA Cystoscopy Voided Blind Molecular beacon Cystoscopy Blind

. [ 25 ] 2010 na Egypt 166/212 na 60.0 (26 – 83) na na Voided RT-PCR, nested RT-PCR Cystoscopy na The main characteristics of the included studies of the included studies The main characteristics

reference standard for the detection and . [19 ] 2006 2004 China 40/30 75.0 66.0 – 81) (51 na na Voided Real-time RT-PCR Cystoscopy na al al al . ] [ 21 2006 2003 – 2005 China 151/100 70.2 59.8 (33 – 84) na na Voided ELISA Cystoscopy na al et

al al et et al et

. [ 24 ] 2008 2003 – 2006 China 115/58 69.6 61.9 (35 – 81) Consecutive na Voided RT-PCR Cystoscopy Blind et 1

et et

et al et et et et surveillance of bladder cancer. However,

et et et many researchers have evaluated multiple TABLE TABLE Study na, not. Smith Smith Ohsawa Shariat Wang Wang Weikert Weikert Hou Moussa Sun Kenney Pu Ziaee Eissa Horstmann Zhao Zhao different assays to ﬁ nd an ideal urine-based

bladder tumour marker. The Food and Drug A FIG. 2. Sensitivity, 95% Cl Administration has approved several tests, Smith et al. [14] (dataset 1) 1.00 (0.89—1.00) A , Meta-analysis forest map of Smith et al. [14] (dataset 2) 1.00 (0.78—1.00) but these tests have not made a signifi cant Ohsawa et al. [15] 0.43 (0.27—0.59) sensitivity of urine-based Shariat et al. [16] 0.64 (0.55—0.73) impact on clinical practice to date, because Wang et al. [17] 0.80 (0.61—0.92) survivin test for the diagnosis of Weikert et al. [18] 0.69 (0.51—0.83) of inadequate performance. There is a need Hou et al. [19] 1.00 (0.91—1.00) bladder cancer. B, Meta-analysis Moussa et al. [20] 0.94 (0.87—0.98) Sun et al. [21] 0.70 (0.62—0.77) forest map of specifi city of to fi nd a highly specifi c and sensitive Ziaee et al. [22] 0.90 (0.68—0.99) Kenney et al. [23] 0.79 (0.59—0.93) urine-based survivin test for the bladder cancer marker to improve detection Pu et al. [24] 0.90 (0.84—0.95) Eissa et al. [25] 0.76 (0.69—0.82) accuracy, reduce patient morbidity and Horstmann et al. [26] 0.53 (0.35—0.71) diagnosis of bladder cancer. Zhao et al. [27] (dataset 1) 0.80 (0.63—0.92) inconvenience, and decrease the follow-up Zhao et al. [27] (dataset 2) 0.76 (0.66—0.84) fi nancial burden. Pooled Sensitivity = 0.77 (0.74 to 0.80) The clinical role of urine survivin tests has Chi-square = 125.66; df = 15 (P < 0.001) 0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 88.1% not been defi ned. The present study showed Sensitivity that urine survivin tests are not suffi ciently B superior at present to replace cystoscopy as Sensitivity, 95% Cl Smith et al. [14] (dataset 1) 0.93 (0.87—0.97) the reference standard because urine Smith et al. [14] (dataset 2) 1.00 (0.48—1.00) < Ohsawa et al. [15] 0.89 (0.52—1.00) survivin tests have a sensitivity 80%. Shariat et al. [16] 0.93 (0.86—0.98) Wang et al. [17] 1.00 (0.90—1.00) Vriesema et al . [ 28 ] reported that 89% of Weikert et al. [18] 1.00 (0.89—1.00) Hou et al. [19] 0.87 (0.69—0.96) the patients would prefer cystoscopy as the Moussa et al. [20] 0.95 (0.88—0.99) Sun et al. [21] 0.85 (0.76—0.91) diagnostic method, if the sensitivity of a Ziaee et al. [22] 0.50 (0.26—0.74) Kenney et al. [23] 0.93 (0.85—0.97) bladder tumour marker was < 90%. However, Pu et al. [24] 0.97 (0.88—1.00) Eissa et al. [25] 1.00 (0.98—1.00) there are good reasons to consider the use Horstmann et al. [26] 0.88 (0.64—0.99) Zhao et al. [27] (dataset 1) 0.77 (0.60—0.90) of urine survivin tests to replace or Zhao et al. [27] (dataset 2) 0.79 (0.68—0.87) complement cytology for detecting and monitoring bladder cancer. The systematic Pooled Specificity = 0.92 (0.90 to 0.93) Chi-square = 102.38; df = 15 (p = 0.0000) review of the literature for urine survivin 0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 85.3% tests showed that the sensitivity for urine Specificity survivin tests is better than cytology. Therefore, although urine survivin tests may not be ready to replace cystoscopy, there Symmetric SROC FIG. 3. may be a role for modifi ed protocols that AUC = 0.9392 Summary ROC curve of = increase the intervals between cystoscopies. SE(AUC) 0.0215 urine-based survivin test for the Q* = 0.8764 In addition, a combination of cytology and diagnosis of bladder cancer. SE(Q*) = 0.0270 urine survivin tests might be suitable for the Sensitivity SROC Curve triage of patients with symptoms of bladder 1 cancer, but we have no data from the 0.9 current meta-analysis to evaluate specifi c 0.8 test combinations because the present analysis included a very limited number of 0.7 studies [ 21,24,25 ] . 0.6 0.5 In the present meta-analysis, we used a bivariate random effects approach to 0.4 summarise estimates of accuracy and LRs, as 0.3 well as a summary ROC curve as outcome. 0.2 An advantage of this bivariate random effects model is that the random effects 0.1 0 nature allows systematic and coincidental 0 0.2 0.4 0.6 0.8 1 differences between studies. LRs summarise 1-specificity information about the test by combining sensitivity and specifi city, and allow estimation of the post-test probability of a tested condition. Further, LRs are intuitive; generate moderate shifts in those a p ositive test rules in pathology), and because LRs indicate how much a given probabilities, values of 2 – 5 generate small hence confi rm the presence of disease [ 31 ] . diagnostic test result will raise or lower the (but sometimes important) changes in Applying these criteria to the present pre-test probability of the target disorder probability, and values of 0.5 – 2.0 alter results showed that a positive urine survivin [ 29 ] . LRs > 10 are estimated to generate probabilities to a small and rarely important test result moderately increased the large, often conclusive changes from pre- to degree [ 30 ] . Tests with high positive LRs may probability of the presence of any stage post-test probabilities; values of 5 – 10 be marked as SpPIn (if sp ecifi city is high, of bladder cancer (positive LR 8.34), and

TABLE 2 Pooled diagnostic performance of urine survivin tests and cytology

Urine survivin * Cytology * Urine survivin † Urine survivin mRNA by RT-PCR ‡ Pooled sensitivity 0.748 (0.712– 0.782) 0.433 (0.393– 0.474) 0.772 (0.745– 0.797) 0.804 (0.770 – 0.835) Pooled speciﬁ city 0.942 (0.918 – 0.960) 0.983 (0.968 – 0.992) 0.918 (0.899 – 0.934) 0.951 (0.931 – 0.966) Pooled positive LR 12.378 (4.291 – 35.701) 25.767 (5.048 – 131.520) 8.341 (4.821 – 14.431) 12.608 (4.300 – 36.969) Pooled negative LR 0.267 (0.193– 0.370) 0.598 (0.534– 0.670) 0.245 (0.176– 0.342) 0.206 (0.120 – 0.354) Pooled DOR 52.984 (15.439 – 181.83) 41.430 (9.291 – 184.7) 49.666 (22.900 – 107.72) 84.343 (27.779 – 256.08) Summary ROC AUC ( se ) 0.862 (0.069) 0.626 (0.132) 0.939 (0.022) 0.954 (0.022) * Q index ( se ) 0.792 (0.067) 0.595 (0.102) 0.876 (0.027) 0.896 (0.031)

DOR, diagnostic OR; SE , standard error. * References: [ 16,18,21,24,25,27 ] ; † References: [ 14 – 27 ] ; ‡ References: [ 14,15,17 – 20,2 2– 26 ] .

also a negative result reduced this remains to be elucidated due to the limited prognosis of patients with stages Ta and probability by a small but signifi cant amount number of patients and heterogeneities in T1 transitional cell cancer of the bladder (negative LR 0.25). series. Prospective validation would be followed for at least 20 years . J Urol required to assess and confi rm its clinical 1995 ; 153 : 1823 – 6 One of the limitations of all meta-analyses value. While the present meta-analaysis 5 Lokeshwar VB , Habuchi T , Grossman is that the quality of the original studies suggests a potential benefi t for urine HB et al . Bladder tumor markers beyond affects the validity of the fi nal results. The survivin tests for the detection of bladder cytology: International Consensus Panel quality of the original studies included here cancer, their value for predicting cancer on bladder tumor markers . Urology was weak. Only a few studies were identifi ed recurrence and progression remains unclear. 2005 ; 66 ( Suppl . 1 ): 35 – 63 as having a consecutive series of patients There was no clear evidence of a difference 6 Raitanen MP , Aine R , Rintala E et al . suspected for bladder cancer with between cytology and urine survivin tests Differences between local and review independent assessment of the marker test for the prediction of tumour recurrence and urinary cytology in diagnosis of bladder and reference standard. Most studies were progression in the longer term. cancer. An interobserver multicenter retrospective studies that were not ‘ blinded ’. analysis . Eur Urol 2002 ; 41 : 284 – 9 It remains unclear in most studies whether 7 Altieri DC . Molecular circuits of or not study inclusion was offered to all CONFLICT OF INTEREST apoptosis regulation and cell division patients on a consecutive basis. In almost all control: the survivin paradigm . J Cell studies, the rationale for the sample size None declared. Biochem 2004 ; 92 : 656 – 63 remained unclear. The exclusion criteria used 8 Margulis V , Lotan Y , Shariat SF . were different in the various studies. Many Survivin: a promising biomarker for studies included healthy patients as a REFERENCES detection and prognosis of bladder control group. When larger numbers of cancer . World J Urol 2008 ; 26 : 59 – 65 healthy patients were used, the specifi city 1 Jemal A , Siegel R , Xu J , Ward E . 9 Adida C , Crotty PL , McGrath J , for the tumour markers was higher. Because Cancer statistics, 2010 . CA Cancer J Clin Berrebi D , Diebold J , Altieri DC . all detection techniques can be affected by 2 0 1 0 ; 60 : 277 – 300 Developmentally regulated expression of the processing and storage time of samples, 2 Sylvester RJ , van der Meijden AP , the novel cancer anti-apoptosis gene RNA quality, the sources of diagnostic Oosterlinck W et al . Predicting survivin in human and mouse reagents and instruments, and operators, recurrence and progression in individual differentiation . Am J Pathol 1998 ; 152 : setting standard operation procedures and patients with stage Ta T1 bladder cancer 43 – 9 techniques is necessary. As yet, MRNA using EORTC risk tables: a combined 10 Ambrosini G , Adida C , Altieri DC . A measurement is not standardised, and analysis of 2596 patients from seven novel anti-apoptosis gene, survivin, clearly defi ned threshold levels for survivin EORTC trials . Eur Urol 2006 ; 49 : 466 – 5 expressed in cancer and lymphoma . Nat measurement do not exist. 3 Babjuk M , Oosterlinck W , Sylvester R , Med 1997 ; 3 : 9 1 7 – 2 1 Kaasinen E , B ö hle A , Palou-Redorta J , 11 Whiting P , Rutjes AW , Reitsma JB , CONCLUSIONS European Association of Urology Bossuyt PM , Kleijnen J . The (EAU) . EAU guidelines on non-muscle- development of QUADAS: a tool for the We systematically reviewed the available invasive urothelial carcinoma of the quality assessment of studies of evidence of urine survivin tests for the bladder . Eur Urol 2008 ; 54 : 303 – 14 diagnostic accuracy included in detection of bladder cancer. The clinical 4 Holm ä ng S , Hedelin H , Anderströ m C , systematic reviews . BMC Med Res utility of urine survivin as a bladder tumour Johansson SL . The relationship among Methodol 2003 ; 3 : 25 marker identifi ed in the present study multiple recurrences, progression and 12 Higgins JP , Thompson SG , Deeks JJ ,

Altman DG . Measuring inconsistency in 20 Moussa O , Abol-Enein H , Bissada NK , 27 Zhao J , Wang ZQ , Wang XY , Yang XJ , meta-analyses . BMJ 2003 ; 327 : 557 – 60 Keane T , Ghoneim MA , Watson DK . He D . Preliminary study of diagnostic 13 Zamora J , Abraira V , Muriel A , Khan Evaluation of survivin reverse utility of molecular beacons in bladder K , Coomarasamy A . Meta-DiSc: a transcriptase-polymerase chain reaction cancer . Urology 2 0 1 0 ; 76 : 512.e8 – 13 software for meta-analysis of test for noninvasive detection of bladder 28 Vriesema JL , Poucki MH , Kiemeney LA , accuracy data . BMC Med Res Methodol cancer . J Urol 2006 ; 175 : 2312 – 6 Witjes JA . Patient opinion of urinary 2006 ; 6 : 3 1 21 Sun Y , He DL , Ma Q et al . Comparison tests versus fl exible urethrocystoscopy in 14 Smith SD , Wheeler MA , Plescia J , of seven screening methods in the follow-up examination for superfi cial Colberg JW , Weiss RM , Altieri DC . diagnosis of bladder cancer . Chin Med J bladder cancer: a utility analysis. Urology Urine detection of survivin and diagnosis (Engl) 2006 ; 119 : 1763 – 71 2000 ; 56 : 793 – 7 of bladder cancer . JAMA 2001 ; 285 : 22 Ziaee SA , Moula SJ , Hosseini 29 Deeks JJ , Altman DG . Diagnostic tests 324 – 8 Moghaddam SM , Eskandar-Shiri D . 4: likelihood ratios . BMJ 2004 ; 329 : 15 Ohsawa I , Nishimura T , Kondo Y et al . Diagnosis of bladder cancer by urine 168 – 9 Detection of urine survivin in 40 survivin, an inhibitor of apoptosis: a 30 Jaeschke R , Guyatt GH , Sackett DL . patients with bladder cancer . J Nippon preliminary report . Urol J 2006 ; 3 : 150 – 3 Users ’ guides to the medical literature. Med Sch 2004 ; 71 : 379 – 83 23 Kenney DM , Geschwindt RD , Kary MR , III. How to use an article about a 16 Shariat SF , Casella R , Khoddami SM Linic JM , Sardesai NY , Li ZQ . Detection diagnostic test. B. What are the results et al . Urine detection of survivin is a of newly diagnosed bladder cancer, and will they help me in caring for my sensitive marker for the noninvasive bladder cancer recurrence and bladder patients? The Evidence-Based Medicine diagnosis of bladder cancer . J Urol 2004 ; cancer in patients with hematuria using Working Group . JAMA 1994 ; 271 : 703 – 171 : 626 – 30 quantitative rt-PCR of urinary survivin . 7 17 Wang H , Xi X , Kong X , Huang G , Ge Tumour Biol 2007 ; 28 : 57 – 62 31 Pewsner D , Battaglia M , Minder C , G . The expression and signifi cance of 24 Pu XY , Wang ZP , Chen YR , Wang XH , Marx A , Bucher HC , Egger M . Ruling a survivin mRNA in urinary bladder Wu YL , Wang HP . The value of diagnosis in or out with ‘ SpPIn ’ and carcinomas . J Cancer Res Clin Oncol combined use of survivin, cytokeratin 20 ‘ SnNOut ’ : a note of caution . BMJ 2004 ; 2004 ; 130 : 487 – 90 and mucin 7 mRNA for bladder cancer 329 : 209 – 13 18 Weikert S , Christoph F , Schrader M , detection in voided urine . J Cancer Res Krause H , Miller K , M ü ller M . Clin Oncol 2008 ; 134 : 659 – 65 Correspondence: Seth P. Lerner, Scott Quantitative analysis of survivin mRNA 25 Eissa S , Swellam M , Shehata H , Department of Urology, Baylor College of expression in urine and tumor tissue of El-Khouly IM , El-Zayat T , El-Ahmady Medicine, 6620 Main St., Suite 1325, bladder cancer patients and its potential O . Expression of HYAL1 and survivin Houston, TX 77030, USA. relevance for disease detection and RNA as diagnostic molecular markers for e-mail: [email protected] prognosis . Int J Cancer 2005 ; 116 : 100 – 4 bladder cancer . J Urol 2 0 1 0 ; 183 : 493 – 8 19 Hou JQ , He J , Wen DG , Chen ZX , Zeng 26 Horstmann M , Bontrup H , Abbreviations : IAP , inhibitor of apoptosis J . Survivin mRNA expression in urine Hennenlotter J et al . Clinical experience protein ; ROC , receiver-operating as a biomarker for patients with with survivin as a biomarker for characteristic ; LR , likelihood ratio ; OR , odds transitional cell carcinoma of bladder . urothelial bladder cancer . World J Urol ratio ; RT , reverse transcriptase ; AUC , area Chin Med J (Engl) 2006 ; 119 : 1118 – 20 2 0 1 0 ; 28 : 399 – 404 under the curve .