Utility of urinary biomarkers in primary haematuria: Systematic review and meta‐analysis

Abstract Objectives To evaluate the diagnostic performance of FDA‐approved urinary biomarkers in the evaluation of primary haematuria for investigation of bladder cancer. Methods The scientific databases MEDLINE, EMBASE, Pubmed and Web of Science were searched to collect studies. Studies that evaluated the diagnostic performance of FDA‐approved urinary biomarkers in investigating patients with primary haematuria without a prior history of bladder cancer were included. Quality of studies was assessed using the JBI Criteria. Bivariate mixed‐effects regression model was used to calculate pooled sensitivities and specificities for each biomarker. Results Eighteen studies were included in the analysis. The biomarkers assessed in these studies were CxBladder, AssureMDx, Bladder Tumour Antigen (BTA), NMP22, UroVysion and Immunocyt/uCyt+. Several biomarkers, such as AssureMDx, CxBladder and Immunocyt, were shown to have better diagnostic performance based on their sensitivity, specificity and diagnostic odds ratio, as well as positive and negative likelihood ratios. Across the six biomarkers, sensitivity ranged from 0.659 to 0.973, and the specificity ranged between 0.577 and 0.833. Conclusion Despite certain biomarkers demonstrated better performance, current diagnostic abilities of the FDA‐approved biomarkers remain insufficient for their general application as a rule out test for bladder cancer diagnosis and as a triage test for cystoscopy in patients with primary haematuria. High‐quality prospective studies are required to further analyse this and also analyse the correct scenario in which urinary biomarkers may be best utilised.

known to have high sensitivities of up to 82% and 100% and high specificities of up to 94% and 97%, respectively. 3 Recent technological developments, such as the introduction of narrow-band imaging and blue-light cystoscopy, have further improved diagnostic capacity of cystoscopy, especially for non-invasive lesions and carcinoma in situ. At present, urine cytology remains the most routinely used urinary marker to investigate for malignancy. However, it has been attributed to low sensitivities of 34-55% with even reduced sensitivity in low grade tumours and considerable inter-and intra-observer variabilities, despite relatively high specificity of approximately 90%. 4 Haematuria itself is a common clinical presentation, occurring in up to 9-18% of the population. It is not pathognomonic for bladder cancer with an approximately 12% patients being investigated for haematuria confirmed to have bladder cancer. 5 This highlights significant issues about how patients should be evaluated as current investigations may be invasive, uncomfortable and morbid, all whilst failing to produce a satisfactory diagnostic yield. One approach is to utilise the diagnostic capacity of urinary markers to determine when to per-  Specimens with more than one green or red urothelial cell are considered to be positive. [8][9][10][11] The bladder tumour antigen (BTA) test employs monoclonal antibodies to detect elevated levels of complement factor H-related protein (CFHrp) in voided urine, which is a degradation product of the basement membrane shown to be released by malignant cells in culture. There are two types of BTA tests, including the qualitative BTA Stat and the quantitative BTA Trak, which is an enzyme-linked immunosorbent assay. [12][13][14][15] Nuclear matrix proteins (NMP) are a group of proteins which provide a structural framework to the nucleus and are involved in DNA replication and RNA synthesis. The NMP22 test in particular detects nuclear mitotic apparatus protein which has been shown to be more abundant in malignant urothelial cells. Upon apoptosis, nuclear mitotic apparatus proteins can be detected in the urine at significantly elevated levels than normal. There are two tests to detect NMP22 levels, including the original quantitative sandwich type immunoassay and the qualitative BladderChek test. 11,12,[16][17][18][19][20][21] AssureMDx isolates DNA from urine samples and analyses these for three mutation genes (FGFR3, TERT and HRAS) and three methylation genes (OTX1, ONECUT2 and TWIST1). 22 UroVysion is another fluorescence-based assay that uses fluorescence in situ hybridisation (FISH) to observe multiple different chromosomal copy numbers and DNA sequences in cell nuclei derived from a urine sample. Various genetic alterations are examined for in this test, including aneuploidy of chromosomes 3, 7 and 17, and the loss of the 9p21 locus, which are 4 chromosomal changes frequently associated with urothelial carcinoma. 11,23 CxBladder extracts and quantifies five mRNA biomarkers (MDA, HOCXA13, CDC2, IGFBP5 and CXCR2) known to be differentially expressed in malignant cells than in normal cells. These biomarkers are quantified in urine samples using reverse transcriptase quantification polymerase chain reaction. 24  "bladder carcinoma," "bladder malignancy," "bladder neoplasm," "bladder tumour," "bladder tumor," "haematuria," "hematuria," "biomarker" or "urinary biomarker," including for each of the FDAapproved biomarkers separately. Boolean operators were utilised to combine the sets of searches. Separate searches were performed by two independent authors (N.S. and D.G.) performing title and abstract screening independent of each other according to our inclusion and exclusion criteria. After collating search results and removing duplicates from the respective searches, full texts of the relevant articles were then reviewed for their quality using the Joanna Briggs Institute (JBI) Criteria for Diagnostic Test Accuracy Studies. Full text articles that were unable to be accessed or obtained through our institutions were excluded. Where necessary, disagreements were resolved in consultation with a senior author (K.S.).

| Inclusion and exclusion criteria
Prospective studies written in English and performed on adult patients of 18 years of age and above who presented with primary macroscopic haematuria without prior diagnosis of bladder cancer were included in our analysis. Patients in the studies needed to be tested with at least one of the FDA-approved biomarkers in addition to cystoscopy, either rigid or flexible. Prospective and retrospective studies were included. No restrictions on date of publications were imposed on the studies. Studies whose patients presented with microscopic haematuria, demonstrated recurrence of bladder cancer, and those without any follow-up cystoscopy were excluded.

| Statistical analysis
For each included study in the meta-analysis, Sensitivity, specificity, positive and negative predictive values were calculated. Pooled sensitivities and specificities and subsequent comparisons of biomarkers were calculated using a bivariate random-effects regression model of meta-analysis as described by Reitsma et al. and Harbord et al. using the restricted maximum likelihood (REML) method of estimating variance. 25,26 Statistical analysis was performed in R (version 3.4, R foundation for statistical Computing, Vienna, Austria). The "mada" package in R was used to design forest plots for pooled specificity, sensitivity and diagnostic odds ratios based on individual biomarkers and individual studies. The analysis was plotted using summary receiver operating characteristics (SROC) curves in R. 27

| RESULTS
A flowchart of our search based on the PRISMA framework was as depicted in Figure 1. Our initial search of the literature on available database yielded 2222 titles/abstracts review, from which 579 were duplicate records and hence were excluded. Following title and abstract screen from 1643 records, 123 articles were selected for fulltext review. Unfortunately, one was not accessible to facilitate fulltext review and was excluded. 28 Seventeen articles met our inclusion criteria, [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24] where results pertaining to each biomarker were extracted with their characteristics were as presented in Table 1 and the calculated sensitivity, specificity and diagnostic odds ratio, as well as the positive and negative likelihood ratio for each of the extracted biomarker as shown in Table 2.
Compared with other biomarkers, NMP22 was discussed in most with eight studies, 11,12,16-21 followed by BTA [12][13][14][15] and Immunocyt [8][9][10][11] in four studies each, CxBladder 24 and UroVysion 11,23 in two studies each, and AssureMDx in one study. Where reported, the age of patients included ranged from 18 to 97 years, with medians ranging from 59 to 69 years. The number of patients included with histopathological diagnosis of bladder cancer varied from one to 245 patients, which comprised 4-73% of the respective recruited cohorts. The sensitivities, specificities, PPV and NPV were as presented in Table 1 Four studies relating to Immunocyt met our inclusion criteria. [8][9][10][11] The pooled sensitivity from the four studies was 0.844 (95% CI T A B L E 1 Characteristics of included studies separated into the different urinary markers each with the calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) Two studies pertaining to the use of Urovysion for investigation of primary haematuria were included 11 and 57-97%, respectively. This was dependent on the different types of cystoscopic adjuncts applied such as blue light and narrow band imaging, which were shown to further improve diagnostic accuracy. 3 Despite the advances, the complication profiles including for discomfort, invasive nature, urinary tract infection, and lower urinary tract symptoms, such as frequency, dysuria, and haematuria, may not support the argument towards the regularity of its use. 29 Hence, the inclusion of urine cytology and urinary markers in the diagnostic algorithm may serve as a triage test for higher risk patients to proceed for cystoscopy.
T A B L E 2 Sensitivity, specificity, diagnostic odds ratio, positive likelihood ratio, and negative likelihood ratio analysed for each urinary biomarker extracted from the included studies  as well as radiation changes. In addition, intra-and inter-observer variability impacting on reproducibility of the results are important to acknowledge. 10,11,32,33 Several studies included in our meta-analysis also reviewed the performance of urine cytology, which again demonstrated the good specificity ranging from 0.945 to 1 but with sensitivity between 0.211 to 0.66. 10,13,15,[18][19][20][21]24