MRI pathway and TRUS‐guided biopsy for detecting clinically significant prostate cancer
Abstract
This is a protocol for a Cochrane Review (Diagnostic test accuracy). The objectives are as follows:
To determine the diagnostic accuracy of, and the concordance between, the MRI pathway and TRUS‐guided biopsy approach for detecting clinically significant prostate cancer in men with a suspicion of prostate cancer, using template‐guided biopsy as the reference standard.
Background
Target condition being diagnosed
Prostate cancer is the most frequently diagnosed solid cancer among men in more developed countries, and after lung cancer the second most frequently diagnosed cancer in men worldwide (Torre 2015). Prostate cancer is the sixth leading cause of cancer death (7.4% of deaths) among men worldwide (Center 2012). However, autopsy studies of men who died of other causes show the existence of a substantial reservoir of latent prostate cancer: ranging from 5% at age < 30 years, to 20% at the age of 50 years and 59% at age > 79 years (Bell 2015). This indicates that a large proportion of prostate cancers are clinically insignificant and will not lead to any complaints or death if left undetected. Still, if clinically insignificant prostate cancer is detected it can be managed by active surveillance and does not necessarily need direct treatment. In contrast, clinically significant prostate cancer has direct therapeutic implications as it may progress, metastasize and lead to prostate cancer specific mortality.
The harm of overdiagnosis mainly lies in overtreatment, as many patients are offered a radical prostatectomy or radiotherapy. The rate of patients with no previous complaints, who report the loss of erections firm enough for intercourse after treatment, is as high as 50% and accompanied by frequent urinary problems, throughout six years of follow‐up (Donovan 2016). Given the sharp increase in prostate‐specific antigen (PSA) testing, prostate cancer diagnoses and the increasing concerns of overdiagnoses and overtreatment as described by Ilic 2013, the distinction between clinically significant and insignificant prostate cancer has become more important. Defining clinically (in‐)significant prostate cancer remains difficult however, and definitions vary in the world literature (Moore 2013a). Established definitions are based on transrectal ultrasound (TRUS)‐guided biopsy and are not limited to histologic parameters, which makes comparing literature difficult. Parameters frequently taken into account are PSA and derivatives, familial history, ethnic profile, Gleason grading as described by Epstein 2010, number or percentage of positive cores and length of cancer in a core (percentage of core involvement) (Epstein 1994; Goto 1996; Harnden 2008; Wolters 2011). Moreover, other clinical parameters such as age and comorbidity also influence the potential for progression and mortality of the individual patient.
Furthermore, with the introduction of multi‐parametric magnetic resonance imaging (MRI) of the prostate and MRI‐targeted biopsy in the diagnostic process, the histological definition of clinically significant prostate cancer based on TRUS‐guided biopsy does not apply any more. As a result of the intention to oversample areas of high suspicion, MRI‐targeted biopsies tend to show longer cancer core length and higher Gleason grading than TRUS‐guided biopsies (Haffner 2011). This results in a drift towards higher risk classification, which is an artefact of the MRI‐targeted sampling method and may prompt men and physicians towards more radical treatment. Based on these assumptions, the International Working Group on Standards of Reporting for MRI‐targeted biopsy studies (START) agreed that definitions of clinical significance in MRI‐targeted biopsy studies should solely focus on histologic definitions (Gleason grade and maximum cancer core length) (Moore 2013a).
Clinical pathway
Opportunistic PSA‐based screening is practiced worldwide and men considered to be at risk of prostate cancer (elevated PSA level, abnormal digital rectal examination, ethnic origin and heredity) are generally advised to have a TRUS‐guided biopsy (Carter 2013; Carroll 2016; Mottet 2017). TRUS‐guided biopsy may be repeated several times in the case of persistent suspicion of clinically significant prostate cancer after previous negative biopsy or during active surveillance of low risk prostate cancer.
Any prostate biopsy is associated with a risk of infection (1% to 8%) and an increasing risk of life‐threatening sepsis (1% to 4%), as a consequence of increasing antibiotic resistance (Loeb 2013; Borghesi 2016). Other associated morbidities include dysuria, hematospermia, haematuria, rectal bleeding, vasovagal episodes and urinary retention (Djavan 2001; Loeb 2013).
Compared to TRUS‐guided biopsy, MRI‐targeted biopsy is only performed when suspected lesions for clinically significant prostate cancer are detected on MRI. Due to selective performance of targeted biopsies the MRI with MRI‐targeted biopsy is able to more accurately detect high‐risk prostate cancer while detecting less low‐risk prostate cancer (Schoots 2015; Siddiqui 2015). Therefore MRI and MRI‐targeted biopsy are increasingly investigated in addition to or as a replacement for TRUS‐guided biopsy, whether in the setting of initial biopsy or repeated biopsy or during active surveillance. Guidelines recommend consideration of the use of MRI, if it is available, especially in the context of persistent clinical suspicion of prostate cancer after previous negative biopsy (Barentsz 2012; Rosenkrantz 2016; Mottet 2017).
Index tests
TRUS‐guided biopsy
TRUS‐guided biopsy is a technique in which the peripheral zone of the prostate is randomly sampled, generally by 8 to 19 systematic TRUS‐guided biopsies, depending on the size of the prostate. TRUS is performed primarily for anatomic guidance, as suspicious lesions for prostate cancer in general cannot be visualised by ultrasound. This approach may therefore result in random and systematic errors, which can lead to hitting insignificant lesions while missing significant lesions (El‐Shater Bosaily 2015). The estimated false negative rate of TRUS‐guided biopsy for any cancer is 25% to 40% (Hu 2012). Also misclassification occurs by not hitting the cancer lesion at its greatest diameter or highest grade, as reclassification is seen in almost half of men when a more accurate biopsy test is applied (Barzell 2007; Taira 2010; Barzell 2012; Taira 2013).
MRI pathway
The MRI pathway comprises two components and can be defined as 1) the performance of an MRI resulting in a suspicion score for the presence of clinically significant prostate cancer, and 2) the performance of only MRI‐targeted biopsies if a suspicious lesion is seen. The MRI pathway is the subject of ongoing research with the intent to improve the detection of clinically significant prostate cancer while reducing the drawbacks of TRUS‐guided biopsy.
Different MRI techniques and MRI systems from different vendors are used worldwide. Techniques to identify and locate suspicious lesions for significant prostate cancer are T2‐weighted imaging (T2W), diffusion weighted imaging (DWI), dynamic contrast enhanced (DCE) imaging and spectroscopy. Different MRI systems (i.e. 1.0, 1.5 and 3.0 Tesla) from different vendors (i.e. General Electric, Philips, Siemens) exist.
In addition, different scoring systems for the suspicion of prostate cancer on MRI have been developed. Some radiologists classify lesions in the prostate as low, moderate, to high suspicion of prostate cancer. Other radiologists use scoring systems from 1 to 5 according to the Likert scale principle; where the presence of clinically significant prostate cancer in a lesion is highly unlikely (1) to highly likely (5), as in the currently most often used Prostate Imaging ‐ Reporting and Data System (PI‐RADS) (Barentsz 2012; Weinreb 2016).
MRI‐targeted biopsy can either be performed with MRI guidance within the MRI scanner (direct in‐bore), or by ultrasound guidance with the use of computer‐based software that overlays the target identified on MRI onto the ultrasound image (software registration), or without the use of software (visual registration) (Moore 2013a). No significant differences in clinically significant prostate cancer detection appear to exist between these navigational approaches (Schoots 2015; Wegelin 2017).
Alternative test(s)
In 1989 the systematic ultrasound guided transrectal sextant prostate biopsy scheme was introduced by Hodge 1989, which replaced digital‐guided prostate biopsy and represented the reference standard until the late 1990s. Since then, different TRUS‐guided biopsy approaches (transrectal or transperineal) with different numbers of additional biopsy cores have been advocated as these influence diagnostic yield (Chun 2010). For example, transrectal saturation biopsy (defined as more than 20 biopsies of the prostate) has the intention of comprehensively sampling the prostate (Kuru 2013). However, most transrectal biopsy approaches do not sample the anterior zones of the prostate and therefore still lack accuracy. Another approach which samples the whole prostate according to a predefined template is the transperineal template‐guided mapping biopsy. This approach is considered to be the most accurate approach to diagnose all prostate cancer. However, intensified biopsy approaches are less frequently used in daily clinical practice due to the widespread idea that they are a high burden to patients and the risk of overdiagnosing insignificant prostate cancer (Jiang 2013).
Furthermore, different ultrasound imaging techniques for localising suspicious lesions in the prostate are also being developed and evaluated; these include contrast enhanced ultrasound, computer‐assisted TRUS, sonoelastography and histoscanning. However, these techniques need further development before considering potential application in daily clinical care (Pummer 2014).
Rationale
To reduce overdiagnosis and overtreatment of clinically insignificant prostate cancer we urgently need more accurate diagnostic methods and better risk stratification for detecting only clinically significant prostate cancer (Alberts 2015). The MRI pathway provides this as it selectively performs targeted biopsies and is able to more accurately detect high‐risk prostate cancer while detecting less low‐risk prostate cancer (Schoots 2015; Siddiqui 2015). Therefore the MRI pathway is increasingly used in daily clinical practice.
However, the methods to prove the diagnostic accuracy of the MRI with and without MRI‐targeted biopsy have so far been poor as a result of absent data for an appropriate reference test. The available systematic reviews on this topic to date, written by Moore 2013b, de Rooij 2014, van Hove 2014, Futterer 2015, Hamoen 2015, Schoots 2015, Valerio 2015 and Gayet 2016, have mainly included men with a positive MRI (suspicious lesion(s) on MRI), thus disregarding MRI‐negative patients. Inevitably, true negative and false negative values of the MRI pathway have not yet been properly ascertained. In addition, TRUS‐guided biopsy or radical whole‐mount surgical specimens have been used as reference standards, which inherently have a number of biases. TRUS‐guided biopsy may miss significant prostate cancer caused by both random and systematic errors. Radical whole‐mount surgical specimens, although accurate, are only available for patients with a positive biopsy who opted for surgery. Therefore patients with a negative MRI or negative biopsy and patients who opted for other treatment modalities, like radiation therapy or active surveillance, are disregarded, leading to considerable selection bias.
For these reasons, a systematic review and meta‐analysis according to the Cochrane principles and guidelines using a more appropriate reference standard is indicated, and will help us to further understand the role of the MRI pathway in clinical practice. The currently available data, investigating both MRI‐positive and ‐negative outcomes using template‐guided biopsy as a reference standard, now allow for more accurate comparisons. Increasing the number of biopsies in a systematic template‐guided approach of the prostate (including anterior zones) results in reduced random and systematic error, decreasing the false negative rates as demonstrated by using TRUS‐guided biopsy. Template‐guided biopsy using a uniform grid taken at 5 mm intervals, can technically only miss those tumours that are smaller than the distance between the adjacent cores (Ahmed 2011; Sivaraman 2015). The sensitivity and negative predictive value for detecting clinically significant cancers of volume ≥ 0.5 cm3 of this technique have both been shown to be 95%, with a sensitivity of 76% for detecting all cancers (Crawford 2005; Ahmed 2011; Crawford 2013; Simmons 2014). Therefore, template‐guided biopsy is a reference standard as close to radical whole‐mount surgical specimens as possible, but without the inherent selection bias of surgery. Although the diagnostic accuracy remains dependent on the intensity of cores taken and core trajectory, as described by Huo 2012, Pham 2015 and Valerio 2015, it is the optimal reference test.
Objectives
To determine the diagnostic accuracy of, and the concordance between, the MRI pathway and TRUS‐guided biopsy approach for detecting clinically significant prostate cancer in men with a suspicion of prostate cancer, using template‐guided biopsy as the reference standard.
Secondary objectives
To investigate what clinical and methodological sources of heterogeneity affect the accuracy of the tests, including type of population (initial biopsy or previous negative biopsy), type of prostate cancer (clinically significant or clinically insignificant), MRI scoring system, MRI suspicion score threshold for MRI‐targeted biopsy, navigational approach of MRI‐targeted biopsy, MRI lesion location, number of biopsy cores (or biopsy density) and core distribution in the reference standard.
Methods
Criteria for considering studies for this review
Types of studies
Retrospective and prospective cross‐sectional studies will be considered. Analysing the literature on the MRI pathway, we can distinguish studies reporting on only the first or both components of the MRI pathway, i.e. the MRI (suspicion score) and the performance of MRI‐targeted biopsy in the presence of a suspicious lesion.
Our primary objective is to analyse studies reporting both components of the MRI pathway. However, when only few studies reporting on both components are found to be eligible, we will consider also accepting studies which only report on the first component of the MRI pathway (i.e. the MRI suspicion score). We will then investigate the first and both components of the MRI pathway separately. Therefore, we will group all studies reporting on MRI suspicion score (with or without reporting MRI‐targeted biopsy) and refer to these as MRI pathway A) MRI suspicion score. Subsequently, we will distinguish only those studies that report on both components of the MRI pathway (MRI and MRI‐targeted biopsy in the presence of a suspicious lesion), and refer to these as MRI pathway B) MRI±MRI‐TBx.
Studies will be included if men, with a suspicion of prostate cancer, receive:
-
one or both components of the MRI pathway and the reference standard; or
-
TRUS‐guided biopsy and the reference standard; or
-
both components of the MRI pathway and TRUS‐guided biopsy.
Studies in which the combination of all three tests are compared either directly or indirectly as in a randomised trial of test accuracy, are considered to represent the highest level of evidence and will also be included. See Figure 1. We will exclude studies in which only MRI‐positive outcomes are investigated, as these studies disregard important information about MRI‐negative patients. Studies with insufficient data to extract a two‐by‐two table to assess the diagnostic accuracy (after contacting the primary author to obtain additional information, if necessary) will also be excluded. We will include studies regardless of their publication status or language of publication.
The MRI pathway comprises two components: MRI (suspicion score) and MRI‐targeted biopsy in the presence of suspicious lesions for clinically significant prostate cancer:
-
a) MRI suspicion score: refers to only the first component of the MRI pathway.
-
b) MRI±MRI‐TBx: refers to both components of the MRI pathway.
Review design
Meta‐analysis of studies reporting on:
The MRI pathway and the reference standard: ·
-
Comparison 1a: MRI suspicion score vs reference standard and all additional biopsies (e.g. MRI‐targeted biopsies).
-
Comparison 1b: MRI±MRI‐TBx vs reference standard.
TRUS‐guided biopsy and reference standard: ·
-
Comparison 2
MRI pathway B) MRI±MRI‐TBx and TRUS‐guide biopsy:
-
Comparison 3
A combination of all tests:
-
Comparison 4: comparing both index tests, either directly with the reference standard or indirectly as a randomised trial of test accuracy.
Participants
Two different groups of patients will be identified as participants: 1) men with an initial prostate biopsy because of clinical suspicion of prostate cancer based on increased PSA or abnormal digital rectal examination, or both; 2) men with a previous negative prostate biopsy with persistent clinical suspicion based on increased PSA or abnormal digital rectal examination, or both. This review will only focus on primary prostate cancer diagnosis. Patients already diagnosed with low‐grade prostate cancer who are on active surveillance will not be included, as insufficient data are available to properly investigate this particular group (Schoots 2015). Patients with recurrent disease will also be excluded, as this leads to multiple biases. There will be no age restriction.
Index tests
The MRI pathway
The MRI pathway is considered as the primary index test as this diagnostic test is novel and is considered to be a potential replacement for TRUS‐guided biopsy. All methods to identify suspicious lesions on MRI (i.e. T2W, DWI, DCE imaging and/or spectroscopy) and all MRI systems from all different vendors will be considered for analysis.
For targeting suspicious lesions we will use the study specified MRI scoring system targeted biopsy thresholds, without excluding any scoring system. The following thresholds are used: at a 5‐point Likert scale the suspicion score ≥ 3, at a 5‐point PIRADS scale the suspicion score ≥ 3, at a 3‐point scale the suspicion score ≥ 2).
Any MRI‐targeted biopsy approach is included. MRI‐targeted biopsy is usually performed by either MRI guidance (‘in bore’) or ultrasound guidance (‘visual/cognitive’ or ‘software’ registered) (Robertson 2013). The number of targeted biopsy cores per lesion will not be limited and will be defined by each study protocol. The two main biopsy approaches (transrectal and transperineal) are considered for analysis.
A positive MRI pathway will be defined as A) an MRI with a suspicion score equal to or greater than the MRI‐targeted biopsy threshold, or B) MRI±MRI‐TBx with the target condition in the MRI‐targeted biopsy cores. Whether the MRI pathway is true or false positive or negative, follows from histopathological confirmation of the target condition in the reference standard biopsies.
TRUS‐guided biopsy
TRUS‐guided biopsy is defined by the performance of a randomly predefined biopsy strategy of 8 to 19 biopsy cores, focusing on the peripheral zone of the prostate. The number of cores is arbitrarily limited to 19, as the definition of saturation biopsy comprises 20 cores and above (Kuru 2013). In clinical practice the number of cores taken is dependent on the volume of the prostate. Studies on ultrasound imaging techniques for localising suspicious lesions (e.g. contrast enhanced ultrasound) will be excluded. The two main biopsy approaches (transrectal and transperineal) will be considered for analysis.
The definition of a positive TRUS‐guided biopsy is histopathological confirmation of the target condition within the TRUS‐guided biopsy cores. Whether the TRUS‐guided biopsy is true or false positive or negative, follows from histopathological confirmation of the target condition in the reference standard biopsies.
Target conditions
The target condition is the detection of clinically significant prostate cancer. To overcome the difficulties in analysing the varying definitions, this review will use study specific definitions, and, if data allows, use Gleason grading 3 + 4 or higher, of any biopsy core length, as definition for clinically significant prostate cancer, as in the START recommendations (Moore 2013a). Furthermore, overall prostate cancer results will be analysed if data allows.
Reference standards
Histopathological diagnosis by template‐guided prostate biopsy will serve as the reference standard. In general, two different template‐guided biopsy techniques are in use: the transperineal template‐guided mapping biopsy (TTMB) and the template‐guided saturation biopsy (TSB). TTMB is defined as “transperineal TRUS‐guided biopsies of the prostate performed with the patient in lithotomy position using a 5‐mm brachytherapy grid, with at least one biopsy from each hole”; and TSB is defined as “20 or more transperineal or transrectal TRUS‐guided biopsies of the prostate performed with the intention to comprehensively sample the whole prostate, according to a predefined core distribution pattern” (Kuru 2013; Sivaraman 2015).
All TTMB and TSB techniques, which sample all (including the anterior) zones of the prostate will be considered for analysis. For the analysis of only the first component of the MRI pathway (the MRI suspicion score) any biopsy cores taken in addition to the reference biopsies, e.g. MRI‐targeted biopsies, will serve as reference biopsies as well. The number and distribution of cores taken will be defined by each study protocol and assessed as a quality property by our tailored Quality Assessment of Diagnostic Accuracy Studies (QUADAS)‐2 quality checklist (Appendix 1) and as a covariate in heterogeneity analysis.
The definition of a positive reference standard is histopathological confirmation of the target condition within the reference biopsy cores.
Search methods for identification of studies
Electronic searches
We will perform a comprehensive search with no restriction on the language of publication or publication status, in the following electronic databases (from inception to present): Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, Embase, Web of Science (Core Collection), Scopus, Pubmed, Google Scholar, WorldCat, ProQuest (ProQuest Dissertations & Theses), OpenGrey, ClinicalTrials and the WHO register. The search strategy for MEDLINE is provided in Appendix 2. For databases other than MEDLINE we will adapt the search strategy accordingly.
The systematic literature search will be conducted with the help of an expert information specialist (librarian) from the Medical Library, Erasmus MC University Medical Center, The Netherlands.
Searching other resources
Additional references will be identified in the Science Citation Index of Web of Science and by manually searching the references of relevant articles. Authors of identified studies may be contacted to ascertain their knowledge of any published or unpublished studies including new, additional studies or works in progress. The trial registers of ClinicalTrials, ICTRP and open trials will be searched for planned or ongoing trials. Conference proceedings will be identified using Embase and Web of Science.
Data collection and analysis
Selection of studies
Potentially eligible studies will be initially selected based on independent title and/or abstract assessment by two review authors (FD, IS) using the criteria set forth in the Criteria for considering studies for this review section.
When eligibility cannot be determined by the title and abstract review, the full text of all potentially relevant studies will be obtained for independent assessment according to the Criteria for considering studies for this review. We plan to contact primary authors to obtain additional information where necessary. Disagreements will be resolved by discussion between FD, MR and IS. The primary search results will be checked for overlapping content and de‐duplication will be performed as described in Bramer 2016 by our information specialist.
Data extraction and management
Two authors (FD, IS) will independently complete a data extraction using a predefined electronic spreadsheet. The following data will be retrieved.
-
General information: title, journal, year, publication status, country and type of hospital.
-
Methodological information: study design (prospective versus retrospective), multicentre, dates of recruitment, men included in previously published cohorts, selection procedure (consecutive, random or convenience), comparisons made, sequence and time between tests.
-
Study population: number of patients (initial biopsy, previous negative biopsy), number of previous negative biopsies, reason and number of exclusions, age, PSA level, prostate volume and clinical stage on Digital Rectal Examination (DRE).
-
MRI: field strength and vendor, sequences for defining targets, surface or endorectal coil, radiologist’s experience, number of (co‐) readers, MRI scoring system, and MRI suspicion score threshold for MRI‐targeted biopsy.
-
MRI‐targeted biopsy: method of registration and guidance, biopsy approach (transperineal, transrectal), anaesthesia, order of tests and blinding.
-
TRUS‐guided biopsy: biopsy approach, anaesthesia, order of tests and blinding.
-
Template‐guided biopsy: technique (TTMB, TSB), anaesthesia, order of tests and blinding.
-
Histopathology: definition of clinically significant prostate cancer, pathologist experience, number of readers, blinding for MRI outcome and biopsy technique.
-
Results: number of men with a suspicious lesion on MRI, number and score of MRI lesions, location of MRI lesions, number of men with MRI‐targeted biopsies, for all biopsy approaches: number of cores taken, biopsy density, core distribution, number and percentage positive cores, number of men in each Gleason score category (e.g. 3+3, 3+4, 4+3, 4+4), number of men with the prespecified target condition.
-
Two‐by‐two cross‐tabulations will be extracted for the target conditions.
The data extraction will be piloted to ensure accuracy and any disagreements will be resolved by discussion between FD, MR and IS.
Assessment of methodological quality
Each included study will be assessed for methodological quality independently by two review authors (FD, IS), using a tailored QUADAS‐2 quality checklist (Whiting 2011). Discrepancies will be resolved by discussion between FD, MR and IS. We will present an overview of our assessment findings by study in a table and the percentage of studies that met the criteria for each QUADAS‐2 item in a graph. The tailoring is detailed in Appendix 1 and summarised below.
Patient selection
Inappropriate exclusions include men of a specific age, PSA value or MRI outcome. The criteria should be clearly stated and patients with a previous diagnosis of prostate cancer or recurrent disease should be distinguishable. Applicability judgements will depend on the degree of suspicion of prostate cancer (initial diagnosis or after a number of previous prostate biopsies).
Index test(s)
Index tests need to be performed without knowledge of the results nor influenced by the performance of the reference standard. When comparing the index tests, the TRUS‐guided biopsy should be performed without the knowledge of MRI results or MRI‐targeted biopsy core trajectory. Applicability judgements will depend on appropriately sampling the prostate.
Reference standard
An insufficient number of biopsy cores, not distributed appropriately to comprehensively sample the whole prostate, is unlikely to correctly classify the target condition. Applicability judgements will depend on performance of the reference standard as commonly done in clinical practice.
Flow and timing domain
The interval between tests should be less than 6 months to be reasonably sure the target condition did not change during the interval. All patients should have received the same reference standard, regardless of the index test outcome.
Statistical analysis and data synthesis
Descriptive analyses
We will present an overview of all available studies in a table, including patient population and test characteristics. The table will be split into groups of studies, as described in Figure 1.
-
One or both components of the MRI pathway and the reference standard.
-
TRUS‐guided biopsy and the reference standard.
-
Both components of the MRI‐pathway and TRUS‐guided biopsy.
-
A combination of all above (i.e. comparing both index tests directly or indirectly as in a randomised trial of test accuracy).
For all studies we will extract two‐by‐two contingency tables for the target condition, containing: true positive, false positive, true negative and false negative values as shown in Figure 2.
Two‐by‐two contingency tables for target condition.
Comparison 1a: MRI suspicion score vs reference standard (if MRI‐targeted biopsies are performed these will serve as additional reference biopsies).
Comparison 1b: MRI±MRI‐TBx vs reference standard.
Comparison 2: TRUS‐guided biopsy vs reference standard.
Comparison 3: MRI±MRI‐TBx vs TRUS‐guided biopsy.
For comparison 1b and comparison 3, a negative test can result in two ways:
-
a negative MRI (thus no MRI‐targeted biopsies are taken).
-
a positive MRI but negative MRI‐targeted biopsy result.
Both negative outcomes should be merged, creating a two‐by‐two from a three‐by‐two contingency table (first and second column).
In comparison 3 the TRUS‐guided biopsy is not a proper reference standard. Therefore we will focus on the number of cancers identified and the concordance and disconcordance between both index tests.
TP = true positive; FP = false positive; TN = true negative; FN = false negative; Dposneg = Discordant MRI positive + targeted biopsies positive and TRUS‐guided biopsies negative; Dnegpos = Discordant MRI positive/negative + targeted biopsies negative and TRUS‐guided biopsies positive.
Study‐specific estimates and 95% confidence intervals (CIs) of sensitivity and specificity will be calculated and displayed in forest plots using Review Manager 5. These graphical displays will demonstrate the variation in diagnostic accuracy between studies. We will also plot the sensitivities and specificities on a receiver operating characteristic (ROC) curve. For comparison number 3 (MRI±MRI‐TBx and TRUS‐guided biopsy), we will focus on the proportion of clinically significant prostate cancer found by each test. We will display the proportion of cancers found by each test in forest plots, showing study specific estimates and 95% CIs. These analyses will be performed on a patient level, as data on a lesion level are rarely available.
Statistical analysis
We will start with an exploration of the ‘study specific’ specificities and sensitivities extracted from the included studies. If sufficient studies have been identified we will perform meta‐analyses, focused on the sensitivity and specificity of the index tests.
For comparison 1.A (MRI suspicion score vs reference standard and all additional biopsies (e.g. MRI‐targeted biopsies)) we will estimate a summary ROC curve using the Rutter and Gatsonis hierarchical summary ROC model because different thresholds to define a positive MRI can be chosen. To draw the summary ROC curve the results will be imported into Review Manager 5.
For comparisons 1.B (MRI±MRI‐TBx vs reference standard) and 2 (TRUS‐guided biopsy vs reference standard) we will perform bivariate analyses because outcomes of the tests will be either positive or negative for the target condition. We will start by calculating initial summary estimates of sensitivity and specificity for the index tests using the bivariate model. Results from these analyses will be imported into Review Manager 5 to draw hierarchic summary ROC plots.
For comparison 3 (MRI±MRI‐TBx vs TRUS‐guided biopsy) we will focus on the number of target conditions identified (concordance and discordance of test results) as described in Schoots 2015 because neither test counts as a valid reference test. The total number of cancers will be calculated as the number of concordant positive results plus the number of discordant results for which either test was positive. The sensitivity of either test will be calculated as the number of positive results of that test divided by the total number of cancers detected. The relative sensitivity of both tests will be used to compare the TRUS‐guided biopsy with the MRI pathway.
We do not expect to find enough studies relating to comparison 4 (a combination of all tests) to analyse them as a separate group. If however sufficient studies are found, we will use each comparison to contribute to the analyses of the index tests as described above.
Once we have insight into the data from the included studies, we will critically evaluate whether the choice of models is correct. All statistical analyses will be performed using Statistical Analysis Software (SAS), Version 9.3 for Windows.
Investigations of heterogeneity
To investigate heterogeneity, variables will be tested by adding them as covariates in our bivariate model or Hierarctical Summary Receiver Operating Characteristics (HSROC) model; we will use likelihood ratio tests in this context to determine the statistical significance of the included variables. The following factors will be included.
-
Type of population (initial biopsy, previous negative biopsy, mixed population).
-
Definition of the target condition.
-
MRI scoring system.
-
MRI suspicion score threshold for performing MRI‐targeted biopsy.
-
Navigational approach of MRI‐targeted biopsy.
-
Average number of MRI lesions and location.
-
Average number of biopsy cores (or biopsy density) and core distribution in the reference standard.
Sensitivity analyses
We will consider risk of bias, concerns for applicability and other methodological properties as assessed by our tailored QUADAS‐2 tool to be potential covariates for examining the reliability of our findings. To determine the reliability, our primary analysis will include all studies but our sensitivity analysis will only include the highest quality studies.
Assessment of reporting bias
We will not undertake assessment of reporting bias. Contrary to the reporting bias seen in intervention studies, reporting of accuracy studies about MRI and prostate biopsy is less likely to be influenced by significant or positive results. Furthermore, there is no evidence of reporting bias in test accuracy reviews nor a reliable method to detect this (Deeks 2005).
The MRI pathway comprises two components: MRI (suspicion score) and MRI‐targeted biopsy in the presence of suspicious lesions for clinically significant prostate cancer:
-
a) MRI suspicion score: refers to only the first component of the MRI pathway.
-
b) MRI±MRI‐TBx: refers to both components of the MRI pathway.
Review design
Meta‐analysis of studies reporting on:
The MRI pathway and the reference standard: ·
-
Comparison 1a: MRI suspicion score vs reference standard and all additional biopsies (e.g. MRI‐targeted biopsies).
-
Comparison 1b: MRI±MRI‐TBx vs reference standard.
TRUS‐guided biopsy and reference standard: ·
-
Comparison 2
MRI pathway B) MRI±MRI‐TBx and TRUS‐guide biopsy:
-
Comparison 3
A combination of all tests:
-
Comparison 4: comparing both index tests, either directly with the reference standard or indirectly as a randomised trial of test accuracy.
Two‐by‐two contingency tables for target condition.
Comparison 1a: MRI suspicion score vs reference standard (if MRI‐targeted biopsies are performed these will serve as additional reference biopsies).
Comparison 1b: MRI±MRI‐TBx vs reference standard.
Comparison 2: TRUS‐guided biopsy vs reference standard.
Comparison 3: MRI±MRI‐TBx vs TRUS‐guided biopsy.
For comparison 1b and comparison 3, a negative test can result in two ways:
-
a negative MRI (thus no MRI‐targeted biopsies are taken).
-
a positive MRI but negative MRI‐targeted biopsy result.
Both negative outcomes should be merged, creating a two‐by‐two from a three‐by‐two contingency table (first and second column).
In comparison 3 the TRUS‐guided biopsy is not a proper reference standard. Therefore we will focus on the number of cancers identified and the concordance and disconcordance between both index tests.
TP = true positive; FP = false positive; TN = true negative; FN = false negative; Dposneg = Discordant MRI positive + targeted biopsies positive and TRUS‐guided biopsies negative; Dnegpos = Discordant MRI positive/negative + targeted biopsies negative and TRUS‐guided biopsies positive.
