Diagnostic Performance of Biparametric MRI for Detection of Prostate Cancer: A Systematic Review and Meta-Analysis
Abstract
OBJECTIVE. The purpose of this study was to perform a systematic review and meta-analysis to estimate the diagnostic performance of biparametric MRI (bpMRI) for detection of prostate cancer (PCa).
MATERIALS AND METHODS. Two independent reviewers performed a systematic review of the literature published from January 2000 to July 2017 by using predefined search terms. The standard of pathologic reference was established at prostatectomy or prostate biopsy. The numbers of true- and false-positive and true- and false-negative results were extracted. The Quality Assessment of Diagnostic Accuracy Studies tool was used to assess the quality of the selected studies. Statistical analysis included pooling of diagnostic accuracy, meta-regression, subgroup analysis, head-to-head comparison, and identification of publication bias.
RESULTS. Thirty-three studies were used for general data pooling. The overall sensitivity was 0.81 (95% CI, 0.76–0.85), and overall specificity was 0.77 (95% CI, 0.69–0.84). As for clinically relevant PCa, bpMRI maintained high diagnostic value (AUC, 0.85; 95% CI, 0.82–0.88). There was no evidence of publication bias (p = 0.67). From head-to-head comparison for detection of PCa, multiparametric MRI (mpMRI) had significantly higher pooled sensitivity (0.85; 95% CI, 0.78–0.93) than did bpMRI (0.80; 95% CI, 0.71–0.90) (p = 0.01). However, the pooled specificity values were not significantly different (mpMRI, 0.77 [95% CI, 0.58–0.95]; bpMRI, 0.80 [95% CI, 0.64–0.96]; p = 0.82).
CONCLUSION. The results of this meta-analysis suggest that bpMRI has high diagnostic accuracy in the detection of PCa and maintains a high detection rate for clinically relevant PCa. However, owing to high heterogeneity among the included studies, caution is needed in applying the results of the meta-analysis.
Prostate cancer (PCa) is the second leading cause of cancer death in men, and its incidence is expected to double by 2030 owing to the aging of the global population [1]. Irrespective of several issues concerning the use of prostate-specific antigen (PSA) as a biomarker for PCa diagnosis, the test is widely used to diagnose PCa all around the globe [2]. The first problem with use of PSA measurement is that several benign prostate lesions, such as benign prostatic hyperplasia and acute prostatitis, can increase serum PSA level, as can manipulation of the prostate. Another problem is related to the increasing failure to detect clinically significant tumors in relation to the extensive use of PSA in the last several years.
Interest has been growing in the contribution of prostatic MRI to improvements in the accuracy of PCa detection and characterization [3]. Combining conventional MRI (multiplanar T1- and T2-weighted imaging) with DWI and dynamic contrast-enhanced MRI (DCE-MRI) is known as multiparametric MRI (mpMRI) [4]. Despite improvements in the diagnostic specificity of mpMRI, the exact beneficial role of contrast-enhanced MRI is debatable [5]. On the one hand, studies [6–8] have shown a substantial benefit for detection of cancer with multiparametric protocols when DCE-MRI is added. On the other hand, a few studies [9–11] have shown no supplemental value of DCE-MRI for detection of cancer. This difference is mainly caused by heterogeneity in study designs in terms of technical criteria and characteristics of patients.
In view of the rapidly increasing demand for MRI in the diagnosis of PCa and the need to optimize imaging protocols with a focus on reducing financial burden and improving patient comfort, some researchers [12, 13] have proposed biparametric (T2-weighted imaging and DWI) MRI (bpMRI) for imaging of PCa. However, the clinical significance of bpMRI of the prostate remains unclear and controversial. Accordingly, we sought to collect as much extractable data as possible by conducting a systematic review and meta-analysis to determine the diagnostic accuracy and technical considerations of bpMRI for the diagnosis of PCa.
Materials and Methods
Literature Search
This meta-analytic review was written with reference to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement [14]. A comprehensive computerized literature search was performed to identify articles examining the diagnostic accuracy of bpMRI for diagnosing PCa. We searched the PubMed, Embase, Scopus, and Cochrane electronic databases and Web of Science for the period January 2000–July 2017. In the electronic search, we used a search algorithm based on a combination of the following keywords: (prostate cancer OR prostatic cancer OR prostate neoplasm OR prostatic neoplasm OR prostate tumor OR prostatic tumor OR prostate carcinoma OR prostatic carcinoma OR PCa) AND (magnetic resonance imaging OR MRI OR MR) AND (biparametric OR bp OR T2-weighted image and DWI OR T2-weighted imaging and DWI). Retrieved titles and abstracts were independently reviewed for relevance by two authors (an abdominal radiologist with nearly 4 years of experience with mpMRI and a radiologist with more than 2 years of experience in performing mpMRI). Full texts of the relevant studies were retrieved for further evaluation. Reference lists of these relevant studies were checked manually to identify other relevant articles. Searching was restricted to publications in English.
Study Selection
The inclusion criteria were as follows: included patients with suspected or diagnosed PCa; bpMRI (at least including T2-weighted imaging and DWI) of the prostate performed as an index test; prostatectomy or prostate biopsy as the reference standard; sufficient data reported to construct 2 × 2 contingency tables with at least 10 patients; and original article. Studies were excluded if any one of the inclusion criteria was not met.
Data Extraction and Quality Assessment
Data extraction was performed independently by two authors using predefined data extraction forms. Finally, Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) was independently performed by the same two authors to ascertain the quality of studies and the likelihood of bias. Results were compared between the authors at every step, and any discrepancies were resolved in consensus review. Assessment of the QUADAS-2 tool was performed with a review manager software program (RevMan, version 5.3, Cochrane Nordic).
Statistical Analysis
Data from the studies were combined to compute the sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio along with 95% CIs across all included studies by use of the bivariate random-effects model. A summary ROC curve was also generated and used to calculate the ROC AUC. Heterogeneity between primary studies was assessed by use of the I2 statistic. Values of I2 equal to 25%, 50%, and 75% were assumed to represent low, moderate, and high heterogeneity. In cases of obvious heterogeneity among studies, univariable meta-regression analysis was performed by use of a random-effects model with several covariates that could have accounted for data heterogeneity. Finally, subgroup analysis according to outcomes of the meta-regression analysis was performed by pairwise comparison. For head-to-head comparison, the individual studies were plotted in an ROC space with connecting lines, as were summary estimates. The bivariate model was used to obtain summary estimates and 95% CIs of sensitivity and specificity in the head-to-head comparisons. A z test for paired data was used in the model to compare sensitivity and specificity.
Publication bias was assessed by means of inspection of the funnel plot for asymmetry, based on the method of Deeks et al. [15]. All statistical analyses were conducted with Stata (version 14.0, StataCorp) and RevMan software.
Results
Study Selection
A total of 498 articles were screened in the primary literature search. Of those, 450 articles were excluded after the titles and abstracts were read, leaving 48 articles for further full-text review. Finally, 33 articles [6, 8, 10, 12, 13, 16–43] were selected for analysis according to the inclusion criteria. A flowchart of the study selection process is shown in Figure 1.
Study Characteristics
In total, 2383 patients with PCa described in the 33 selected articles were included in our meta-analysis. The patient characteristics are presented in Table 1. The sizes of the study populations ranged from 23 to 204 patients. The patients had median ages of 58–70 years. The median PSA levels were 4.6–14 ng/mL, and the Gleason scores, 6–10. For 1139 of the 2383 patients, no data regarding previous biopsies were reported; 597 had not undergone previous biopsy; and the others had at least one negative biopsy result.
| First Author | Reference | Year of Publication | Country | Duration of Recruitment | No. of Patients | No. of Patients with PCa | Age (y) | PSA (ng/mL) | Gleason Score | No. of Previous Biopsies | PCa Diagnosis Before MRI | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Median | Range | Median | Range | Median | Range | |||||||||
| AbdelMaboud | [16] | 2014 | Egypt | 12/2010–8/2012 | 36 | 31 | 61a | 50–72 | NR | >4 | NR | NR | 0 | No |
| Barth | [17] | 2017 | Switzerland | 7/2013–3/2015 | 63 | 38 | 65.2a | 51.2–78.2 | 9.2a | 0.3–32.4 | 7 | 5–9 | NR | No |
| Brock | [18] | 2015 | United States | 9/2012–10/2012 | 45 | 45 | 66a | NR | 9.2a | NR | 7 | ≥6 | NR | Yes (all) |
| Delongchamps | [6] | 2011 | France | 10/2009–5/2010 | 58 | 58 | 62 | 49–74 | 6.8a | 4–9.9 | NR | NR | 1–3 | Yes (all) |
| Doo | [19] | 2012 | Korea | 7/2007–7/2011 | 51 | 51 | 63 | 50–72 | 11.5 | 4.23–43.83 | NR | 6–10 | NR | Yes (all) |
| Fascelli | [12] | 2016 | Turkey | 12/2012–12/2014 | 59 | 44 | 64.3a | 45.0–84.9 | 6.6a | 0.9–43.3 | 7 | NR | 0 | No |
| Franiel | [20] | 2011 | Germany | 12/2008–12/2009 | 54 | 21 | 68 | 49–78 | 12.1 | 3.3–65.2 | 7 | 6–10 | 1–6 | No |
| Haider | [21] | 2007 | Canada | 5/2005–5/2006 | 49 | 49 | 61 | 46–75 | 5.375 | 0.9–26 | 6 | 6–10 | NR | Yes (all) |
| Jung | [22] | 2013 | Korea | 1/2008–4/2010 | 156 | 72 | 59.2 | 42–79 | 5 | 0.2–78.1 | 7 | ≥6 | NR | No |
| Katahira | [23] | 2011 | Japan | 11/2004–3/2008 | 201 | 201 | 70 | 43–80 | 8.6 | 2.61–114 | 7 | 4–10 | ≥1 | Yes (all) |
| Kitajima | [10] | 2010 | Japan | 3/2008–1/2009 | 53 | 30 | 69 | 56–84 | 11.1 | 4.2–112.1 | NR | NR | 0 | No |
| Kitamura | [24] | 2014 | Japan | 2/2009–7/2013 | 54 | 54 | 62.7a | NR | 5.7 | 4.4–7.6 | 7 | ≥6 | ≥1 | Yes (all) |
| Lawrence | [25] | 2014 | UK | 2/2012–6/2012 | 39 | 16 | 64a | 47–77 | 10 | 1.2–36 | NR | 6–9 | ≥1 | No |
| Lim | [26] | 2009 | Korea | 3/2005–2/2007 | 52 | 52 | 65a | 48–76 | 10.5a | 1.2–79.6 | 7.4a | 6–9 | NR | Yes (all) |
| Morgan | [27] | 2007 | UK | NR | 54 | 54 | 68 | 52–80 | 9.8 | 2.3–46 | NR | 6–8 | NR | Yes (all) |
| Mussi | [28] | 2017 | Brazil | 9/2013–12/2014 | 118 | 68 | NR | NR | 4.6 | 3.8–7.0 | NR | ≥6 | NR | No |
| Naiki | [29] | 2011 | Japan | 1/2006–10/2008 | 35 | 35 | 67.7a | 49–78 | 12.8a | 2.78–67.3 | 7 | 5–10 | NR | No |
| Rais-Bahrami | [13] | 2015 | United States | 8/2007–12/2012 | 143 | 84 | 60.7a | 41–80 | 6.8a | 0.1–51.1 | NR | NR | 0 | No |
| Rinaldi | [30] | 2012 | Italy | 6/2010–12/2010 | 41 | 36 | 69a | 57–80 | 15.15a | 5.98–133 | NR | NR | NR | Yes (some) |
| Rosenkrantz | [31] | 2011 | United States | NR | 42 | 42 | 62a | 47–76 | 6.2a | 1.3–32.5 | 6.5 | 6–9 | NR | Yes (all) |
| Scialpi | [32] | 2017 | Italy | 1/2014–12/2015 | 41 | 41 | 64.5a | 53–78 | 6.8 | 1.5–39.3 | NR | ≥6 | NR | Yes (all) |
| Shimofusa | [33] | 2005 | Japan | 2/2003–11/2003 | 60 | 37 | 71a | 54–82 | 21.8a | 4.5–130 | NR | NR | 0 | Yes (all) |
| Stanzione | [34] | 2016 | Italy | NR | 82 | 34 | 65a | 43–84 | 8.8a | NR | 7 | 6–9 | NR | No |
| Tanimoto | [8] | 2007 | Japan | 1/2005–5/2005 | 83 | 44 | 67.4a | 53–87 | 19.4a | 4.3–332.1 | 6.9a | 6–10 | NR | No |
| Thestrup | [35] | 2016 | Denmark | 1/2014–6/2014 | 204 | 68 | 65 | 45–75 | 14 | 2.2–120 | 6 | NR | 0–3 | Yes (some) |
| Ueno | [36] | 2013 | Japan | 1/2010–2011.08 | 80 | 80 | 66.5a | 50–77 | 9.51a | 2.9–49 | 7 | 6–9 | NR | Yes (all) |
| Ueno | [37] | 2013 | Japan | 1/2010–6/2011 | 73 | 73 | 66a | 50–77 | 9.51a | 2.9–49 | 7 | 6–9 | NR | Yes (all) |
| Vargas | [38] | 2011 | United States | 9/2008–5/2009 | 51 | 51 | 58 | 46–74 | 5.3 | 0.4–62.2 | NR | 6–8 | NR | Yes (all) |
| Vilanova | [39] | 2011 | Spain | 5/2008–9/2009 | 70 | 38 | 63.5a | 43–87 | 7.4 | 4–17.20 | 7 | 5–8 | 0 | No |
| Wang | [40] | 2016 | China | 12/2011–1/2013 | 133 | 60 | 68 | 46–92 | 6.9 | 0.2–19.8 | 7 | NR | 0 | Yes (some) |
| Yağci | [41] | 2011 | Turkey | 9/2007–3/2009 | 43 | 21 | 66a | 49–79 | 9.1 | 1.4–120 | 7 | 6–10 | 0 | No |
| Yoshimitsu | [42] | 2008 | Japan | 1/2000–3/2004 | 37 | 37 | 66a | 56–75 | 11.9a | 0.7–54.8 | NR | NR | ≥1 | Yes (all) |
| Yoshizako | [43] | 2008 | Japan | NR | 23 | 23 | 68 | 52–81 | NR | NR | 7 | 6–9 | NR | Yes (all) |
Note—PCa = prostate cancer, PSA = prostate-specific antigen level, NR = not reported.
a
Mean.
The principal characteristics of the 33 selected articles are listed in Table 2. Among these articles were 10 prospective and 23 retrospective studies. The publication years of the eligible studies ranged from 2007 to 2017. An MRI field strength of 3 T was used in 10 studies, 1.5 T in 19 studies, and a combination of 3 T and 1.5 T in two studies. An endorectal coil was used in only 14 of the studies. The b values were classified on two scales. High b values (≥ 1400 s/mm2) were used in eight studies and low b values (< 1400 s/mm2) in 25 studies.
| First Author | Reference | Design | Consecutive Enrollment | Reference Standard | MRI-Reference Standard Interval | No. of Readers | Blinding | Field Strength (T) | b Value (s/mm2) | Endorectal Coil | Localization | Type of Analysis |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AbdelMaboud | [16] | Prospective | NR | TRUSGB and RP | NR | NR | NR | 1.5 | 0, 1000 | No | PZ | Patient |
| Barth | [17] | Prospective | Yes | TTSB | ≤ 6 mo | 3 | Yes | 3 | 0, 50, 1000 or 100, 600, 1000 | Yes | Whole | Lesion |
| Brock | [18] | Prospective | Yes | RP | 28–59 d | 1 | Yesa | 1.5 | NR | Yes | Whole | Lesion |
| Delongchamps | [6] | Prospective | Yes | RP | ≥ 6 wk | 2 | Yesa | 1.5 | 0, 800 | Yes | PZ, TZ, whole | Lesion |
| Doo | [19] | Retrospective | NR | RP | ≤ 6 mo | 2 | Yesa | 3 | 0, 1000 | No | Whole | Lesion |
| Fascelli | [12] | Retrospective | Yes | MRI-TRUS biopsy | NR | NR | NR | NR | NR | No | Whole | Patient |
| Franiel | [20] | Prospective | Yes | STRUSGB and MRGB | NR | 1 | NR | 1.5 | 0, 100, 400, 800 | Yes | Whole | Lesion |
| Haider | [21] | Prospective | NR | RP | ≥ 6 wk | 1 | Yes | 1.5 | 600 | Yes | PZ, TZ, Whole | Lesion |
| Jung | [22] | Retrospective | Yes | RP | ≤ 6 m | 2 | Yes | 1.5 or 3 | 0, 1000 | Yes | TZ | Patient |
| Katahira | [23] | Retrospective | Yes | RP | ≤ 2 m | 3 | Yesa | 1.5 | 1000, 2000 | No | PZ, TZ, Whole | Lesion |
| Kitajima | [10] | Retrospective | Yes | STRUSGB | 10–41 d | 2 | Yes | 3 | 0, 1000 | No | PZ, TZ, Whole | Lesion |
| Kitamura | [24] | Prospective | Yes | TRUSGBandRP | 24.8–54.5 d | 1 | Yesa | 1.5 | 0, 1000 | Yes | Whole | Lesion |
| Lawrence | [25] | Retrospective | Yes | MRI-TRUS biopsy | ≥ 9m | 2 | Yes | 1.5 or 3 | 0, 1000, 1400 | No | PZ, TZ | Lesion |
| Lim | [26] | Retrospective | NR | RP | 2–38 d | 3 | Yesa | 1.5 | 0, 1000 | Yes | Whole | Lesion |
| Morgan | [27] | Prospective | Yes | TRUSGB | 1–90 d | 2 | Yes | 1.5 | 50, 300, 500, 800 | Yes | Whole | Lesion |
| Mussi | [28] | Retrospective | NR | MRI-TRUS biopsy | ≤ 6 mo | 2 | Yes | 3 | 50, 400, 800, 1500 | No | Whole | Lesion |
| Naiki | [29] | Retrospective | NR | TRUSGBandRP | NR | 1 | Yes | 1.5 | 0, 800 | Yes | PZ, TZ, whole | Lesion |
| Rais-Bahrami | [13] | Retrospective | NR | MRI-TRUS biopsy | NR | NR | NR | NR | NR | NR | Whole | Patient |
| Rinaldi | [30] | Prospective | NR | TRUSGB | ≥ 1 mo and ≤ 3 mo | 2 | NR | 1.5 | 0, 250, 500, 750, 1000 | Yes | PZ, CZ, whole | Lesion |
| Rosenkrantz | [31] | Retrospective | Yes | RP | 19–191 d | 2 | Yes | 1.5 | 0, 500, 1000 | No | PZ | Lesion |
| Scialpi | [32] | Retrospective | NR | TRUSGB and RP | ≤ 45 d | 2 | Yes | 3 | 0, 2000 | No | PZ, TZ, whole | Lesion |
| Shimofusa | [33] | Retrospective | Yes | RP | ≤ 6 mo | 3 | Yes | 1.5 | 0, 1000 | No | Whole | Patient |
| Stanzione | [34] | Retrospective | NR | TRUSGB | ≤ 1 mo | 2 | Yes | 3 | 0, 400, 2000 | No | Whole | Patient |
| Tanimoto | [8] | Retrospective | Yes | RP | ≤ 4 mo | 2 | NR | 1.5 | 0, 1000 | No | Whole | Patient |
| Thestrup | [35] | Retrospective | NR | TRUSGB and MRGB | ≤ 2 mo and ≥ 6 wk | 2 | Yes | 3 | 0, 100, 800, 2000 | No | Whole | Patient |
| Ueno | [36] | Retrospective | Yes | RP | NR | 2 | Yes | 3 | 0, 1000, 2000 | No | PZ, TZ, whole | Lesion |
| Ueno | [37] | Retrospective | Yes | RP | 3–343 d | 2 | Yes | 3 | 0, 1000, 2000 | No | PZ, TZ, Whole | Lesion |
| Vargas | [38] | Retrospective | Yes | RP | ≤ 6 mo | 2 | Yesa | 3 | 0, 700, 1000 | Yes | Whole | Lesion |
| Vilanova | [39] | Retrospective | NR | TRUSGB and RP | 13 ± 9 d | 3 | Yes | 1.5 | 0, 1000 | Yes | Whole | Patient |
| Wang | [40] | Prospective | Yes | TRUSGB | ≤ 2 mo | 2 | Yes | 1.5 | 0, 1000 | No | PZ | Lesion |
| Yağci | [41] | Retrospective, prospective | Yes | STRUSGB | ≤ 7 d | 1 | Yes | 1.5 | 800 | Yes | PZ | Lesion |
| Yoshimitsu | [42] | Retrospective | NR | TRUSGB | 6–9 wk | 2 | Yesa | 1.5 | 0, 500, 1000 | No | PZ, TZ, whole | Lesion |
| Yoshizako | [43] | Retrospective | NR | RP | 1–7 wk | 2 | Yesa | 1.5 | 0, 1000 | No | TZ | Lesion |
Note—NR = not reported, TRUSGB = transrectal ultrasound–guided biopsy, RP = radical prostatectomy, PZ = peripheral zone, TTSB = transperineal template saturation biopsy, TZ = transition zone, MRI-TRUS = fusion of MRI and transrectal ultrasound images, STRUSGB = systematic transrectal ultrasound-guided biopsy, CZ = central zone.
a
Blinded but aware that patients had biopsy-proven prostate cancer.
Radical prostatectomy was used in 20 studies. Transrectal ultrasound (TRUS)-guided biopsy was used in eight studies, MRI-TRUS fusion–guided in four studies, and standardized transperineal template saturation biopsy in one study. Assessment was performed on a per-patient basis in nine studies and a per-lesion basis in 24 studies.
Assessment of Study Quality and Publication Bias
The methodologic quality of all the included studies was evaluated in accordance with QUADAS-2. The results of the QUADAS-2 evaluation are shown in Figure 2. Regarding the patient selection domain, 16 studies [10, 16, 18, 19, 21, 23, 24, 26, 27, 29, 31, 32, 36, 37, 42, 43] had high risk of bias because consecutive enrollment was not used or the exclusion criteria were not appropriately met. Regarding the index test domain, four studies had a high level of concern about applicability. One of them had unclear research background [43], and the other three did not apply the complete MRI parameters [12, 13, 18]. Regarding the reference standard domain, we considered radical prostatectomy or MRI-TRUS fusion–guided targeted biopsy the reference standard. In comparison with the reference standard, TRUS-guided biopsy or transperineal biopsy was considered to have high risk of bias. Therefore, the risk of bias regarding the reference standard was high in eight studies [10, 20, 27, 30, 34, 40–42] and low in 25 studies [6, 8, 12, 13, 16–19, 21–26, 28, 29, 31–33, 35–39, 43]. For flow and timing, there was no high risk of bias in any of the included studies.
Results of the funnel plot asymmetry test for publication bias analyzed by means of linear regression of the log odds ratio on effective sample size were not significant (p = 0.67), and the slope was not significant, suggesting no major publication bias (Fig. 3).
Overall Diagnostic Accuracy
Diagnostic performance showed that the sensitivity in individual studies ranged from 0.52 to 1.00 and the specificity from 0.12 to 0.98 (Fig. 4). The pooled sensitivity was 0.81 (95% CI, 0.76–0.85; I2 = 0.91); specificity, 0.77 (95% CI, 0.69–0.84; I2 = 0.97); positive likelihood ratio, 3.5 (95% CI, 2.6–4.7); negative likelihood ratio, 0.25 (95% CI, 0.20–0.31); and diagnostic odds ratio 14 (95% CI, 10–20) for bpMRI in detecting PCa. The summary ROC curve had an AUC of 0.86 (95% CI, 0.83–0.89) (Fig. 5A). The value of bpMRI in detection of clinically relevant PCa (defined as any cancer that met one of three criteria: Gleason pattern ≥ 4, tumor volume ≥ 0.5 mL, Gleason score ≥ 7) was evaluated in 10 studies. The results showed that bpMRI had high diagnostic value, evidenced by an AUC of 0.85 (95% CI, 0.82–0.88), sensitivity of 0.81 (95% CI, 0.69–0.89), and specificity of 0.74 (95% CI, 0.54–0.88) (Fig. 5B).
Fig 5A —Summary ROC (SROC) curves with prediction and confidence contours. Numbers in circles indicate numbers assigned to given articles in bivariate model.
A, Prostate cancer.
Fig 5B —Summary ROC (SROC) curves with prediction and confidence contours. Numbers in circles indicate numbers assigned to given articles in bivariate model.
B, Clinically relevant cancer.
Subgroup Analyses and Head-to-Head Comparison
Several potential variables, including study design (prospective versus retrospective), patient enrollment (consecutive versus not consecutive), reference standard (high risk of bias versus low risk of bias), MRI reference standard reported (yes or no), whether the readers were blinded to histologic findings (blinded versus not blinded), b value (≥ 1400 s/mm2 versus < 1400 s/mm2), and MRI field strength (3 T versus 1.5 T) showed significant independent association with sensitivity (p < 0.05 for all), whereas only reference standard (high risk of bias vs low risk of bias) and analysis type (per patient versus per lesion) had a significant independent association with specificity (p < 0.05) (Table 3).
| Subgroup | No. of Studies | Pooled Sensitivity | p | Pooled Specificity | p |
|---|---|---|---|---|---|
| Design | < 0.001 | 0.13 | |||
| Prospective | 10 | 0.77 (0.67–0.86) | 0.77 (0.64–0.90) | ||
| Retrospective | 23 | 0.82 (0.77–0.88) | 0.77 (0.69–0.86) | ||
| Enrollment | < 0.001 | 0.19 | |||
| Consecutive | 19 | 0.78 (0.71–0.84) | 0.79 (0.70–0.87) | ||
| Not consecutive | 14 | 0.84 (0.78–0.90) | 0.75 (0.63–0.87) | ||
| Reference standard | < 0.001 | 0.02 | |||
| Low risk of bias | 25 | 0.80 (0.75–0.86) | 0.75 (0.66–0.84) | ||
| High risk of bias | 8 | 0.82 (0.73–0.91) | 0.83 (0.72–0.94) | ||
| MRI-reference standard interval | < 0.001 | 0.18 | |||
| Reported | 27 | 0.78 (0.73–0.84) | 0.81 (0.75–0.87) | ||
| Not reported | 6 | 0.89 (0.82–0.96) | 0.51 (0.29–0.74) | ||
| Blinding to histologic findings | < 0.001 | 0.23 | |||
| Blinded | 27 | 0.78 (0.73–0.83) | 0.81 (0.75–0.87) | ||
| Not reported | 6 | 0.91 (0.85–0.97) | 0.53 (0.30–0.76) | ||
| b value | < 0.001 | 0.10 | |||
| High b value (≥ 1400 s/mm2) | 8 | 0.84 (0.77–0.91) | 0.72 (0.58–0.86) | ||
| Low b value (< 1400 s/mm2) | 22 | 0.77 (0.72–0.83) | 0.79 (0.76–0.89) | ||
| MRI field strength | < 0.001 | 0.17 | |||
| 3 T | 10 | 0.82 (0.73–0.89) | 0.81 (0.66–0.89) | ||
| 1.5 T | 19 | 0.79 (0.70–0.87) | 0.77 (0.71–0.86) | ||
| Localization analyzed | 0.11 | 0.19 | |||
| Whole prostate | 27 | 0.81 (0.76–0.86) | 0.77 (0.69–0.85) | ||
| Peripheral or transitional zone | 6 | 0.77 (0.65–0.90) | 0.80 (0.64–0.96) | ||
| Analysis type | 0.10 | 0.03 | |||
| Per patient | 9 | 0.88 (0.82–0.94) | 0.68 (0.50–0.85) | ||
| Per lesion | 24 | 0.77 (0.72–0.83) | 0.80 (0.73–0.87) |
Note—Values in parentheses are 95% CI.
Head-to-head comparison between b pMRI and mpMRI was provided in 11 studies. The pooled sensitivity showed a significant difference between bpMRI and mpMRI (bpMRI, 0.80 [95% CI, 0.71–0.90]; mpMRI, 0.85 [95% CI, 0.78–0.93]; p = 0.01). The specificity, however, did not show a significant difference between these two groups (bpMRI, 0.80 [95% CI, 0.64–0.96]; mpMRI, 0.77 [95% CI, 0.58–0.95]; p = 0.82) (Fig. 6).
Discussion
The results of our meta-analysis show that the diagnostic accuracy of bpMRI for detecting PCa is high with a sensitivity of 0.81 and a specificity of 0.77. Comparing our results with the outcome of a published meta-analysis (only one article) of bpMRI for detecting PCa [44], we found that a trend toward higher sensitivity and lower specificity can be inferred. In that study, the investigators evaluated 10 studies of DWI in combination with T2-weighted imaging and pooled sensitivity of 0.76 (95% CI, 0.65–0.84) and specificity of 0.82 (95% CI, 0.77–0.87). In our study, we included several more recently published bpMRI studies, which might have affected the sensitivity and specificity of bpMRI for detecting PCa. Notably, a large sample size (2383 vs 627 patients) included in the current study could allow the robust conclusion that bpMRI can be a reliable tool for detecting PCa.
We found significant heterogeneity among the included studies. To deal with this issue, we used meta-regression to explore the sources of heterogeneity and then performed the subgroup analysis. Interestingly, the characteristics of the study affected only sensitivity. There is higher sensitivity in high-bias (e.g., retrospective study, nonconsecutive enrollment of patients) research than in low-bias research, because high-bias studies overestimate the accuracy of test methods.
Our study shows significantly higher sensitivity and specificity for high b values (≥ 1400 s/mm2) than low b values for detecting PCa, but the difference in specificity did not reach statistical significance. It is well known that one of the most important acquisition parameters influencing results obtained at DWI is the selection of b value. For the purpose of detecting PCa, acquisition of high-b-value images is recommended in Prostate Imaging Reporting and Data System (PI-RADS) version 2. The guideline is a b value of at least 1400 s/mm2, or up to 2000 s/mm2 if the signal-to-noise ratio remains adequate. On images with a high b value (≥ 1400 s/mm2), the signal intensity of the normal prostate tissue is low, whereas the signal intensity of PCa is high, representing restricted diffusion. In comparison, at a b value less than 1400 s/mm2, the signal intensity of the foci in benign prostatic hyperplasia and prostatitis is high, obscuring the conspicuity of tumors [45]. However, as the b value increases, the signal-to-noise ratio decreases, so that selection of the optimum high b value may depend on magnetic field strength, software, and manufacturer. Evidence [46, 47] suggests that use of high-b-value computed or extrapolated DWI (e.g., incoherent motion, stretched exponential, and diffusional kurtosis) may avoid artifacts and may improve lesion-to-background contrast ratios for PCa detection.
From head-to-head comparison for detection of PCa, our research showed the higher pooled sensitivity for mpMRI than for bpMRI, but the difference in specificity was not significant. From PI-RADS version 1 to version 2, the interpretation of DCE-MRI was greatly simplified and expressed as a binary assessment, according to either the presence or absence of early focal enhancement. Although DCE-MRI has a limited role in the diagnosis of PCa, some researchers insist that DCE-MR images should still be carefully evaluated for potential abnormalities, because inspection of DCE-MR images may help identify the subtle lesions that may be missed in other pulse sequences. This may be why our study showed that bpMRI has lower sensitivity than mpMRI does.
As many as one-third of patients with a diagnosis of PCa had an index tumor volume less than 0.5 cm3 and a Gleason score less than 7. These tumors are more likely to remain latent in long-term follow-up and are considered clinically insignificant or indolent. Patients with clinically relevant cancers, however, have a high probability of dying of PCa within 10 years. Therefore, many researchers have proposed that exploring clinically relevant cancer is necessary. In the current study, we extracted data to investigate the value of bpMRI for detection of clinically relevant PCa and found high value of bpMRI (AUC, 0.85; 95% CI, 0.82–0.88). The results of Vargas et al. [48] showed that DCE-MRI provided incremental value of only 3% for peripheral zone tumors and no benefit for transitional zone tumors in a multiparametric protocol when the PI-RADS algorithm was used for interpretation. Furthermore, Barth et al. [17] found that for the detection of clinically relevant PCa, no significant difference was found in the diagnostic performance of a bpMRI protocol consisting of T2-weighted imaging and DWI compared with that of a traditional mpMRI protocol. In the current study, we could not directly compare bpMRI with mpMRI in the diagnosis of clinically relevant PCa because of insufficient data. Well-designed prospective studies with large sample sizes are needed to prove whether bpMRI equals or outweighs mpMRI in the diagnosis of clinically relevant PCa.
There were several limitations to our study. First, the included studies were heterogeneous in their methods, which affected the general applicability of the summary estimates. However, such methodologic variability was useful for the subgroup analyses we performed to identify factors that could improve the diagnostic accuracy of bpMRI in the future. Second, we did not investigate quantitative apparent diffusion coefficient (ADC) analysis, which is thought to improve the diagnostic performance of bpMRI. De Cobelli et al. [49] found that decreasing ADC values are a strong risk factor, independent of biopsy features, for the presence of poorly differentiated PCa. Therefore, incorporating ADC values into the composite equations may further improve the value of bpMRI. Third, until recently there has been no consensus on the definition of clinically relevant PCa, which might have led to unreliable conclusions in our study. Finally, we could not identify a scoring system for bpMRI because of the heterogeneity of studies. Further study is necessary to set up such a scoring system comparable to PI-RADSv1 and PI-RADSv2.
Conclusion
Our meta-analysis showed that bpMRI is an accurate diagnostic imaging tool in the detection of PCa. However, owing to the lower sensitivity of bpMRI than mpMRI and high heterogeneity of the included studies, caution is needed in using it clinically.
Acknowledgment
We thank Sushant Kumar Das for improving the use of English in this article.
References
1.
Maddams J, Utley M, Møller H. Projections of cancer prevalence in the United Kingdom, 2010–2040. Br J Cancer 2012; 107:1195–1202
2.
Lippi G, Montagnana M, Guidi GC, Plebani M. Prostate-specific antigen-based screening for prostate cancer in the third millennium: useful or hype? Ann Med 2009; 41:480–489
3.
Hoeks CM, Barentsz JO, Hambrock T, et al. Prostate cancer: multiparametric MR imaging for detection, localization, and staging. Radiology 2011; 261:46–66
4.
Weinreb JC, Barentsz JO, Choyke PL, et al. PI-RADS Prostate Imaging Reporting and Data System: 2015, version 2. Eur Urol 2016; 69:16–40
5.
Scialpi M, Falcone G, Scialpi P, D'Andrea A. Biparametric MRI: a further improvement to PIRADS 2.0? Diagn Interv Radiol 2016; 22:297–298
6.
Delongchamps NB, Beuvon F, Eiss D, et al. Multi-parametric MRI is helpful to predict tumor focality, stage, and size in patients diagnosed with unilateral low-risk prostate cancer. Prostate Cancer Prostatic Dis 2011; 14:232–237
7.
Tamada T, Sone T, Higashi H, et al. Prostate cancer detection in patients with total serum prostate-specific antigen levels of 4-10 ng/mL: diagnostic efficacy of diffusion-weighted imaging, dynamic contrast-enhanced MRI, and T2-weighted imaging. AJR 2011; 197:664–670
8.
Tanimoto A, Nakashima J, Kohno H, Shinmoto H, Kuribayashi S. Prostate cancer screening: the clinical value of diffusion-weighted imaging and dynamic MR imaging in combination with T2-weighted imaging. J Magn Reson Imaging 2007; 25:146–152
9.
Delongchamps NB, Rouanne M, Flam T, et al. Multiparametric magnetic resonance imaging for the detection and localization of prostate cancer: combination of T2-weighted, dynamic contrast-enhanced and diffusion-weighted imaging. BJU Int 2011; 107:1411–1418
10.
Kitajima K, Kaji Y, Fukabori Y, Yoshida K, Suganuma N, Sugimura K. Prostate cancer detection with 3 T MRI: comparison of diffusion-weighted imaging and dynamic contrast-enhanced MRI in combination with T2-weighted imaging. J Magn Reson Imaging 2010; 31:625–631
11.
Schimmöller L, Quentin M, Arsov C, et al. MR-sequences for prostate cancer diagnostics: validation based on the PI-RADS scoring system and targeted MR-guided in-bore biopsy. Eur Radiol 2014; 24:2582–2589
12.
Fascelli M, Rais-Bahrami S, Sankineni S, et al. Combined biparametric prostate magnetic resonance imaging and prostate-specific antigen in the detection of prostate cancer: a validation study in a biopsy-naive patient population. Urology 2016; 88:125–134
13.
Rais-Bahrami S, Siddiqui MM, Vourganti S, et al. Diagnostic value of biparametric magnetic resonance imaging (MRI) as an adjunct to prostate-specific antigen (PSA)-based detection of prostate cancer in men without prior biopsies. BJU Int 2015; 115:381–388
14.
Moher D, Shamseer L, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015; 4:1
15.
Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005; 58:882–893
16.
AbdelMaboud NM, Elsaid HH, Aboubeih EA. The role of diffusion-weighted MRI in evaluation of prostate cancer. Egypt J Radiol Nucl Med 2014; 45:231–236
17.
Barth BK, De Visschere PJ, Cornelius A, et al. Detection of clinically significant prostate cancer: short dual-pulse sequence versus standard multi-parametric MR imaging—a multireader study. Radiology 2017; 284:725–736
18.
Brock M, Roghmann F, Sonntag C, et al. Fusion of magnetic resonance imaging and real-time elastography to visualize prostate cancer: a prospective analysis using whole mount sections after radical prostatectomy. Ultraschall Med 2015; 36:355–361
19.
Doo KW, Sung DJ, Park BJ, et al. Detectability of low and intermediate or high risk prostate cancer with combined T2-weighted and diffusion-weighted MRI. Eur Radiol 2012; 22:1812–1819
20.
Franiel T, Stephan C, Erbersdobler A, et al. Areas suspicious for prostate cancer: MR-guided biopsy in patients with at least one transrectal US-guided biopsy with a negative finding—multiparametric MR imaging for detection and biopsy planning. Radiology 2011; 259:162–172
21.
Haider MA, van der Kwast TH, Tanguay J, et al. Combined T2-weighted and diffusion-weighted MRI for localization of prostate cancer. AJR 2007; 189:323–328
22.
Jung SI, Donati OF, Vargas HA, Goldman D, Hricak H, Akin O. Transition zone prostate cancer: incremental value of diffusion-weighted endorectal MR imaging in tumor detection and assessment of aggressiveness. Radiology 2013; 269:493–503
23.
Katahira K, Takahara T, Kwee TC, et al. Ultra-high-b-value diffusion-weighted MR imaging for the detection of prostate cancer: evaluation in 201 cases with histopathological correlation. Eur Radiol 2011; 21:188–196
24.
Kitamura K, Muto S, Yokota I, et al. Feasibility of multiparametric prostate magnetic resonance imaging in the detection of cancer distribution: histopathological correlation with prostatectomy specimens. Prostate Int 2014; 2:188–195
25.
Lawrence EM, Tang SY, Barrett T, et al. Prostate cancer: performance characteristics of combined T2W and DW-MRI scoring in the setting of template transperineal re-biopsy using MR-TRUS fusion. Eur Radiol 2014; 24:1497–1505
26.
Lim HK, Kim JK, Kim KA, Cho KS. Prostate cancer: apparent diffusion coefficient map with T2-weighted images for detection: a multireader study. Radiology 2009; 250:145–151
27.
Morgan VA, Kyriazi S, Ashley SE, DeSouza NM. Evaluation of the potential of diffusion-weighted imaging in prostate cancer detection. Acta Radiol 2007; 48:695–703
28.
Mussi TC, Martins T, Garcia RG, Filippi RZ, Lemos GC, Baroni RH. Are dynamic contrast-enhanced images necessary for prostate cancer detection on multiparametric magnetic resonance imaging? Clin Genitourin Cancer 2017; 15:e447–e454
29.
Naiki T, Okamura T, Nagata D, et al. Preoperative prediction of neurovascular bundle involvement of localized prostate cancer by combined T2 and diffusion-weighted imaging of magnetic resonance imaging, number of positive biopsy cores, and Gleason score. Asian Pac J Cancer Prev 2011; 12:909–913
30.
Rinaldi D, Fiocchi F, Ligabue G, Bianchi G, Torricelli P. Role of diffusion-weighted magnetic resonance imaging in prostate cancer evaluation. Radiol Med (Torino) 2012; 117:1429–1440
31.
Rosenkrantz AB, Mannelli L, Kong X, et al. Prostate cancer: utility of fusion of T2-weighted and high b-value diffusion-weighted images for peripheral zone tumor detection and localization. J Magn Reson Imaging 2011; 34:95–100
32.
Scialpi M, Prosperi E, D'Andrea A, et al. Bipara-metric versus multiparametric MRI with non-endorectal coil at 3T in the detection and localization of prostate cancer. Anticancer Res 2017; 37:1263–1271
33.
Shimofusa R, Fujimoto H, Akamata H, et al. Diffusion-weighted imaging of prostate cancer. J Comput Assist Tomogr 2005; 29:149–153
34.
Stanzione A, Imbriaco M, Cocozza S, et al. Biparametric 3T magnetic resonance imaging for prostatic cancer detection in a biopsy-naïve patient population: a further improvement of PI-RADS v2? Eur J Radiol 2016; 85:2269–2274
35.
Thestrup KC, Logager V, Baslev I, Møller JM, Hansen RH, Thomsen HS. Biparametric versus multiparametric MRI in the diagnosis of prostate cancer. Acta Radiol Open 2016; 5:2058460116663046
36.
Ueno Y, Takahashi S, Kitajima K, et al. Computed diffusion-weighted imaging using 3-T magnetic resonance imaging for prostate cancer diagnosis. Eur Radiol 2013; 23:3509–3516
37.
Ueno Y, Kitajima K, Sugimura K, et al. Ultra-high b-value diffusion-weighted MRI for the detection of prostate cancer with 3-T MRI. J Magn Reson Imaging 2013; 38:154–160
38.
Vargas HA, Akin O, Franiel T, et al. Diffusion-weighted endorectal MR imaging at 3 T for prostate cancer: tumor detection and assessment of aggressiveness. Radiology 2011; 259:775–784
39.
Vilanova JC, Barceló-Vidal C, Comet J, et al. Usefulness of prebiopsy multifunctional and morphologic MRI combined with free-to-total prostate-specific antigen ratio in the detection of prostate cancer. AJR 2011; 196:[web]W715–W722
40.
Wang X, Wang JY, Li CM, et al. Evaluation of the Prostate Imaging Reporting and Data System for magnetic resonance imaging diagnosis of prostate cancer in patients with prostate-specific antigen <20 ng/ml. Chin Med J (Engl) 2016; 129:1432–1438
41.
Yağci AB, Ozari N, Aybek Z, Düzcan E. The value of diffusion-weighted MRI for prostate cancer detection and localization. Diagn Interv Radiol 2011; 17:130–134
42.
Yoshimitsu K, Kiyoshima K, Irie H, et al. Usefulness of apparent diffusion coefficient map in diagnosing prostate carcinoma: correlation with stepwise histopathology. J Magn Reson Imaging 2008; 27:132–139
43.
Yoshizako T, Wada A, Hayashi T, et al. Usefulness of diffusion-weighted imaging and dynamic contrast-enhanced magnetic resonance imaging in the diagnosis of prostate transition-zone cancer. Acta Radiol 2008; 49:1207–1213
44.
Wu LM, Xu JR, Ye YQ, Lu Q, Hu JN. The clinical value of diffusion-weighted imaging in combination with T2-weighted imaging in diagnosing prostate carcinoma: a systematic review and meta-analysis. AJR 2012; 199:103–110
45.
Rosenkrantz AB, Hindman N, Lim RP, et al. Diffusion-weighted imaging of the prostate: comparison of b1000 and b2000 image sets for index lesion detection. J Magn Reson Imaging 2013; 38:694–700
46.
Maas MC, Fütterer JJ, Scheenen TW. Quantitative evaluation of computed high b value diffusion-weighted magnetic resonance imaging of the prostate. Invest Radiol 2013; 48:779–786
47.
Verma S, Sarkar S, Young J, et al. Evaluation of the impact of computed high b-value diffusion-weighted imaging on prostate cancer detection. Abdom Radiol (NY) 2016; 41:934–945
48.
Vargas HA, Hötker AM, Goldman DA, et al. Updated prostate imaging reporting and data system (PIRADS v2) recommendations for the detection of clinically significant prostate cancer using multiparametric MRI: critical evaluation using whole-mount pathology as standard of reference. Eur Radiol 2016; 26:1601–1612
49.
De Cobelli F, Ravelli S, Esposito A, et al. Apparent diffusion coefficient value and ratio as noninvasive potential biomarkers to predict prostate cancer grading: comparison with prostate biopsy and radical prostatectomy specimen. AJR 2015; 204:550–557
Information & Authors
Information
Published In
Copyright
© American Roentgen Ray Society.
History
Submitted: August 23, 2017
Accepted: December 10, 2017
Version of record online: June 12, 2018
Keywords
Authors
Funding Information
Supported by the Health and Family Planning Commission of Chengdu (Sichuan, China) (grant 2015080) and Sichuan Province (grant 17PJ430), Youth Innovation Research Fund of Sichuan Medicine (grant Q14004).
Metrics & Citations
Metrics
Citations
Export Citations
To download the citation to this article, select your reference manager software.