Diagnostic Performance of PSA Density for Detecting Clinically Significant Prostate Cancer in the Era of MRI: A Systematic Review and Meta-analysis

Abstract

Context

There has been a dramatic increase in the use of prostate magnetic resonance imaging (MRI) in the diagnostic workup. With prostate volume calculated from MRI, prostate-specific antigen density (PSAD) now is a ready-to-use parameter for prostate cancer (PCa) risk stratification before prostate biopsy, especially among patients with negative MRI or equivocal lesions.

Objective

In this review, we aimed to evaluate the diagnostic performance of PSAD for clinically significant prostate cancer (CSPCa) among patients who received MRI before prostate biopsy.

Evidence acquisition

Two investigators performed a systematic review according of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) criteria. Studies (published between Jan 01, 2012, and Dec 31, 2021) reporting the diagnostic performance (Outcomes) of PSAD (Intervention) for CSPCa among men who received pre-biopsy prostate MRI and subsequent prostate biopsy (Patients), using biopsy pathology as the gold standard (Comparison) were eligible for inclusion.

Evidence synthesis

1,536 papers were identified in PubMed, Scopus, and Embase. Of these, 248 studies were reviewed in detail, and 39 were qualified. The pooled sensitivities (SENS) and specificities (SPEC) for diagnosing CSPCa among patients with positive MRIs were 0.87 and 0.35 for PSAD of 0.1, 0.74 and 0.61 for PSAD of 0.15, and 0.51 and 0.81 for PSAD of 0.2 ng/ml/ml. The pooled SENS and SPEC for diagnosing CSPCa among patients with negative MRIs were 0.85 and 0.36 for PSAD of 0.1, 0.60 and 0.66 for PSAD of 0.15, and 0.33 and 0.84 for PSAD of 0.2 ng/ml/ml. The pooled SENS and SPEC among patients with PI-RADS 3 or Likert 3 lesions were 0.87 and 0.39 for PSAD of 0.1, 0.61 and 0.69 for PSAD of 0.15, and 0.42 and 0.82 for PSAD of 0.2 ng/ml/ml. The post-test probability for CSPCa among patients with negative MRIs was 6% if PSAD<0.15 and dropped to 4% if PSAD<0.10 ng/ml/ml.

Conclusions

In this systematic review, we quantitatively evaluated the diagnosis performance of PSAD for CSPCa in combination with prostate MRI. It demonstrated a complementary performance and predictive value, especially among patients with negative MRI and PI-RADS 3 or Likert 3 lesions. Integration of PSAD into decision making for prostate biopsy may facilitate improved risk-adjusted care.

Patient summary

PSAD is a ready-to-use parameter in the era of increased MRI use in CSPCa diagnosis. Findings suggest that the chance of having CSPCa was very low (4% or 6% for those with a negative pre-biopsy MRI or PIRADS (Likert) Score 3 lesion, respectively, if the PSAD<0.10ng/ml/ml), which may lower the need for biopsy in these patients.

1. Introduction

Recently, the use of prostate magnetic resonance imaging (MRI) has been dramatically increased in the diagnostic pathway for PCa [1]. It enhances PCa risk stratification so that prostate biopsy might be omitted in low-risk patients such as those with PIRADS 1 and 2 lesions [2]. In addition, subsequent biopsy guided by MRI combined with systematic biopsy can identify more clinically significant PCa (CSPCa), compared to systematic biopsy alone [3, 4]. Considering that MRI along or with subsequent targeted biopsy has a low positive predict value (PPV) [5], and the high negative predict value (NPV) still will miss a number of CSPCa [6], another useful parameter is needed as a complementary to MRI to increase the confidence of either performing or omitting a biopsy.

Prostate-specific antigen (PSA) has long been used in clinical practice for prostate cancer (PCa) screening. However, an elevated PSA level is not cancer-specific and could be due to a simple age-related benign prostatic hyperplasia. The lack of specificity and low diagnostic efficacy prompted another parameter, PSA density (PSAD), as a more reliable predictor for PCa. While digital rectal examination and trans-rectal ultrasound (TRUS) can be used to measure prostate volume (PV), the accuracy is insufficient and variation between performers exists [7][8]. Moreover, TRUS is often performed concurrently with biopsy and rarely performed as a separate procedure to assess size, thus, acquired too late to guide whether to perform biopsy. With MRI and the PV, PSAD become a ready-to-use parameter before biopsy, to further help stratify patients.

Despite the usefulness of PSAD, there is a lack of agreement on the optimal cutoff for avoiding prostate biopsy. The European Association of Urology (EAU) guideline suggests PSAD over 0.1–0.15ng/ml/ml may be predictive of cancer, while PSAD lower than 0.99ng/ml/ml were less likely (4%) to be diagnosed with CSPCa [9]. Specifically, the systematic review [10] cited by EAU on PIRADS 3 lesions was limited due to the absence of quantitative synthesis to justify the use of 0.15 ng/ml/ml as the optional cutoff. In addition, the recommendation by EAU on risk-stratification for CSPCa using PSAD was based on a systematic review study [11] using only prevalence of CSPCa from included studies, which did not incorporate the diagnostic performance of PSAD for cancer prediction. Thus, the quantitative risk stratification with different PSAD cutoffs is worth further investigation. In this study, we aimed to systematically review the literature reporting the performance of PSAD in diagnosing CSPCa and quantitatively calculate the risk of having CSPCa with different PSAD cutoffs and MRI results.

2. Evidence acquisition

2.1. Objective

The objective is to quantitatively evaluate the performance of PSAD in the diagnosis of CSPCa among patients with pre-biopsy MRI, using the biopsy pathology as the reference standard. In addition, we performed a subgroup analysis to further demonstrate its performance among different populations.

2.2. Search strategy

A systematic review of the literature was performed following the guideline from the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [12]. This study was registered in the PROSPERO database (registration number CRD42022287891). PubMed, Scopus, and Embase were searched for eligible studies published between Jan 01, 2012, and Dec 31, 2021, using the search query (PSA Density) AND (prostate cancer) AND (MRI), search details are listed in Supplementary 1. We conducted this review using the PICO principle, which was the diagnostic performance (Outcome) of PSAD (Intervention) for CSPCa among patients who received pre-biopsy prostate MRI and subsequent prostate biopsy (Patients), using the biopsy pathology (Comparison) as the gold diagnosis standard for diagnosis.

2.3. Inclusion and exclusion criteria

We selected studies based on the following proposed PCa diagnostic algorithm: patients (biopsy naïve or with previously negative biopsy) with any indication for prostate biopsy who received prostate MRI; this determined the subsequent biopsy procedure: for patients with negative MRI, only systematic biopsy (SB) was performed; otherwise, lesion-targeted biopsy (TB) with or without concurrent SB was performed. All studies published between Jan 1, 2012 (the year when PI-RADS v1.0 was first introduced, which was updated as PI-RADS v2.1 in 2019) and Dec 31, 2021, reporting the diagnostic performance of PSAD for CSPCa among patients following the above algorithm were eligible for review. Inclusion criteria were as follows: written in English; full texts available; biopsy-naïve patients or patients with previous negative biopsies; cancer detection rate reported on the patient level. Studies investigating multiple PSAD cut-offs were also included [13]. We preferred the PV calculation based on the ellipsoid formula, which is (1/6 × π × height × width × length) if multiple equations were evaluated [14]. We did not set any criteria for the MRI parameters and MRI-guided biopsy procedure (MRI/US fusion biopsy, MRI-guided cognitive biopsy, and others). Exclusion criteria were studies: published as reviews, correspondences, commentaries, editorials, book chapters, and meeting abstracts; only reporting Area Under Curve (AUC) without a 2-by-2 table or equivalent data; using prostatectomy or TURP specimen as pathology reference. Studies were also excluded if they indicated that patients received treatments or procedures (such as 5α-reductase, previous TURP, and others) that influence on the PSA or PV. Furthermore, studies that did not report PSAD performance stratified by MRI results, or with a sample size of fewer than 50 patients were included in the summary (i.e., features of the studies) but excluded from subsequent meta-analyses. We also scanned the references of all papers included for additional eligible studies. All study designs as original research were accepted. If more than one report of the same cohort was found, only the most up-to-date publication was included in the analysis.

2.4. Study definitions

The primary outcome of interest is to summarize the diagnostic performance of PSAD for predicting CSPCa among biopsy-naiïve patients or patients with previously negative biopsies. We allowed the definition for CSPCa from each study. In case the definition was not specified, but Gleason Score (GS) or Gleason Group (GG) was given, cancer with a GS of ≥3+4 (or 7) or a GG≥2 was considered clinically significant. A positive MRI was defined by each study or was considered positive if any MRI visible lesion was targeted (PIRADS or Likert 3–5 lesions). In addition, the PI-RADS system was considered equal to the Likert score system in this study [15].

2.5. Data extraction

Two investigators (SW and JK) independently screened all abstracts and full texts. Any disagreements were resolved by discussion between the two and a third investigator (MMS) until consensus was reached. The following information was extracted from eligible studies into a pre-designed excel sheet: first author, year of publication, country or area, study design, study period, number of patients, patient characteristics including mean or median age, PSA level, PV, PSAD, MRI parameters, patient cohort compositions, biopsy approach, CSPCa detection rate, PSAD performance including true positive (TP), true negative (TN), false positive (FP), and false negative (FN).

2.6. Quality and bias assessment.

The quality and bias assessment was carried out by the above two investigators independently, referring to the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool for diagnostic accuracy studies, which includes 14 questions covering 4 domains (patient selection, index test, reference standard, and flow/timing) [16].

2.7. Data analysis

2.7.1. Descriptive summary

We first summarized descriptive characteristics of eligible studies, including the cutoff values, Sensitivity (SENS), and Specificity (SPEC) of PSAD for predicting CSPCa. Usually, a 2-by-2 table includes the numbers of TP, TN, FP, and FN. Reviewers would reconstruct the 2-by-2 table from available data if no direct information was presented (Supplementary 2, 2-by-2 table demonstration). If more than one PSAD cut-off value was reported, each was included for subsequent summary and meta-analysis.

2.7.2. Heterogeneity and threshold effect evaluation

As a continuous parameter, PSAD was evaluated at a mixture of different cutoffs, indicating a possible threshold effect, which will cause heterogeneity. Thus, Higgins’ I² statistics were only used when we evaluated the heterogeneity of studies that reported the performance for the same PSAD cutoff. [17] [18]. A value of 0% indicates no observed heterogeneity, and values greater than 50% may be considered substantial heterogeneity. The causes of significant heterogeneity were explored using subgroup analyses. A summary receiver operating characteristic (SROC) curve was plotted for each analysis combining different PSAD cutoffs and MRI results. From the plot, heterogeneity can also be visualized and evaluated by the distribution of the included studies within the 95% confidence contour/prediction contour and their proximity to the summary curve.

2.7.3. Data synthesis and calculation of post-test probability

Forest plots with 95% CIs were calculated and depicted. We evaluated the pooled SENS and SPEC at different PSAD cutoffs (0.10, 0.15, and 0.20 ng/ml/ml) using the bivariate random-effects mode. The interpretation of a test depends on both the prevalence of the disease that is currently under investigation, and the performance of a test used. Since our aim is to quantitatively calculate the possibility of CSPCa when using PSAD, the ability of PSAD to make a correct diagnosis (either TP or TN) is critical and will influence the probability after the test. Herein, we calculate post-test probability using the Bayes’ Theorem, including both pre-test probability (prevalence observed in the studies) and the likelihood ratio (LR, the probability of correctly predicting disease in ratio to the probability of incorrectly predicting disease):

• Pre-test probability=prevalence= TP/TP+FP+TN+FN

• Pre-test odds=prevalence/(1-prevalence)

• Post-test odds=pre-test odds × Positive LR or Negative LR

• Post-test probability=post-test odds/(1+post-test odds)

• Positive LR=SEN/(1-SPE)

• Negative LR=(1-SEN)/SPE

2.7.4. Publication bias

Deeks’ funnel plot was used to assess the likelihood of publication bias, with a p value<0.1 indicating possible bias. We conducted this evaluation at the level of subgroup studies to investigate this effect.

2.7.5. Analyses at different PSAD cutoffs

Studies that only reported the PSAD performance among all patients (not stratified by MRI results) were presented in Table 1. We performed meta-analyses in studies reporting the diagnostic performance of PSAD among patients with positive pre-biopsy MRI, negative pre-biopsy MRI or with PI-RADS 3 or Likert 3 lesions, combined with commonly used PSAD cutoffs (0.1, 0.15, and 0.2ng/ml/ml). All statistical analyses were conducted using Stata 16.0 for Mac OS (Statistics and plotting), Revman 5.4 (Quality evaluation and plotting), and Meta Disc (statistics). Statistical significance for all analyses (except publication bias) was defined as two-sided p<0.05.

Table 1.

Characters of studies included in this review.

Author	Country	Study design	Mean/median PSAD	MRI field strength	MRI parameters	% of MRI Results cohorts	Bx approach (TR or TP)	Defination of CSPCa	PSAD cutoffs
Abdi 2015 [19]	Canada	Retrospective	0.21	1.5T	mpMRI	3–5: 100%	TR	Epstein criteria: GS≥7, or >3 biopsy cores positive, or at least one biopsy core with > 50% involvement	0.150
Fascelli 2015 [20]	USA	Retrospective	0.16	N/A	mpMRI	Positive: 100%	TR	GS≥7	0.150
Niu 2017 [21]	China	Retrospective	0.17	3T	mpMRI	Mean score: 3.3	TR	GS≥7	0.160
Hansen 2017 [22]	UK+Germany	Prospective	0.15	1.5 or 3T	mpMRI	1–2: 30%, 3: 26%, 4: 21%, 5: 24%	TP	GS≥7 or GG≥2	0.150
Washino 2017 [23]	Japan	Retrospective	0.26	1.5 or 3T	mpMRI	1–2: 44.4%, 3: 14.9%, 4–5: 40.6%	TP	GS≥7 and/or a maximum cancer core length of ≥4 mm	0.15, 0.30
Hansen 2018 [24]	UK+German+Australia	Prospective	0.15	1.5 or 3T	mpMRI	1–2: 29%, 3: 19%, 4–5: 52%	TP	GS≥7	0.10, 0.15, 0.20
Cuocolo 2018 [25]	Italy	Retrospective	0.14	N/A	bp-MRI	1–2: 48.2, 3–5: 51.8%	TR	ISUP≥3	0.140
Borkowetz 2019 [26]	Germany	Retrospective	0.15	3T	mpMRI	2: 11%, 3: 35%, 4: 36, 5: 18%	TP and TR	Epstein criteria: GS≥7, or >3 biopsy cores positive, or at least one biopsy core with > 50% involvement	0.200
Boesen 2019 [27]	Denmark	Prospective	0.12	3T	bpMRI	1–2: 37%, 3: 15%, 4–5: 48%	TR	GS≥7	0.10. 0.15, 0.20
Oishi 2019 [28]	Japan	Prospective	N/A	3T	mpMRI	1–2: 100%	TR	GS≥7	0.10, 0.15
Elkhoury 2019 [29]	USA	Prospective	N/A	3T	mpMRI	Non-visible lesions: 100%	TR	GS≥7	0.150
Kim, L 2020 [30]	UK	Prospective	N/A	1.5 or 3T	mpMRI	1–2: 33%, 3–5: 77%	TR or TP	GG≥2, or Cambridge Prognostic Group≥3	0.10, 0.15, 0.20
Stevens 2020 [31]	USA	Retrospective	0.17	3T	mpMRI	1–2: 23.6%, 3: 23%, 4: 34%, 5: 19.4%	TR	GS≥7	0.10, 0.15, 0.20
Falagario 2020 [32]	Europe	Retrospective	0.13	1.5 or 3T	bpMRI or mpMRI	1–2: 19.1%, 3: 27.2%, 4: 32.3%, 5: 21.3%	TR	GS≥7	0.10, 0.15, 0.20
Sokhi 2020 [33]	UK	Retrospective	N/A	1.5T	mpMRI	1–2: 29%, 3: 8.1%, 4: 21.4%, 5: 42.9%	TR	GS≥7	0.12, 0.15
Takeshima 2020 [34]	Japan	Retrospective	0.27	1.5T	mpMRI	1–5: 100%	TR	ISUP≥2, bilateral cancer, or ≥3 positive cores	0.250
Zhang 2020 [35]	China	Retrospective	0.21	3T	mpMRI	1–2: 100%	TR	Epstein criteria: GS≥7, or >3 biopsy cores positive, or at least one biopsy core with > 50% involvement	0.15, 0.20
Knaapila 2020 [36]	Finland	Retrospective	0.18	1.5T or 3T	bpMRI	1–2: 23%, 3: 18%, 4–5: 59%	TR	GS≥7	0.10, 0.15, 0.20
Kinnaird 2020 [37]	USA	Retrospective	0.12	3T	mpMRI	1–2: 26%, 3:36%, 4: 29%, 5: 10%	TR	GG≥2	0.100
Deniffel 2020 [38]	USA	Retrospective	0.11	3T	mpMRI	1–2: 54.6%, 3: 27.3%, 4–5: 18.4%	TR	GS≥7 or GG≥2	0.078
Lim 2020 [39]	Canada	Retrospective	0.186	3T	mpMRI	3: 100%	TR	GS≥7 or GG≥2	0.215
Kim, KH 2020 [40]	South Korea	Retrospective	0.165	3T	mpMRI	3: 100%	TR	GS≥3+4	0.180
Zhang 2020 [41]	China	Prospective	0.21	3T	mpMRI	3: 100%	TR	Epstein criteria: GS≥4+3, a GS of 3+4 with PSA >10, >3 positive cores, or ≥1 bx core with >50% involvement.	0.15, 0.20
Liang 2021 [42]	China	Retrospective	0.21	3T	bpMRI	1–2: 100%	TR	GS≥7	0.10, 0.15
Girometti 2021 [43]	Italy	Retrospective	0.12	3T	mpMRI	1–2: 24.4%, 3: 8.9%, 4: 38.2%, 5: 28.5%	TP	GS≥7	0.10, 0.15, 0.20
Kim 2021 [44]	Republic of Korea	Retrospective	N/A	3T	mpMRI	3: 46.4%, 4: 34%, 5: 19.6%	TR	GS≥7	0.200
Russo 2021 [45]	Italy	Clinical trial	N/A	1.5T	mpMRI or fast MRI	3: 100%	TR	GG≥2	0.120
Sekito 2021 [46]	Japan	Retrospective	0.24	1.5T	bpMRI	Positive: 76.6%, negative: 23.4%	TR	GG≥2	0.230
Stanzione 2021 [14]	Italy	Retrospective	0.149	3T	mpMRI	1–5: 100%	TR	ISUP>2	0.150
Avolio 2021 [47]	Italy	Retrospective	0.11	1.5 or 3T	mpMRI	3: 100%	TR or TP	GS≥7	0.150
Yu 2021 [48]	Republic of Korea	Retrospective	0.15	3T	mpMRI	1–2: 100%	TP	GS≥7	0.10, 0.15, 0.20
Görtz 2021 [49]	Germany	Prospective	0.1	3T	mpMRI	3: 100%	TP	D’Amico risk classification: clinical stage≥T2b or PSA>10 ng/ml or GS≥7	0.10, 0.15
Stonier 2021 [50]	UK	Retrospective	N/A	1.5 or 3T	mpMRI	1–2: 48.3%, 3: 51.7%	TP or TR	GG≥2	0.12. 0.15, 0.20
Zheng 2021 [51]	USA	Retrospective	0.11	3T	mpMRI	1–2: 100%	TR	GS≥7	0.10, 0.15
Pan 2021 [52]	China	Retrospective	0.28	1.5T	bpMRI	1: 4.7%, 2: 24.7%, 3: 28.5%, 4: 29.4%, 5: 12.6%	TP	GS≥7	0.29, 0.40
Buisset 2021 [53]	Franxe	Retrospective	0.13	1.5T	mpMRI	1–2: 100%	TR	GG1 and maximum cancer core length >5 mm, or positive systematic biopsy cores ≥3, and any GG≥2	0.10, 0.14, 0.15
Lophatananon 3032 [54]	UK	Retrospective	0.11	1.5 or 3T	mpMRI	Score 1–5: 100%	N/A	GG≥2	0.10, 0.15
Kaufmann 2021 [55]	Switzerland	Prospective	0.14	3T	mpMR	1–2: 31.6%, 3: 22%, 4: 31%, 5: 15%	TP	GS≥7	0.07, 0.10, 0.15, 0.20
Gan 2021 [56]	USA	Retrospective	N/A	1.5 or 3T	bpMRI or mpMRI	1–2: 100%	N/A	GS≥7	0.10, 0.15, 0.20

3. Evidence synthesis

Our initial search returned 1,536 records, of which 248 with full text underwent detailed review.

Finally, 39 studies were eligible for inclusion in this review (Figure 1). Critical characteristics for each study were summarized in Table 1 (Supplementary 3 shows the full table with detailed information). The numbers of studies reporting the PSAD performance in the diagnosis of CSPCa were as follows: with all MRI results (n=19), positive MRI (n=13), negative MRI (n=18), and PI-RADS 3/Likert 3 lesions (n=19).

Figure 1.

Flowchart for the selection of eligible studies in this review following PRISMA.

Regarding the study design, only one study was a clinical trial, 9 studies were prospective, 29 were retrospective. 15 studies included only biopsy naïve patients, 6 studies included only patients with negative previous biopsy, and 18 studies included both biopsy naïve patients and patients with negative previous biopsy. 22 studies evaluated multiple PSAD cutoffs. Among the 74 cutoffs, 0.15ng/ml/ml was the most evaluated (27/74, 36%), followed by 0.1ng/ml/ml (17/74, 23%) and 0.2ng/ml/ml (15/74, 20%). MpMRI was used in 33 studies (33/39, 85%), 3T field strength in 30 studies (30/39, 77%), and PI-RADS score system in 35 (35/39, 90%) studies. Most studies used GS≥7 or GG≥2 (with/without additional diagnostic criteria such as the number of positive cores, length, or percentage of cancer tissue) as the definition of CSPCa. In terms of the biopsy approach, 25 (25/39, 64%) studies used transrectal (TR), 8 (8/39, 21%) used transperineal (TP), and 4 (4/39, 10%) used either one or both within their cohorts.

3.1. Methodological quality of included studies

Overall, the included studies had low to moderate risk of bias (Figure 2). The most concerning domain was the patient selection as 51% (20/39) of the included studies did not indicate their selection as either consecutive or random. Especially among retrospective studies, the clinical decision to biopsy might have already been affected by PSAD, resulting in patient selection bias. Regarding the index and reference test domains, all studies did not mention if the interpretation was blinded. However, we consider it as low risk because the positivity or negativity of PSAD would not change once the cutoff was set, and the chance that any pathologists referred to PSAD for the biopsy histology reading was extremely low and unrealistic. As for the flow and timing domain, even though no studies indicated the interval between PSA test and prostate biopsy, it was considered as low risk since all included patients were supposed to receive timely medical care if there had any indications for a subsequent prostate biopsy.

Figure 2.

Methodological quality evaluation for risk of bias and applicability concerns using the QUADAS-2 tool.

3.2. The performance of PSAD among patients with positive pre-biopsy MRI

We evaluated the performance of PSAD in subgroups stratified by pre-biopsy MRI results as this would be more clinically relevant and informative. Among patients with positive MRIs, 13 studies evaluated the performance of 27 PSAD cutoffs in the diagnosis of CSPCa (Supplementary 3). The pooled SENSs and SPECs were 0.87 (95% CI 0.83–0.91, I²=83%) and 0.35 (95% CI 0.27–0.44, I²=91%) for PSAD cutoff of 0.1, 0.74 (95% CI 0.62–0.83, I²=92%) and 0.61 (95% CI 0.50–0.71, I²=92%) for PSAD cutoff of 0.15, 0.51 (95% CI 0.43–0.58, I²=93%) and 0.81 (95% CI 0.75–0.85, I²=93%) for PSAD cutoff of 0.2 (Figure 3). Post-test probabilities for CSPCa were 26% (PSAD<0.1) vs 57% (PSAD≥0.1) (p<0.001), 30% (PSAD<0.15) vs 65% (PSAD≥0.15) (p<0.001), and 34% (PSAD<0.20) vs 69% (PSAD≥0.20) (p<0.001) (Figure 4).

Figure 3.

Forest plots of the pooled SENS and SPEC for positive MRI, negative MRI, and PIRADS (Likert) 3 lesions, combined with PSAD cutoffs of 0.10, 0.15, and 0.20, respectively.

Figure 4.

Diagnostic performance of PSAD at different cutoffs and post-test probabilities for the prediction of CSPCa. Columns and percentages represent the possibility of having CSPCa at each PSAD cutoff (0.1, 0.15 and 0.20ng/ml/ml).

3.3. The performance of PSAD among patients with negative pre-biopsy MRI

For patients with negative pre-biopsy MRIs, 18 studies reported the performance of 44 PSAD cutoffs in the diagnosis of CSPCa (Supplementary 3). The pooled SENSs and SPECs were 0.85 (95% CI 0.77–0.90, I²=31%) and 0.36 (95% CI 0.29–0.45, I²=92%) for PSAD cutoff of 0.1, 0.60 (95% CI 0.48–0.72, I²=74%) and 0.66 (95% CI 0.58–0.74, I²=96%) for PSAD cutoff of 0.15, 0.33 (95% CI 0.27–0.40, I²=0%) and 0.84 (95% CI 0.73–0.91, I²=98%) for PSAD cutoff of 0.2 (Figure 3). Post-test probabilities for CSPCa were 4% (PSAD<0.1) vs 13% (PSAD≥0.1) (p<0.001), 6% (PSAD<0.15) vs 17% (PSAD≥0.15) (p<0.001), and 8% (PSAD<0.20) vs 18% (PSAD≥0.20) (p<0.001) (Figure 4).

3.4. PSAD performance among patients with PIRADS 3 or Likert 3 lesions

Specifically, we evaluated patients with PIRADS 3 or Likert 3 lesions on MRI. 19 studies investigated 37 PSAD cutoffs for their performance in the diagnosis of CSPCa (Supplementary 3). The pooled SENSs and SPECs were 0.87 (95%CI 0.82–0.91, I²=0%) and 0.39 (95% CI 0.29–0.50, I²=87%) for PSAD cutoff of 0.1, 0.61 (95% CI 0.50–0.71, I²=74%) and 0.69 (95% CI 0.62–0.76, I²=91%) for PSAD cutoff of 0.15, 0.42 (95% CI 0.32–0.53, I²=79%) and 0.82 (95% CI 0.75–0.87, I²=92%) for PSAD cutoff of 0.2 (Figure 3). Post-test probabilities for CSPCa were 6% (PSAD<0.1) vs 23% (PSAD≥0.1) (p<0.001), 12% (PSAD<0.15) vs 33% (PSAD≥0.15) (p<0.001), and 15% (PSAD<0.20) vs 36% (PSAD≥0.20) (p<0.001) (Figure 4).

3.5. Publication bias

As shown in Deeks’ funnel plots (Supplementary 4) suggest publication biases existed in the following meta-analyses: PSAD performance in the diagnosis of CSPCa among patients with positive MRIs (p=0.060) and PI-RADS 3 or Likert 3 lesions (p=0.013), while no publication bias was observed from studies reporting negative MRI (p=0.5).

3.6. Heterogeneity evaluation

Considerable heterogeneity was observed in all analyses, which was demonstrated by the exceptionally high I² values. Consequently, additional subgroup analyses were conducted, categorized by study design, biopsy history, MRI parameter, biopsy approach, and MRI report system. These findings are presented in Supplementary 5. No significant factors were identified as primary sources of heterogeneity, given that the I² values remained high even in subgroup analyses. The MRI reporting systems used varied substantially across studies, and subgroup analysis was restricted to those that used PIRADS V2, which yielded a stable and comparable pooled SENS and SPEC. Nevertheless, most studies on the SROC plot were close to the summary ROC curve (Supplementary 6).

4. Discussion

In this systematic review and meta-analysis, we were able to report the diagnostic performance of PSAD combined with pre-biopsy MRI at certain PSAD cutoffs. Patients with a PSAD value of <0.10 ng/ml/ml combined with negative MRI or PI-RADS 3/Likert 3 lesions, respectively, had a 4% or 6% chance of CSPCa detection from prostate biopsy. The chance increased to 6% or 12% if PSAD <0.15 ng/ml/ml with negative MRI or equivocal lesions, respectively. This complementary function of PSAD to MRI is critical in PCa risk stratification and decision making for prostate biopsy. Using the Bayes’ Theorem, this study provides the quantitative risk evaluation for different PSAD cutoffs combined with MRI results for the prediction of CSPCa in the MRI era, especially when PSAD cutoff of 0.15 ng/ml/ml is widely recommended without support of high-level evidence [57].

MRI has been increasingly used for the risk-stratification and diagnosis for PCa, and its performance has been well evaluated. Sathianathen et al. reported in a recent meta-analysis covering studies from 2016 to 2019 that the pooled NPV for CSPCa among biopsy-naïve patients was 90.8% (95%CI 88–93%) [6]. Considering that the variation among different settings will have impact on the MRI performance, the percentage of patients that can safely omit biopsy is still not insignificant. In addition, another recent meta-analysis covering studies from 2015 to 2020 showed that the pooled PPV of MRI in predicting CSPCa is only 40% (95%CI 36–43%) [5]. Therefore, combining pre-biopsy MRI with other clinical information, such as PSAD in this study, is critical for clinical decision-making.

Results from this study showed that the PSAD performs well in “ruling-out” CSPCa among patients with negative MRI and equivocal lesions. When using PSAD of 0.15 ng/ml/ml as the cutoff, the post-test probability of having CSPCa for a patient with a negative MRI scan can be as low as 6%, given a pre-test probability of 10%. Lowering the cutoff to 0.10 ng/ml/ml, the post-test probability of CSPCa was only 4%. This is of clinical significance because it suggests that many patients with <0.1 ng/ml/ml PSAD may avoid invasive prostate biopsy. This is often seen in patients with large prostates and only slightly elevated PSAs, yet negative MRIs.

Whether to biopsy PI-RADS 3 or Likert 3 lesions is still controversial because of the low possibility of CSPCa in this “grey zone” [58][59]. Maggi et al. reported in a systematic review covering studies from 2009–2019 that the prevalence of CSPCa detection rate was 18.5% (95% CI 16.6–20.3%) among patients with PIRADS 3 lesions. The authors suggested that the evaluation of clinical predictors, other than MRI alone, is a main aspect in terms of biopsy decision making [10]. In this study, we found that the likelihoods of having CSPCa in patients with only PI-RADS 3 or Likert 3 lesions was 12% with PSAD<0.15 ng/ml/ml and 6% with PSAD<0.10 ng/ml/ml. There have been several proposals for the management of PIRADS 3 lesions, such as no biopsy at the initial visit followed by MRI monitoring or adding lesion size/biogenetic markers as complementary parameters when considering biopsy [58]. Further work is needed, however the value of PSAD to risk-stratify equivocal lesions on MRI may potentially be beneficial for avoiding biopsy.

Compared to the “rule-out” function of PSAD in predicting CSPCa among patients with negative MRI, its use among patients with positive MRI can help “rule-in” CSPCa. A patient with a positive MRI and positive PSAD test can have a twice higher chance of harboring CSPCa (26% vs 57% for PSAD 0.1, 30% vs 65% for PSAD 0.15, and 34% vs 69% for PSAD 0.2). This possibility is assumed to be higher if only PIRADS 4 and 5 lesions are considered. Even though all guidelines recommend prostate biopsy for PI-RADS 4 or 5 lesions regardless of PSAD, PSAD can still provide further information on risk stratification among this group of patients and increase the confidence of performing prostate biopsy in predicting CSPCa.

If the combination of PSAD and MRI is used to decide where or not to proceed with biopsy, then the follow-up plan after omitting biopsy will be critical since there is chance of either missing CSPCa or disease progression. Following the change of PSAD and MRI is a reasonable and practical way to decide if prostate biopsy needs to be re-considered. An example that we can take from the active surveillance follow-up schedule is to monitor PSA (hence PSAD) and its kinetics, and MRI change (Prostate Cancer Radiologic Estimation of Change in Sequential Evaluation Recommendation, PRECISE) [60]. Other factors might also be taken into consideration such as previous biopsy results, family history, genomic risk, etc.

Our study is not free of limitations. Firstly, we observed significant heterogeneity across the analyses. While subgroup analyses were unable to identify factors explaining the heterogeneity, SROC plots indicated that, for most of the cutoff and MRI result combinations, the majority of the studies tended to fall within the 95% confidence contour. Additionally, most studies outside the contour remained close to the summary curve. As a second limitation, studies in this review included both biopsy-naïve and previously biopsy-negative patients. This may influence the cancer detection rate because studies report that patients with previously negative biopsy have a lower CSPCa detection rate than biopsy-naïve patients [61]. Third, the variation of the definition for CSPCa in each study may also influenced the cancer detection rate. However, this effect was subtle as most studies considered GS≥7 or GG≥2 as CSPCa, with or without additional diagnostic criteria such as the number of positive cores, length, or percentage of cancer tissue. Four, we did not set strict criteria for MRI parameters, and any forms of MRI evaluated with different versions of imaging standards were included. There is still ongoing debate on whether to use bpMRI vs mpMRI [62][63] and 1.5T vs 3T [64][65], and the variability in MRI interpretation may also have impacted the outcome. Fifth, we noticed publication biases among studies evaluating PSAD performance in the diagnosis of CSPCa among patients with positive MRIs and PI-RADS (Likert) 3 lesions. Lastly, no optimal PSAD cut-off can be generated, however, certain PSAD cut-offs evaluated in this study are noteworthy and could provide references during clinical practice. These limitations imply that the results of our study should be interpreted with caution, as the studies included a mixed patient populations, diagnostic techniques, and other undetected factors.

5. Conclusion

In this systematic review, we evaluated the diagnostic performance of PSAD for CSPCa in combination with mpMRI. With negative MRI, PSAD value of <0.10 and <0.15 resulted in only 4% and 6% chance of finding CSPCa from prostate biopsy. With equivocal lesions on MRI, PSAD value of <0.10 and <0.15 resulted in 6% and 12% chance of CSPCa. The data suggest that with negative MRI or PI-RADS 3 or Likert 3 lesions combined with a low PSAD value, the risk of having CSPCa is low, and biopsy may be avoided. Integration of PSAD into decision making for prostate biopsy may facilitate improved risk-adjusted care.