Document Type : Systematic Review/ Meta-analysis
Authors
1 Urology Research Center, Tehran University of Medical Sciences, Tehran, Iran
2 Chronic Diseases Research Center, Endocrinology and Metabolism Population Sciences Institute, Tehran University of Medical Sciences, Tehran, Iran
3 Department of Internal Medicine, Faculty of Medicine, Sina Hospital, Tehran University of Medical Sciences, Tehran Iran
Abstract
Highlights
Keywords
Introduction
Prostate cancer (PCa) is one of the most prevalent cancers and is the second source of death in American men with a global incidence of 49.6 per 10000 persons (1). Prostate-Specific Antigen (PSA) is being commonly used for the detection of PCa as it is cheap and easy to access; however, its sensitivity and specificity are not adequate as an ideal tumor marker (2). Other diagnostic modalities including PSA density, prostate health index (PHI), prostate imaging reporting and data system (PI-RADS), and prostate cancer antigen3 (PCA3) have gained a great deal of attention as accurate diagnostic tests (3-5).
PCA3 is a urinary biomarker that is overexpressed in PCa and has been shown to have more specificity compared to PSA, especially in repeated biopsy patients (6). PHI is a combination formula of total PSA, free PSA, and proPSA and enjoys greater specificity than PSA, especially for clinically significant PCa detection. Also, increased PHI is an index of cancer recurrence after radical prostatectomy (7, 8). PI-RADS is another diagnostic method for PCa with good accuracy, though it has a significant heterogeneity (9). PI-RADS is a scoring system for each MRI sequence and is reported as a per lesion score. Investigations of early PCa diagnosis have indicated different methods with each having limitations despite good accuracy (9, 10). A prostate biopsy is a gold standard for PCa diagnosis (11) and many papers have been published comparing its accuracy versus diagnostic accuracy of PCA3, PHI, and PI-RADS individually (6, 12, 13). The object of this study was to compare three methods together based on their sensitivity, specificity, and accuracy to determine an overall picture of accurate PCa diagnosis.
Methods
The protocol of the study was saved in PROSPERO; International Prospective Register of Systematic Reviews, with
ID= CRD42018089099.
Search strategy
The systematic search was conducted in the international databases including PubMed, Scopus, and Web of Science from January 2000 to Feb 2018 according to PRISMA guideline (14). The search terms were: (prostate or prostatic) AND (cancer or carcinoma or neoplasm or malignancy or tumor) AND (assessment or diagnosis or (sensitivity and specificity) or detection) AND (biopsy or pathology or histopathology) AND [ (“Prostate Imaging Reporting and Data System”) or PI-RADS or PIRADS) OR (“Prostate Cancer Antigen3” or PCA3 or dd3 or upm3 or “differential display code 3 antigen") OR (Prostate Health Index or PHI or “[-2] proPSA). Additionally, the reference list of each relevant article was reviewed. Furthermore, gray literature, such as reports and conference presentations were checked.
Study selection
We included three groups of articles in this systematic review to evaluate the diagnostic tests of PCa. The common criteria included 1) enrolled patients with suspected or early diagnosis PCa; 2) for comparison, a gold standard based on the histopathological examination of biopsy; 3) sufficient data to calculate true positive (TP), false negative (FN), false positive (FP) and true negative (TN) values for PCa diagnosis; and 4) studies that were original articles. On the other hand, the specific criteria were: a) for PI-RADS V2 test, MRI of the prostate including all sequences performed and assessed by a PI-RADS scoring system; 2) for PCA3 test, all patients undergoing PCA3 testing before biopsy; and 3) for PHI test, calculating of this index using the formula of serum levels of fPSA, tPSA, and p2PSA.
Studies with no usable data, receiving therapy, aggressive PCa, non-English full-text papers, and studies with overlapping patient populations were excluded. Also, review articles, letters to editors, animal studies, and case-report studies were excluded.
Two researchers independently performed the screening process based on titles, abstracts, and then full texts of selected papers. Possible disagreements were resolved by consensus.
Data extraction and quality evaluation
We extracted the following data from each paper: the name of the first author, the year of publication, country of data collection, study design, patient age, number of PCa, descriptions of the diagnostic tests, and cut-off values. For each study, values of TP, FN, FP, TN, sensitivity, specificity, and area under the curve (AUC) were extracted if available.
The Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) was used for the quality assessment of included studies (15). Data extraction and quality assessment were performed by two separate researchers and all disagreements were resolved by a third reviewer.
According to each diagnostic test (PIRADS, PCA3, and PHI), the sensitivity, specificity, and diagnostic odds ratio (DOR) were obtained for every study and subsequently pooled. The heterogeneity of included studies was assessed by Q test and I2statistic. If the results of the Q test were significant and I2> 50%, the random effect model was selected. We also assessed the test performance using the summary receiver operation characteristics (SROC) curve and AUC. The Deeks funnel plot was used to assess potential publication bias. We used Fagan’s nomogram to estimate the clinical value of the diagnostic test.
All meta-analysis methods were performed by STATA (Release 14. statistical software. College Station, Texas: STATA Corp LP).
Results
Literature search and study selection
As displayed in Figure 1, the literature searches initially identified 1702 articles. According to inclusion criteria, we assessed titles and abstracts, where 485 articles (PHI=90, PI-RADS=254, and PCA3=141 papers) were selected. After reviewing the full texts, a total of 151 articles (PHI=33, PI-RADS=71, and PCA3=45 papers) remained and finally, 26 studies on PHI, 24 articles on PI-RADS, and 26 papers on PCA3 were eligible for the meta-analysis. The general characteristics of the included studies are presented in Table 1.
Figure 1. Flow diagram of the study selection process
Table 1-1. Characteristics of studies included in the meta-analysis according to PI-RADS
First Author |
Year |
Country/Region |
Sample Size |
PCa number |
Patient Age (Yr) (mean/ median) |
Cutoff value |
TP |
FP |
FN |
TN |
AUC (95% CI) |
Garcia-Reyes, K(12) |
2013-2016 |
USA |
178 |
152 |
64.7(44-83) |
NA |
95 |
4 |
57 |
22 |
0.830 |
Xu, N (17) |
2015-2017 |
China |
528 |
137 |
65(52-82) |
3.0 |
122 |
148 |
15 |
243 |
0.836±0.03 |
Borkowetz, A (18) |
2015 |
Germany |
214 |
111 |
63(40-75) |
4.0 |
92. |
38 |
19 |
65 |
NA |
Rosenkrantz, A. B.(19) |
NA |
USA |
60 |
30 |
NA |
3.0 |
20 |
9 |
10 |
21 |
NA |
Kuhl, C. K (20) |
2013-2016 |
Germany |
542 |
180 |
64.8(42-80) |
NA |
156 |
43 |
24 |
319 |
NA |
Kim, S. H (21) |
2014-2016 |
Korea |
295 |
160 |
67(51-79) |
4.0 |
144 |
27 |
16 |
108 |
NA |
Gao, G (22) |
2008-2010 |
China |
71 |
35 |
68.8±8.9 |
4.0 |
30 |
2 |
5 |
34 |
0.906±0.03 |
Nougaret, S (23) |
2007-2014 |
France |
371 |
292 |
60(41-81) |
4.0 |
95 |
44 |
12 |
220 |
NA |
Wang, R (24) |
2002-2009 |
China |
1478 |
507 |
70(65-75) |
3.0 |
412 |
100 |
95 |
871 |
0.931 |
Furuya, K(25) |
2012-2013 |
Japan |
50 |
33 |
68.5 (53-82) |
NA |
21 |
8 |
12 |
9 |
0.5830(0.435-0.731) |
Kasel-Seibert, M (26) |
2013-2015 |
Germany |
82 |
31 |
65(48-88) |
4.0 |
24 |
10 |
7 |
41 |
0.83±0.08 |
El-Samei, H (27) |
2014-2015 |
Egypt |
55 |
38 |
62(51-79) |
3.0 |
35 |
1 |
3 |
16 |
NA |
Feng, Z. Y (28) |
2013-2015 |
China |
401 |
150 |
64.4(34-88) |
3.0 |
144 |
40 |
6 |
211 |
0.942±0.03 |
Wang, X (29) |
2011-2013 |
Italy |
133 |
60 |
68±7.9 |
5.0 |
32 |
8 |
28 |
65 |
0.749±0.02 |
Polanec, S (30) |
2011-2015 |
Austria |
65 |
33 |
65.3(62-87) |
3.0 |
32 |
20 |
1 |
12 |
0.75 |
Sahibzada, I(31) |
2008-2011 |
UK |
200 |
111 |
69.8(59-86) |
NA |
41 |
13 |
70 |
76 |
NA |
Radtke, J. P (32) |
2013 |
Germany |
294 |
150 |
64(60-71) |
3.0 |
112 |
48 |
38 |
103 |
NA |
Rastinehad, A. R (33) |
2012-2014 |
USA |
312 |
202 |
65.1(60-70) |
3.0 |
191 |
75 |
11 |
35 |
0.702 |
Wang R (34) |
2002-2014 |
China |
142 |
55 |
68.6(26-91) |
3.0 |
50 |
20 |
5 |
67 |
0.90±0.05 |
Grey, A .D.R (35) |
2012-2013 |
UK |
201 |
77 |
64.5±7.1 |
3.0 |
74 |
50 |
3 |
74 |
0.89 |
Baur, A .D. J (36) |
2008-2012 |
Germany |
55 |
18 |
66(54-78) |
4.0 |
14 |
3 |
4 |
34 |
0.93 |
Roethke, M. C (37) |
2011-2012 |
Germany |
64 |
27 |
64.5(49-77) |
NA |
18 |
3 |
9 |
34 |
0.848±0.11 |
Junker, D (38) |
2011-2013 |
Austria |
73 |
39 |
62±7.8 |
NA |
35 |
13 |
4 |
21 |
0.86±0.08 |
Schimmöller, L (39) |
2011-2012 |
Germany |
67 |
28 |
66.8±7.5 |
10.0 |
24 |
13 |
4 |
26 |
NA |
TP: True Positive; FN: False Negative; FP: False Positive; TN: True Negative; AUC: Area Under the Curve; PI-RADS: Prostate Imaging Reporting & Data System; PCa ;Prostate Cancer
Table 1-2. Characteristics of studies included in the meta-analysis according to PHI
First Author |
Year |
Country/Region |
Sample Size |
PCa number |
Patient Age (Yr) (mean/ median) |
Cutoff value |
TP |
FP |
FN |
TN |
AUC (95% CI) |
Catalona, W.J (40) |
2003-2009 |
USA |
892 |
430 |
62.8±7 |
24.1 |
387 |
334 |
53 |
118 |
0.72 |
Seisen, T (41) |
2013 |
France |
138 |
39 |
63.4(44-83) |
40 |
26 |
26 |
13 |
73 |
0.8 |
Lazzeri, M (13) |
NA |
European |
262 |
136 |
67.3±8.1 |
43.7 |
123 |
76 |
13 |
50 |
0.81(0.75-0.85) |
Filella, X (42) |
2011-2013 |
Spain |
354 |
175 |
68(38-88) |
31.94 |
158 |
135 |
17 |
49 |
0.73 |
Stephan, C (43) |
2009-2012 |
Germany |
246 |
110 |
65(41-81) |
NA |
99 |
107 |
11 |
29 |
0.68(0.62-0.74) |
Al Saidi, S.S (44) |
2014-2015 |
Oman |
136 |
28 |
66(45-90) |
41.9 |
23 |
21 |
5 |
87 |
NA |
Porpiglia, F(5) |
2011-2013 |
Italy |
170 |
52 |
65(60-70) |
48.9 |
21 |
26 |
31 |
92 |
NA |
Loeb, S (45) |
2004-2009 |
USA |
658 |
324 |
63(50-84) |
27 |
292 |
220 |
32 |
104 |
0.707(0.665-0.73) |
Ng, C.F (46) |
2008-2013 |
China |
230 |
21 |
65.9(50-79) |
26.54 |
19 |
105 |
2 |
104 |
0.781(0.67–0.897) |
Osredkar, J (47) |
2013-2014 |
Slovenia |
110 |
36 |
67(63-72) |
25.6 |
32 |
53 |
4 |
21 |
0.742(0.65-0.82) |
Lazzeri, M(48) |
2011-2012 |
European |
646 |
264 |
64.2±7.2 |
27.6 |
238 |
308 |
26 |
74 |
0.67(0.64–0.71) |
Friedl, A (49) |
2014-2016 |
Austria |
112 |
62 |
67(61-72) |
40 |
57 |
34 |
5 |
16 |
0.79 |
Na, R (50) |
2013-2014 |
China |
660 |
136 |
66.95±8.89 |
28.0 |
127 |
259 |
9 |
265 |
0.87(0.83-0.90) |
Ferro, M (51) |
NA |
Italy |
300 |
108 |
65(50-73) |
31.6 |
97 |
115 |
11 |
77 |
0.77(0.72 - 0.83) |
Tan, L.G.L (52) |
2012-2014 |
Singapore |
157 |
30 |
65.4±6.46 |
26.75 |
27 |
53 |
3 |
74 |
0.7937(0.71–0.88) |
Vukovic, I (53) |
2012-2014 |
Serbia |
129 |
65 |
64±6.6 |
27.48 |
59 |
47 |
6 |
17 |
0.68(0.59-0.77) |
Ferro, M(54) |
2010 |
Italy |
151 |
48 |
64.5(48-87) |
38.7 |
41 |
40 |
7 |
63 |
0.77 |
Fuchsova, R (55) |
2010-2013 |
Czech |
263 |
113 |
66.5(50-83) |
37.0 |
102 |
71 |
11 |
79 |
0.79 |
Yu, G.P (56) |
2012-2013 |
China |
261 |
97 |
67(25-91) |
38.59 |
88 |
71 |
9 |
93 |
0.85 (0.79–0.91) |
Lughezzani, G (57) |
NA |
European |
883 |
365 |
64.5±7.7 |
NA |
328 |
418 |
37 |
100 |
0.68(0.64-0.72) |
Furuya , K (25) |
2012-2013 |
Japan |
50 |
33 |
68.5(53-82) |
38.7 |
21 |
4 |
12 |
13 |
0.79 (0.67–0.92) |
Scattoni ,V(58) |
2011-2012 |
Italy |
211 |
73 |
67.5±7.5 |
28.3 |
66 |
95 |
7 |
43 |
0.69(0.59–0.79) |
Mearini, L (59) |
2012 |
Italy |
275 |
86 |
65.4±6.8 |
37.1 |
79 |
116 |
7 |
73 |
0.76(0.71 – 0.81) |
Park, H (60) |
2015-2016 |
Korea |
246 |
125 |
69.6±8.7 |
22.9 |
112 |
38 |
13 |
83 |
0.76(0.69–0.84) |
Lazzeri, M (61) |
2010-2011 |
Italy |
222 |
71 |
63.9±7.1 |
28.8 |
64 |
113 |
7 |
38 |
0.67(0.61–0.73) |
Lughezzani, G (62) |
2010-2011 |
Italy |
729 |
280 |
64.3±7.8 |
NA |
252 |
328 |
28 |
121 |
0.70(0.66-0.73) |
TP: True Positive; FN: False Negative; FP: False Positive; TN: True Negative; AUC: Area Under the Curve; PHI: Prostate Health Index; PCa: Prostate Cancer
Table 1-3. Characteristics of studies included in the meta-analysis according to PCA3
First Author |
Year |
Country/Region |
Sample Size |
PCa number |
Patient Age (Yr) (mean/ median) |
Cutoff value |
TP |
FP |
FN |
TN |
AUC (95% CI) |
Haese, A(63) |
2008-2009 |
European |
463 |
128 |
64.4±6.6 |
35 |
60 |
94 |
68 |
241 |
0.66 |
Salami, S.S (64) |
NA |
USA |
45 |
15 |
64.5 |
NA |
14 |
19 |
1 |
11 |
0.65 (0.54–0.76) |
Marks, L.S (65) |
2004-2006 |
USA |
233 |
60 |
64(45-83) |
35 |
35 |
48 |
25 |
125 |
0.67 (0.60–0.76) |
Van Gils, M.P.M.Q (66) |
NA |
Netherland |
534 |
174 |
64.3±7.2 |
58 |
113 |
122 |
61 |
238 |
0.65 (0.58-0.72) |
Panebianco, V(67) |
2009-2010 |
Italy |
41 |
28 |
60.3(48-69) |
35 |
20 |
4 |
8 |
9 |
0.75 (0.60–0.87) |
Salagierski , M(68) |
2011 |
Poland |
80 |
24 |
66.2(50-81) |
35 |
18 |
25 |
6 |
31 |
0.72 |
Adam, A(69) |
2010 |
South African |
105 |
44 |
67(35-89) |
35 |
30 |
20 |
14 |
41 |
0.70 (0.60–0.81) |
Deras, I.L (70) |
NA |
North American |
507 |
206 |
64(32-89) |
35 |
111 |
78 |
95 |
223 |
0.70 |
Aubin, S.M.J (71) |
NA |
USA |
1072 |
190 |
NA |
35 |
92 |
189 |
98 |
693 |
0.69(0.65-0.74) |
Goode, R.R (72) |
NA |
New York |
456 |
88 |
66(41-90) |
35 |
64 |
76 |
31 |
285 |
0.77 |
Hessels, D (73) |
2003-2006 |
Netherland |
336 |
134 |
63(38-83) |
35 |
82 |
53 |
52 |
149 |
0.72 (0.66–0.77) |
De Luca, S (74) |
2011 |
Italy |
432 |
114 |
68(41-82) |
35 |
92 |
228 |
22 |
90 |
NA |
Gittelman, M.C (3) |
2013 |
USA |
466 |
102 |
NA |
25 |
79 |
156 |
23 |
208 |
0.74 |
Nyberg, M (75) |
2008 |
Sweden |
62 |
18 |
63 |
35 |
12 |
24 |
6 |
20 |
0.74 |
Ochiai, A(76) |
2009-2011 |
Japan |
633 |
264 |
67(42-89) |
35 |
176 |
105 |
88 |
264 |
0.74 |
Ramos, C.G (77) |
2009-2010 |
chile |
64 |
25 |
62.1(44-83) |
35 |
13 |
5 |
12 |
34 |
0.77 |
Pepe, P (78) |
2009-2011 |
Italy |
74 |
27 |
NA |
35 |
19 |
27 |
8 |
20 |
0.66 |
Wu, A.K (79) |
2012 |
USA |
103 |
37 |
63.5±7.4 |
35 |
14 |
15 |
23 |
51 |
0.64 (0.53–0.75) |
Wang, R (80) |
2006-2007 |
USA |
187 |
87 |
62(44-86) |
35 |
46 |
20 |
41 |
80 |
NA |
Busetto, G.M (81) |
2010-2012 |
Italy |
163 |
68 |
66.4±5.3 |
35 |
46 |
48 |
22 |
47 |
0.59 (0.51–0.66) |
De la Taille, A (82) |
2008-2009 |
European |
515 |
207 |
63±7.6 |
35 |
132 |
74 |
75 |
234 |
0.76 |
Stephan, C (43) |
2009-2012 |
Germany |
246 |
110 |
65(41-81) |
|
99 |
91 |
11 |
45 |
0.74 |
Ferro, M (54) |
2010 |
Italy |
151 |
48 |
64.5(48-87) |
32.5 |
39 |
44 |
9 |
59 |
0.71 |
Scattoni, V (58) |
2011-2012 |
Italy |
211 |
73 |
67±7.5 |
31.5 |
66 |
106 |
7 |
32 |
0.57 |
Porpiglia, F(5) |
2011-2013 |
Italy |
170 |
52 |
65(60-70) |
32.5 |
34 |
29 |
18 |
89 |
NA |
Seisen, T (41) |
2013 |
France |
138 |
39 |
63.4(44-83) |
35 |
24 |
41 |
15 |
58 |
0.55 |
TP: True Positive; FN; False Negative; FP; False Positive; TN: True Negative; AUC: Area Under the Curve; PCA3: Prostate Cancer Antigen3; PCa: Prostate Cancer
Study characteristics
The individual characteristics of the included studies are summarized in Table 1-1 to 1-3. The mean/median age, sample size, TP, TN, FN, FP, and AUC are reported in the tables. A total of 5931 patients (2656 PCa and 3275 non-PCa), 8491 patients (3307 PCa and 5184 non-PCa), and 7487 subjects (2362 PCa and 5125 non-PCa) were included in the pooled analyses for PI-RADS, PHI, and PCA3, respectively.
Quality assessment
All studies in each group (PI-RADS, PHI, and PCA3) were assessed using the QUADAS-2. The results of this assessment are shown in Figure 2. Overall, the quality of the studies was moderate.
Figure 2. The assessment of the methodological quality of the graph for all included studies according to diagnosis tests, A= PHI, B=PI-RADS, C= PCA3Diagnostic accuracy of tools for overall PCa
The combined sensitivity and specificity were 0.76 (95% CI, 0.71–0.81) and 0.84 (0.78–0.90) for PI-RADS, 0.48 (0.43–0.54) and 0.85 (0.80–0.89) for PHI, and 0.49 (0.44–0.54) and 0.79 (95% CI, 0.76–0.82) for PCA3, respectively (Figure 3). Also, diagnostic odds ratio was 17.57 (11.52-26.80) for PI-RADS, 3.70 (3.14-4.36) for PCA3 and 5.28 (4.03-6.93) for PHI. Figure 4 illustrates the hierarchic summary ROCs (HSROC) plot with 95% CI area and summary points of tools. The derived AUC from the HSROCs were 0.86 (0.83-0.89) for PI-RADS, 0.72 (0.68-0.76) for PCA3, and 0.70 (0.66-0.74) for PHI.
Figure 3. Forest plots of pooled sensitivities and specificities of PHI(A), PI-RADS(B), and PCA3(C) for the diagnosis of PCa
Figure 4. The SROC curve of diagnostic tests for PCa. A= PHI, B=PI-RADS, C= PCA3
We used Fagan's nomogram to find the posttest probability, for which we simulated a set with a prevalence of 26% for PCa based on the studies (16). Accordingly, in this model, the probability of someone having PCa and not being detected by the PHI tool was 18%. In the same vein for the PI-RADS, a negative result was associated with 9% of individuals with PCa. Eventually, for PCA3, it was 18% (Figure 5). On the other hand, the posttest probability of cancer patients with a positive test was 53%, 63%, and 45% for PHI, PI-RADS, and PCA3, respectively. All these suggest that among these tests, PI-RADS is more specific in the diagnosis of PCa.
Figure 5. Fagan diagram assessing the diagnostic value of tests for PCa. A= PHI, B=PI-RADS, C= PCA3
Publication bias
The asymmetry of Deek’s plot was used to detect possible publication bias. The results revealed no publication bias for PCA3 and PI-RADS (Figure 6).
Figure 6. Linear regression test of funnel plot asymmetry for PCA3 (A), PI-RADS (B), and PHI(C)
Discussion
In this meta-analysis, we assessed the diagnosis of PCa by three tools, PI-RADS, PHI, and PCA3. The results of this study showed that these techniques have acceptable validity indices to detect PCa.
The diagnostic value of common tests in PCa detection is still challenging. Despite the widespread use of PSA as a biomarker for unnecessary biopsies and detection of PCa, its use is far from ideal due to its low specificity (83). Therefore, a US preventive services task force recommends other biomarkers and tools with high sensitivity and specificity for diagnosis (84). Meanwhile, despite antibiotic prophylaxis, a biopsy may be related to a few problems, such as bleeding, urinary retentions, or infections (85, 86). Therefore, it is necessary to evaluate the diagnostic value of tools for reducing useless biopsies.
Although studies showed many tests to diagnose PCa, three tests with high sensitivity and accuracy are used for the early detection of prostate cancer. Recently, multi-parametric MRI, which includes anatomical T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), dynamic contrast-enhanced (DCE) imaging, and magnetic resonance spectroscopy (MRS), is becoming increasingly popular for detection, staging, and treatment planning of PCa (87). To decrease variability and establish wide acceptance and implementation in practice, the European Society of Urogenital Radiology (ESUR) published a guideline known as PI-RADS (88). In 2015, an updated version (PI-RADS V2) was developed that improved the sensitivity and specificity of V1 for assessing PCa (89). Our meta-analysis only included studies employing PI-RADS V2 for detecting overall PCa.
Other methods for diagnosing PCa are serum markers. It has been proposed that measurement of the precursor PSA isoform and its derivatives might improve the detection of PCa. The PHI is a comprehensive test that includes serum p2PSA, free PSA, and total PSA, and can increase the diagnostic accuracy of tPSA in detecting PCa (90). Some studies demonstrate that PHI may be better than tPSA alone in initial or repeat setting (91, 92). Another diagnostic tool is a urine marker called PCA3, a non-coding RNA with an over-expressed PCa cell (93, 94).
A few studies have found that PCA3 is valuable for PCa screening and decreasing negative biopsies (82, 95).
In 2013, a systematic review showed that PHI's pooled sensitivity and specificity were 90% and 31.6%, respectively. The results of this study indicated that the accuracy of PCa detection improved using PHI (96). Another meta-analysis study suggested that the urine PCA3 has acceptable sensitivity (62%), specificity (75%), and a moderate level of accuracy (AUC=0.75) in PCa diagnosis (97). The finding of a meta-analysis revealed that PI-RADS version 2 has (9) good precision in PCa with great sensitivity and modest specificity (9).
Our results showed approximately high specificity of three tests, PI-RADS, PHI, and PCA3, ranging between 0.79 and 0.85. Few studies assess the prognostic performance of these three tests in patients with PCa. In a cohort study, PI-RADS resulted in the highest value in the accuracy for predicting PCa compared with PCA3 and PHI (AUC=0.78) (98). Another study showed that PI-RADS has a high diagnostic value in identifying PCa compared with PCA3 and PHI (AUC=0.936) (5).
To the best of our knowledge, our study is the first systematic review and meta-analysis to evaluate and compare the performance of PI-RADS, PHI, and PCA3 with each other. Our results suggested that PI-RADS V2 yielded the highest AUC value (0.86) and that this model is superior to the other models in terms of performance, including PCA3 or PHI (DOR= 17.57 for PI-RADS, 3.70 for PCA3, and 5.28 for PHI). Indeed, the high variations of values for sensitivity and specificity have been revealed in the DOR, with greater values indicating a better discriminatory diagnostic test (99). Also, the post-test probability of PI-RADS was higher than that of PCA3 or PHI, indicating a relatively good clinical value of the PI-RADS test. On the other hand, if the patient test is negative, the post-test probability of having PCa would be 9%, and if the patient test is positive, the post-test probability of having PCa would be 63% for PI-RADS.
It is important to note that although MRI requires expert interpretation, has high inter-observer variability, and is expensive; it offers spatial information on tumors.
Our meta-analysis had some limitations which should be taken into account. The results of this study showed heterogeneity compromising the overall accuracy. Although there is one gold standard (biopsy), and studies were included only based on the number of patients in all studies, lack of blinding, not selecting patients with the same criteria across all studies, and different cut-off points for tools caused heterogeneity. Also, we expanded our searches in several databases to avoid publication bias. Another limitation was a potential publication bias, as non-English studies were excluded. The small number of trials and the marked differences between the methodology of tests may have yielded moderate quality.
Nevertheless, our study provided the most up-to-date evidence on the important tests of PCa diagnosis. However, large multicenter randomized trials or cohort studies with similar methodologies should be performed to evaluate the diagnostic value of PI-RADS, PHI, and PCA3 in predicting PCa in the future.
Conclusions
Based on the results of this review, the clinical application of these non-invasive methods of early detection of PCa would reduce unnecessary biopsies. Available evidence suggests that the PI-RADS is superior in PCa diagnosis with high sensitivity, specificity, and AUC compared to PHI and PCA3.
Authors’ Contribution
All authors contributed equally.
Acknowledgments
Special thanks to the Urology Research Center (URC), Tehran University of Medical Sciences (TUMS).
Conflict of interest
All authors declare that there is not any kind of conflict of interest.
Funding
There is no funding.
Ethical Statement
Not applicable.
Data availability
Data will be provided by the corresponding author on request.
Abbreviation
AUC Area under curves
DCE Dynamic contrast-enhanced
DOR Diagnostic odds ratio
DWI Diffusion-weighted imaging
ESUR European Society of Urogenital Radiology
FN False negative
FP False positive
HSROCs Hierarchic summary ROCs
MRS Magnetic resonance spectroscopy
PCa Prostate cancer
PCA3 Prostate cancer antigen 3
PHI Prostate health index
PI-RADS Prostate imaging reporting & data system
PSA Prostate-specific antigen
QUADAS Quality assessment of diagnostic accuracy studies
SROC Summary receiver operation characteristics
TN True negative
TP True positive
T2WI T2-weighted imaging