Prostate Cancer Screening with PSA and MRI Followed by Targeted Biopsy Only

Abstract

BACKGROUND

Screening for prostate cancer is burdened by a high rate of overdiagnosis. The most appropriate algorithm for population-based screening is unknown.

METHODS

We invited 37,887 men who were 50 to 60 years of age to undergo regular prostate-specific antigen (PSA) screening. Participants with a PSA level of 3 ng per milliliter or higher underwent magnetic resonance imaging (MRI) of the prostate; one third of the participants were randomly assigned to a reference group that underwent systematic biopsy as well as targeted biopsy of suspicious lesions shown on MRI. The remaining participants were assigned to the experimental group and underwent MRI-targeted biopsy only. The primary outcome was clinically insignificant prostate cancer, defined as a Gleason score of 3+3. The secondary outcome was clinically significant prostate cancer, defined as a Gleason score of at least 3+4. Safety was also assessed.

RESULTS

Of the men who were invited to undergo screening, 17,980 (47%) participated in the trial. A total of 66 of the 11,986 participants in the experimental group (0.6%) received a diagnosis of clinically insignificant prostate cancer, as compared with 72 of 5994 participants (1.2%) in the reference group, a difference of −0.7 percentage points (95% confidence interval [CI], −1.0 to −0.4; relative risk, 0.46; 95% CI, 0.33 to 0.64; P<0.001). The relative risk of clinically significant prostate cancer in the experimental group as compared with the reference group was 0.81 (95% CI, 0.60 to 1.1). Clinically significant cancer that was detected only by systematic biopsy was diagnosed in 10 participants in the reference group; all cases were of intermediate risk and involved mainly low-volume disease that was managed with active surveillance. Serious adverse events were rare (<0.1%) in the two groups.

CONCLUSIONS

The avoidance of systematic biopsy in favor of MRI-directed targeted biopsy for screening and early detection in persons with elevated PSA levels reduced the risk of overdiagnosis by half at the cost of delaying detection of intermediate-risk tumors in a small proportion of patients. (Funded by Karin and Christer Johansson’s Foundation and others; GÖTEBORG-2 ISRCTN Registry number, ISRCTN94604465.)

Appropriate screening for prostate cancer remains a controversial issue. The European Randomized Study of Screening for Prostate Cancer (ERPSC) and the randomized GÖTEBORG-1 trial, initiated in the 1990s, were based on testing for prostate-specific antigen (PSA) followed by systematic prostate biopsy guided by transrectal ultrasonography. Both trials showed a significant reduction in prostate cancer mortality; however, a high risk of overdiagnosis was also noted.^1–4 The Prostate Lung Colorectal and Ovarian (PLCO) trial in the United States and the Cluster Randomized Trial of PSA Testing for Prostate Cancer (CAP) in the United Kingdom were conducted on the basis of PSA testing and systematic biopsy. Neither trial showed an effect on prostate cancer mortality, yet screening was associated with overdiagnosis in both trials.^5,6 Furthermore, although the randomized Scandinavian Prostate Cancer Group Number 4 (SPCG-4) trial (which included 18 years of follow-up) documented greater treatment efficacy of radical prostatectomy than with observation among patients with early, localized prostate cancer,⁷ such improved efficacy was not shown in the Prostate Cancer Intervention Versus Observation Trial (PIVOT), with 22 years of follow-up,⁸ and the Prostate Testing for Cancer and Treatment (ProtecT) trial, with 10 years of follow-up.⁹

However, the high rate of overdiagnosis is regarded as the main obstacle to the recommendation of population-based screening for prostate cancer.^10–12 The cause of overdiagnosis is the high prevalence of small, low-grade prostate cancers in the adult population; approximately 50% of men older than 60 years of age have such tumors.^13,14 These tumors are often indolent in nature and show slow or no progression. In addition, the PSA test has low specificity. The positive predictive value of a PSA level of 3 ng per milliliter or greater as an indicator of a Gleason score of 3+4 or greater is estimated to be 16%.⁴ (The Gleason score for each patient was based on the Gleason score from each biopsy sample, which was defined as the sum of the most common Gleason grade plus the highest remaining Gleason grade; scores range from 6 [lowest grade of cancer] to 10 [highest].) Thus, in a screening program that involves systematic biopsy, a majority of patients without clinically significant cancer will undergo systematic biopsy that will accidentally detect clinically insignificant cancers.⁴

Targeted biopsy of suspected lesions that are shown on magnetic resonance imaging (MRI) has been suggested as a means of reducing overdiagnosis of prostate cancer; this approach has been shown to be noninferior to systematic biopsy in patients with elevated PSA.^15–17 However, there is not a worldwide consensus as to whether systematic biopsy can be omitted.^18,19

We designed the randomized GÖTEBORG-2 trial to evaluate whether a screening algorithm that included PSA testing followed by targeted biopsy alone in patients with positive MRI results would result in less overdiagnosis than screening according to current recommendations, in which all persons with elevated PSA levels are referred for systematic biopsy irrespective of the outcome of MRI.

METHODS

TRIAL DESIGN

GÖTEBORG-2, which has been described previously,²⁰ is a large, population-based, randomized screening trial to evaluate several different research questions with regard to prostate cancer screening and diagnosis (see the protocol and Section S2 of the Supplementary Appendix, which are both available with the full text of this article at NEJM.org). The main objective of the trial was to evaluate whether omitting systematic biopsy for all men with an elevated PSA level (≥3 ng per milliliter) and performing targeted biopsy of only MRI-positive lesions reduced the risk of detecting clinically insignificant prostate cancer (defined as a Gleason score of 3+3) while still detecting clinically significant cancer.

Data were prospectively recorded in a central database hosted at the Unversity of Gothenburg. Data were analyzed and reported only according to group levels, and no data can be linked to any individual participant.

PARTICIPANTS AND RANDOMIZATION

We identified men in the Swedish Population Register who were 50 to 60 years of age and were living in Gothenburg, Sweden, or its 10 surrounding municipalities in the 2015–2020 period. Of the men who met the inclusion criteria, 38,775 were randomly assigned to a screening group and 19,450 to a control group that did not receive screening (and is not included in this report). Invitations to the trial began on September 1, 2015, and randomization ended in January 2020 (Fig. S1). Persons who had prevalent prostate cancer or who had emigrated or died before their randomization date were excluded.

Men in the screening group received an invitation letter with information about the trial. Those who agreed to participate received a PSA test at one of several primary care facilities. Participants were sequentially assigned by an external password-protected computer system, in a 1:1:1 ratio, to one of three trial groups. Only men who participated in PSA testing were assigned to a trial group.

OVERSIGHT

In January 2015, the trial was approved by the regional ethics committee at the University of Gothenburg, which permitted prospective enrollment of men without the provision of written informed consent. All the enrollees received written information that participation was voluntary and that they could opt out at any time by contacting the trial administrators.

AUTHOR CONTRIBUTIONS

The first and second authors had full access to all trial data and vouch for the completeness and accuracy of the data and for the fidelity of the trial to the protocol. All the authors were involved in data interpretation. The first, second, and last authors wrote the first draft of the manuscript, and all the authors made the decision to submit the manuscript for publication. All the authors contributed to writing the version of the manuscript submitted for publication.

INTERVENTIONS

Reference Group

The reference group (5994 participants) consisted of men assigned to trial group 1. Men in this group who had a PSA level of 3 ng per milliliter or higher underwent further evaluation with MRI of the prostate followed by systematic biopsy regardless of MRI results. Targeted biopsy was added if MRI revealed suspicious lesions, defined as a score of 3 to 5 as measured by the Prostate Imaging Reporting and Data System (PI-RADS), version 2 (scores range from 1 to 5, with higher scores indicating more clinically suspicious lesions).²¹

Experimental Group

The experimental group (11,986 participants) consisted of men assigned to trial groups 2 and 3. Participants in trial group 2 who had a PSA level of 3 ng per milliliter or higher underwent further evaluation with MRI of the prostate; in participants in whom suspicious lesions were found, MRI evaluation was followed by targeted biopsy alone. Systematic biopsy with or without targeted biopsy was performed only in participants whose PSA level was 10 ng per milliliter or higher, irrespective of the MRI result.

The screening procedure for trial group 3 was identical to that of trial group 2, except that the PSA cutoff for MRI evaluation was lower — 1.8 ng per milliliter or higher. In this report, the analysis includes MRI and biopsy results only from participants with a PSA level of 3 ng per milliliter or higher.

MRI SCREENING

Multiparametric MRI that was compatible with PI-RADS²¹ was performed at Sahlgrenska University Hospital, with the use of a 3-tesla scanner (Philips Medical Systems) with a pelvic phased-array coil (Section S3 and Table S1). All images were read in a blinded fashion with consensus by two of three authors who are specialist radiologists with more than 5 years of experience with prostate MRI. A positive MRI lesion was defined as one with a PI-RADS score of 3 to 5. Each lesion was assigned a location according to a 24-sector template.²² External validation of MRI readings was performed in a sample of 100 examinations (Section S4 and Tables S2A and S2B).

CLINICAL EVALUATION

Clinical evaluation was performed by one of five experienced urologists and included digital rectal examination for clinical tumor staging and transrectal ultrasound examination followed by prostate biopsy guided by transrectal ultrasonography, unless contraindicated. Assessments of clinical stage and transrectal ultrasound examinations were performed by investigators who were unaware of group assignment or PSA and MRI results. After group assignment and screening results were disclosed to investigators, biopsy was performed in accordance with the protocol with regard to the participant’s study group. Systematic transrectal ultrasound-guided biopsies (10 to 12 cores) were obtained from the peripheral zone, and 4 cores were obtained from MRI-positive lesions in biopsies directed by a cognitive MRI–transrectal-ultrasound fusion technique.

Participants with low-grade prostate cancer (mainly with Gleason 3+3 cancer but also some with Gleason 3+4 cancer) that was detected by targeted biopsy in the experimental group were invited to undergo follow-up systematic biopsy. The Gleason score was thus based on both targeted and systematic biopsies after a cancer diagnosis in both groups in order to avoid sampling bias due to different primary biopsy techniques.^15,23 Treatment decisions were made according to routine clinical practice and after shared decision making with the patient.

LABORATORY AND PATHOLOGICAL ANALYSES

All samples were analyzed for total PSA (Advia XPT, Siemens Health Care Diagnostics) at a central laboratory. All prostate specimens (biopsies and radical prostatectomy specimens) were primarily assessed by one experienced pathologist. The pathology report of biopsies for each core included the Gleason score, the extent of tumor growth in millimeters, and the semiquantitative percentage of Gleason grades 4 and 5. Prostate biopsy specimens that revealed the presence of cancer were also reviewed independently by two external, experienced uropathologists who were unaware of group assignment and clinical data.

The final Gleason score for each biopsy was assessed from the verdicts of the three pathologists. If at least two pathologists agreed, the Gleason score was set. If the verdicts of all three pathologists were discordant, the middle score was chosen. The pathology review process was invalid or inconclusive in 16 cases for which the verdict from the local pathologist was used to assign the Gleason score (Section S5).

ADVERSE EVENTS

Adverse events in participants who had undergone prostate biopsy were recorded. Participants attended a follow-up appointment with the urologist 3 to 4 weeks after the biopsy to receive the pathological findings. Shortly before this consultation, participants filled out an electronic questionnaire about their experience with the screening procedures and about the adverse events they had had.

TRIAL OUTCOMES

The primary outcome was clinically insignificant prostate cancer, defined as a Gleason score of 3+3. The secondary outcome was clinically significant prostate cancer, defined as a Gleason score of 3+4 or higher. Other established and alternative definitions of clinically insignificant and significant cancers were also evaluated (Table S4).

STATISTICAL ANALYSIS

We determined that a sample size of 36,000 men would provide the trial with 80% power to show the superiority of MRI-targeted biopsy alone as compared with systematic biopsy with respect to the reduction in clinically insignificant prostate cancer, with the use of a two-sided test at a significance level of 0.05 for the primary outcome. These calculations were based on the assumption that the strategy of using MRI-targeted biopsy alone would reduce the overdiagnosis of clinically insignificant cancer by 50% (Section S7).

We performed analyses on the results of the first screening round of the GÖTEBORG-2 trial (Fig. 1). The main analysis, performed according to the intention-to-treat principle, examined data from all the participants in the two groups, regardless of whether they had adhered to the protocol. Two variants of per-protocol analyses were performed as sensitivity analyses. For purposes of the intention-to-treat analysis, participants who had an indication for biopsy but did not undergo biopsy were considered not to have cancer. To investigate the importance of these missing values, we performed another sensitivity analysis based on multiple imputation of the primary outcome (Section S8).

Figure 1. Randomization and Enrollment.

Men 50 to 60 years of age identified from a population registry were invited to screening. Those who agreed to undergo screening and prostate-specific antigen (PSA) testing were randomly assigned to the reference group or to the experimental group. Men in both groups who had a PSA level of at least 3 ng per milliliter were offered multiparametric magnetic resonance image (MRI) screening. Participants in the reference group underwent systematic prostate biopsies (10 to 12 cores) irrespective of their MRI results and, possibly, targeted biopsies (4 cores) in case of positive MRI findings (Prostate Imaging Reporting and Data System [PI-RADS] score 3 to 5; scores range from 1 to 5, with higher scores indicating more clinically suspicious lesions). Men in the experimental group underwent only targeted prostate biopsy if the MRI showed suspicious lesions (PI-RADS score, 3 to 5) unless the PSA level was 10 ng per milliliter or higher.

The per-protocol analysis was restricted to participants who had strictly adhered to the diagnostic pathway specified in the protocol. The modified per-protocol analysis included data from participants who had strictly adhered to the diagnostic pathway, but excluded results from possible systematic biopsies in the experimental group if those biopsies had been performed during the diagnostic biopsy phase of the trial and the targeted biopsies were benign.

We evaluated the primary and secondary outcomes by comparing the relative risk and difference in cancer detection between the experimental group and the reference group and calculated two-sided 95% confidence intervals. The biopsy results in the reference group were evaluated in two ways: one that included all cancers that had been detected (results from MRI-targeted and systematic biopsy) and one that included cancers that were detected when targeted biopsy was disregarded (results from systematic biopsy only). The score method was used for testing and calculating confidence intervals.²⁴ The widths of the confidence intervals were not adjusted for multiplicity and should not be used in place of hypothesis testing. Significance testing was performed only for the main definition of clinically insignificant cancer (Gleason score, 3+3). A P value of less than 0.05 was considered to indicate statistical significance. Statistical analyses were conducted with the use of R statistical software, version 4.0.3 (R Foundation).

RESULTS

TRIAL PARTICIPANTS

Of the total 38,775 men who were randomly assigned to the screening group, 37,887 were invited to participate in the first round of the trial; 17,980 (47%) underwent PSA testing (Fig. 1). Pretrial PSA testing was common (reported in 32% of the participants), but the percentage who had undergone a pretrial prostate biopsy was very low (1%) in the two groups (Table 1).

Table 1.

Characteristics of the Participants at Randomization.^*

Characteristic	Reference Group (N = 5994)	Experimental Group (N = 11,986)
Median age (IQR) — yr	56 (52–59)	56 (52–59)
Family history of prostate cancer — no. (%)
Yes	630 (11)	1264 (11)
No	2754 (46)	5554 (46)
Missing data	2610 (44)	5168 (43)
Pretrial PSA test — no. (%)
Yes	1948 (32)	3878 (32)
No	2513 (42)	5012 (42)
Missing data	1533 (26)	3096 (26)
Pretrial prostate biopsy — no. (%)
Yes	75 (1)	170 (1)
No	4427 (74)	8806 (73)
Missing data	1492 (25)	3010 (25)
Median PSA level (IQR) — ng/ml	0.8 (0.5–1.4)	0.8 (0.5–1.4)
PSA level — no. (%)
<3 ng/ml	5589 (93)	11,190 (93)
3 to 9.9 ng/ml	389 (6)	737 (6)
≥10 ng/ml	16 (<1)	59 (<1)

^*Pretrial prostate-specific antigen (PSA) levels and biopsy information were obtained from electronic questionnaires sent to the men who had been invited to participate in the trial. IQR denotes interquartile range.

Approximately 7% of the participants in the two groups had an elevated PSA level (≥3 ng per milliliter); less than 1% had a PSA level of 10 ng per milliliter or higher. Of the participants who had an elevated PSA, 95% underwent MRI. Of those who had an indication for biopsy, 348 of 405 participants (86%) in the reference group and 300 of 335 (90%) in the experimental group underwent biopsy.

OUTCOMES

A total of 66 of the 11,986 participants in the experimental group (0.6%) received a diagnosis of clinically insignificant prostate cancer, as compared with 72 of 5994 participants (1.2%) in the reference group, a difference of −0.7 percentage points (95% confidence interval [CI], −1.0 to −0.4; relative risk, 0.46; 95% CI, 0.33 to 0.64; P<0.001). When we analyzed the results of systematic biopsy alone in the reference group, the corresponding relative risk was 0.45 (95% CI, 0.32 to 0.63). Clinically significant cancer was detected in 110 participants (0.9%) in the experimental group, as compared with 68 (1.1%) in the reference group (relative risk, 0.81; 95% CI, 0.60 to 1.10). Results of analyses and outcomes are shown in Table 2. Results of analyses of clinically insignificant and significant cancers according to alternative definitions of outcomes are included in Table S4.

Table 2.

Prostate Cancer Detection Analyses and Outcomes.

Analysis and Outcome	Reference Group (N = 5994)*		Experimental Group (N = 11,986)	Experimental vs. Reference Group, MRI-Targeted and Systematic Biopsy		Experimental vs. Reference Group, Systematic Biopsy
	MRI-Targeted and Systematic Biopsy	Systematic Biopsy	MRI-Targeted Biopsy	Relative Risk (95% CI)	Between-Group Difference (95% CI)	Relative Risk (95% CI)	Between-Group Difference (95% CI)
					percentage points		percentage points
Primary analysis †
PSA ≥3 ng/ml — no. (%)	405 (6.8)	405 (6.8)	796 (6.6)
Biopsy indication — no. (%)	405 (6.8)	405 (6.8)	335 (2.8)
Biopsy performed and PSA ≥3 ng/ml — no. (%)	348 (5.8)	347 (5.8)	301 (2.5)‡
Any prostate cancer— no. (%)	140 (2.3)	131 (2.2)	176 (1.5)	0.63 (0.50 to 0.78)	−0.9 (−1.3 to−0.4)	0.67 (0.54 to 0.84)	−0.7 (−1.2 to−0.3)
Clinically insignificant cancer— no. (%)	72 (1.2)	73 (1.2)	66 (0.6)	0.46 (0.33 to 0.64)§	−0.7 (−1.0 to−0.4)	0.45 (0.32 to 0.63)	−0.7 (−1.0 to−0.4)
Clinically significant cancer— no. (%)	68 (1.1)	58 (1.0)	110 (0.9)	0.81 (0.60 to 1.10)	−0.2 (−0.6 to 0.1)	0.95 (0.69 to 1.31)	−0.1 (−0.4 to 0.2)
Clinically insignificant cancer, with correction for missing biopsy data — no. (96)¶	80 (1.3)	81 (1.4)	75 (0.6)	0.47 (0.33 to 0.65)	−0.7 (−1.0 to−0.4)	0.46 (0.33 to 0.64)	−0.7 (−1.0 to−0.4)
Per-protocol analysis — no./total no. (%)†
PSA ≥3 ng/ml	336/5925 (5.7)	347/5936 (5.8)	734/11,924 (6.2)
Biopsy performed and PSA ≥3 ng/ml	336/5925 (5.7)	347/5936 (5.8)	274/11,924 (2.3)
Any prostate cancer	135/5925 (2.3)	131/5936 (2.2)	159/11,924 (1.3)	0.59 (0.47 to 0.74)	−0.9 (−1.4 to−0.5)	0.60 (0.48 to 0.76)	−0.9 (−1.3 to −0.5)
Clinically insignificant cancer	68/5925 (1.1)	73/5936 (1.2)	59/11,924 (0.5)	0.43 (0.30 to 0.61)	−0.7 (−1.0 to−0.4)	0.40 (0.29 to 0.57)	−0.7 (−1.1 to−0.4)
Clinically significant cancer	67/5925 (1.1)	58/5936 (1.0)	100/11,924 (0.8)	0.74 (0.55 to 1.01)	−0.3 (−0.6 to 0.0)	0.86 (0.62 to 1.19)	−0.1 (−0.5 to 0.2)
Modified per-protocol analysis — no./total no. (%)†
PSA ≥3 ng/ml	336/5925 (5.7)	347/5936 (5.8)	738/11,928 (6.2)
Biopsy performed and PSA ≥3 ng/ml	336/5925 (5.7)	347/5936 (5.8)	278/11,928 (2.3)
Any prostate cancer	135/5925 (2.3)	131/5936 (2.2)	156/11,928 (1.3)	0.57 (0.46 to 0.72)	−1.0 (−1.4 to−0.6)	0.59 (0.47 to 0.75)	−0.9 (−1.3 to −0.5)
Clinically insignificant cancer	68/5925 (1.1)	73/5936 (1.2)	57/11,928 (0.5)	0.42 (0.29 to 0.59)	−0.7 (−1.0 to−0.4)	0.39 (0.27 to 0.55)	−0.8 (−1.1 to−0.5)
Clinically significant cancer	67/5925 (1.1)	58/5936 (1.0)	99/11,928 (0.8)	0.73 (0.54 to 1.00)	−0.3 (−0.6 to 0.0)	0.85 (0.62 to 1.18)	−0.1 (−0.5 to 0.1)

^*Findings in the reference group were analyzed according to all cancers detected (results of both MRI-targeted and systematic biopsies) and cancers detected when targeted biopsies were disregarded (evaluation of results of systematic biopsies alone). Analyses of cancer detection in a third subgroup of the reference group (results of MRI-targeted biopsies only) are shown in Table S7.

^†The primary analysis, performed according to the intention-to-treat principle, included data from participants who had undergone randomization and been assigned to either group, regardless of whether they adhered to the protocol. The per-protocol analysis was restricted to participants who had strictly adhered to the diagnostic pathway specified in the protocol. The modified per-protocol analysis included data from participants who had strictly adhered to the diagnostic pathway, but it excluded results from possible systematic biopsies in the experimental group if they were performed during the diagnostic biopsy phase of the trial and if results of the targeted biopsies were benign.

^‡One participant underwent a biopsy but did not have an indication for biopsy as described in the protocol.

^§P<0.001.

^¶The median number of insignificant cancers was calculated for 100 imputed data sets.

The outcome of prostate biopsy according to study group and PI-RADS score is shown in Table 3. Primary treatment approaches are shown in Table 4.

Table 3.

MRI Result and Detection of Cancer.^*

Group and Result	No Biopsy	Benign Biopsy	Gleason Score 3+3	Gleason Score 3+4	Gleason Score 4+3	Gleason Score >4+3
	number/total number (percent)
Reference group
MRI not performed	14/21 (67)	5/21 (24)	1/21 (5)	0	0	1/21 (5)
MRI incomplete or not assessable	0	2/2 (100)	0	0	0	0
PI-RADS 1–2	37/240 (15)	154/240 (64)	40/240 (17)	9/240 (4)	0	0
PI-RADS 3	4/41 (10)	24/41 (59)	7/41 (17)	5/41 (12)	1/41 (2)	0
PI-RADS 4	2/85 (2)	23/85 (27)	23/85 (27)	25/85 (29)	4/85 (5)	8/85 (9)
PI-RADS 5	0	0	1/16 (6)	10/16 (62)	2/16 (12)	3/16 (19)
Total participants	57/405 (14)	208/405 (51)	72/405 (18)	49/405 (12)	7/405 (2)	12/405 (3)
Experimental group
MRI not performed	20/34 (59)	7/34 (21)	4/34(12)	2/34 (6)	0	1/34 (3)
MRI incomplete or not assessable	2/7 (29)	1/7 (14)	3/7 (43)	1/7 (14)	0	0
PI-RADS 1–2	467/488 (96)	19/488 (4)	0	1/488 (<1)	1/488 (<1)	0
PI-RADS 3	3/65 (5)	38/65 (58)	15/65 (23)	8/65 (12)	1/65 (2)	0
PI-RADS 4	2/150 (1)	56/150 (37)	38/150 (25)	43/150 (29)	6/150 (4)	5/150 (3)
PI-RADS 5	1/52 (2)	4/52 (8)	6/52 (12)	22/52 (42)	6/52 (12)	13/52 (25)
Total participants	495/796 (62)	125/796 (16)	66/796 (8)	77/796 (10)	14/796 (2)	19/796 (2)

^*The Gleason score for each patient was based on the Gleason score from each biopsy sample, which was defined as the sum of the most common Gleason grade plus the highest remaining Gleason grade; scores range from 6 (lowest grade of cancer) to 10 (highest). Prostate Imaging Reporting and Data System (PI-RADS) scores range from 1 to 5, with higher scores indicating more clinically suspicious lesions.

Table 4.

Treatment According to Risk Group.^*

Group and Treatment	Very Low Risk	Low Risk	Intermediate Risk	High Risk	Advanced	Total
	number/total number (percent)
Reference group
Any prostate cancer	30/5994 (0.5)	40/5994 (0.7)	60/5994 (1.0)	10/5994 (0.2)	0	140/5994 (2.3)
Active surveillance	30/30 (100.0)	30/40 (75.0)	13/60 (21.7)	0	—	73/5994 (1.2)
Radical prostatectomy	0	10/40 (25.0)	35/60 (58.3)	6/10 (60.0)	—	51/5994 (0.9)
Radiation therapy	0	0	11/60 (18.3)	4/10 (40.0)	—	15/5994 (0.3)
Other treatment	0	0	1/60 (1.7)	0	—	1/5994 (<0.1)
Unknown	0	0	0	0	—	0
Experimental group
Any prostate cancer	19/11,986 (0.2)	43/11,986 (0.4)	100/11,986 (0.8)	12/11,986 (0.1)	2/11,986 (<0.1)	176/11,986 (1.5)
Active surveillance	16/19 (84.2)	28/43 (65.1)	14/100 (14.0)	0	0	58/11,986 (0.5)
Radical prostatectomy	3/19 (15.8)	14/43 (32.6)	68/100 (68.0)	9/12 (75.0)	0	94/11,986 (0.8)
Radiation therapy	0	0	14/100 (14.0)	3/12 (25.0)	0	17/11,986 (0.1)
Hormonal treatment	0	0	1/100 (1.0)	0	2/2 (100.0)	3/11,986 (<0.1)
Other treatment	0	1/43 (2.3)	2/100 (2.0)	0	0	3/11,986 (<0.1)
Unknown	0	0	1/100 (1.0)	0	0	1/11,986 (<0.1)

^*Risk groups were defined with the use of a modified National Comprehensive Cancer Network classification. Very low risk was defined as having stage T1c, Gleason score of 3+3, maximum two sectors cancer, PSA level less than 10 ng per milliliter, PSA density of less than 0.15 ng per milliliter per gram, and lesions that are not N1 and not M1. Low risk was defined as having cancer staged at T1c to T2a or b, Gleason score of 3+3, PSA level of less than 10 ng per milliliter, and lesions that are not N1 and not M1 but do not belong in the very-low-risk group. Intermediate risk was defined as having cancer staged at T2c or a Gleason score of 7 or less or a PSA level of 10 to 20 ng per milliliter, and lesions that are not N1 and not M1 but do not belong in the high-risk group. High risk was defined as having cancer staged at T3 to T4 or a Gleason score greater than 7 or a PSA level greater than 20 ng per milliliter or more than 4 sectors with lesions of Gleason score greater than 7, and lesions that are not N1 or M1. Advanced risk was defined as N1 or M1 lesions or a PSA level greater than 100 ng per milliliter.

Clinically significant cancer that was detected by systematic biopsy alone was diagnosed in 10 participants in the reference group; 9 of the participants had negative MRIs and 1 had false positive findings on MRI. Gleason 3+4 disease was diagnosed in all 10 participants; tumors of less than 5% Gleason pattern 4 were diagnosed in 6 participants, clinical stage T1c disease was diagnosed in 7 participants, and stage T2a or T2b disease was diagnosed in 3. The disease was managed primarily with active surveillance in 6 participants and was treated with radical prostatectomy in 3 and radiation therapy in 1.

In the experimental group, cancer was diagnosed in 128 participants who had a PSA level of less than 10 ng per milliliter by means of targeted diagnostic biopsies only. Of these 128 men, 72 had a Gleason score of 3+3, 62 of whom (86%) underwent subsequent confirmatory systematic biopsy. Of those 62 participants, Gleason scores were upgraded in 16 (26%) — 15 to Gleason 3+4 and 1 to Gleason 3+5 (Table S5A). Gleason 3+3 lesions that had been detected by systematic biopsy differed only in tumor extension from those that had been detected by targeted biopsy of suspicious lesions shown on MRI, with greater volume measured in tumors that were visible on MRI (Table S5B).

SAFETY

Mild adverse events (most commonly hematuria and hematospermia) were common in the two groups. Seven participants (0.1%) in the reference group received prescriptions for antibiotic agents for outpatient treatment of urinary tract infections within 30 days after biopsy, as compared with three participants (0.03%) in the experimental group. A total of five participants were hospitalized within 30 days after biopsy — four (0.07%) in the reference [66, and one for acute hypertension) and one (0.008%) in the experimental group (for urosepsis). No deaths were reported within 30 days after biopsy (Table S6). Protocol deviations are shown in Table S8.

DISCUSSION

Pathways for the diagnostic evaluation of an elevated PSA level vary among guidelines. For example, the American Association of Urology advises that systematic biopsy be performed in conjunction with MRI-targeted sampling,²⁵ and the European Association of Urology recommends systematic biopsies in all men who have positive MRI results if they have not undergone a biopsy and in men who have negative MRI results if the clinical suspicion for cancer is high.¹⁹

Our trial showed that changing the diagnostic algorithm to include prebiopsy MRI in all participants with elevated PSA levels and only targeted biopsy of lesions with a PI-RADS score of 3 to 5, and no biopsy of lesions with PI-RADS scores of 1 to 2 reduced the risk of detecting Gleason 3+3 cancers by half. Since the frequent detection of small Gleason 3+3 cancers after PSA screening is regarded as a major contributor to the high incidence of potentially harmful overdiagnosis of prostate cancer, this finding is of importance. This strategy also considerably reduced the percentage of biopsies in participants in the experimental group who had elevated PSA levels. Although the incidence of severe adverse events after prostate biopsy in our trial was low — probably owing to the relatively young age of the participants — prostate biopsy is an unpleasant and potentially risky procedure.²⁶

Almost all evaluations of prostate cancer screening programs have concluded that no evidence suggests a net benefit of screening because of the high rate of overdiagnosis.^10–12 A major ethical dilemma is that prostate cancer screening leads to cancer diagnoses in many patients, a large proportion of whom undergo radical treatment and risk permanent side effects in the treatment of a disease that may never have progressed during their lifetime.²⁷ Nevertheless, prostate cancer constitutes a major health problem worldwide, and regular PSA screening can decrease the risk of death from prostate cancer.^1,3 Furthermore, the PSA test is widely available worldwide, and opportunistic screening is ubiquitous.

In our trial, clinically insignificant cancer was the most common cancer finding in participants who underwent systematic biopsy (72 of 140 participants [51%] in the reference group). This percentage is higher than those found in previous studies, including among patients referred for biopsy in clinical populations such as those in the PRECISION (Prostate Evaluation for Clinically Important Disease: Sampling Using Image Guidance or Not?) trial and the National Institutes of Health Trio Study.^15,28 However, the percentage of clinically insignificant cancer in our trial was similar to or lower than that in other screening trials in which participants underwent biopsy without previous clinical evaluation. For example, the percentage of Gleason score 3+3 tumors was 72% in the ERSPC trial¹ and 66% in the PLCO trial.⁹ Subsequent studies also showed high incidences of detection of Gleason 3+3 cancer when all the participants with elevated PSA levels underwent systematic biopsy (77% of all cancers detected in the CAP trial²⁹ and 41% in Stockholm-3³⁰).

The likelihood of diagnosing Gleason 3+3 cancers in our trial was reduced by 54%, but was not negligible (38%), among participants in the experimental group who underwent MRI-targeted biopsy alone. However, Gleason 3+3 cancers that were detected in the experimental group were in general larger than those detected in the reference group and may potentially have had clinical importance, especially in this relatively young age group in which grade progression might occur with increasing age.^31,32 A recent study also showed that Gleason 3+3 tumors that are visible on MRI have a higher risk of progressing than MRI-negative cancers.³³ Because the definition of overdiagnosis is controversial,^32,34 we performed three sensitivity analyses using three alternative definitions that also took into account tumor size. When we used the alternative definitions of outcomes in analyses, we found lower percentages of clinically insignificant cancer than we did in the primary analysis, especially in the experimental group (Table S4).

Our trial showed that 19% fewer clinically significant cancers were detected in the experimental group than in the reference group (relative risk, 0.81; 95% CI, 0.60 to 1.10). Clinically significant cancers in the reference group that were detected by systematic biopsy alone (in 10 participants) were mostly small and all had a Gleason score of 3+4, with a small amount of Gleason pattern 4 detected. Six of the 10 participants had disease that was managed with active surveillance, and 4 underwent radical treatment. Delayed diagnosis in these participants is unlikely to be detrimental, but future follow-up is important.³⁵

In this trial, all participants in the reference group were offered evaluation with MRI; however, because we separated data on systematic and MRI-targeted biopsies, we could compare results of systematic biopsies separately in the reference group, disregarding MRI results. The diagnoses in 10 of the 68 participants (15%) with clinically significant cancer would have been missed or delayed if only systematic biopsy had been performed, but the number of clinically insignificant cancers that were detected remained unchanged.

Our trial has several strengths, including its large size and population-based design. Sampling bias was diminished by the use of standardized biopsy templates in the two groups. Diagnostic bias was minimized by external, blinded review of all biopsy specimens. Limitations of the trial include the relatively young participant age and the single-center design, which may limit generalizability. Whether newer biopsy techniques, such as transperineal biopsy and image-guided fusion technology, may improve the diagnostic performance of the screening algorithm is unknown. At a minimum, such methods are expected to eventually produce an even larger effect than that seen in our trial.

Randomized, controlled trials that compare the effectiveness of transperineal biopsy versus biopsy guided by transrectal ultrasonography with regard to prostate cancer detection and infectious complications are currently ongoing. The feasibility and scalability of transperineal biopsy for population-based screening remains to be studied. A beneficial feature of transrectal biopsy is that it can be performed under local anesthesia in a urologist’s office. Furthermore, multiparametric MRI was used in the first screening round, reported here; however, since biparametric MRI without contrast medium has been shown to be noninferior to multiparametric MRI for prostate cancer detection and for reducing the incidence of false positive results,³⁶ the trial protocol has been amended to use biparametric MRI starting with the second screening round. In our analysis of the first round of screening in the current trial, we found that a screening algorithm that included PSA measurement followed by MRI evaluation and targeted biopsy only, as compared with systematic biopsy of all participants with elevated PSA, led to a substantial reduction of overdiagnosis at the cost of missing a limited number of clinically significant cancers.