Digital Rectal Examination in Stockholm3 Biomarker-based Prostate Cancer Screening

Take Home Message A man with a positive digital rectal examination (DRE) has more than three times higher risk of harboring clinically significant prostate cancer on biopsy than a man with a negative DRE. We support the use of DRE as a second-line test after primary biomarker-driven testing strategies.


Introduction
Prostate-specific antigen (PSA) and digital rectal examination (DRE) have been cornerstones in the workup for detection of prostate cancer since the early 1990s [1]. In the European Randomised Study of Screening for Prostate Cancer (ERSPC) and the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO), the two largest prostate cancer screening trials, DRE and PSA were initially used as screening tests. European and US guidelines state that a suspicious DRE is associated with an elevated risk of cancer and is an indication for biopsy [2,3]. Since PSA <3 ng/ml has a low positive predictive value (PPV) for detection of clinically significant prostate cancer among men with suspicious DRE, use of DRE as a ''tool'' in screening has mostly been recommended after an initial PSA screening test [4][5][6]. DRE is one of the predictors in most clinical risk calculators for prostate cancer, including the ERSPC risk calculator [7], the Prostate Biopsy Collaborative Group calculator [8], and 4K. However, the performance of DRE for prostate biopsy decisions in early detection settings that include the use of polygenic markers in prediction models has not been fully elucidated.
The Stockholm3 test is a multivariable biomarker and clinical predictive model that decreases the number of biopsies needed to find International Society of Urological Pathology grade group !2 (GG !2) prostate cancer by 32% when maintaining the same sensitivity as a PSA cutoff of 3 ng/ml [9]. Apart from two clinical parameters that require clinical workup (DRE and prostate volume), all the other parameters in the model can be collected via a web questionnaire and a blood test.
We therefore sought to analyze the value of performing DRE as part of prostate cancer screening using data from a contemporary screening-by-invitation cohort of men. We evaluated the association between DRE status and the risk of clinically significant prostate cancer on transrectal systematic prostate biopsy and how the predictive performance of the Stockholm 3 model without DRE and prostate volume compares to the full model.

Patients and methods
The data used in this study were collected prospectively between 2012 and 2015 in the population-based STHLM3 diagnostic study [9] Our main definition of clinically significant prostate cancer was GG !2 cancer on systematic transrectal ultrasound-guided prostate biopsy.
We also analyzed outcomes for GG !3 and any-grade prostate cancer. All biopsies were reviewed by a single pathologist (L.E.).
The STHLM3 study is registered on ISRCTN.com (ISRCTN84445406) and was approved by the institutional ethics review board.

Statistical analysis
A two-sided t test and v 2 test were used to assess differences in baseline variables between the suspicious and nonsuspicious DRE groups.  To assess the performance of a model using DRE in a clinical setting, we normalized results to a population of 1000 men with PSA !3 ng/ml.
We calculated the reductions in the number of men undergoing biopsy and number of significant cancer cases found when biopsying men with PSA of either >4 or 3-4 ng/ml and with suspicious DRE in comparison to using PSA >3 ng/ml or >4 ng/ml alone as the cutoff. Statistical analysis was performed using Stata v14.2 (Stata Corp., College Station, TX, USA). Figure 1 outlines the inclusion of men invited to participate in the STHLM3 screening-by-invitation study (n = 58 818). The participation rate was 41% and biopsy compliance was 72% (men biopsied/men recommended biopsy
In the multivariable regression model including prostate volume and the modified Stockholm3 model without prostate volume and DRE status as predictors, the odds ratio for the risk of GG !2 cancer for DRE + status was 2.88 (95% CI 2.32-3.57; Table 3). The increase in AUC was small but statistically significant when DRE was added as a predictor to a  Table 4). Figure 2 shows the probability of finding any-grade cancer, GG !2 cancer, and GG !3 cancer by DRE status and PSA level. For the PSA range 3-20 ng/ml, the absolute risk difference for men with DRE + status was 22-33% for GG !2 cancer (Fig. 2B) and 13-32% for GG !3 cancer (Fig. 2C). A similar association was seen on stratification by PSA density (<0.15 vs !0.15 ng/ml/cm 3 ; Supplementary Fig. 1).
Analysis revealed no significant interaction between DRE and age, Stockholm3 result, or PSA for GG !2 cancer as the outcome (Supplementary Table 2).
In the sensitivity analysis in which cases with missing data for prostate volume were excluded from the multivariable regression, the results did not change materially (Supplementary Table 3).

3.4.
Proposed model and head-to-head comparison Supplementary Table 1 shows predictions normalized to a population of 1000 men with PSA !3 ng/ml considered for further workup in a prostate cancer screening setting. We found that using a PSA threshold of either !4 ng/ml or 3-4 ng/ml and DRE+ status would lead to 40% fewer biopsies, but at the cost of 24% fewer GG !2 cancer cases identified. Using these data, 49 men with PSA of 3-4 ng/ml would need to undergo DRE, resulting in biopsy for four men that would detect 1.9 cancers and one clinically significant cancer.

Discussion
Our analysis of men from a large population-based screening study shows that DRE has sensitivity of 28% and specificity of 93% for predicting GG !2 cancer among men with PSA !3 ng/ml. These results are congruent with previous findings [11,12]. With specificity of 93% and a PPV of 46%, a suspicious DRE can aid in identifying men at higher risk of clinically significant cancer in this group.
We found that approximately one in ten men with PSA !3 ng/ml had a suspicious DRE and the proportion increased with higher PSA level. For men with a suspicious DRE the risk of having ISUP !2 cancer was nearly 50%, which is more than three times higher than the risk for men with a nonsuspicious DRE. In comparison to findings from the ERSPC (35%) [7] and PLCO (15.6%) [13] trials, we found that DRE had a higher PPV for GG !2 cancer. The difference is mostly explained by differences in population selection, mainly because we used a PSA cutoff of !3 ng/ ml. In the DRE + group the proportion of men with a previous negative biopsy was 42% less than in the DRE À group. This suggests that at a population level, the extent of prescreening for cancer before possible inclusion in this study could have been greater for the DRE + group. This could possibly explain the lower DRE + rate in our study than in previous screening trials. For instance, more than 20% of the men in the ERSPC Rotterdam cohort had a suspicious DRE in each of the different screening rounds [11].
Our results show that a DRE suspicious for cancer is strongly correlated with higher-grade cancer and greater cancer length on biopsy, confirming previous data on a larger volume of cancer in the prostate for men with a suspicious DRE [2]. The PCPT and PLCO trials showed an interaction between DRE status and PSA level for the risk of GG !2 cancer [4,14], but our data do not support such an interaction.
Our results support the recommendation that patients with a suspicious DRE should be investigated further for prostate cancer. Furthermore, the association with highergrade cancer, greater cancer length on biopsy, and by definition a minimum of T2 stage would suggest higher clinically significance among DRE + GG !2 cancers. The increase in AUC was statistically significant when DRE and prostate volume were added as predictors to the simplified Stockholm3 model using only blood test and questionnaire parameters (0.785 vs 0.752; p = 0.001). The small increase reflects the fact that many men would need the extra screening steps of DRE and measurement of prostate volume to find one significant cancer regardless of the probability threshold chosen as a cutoff for biopsy.
There are some practical concerns for DRE; because of its low sensitivity, DRE cannot be used as a ''rule out'' test, with an expected high false-negative rate. More than half of patients undergoing DRE experience non-negligible discomfort [15]. According to the data in Table 4, we can assume a relatively high number needed to screen among men with a lower risk threshold according to PSA and other parameters than in our population, such as PSA <3.0 ng/ml. We deliberately chose to exclude men in the STHLM3 study with PSA of 1-2.99 ng/ml because of the selection bias this would have introduced, since these men were biopsied on the basis of elevated risk according to the Stockholm3 model.
Optimal incorporation of DRE in screening for prostate cancer in relation to multiparametric magnetic resonance imaging (mpMRI) and guided biopsy needs further study. Morote et al [16] showed that use of mpMRI did not decrease the number of men needing biopsy in the subgroup with PSA !10 ng/ml and DRE + in a study of 768 men scheduled for prostate biopsy. Use of DRE as a secondline test to provide clinical direction regarding patients who have had negative mpMRI and/or prostate biopsies is beyond the scope of this study, but there is recent evidence that mpMRI is more beneficial for DRE À than for DRE + men in finding clinically significant prostate cancer [17].
The main strength of our study is the large populationbased invitational setting. DRE was performed by urology specialists, blinded to the PSA and Stockholm3 test results, and there was a clear definition of DRE + status in terms of T stage. Biopsy compliance for patients for whom biopsy was recommended was moderate at 71% (Fig. 1). A urologist performed systematic biopsies according to a predefined scheme. The pathology analysis was performed by an expert uropathologist who was blinded to T-stage assessment. This makes the risk of selection bias and misclassification bias low. The lack of mpMRI information in our study is the main limitation. We cannot say what independent predictive value DRE has in a risk model that includes mpMRI information. Our definition of a true positive as GG !2 cancer in one round of 10-12 systematic peripheral-zone transrectal biopsies is another limitation. The PROMIS study showed that systematic transrectal biopsy diagnosed only 48% of clinically significant prostate cancers in comparison to template prostate mapping biopsy [18]. Up to 30% of cancers are in the anterior transitional zone and these are less likely to be palpable [19]. Our biopsy scheme is not specifically aimed at the anterior zone, and this is a possible area of bias. In contemporary recommendations for early detection, mpMRI targeted biopsy is strongly recommended and has been established as a standard of care in settings where mpMRI is available [3,20]. From this study we cannot say how DRE performs with regard to any long-term oncological results such as survival or upgrading at radical prostatectomy, or how DRE performs among men with PSA <3 ng/ml.

Conclusions
DRE suspicious for cancer is highly associated with elevated risk of significant prostate cancer. Thus, prostate biopsy should be recommended for men with abnormal DRE findings. DRE and prostate volume add significant predictive value to the performance of the Stockholm3 model.
Author contributions: Joel Andersson had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Analysis and interpretation of data: Eklund, Andersson, Palsdottir.
Drafting of the manuscript: Andersson.
Financial disclosures: Joel Andersson certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Henrik Grönberg has five patents pending that are related to prostate cancer diagnostics, has patent applications licensed to A3P, and might receive royalties from sales related to these patents. Martin Eklund is named on four of these five patent applications. The remaining authors have nothing to disclose.