The comparative effectiveness of mpMRI and MRI-guided biopsy vs regular biopsy in a population-based PSA testing: a modeling study

The benefit of prostate cancer screening is counterbalanced by the risk of overdiagnosis and overtreatment. The use of a multi-parametric magnetic resonance imaging (mpMRI) test after a positive prostate-specific antigen (PSA) test followed by magnetic resonance imaging-guided biopsy (MRIGB) may reduce these harms. The aim of this study was to determine the effects of mpMRI and MRIGB vs the regular screening pathway in a population-based prostate cancer screening setting. A micro-simulation model was used to predict the effects of regular PSA screening (men with elevated PSA followed by TRUSGB) and MRI based screening (men with elevated PSA followed by mpMRI and MRIGB). We predicted reduction of overdiagnosis, harm-benefit ratio (overdiagnosis per cancer death averted), reduction in number of biopsies, detection of clinically significant cancer, prostate cancer death averted, life-years gained (LYG), and quality adjusted life years (QALYs) gained for both strategies. A univariate sensitivity analysis and threshold analysis were performed to assess uncertainty around the test sensitivity parameters used in the MRI strategy.In the MRI pathway, we predicted a 43% reduction in the risk of overdiagnosis, compared to the regular pathway. Similarly a lower harm-benefit ratio (overdiagnosis per cancer death averted) was predicted for this strategy compared to the regular screening pathway (1.0 vs 1.8 respectively). Prostate cancer mortality reduction, LY and QALYs gained were also slightly increased in the MRI pathway than the regular screening pathway. Furthermore, 30% of men with a positive PSA test could avoid a biopsy as compared to the regular screening pathway. Compared to regular PSA screening, the use of mpMRI as a triage test followed by MRIGB can substantially reduce the risk of overdiagnosis and improve the harm-benefit balance, while maximizing prostate cancer mortality reduction and QALYs gained.

Screening strategy. In our previous study, we compared more than 200 population-based prostate cancer screening strategies, and we found that screening with 3 years interval at ages 55-64 would be the optimum screening strategy 19 . All men with an elevated serum PSA level (cut-off 3 ng/mL) were referred to TRUSGB in that study. Those who were positive at TRUSGB were assigned to either RP, RT or AS according to the treatment distribution mentioned before. The biopsy compliance rate after a positive screen test result was assumed to be 90%, with a sensitivity of 90% as observed in the ERSPC Rotterdam data 20,21 . An 80% screening attendance rate was assumed. The total number of biopsies was calculated by using the number of screen detected cancers and a mean positive predictive value of 22.7% of a biopsy in the screen arm of the ERSPC 21 and by using the number of clinically detected cancers and the positive predictive value of 35.8% of a biopsy in the control arm 22 .
In the present study, we included mpMRI as a triage test to this screening strategy (screening with 3-year intervals at ages 55-64) for those men with an elevated PSA level (cut-off 3 ng/mL) before referring them to a biopsy (MRIGB) (Fig. 1). PIRADS scores of 3-5 were considered positive for the mpMRI test. It is important to note that we did not use a combined biopsy, rather those men with positive mpMRI tests were subjected only to an MRI-guided biopsy (no systematic biopsy). The same screening attendance and biopsy compliance were assumed as in the regular pathway. A positive predictive value of 58% 23 was assumed to calculate the total number of biopsies in this strategy. Men positive at MRIGB were assigned to the same treatment options as in TRUSGB. Grade specific sensitivity values for mpMRI and MRIGB were mainly based on literature that used meta-analysis (Table 1). Although a very recent meta-analysis reported by Drost et al. 24 was not included in our study, most of the test sensitivity parameter values reported in that study are within the range of the values that we used for our sensitivity analysis. We also accounted for misclassification of grades both in the MRIGB and regular biopsy. In our study misclassification of grades represents only wrong classification of clinically significant cancer in to insignificant cancer at biopsy. For the MRIGB we used an 8.7% misclassification rate based on Ahdoot et al. 2  www.nature.com/scientificreports/ gained, life-years gained (LYG) per death averted, QALYs gained and QALYs gained per death averted. In this study, clinically significant prostate cancer was defined as Gleason score 7 or more and clinically insignificant cancer as Gleason score 6 and less 5 . In both screening strategies, a hypothetical cohort of 10 million men was simulated over a lifetime period. All the results are reported per 1000 men.

Quality of life.
The quality adjusted life-years (QALYs) were calculated based on the utility estimates of given health states where patients remain for a certain period of time. The utility values range between 0 (death) and 1 (perfect health), and one minus the utility value gives a loss in utility at each health state. By multiplying the number of men in a given health state with the loss in utility and the duration of the health state, the loss in quality of life was calculated. The utility estimate (0.96) and duration (1 week) for mpMRI were based on Grana et al. 27 . There is evidence that MRIGB is associated with less frequent adverse outcomes compared with TRUSGB 3 . Therefore, we assumed 50% lower disutility for MRIGB compared with that of TRUSGB. All other utilities and durations were based on our previous study 12 (Appendix_Table S1).   www.nature.com/scientificreports/ interpretation of both mpMRI and MRIGB are highly influenced by the specialists (radiologist or urologist) skills, we varied the test sensitivity parameters for the analysis using the 95% confidence intervals indicated in Table 1. A threshold analysis was also performed on QALYs per death averted by changing the baseline sensitivity values of the mpMRI and MRIGB simultaneously.

Results
Base model. The total numbers of men referred to a biopsy were 396 and 278 for the regular and the MRI pathway respectively, a 30% reduction (Table 2). Our model predicted 16 overdiagnosed cases for the regular pathway and 9 (43% reduction) for the MRI pathway (overdiagnosed cancer was defined as a prostate cancer detected during screening but would not have been clinically diagnosed during the man's life time in the absence of screening). The model predicted a 2.7% higher prostate cancer mortality reduction for the MRI pathway than the regular pathway (8.77 vs 8.53). The MRI based screening was also associated with a lower harm-benefit ratio (overdiagnosis per cancer death averted) than the regular screening (1.0 vs 1.8). Our model predicted a higher LY gained (85 vs 81.6) and QALYs gained (80.2 vs 77) in the MRI pathway than the regular screening pathway. Clinically significant prostate cancer was detected in 51.3 men in the MRI pathway, as compared with 36 in the regular pathway (30% increment in the detection rate of clinically significant prostate cancer). In contrary, fewer men were diagnosed with clinically insignificant prostate cancer in the MRI pathway than the regular pathway (59 vs 80.8), which resulted in a 27% reduction. However, the MRI pathway was also associated with an 11% risk of missing clinically significant cancer due to not performing biopsy in the mpMRI negative patients.

Sensitivity analysis.
After varying the baseline sensitivity values of the MRI pathway, using the 95% confidence intervals or ranges, the harm-benefit ratio (overdiagnosis per death averted) remained lower in the MRI pathway than the baseline value (1.8) of the regular pathway (Fig. 2). The threshold analysis indicated that when the baseline test sensitivity values of the MRI pathway were changed by 14% simultaneously, the QALYs/death averted became the same for the two strategies (Fig. 3). To be the QALYs per death averted in favour of the MRI pathway, the sensitivity of mpMRI and MRIGB for clinically significant prostate cancer should be higher than 81% and 78% respectively; whereas, for that of clinically insignificant prostate cancer it should be lower than 84% and 50% respectively.

Discussion
The benefit of prostate cancer screening in reducing advanced stage disease or mortality is counterbalanced by the risk of overdiagnosis and overtreatment 28 . In our study, when mpMRI was applied after a positive PSA test and followed by MRIGB, the risk of overdiagnosis was decreased substantially (by 43%) compared with the regular screening. This result can be taken confirmatory for previous studies that proposed the use of mpMRI and MRIGB as a potential means to reduce the risk of overdiagnosis. The lower harm-benefit ratio predicted in the present study could also inform policymakers about the role of MRI in a population-based prostate cancer screening.
When the MRI pathway was used instead of the regular pathway, 30% of men avoided biopsies. A recent study by Kasivisvanathan et al. 5 reported a 28% biopsy reduction due to the use of mpMRI and MRIGB. As compared to the regular pathway, the MRI pathway was also associated with a 30% higher detection rate and 27% lower detection rate for clinically significant and insignificant prostate cancer, respectively. A meta-analysis 29 concluded that MRIGB has a higher detection rate for clinically significant prostate cancer and a lower detection rate for insignificant cancer compared with TRUSGB. More specifically, Siddiqui et al. 30 reported MRIGB increases the detection of high risk cancer by 30% (compared to TRUSGB), and Leest et al. 31 indicated TRUSGB would over detect insignificant cancer in 20%. The number (percentage) of clinically significant cancers reported in our study (in both pathways) are lower than the number reported by Kasivisvanathan et al. 2018, who used the same definition. The main reason for this discrepancy could be the difference in population characteristics of the two studies. For instance, the upper age limit included in the present study was 64 years, whereas in Kasivisvanathan et al. 2018 the mean age was 64 ± 7. Therefore, the older age groups in the Kasivisvanathan et al. 2018 may contribute to the higher number of high grade cancers (grade 7 and above) than reported in our study. Although it is difficult to directly compare our results with the above studies (because of differences such as, population characteristics, follow-up period and screening strategy), the general conclusion is the same: the use of mpMRI and MRI guided biopsy is superior over that of the regular pathway.
Using of the MRI pathway resulted in an increased LYG, QALYs gained, and prostate cancer death averted compared to the regular pathway. The increased in LYG and mortality benefit in the MRI pathway can be explained by the increased detection of clinically significant cancer (by about 30%), and the lower misclassification rate of grades by MRIGB (compared to TRUSGB), that were included in our model. On the other hand, the lower detection rate of clinically insignificant cancer in the MRI pathway could explain the higher QALYs gained. However, the MRI pathway also failed to detect around 11% of clinically significant cancer, that would be detected in the regular pathway, and this could explain the smaller difference in mortality benefit between the two strategies. This percentage is in agreement with a previous study by Pokorny et al. 6 . The small QALYs difference reported between the two strategies may raise a question of whether the MRI-pathway can be an efficient strategy, especially in relation to the initial additional expenditures required in the MRI-pathway. However, a substantial amount of biopsies were avoided as a result of using the MRI pathway, and this could compensate for the additional expenditures. www.nature.com/scientificreports/ Our prediction of the lower harm benefit ratio (overdiagnosis per death averted) for the MRI pathway than the regular pathway was robust to the sensitivity analysis (Fig. 2). It is also important to note from the figure that, increasing the sensitivity of mpMRI and MRIGB for high grade cancer resulted in a more better harm benefit ratio, and lowering theses sensitivities relatively worse the ratio. In contrary, lowering the sensitivity of mpMRI and MRIGB for low grade cancer makes the ratio more better, and increasing these sensitivities makes the ratio relatively worse. The threshold analysis showed that when the baseline test sensitivity values of the MRI pathway were changed by 14% simultaneously (this means increasing the sensitivities of mpMRI and MRIGB for low grade cancer and decreasing for high grade cancer by 14% simultaneously), the QALYs per death averted became the same for the two strategies. This may signify the importance of adhering to proper imaging protocol as well Table 2. Predictions of the effects of prostate cancer screening for men between age 55-64 at 3 years intervals using regular pathway and MRI pathway, per 1000 men. a Clinically insignificant cancer was defined as Gleason score 6 and below (it contains both screen detected and interval cancer). b Clinically significant prostate cancer was defined as Gleason score 7 or more (it contains both screen detected and interval cancer). c Overdiagnosed cancer was defined as a prostate cancer detected during screening but would not have been clinically diagnosed during the man's life time in the absence of screening. An important strength of this study is that we were able to quantify the effect of MRI based prostate cancer screening on the risk of overdiagnosis, which is obviously not observable in trial studies. We also quantified the effect of the MRI pathway on the harm-benefit ratio (overdiagnosis per death averted) as compared to the regular pathway, which was also not reported in previous studies. Furthermore, we were able to evaluate the MRI pathway in a population based screening setting. Although our model is calibrated to the Dutch prostate cancer incidence, the results may also be extrapolated to other western populations with similar prostate cancer incidence trends. Study shows that in Western Europe, the incidence of prostate cancer has been on the rise 33 .
Our study has also certain limitations. First, we assumed the same mortality benefit for radiation therapy as that of radical prostatectomy, since there is no clinical trial that compared the two treatment directly. We also assumed that treatment options will not change in both strategies. However, treatment behavior may change in the future, such as more active surveillance than now. Cost is another important factor which was not included in this study. However, avoidance of biopsies and subsequent biopsy related complications and treatment costs, probably make the MRI pathway cost-effective or at least compensate its additional costs. Various studies, though not population-based studies, indicated that the inclusion of mpMRI after a positive PSA test followed by MRIguided biopsy is cost-effective compared to a regular prostate cancer screening pathway [34][35][36] . Future studies are needed to evaluate this in a population-based screening settings. Lastly, a probabilistic sensitivity analysis was not included in our study: only a one-way sensitivity analysis and threshold analysis were included.
In conclusion, our modeling results indicated that the use of mpMRI after a positive PSA test followed by MRIGB can substantially reduce the risk of overdiagnosis and improve the harm-benefit ratio, while maximizing prostate cancer mortality reduction and QALYs gained, as compared to the regular screening pathway.
Received: 15 September 2020; Accepted: 5 January 2021 Figure 3. A threshold analysis diagram indicating the QALYs per cancer death averted continues to be in favor of the MRI pathway when the sensitivity values of the MRI pathway were changed simultaneously by up to 14% (this means increasing the sensitivities of mpMRI and MRIGB for low grade cancer and decreasing for high grade cancer by up to 14% simultaneously). Increasing the sensitivity for low grade cancer means detecting more Gleeson 6 cancer and decreasing the sensitivity for high grade cancer means detecting fewer clinically significant cancer which reduce the QALY per death averted.