Active Surveillance Strategies for Low-Grade Prostate Cancer: Comparative Benefits and Cost-effectiveness

Abstract

Background

Active surveillance (AS) is the recommended treatment option for low-risk prostate cancer (PC). Surveillance varies in MRI, frequency of follow-up, and the Prostate Imaging Reporting and Data System (PI-RADS) score that would repeat biopsy.

Purpose

To compare the effectiveness and cost-effectiveness of AS strategies for low-risk PC with versus without MRI.

Materials and Methods

This study developed a mathematical model to evaluate the cost-effectiveness of surveillance strategies in a simulation of men with a diagnosis of low-risk PC. The following strategies were compared: watchful waiting, prostate-specific antigen (PSA) and annual biopsy without MRI, and PSA testing and MRI with varied PI-RADS thresholds for biopsy. MRI strategies differed regarding scheduling and use of PI-RADS score of at least 3, or a PI-RADS score of at least 4 to indicate the need for biopsy. Life-years, quality-adjusted life-years (QALYs), and incremental cost-effectiveness ratios were calculated by using microsimulation. Sensitivity analysis was used to assess the impact of varying parameter values on results.

Results

For the base case of 60-year-old men, all strategies incorporating prostate MRI extended QALYs and life-years compared with watchful waiting and non-MRI strategies. Annual MRI strategies yielded 16.19 QALYs, annual biopsy with no MRI yielded 16.14 QALYs, and watchful waiting yielded 15.94 QALYs. Annual MRI with PI-RADS score of at least 3 or of at least 4 as the biopsy threshold and annual MRI with biopsy even after MRI with negative findings offered similar QALYs and the same unadjusted life expectancy: 23.05 life-years. However, a PI-RADS score of at least 4 yielded 42% fewer lifetime biopsies. With a cost-effectiveness threshold of $100 000 per QALY, annual MRI with biopsy for lesions with PI-RADS scores of 4 or greater was most cost-effective (incremental cost-effectiveness ratio, $67 221 per QALY). Age, treatment type, risk of initial grade misclassification, and quality-of-life impact of procedural complications affected results.

Conclusion

The use of active surveillance (AS) with biopsy decisions guided by findings from annual MRI reduces the number of biopsies while preserving life expectancy and quality of life. Biopsy in lesions with PI-RADS scores of 4 or greater is likely the most cost-effective AS strategy for men with low-risk prostate cancer who are younger than 70 years.

© RSNA, 2021

Online supplemental material is available for this article.

An earlier incorrect version appeared online. This article was corrected on August 25, 2021.

Summary

MRI in active surveillance of men with low-risk prostate cancer resulted in a higher quality-adjusted life expectancy and increased cost-effectiveness compared with watchful waiting or surveillance without MRI.

Key Results

• ■ A Prostate Imaging Reporting and Data System (PI-RADS) score of at least 4 as a biopsy threshold, instead of a score of at least 3, at annual MRI had a lower incremental number of lifetime biopsies per life-year while preserving high-grade cancer detection.

• ■ The Prostate Cancer Research International Active Surveillance schedule of MRI (years 1, 3, 7, and 10; and every 5 years thereafter) with PI-RADS score of 4 as a biopsy threshold minimized excess prostate biopsies.

• ■ The active surveillance strategy with the highest economic value was annual MRI with PI-RADS score of at least 4 as a threshold for repeat biopsy.

Introduction

Prostate cancer (PC) is the most commonly diagnosed malignancy and second most common cause of cancer-related mortality in men in the United States (1). However, a large proportion of PC is classified as low risk (2,3). As a result, conservative management with active surveillance (AS) and watchful waiting has emerged as increasingly common and recommended for low-risk PC (4–6) because efficacy was demonstrated in randomized trials (7,8). For men electing for AS, to our knowledge there is no consensus regarding the optimal frequency and composition of follow-up testing (eg, prostate-specific antigen [PSA] testing, MRI, and prostate biopsy), and there has been little comparative evaluation of the economic impact and quality-of-life considerations for different testing regimens.

Most previously published cost-utility analyses have focused on non-MRI–based strategies (9–12), with one exception wherein AS with MRI every 5 years improved the quality-adjusted life expectancy (LE; QALE) compared with annual biopsy–based strategies and was also less costly than the use of more frequent MRI intervals (13). However, this study was relatively narrow in the scope of MRI application and indications for biopsy. Specifically, there is no consensus on the Prostate Imaging Reporting and Data System (PI-RADS) version 2 (14) threshold for biopsy when patients undergo prostate MRI during AS. Although practice guidelines do not explicitly support deferring biopsy for a PI-RADS score of 3 with prior negative biopsy results (15) and although a PI-RADS score of 3 is a commonly applied biopsy criterion, to use a threshold PI-RADS score of 4 in certain settings may result in the optimal net benefit through a modest increase in specificity (16–18). The differential specificity and net harms of false-positive findings at MRI resulting in unnecessary biopsies are particularly relevant for shared decision-making in PC treatment because some men prefer to avoid additional biopsies as a reasonable trade-off for a more probable biopsy result of high-grade PC.

To guide policy for prostate MRI use, which may be tied to its value in the treatment of PC, decision analysis offers an assessment of population-level utility and the conditions of this value in generalizable terms. To assess the comparative costs, effectiveness, and benefits and harms of prostate MRI in AS, including use of different PI-RADS scores as the criterion for biopsy, we developed a microsimulation model. We sought to evaluate choices of strategies for AS and thresholds for biopsy on the basis of considerations of quality of life, LE, and value. Our purpose was to compare the effectiveness and cost-effectiveness of AS strategies with and without prostate MRI in hypothetical men with a diagnosis of localized, low-risk PC.

Materials and Methods

Model Overview

This mathematical modeling study did not involve patient data collection and institutional review board approval was not required. A state-transition microsimulation model was constructed for a lifetime horizon by using TreeAge Pro version 2019 (TreeAge Software). The prior natural history cohort model, reported by Loeb et al (19) for low-risk PC and AS, included biopsy with MRI in one tested strategy for analysis of LE and QALE. Briefly, the natural history model accounted for both the correct diagnosis of a low-grade (Gleason score of 6) prostate tumor and the possibility of the presence of an actual high-grade prostate tumor, either present at the beginning of the model or developing over time. The risk of misclassifying a tumor as low-risk PC at the baseline evaluation was on the basis of prospective studies about AS in which the patients underwent a repeat biopsy or prostatectomy within 6 months that showed a high-grade tumor. The rate of progression to a high-grade tumor was estimated by using prospective data on watchful waiting and a Bayesian modeling study (20–24). The natural history disease model was then validated against independent prospective patient data about AS and watchful waiting (Appendix E1 [online]). For the current comparative effectiveness and economic evaluation, we constructed a decision tree of multiple MRI strategies, incorporated the costs and health utilities (quality-of-life measures) associated with health events, and enabled tracking of individual hypothetical patients to increase the representation of patient heterogeneity and to summarize intermediate oncologic outcomes.

The base case population was 60-year-old men with a diagnosis of low-risk, localized PC. Additional age groups were modeled from ages 50 to 70 years. Low-risk PC was defined by using the D’Amico risk classification system (biopsy Gleason score of 6, clinical stage of T1c–T2a, and PSA level of ≤ 10 ng/mL) in patient selection for conservative treatment. The simulated patients were either classified accurately as having low-grade disease or were misclassified and had undetected higher-grade PC (Gleason score > 6). At the end of each monthly cycle, patients proceeded to the treatment pathway indicated by the clinically assumed disease state, which included continuing AS, undergoing prostatectomy for intermediate or high-grade PC or the progression to metastasis, or death. An alternative treatment scenario of brachytherapy for progression of disease with AS was also applied. Causes of death included PC, all-cause mortality, or surgical mortality.

Diagnostic Strategies for AS

We compared multiple AS strategies, varying in their use of diagnostic tests and schedules, with watchful waiting (Fig 1). Simulated patients undergoing watchful waiting did not undergo testing unless symptomatic disease developed, at which time patients were treated for advanced-stage PC. The performance characteristics of PSA kinetics were modeled by using reported performance levels in prospective studies of AS (19,25). AS without prostate MRI included PSA testing every 6 months and a yearly biopsy on the basis of the Johns Hopkins AS protocol (26). MRI in the prostate with AS included different schedules and PI-RADS thresholds for indication of biopsy and assumed the use of contrast enhancement as recommended by PI-RADS (27). AS strategies with prostate MRI included annual MRI and systematic biopsy regardless of the findings at MRI (wherein MRI might be used to identify suspicious lesions to target, with biopsy pursued even after negative findings at MRI); annual MRI with systematic and targeted biopsy of lesions with a PI-RADS of 3 or greater (with a score of 3 indicating that the clinically significant tumor is equivocal); annual MRI with systematic and targeted biopsy of lesions with a PI-RADS score of at least 4; and annual MRI only for rising PSA levels, with biopsy performed for findings with a PI-RADS score of at least 3. Finally, the least invasive protocol entailed PSA testing every 6 months and MRI scheduled according to the Prostate Cancer Research International Active Surveillance (PRIAS) protocol (ie, MRI at years 1, 3, 7, and 10; and every 5 years thereafter), with biopsy performed only in lesions with a PI-RADS score greater than or equal to 4 (21).

Figure 1:

Simplified schematic of active surveillance strategies and testing consequences in the model. These strategies were compared with each other and watchful waiting (not depicted) in which simulated patients were not tested or treated until they presented with clinical progression. Unless otherwise specified, prostate-specific antigen (PSA) status did not influence whether MRI was performed. The strategies were as follows: (1) PSA every 6 months and requisite annual biopsy, no MRI; (2) PSA and annual MRI, annual biopsy regardless of MRI results; (3) no MRI if PSA is stable, otherwise annual MRI by using Prostate Imaging Reporting and Data System (PI-RADS) of at least 3 to determine need for biopsy; (4) MRI on Prostate Cancer Research International Active Surveillance (PRIAS) schedule (at years 1, 3, 7, 10, and then every 5 years) by using PI-RADS of at least 4 to determine need for biopsy; (5) annual MRI by using PI-RADS of at least 4 to determine need for biopsy; and (6) annual MRI by using PI-RADS of at least 3 to determine need for biopsy MRI+ = positive MRI result, MRI- = negative MRI result.

Model Inputs

The model input selection entailed a systematic literature search and an appraisal of study quality. We incorporated patient age, the comorbidity burden as related to all-cause mortality risk, the histologic grade of misclassified cancer, the risks of cancer progression, and termination of surveillance at age 75 years (4). All-cause mortality rates were derived from U.S. life tables adjusted a priori by using an established multiplier of 0.45 to account for the highly selective use of an AS patient population with a lower comorbidity burden and localized PC (26). The mortality multiplier was tested across a range of values to assess the impact of competing mortality risks on the comparative effectiveness of strategies. Transition probabilities and state-specific utilities were derived from the literature specific to those health states and national registries. Represented harms included complications from biopsy, short-term and long-term complications of treatment, and development of metastasis. The values for sensitivity and specificity of a PI-RADS score of at least 3 in the model were 96% and 29%, respectively (28) (Table 1). In strategies with a PI-RADS score of at least 4 as the threshold, the sensitivity and specificity were 90% and 62%, respectively, from previous meta-analysis (28). Costs for diagnostic tests, treatment, and office visits were derived from 2019 Medicare reimbursements or were adjusted to 2019 U.S. dollars. Validation of the low-risk PC natural history model was reported previously (19).

Table 1:

Major Parameter Values for Model Inputs

Outcomes

The primary outcomes evaluated were LEs, QALEs, lifetime costs, and incremental cost-effectiveness ratios accumulating over monthly cycles. QALE is a composite measure of the length and quality of life; 1 quality-adjusted life-year (QALY) is defined as 1 year lived in perfect health (58). Tabulated intermediate outcomes included the number of biopsies and PC progression to metastasis.

Analysis

In addition to calculating the primary outcomes, we identified a set of surveillance strategies that were projected to require more biopsies and provide a lower LE than another strategy of AS. These strategies were considered dominated because of their relative inefficiency in terms of the incremental number of biopsies required per gain in life-year. We constructed an efficient frontier of the remaining strategies, which is the line connecting testing strategies that provide the largest incremental increase in life-year gained per prostate biopsy.

For cost-effectiveness, a willingness-to-pay threshold of $100 000 was applied for the base case analysis and then varied from $0–$200 000; the strategy resulting in the highest incremental cost-effectiveness ratio but resulting in less than $100 000 per QALY was considered the most cost-effective strategy. The net monetary benefit for each strategy is a standard metric that combines cost, effectiveness, and willingness to pay into a single measurement (59). We applied 3% annual discounting of costs and QALYs for the base case scenario.

We evaluated an alternative treatment scenario of brachytherapy for tumor-grade progression during AS because of its recent recognition as a treatment option for intermediate-risk disease that has efficacy similar to and a lesser quality-of-life impact than prostatectomy (Appendix E1 [online]) (4,60,61). By using sensitivity analysis, we evaluated the stability of results over changes in model estimates. One- and two-way deterministic sensitivity analyses were used to test the robustness of strategy rankings across a range of variable values. In addition, probabilistic sensitivity analysis was used to estimate the impact of parameter uncertainty by varying the values of all major parameters simultaneously by using prespecified distributions to represent variables (Table E1 [online]). Each distribution was sampled 10 000 times with 1000 iterations per sampling. Value-of-information analysis was conducted to assess whether a reduction in uncertainty would impact the economic favorability of a given strategy, enabling identification of high-priority research questions (62).

Results

Base Case Analysis

For the base case of 60-year-old men and a treatment scenario of robotic prostatectomy, three strategies offered equivalent LE and QALE outcomes: annual MRI with a PI-RADS score of at least 3 used as the criterion for biopsy, annual MRI with a PI-RADS score of at least 4 used as the criterion for biopsy, and annual MRI with biopsy regardless of MRI findings. These strategies yielded essentially equivalent LE values of 23.05 life-years and 16.19 QALYs.

The number of biopsies was lowest in the strategy that used a PRIAS MRI schedule and a PI-RADS score of at least 4 as the biopsy threshold, and the number of biopsies was highest in strategies that used an annual biopsy without MRI (Table E2 [online]). Compared with the use of PI-RADS score of at least 3, the use of PI-RADS score of at least 4 as the positive-test-result criterion at annual MRI reduced the number of biopsies from 55 543 to 32 132 per 100 000 patients (a 42% reduction), with a similar metastatic disease case decrease during surveillance (129 vs 125, respectively). Thus, the use of a PI-RADS score of 3 as the test-result threshold was dominated when considering the harms of additional procedures and no comparative gain in the LE (Fig 2). Strategies preferred on the efficient frontier using criteria of incremental gains in LE with increased numbers of biopsies included the strategy of a PRIAS MRI schedule with a PI-RADS score of 4 as the biopsy threshold and annual MRI with a PI-RADS score of 4 as the biopsy threshold.

Figure 2:

Efficient frontier according to the lifetime number of prostate biopsies and the life expectancy gain per 1000 patients. Strategies at the frontier (line) provided comparative gains in life-years for additional biopsies performed. PI-RADS = Prostate Imaging Reporting and Data System version 2, PRIAS = Prostate Cancer Research International Active Surveillance, PSA = prostate-specific antigen.

When explicitly considering costs and applying a cost-effectiveness threshold of $100 000 per QALY, performing AS with annual MRI and biopsy only in lesions with a PI-RADS score of at least 4 was the most cost-effective strategy, resulting in an incremental cost-effectiveness ratio of $67 221 per QALY and the highest net monetary benefit (Fig 3, Table E3 [online]). Meanwhile, annual MRI with a PI-RADS score of at least 3 as the criterion for biopsy yielded LE and QALE values that were equivalent to those obtained by using annual MRI with a PI-RADS score of at least 4 as the criterion for biopsy, but the former strategy resulted in a higher cost ($2352) and was therefore considered a dominated strategy (Table 2). Both watchful waiting and MRI surveillance with the PRIAS protocol yielded a lower QALE and were also less expensive than the annual MRI strategies. For example, the PRIAS schedule of MRI cost $45 171 (95% CI: $38 450, $52 488) and resulted in a QALE of 16.14 QALYs (95% CI: 16.02, 16.26). The annual biopsy without MRI strategy was dominated because it resulted in fewer QALYs and higher costs than the most favorable strategies. Higher willingness-to-pay thresholds further increased the likelihood that the most cost-effective strategy would be the use of annual MRI with PI-RADS score of at least 4 as the criterion for biopsy (Fig E1 [online]). Annual MRI with PI-RADS score of at least 4 as the biopsy threshold remained the favored strategy, with cost-effectiveness thresholds varying up to $200 000 per QALY. Watchful waiting was the most cost-effective strategy only for thresholds under $75 000 per QALY.

Figure 3:

Cost-effectiveness analysis graph for a 60-year-old man with newly diagnosed low-risk prostate cancer for willingness to pay (WTP) $1000 000 per quality-adjusted life-year (QALY). AS = active surveillance, PI-RADS = Prostate Imaging Reporting and Data System version 2, PRIAS = Prostate Cancer Research International AS, PSA = prostate-specific antigen.

Table 2:

Cost-effectiveness Analysis Results for Active Surveillance Strategies in 60-year-old Men with Newly Diagnosed Low-Risk Prostate Cancer

Patient age at diagnosis and the treatment option used for tumors progressing to a Gleason score of 7 or more affected the economic utility of prostate MRI and AS protocols (Table 3). For men aged 50–65 years, annual MRI with systematic and targeted biopsy for PI-RADS score of at least 4 was most cost-effective strategy, whether the tumor was treated with surgery or brachytherapy. At ages 65–70 years, MRI AS strategies that used prostatectomy for a Gleason score of at least 7 exceeded the cost-effectiveness threshold of $100 000 per QALY, although there was a continued LE benefit (of 0.17 life-years) compared with watchful waiting. A steep increase in the incremental cost-effectiveness ratio at ages older than 65 years to over $1 million was mainly driven by a minimal incremental benefit in QALYs (0.005 QALYs) compared with watchful waiting, which was because of prostatectomy-related complications. In patients older than age 70 years, watchful waiting extended the QALE compared with AS when the treatment was radical prostatectomy.

Table 3:

Incremental Cost-effectiveness Ratios by Patient Age at Diagnosis with Prostatectomy as Treatment for Prostate Cancer Progression

The alternative treatment scenario of brachytherapy was evaluated, given the impact of prostatectomy complications on QALE and the less lengthy quality-of-life impact reported for radiation therapy (57). In men aged 50–70 years, AS with MRI with PI-RADS score of at least 4 as the repeat-biopsy threshold remained the most cost-effective treatment option (eg, incremental cost-effectiveness ratio of $64 476 per QALY in men aged 65 years) (Table E4 [online]). The PRIAS schedule with a PI-RADS score of at least 4 and treatment with brachytherapy was another cost-effective option (eg, incremental cost-effectiveness ratio of $63 000 per QALY in men aged 65 years). The results were robust to the range of competing all-cause mortality rates that were tested. The only exception was that in men aged 70 years who had average mortality risk, the only cost-effective strategy was the PRIAS schedule with a PI-RADS score of at least 4. Notably, all simulated subgroups had an LE over 10 years. Treatment recommendations regarding both treatment scenarios are in Figure 4.

Figure 4:

Decision-making algorithm on the basis of quality-adjusted life expectancy and cost-effectiveness when comparing watchful waiting and multiple active surveillance options with or without prostate MRI. PI-RADS = Prostate Imaging Reporting and Data System version 2, PRIAS = Prostate Cancer Research International Active Surveillance. * = Severity as determined by providers, given wide variability in life expectancy across types of comorbid conditions. Reflects patients with life expectancy of at least 10 years.

Results of Sensitivity Analysis

Model results were affected by the probability of complications after biopsy, initial tumor-grade misclassification, quality of life during surveillance of low-risk PC (related to anxiety or depression), and utility outcomes because of complications of prostatectomy (Table E5 [online]). For example, given a low risk of PC grade misclassification (<15%), watchful waiting provided the most value at the population level (Figs 5–E2 [online]). Among the varying levels of patient adherence to MRI and biopsy, perfect adherence further strengthened the favorability of annual MRI with PI-RADS score of at least 4 as a biopsy criterion.

Figure 5:

Two-way sensitivity analysis, varying the sensitivity of MRI by using a Prostate Imaging Reporting and Data System (PI-RADS) version 2 score greater than or equal to 4 and the risk of initial tumor-grade misclassification. A low risk of tumor-grade misclassification results in less frequent MRI and biopsy performance (Prostate Cancer Research International AS protocol with MRI), showing greater value; however, low risk favors watchful waiting. *Base case values are marked. AS = active surveillance, PSA = prostate-specific antigen.

Across the broad range of tested sensitivities and specificities of MRI with PI-RADS score of at least 4, this positive-test-result criterion remained the most cost-effective (Fig E3 [online]). Even when accounting for parameter uncertainty in probabilistic sensitivity analyses, the QALE was no greater with use of a PI-RADS score of at least 3 versus a PI-RADS score of at least 4 as a biopsy criterion; use of the former criterion remained more costly across all tested age groups (Table E6 [online]).

Finally, value-of-information analysis showed that the expected value of perfect information was $15 527 per patient. This was mostly attributable to utilities related to treatment, surveillance, and uncertainty regarding patient adherence to the testing schedule. On the basis of an estimated 82 500 men in the United States who are diagnosed with low-risk PC each year (63), the expected value of perfect information for the population of current and future patients was $6.0 billion over 5 years.

Discussion

Recent studies support the use of prostate MRI as a case-finding tool for clinically significant tumors and as a tool for reducing the number of biopsies for men undergoing long-term active surveillance (AS) (64,65). By considering the recognized options of surgery or radiation for progression of prostate cancer during AS for men aged 70 years or younger, we found that use of prostate MRI to determine the need for biopsy improved the quality-adjusted life expectancy (LE) and is more cost-effective compared with non-MRI strategies. For example, the use of annual MRI and a Prostate Imaging Reporting and Data System (PI-RADS) score of at least 4 yielded an additional 0.04 quality-adjusted life-years (QALYs) compared with AS with only prostate-specific antigen (PSA) testing and repeat biopsies. The former strategy had an incremental cost-effectiveness ratio of $67 221 per QALY, whereas the use of PSA testing and biopsies was both less effective and more costly (ie, it was a dominated strategy). Important for patient care and policy decisions, our findings support a reasonable set of choices for MRI AS schedules and PI-RADS thresholds for repeat biopsies, depending on strength of preferences for avoiding repeat testing and unnecessary biopsies. The greatest efficiency in terms of LE gain per additional prostate biopsy was found with the Prostate Cancer Research International Active Surveillance schedule of surveillance and annual MRI, with both applying a PI-RADS score of at least 4 as the biopsy threshold. Annual MRI with a PI-RADS score of at least 4 as the biopsy threshold yielded essentially the same LE (23.05 life-years) and resulted in an estimated 40% reduction in biopsies compared with the threshold of a PI-RADS score of at least 3.

Modeling results were mainly driven by higher specificity and similar sensitivity of a PI-RADS score of 4 compared with a PI-RADS score of 3 and by the quality-of-life impact associated with biopsies. Regarding the recent broad support for shared decisions about the trade-offs of treatment options for PC, including variations of AS, these choices of surveillance protocols with annual MRI may be weighed by using the patient preference regarding biopsy avoidance and maximizing the detection of disease progression. The risks of missing PC progression with a PI-RADS score of at least 4 as the threshold for repeat biopsy rather than the commonly applied PI-RADS score of at least 3 are offset by the quality-of-life harms from procedures that do not result in an LE benefit across a broad risk range of tumor-grade misclassification. A prior prospective study (66) regarding the diagnostic yield enabled by PI-RADS scores in AS also reported the potential for a substantial reduction in repeat biopsies and a miss rate of 18% for disease with a Gleason score of 7 (any volume) when lesions with a PI-RADS score of 3 were monitored instead of undergoing a repeat biopsy. More certainty about patient adherence to surveillance (eg, as may be driven by the tolerability of undergoing MRI and procedures or by the social determinants of health) represented a high-priority research area according to our findings. In addition, although this model does not explicitly evaluate differences in patient acceptability or access to tests, an important implication of our findings is that a lack of access to prostate MRI as an option for AS could potentially worsen disparities in PC outcomes.

Another important treatment implication of our results is the lack of impact on relative economic favorability, with further improved specificity from the base case value of 62% demonstrated with the use of a PI-RADS score of 4 as the positive-test-result criterion. Further improvements in specificity did not outright favor the PRIAS schedule of MRI and biopsy unless the risk of initially misclassifying a high-grade tumor was low, within the range of 16%–20%. The use of negative biomarker results for risk evaluation, such as those from an initial MRI examination, the prostate health index, and the four-kallikrein panel, were associated with a high-grade tumor risk in the range of 11%–20%, and it is feasible that with further validation, risk-stratification at the outset of AS decision-making could support the use of less frequent testing after the 1st year (67).

Prior cost-effectiveness analyses have compared AS protocols, including AS with prostate MRI, but without considering different test thresholds for biopsy or reasonable choices among MRI schedules (13). Sathianathen et al (13) reported the value of incorporating MRI into AS in men 50 years with low-risk PC but favored a schedule of every 5 years for extending QALE at a lower cost than alternative schedules. Their overall reported LE and QALE were lower than those in our model, which is accounted for at least in part by higher all-cause mortality rates and differing assumptions about testing adherence (13). Our assumptions about all-cause mortality in this population were on the basis of prospective data regarding AS, in which men were less likely to be afflicted by multimorbidity and were therefore more appropriate as candidates for surveillance versus as candidates for watchful waiting (26). Still, a consistent conclusion was that performing biopsies less frequently on the basis of MRI results is favorable in terms of quality of life and economic considerations. Patel et al (68) reported that requiring a biopsy during AS after obtaining prostate MRI findings was more cost-effective than performing a transrectal US-guided biopsy without the use of MRI. Our results suggest the use of a PI-RADS score of 4 as the threshold for biopsy on an annual basis rather than the PRIAS schedule shows comparatively favorable value because of the risk of undetected high-grade tumor.

Our study had limitations. The limitation of a simulation modeling approach included simplifications of the disease course and the inherent uncertainty in assumptions about the natural history of PC. The model representation of natural history was biologically plausible and was on the basis of AS trials and large observational series. Sources were selected for applicable populations and quality of design. However, it was possible for the primary literature to have been affected by patient selection or by differences in variables such as the reference standard in prostate MRI studies. The impact of such uncertainty was tested by using extensive sensitivity analysis. In sensitivity analysis, sensitivity and specificity were likely correlated but the literature did not provide sufficient evidence for a summary receiver operating characteristic curve for PI-RADS scores in AS. Although explicit differences in imaging techniques were not directly represented, the potential impact of these differences was indirectly addressed by varying both sensitivity and specificity across a plausible range to view the impact on both effectiveness and cost. In addition, we assumed that MRI performance in each year was conditionally independent, although even with an interval of a year or more there may not be true conditional independence. Sensitivity and specificity may possibly increase on the basis of prior information (eg, a new lesion with a PI-RADS score of 4 in which previous MRI scans showed negative results or a stable lesion with a PI-RADS score 3 with a prior negative biopsy result). The model assumed conservatively that the test characteristics did not further improve with prior information. Future clinical studies may assess changes in test performance over time during AS with MRI. Finally, the model did not allow specific selection of the comorbidity burden, although such a feature would support more personalized decision-making. Instead, we modeled the lower comorbidity-related mortality reflected in patients in AS trials and then varied this assumption in the sensitivity analysis. Although decision-modeling results did not enable individual-level prediction of the LE or diagnostic yield, the risk of tumor-grade misclassification at baseline, as presented, may be considered as a general guide for the selection of an AS protocol. A set of strategies on the basis of diagnostic efficiency can be considered along with patient-level preferences and, in general, more frequent biopsies, especially without MRI. A lower MRI-result threshold for biopsy did not improve the LE or QALE.

In conclusion, annual MRI surveillance options resulted in life expectancy (LE) and quality-adjusted LE (QALE) benefits compared with less frequent prostate MRI or with active surveillance (AS) without prostate MRI. A Prostate Imaging Reporting and Data System (PI-RADS) version 2 score of 4 as a biopsy criterion at annual MRI for AS of low-risk prostate cancer resulted in the performance of substantially fewer biopsies than the use of a PI-RADS score of 3 and demonstrated an essentially equivalent LE and QALE. The use of the Prostate Cancer Research International Active Surveillance schedule resulted in a slightly lower LE and QALE, although the performance of far fewer biopsies may be appealing to some stakeholders for diagnostic efficiency. Across a wide age range, the use of annual MRI with a PI-RADS score of at least 4 as the biopsy threshold was the most cost-effective strategy. The use of a PI-RADS score of at least 4 as a biopsy criterion remained favorable across a wide range of sensitivity, specificity, and risk levels for tumor-grade misclassification at the time of treatment selection. For men over age 70 years, watchful waiting resulted in a slightly lower QALE than surveillance, a difference driven by the quality-of-life harms from procedural complications.