Michael B. Nichol, Department of Clinical Pharmacy and Pharmaceutical Economics and Policy, University of Southern California, 3335 S. Figueroa Street, Unit A, Los Angeles, CA 90007, USA. e-mail: email@example.com
Study Type – Diagnostic (cost effectiveness)
Level of Evidence 2b
What's known on the subject? and What does the study add?
The Beckman Coulter prostate health index (phi) was developed as a combination of serum prostate specific antigen (PSA), free PSA and a PSA precursor form [−2]proPSA to calculate the probability of prostate cancer and was used as an aid in distinguishing prostate cancer from benign prostatic conditions for men with PSA test 2–10 ng/mL and non-suspicious digital rectal examination. Phi has been shown to improve diagnostic accuracy in prostate cancer detection compared with total and free PSA. An earlier 1-year budget impact analysis revealed it to be a complementary approach to current prostate cancer screening strategies.
The current study evaluated the cost-effectiveness of early prostate cancer detection with phi in combination with a PSA test compared with a PSA test alone from the US societal perspective. The model with over 25 annual screening cycles for men aged 50–75 years indicated that PSA plus phi dominated the PSA test alone in prostate cancer detection and consequent treatment. PSA plus phi may be an important strategy for prostate cancer detection.
• To evaluate the cost-effectiveness of early prostate cancer detection with the Beckman Coulter Prostate Health Index (phi) (not currently available in the USA) adding to the serum prostate-specific antigen (PSA) test compared with the PSA test alone from the US societal perspective.
PATIENTS AND METHODS
• Phi was developed as a combination of PSA, free PSA, and a PSA precursor form [−2]proPSA to calculate the probability of prostate cancer and was used as an aid in distinguishing prostate cancer from benign prostatic conditions for men with a borderline PSA test (e.g. PSA 2–10 ng/mL or 4–10 ng/mL) and non-suspicious digital rectal examination.
• We constructed a Markov model with probabilistic sensitivity analysis to estimate expected costs and utilities of prostate cancer detection and consequent treatment for the annual prostate cancer screening in the male population aged 50–75 years old.
• The transition probabilities, health state utilities and prostate cancer treatment costs were derived from the published literature. The diagnostic performance of phi was obtained from a multi-centre study. Diagnostic related costs were obtained from the 2009 Medicare Fee Schedule.
• Cost-effectiveness was compared between the strategies of PSA test alone and PSA plus phi under two PSA thresholds (≥2 ng/mL and ≥4 ng/mL) to recommend a prostate biopsy.
• Over 25 annual screening cycles, the strategy of PSA plus phi dominated the PSA-only strategy using both thresholds of PSA ≥2 ng/mL and PSA ≥4 ng/mL, and was estimated to save $1199 or $443, with an expected gain of 0.08 or 0.03 quality adjusted life years, respectively.
• The probabilities of PSA plus phi being cost effective were approximately 77–70% or 78–71% at a range of $0–$200 000 willingness to pay using PSA thresholds ≥2 ng/mL and ≥4 ng/mL, respectively.
• The strategy PSA plus phi may be an important strategy for prostate cancer detection at both thresholds of PSA ≥2 ng/mL and PSA ≥4 ng/mL to recommend a prostate biopsy compared with using PSA alone.
European Randomized Study of Screening for Prostate Cancer
prostate health index
net monetary benefit
willingness to pay
quality adjusted life year.
Although screening of men for prostate cancer with the serum PSA test remains controversial , the American Urological Association (AUA) recommends annual screening for men aged 40 and older who have a life expectancy of at least 10 years, after receiving information regarding potential risks and benefits of prostate cancer screening . The 2010 update of American Cancer Society prostate cancer early detection guidelines emphasize the importance of informed or shared decision making regarding screening options . Two large prospective ongoing randomized controlled trials are collecting data for evaluating the efficacy and effectiveness of prostate cancer screening; however, preliminary data published in 2009 showed inconsistent results. The US Prostate, Lung, Colorectal and Ovarian (PLCO) study demonstrated no benefit to screening , whereas the European Randomized Study of Screening for Prostate Cancer (ERSPC) demonstrated a 20–31% mortality reduction in the screened population, with an associated high risk of over-diagnosis and over-treatment [5,6]. A study conducted by Wever and colleagues  comparing PSA screening for prostate cancer between the US population in the PLCO and the Rotterdam section of the ERSPC suggested that the efficacy of PSA screening in detecting prostate cancer was lower in the USA than in the ERSPC – Rotterdam because of the lower sensitivity of PSA testing in the USA.
The traditional PSA threshold level of 4 ng/mL is widely adopted for further prostate cancer evaluation, including the consideration of biopsy in the USA [2,7]. However, PSA levels lower than 4 ng/mL are sometimes associated with prostate malignancy [8,9], so increased test sensitivity may be useful. Currently, using a 2 ng/mL cutoff for recommending prostate biopsy has been shown to have advantages, although lowering the biopsy threshold may increase detection of ‘indolent’ lesions [8–11].
The value of prostate cancer screening remains uncertain. It may reduce morbidity and mortality if early detection leads to early treatment of curable disease . However, treatment for screening-detected prostate cancer may also produce over-treatment of patients with indolent conditions, and may produce temporary and permanent consequences such as urinary, bowel and/or sexual problems [13–15]. Likewise, biopsies that are found to be negative for cancer often lead to undesirable psychological effects in screened men . To minimize these screening limitations, methods to increase the specificity for clinically relevant prostate cancer may be useful.
Several PSA derivatives or biomarkers and test methods may enhance test specificity, including percent free PSA, complexed PSA, PSA velocity, PSA density, and age-specific PSA ranges . Currently, a precursor form of PSA ([−2]proPSA), measured using the Access Hybritech p2PSA assay, , is being tested for use with Access Hybritech PSA and Access Hybritech free PSA to calculate a prostate cancer risk index, especially for individuals ≥50 years of age with PSA 2–10 ng/mL and non-suspicious DRE [18–20]. A multi-centre clinical study sponsored by Beckman Coulter suggests that the prostate health index (phi) doubled the specificity of prostate cancer detection relative to percent free PSA for individuals with PSA 2–10 ng/mL at 95% sensitivity . A 1-year budget impact analysis of phi indicated that phi as a complementary approach to current prostate cancer screening strategies resulted in healthcare cost savings . The index with [−2]proPSA was recently approved in Europe; however, it is not currently available for commercial distribution in the USA. The purpose of this study is to evaluate the cost-effectiveness of early prostate cancer detection with phi adding to the PSA test compared with the PSA test alone under two PSA threshold values (≥2 ng/mL or ≥4 ng/mL) to recommend a prostate biopsy from the US societal perspective.
MATERIALS AND METHODS
A Markov model was constructed with three health states (no prostate cancer, detected prostate cancer, and death). The individual males started prostate cancer screening at 50 years of age from the no prostate cancer health state, and then could move to prostate cancer or remain in the same health state depending on the PSA testing result. They could also transition to non prostate cancer related death. The individuals in the prostate cancer health state could move to cancer related or non cancer related death, or stay in the detected prostate cancer health state.
Two PSA threshold values (2 and 4 ng/mL) were used for recommending prostate biopsy or an additional reflex test in the model. Individuals have three possible test results.
1The test is negative (PSA < threshold value). We assumed that this result indicates no cancer detection and biopsy is not ordered.
2The test is positive (PSA >10 ng/mL). An individual is referred to a urologist and a biopsy is used to confirm prostate cancer.
3The test is borderline (PSA value is between the threshold value and 10 ng/mL). Two testing strategies (PSA alone vs PSA plus phi) were compared. For the strategy of PSA test alone, an individual is referred to a urologist and repeats a PSA test, and then receives a biopsy for prostate cancer diagnosis. For this model, we assumed the second PSA test result is the same as that in the first test. For the strategy of PSA plus phi, both free PSA and [−2]proPSA would be performed as reflex tests and the phi score could be calculated. Based on the study of phi, a score higher than 25 corresponds to approximately 30% of the weighted average relative risk of prostate cancer. Thus, this value represents a positive test used for a urologist visit referral and subsequent prostate biopsy. Conversely, a phi score of 25 or less is considered test negative, and a prostate biopsy would not recommended.
Individuals with a positive PSA test but without biopsy-confirmed cancer remained at the no prostate cancer health state and continued annual prostate cancer screening. We assumed that missed cancers in previous screenings could only be found in subsequent screenings. Research indicates that most prostate cancers detected at 2–4 years after an initial screen (first round) will be curable [22–27]. Therefore, we assumed that the detected missed cancers from a previous screening have similar clinical characteristics to the cancers detected in regular screening.
The individuals with biopsy-confirmed prostate cancer moved to the detected prostate cancer health state and received treatment, which may be a radical prostatectomy, or radiation, or androgen deprivation therapy as appropriate. We have not explicitly modelled individual treatments, and watchful waiting was not considered in the model because choice of treatment is likely to vary between individual patients, depending on their physician, age, health status and personal preferences regarding potential benefits and adverse effects.
Each cycle in the model was set to 1 year. The model was iterated until the individual reached age 75 years old or died, whichever came first. The screening from age 50 to 75 years was based on three considerations: (i) the US Food and Drug Administration has approved the use of the PSA test to aid in the detection of prostate cancer in men 50 years of age or older ; (ii) several recommendations for prostate cancer screening initiate PSA screening at 50 years of age in average risk men [10,29–31]; and (iii) the United States Preventive Services Task Force recommends against screening for prostate cancer in men aged 75 years or older . Although some practice guidelines recommend screening starting early at age 40, phi is intended to be used for males aged 50 years or older and therefore we did not model individuals starting screening at 40 years old. We also performed two analyses for men aged 50–64 years and 65–75 years. A discount rate of 3% was applied to costs and health utilities. An incremental cost–effectiveness ratio was calculated. A dominant strategy is defined as a strategy that is less expensive but more effective than the other strategy.
DATA SOURCE AND INPUT DATA
The probability of having a positive PSA test was derived from the age-specific prevalence for each PSA range according to the population-based US data . Among individuals who had a positive PSA test result, the probability of prostate cancer upon biopsy was derived from published studies (Table 1). The study of phi provided data for the probability of having a positive test for phi (a score ≥25) and the probability of prostate cancer detection among those biopsied . Since the study was only published for the sample of men with PSA 2–10 ng/mL , data for PSA 4–10 ng/mL were provided by Beckman Coulter based on the additional analysis in a subsample.
Table 1. Model inputs for probabilities and health state utilities
Base-case estimate (sd)
Range used in sensitivity analysis
Note: Standard deviation of the input data was calculated from the formula (high range – low range)/(2 × 1.96).
(Probability of PSA 2–10 ng/mL) = (probability of PSA ≥2 ng/mL) − (probability of PSA >10 ng/mL).
Utility was estimated from linear models.
Probability of PSA test results for different PSA levels
We assumed that the prostate biopsy rate was 50% (10% higher than the biopsy rate in the PLCO trial ) for the base-case analysis in men with PSA testing positive. The prostate cancer related mortalities under two common treatment options (brachytherapy and radical prostatectomy) were derived from a systematic literature review as part of the study conducted by Hayes et al. . Age-specific mortality due to causes other than prostate cancer were taken from 2004 US life tables .
The costs were expressed in 2009 US dollars. PSA test related costs were based on the national 2009 Medicare Fee Schedule (Table 2). Office visit costs included one primary care visit for the individuals with a negative PSA test, or one primary care visit and at least two urologist visits for the individuals who were referred to a urologist. We assumed that a man with a positive PSA test at the primary care visit sees a urologist for the prostate biopsy and then comes back to the urologist at least one time for a follow-up visit.
Table 2. Model input for costs in 2009 US dollars
Base-case estimate, $ (sd)
Range used in sensitivity analysis, $
Costs data (except for prostate cancer related treatment costs) were based on the 2009 Medicare Fee Schedule. Prostate cancer related treatment costs were obtained from available published literature and are presented in 2009 US dollars. The ranges tested in sensitivity analyses were decreasing and increasing 25% of base-case costs.
The cost of [−2]proPSA has been estimated from discussions with Beckman Coulter Inc.
The prostate biopsy cost was the integrated cost from 12-core prostate biopsy, echography-guided for biopsy, transrectal ultrasound, tissue examination by a pathologist, and three immunohistochemistry stains.
Different PSA blood test cost components were applied to the PSA test strategies according to the PSA test result. If PSA < threshold value or >10 ng/mL, one PSA test was included. If PSA is between threshold and 10 ng/mL, PSA test costs were calculated as the following: (i) two PSA test costs (one at the primary care visit and the other at the urologist visit) for the PSA strategy; (ii) one PSA, one free PSA and one [−2]proPSA for the strategy of PSA plus phi. We assumed that a routine urinalysis is performed to identify possible reasons for a positive PSA or phi test. Because the prostate biopsy involved multiple procedures, the cost was integrated from 12-core prostate biopsy, echography-guided biopsy, transrectal ultrasound, tissue examination by a pathologist, and three immunohistochemistry stains.
Prostate cancer treatment costs were derived from published data and expressed in 2009 US dollars. We assumed the same cancer treatment costs for both strategies and applied gross costs for prostate cancer treatment. The PLCO cancer screening trial reported that 95.5% of prostate cancers were classified as stage II at 10 years follow-up . We assumed that most detected prostate cancers would be in stage II under an annual screening strategy without previous history of prostate cancer. Therefore, long-term stage II prostate cancer related costs were apply to the model, which included the costs for 6 months of the initial phase, the last 12 months of terminal phase, and continuous treatment between the initial and terminal phases .
We used weights from a meta-analysis conducted by Bremner et al.  to assign prostate cancer utilities in the model. The meta-analysis incorporated multiple factors including cancer stages, symptoms, disease severity, and scaling methods from 173 different utilities in 23 studies. In the reference case, we applied the utility for a person who experiences mild severity of multi-attribute symptoms from a non-metastatic prostate cancer. Since a procedure of biopsy could result in complications and impacted patients' quality of life [39–41]. We also used prostate biopsy related disutility in the model for individuals undergoing biopsies .
Sensitivity analyses were performed by varying base-case parameters of probability, utility and cost values. We used net monetary benefit (NMB) as the expected outcome, combining the outcomes cost and effect at a specified willingness to pay (WTP) threshold, e.g. $50 000 per quality adjusted life year (QALY) in our models . NMB is calculated using the formula NMB = E*WTP − C, where E represents effectiveness and C represents cost. In the one-way sensitivity analysis, we investigated the NMB effect of changes in individual base-case parameters across possible ranges of value to examine parameter uncertainties. We also conducted a second-order Monte Carlo simulation under the assumption of accordant beta distribution in chance nodes for variables of transition probabilities and utilities, and accordant gamma distribution for prostate cancer treatment cost variables. The key parameters were simultaneously and randomly varied over the appropriate probability distributions using Monte Carlo simulation with 10 000 iterations. Based on the simulation results, we constructed a cost-effectiveness acceptability curve, and also report a plot of the probability that a PSA screening strategy option is cost effective as a function of WTP .
All analyses were performed using the TreeAge Pro 2009 program (TreeAge Software, Williamstown, MA, USA).
Table 3 summarizes the results of the base-case analysis. Over 25 screening cycles from age 50 to 75 years, the mean total costs per man were estimated at $18 051 or $7687 for the PSA alone testing strategy compared with $16 852 or $7244 for the strategy of PSA plus phi depending on the PSA threshold of 2 or 4 ng/mL, respectively. The expected effects were 15.01 or 15.72 QALYs for the strategy of PSA testing alone, and 15.09 or 15.75 QALYs for the strategy of PSA plus phi, for a PSA threshold of 2 or 4 ng/mL, respectively. The strategy of PSA plus phi dominated the PSA test alone for prostate cancer detection and consequent treatment. It was estimated to save $1199 or $443 with an expected gain of 0.08 or 0.03 QALYs for a PSA threshold of 2 or 4 ng/mL respectively.
Table 3. Discounted absolute and incremental costs and effectiveness and incremental cost–effectiveness ratios in base-case analysis for screening age 50–75 years, 50–64 years and 65–75 years
Total cost ($)
Incremental cost ($)
Total effectiveness (QALYs gained)
Incremental effectiveness (QALYs gained)
Cost–effectiveness ratio ($/QALY)
Costs are presented in 2009 US dollars. NMB was calculated given a WTP of $50 000/QALY.
The strategy PSA plus phi was also the dominant strategy for the models of prostate screening from age 50–64 years and age 64–75 years for both PSA thresholds 2 and 4 ng/mL (Table 3).
One-way sensitivity analysis revealed that results are robust to changes in parameter values. Discount rate and screening start and stop age were the most notable variables that significantly affected the expected outcome presented as NMB given a WTP of $50 000/QALY in the models. Figure 1 shows tornado diagrams for the other 10 estimated sensitivity analysis parameters with the largest impact on expected outcome at the decision node. These variables included prostate biopsy rate, probability of prostate cancer related death, and utility of prostate cancer state.
The second-order Monte Carlo simulation analysis showed that PSA plus phi was approximately 69.5% or 70.9% of the total iterations and was the dominant strategy at a given WTP of $50 000/QALY for the PSA threshold 2 ng/mL (Fig. 2A) or 4 ng/mL (Fig. 2B), respectively. The cost-effectiveness acceptability curves showed that the strategy of PSA plus phi had a higher probability of being cost effective than the PSA test alone across ranges of WTP ($0–$200 000/QALY) which were approximately 77–70% or 78–71% for the PSA thresholds 2 or 4 ng/mL, respectively (Fig. 3).
This research presents the costs and benefits of annual prostate cancer screening using different PSA test strategies. Overall, in the base-case, PSA plus phi was a less costly diagnostic strategy to detect prostate cancers compared with the PSA test alone, and effectiveness was slightly higher, regardless of thresholds for recommending a prostate biopsy. The strategy of PSA plus phi can be expected to increase true positive and reduce false positive tests in men aged 50–75 years, which may decrease the costs and disutility related to the prostate biopsy procedure. It should be noted that increased total costs of screening due to additional costs of free PSA and [−2]proPSA added to the original PSA cost could be offset by reducing unnecessary prostate biopsies, if it is adopted by insurance providers.
We noticed that the expected total costs in the model with PSA threshold ≥2 ng/mL were greater, but the expected QALYs were slightly lower than those in the model with PSA threshold ≥4 ng/mL (Table 3). PSA plus phi in the model with PSA threshold ≥2 ng/mL generated higher cost savings (−$1199 vs −$443), more QALYs gained (0.08 vs 0.03) and greater incremental net monetary benefit ($5199 vs $1943) than the model with PSA threshold ≥4 ng/mL compared with the PSA test alone for modelling age 50–75 years. A lower PSA threshold could have more men transitioning from the no prostate cancer state to the detected prostate cancer state than with a higher PSA threshold. Therefore, the model with low PSA threshold would have higher costs in both biopsies and prostate cancer related treatment but lower QALYs gained than the model with higher PSA threshold. Because multiple data sources were used in deriving the prostate cancer probabilities, resulting in overlapping values for PSA 2–10 ng/mL and 4–10 ng/mL, we used a fixed 25% probability of prostate cancer and a fixed biopsy rate of 50% as the base case input in both PSA threshold models. The model with PSA threshold 2–10 ng/mL may overestimate the number of biopsies and prostate cancers. Thus, a direct comparison in costs and effectiveness between the models with the two PSA thresholds should be treated with caution.
Our model has several limitations owing to the availability and uncertainty of the model parameter values, and is based on model assumptions driven by limited data. First, the model may suffer from prostate cancer early detection biases on over-diagnosis and over-treatment which have been criticized in the literature, as the model quantifies benefits from early detection even for prostate cancers which may not progress until death . Second, we assumed that screened cohorts in each screening cycle have an age-specific probability of positive or negative tests, and did not account for the previous year's test result; this may overestimate or underestimate the transition probability of prostate cancer detection and lead to inflated positive or negative results. Availability of consecutive annual screening data could ameliorate this problem. Third, we did not account for prostate treatment side effects, as there are no studies linking screening type with early stage cancer and consequent effects. Lastly, the 2009 AUA PSA Best Practice Policy no longer recommends a single threshold value of PSA which should prompt prostate biopsy . Instead, the prostate biopsy decision should consider PSA and DRE results, as well as other factors in the testing decision. However, we could not consider individual risk factors to identify cancers and consequent treatment due to the lack of data linking these variables.
In conclusion, this decision analysis demonstrated that adding phi to borderline PSA results under both thresholds 2 and 4 ng/mL produced total cost savings and increased effectiveness compared with the PSA alone test. Results were robust across multiple parameter uncertainties. PSA plus phi was the dominant strategy across ranges of WTP ($0–$200 000/QALY gained). However, decision uncertainty surrounding the adoption of PSA plus phi should be considered. Future studies should address the long-term benefits of adding phi, and the relationship between screening strategy and outcomes.
Partial contents of this research have been presented at the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) 16th Annual International Meeting, Baltimore, MD, USA, May 2011. This study was funded by Beckman Coulter Inc.
CONFLICT OF INTEREST
Research funding for this study was provided by Beckman Coulter Inc. as part of an unrestricted research grant to Kaiser Permanente. Dr Nichol has received a research sub-contract from Kaiser Permanente; Ms Wu has been paid from the research grant from Kaiser Permanente in connection with this paper; Mr Denham is an employee of Beckman Coulter Inc. and Ms Huang was an employee of Beckman Coulter Inc. at the time this research was completed. Dr Jacobsen has received a research grant from Beckman. Two coauthors from Beckman Coulter (Denham and Huang) participated in study design and manuscript development. Final approval of the manuscript was the responsibility of the corresponding author.