Evidence review

Clinical and technical evidence

Regulatory bodies

A search of the Medicines and Healthcare Products Regulatory Agency website revealed no manufacturer Field Safety Notices or Medical Device Alerts for this device. No reports of adverse events were identified from a search of the US Food and Drug Administration (FDA) database: Manufacturer and User Device Facility Experience (MAUDE).

Clinical evidence

Of the 19 relevant papers identified, 10 were excluded because they were abstracts, based on intention or examined overlapping populations. Consequently, 9 studies are included in this briefing. Of these, 6 were prognostic value studies, 2 studied clinical utility (the effect of Prolaris results on clinicians' treatment decisions) and 1 was an analytical validation study. In these studies, the Prolaris score is referred to as the cell cycle progression (CCP) score and the previous scoring system is used (scoring range −3 to +7). According to this system, for patients in the D'Amico low risk category, a CCP score below −0.7 indicates that the cancer is less aggressive than the average cancer in this risk category. A CCP score above 0.3 indicates a more aggressive cancer. Similarly, for the D'Amico intermediate risk category, CCP scores below −0.9 and above 0.1 indicate that the cancer is less aggressive and more aggressive respectively than the average cancer in this risk category.

Prognostic value

Bishoff et al. (2014) studied the prognostic utility of CCP scores generated from tissue samples in 582 men who had had radical prostatectomies in 3 patient cohorts (2 in the US and 1 in Germany). The score was derived from a diagnostic or simulated biopsy (taken randomly from a post-operative, FFPE tumour block) that was analysed at Myriad Genetics. Time to biochemical recurrence (BCR) and time to metastasis were measured. Combined analysis of all patients showed that the CCP score was a strong predictor of biochemical recurrence; the hazard ratio was 1.60 (95% confidence interval [CI] 1.35 to 1.90; p=2.4×10−7) by univariate analysis and 1.47 (95% CI 1.23 to 1.76, p=4.7×10−5) by multivariate analysis. Similarly, a combined analysis in 12 men with metastatic prostate cancer showed that the CCP score was predictive of metastatic disease. The hazard ratio was 3.35 (95% CI 2.89 to 9.92; p=2.1×10−8) by univariate analysis.

Cooperberg et al. (2013) aimed to validate the use of the CCP score to predict radical prostatectomy outcomes in 413 men in the US. The study assessed the CCP score for prognostic utility and generated prediction models based on CCP only, the CAPRA-S score, and the combined CCP and CAPRA-S score. The hazard ratio of the CCP score was 2.1 (95% CI 1.6 to 2.9, p<0.001) and after combination with the CAPRA-S score it was 1.7 (95% CI 1.3 to 2.3, p<0.001). The CCP score correlated weakly but significantly with the CAPRA-S score (r=0.21, p<0.001). When the CCP score and CAPRA-S variables were combined (to provide the patient's 10-year biochemical recurrence risk), a decision curve analysis demonstrated that the combined model was more predictive than CAPRA-S alone.

Cuzick et al. (2011) assessed the prognostic value of the CCP score in 2 cohorts (1 from the US and 1 from the UK) of patients with prostate cancer. Patients from the US cohort had had radical prostatectomies (n=410). The UK cohort were patients with clinically localised prostate cancer diagnosed following transurethral resection of the prostate (TURP) and managed by watchful waiting, randomly selected from 6 registries (n=337). Patients who had radical prostatectomies were evaluated for time to biochemical recurrence and patients who had TURP were evaluated for time to death. Median follow-up time was 9.4 years for the radical prostatectomy group and 9.8 years for TURP group. Hazard ratios showed that the CCP score was predictive of outcomes in both cohorts. After radical prostatectomy, the CCP score alone was useful for predicting biochemical recurrence as assessed by the univariate analysis, and in combination with tumour and patient data, as assessed by multivariate analysis. The hazard ratios were 1.89 (95% CI 1.54 to 2.31; p=5.6×10−9) and 1.77 (95% CI 1.40 to 2.22; p=4.3×10−6) respectively. In the TURP cohort, the CCP score was the most important variable for prediction of time to death from prostate cancer in both univariate analysis (hazard ratio 2.92, 95% CI 2.38 to 3.57, p=6.1×10−22) and multivariate analysis (hazard ratio 2.57, 95% CI 1.93 to 3.43; p=8·2×10−11).

Cuzick et al. (2012) examined the prognostic value of the CCP score compared with other variables, including the Gleason score, PSA level and clinical stage, in a cohort of 349 patients who had conservatively treated localised prostate cancer which had been diagnosed by needle biopsy. The median CCP score was 1.03 (interquartile range from 0.41 to 1.74) and was associated with a 2.02‑fold increase in risk of cancer-related mortality in the univariate analysis (χ2=37.6, 95% CI 1.62 to 2.53, p=8.6×10−10). The risk of death from prostate cancer at 10 years after diagnosis was associated with an increased CCP score. For example, for a CCP score of less than 0, the estimated rate of death from prostate cancer was 19.3%. For a CCP score greater than 3, the estimate was 74.9%. The CCP score was a stronger prognostic factor than the Gleason score or PSA level. Multivariate analysis hazard ratio for CCP score was 1.65 (95% CI 1.31 to 2.09, p=2.6×10−5).

Cuzick et al. (2015) assessed the prognostic value of the CCP score in predicting the 10‑year risk of cancer-related mortality, both independently and in combination with standard clinical variables used to determine the CAPRA score (such as Gleason score, PSA level and clinical stage). A cohort of patients (n=761) with clinically localised prostate cancer diagnosed by needle biopsy was selected from 3 UK registries. Using univariate analysis, a 1-unit increase in CCP score was associated with a hazard ratio of 2.08 (95% CI 1.76 to 2.46; p=6.0x10−14). Using multivariate analysis, the CCP score hazard ratio was 1.76 (95% CI 1.44 to 2.14; p=4.2x10−8), whereas the CAPRA score hazard ratio was 1.29 (95% CI 1.18 to 1.42; p=4.6x10−9). The CCR score (combination of CCP and CAPRA scores) was most predictive of cancer related mortality, with a hazard ratio of 2.17 (95% CI 1.83 to 2.57; p=4.1x10−21).

Freedland et al. (2013) evaluated the prognostic utility of the CCP score in patients with prostate cancer who had external beam radiation therapy less than 2 years after biopsy. The authors analysed time to biochemical recurrence in a US-based population (n=141). The median CCP score was 0.12 and the hazard ratio for biochemical recurrence was 2.55 (95% CI 1.43 to 4.55) for a 1-unit increase in CCP score (p=0.0017). The multivariate analysis had similar results. Freedland et al. (2013) concluded that CCP was a statistically significant predictor of outcome for patients who had external beam radiation therapy and that the test provided greater prognostic information than other clinical parameters.

Clinical utility

Crawford et al. (2014) studied how the CCP score affected clinicians' treatment recommendations for 331 patients diagnosed with prostate cancer after biopsy in the US. The main evaluations were:

  • change in treatments recommended before and after the test (that is, the change between interventional and non-interventional therapy options)

  • the overall direction of change (to a more aggressive or less aggressive treatment).

Most patients had cancers classified as being low (43.5%) or intermediate risk (44.1%) for 10-year cancer related mortality. The average CCP score was −0.69±0.82 with an average risk of 10-year mortality with conservative management of 3.5%. Overall, 65% of clinicians changed their treatment recommendation based on the results of the CCP score. There was a reduction in therapeutic burden in 40% of cases (122/305) and an increase in 24.9% of cases (76/305). The authors concluded that the study demonstrates high clinical utility for CCP scoring among clinicians.

In a prospective registry study, Shore et al. (2016) evaluated how CCP score affected shared treatment decision-making for 1,206 patients with newly diagnosed prostate cancer. Four sequential surveys tracked changes to the initial therapy: before the initial CCP (Prolaris) test; after clinical review of the CCP score; after shared clinician/patient review of the test results; and after at least 3 months of clinical follow-up (actual treatment). There was a significant reduction in the treatment burden recorded at each successive evaluation (p<0.0001). The mean number of treatments per patient decreased from 1.72 before the CCP score was determined to 1.16 in clinical follow-up. The CCP score resulted in a change in treatment in 47.8% of patients. Of these changes, 72.1% were reductions and 26.9% were increases in treatment burden, measured as the total number of treatment options recommended or administered per patient. For each clinical risk category there was a significant change in treatment modality (intervention versus non-intervention) before the CCP test compared with after CCP testing (p=0.0002). The authors concluded that the CCP score had a significant impact in helping clinicians and patients to reach shared treatment decisions.

Analytical validation

One study demonstrated the analytical validity of the CCP (Prolaris) score. Warf et al. (2015) examined the precision of the CCP score, the stability of stored RNA, the yield of RNA extraction (from FFPE tissue), the linearity of the score (in relation to RNA concentration), the amplification efficiency of genes within the CCP score and the dynamic range over which this gene expression signature could produce valid CCP scores in both prostatectomy and needle biopsy samples. The authors concluded that the CCP score is reproducible and robust, its linear and dynamic range exceeds the parameters utilised in the clinical setting (indicating that it is suitable for use) and it is analytically validated for use on FFPE prostate biopsy samples and radical prostatectomy specimens.

Recent and ongoing studies

Two ongoing or in-development trials using Prolaris were identified in the preparation of this briefing.

  • NCT02209584 is a US-based open registry with the aim of measuring the impact of Prolaris on treatment decisions after biopsy in newly diagnosed prostate cancer patients. It is sponsored by the manufacturer and was expected to be completed in September 2015.

  • NCT02454595 is a US-based open registry with the aim of measuring the impact of Prolaris in selecting first-line therapy for newly diagnosed, treatment-naive patients with early-stage localised prostate cancer. It is sponsored by the manufacturer and is estimated to be completed in November 2016.

Costs and resource consequences

Two abstracts (Crawford et al. 2015, de Pouvourville 2015) providing economic evidence on Prolaris were identified. Crawford et al. (2015) quantified the economic impact of the CCP (Prolaris) test in the US healthcare setting using a hypothetical cohort of patients with localised prostate cancer (of all risk types) over 10 years. Management and progression assumptions were made based on published clinical data and interviews with clinicians (Crawford et al. 2015). The study found that using the CCP score over 10 years reduced per patient costs by about £1,938. The authors concluded that the savings were a result of increased use of active surveillance in low- and intermediate-risk patients with less aggressive disease, as well as reduced progression rates in high-risk patients. De Pouvourville (2015) evaluated the cost effectiveness of using the CCP score in France using a Markov model. They compared the treatment decisions based on diagnosis with and without the CCP score in patients with localised low-risk prostate cancer. Direct medical costs were calculated from public data sources. The study found that in the long term (the time period was not defined in the abstract), using the test at a hypothetical price of £1,502 was a dominant strategy, with a lower limit lifetime discounted cost of £1,284 and an incremental discounted quality-adjusted life-year gain of 0.23.

If the adoption of Prolaris led to more accurate risk stratification, it could avoid the need for chemotherapy in some patients. The use of Prolaris will not require changes in the organisation or delivery of current services, and no additional facilities or technology will be needed. Sample preparation requirements are exacting and will require pathology resources to enable the test to be used. The product is not currently used in the NHS but is used in UK private practice.

Strengths and limitations of the evidence

The evidence for clinical validity and prognostic value of Prolaris is based on the retrospective analyses of archived material, mainly from registries. The exception to this is Cooperberg et al. (2013), who collected specimens prospectively and then used retrospective blinded evaluation design for validation (Pepe et al. 2008).

Crawford et al. (2014) and Shore et al. (2016) prospectively examined clinicians' treatment decisions after receiving the CCP score for their patients. Clinical utility studies should show that changes in treatment ultimately translate to benefits for patients but clinical effectiveness outcomes to validate clinical utility results were not included in the 2 studies. Questionnaires administered after a clinical procedure may introduce the risk of recall bias, because clinicians' recollections of how they planned to manage individuals' care before receiving the test results may have been skewed by the results themselves. Crawford et al. (2014) and Shore et al. (2016) eliminated the risk of recall bias by doing a pre-test survey to assess how the clinicians planned to manage their patients' care, as well as a post-test survey done after the clinicians saw the CCP results.

The UK cohorts in Cuzick et al. (2011, 2012 and 2015) included patients with high-risk prostate cancer who were conservatively managed, for example by watchful waiting, which is not representative of current prostate cancer treatment in the NHS.

Cuzick et al. (2012 and 2015), Bishoff et al. (2014) and Freedland et al. (2013) used biopsy samples to evaluate the CCP test. In the study by Cooperberg et al. (2013), samples were taken from the largest tumour area of prostatectomy specimens which may have limited the heterogeneity of the sampled tissue leading to biased results. None of the studies explicitly stated the number of biopsy cores assessed and therefore the effect of tumour heterogeneity cannot be discounted.

None of the studies discussed the power calculations used to justify their sample sizes. The patient cohorts were mainly based in the US. Three patient cohorts were taken from UK registries, and there was 1 German cohort. Incidence rates and standard treatment for prostate cancer vary by country, which may limit the generalisability of the results to the UK population. Most studies had relatively large sample sizes, ranging from 141 to 761 patients. In contrast, the metastasis group in the Bishoff (2014) study included only 12 patients with metastatic cases of prostate cancer. Low sample sizes can bias the power and reliability of statistical findings.

The generalisability to the NHS of the studies by de Pouvourville (2015) and Crawford et al. (2014) is limited due to their respective settings (France and the US), as well as the lack of detail about which specific costs were used in the calculations and how the models were constructed. The Crawford et al. (2014) study is inherently limited because it used a hypothetical cohort, which assumes that the likelihood of advancing to a different health state is homogeneous across the population (independent of time or past health states). Additionally, the hypothetical cost used by de Pouvourville (2015) was slightly lower than the actual cost of the Prolaris test (£1,502 compared with £1,800 respectively), which could affect the results of the cost analysis.

All of the clinical studies presented in this briefing received some or all of their funding from the manufacturer. All 9 publications included authors employed by Myriad Genetics. The manufacturer's involvement in these publications introduces the potential for bias in reporting the outcomes. The 2 economic abstracts did not mention a funding source.