Diagnostic pathway

We mapped the current strategy (biopsy all) and the proposed diagnostic strategy for men with normal DRE and PSA levels 4–10 ng/ml using information supplied by Hong Kong clinicians (JT, PC and CFN). For the proposed strategy, we estimated the costs and consequences of three different cut-off levels for the test. These pathways were set out as decision trees with a time horizon covering the diagnostic process up to biopsy (see Fig 1). A Hong Kong public health service perspective was adopted. Health outcomes differ between the strategies due to the direct impact of the biopsy and adverse effects following a proportion of biopsies and the proportion of cancer cases that are missed. Health outcomes relating to the biopsy procedure itself or complications of the procedure are not considered in this study but they would be positive under each of the testing strategies as they reduce as the number of biopsies reduces. We calculate and present the proportion of missed cancers which would likely result from each testing strategy as well as the costs of the alternative strategies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Decision tree illustrating diagnostic strategies for men with suspected prostate cancer, normal digital rectal examination and prostate specific antigen level 4–10 in Hong Kong.

DRE–Digital Rectal Examination, PHI–Prostate Health Index, PSA—Prostate Specific Antigen, TRUS–Transrectal Ultrasound-guided biopsy. Diamond represents a decision node. Circle represents a probability node.

https://doi.org/10.1371/journal.pone.0215279.g001

Fig 1 shows the current strategy under which all patients undergo TRUS biopsy and the proposed testing strategy. In the proposed pathway (the lowest branch of the decision tree) all patients are tested using PHI. Those with a positive result undergo TRUS biopsy. Those with PHI scores under the threshold (negative result) do not undergo biopsy. In both positive and negative test result arms there is the possibility of the test result being correct or incorrect resulting in missed cases of cancer as well as a proportion of unnecessary biopsies.

Data sources

Clinical data to populate the decision trees and costing data for all costs apart from the PHI test were collected as part of the study reported by Chiu and colleagues [2] (see Table 1 for inputs to the model). Sensitivity and specificity for three different thresholds of PHI score are shown in Fig 2. Any accident and emergency attendances and length of hospital stay (where appropriate) following biopsy were recorded. This resource usage was valued using costs from the Annual Report for 2016–7 of the Hospital Authority [6]. The PHI test cost was based on the cost of the test at a Hong Kong clinic to a private user. Costs are not discounted given the short time horizon of the economic evaluation. All data were fully anonymized before access. The ethics committee waived the requirement for informed consent in this study.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Inputs to the model and data sources.

https://doi.org/10.1371/journal.pone.0215279.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Test performance of Prostate Health Index at cut-off levels of 25, 35 and 55.

Figures in boxes are the cut-off levels for the Prostate Health Index test from Chiu et al study [2]. Sensitivity/specificity for any grade cancer at different cut-off levels are 88.7%/36.5% at 25, 61.3%/77.5% at 35 and 12.9%/97.4% at 55.

https://doi.org/10.1371/journal.pone.0215279.g002

Sensitivity and threshold analysis

One-way sensitivity analysis was undertaken whereby input parameters were varied in turn by 50% of base case value in both directions (whilst all other inputs to the model were held constant) to determine the impact of an over or under-estimation on the base-case results. This form of sensitivity analysis was undertaken as this allowed decision makers to assess the individual impact of each of the input parameters. Where sensitivity analysis showed that the result was sensitive to an individual parameter threshold analysis was undertaken. This involved varying the range to determine the highest level at which the cost savings would be reduced to zero.

Summary of findings

This study found that the adoption of the PHI test for patients with negative DRE and PSA score 4–10 ng/ml in Hong Kong has the potential to deliver significant cost savings although there are implications with a proportion of all grade cancers missed. The cost savings arise because the PHI test stratifies men into those requiring TRUS biopsy and those who can avoid biopsy and enter the monitoring programme. As biopsy and the adverse events associated with it are expensive to deal with, a strategy which avoids a relatively small proportion of biopsies has the potential to deliver savings which exceed the costs of testing all the patients in this population but this must be balanced with the risks of missed cases and the longer term cost and clinical outcomes. The cost savings are sensitive to the cost of the test and the biopsy as well as the specificity of the PHI test in this population. This information is useful to decision-makers as the costs of the biopsy are known, the cost of the test would be known prior to introduction and cost savings are likely even at relatively low level of specificity (break-even level at 24% specificity compared to the base case of 78%). This information allows future evidence generation and decision-making to focus on the consequences of introducing PHI rather than the costs.

At the base case cut-off of 35, the study data [2] indicated that cancer in 4.2% of the population (including 0.53% with high grade cancers) may be missed. However, applying a cut-off of 55 to study data, prostate cancer in 9.5% of the population would have been missed including 2.1% with high grade cancer. If the cut-off is reduced to 25 around a third of biopsies could be avoided with just over 1% of all cancer cases missed including less than 0.2% of high-grade cancers.

How these results compare to previous studies

The costs results are broadly consistent with three previous economic evaluations of PHI. The first study by Nichol et al [7] was a budget impact analysis of PHI plus total PSA and percent free PSA compared to PSA alone. This study evaluated the impact on 1 year total costs of PHI plus PSA to PSA alone in a screening programme from a US societal perspective in men 50–75 years old. Using thresholds for PHI testing of 2ng/ml and 4ng/ml they estimated cost savings of $356,647 and $94,219 respectively in a notional insurance company cohort of 100,000 men. 90% of the overall savings came from avoiding unnecessary biopsies. A further study by Nichol et al [8] extended their previous analysis to a cost-utility analysis with a 25-year time horizon. This extended analysis found that PHI plus PSA dominated PSA alone strategy for both 2ng/ml and 4ng/ml thresholds delivering cost savings of $1,199 and $443 respectively together with utility gains of 0.08 and 0.03. Both Nicholl et al studies used data relevant to the US population and are not directly applicable to a Chinese population. The final economic evaluation study identified was Heijnsdijk et al [9] who assessed the cost-effectiveness of using a PHI cut-off of 25 as an add-on to PSA with a cut-off of 3ng/ml in a European screening population aged 50–75. This study found a reduction in negative biopsies of 23%, a reduction of 17% in costs of diagnosis and 1% in total cost of prostate cancer.

A recent study [10] extended the Hong Kong cohort from Chui et al [2] with cohorts from Taiwan, Singapore and Shanghai to make a total of 1,150 men. This later study found higher prevalence of all PCa (13.1% compared to 10.9% in this study) and high grade PCa (5.7% compared to 2.8% in this study). Introduction of a test such as PHI into a population where there was a higher prevalence would mean that less biopsies could be avoided for the same risk of false negatives.

Nicholl at al [7] suggests that the negative consequences of missed cases are limited as they are likely to be found in subsequent screenings [11]. Effectively false negatives represent delayed rather than missed diagnoses. Moreover, cancers missed tend to have relatively low Gleason scores and most cancers found 2–4 years after an initial screen are still curable [12–18]. It is proposed that men below the PHI score of 35 would undergo an annual PSA test until the age of 78 when mean survival from PCa exceeds life expectancy [19,20]. However, the extent and grade of false negatives would require further study.

Strengths of this study

This study is a preliminary cost-consequence analysis that indicates that the PHI test is potentially cost-effective in that it is not “dominated” by current practice (i.e. costs more with worse outcomes) [21]. This represents a necessary but not sufficient condition for cost-effectiveness. As far as we are aware, it is the first study to examine the potential economic impact of introducing the PHI test in an Asian population.

The study demonstrates the use of simple economic evaluation in a preliminary assessment of a diagnostic technology using local data. Another strength of the study is that it was relatively quick and resource light as a result of the simplicity of the model and the availability of locally relevant data from a previous study. Evidence was taken from a single clinical study [2] and micro-costings from a single hospital. One-way sensitivity analysis was appropriate as it allowed decision makers to assess the importance of individual parameters.

Limitations

A significant limitation of this study is that data have been taken from a retrospective analysis of a cohort taken from a single clinical study [2].

A simplifying assumption was made that all men in the population currently undergo TRUS biopsy. In a proportion of cases patients and clinicians decide that biopsy is not their preferred option. In order to change the conclusion of our base case analysis (at a PHI cut-off of 35) just under 50% of men would need to refuse biopsy. We believe the proportion of men not undergoing biopsy is substantially lower than this.