Exploring the use of a new risk grouping to assess ‘over-treatment’ for the National Prostate Cancer Audit
Melanie Morris, NPCA Lead Epidemiologist
The NPCA’s aim is to stimulate and support local and national quality improvement interventions and – through annual re-auditing – assess the impact of these interventions. A key component is the ‘outlier process’, whereby the NPCA identifies where improvement is required.
Appropriate performance measures are essential when trying to understand the quality of care received by patients and how this varies by hospital. Many men experience treatment-related side effects after radical surgery or radiotherapy for localised prostate cancer including detrimental impacts on urinary, bowel and/or sexual functioning. To help us to understand the impact of these side-effects after radical treatment, the NPCA has developed ‘treatment-outcome’ measures that are meaningful and relevant to patients and clinicians .
Over the years, the NPCA team have developed various indicators, validating and testing their use before integrating them into the audit itself. Now we are considering the implications of moving in our analyses from the use of our current three-tiered risk stratification system to one with five tiers. In our short report, published in February 2021, we lay out the impact this would have on one area of the audit’s reporting: the assessment of the potential ‘over-treatment’ of men with low-risk disease.
What is the current classification system and how is it used?
The three-tiered system of low-, intermediate- and high-risk/locally advanced groups is advocated in the 2019 NICE guidelines and the NPCA has used a similar system, based on a modified D’Amico classification, for several years. We take the anonymised data on each man’s PSA level, Gleason score and the size of his tumour (gathered by each Trust or Health Board, and provided to us by our data partners), and combine them in an established algorithm to assign him to one of these three risk groups or to metastatic disease status.
We use the stratification to restrict our analyses to the relevant men: for instance, we examine the proportion of men with high-risk/locally advanced disease receiving radical prostate cancer therapy. This gives us a measure of the potential ‘under-treatment’ of men who should be eligible for this treatment.
We also use the risk groups to examine the proportion of men classified as having low-risk disease who receive radical treatment and thus the potential for ‘over-treatment’ in this group. These proportions have been coming down steadily over recent years: from 12% in 2016 to 4% in this year’s and last year’s Reports. There is a range across providers, however, with some still as high as 16%. The short report examined how the new five-tiered stratification system would impact these estimates of ‘over-treatment’.
What is the new system?
Several risk stratification tools exist which use the same clinical measures as our current three-tiered system and studies have shown that these have differing abilities to predict prostate cancer mortality and other outcomes (1-3). A five-tiered risk stratification system better describes the range of disease presentation and more closely predicts outcomes for patients, for example by splitting the intermediate group into ‘favourable’ and ‘unfavourable’ prognoses. Five-tiered systems are already used in the USA and one has recently been developed using ‘real world’ UK registry data by Vincent Gnanapragasam and colleagues called the Cambridge Prognostic Group (CPG) classification (4), the only system derived in the UK. It has now been tested in over 80,000 men across three countries, and Matt Parry and other members of the NPCA team collaborated with Vincent to explore its relevance to the NPCA (5).
CPG1 encompasses more men than the existing low-risk group, including some men who previously would have been in the intermediate-risk group: it comprises men who have a Gleason score of 6 and a PSA <10ng/ml and a tumour at stage T1 or T2.
Another major difference is that Gleason score 7 is subdivided in the CPG, picking up the difference, for instance, between one made up of 3+4 (more grade 3 cells than grade 4 cells, more ‘favourable’) versus one made up of 4+3. This difference is reflected in the split of CPG2 and CPG3, which are both within the existing intermediate-risk group.
The CPG also has two tiers which correspond to the current high-risk group, dividing men with higher grade prostate cancer (Gleason score 9-10) or a stage T4 tumour, or those with a combination of Gleason score 8, PSA >20ng/ml or Stage 3 (making up CPG5), from men with only one of these last features of their cancer, who generally have better prognosis (CPG4).
Why might the CPG be preferable for use in the audit assessment of ‘over-treatment’?
The ability to subdivide our cohorts of men more precisely is beneficial to the analysis of trends in treatment. Focusing on our assessment of the ‘over-treatment’ indicator, we wanted to examine what impact using the CPG classification would have by comparing it to using the three-tiered risk grouping for the same cohort of men.
Importantly, CPG1 included more men: 15% of men in the cohort compared to 6% into the low-risk group. By their clinical characteristics, all of these men would have been eligible for active surveillance, rather than radical treatment. So, by using the CPG classification we not only have a more robust measure due to larger numbers, but one that encompasses more clinically relevant patients for the indicator.
In our comparison, 4% of men classified as having low-risk disease had radical treatment, compared to 10% of those classified in the CPG1 group. The socio-demographic characteristics of men in both of these lowest risk groups were similar to each other, whether they received radical treatment or not, showing that the differences in treatment receipt did not originate there.
By including more eligible men in the indicator, we can also better highlight for providers where they might need to examine practice and make changes. Using CPG1, we found more variation was apparent between providers, and five were highlighted as being outside the expected range of values, whereas only one (a different one) was outside these limits when restricting to the low-risk group.
What are the possible challenges?
We rely on the completeness and accuracy of the data we receive to make these classifications and so the audit, as always, is only as good as the data we put in. We know that during the pandemic data gathering and collating has been extra challenging and we may need to produce our next audit report on the basis of reduced datasets, but going forward we expect to be able to use the full and complete data as before, allowing the CPG classification to be implemented.
We also know that we still need to consider men whose disease is node positive (who were previously included in the high-risk/locally advanced group, but are not encompassed by the CPG classification) and men with metastatic disease. If we adopt the CPG classification throughout the audit, the treatments given to these groups of men and their outcomes, will continue to be monitored separately.
Finally, we know that any post-hoc classification system may not exactly reflect clinical decisions that were made at the time; these, of course, will consider various factors beyond the clinical characteristics. We don’t pretend to advise here about the utility of the different systems for clinical practice, but we are heartened to hear anecdotally that the five-tiered CPG system is being used more and more by clinicians, and that NICE are conducting a review into including it in their guidelines.
We wouldn’t expect the radical treatment of men with the lowest risk prostate cancer to ever reach zero, as patient choice and other considerations will always lead to some men receiving that treatment. But these risk stratification systems are our best resource for judging the appropriateness of treatments that the data show us were given to those men, and the CPG classification appears our best bet to do that robustly and for a wide range of indictors.
(1) Zelic R, Garmo H, Zugna D, et al. Predicting Prostate Cancer Death with Different Pretreatment Risk Stratification Tools: A Head-to-head Comparison in a Nationwide Cohort Study. Eur Urol. 2020 Feb;77(2):180-188.
(2) Gnanapragasam VJ, Barrett T, Thankapannair V, et al. Using prognosis to guide inclusion criteria, define standardised endpoints and stratify follow-up in active surveillance for prostate cancer. BJU Int. 2019;124(5):758-67.
(3) Gnanapragasam VJ, Lophatananon A, Wright KA, Muir KR, Gavin A, Greenberg DC (2016) Improving Clinical Risk Stratification at Diagnosis in Primary Prostate Cancer: A Prognostic Modelling Study. PLoS Med. 2016. 13(8): e1002063.
(4) Gnanapragasam VJ, Bratt O, Muir K, et al. The Cambridge Prognostic Groups for improved prediction of disease mortality at diagnosis in primary non-metastatic prostate cancer: a validation study. BMC Med. 2018;16(1):31.
(5) Parry MG, Cowling TE, Sujenthiran A, et al. Risk stratification for prostate cancer management: value of the Cambridge Prognostic Group classification for assessing treatment allocation. BMC Med. 2020;18(1):114.
Last updated: 27 April 2022, 8:42am