This study has demonstrated that the five-tiered CPG classification distributes men in groups that are about equal in size. The subdivision of men with intermediate-risk disease into CPG2 and CPG3, and of men with high-risk disease into CGP4 and CPG5, gives a finer and more clinically relevant degree of granularity compared to the traditional three-tiered risk stratification system. It is therefore more informative for studies, such as the National Prostate Cancer Audit, that aim to evaluate the appropriateness of treatment selection amongst providers of prostate cancer services at the regional or national level.
By using the CPG classification, we also demonstrated the considerable between-hospital variation in radical treatment rates observed for men in CPG2 across England, which indicates a lack of consensus in the management of men with ‘favourable’ intermediate-risk disease (i.e. those in CPG2). A potential explanation for this discrepancy may be due to the differences in the uptake of magnetic resonance imaging techniques and image-guided biopsies. However, we could not explore this further due to a lack of imaging data within the data sources that were available.
The recently updated NICE guidelines, published in 2019, currently advise active surveillance for men with intermediate-risk disease only if they opt not to have radical treatment. It is important to note that the NICE guidelines do not make a distinction between favourable (CPG2) and unfavourable (CPG3) intermediate-risk disease. Without this distinction, a clinical emphasis is placed on treatment and not on surveillance for men in CPG2 [1]. Equally, although the European Association of Urology guidelines, published in 2017, make reference to the subdivision of intermediate-risk disease into those with favourable or unfavourable disease, the inclusion criteria for active surveillance do not include patients with favourable intermediate-risk disease [20].
However, other guidelines, such as those supported by Prostate Cancer UK, do indicate that active surveillance is suitable for men who have favourable intermediate-risk disease and a PSA density of ≤ 0.2 ng/ml2 [21]. North American guidelines also recommend active surveillance as a primary management option for men with favourable intermediate-risk disease. For example, the American Urological Association recommend offering active surveillance to ‘select’ patients but adds the caveat that patients should be informed that this comes with an increased risk of developing metastases without defining by how much [7]. Also, the National Comprehensive Cancer Network recommends active surveillance specifically for men with favourable intermediate-risk disease but adds that less than 50% of biopsy cores should be positive [8].
Using the CPG classification also sheds light on whether men who received radical treatment had surgery or radiotherapy. Radiotherapy appears to be a more frequent treatment for men with more high-risk disease. Eighty per cent of men in CPG5 underwent radiotherapy compared to 59% in CPG4. Men in CPG5 represent those with more locally advanced disease, for whom surgery may not be considered appropriate. The CPG classification reveals the differences in how men with ‘high-risk’ disease are apparently selected for therapy when subdivided into CPG4 and CPG5, which studies that used the traditional three-tiered classification fail to demonstrate [22].
An important observation from our study is the relationship between age and CPG at diagnosis. The proportion of men diagnosed with CPG5 aged 70 or older was double that of those diagnosed with CPG1, indicating a clear progression of prostate cancer stage with age and that age at diagnosis is a major risk factor for aggressive disease. This observation highlights the complexity of the association of age and disease aggressiveness on the one hand and treatment selection on the other [23, 24], given the clinical imperative to avoid early surgical complications in older patients and to reduce the need for later salvage treatments in patients with higher risk disease [25].
We also found that by using the CPG classification, the rates of surgery exceed radiotherapy only in men with CPG1 and CPG2 disease. Interestingly, in a prior publication, we showed an association between being diagnosed at a hospital with surgical services onsite and being more likely to receive surgery for their prostate cancer, than if these services were not available onsite [24].
A key strength of our study is that we used a contemporary national English cohort, ensuring that treatment patterns were representative of the current nationwide practice. Inclusion of more than 60,000 men who could be classified into a CPG enabled a reliable assessment of treatment patterns.
A major limitation of our work is that we were reliant on the accuracy of the clinical coding in the routinely collected hospital data. However, the accuracy of these data has been shown to be high when compared to clinical notes and is sufficiently robust to support its use in research [26]. It must also be noted that the sub-classification of T stage (i.e. T2a, T2b, and T2c) was not available. However, it is unlikely that the use of this further sub-classification would be beneficial given that this level of staging is known to be frequently inaccurate [27, 28].
A further weakness is that we were unable to classify 27% into a CPG due to missing staging data. This demonstrates a trade-off between a better classification system and a higher level of data completeness when using the five-tiered CPG instead of the three-tiered system. Moreover, varying levels of missing data could also introduce bias. For example, the proportion of men who could not be classified into a CPG varied between hospitals. For hospitals that were identified as outliers with higher than expected radical treatment rates for men in CPG2, we could not classify 21.4% into a CPG group, whereas the corresponding figure was 18.7% for hospitals identified as outliers with a lower than expected radical treatment rate.