Assessment of bone metastasis status is key for the management of prostate cancer (PCa) (NICE, 2014a; Fizazi et al, 2015). Clear evidence exists for staging investigations such as isotope bone scintigraphy (BS) in patients with high-risk PCa while precluding its routine use in low-risk disease (NICE, 2014a). However, the utility of BS in intermediate-risk disease is unclear with EAU, AUA and NICE offering conflicting guidance (Heidenreich et al, 2014; NICE, 2014a).

The International Society of Uro-Pathology (ISUP) have recently approved a new PCa ‘Grade Group’ classification system to improve correlation of Gleason grade to biochemical recurrence (Epstein et al, 2016). Concurrently, multiparametric (mp)-MRI has emerged as a crucial tool in local staging of PCa, allowing improved discrimination between T2 and T3 disease compared with clinical nomograms (Turkbey et al, 2013; Lawrence et al, 2014). However, current guidelines on bone staging do not mandate use of mpMRI information or Grade Groups (NICE, 2014a). Therefore, we investigated the ability of mpMRI-staging information or the new Grade Group system to help predict bone metastatic status and refine the use of bone staging investigations.

Materials and methods

A radiology database was searched for all PCa patients undergoing baseline mpMRI prostate and BS from January 2010 to May 2015 in our tertiary centre (study registration CUH/3927). In cases of equivocal BS, the final status was recorded using a combination of clinical follow-up and/or any further radiological investigations. mpMRI was performed on a 3T Discovery MR750-HDx or 1.5T MR450 system (GE Healthcare, Waukesha, WI, USA) with a surface phased-array coil, including standard anatomical and functional diffusion-weighted imaging using multiple b-values, as previously described (Lawrence et al, 2014). All studies were reported by expert uro-radiologists and reviewed in a multi-disciplinary team (MDT) setting.

Gleason score was assessed according to the ISUP 2005 recommendations (Epstein et al, 2005) and recorded alongside the number of positive cores and percentage involvement of tissue. All cases were reported by a specialist uropathologist and reviewed a second time by another uro-patholigist prior to discussion at a specialist MDT. The core with the highest grade was used to devise the Grade Group.

Patients were first categorised to low- (T1–T2a, Gleason 6 and PSA<10 ng ml−1), intermediate- (T2b–c and/or Gleason=7 and/or PSA 10–20 ng ml−1) and high-risk (T3-T4, or Gleason 8–10 or PSA level >20 ng ml−1) according to NICE 2008 guidelines (NICE, 2014b). Patients were subsequently re-categorised according to the new Grade Group system (Epstein et al, 2016) and a novel five stratum Prognostic Risk Grouping system developed in our centre integrating PSA, Grade Group and mpMRI staging (Gnanapragasam et al, 2016; Table 1).

Table 1 Proposed new prostate cancer Prognostic Risk Group criteria taken from Gnanapragasam et al (2016) and using the new Grade Group system (Epstein et al, 2016)

Contingency tables were constructed with expected frequency for bone metastasis and adjusted residuals calculated for each risk system (Supplementary Table S1). Pearson’s chi-squared test was used to examine the differences in observations. Comparison of sensitivities and specificities of bone metastasis between systems was made using McNemar’s test. To compare positive and negative predictive values, we used a generalised score statistic in R 3.1.2 (Leisenring et al, 2000; Stock and Hieschler, 2014). All other statistical analysis was performed in Stata14 (StataCorp LP, College Station, TX, USA).


Four hundred and thirty-eight patients underwent mpMRI and BS. Mean age (±s.d.) was 67.1 years (±6.7) and mean PSA 21.3 ng ml−1 (±48.1). In this cohort, 37 patients had bone metastases (8.4%); Tables 2A and B. The specificity of BS was 86.5%. Fourteen out of 37 (37.8%) men with bone metastases also had evidence of pelvic metastasis on prostate MRI. Using NICE intermediate-risk as a threshold for performing BS, 426 BS would have been performed. Assessing this risk stratification as if it were a diagnostic test demonstrated a sensitivity of 100% but specificity of only 3.0%. Comparisons of other decision models were made against this standard (Table 3).

Table 2A Distribution and bone metastasis status of men categorised according to NICE classification, by clinical parameters alone and with mpMRI-staging information integrated
Table 2B Distribution and bone metastasis status of men categorised by Grade Group score (Epstein et al, 2016) and the proposed Prognostic Risk Groups (Gnanapragasam et al, 2016)
Table 3 Summary statistics for each stratification system and risk group using the stated cutoffs, as if it were a diagnostic test for bone metastasis

Adding mpMRI to current NICE risk groups

The distribution of patients by NICE classification is shown in Table 2A. By these categories, 33 out of 37 men with bone metastases were classified as high risk and 4 as intermediate risk. All 4 intermediate-risk patients had PSA<20.0 ng ml−1 but Gleason 4+3 disease. mpMRI staging re-categorised three of these patients to high risk (Supplementary Figures S1 and S2), improving the sensitivity of high risk for detecting bone metastases to 97.3% from 89.2% and specificity compared with intermediate risk to 24.2% from 3.0% (P<0.01). Using mpMRI-defined high-risk disease as a threshold for BS would have reduced the number of scintigrams performed by 98 (22.4%), with a single missed diagnosis.

Applying the new Grade Group scores to predict bone metastases

Using the new Grade Group scores in isolation, there was a progressive increase in bone metastases detection from 0 out of 43 (0%) for Grade Group 1 to 20 out of 121 (16.5%) for Grade Group 5 (Gleason 9–10); P<0.001, Table 2B. Using BS only for patients with Grade Group scores 3 significantly improved sensitivity to 97.3% and significantly improved specificity to 37.7% (P<0.01); Table 3. Using this cutoff would have reduced the number of BS by 152 (34.7%), with a single missed diagnosis.

Combining mpMRI and Grade Group scores in a novel risk-stratification system

This model defines five prognostic risk strata for prostate cancer (Table 1; Gnanapragasam et al, 2016). Within this model, no men in Risk Groups 1 or 2 had bone metastasis. Bone metastasis rate increased progressively in Risk Groups 3 (2.13%), 4 (5.8%) and 5 (13.7%); P=0.004. Using Risk Group 4 as a threshold for bone staging demonstrated improvement in specificity to 23.9% (P<0.01) and 97.3% sensitivity, with a 22.1% reduction in the need for BS.


Detection of bone metastasis confers a significantly worse prognosis in men with PCa and is thus an important part of the staging work-up (Fizazi et al, 2015). In our study, we showed that only 8.4% of men had bone metastasis out of the 438 men scanned, emphasising the need to refine use of this resource-intensive investigation. Of note, we did not find that a high PSA alone was a good discriminator; in our cohort, we identified 4 men with bone metastases and a PSA<20 ng ml−1, which is at odds with other reports (McArthur et al, 2012).

In this study, we have demonstrated evidence that mpMRI staging provides a useful adjunct in appropriately identifying men who will benefit most from BS. The integration of mpMRI alongside traditional biochemical and pathological markers, to re-categorise patients as high-risk disease, would have led to a 99.0% NPV for bone metastases at this threshold, alongside a significant increase in specificity to 24.2% and reduction in BS use. We also provide early validation of the new histological Grade Group system, with BS yield being higher in Grade Group 3 (bone metastasis rate of 5.8%) compared with Grade Group 2 (0.6%). Setting a threshold for BS of Grade Group 3 would have reduced the number of BS performed by 152 (34.7%), with a single false negative. This correlates with data used to inform the 2015 update to EAU guidelines (Heidenreich et al, 2014) and should encourage adoption of these proposed grading groups (Epstein et al, 2016).

It should be noted that grade and stage data are not used in isolation in clinical management. We therefore tested the proof-of-principle of a new prognostic risk model incorporating the new Grade Group system, mpMRI and biochemical information. This refinement demonstrated promising results with an NPV of 100% for Risk Groups 1 and 2. Taken together, our data demonstrate that the combined use of more accurate mpMRI staging and histological grade stratification better defines men who benefit most from bone staging investigations. These promising results show potential for reductions in the use of BS by up to a third while maintaining sensitivity and NPV above 97% and 99%, respectively.

The main limitation of our study is its retrospective nature and a selected population only including men undergoing both mpMRI and BS. The relatively small absolute numbers of men with intermediate-risk disease or men with bone metastases in this cohort should lead to caution when interpreting sensitivity or specificity values in isolation. Additionally, bone scintigraphy can be questioned as an ideal standard for diagnosing bone metastases; however, our data reflects current clinical practice and should be useful in guiding clinical management. PET-CT using fluorine or choline tracers has been shown to be a superior, albeit more expensive option for assessing metastatic bone involvement (Fuccio et al, 2012). Our results would require further validation in external cohorts in a prospective study and preferably using these modalities.


We demonstrate for the first time that the new histological Grade Group system and mpMRI staging more accurately identified men at risk of harbouring bone metastases. Importantly, our data strongly suggest that men without histology in Grade Group 3 and/or radiological T3 disease could safely avoid routine bone staging.