Radiologists and clinicians involved in advanced prostate cancer management and in development of imaging biomarkers.Bone metastases (BM) are highly prevalent in metastatic castration-resistant prostate cancer (mCRPC). There is an urgent need for robust radiological criteria for bone response in mCRPC with Prostate Cancer Working Group (PCWG) criteria only defining bone progression [1]. DWI is a promising assessment tool for BM detection and response assessment [2, 3]. We hypothesized that changes in the median apparent diffusion coefficient (mADC) and volume of BM, quantified by whole body (WB) DWI, are response indicators in mCRPC. We conducted a phase II trial of the PARP inhibitor olaparib in mCRPC (TOPARP-A CRUK/11/029); the primary endpoint of the trial was response to therapy defined using standard imaging (RECIST by CT scans), PSA decrease of ≥50%, and/or conversion of circulating tumour cell counts (CTC) from ≥5 to <5[4]. In patients treated at the Royal Marsden, optional WB-DWI was performed pre-treatment and 12-weekly on a 1.5 T MRI Siemens Avanto (Erlangen, Germany) (See Table 1). Mono-exponential ADC maps were generated using Siemens system software. Images were processed and analysed with open-access imaging assistant software (Osirix v5.6). Four different segmentation techniques were explored: 1) Skeletal DWI signal abnormality, including all areas of bone DWI signal abnormality in keeping with bone metastases (based on DWI and T1w) observed from C4 to lesser trochanters (spine and pelvis), 2) Entire volume of 5 target lesions, 3) Central slice of 5 target lesions and 4) Entire skeleton including normal and abnormal bone marrow from C4 to lesser trochanters (spine and pelvis). The 5 target lesions were chosen using the following criteria: maximum axial dimension > 1cm; well-defined lesion border; and representing different skeletal areas. The association between volume, diameter, and mADC changes between baseline and 12-weeks and binary response to treatment was assessed using logistic regression. A one-way analysis of variance (ANOVA) was pursued to assess differences in the distribution of the medians for each segmentation technique.33/50 patients consented to the WB-DWI substudy of whom 21 had WB-DWI at baseline and at 12-weeks. Of these 29% (6/21) were classified as responders as per trial criteria. Change in BM volume, diameter and mADC for responders and non-responding patients were assessed (Figure 1). Change in volume of BM DWI signal abnormality at 12-weeks associated with response to therapy (p<0.01). Similarly, change in volume or diameter of the 5-target BM (entire lesion and central slice) also associated with response (p=0.04 and p<0.01 respectively). The mADC change at 12-weeks, when segmenting all areas of BM DWI signal abnormality, also significantly associated with response (p=0.04) (Table 2; figure 2); when analysing mADC in 5-target BM (either entire lesion or central slice), an association with response was observed but was not statistically significant (p=0.06 and p= 0.08 respectively). Changes in mADC when segmenting the entire skeleton (normal and abnormal bone marrow) did not associate with response (p=0.50). Analysis of the variance of mADC at baseline, 12 weeks and percentage change in ADC and volume/diameter using the four different techniques did not show significant differences (p>0.05 for all parameters).Changes in BM DWI signal can reliably assess response to therapy, with changes in volume and mADC being promising response biomarkers. Simpler analyses (e.g. analysing 5-lesions) warrant further evaluation. Larger patient numbers and prospective validation of these data is on going.Changes in volume and mADC of BM assessed by DWI correlate with response to anticancer therapy in mCRPC.