To evaluate tumoral T1 relaxation times using multiple flip angle method prior to, and early during, systemic chemotherapy for prediction of response to therapy.

Oncological therapies are associated with variable efficacy, side-effects and financial cost. Radiological tests are increasingly used for assessing treatment response, usually measuring lesional size changes. However, therapy related size change may not occur for several treatment cycles, whilst emerging therapies may not result in volume change [1].

Preclinical evidence suggests changes of the order of 20% in responding tumor T1 relaxation time precede size change [2]. Repeatability of a whole body B1 corrected multiple flip angle (MFA) method has been reported in healthy volunteers with co-efficients of variation across tissue types between 5-10% when selecting two appropriate flip angles [3].

Institutional approval was obtained for enrolling oncology patients with histologically proven and metastatic breast cancer, soft tissue sarcoma or localised extraosseous Ewing's sarcoma prior to commencing systemic chemotherapy treatment between January 2014 and June 2015 (n=59). Those with non-measurable disease, prior chemo/radiotherapy and/or contra-indication to MR were excluded (n=44).

Fifteen patients (mean age 54.4 years) underwent baseline whole body 3.0T MRI (Fig 1) prior to commencing systemic chemotherapy (median interval 4 days) covering vertex to mid thigh. Volumetric T1 maps (Fig 2) were generated using dual flip angle (2.5 and 15 degrees) 2-point modified (m)Dixon 'in phase' images by a linear fitting algorithm [4] with MATLAB (v7.13) incorporating a separately acquired dual TR B1 map [5]. Apparent diffusion coefficient (ADC) was calculated by mono-exponential curve fitting of mean signal intensities of all b-values from diffusion weighted sequences.

Baseline T1 study

Two experienced radiologists in consensus identified measurable soft tissue lesions in each patient with the benefit of prior imaging, though blinded to subsequent outcome information, recording single largest diameter of each lesion using anatomic sequences according to established 'Response Evaluation Criteria in Solid Tumors' (RECIST 1.1) 'target' lesion criteria [6]. Each lesion was contoured as a region of interest (ROI) generating baseline pixel-by-pixel T1 and ADC median values. In total 57 lesions (median 4/patient; 1-8) were defined (Fig 3).

Lesions were followed after the end of treatment (post six cycles of chemotherapy) and categorised into 'partial response' (>30% decrease in maximum diameter), progressive disease (>20% increase) or stable disease [4]. Partial responding lesions were compared to non-responding lesions (i.e. progressive disease + stable disease) in all analyses.

Early Follow-up Study

Seven patients (46% original cohort) were rescanned after receiving two cycles of chemotherapy with total of 28 lesions (median 4/patient; 1-7). Two radiologists again contoured target lesion ROIs with reference to original MRI though blinded to subsequent outcome data. Post second cycle, T1 and ADC median values were obtained and percentage changes compared to baseline were calculated for each lesion.

Per lesion differences between partial and non-responding lesions in baseline T1/ADC, early post second cycle T1/ADC and percentage change from baseline were analysed using Mann Whitney U test (statistical significance p<0.05) and univariate accuracy for prediction of partial response determined by receiver operating characteristic (ROC) area under curve (AUC) analysis.

Figure 3 summarises median values and ranges of baseline and post 2 cycle T1 and ADC values and percentage changes post 2 cycles. At baseline, partial responders had slightly higher median T1 of 2765ms versus 2267ms in non-responders (p=0.09) and significantly lower median ADC 0.89x10^-3mm²/s versus 1.12x10^-3mm²/s (p=0.02).

Percentage change in T1 value after two cycles chemotherapy in partial responding lesions was significantly lower at -64% versus +9% in non-responders (p=0.001) and ADC significantly higher at +70% versus 9.8% (p=0.003). The actual ADC value post 2 cycles was also significantly higher in partial responding lesions at 1.77 versus 1.28x10^-3mm²/s (p=0.001), whilst actual T1 value was not significantly different (p=0.17)- summarised in Fig 4.

Percentage change in lesion T1 values after two cycles was the best predictor of partial response with ROC-AUC 0.89; followed by the actual ADC value post 2 cycles, percentage change in ADC from baseline and baseline ADC with ROC-AUC of 0.86, 0.84 and 0.70 respectively (Fig 5).

Prediction of eventual partial response by early lesion T1 value change is comparable with more widely used quantitative imaging biomarkers such as ADC value. As previously reported in pre-clinical models, our in-vivo human T1 changes may reflect a decrease in remaining viable/proliferating tumour cells due to cell destruction releasing proteins and/or metals (causing T1 relaxation) [7]. This additional quantitative information can be obtained on a whole body basis within a clinically feasible total scan duration and may provide complementary information by highlighting different intrinsic tissue properties.