Multi-parametric MRI for predicting disease recurrence or death in breast cancer patients following neo-adjuvant chemotherapy.
Elizabeth Anne Maxine O'Flynn1, Maria A Schmidt1, David Collins1, James D'arcy1, and Nandita M deSouza1

1Cancer Research UK Cancer Imaging Centre, Institute of Cancer Research and Royal Marsden Hospital, Sutton, United Kingdom

Synopsis

The aim of this pilot study was to explore whether functional MRI metrics from a multi-parametric acquisition can predict for disease recurrence or death in breast cancer following neo-adjuvant chemotherapy. 60 months after commencement of the study, a lower R2* value was found at baseline in the women who are alive and disease free, suggesting that these tumours were less hypoxic and better oxygenated than those who developed metastatic disease or died in the interim

Background and purpose

Multi-parametric MRI is increasingly used in assessing response to neoadjuvant chemotherapy in breast cancer however it’s potential as a predictor of outcome remains unexplored. This pilot study explores whether functional MRI metrics from a multi-parametric acquisition can predict for disease recurrence or death in breast cancer following neoadjuvant chemotherapy.

Methods

Research ethics committee approval and patient written informed consent were obtained. 42 women (median age 52 years, range 32-82 years) with proven breast cancer (40 invasive ductal carcinomas and 2 invasive lobular carcinomas) underwent breast MRI prior to neoadjuvant chemotherapy on a 3.0T Philips Achieva MRI scanner (Best, Netherlands) between September 2010 - August 2013. A DW- sequence (sagittal single shot echoplanar sequence, with SPAIR and a slice-selection gradient reversal (SSGR) method for fat suppression with 4 b values (0, 100, 700, 1150mm2/s) (TR/TE=3771/66 ms, flip angle 90°, 180 mm FOV, 3 mm slice thickness with 0 slice gap, one excitation and a 1.96x2.02x3mm acquisition voxel)), T2* sequence (sagittal gradient echo (FFE) with 12 echoes (TR/TE=1400/4.6ms, echo spacing 6.9ms, flip angle 18⁰, slice thickness 3mm, 180mm FOV, a 1.22x1.2x3mm acquisition voxel reconstructed to 0.94x0.94x3mm) and a dynamic contrast enhanced (DCE) sequence (sagittal 3D gradient echo sequence with a temporal resolution of 2.5seconds (TR/TE=4.5/2.3 ms, flip angle 16°, 180mm FOV, 3mm slice thickness with 0 slice gap, one excitation and a 2.37x2.4x6mm acquisition voxel reconstructed to 0.94x0.94x3mm)) following intravenous injection of 0.2ml/kg of gadoterate meglumine (n=17) or gadopentate dimeglumine (n=25) were performed covering the tumour-containing breast. 1 patient declined intravenous contrast and had a non-contrast study acquiring ADC and R2* data only. Regions of interest (ROI) were drawn manually slice-by-slice on an early subtracted DCE image using in-house software (MRIW, Institute of Cancer Research, London). A modified Tofts pharmacokinetic model was used to estimate kinetic parameters Ktrans, ve and kep. The ROI’s were aligned to the corresponding ADC and R2*maps using in-house software (Adept Institute of Cancer Research, London). The ADC maps were computed from mono-exponential fitting of signal intensity for all 4 b values. R2* was computed using echo times (4.6–59.81ms). Values for ADC, R2*, Ktrans, ve, kep and IAUGC were recorded pixel-by-pixel for each patient and the median documented. Non-enhancing pixels were excluded in this part of the analysis, to prevent bias of results towards 0. The number of enhancing pixels per slice was recorded and multiplied by the reconstructed voxel size (0.94x0.94x3mm) to determine tumour volume for each time point. The enhancement fraction (EF) was calculated as total number of enhancing pixels/total number of pixels for each patient. 60 months after commencement of the study, 6 women had died and 2 had developed metastatic disease (2 in this group achieved pathological complete response at final surgery (pCR), while 34 were alive without disease (12 of these women achieved pCR at final surgery). An independent t-test was used to determine any difference in the mean of each parameter at baseline and binary logistic regression to determine if any parameter could predict the likelihood of metastatic disease or death at baseline.

Results

At baseline, mean R2* values were significantly lower in women who were alive without disease (45.1+/-9.6 sec-1) compared to those who were alive with the disease or dead (51.3 +/-5.6sec-1) (p=0.028) (figure 1). No other baseline imaging metric showed any significant difference between the two groups. Univariate logistic regression showed no association between any parameter and disease recurrence or death following neoadjuvant chemotherapy.

Discussion

60 months after commencement of this neoadjuvant response study, a lower baseline R2* value was found in women who are alive and disease free, suggesting that these tumours were less hypoxic and better oxygenated than those who developed metastatic disease or died in the interim. Better oxygenated tumours may well be associated with improved disease free survival or overall survival, with insufficient patient numbers in this pilot to show association with disease recurrence or death on univariate regression analysis.

Conclusion

A lower baseline R2* value was found in women who are alive and disease free following neoadjuvant chemotherapy. No parameter was found to predict for disease recurrence or death from a pilot multi-parametric acquisition.

Acknowledgements

CRUK and EPSRC support to the Cancer Imaging Centre at ICR and RMH in association with MRC & Dept of Health C1060/A10334, C1060/A16464 and NHS funding to the NIHR Biomedicine Research Centre and the Clinical Research Facility in Imaging.

References

No reference found.

Figures

Table showing the difference in baseline imaging metrics between women who are alive without disease and those who are alive with disease or dead.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2728