Optical spectroscopy identifies responders

09 Jul 2009

Measuring haemoglobin concentration using optical spectroscopy could provide an early predictor of a tumour's response to chemotherapy.

Diffuse optical spectroscopy (DOS) has the potential to predict treatment response in patients undergoing neoadjuvant chemotherapy for locally advanced breast cancer. Pilot studies have raised hopes that DOS could show quickly whether this strategy of shrinking tumours prior to surgery is working. An early alert could spare patients the discomfort of unnecessary chemotherapy, as well as reduce needless delays before surgery. It could also provide encouragement to patients whose drug therapy is working that the side effects are worth tolerating.

Doctors need a clear yes/no message if they are to use spectroscopic data to guide patient management. With this in mind, researchers from Dartmouth Medical School and the Thayer School of Engineering at Dartmouth College (Hanover, NH) have evaluated two methods of summarizing DOS data when discriminating between complete and incomplete responders (Radiology doi: 10.1148/radiol.2522081202).

A total of 16 women with locally advanced breast cancer were enrolled in the study. All were imaged with mammography and MRI before drug treatment was started, and then again with MRI prior to surgery. DOS imaging was carried out before, several times during, and after drug treatment. Resected tissue specimens from the surgery were used to classify patients as complete or incomplete responders.

Two different definitions of the tumour region-of-interest (ROI) were used when analysing the spectral data. One approach was to define a fixed-size ROI based on a radiologist's interpretation of pre-therapy MR images. The alternative was to base the size and position of the ROI on measurements of total haemoglobin concentration (Hb_T) acquired from DOS imaging. The ROI area, which could potentially alter during treatment, was defined as the region with Hb_T values larger than half of the maximum.

Shudong Jiang and colleagues determined average values of Hb_T, blood oxygen saturation, water fraction, and optical scattering amplitude and scattering power from all reconstructed DOS images, inside and outside the tumour ROI. They also gathered average values of the same parameters for each patient's contralateral, healthy breast so that readings could be normalized.

Comprehensive sets of sequential DOS data were collected for seven of the study's participants, with two cases studied in detail in the Radiology paper. Just one of the multiple parameters (Hb_T change) was shown to be significantly different during the early cycles of treatment, when comparing patients who had responded completely to pre-surgical chemotherapy with those whose response was deemed incomplete. The change in Hb_T for complete responders was noticeably greater when the tumour ROI was based on DOS imaging, not MRI.

"The approach of using a fixed-size ROI analysis on the basis of pre-treatment MR imaging findings might appear to be best in terms of establishing an objective definition of the ROI, but it does not account for changes in tumour volume that may occur during treatment when serial MR imaging is not available," the authors commented. "Because tomographic DOS provides spatial mapping of spectral data, it is attractive to use this information to define the ROI."

The discovery that Hb_T change could predict tumour response so early in the chemotherapy cycle was an "interesting surprise", Jiang told our sister website medicalphysicsweb. However, no firm conclusions will be drawn until clinical data from additional patients have been analysed.

According to Jiang, this strategy of forecasting the value of neoadjuvant therapy need not be limited to breast cancer. Dartmouth researchers are also exploring whether DOS could be used to monitor therapy and predict response in head-and-neck cancers. "As neoadjuvant therapy use increases in other types of cancer, the ability to detect the success or not in early days following administration will likely become a very important patient management tool," she said.