What if we could predict breast cancer? Through plasma analysis, scientists have your odds – with 82% accuracy, years before disease onset.
Breast cancer is the major cause of death in the first decade after menopause in women. Early intervention is one of the key game changers. An important focus in cancer research is therefore the pursuit of techniques for detection of malignancy before the clinically detectable disease stage.
Many western countries recommend mammography for tumor detection prior to clinical symptoms, although the burden of over-diagnosis and unnecessary intervention in non-malignant tumors are points of contention. To counter this problem, a research team investigated using plasma samples for forecasting breast cancer ahead of any indication by mammography.
In a study published in May 2015, scientists looked at 57,073 healthy men and women who joined their study between 1993 and 1997. Plasma samples collected from these subjects were cryopreserved. By Dec 31, 2000, 418 female subjects who had developed breast cancer were identified. From the majority remaining cancer-free, a matching number were chosen as healthy female controls.
Scientists subsequently analyzed plasma samples from selected subjects, dividing these into 1) those who developed breast cancer, and 2) those who remained healthy. The interest was in metabolites – the intermediates and products of metabolism present in plasma, and whether they could be correlated with the outcome of health vs. disease years later.
Plasma metabolites were analyzed with nuclear magnetic resonance spectroscopy (NMR). Though quantitative and a highly reproducible technique, NMR in this exploratory manner posed a practical hindrance because of the breadth of information. Which metabolites were of significance? Which metabolites were noise?
Scientists sorted through the plasma sample NMR data and also included 47 other variables for lifestyle and phenotype of subjects. This meant a total of 176 variables were analyzed to look for correlation with cancer or health. The expansive data from a huge cohort provided a statistical power of inference that is not always feasible.
When hormone replacement therapy was looked at alone, no significant correlation was seen. Thus, rather than using single variables, scientists explored groups of relevant variables to reflect biological patterns related to the given endpoint. In this manner, 27 of the 176 original variables were pinpointed to forecast breast cancer.
Scientists found that NMR data from plasma metabolites alone could predict breast cancer with an error of 22%, highlighting the importance of this quantitative and reproducible parameter. NMR data in combination with other variables encompassing a "27-point biocontour" further brought down error to 18% for predicting breast cancer.
- Rasmus Bro, Maja H. Kamstrup-Nielsen, Søren Balling Engelsen, Francesco Savorani, Morten A. Rasmussen, Louise Hansen, Anja Olsen, Anne Tjønneland, Lars Ove Dragsted. Forecasting individual breast cancer risk using plasma metabolomics and biocontours. Metabolomics. 2015;10 Mar. doi:10.1007/s11306-015-0793-8.