Osteosarcoma is the most common malignant bone
tumor that affects hundreds of children and young adults every year. The major prognostic factor in patients with localized
osteosarcoma is the development of resistance towards pre-operative
chemotherapy. However, modifications of post-operative
chemotherapy based on the histological response have not significantly improved the outcome of patients. Thus, it would be of tremendous clinical value if the poor responders could be identified at the time of diagnosis, so that ineffective
therapy can be prevented and intensified or alternative
therapy could be provided to improve their outcome. We hypothesized that plasma proteomic profiles could be used to distinguish good from poor responders prior to the start of treatment. In order to test this hypothesis, we analyzed the proteomic profiles in two sets of plasma samples (n=54) from
osteosarcoma patients collected before (n=27) and after (n=27) pre-operative
chemotherapy. Using a linear support vector machine algorithm and external leave-one-out cross validation, we developed two classifiers that classified good and poor responders with an equal accuracy of 85% (p<0.01 after 5000 permutations) in both sets of plasma samples. In order to understand the biological basis of the classifiers, we further identified and validated two
plasma proteins,
serum amyloid protein A and
transthyretin, in the classifiers. Our results suggest that plasma proteomic profiles can predict
chemotherapy response before treatment as accurately as
after treatment. Our study could lead to the development of a simple blood test that can predict
chemotherapy response in
osteosarcoma patients. Since the two identified
proteins are involved in innate immunity, our findings are corroborated by the notion that boosting the innate immunity in conjunction with
chemotherapy, achieves a better anti-
tumor activity, thus improving the overall survival of
osteosarcoma patients.