There are various measuring tools to evaluate the therapeutic effectiveness of acupuncture for neck pain caused by cervical spondylosis, such as NPQ and MPQ. However, the outcomes are challenged because cervical spondylosis can be subdivided into different sub-types due to different pathological diagnosis. Therefore, a new algorithm is needed to analyze the difference of therapeutic effectiveness among diagnostic subtypes. We proposed Kernel Canonical Correlation Analysis (KCCA), which has been successfully applied in many statistical learning tasks, as a potential approach to discover the underlying relationship between different effective outcome measures and diagnostic sub-types in clinical practice. The application of kernel mapping on the basis of correlation analysis provides a nonlinear relationship expression between input variables. The proposed method is applied to the clinical data from a multi-center randomized controlled trial (RCT) on acupuncture for neck pain caused by cervical spondylosis, and the result shows that it is effective and capable to dramatically improve the correlation between diagnostic sub-types and clinical outcome measures.