Design of medical sensor implants is very challenging due to numerous restrictions in terms of size and biological acceptability. In addition, large absorption and reflection of electromagnetic waves in body tissues significantly degrades the communication link capabilities. The usual assumption in calculating the link budget is that the permittivity and conductivity of the body tissues are known. However, biological tissues show considerable diversity in structure and consequently in dielectric properties. The first aim of this article is to determine how much the uncertainty of the body tissues' constitutive parameters affects the prediction of the achievable radiation efficiency of the implant. The second aim is to understand the loss mechanism and thus which tissue permittivity and/or conductivity has the dominant influence on losses. The analysis is done by considering a spherical multilayer human phantom and applying spherical wave decomposition proposed in a previous article. The stochastic collocation method is used to estimate the uncertainty of power density outside the human body, and sensitivity on the uncertainty of each tissue parameter is performed using the analysis of variance (ANOVA) approach. Several test cases were considered and the results clearly indicate which constitutive parameters have dominant effect on uncertainty.