||Microstrip antennas are an attractive option for biotelemetry communication between implanted medical devices and external control systems due to their small size and ability to be integrated with the implanted device. These antennas may be placed in the shoulder or chest cavity for cardiac devices, under the skin behind the ear for artificial hearing devices, or under the lens of the eye for retinal prostheses. Thus, the antenna's actual operating environment can have substantial variation due to placement location, differences between individuals, and also from uncertainties in the electrical properties of the biological tissues. These variations can be described statistically, and the purpose of this thesis is to evaluate the resultant variation in the antenna performance from these variations. In this thesis, we apply an analytical technique called the variational method to solve for the effective relative permittivity, ereff, and effective conductivity, aeff, of a microstrip antenna placed under muscle, fat, and skin layers with statistically variable thicknesses and electrical properties. The variational method was used to solve the quasistatic Poisson's equation for electric potential in a multilayered medium. The resulting integral expression was then solved numerically using a Monte Carlo simulation method by which the effects of tissue variations, discussed above, were studied. These simulations were used to calculate ereff and aeff for thousands of different configurations of dielectric layer thickness and constitutive parameters. These values were then used to quantify detuning effects such as shifts in resonant frequency and the resultant reduction in useable bandwidth of the implanted antenna. The simulations used two types of input data. First, the measured means and standard deviations for the constitutive parameters, sr and o, for muscle, fat, and skin tissues were used to generate randomly varying, normally distributed, electrical properties. Second, superstrate and tissue thicknesses were randomly chosen (uniformly distributed) within ranges considered appropriate for the particular area in the body being studied. The simulation data were statistically analyzed by fitting their empirical distributions to the lognormal distribution function. The parameters of these fitting functions were estimated and used to calculate confidence intervals on expected variations in resonant frequency. The effect of superstrate thickness on reduction of this variation was demonstrated.