Feasibility of a non-invasive temperature profiling technique is studied for use in clinical hyperthermia environment, where a water-filled bolus is used for cooling the surface of treatment region of patient's body during the electromagnetic heating for hyperthermia treatment of cancer. We intend to measure the temperature distribution in the treatment region using the microwave radiometry. For the temperature measurement to be compatible with the heating, the radiometric measurement has to be made via an antenna placed on the bolus. The presence of the bolus between the radiometer antenna and the body surface adversely affects the radiometric measurement, since it attenuates the thermal radiation from the body on one hand and generates that of its own on the other hand. Hence, the presence of the bolus degrades the radiometric measurement. This peper describes an analysis of the degradation due to the bolus by applying a previously developed temperature profiling method, which allows one to calculate the confidence interval of a temperature-varsus-depth profile in tissue estimated from a set of radiometric measurement data. We analyzed the experimental data obtained on an agar phantom with using an experimental radiometer operating at center frequencies of 1.2, 1.8, 2.5, 2.9, 3.6 GHz with a 0.4-GHz bandwidth for respective frequency bands. The radiometer had the brightness temperature resolution of about 0.07 K for a 5-second integration time. Results showed that the precision of tissue temperature was about +/-0.8 K at 2 cm and +/-2.4 K at 4 cm from the surface into phantom tissue which was uniform with muscle-equivalent electrical characteristics. The degradation due to a 1-cm thick bolus filled with the distilled water was 0.1 K at 2 cm and about 0.2 K at 4 cm from the surface, indicating that the presence of bolus could be tolerated. In the present work, the precision is defined by plus or minus one half of the confidence interval at a 70-percent-confidence level, which roughly corresponds to the error bar defined by plus and minus one standard deviations of a normal distribution. It was also shown by means of the numerical simulation that the precision would be improved to +/-0.4 K at 2 cm and +/-1.2 K at 4 cm if the brightness temperature resolution of the radiometer was improved to about 0.03 K.