Based on the measured data from the construction site, the prediction uncertainty of the CEBFIP MC90 due to the variation of the calculation parameters (i.e. the external uncertainty) and their sensitivity are analyzed with the help of statistical method and Latin Hypercube sampling method. Three conclusions can be deduced from the analysis. Those are a)the external uncertainty and its decreasing rate of the CEBFIP MC90 reduce with the time increasing, b)the creep coefficient increases with the increasing temperature while decreases with the increase of other calculation parameters, c)in accordance with the degrees of effect to the creep uncertainty, the first three calculation parameters are relative humidity, loading age and temperature, if the variation of the loading age is not taken into acount, the uncertainty of the creep pridection will be underestimated. Also, a method to analyze the uncertainty of the bridge structure due to the creep prediction uncertainty is proposed after the creep uncertain factor, which is the standard of the uncertainty of the creep, is introduced. An exmaple is analyzed in this paper using the random sampling method and the method presented in this paper. Through drawing a comparison between the results from different methods, the method presented in this paper appears to be feasible and efficient.