When you have a certain clinical set, you take available patients. If you get less than 30 patients, you use non-parametric methods for computations of results, if you have more patients you use parametric methods. Median is not the best measure. -- I looked up sample sizes used previously and they varied between 4 to 90 patients!

Hello

i agree with Beatrice 100%

Article Sample Size Determination

Good Luck

You should use the "calculadora amostral", available online. Put your data in blanks and you will have a idea about your "n".After you have some results of your new experiments (pilot model), put your data again and check if the "n" is enough

http://www.publicacoesdeturismo.com.br/calculoamostral/

http://comentto.com/blog/calculadora-amostral

I suggest you look into certain statistical packages like Epiinfor or others with the help of a statistician. They contain facilitation to calculate sample size. In your case, you follow up patients with certain tumours and make serial assessment of the marker you study (you need at least two readings to see the change in the marker value(level). Of course you need to specify your objective. Do you like to measure just trend in the change (increasing, declining or unchanged) or your are looking for a significant change