Methodologic problems in clonogenic assays of spontaneous human tumors