[Observations on the number of sputum samples used in establishing the bacteriological diagnosis of pulmonary tuberculosis].