Yes, by performing a non-controlled, non-blinded, non-randomized trial. And by keeping track in a non-systematic manner the course of a disease that naturally waxes and wanes upon which is superimposed his own attempts to alter that course. Under those circumstances, when confirmation and attribution bias are allowed free rein, it would be very difficult not to be able to identify a treatment as "effective". It is, after all, exactly how medicine was practised until 150 years ago and allowed us to confirm the effectiveness of bloodletting and heavy metal poisoning in the treatment of disease.