The multi-crossover model for classifying patients as responders to a given treatment.

The strength and the validity of the multi-crossover model (MCO model) were investigated to optimalize the procedure for correctly classifying individual responders to a given therapy. One hundred and fifteen patients with non-ulcer dyspepsia from seven Norwegian hospitals were included in a 6-week double-blind MCO-designed trial with alternating weekly treatments with ranitidine and placebo. An individual effect score (X score) was calculated on the basis of the number of times the active drug was associated with less symptoms than the preceding or following placebo period. Patients categorized as responders (X score greater than or equal to 4) or unclassifiables (X scores = 2 or 3) after the MCO period continued single-blind active treatment for 4 weeks and were then reclassified by means of cross-tabulation of efficacy and adverse effects. Eighty-five per cent of the MCO responders and 62% of the 42 MCO unclassifiables were reclassified as responders. The reclassified responders were then included in a single-blind follow-up placebo period until relapse or for a maximum of 8 weeks. The relapse rate was significantly greater (p less than 0.01) and the time to relapse significantly shorter (p less than 0.01) in the group of MCO responders than in the MCO unclassifiables. The large response and relapse rates in the group of MCO responders verify that the MCO model is a reliable method for correctly classifying responders to treatment. Our results indicate that by including patients with an X score of 3 in the definition of 'responder', the MCO model could be modified to optimalize the procedure for correct classification.