In response to the extensive development of upland rice on the hillsides of the Madagascan highlands, alternative cropping systems based on conservation agriculture have been recommended to halt declining soil fertility and to limit erosion. To assess the efficiency of these cropping systems in limiting rice blast disease and to measure their yield performance, an experiment was set up in 2003 at Andranomanelatra (1640 m a.s.l.) in the Madagascan highlands. The rice crop was planted every second year following oat (Avena sativa) after common bean (Phaseolus vulgaris), with both conventional tillage and no tillage. For each cropping system, two levels of fertilization were used: (i) organic and (ii) organic + mineral fertilization. The level of blast epidemic was measured on two different cultivars over a 5-year period. Disease severity was significantly lower in the no-tillage cropping system than in the conventional tillage system. Mineral fertilization increased the level of blast. A significant interaction between cropping system and fertilization indicated that the impact of fertilization differed with the cropping system. When the level of blast was low, yield was higher in the conventional cropping system but as soon as blast level increased, yield was better in the no-tillage cropping system. Possible interactions between cropping system and blast epidemics are explained and the problem of high performance but risky cropping systems is discussed.