Figure 3. Frequency of best policies over key parameters.
An aggressive policy (dark blue) is deemed best if the Spearman correlation S between treatment and resistance is S < −0.7, moderate (light blue) is deemed best if S > 0.7 and the classification is neutral (medium blue) otherwise. When the DR strain has a lower growth rate (LamR), an aggressive policy is more likely best because more of the DR strain's population arises through resistance acquisition from the DS population. In this case, reducing the DS strain also reduces DR. Conversely, when ΛR (LamR) is high the DR strain is a more robust competitor and a moderate policy is more frequently best. Similarly, when the DR strain has a low MIC (mR), it is a less robust competitor. In this case, an aggressive policy is more frequently best than when mR is high (second panel). The third panel shows that when the immune system is strong (high kp), an aggressive policy is more frequently best, because again more of the DR population increases are driven by acquisition from DS, due to immune suppression of DR growth. A plot with η on the horizontal axis is very similar to this one. Finally, the right plot shows that when the DS growth rate (LamS) is low, an aggressive strategy is more often best to minimize resistance; this depends on the ability of therapy to prevent the emergence of resistance.