Based on 52 studies with samples mostly from English-speaking countries, the current study used Bayesian network meta-analysis to investigate the intervention effectiveness of different reading comprehension strategy combinations on reading comprehension among students with reading difficulties in 3rd through 12th grade. We focused on commonly researched strategies: main idea, inference, text structure, retell, prediction, self-monitoring, and graphic organizers. Results showed (1) instruction of more strategies did not necessarily have stronger effects on reading comprehension; (2) there was no single reading comprehension strategy that produced the strongest effect; (3) main idea, text structure, and retell, taught together as the primary strategies, seemed the most effective; and (4) the effects of strategies only held when background knowledge instruction was included. These findings suggest strategy instruction among students with reading difficulties follows an ingredient-interaction model--that is, no single strategy works the best. It is not "the more we teach, the better outcomes to expect." Instead, different strategy combinations may produce different effects on reading comprehension. Main idea, text structure, and retell together may best optimize the cognitive load during reading comprehension. Background knowledge instruction should be combined with strategy instruction to facilitate knowledge retrieval as to reduce the cognitive load of using strategies. [This is the online version of an article published in "Review of Educational Research."]