Can Large Language Models Match the Conclusions of Systematic Reviews?