Incorporating variations in the quality of individual randomized trials into meta-analysis