ALT A continuum of bias in study design. At the top, a double-headed arrow is labeled “unbiased” on the left with “ideal randomized controlled trial” as an example and “intractably biased” on the right with “relying on ‘gut feeling’” as an example. Below the arrow are 4 overlapping bars with relative positions of 4 main study types. From left to right: explanatory randomized trials, pragmatic randomized trials, observational studies, simulation studies. Each study type covers a range of quality. For each of the four types listed, under the very best version of the design possible that design is better than those listed to the right of it. But, the very best version of any design to the right may be better than a bad version of the design further to the left. Below that, potential biases are listed left to right: lack of generalizability, loss to follow-up, non-adherence, baseline confounding, ill-defined uncertainty. When a bias occurs, all biases further left are also likely.