Lack of access to the full set of metadata variables associated with these earlier studies is crippling, as interpretation of the observational data can only happen within the context of the collected variables. From a practical standpoint, if a systematic pattern is observed in the data, but there are not any variables that explain the pattern, then the researcher cannot support a hypothesis about the pattern without collecting new information (which may be impractical and impossible or introduce recall bias). Similarly, confidence in the face of confounding variables is reduced if only a limited number of variables are tracked. As a concrete example, if researchers broadly characterize subjects by diet type (e.g., vegan) and observe an effect, the researchers will be unable to assess whether the effect is due to the diet type itself, differential fiber consumption, protein source, etc., unless these other variables are recorded. Given that researchers typically do not know the answer in advance of a study, it is imperative that study designs strive to collect as much information as feasible.
