A risk ratio or difference from a meta-analysis is as many as ten steps awa
y from the unobservable causal risk ratios and differences in target popula
tions. The steps are like lenses, filters, or other fallible components of
the epidemiologist's "telescope" for observing populations. Each step is an
other domain where different biases can be caused. How biases combine acros
s domains in the production of epidemiologic evidence can be quickly explai
ned to nonepidemiologists by using a sequence of causal arrow diagrams with
easy notation: (a) agent of interest, (b) background risk factors, (c) cor
related causes, (d) diagnosis, (e) exposure measurement, (f) filing of data
, (g) grouping of cohorts, (h) harvesting of cases and controls, (i) interp
retations of investigators, (j) judgments of journals, and (k) knowledge of
meta-analysts. For epidemiologists, this article serves as a review of ide
as about confounding, information bias, and selection bias and underscores
the need for routinely analyzing the sensitivity of study findings to multi
ple hypothesized biases.