Data recycling and misreading: two potential errors in pooled data from small studies.

Data from studies on small numbers of patients are often pooled to obtain more significant numbers with which to estimate response rates to chemotherapy. This has several problems that are well recognized. Two additional types of errors are (1) data recycling, in which the same data are included twice or more, and (2) data misreading, in which mistakes frequently occur in quoting the small numbers involved. Both errors are easily avoided by careful reading of the original reports.