Figure 2.

p Values of Joint Association Tests When the Overlapping Control Samples Are Shared or Split between the CATIE and GAIN Studies

“Sharing (robust)” and “sharing (naive)” pertain to the use of the robust and naive variance estimators when the overlapping control samples are shared in the meta-analysis. Under option 1, all 199 overlapping controls are assigned to CATIE; under option 2, all 199 overlapping controls are assigned to GAIN; under option 3, a total of 99 overlapping controls are randomly assigned to GAIN and 100 to CATIE; under option 4, a total of 29 overlapping controls are randomly assigned to GAIN and 170 to CATIE.