Key points are not available for this paper at this time.
I(2) estimates need to be interpreted with caution when the meta-analysis only includes a limited number of events or trials. Confidence intervals for I(2) estimates provide good coverage as evidence accumulates, and are thus valuable for reflecting the uncertainty associated with estimating I(2).
Thorlund et al. (Wed,) studied this question.