Key points are not available for this paper at this time.
The point estimate I(2) should be interpreted cautiously when a meta-analysis has few studies. In small meta-analyses, confidence intervals should supplement or replace the biased point estimate I(2).
Paul T. von Hippel (Mon,) studied this question.