Control intervention matters: an important but often ignored factor affecting clinical heterogeneity in meta-analyses of acupuncture trials.