On 2017 Aug 17, Jesper M Kivelä commented:

This re-analysis He FJ, 2011 of an earlier version Cochrane review Taylor RS, 2011 considering cardiovascular events (CVD) at longest follow-up is based on wrong numbers extracted from trials (namely TOHP 1 trial). There were 17 CVD of total 231 (not 321) patients in intervention group and 32 CVD of total 311 patients in control group in TOHP I (Table 2) Cook NR, 2007. See also my comment in current version of Cochrane review Adler AJ, 2014. Contrary, different numbers were extracted by O´Donnell et al O'Donnell MJ, 2013 in their analysis of the same set of 4 studies than used in He FJ, 2011.

In current version of Cochrane review, average risk ratio was 0.81 (95% CI 0.66 to 0.98) for CVD at longest follow-up based on 6 studies (Analysis 1.5) Adler AJ, 2014. However, between-study variance was estimated to be zero which is implausible based on empirical evidence from Cochrane Database of Systematic Reviews Turner RM, 2012, Turner RM, 2015. In other words, a small amount of between-study variance should be expected even for end-points like all-cause mortality Turner RM, 2012, Turner RM, 2015.

On 2017 Aug 17, Jesper M Kivelä commented:

This re-analysis He FJ, 2011 of an earlier version Cochrane review Taylor RS, 2011 considering cardiovascular events (CVD) at longest follow-up is based on wrong numbers extracted from trials (namely TOHP 1 trial). There were 17 CVD of total 231 (not 321) patients in intervention group and 32 CVD of total 311 patients in control group in TOHP I (Table 2) Cook NR, 2007. See also my comment in current version of Cochrane review Adler AJ, 2014. Contrary, different numbers were extracted by O´Donnell et al O'Donnell MJ, 2013 in their analysis of the same set of 4 studies than used in He FJ, 2011.

In current version of Cochrane review, average risk ratio was 0.81 (95% CI 0.66 to 0.98) for CVD at longest follow-up based on 6 studies (Analysis 1.5) Adler AJ, 2014. However, between-study variance was estimated to be zero which is implausible based on empirical evidence from Cochrane Database of Systematic Reviews Turner RM, 2012, Turner RM, 2015. In other words, a small amount of between-study variance should be expected even for end-points like all-cause mortality Turner RM, 2012, Turner RM, 2015.