Reader Comments

Post a new comment on this article

Correction

Posted by hpnoto on 07 Sep 2018 at 07:29 GMT

The number of the all-cause deaths in the abstract and on page 3 should have been 24,659 (9.1%), rather than 15,981 (5.9%). The result of meta-analysis, however, is unaffected.

No competing interests declared.