Quality assessment
Six studies were assessed as good quality, four studies as fair quality and one study as poor quality (Table S1). With one exception22, all studies scored at least two out of four stars for selection criteria and two out of three stars for outcome criteria. Most frequent source of risk of bias included ‘no control group’ (four studies did not include a control group), ‘selection of controls’ and ‘adequacy of follow up’, since nine studies were retrospective cohorts.