Figure 1.

PRISMA Flowchart of the study selection process.Two studies used the same prostate cancer cohort but one reported on circulating 25OHD and the other on genetic variants, and so both were retained. (Holt et al,2010, 2013) Three publications used the same initial cohort of lung cancer patients but two reported on different subpopulations of patients (according to disease stage) and so were retained, (Zhou et al, 2007; Heist et al, 2008), while a third reported on different exposures to the first two and so was also retained (Zhou et al, 2006). Finally, four studies reported on the same melanoma patient cohort (Newton-Bishop et al, 2009, 2015; Field et al, 2013; Davies et al, 2014) (one paper scored lower in NOS scoring was excluded (Field et al, 2013), while the remaining three, which reported different exposure or outcomes were retained. ^§Only a single study reported impact of circulating vitamin D-binding protein levels on outcome and so could not be included in the meta-analysis. *Includes only exposures and outcomes included in MA. Articles may report on multiple exposure-outcome pairs hence the sum of the pairs is greater than the number of articles included. For example, several papers studied the effect of more than one SNP for example, Zgaga et al, (Zgaga et al, 2014), while many papers studied the impact on both overall survival or progression-free survival for example, Lohman et al (Lohmann et al, 2015). However, where multiple estimates were extracted, no patient was included more than once for a certain exposure or outcome. † Study authors were contacted to provide HR, RR or OR when not reported; 13 did not respond. ¥ One study (Vrieling et al, 2011) used the same breast cancer cohort as a later, larger study (Vrieling et al, 2014) and as both had the same NOS score, the newer study was included. 25OHD: 25-hydroxyvitamin D; DBP: vitamin D binding protein; HR: hazard ratio; PSA: prostate specific antigen; WOK: Web of Knowledge.