Bibian van der Voorn

144 CHAPTER 10 6,923 records identified through database searching 3 additional articles identified through other sources 6,158 records, after removal of duplicates 5,744 records were not eligible for inclusion 414 full-text articles assessed for eligibility 79 studies were included 308 full-text articles were not eligible for inclusion 21 studies in descriptive analysis 27 full-text articles excluded due to lack of data 106 full-text articles were eligible for inclusion 58 studies in meta-analysis FIGURE 2.Thisflowchartpresentsthedifferentphasesofthesystematicreviewandmeta-analysis,conformthePRISMA-statement. (www.prisma-statement.org ) DESCRIPTION OF INCLUDED STUDIES Studies were conducted in Europe (n = 36), North-America (n = 37), Asia (n = 3), South-America (n = 2) or Africa (n = 1), and were published between 1973 and 2016. Sample sizes ranged from 11 to 2,824 subjects, with seven studies having a sample size >500 subjects. Study designs were as follows: randomized placebo-controlled (n = 2), prospective observational (n = 29), non-randomized intervention, i.e., stress tests (n = 15), cross-sectional (n = 16), longitudinal (n = 11) and case-control (n = 6). All studies that assessed serum or salivary cortisol used immunoassays, except for one that used high-performance liquid chromatography (HPLC). Studies that assessed 24h-urine cortisol used immunoassays (n = 4), gas chromatography–mass spectrometry (n = 3), HPLC (n = 1), and liquid chromatography-UV detection (n = 1). Twenty-two studies (28%) did not collect morning samples, of which 11 did not report the time of collection and 11 described specifically that samples were collected in the afternoon. Online supplementary File 1 presents the data extracted from the articles included in the meta-analysis. Three out of 21 studies (14%) included in the descriptive analysis had no high bias risk ( Table 1 ), while 16 out of 58 studies (28%) included in the meta-analysis had no high bias risk ( Figure 1 ).