Salivary cortisol measures across the clinical stages of psychosis: An individual participant data (IPD) meta-analysis

Abstract

Background

Studies of salivary cortisol levels in psychosis have yielded inconsistent findings, which may be attributable to heterogeneity in cortisol measurement, illness stage, and approaches to dealing with sampling factors and potential confounders. To address these issues, we performed an individual participant data (IPD) meta-analysis comparing individuals at different stages of psychosis to controls using five different salivary cortisol measures and explored potential effect modifiers.

Methods

Salivary cortisol data from five London-based cohorts were used to derive the cortisol awakening response, total daytime cortisol output, basal cortisol, and diurnal slope measures (wake-to-evening and peak-to-evening). Linear regression models were first performed to obtain standardised beta coefficients (β), representing the difference in each cortisol metric between each clinical stage group (cases) and healthy individuals (controls) after accounting for relevant sampling factors; we then used random-effects meta-analyses and meta-regression models to investigate the effect of psychosis stage and sample characteristics on effect sizes.

Results

Data were available for 352 individuals distributed across psychosis clinical stages (1a – distress disorder: N = 35; 1b – clinical high-risk for psychosis: N = 90; 2a – first-episode psychosis: N = 197; 2b – single episode remitted: N = 5; 3 – relapsing/remitting illness: N = 18; 4 – severe and persistent illness: N = 7) and 292 controls. A significant overall main effect of clinical stage on peak-to-evening diurnal slope was observed (χ²=12.83, p = 0.025), with both the clinical high-risk (β=0.21, 95 % CI: 0.06, 0.36) and first-episode psychosis (β=0.20, 95 % CI: 0.10, 0.31) groups characterised by flatter slopes than controls. The clinical stage groups and controls did not differ on any other cortisol measure. Several sample characteristics were significantly associated with diurnal slope effect sizes, but after accounting for clinical stage, only the association between mean age in cases and wake-to-evening diurnal slope retained significance.

Conclusion

Clinical high-risk and first-episode psychosis participants differed from healthy controls in the peak-to-evening diurnal cortisol slope. This measure has not been examined in these populations before, and its potential predictive and prognostic utility for psychotic disorders merits further investigation.