Use of nonprescription medications over the past 24 hours was assessed via self-report. Samples were stored at -20 °C, and rapidly thawed and centrifuged prior to assay using a highly sensitive enzyme immunoassay . All samples were assayed in duplicate with intra- and inter-assay coefficients of variation less than 10% and 15%, respectively.All analyses were performed using Stata Version 15 . The number of days between stress measure completion and cortisol collection could not be computed for ∼6% of the sample due to missing assessment dates. In such cases, missing values for the three time-lapse variables were imputed using the median number of days across the entire sample. The imputed data variables were used in all subsequent analyses. Ladder and gladder commands were used to identify transformations yielding normally distributed continuous variables. Subsequently, age, cortisol, daily stressor average distress scores, and life event exposure scores were log-transformed, daily stressor exposure scores were square-root transformed, whilst life event average distress scores did not require transformation. There were no transformations that could improve the distribution of the assessment time-lapse variable, therefore a five-level categorical variable was created cortisol collected before stress measurement; assessments completed on the same day; cortisol 1-10 days after stress measurement; cortisol 11-30 days after; and cortisol > 30 days after. Next, we examined correlations between salivary cortisol and sampling variables, namely, time of first sample collection, number of samples collected , and use of non-prescription medications in past 24 hours.
To remove the influence of relevant factors ,marijuana grow system cortisol values for the entire sample were regressed on sampling time, cough/cold medication use, and corticosteroid use to obtain standardised residuals. The resulting variable was used for all subsequent analyses. Group differences in demographic variables were examined using one-way analysis of variance, Kruskall Wallis, and chi-squared tests. To identify potential confounders, associations among demographic factors, cortisol, and psychosocial stress measures were examined using within-group Pearson’s correlations , biserial correlations , and chi-squared tests . Associations of group status with basal cortisol and psychosocial stressors were next examined, with adjustment for factors that were found in the above steps to be significantly associated with basal cortisol and/or any stressor in any group. Analyses of covariance , were employed for basal cortisol and continuous stressor measures, with estimated means and standard errors derived from these models. For trauma exposure, a logistic regression model was used to test the association with group status ; pairwise comparisons and adjusted trauma prevalence rates and associated SEs for each group were derived from the logistic model.To test whether stressor-cortisol concordance was moderated by time-lapse between assessments, correlations between cortisol and psychosocial stressors were examined within each time-lapse category. Linear regression models were used to test associations between individual psychosocial stressors and salivary cortisol in each group. Owing to multi-collinearity, each stressor was examined in a separate model; all models were adjusting for potential confounders identified in the above steps. To facilitate comparison of stressor-cortisol concordance across groups, from these adjusted models, we obtained standardised beta coefficients for each psychosocial stress measure and computed SEs for these coefficients [SEStβ = SEβ ].
Stβ coefficients for each stressor were then pooled using the ‘meta’ command to obtain an overall measure of stressor-cortisol concordance.Age was positively associated with life event exposure and life event distress in controls and all four CHR groups, with daily stressor distress in CHR remitted, symptomatic, and progressed groups, and with cortisol in controls, CHR remitted, and CHR converted groups . Female sex was likewise positively associated with life event exposure in controls, daily stressor exposure in the CHR remitted group, daily stressor distress in CHR symptomatic individuals, and with all five psychosocial stressor measures in the CHR progression of positive symptoms group. In contrast, ethnicity was not associated with any stress measure or cortisol in any group. With regards to psychotropic medication, antipsychotic use at baseline was negatively correlated with daily stressor exposure in the CHR remitted group and with life event exposure and trauma in the CHR progressed group, but positively associated with daily stressor distress in CHR individuals who later converted to psychosis; similarly, other psychotropic medication was negatively correlated with daily stressor exposure in the remitted group but positively correlated with daily stressor and life event distress variables in the converter group. Current cannabis use was associated positively with daily stressor exposure, life event distress, and trauma in the control group and with life event exposure in remitted and symptomatic groups. The above analyses identified the following baseline factors as potential confounders in the relationship between stress and cortisol: age, sex, current antipsychotic use, current other psychotropic medication use, and current cannabis use. However, owing to multicollinearity issues , all models included antipsychotic use only as a covariate, with sensitivity analyses performed using other psychotropic medication in place of antipsychotic use. We were additionally concerned that controlling for current cannabis use might obscure important relationships between stress and cortisol , given that recent evidence indicates that stress can precipitate cannabis use in healthy and clinical samples , and therefore included cannabis use as a covariate in sensitivity analyses only.
ANCOVAs and logistic regression indicated significant main effects of group status on basal cortisol and all stress measures after adjustment for age, sex, and antipsychotic use at baseline . Post-hoc comparisons indicated that only CHR convertors were characterised by elevated basal cortisol compared to controls , no other group differences were observed. With regards to continuous stress measures , all four CHR subgroups were characterised by significantly higher scores relative to controls; for daily stressor exposure only, symptomatic, progressed, and converted groups also showed significantly higher scores relative to remitted CHR youth. To confirm that the greater exposure to life events observed in CHR subgroups was not simply due to events that could be caused by illness, we additionally compared groups on exposure to independent life events and observed a significant main effect of group status . Post-hoc tests indicated that all CHR subgroups, except for the remitted group, reported increased exposure to independent life events compared to controls. Finally, childhood trauma was more common in all CHR groups compared to controls but did not distinguish among CHR subgroups. All results were largely unchanged when cannabis use was included as an additional covariate and other psychotropic medication use was additionally included in place of antipsychotic medication, with the exception that CHR remitted youth no longer showed significantly greater life event exposure compared to controls.In line with predictions, stressor-cortisol concordance varied according to the lapse-of-time between assessments : When acquired on the same day as saliva sampling, all stress measures showed significant, positive correlations with cortisol , except for life event distress scores which were positively correlated but not significantly. In contrast, stress measures were not significantly correlated with cortisol in any other time-lapse category except for life event exposure and cortisol which were positively associated when the lapse-of time was 31 days or longer. To account for the moderating effect of time-lapse between assessments, interaction terms were additionally included in subsequent regression models.The current study aimed to further characterise the nature of HPA axis abnormalities among individuals at-risk for psychosis by examining psychosocial stressors, basal cortisol,cannabis vertical farming and the concordance between these measures in a large sample of CHR youth categorised according to clinical status at the two-year follow-up. In line with hypotheses, all CHR groups were characterised by significantly greater psychosocial stressor exposure and distress relative to healthy controls; however, only those who converted to psychosis demonstrated elevated basal cortisol levels. In contrast to expectations, whilst CHR converters showed the greatest degree of stressor-cortisol concordance when pooled across stressors, confidence intervals substantially overlapped with the control group; moreover, the degree of concordance among CHR youth who remitted, remained symptomatic, or whose positive symptoms had progressed at follow-up was lower than that observed in the control group. After adjustment for potential confounders and correction for multiple testing, only CHR converters showed elevated basal cortisol relative to healthy controls. This finding cannot be simply attributed to greater stressor exposure or distress experienced by CHR converters relative to controls, as these features characterised all CHR subgroups. This elevation might instead reflect an amplification of the normative adolescent increase in cortisol secretion , or metabolic abnormalities [more common among CHR youth ], independent of stress exposure. Consistent with a recent meta-analysis , pairwise comparisons showed that basal cortisol levels did not distinguish CHR converters from CHR non-converters. Whilst this suggests that within the CHR population, baseline cortisol levels do not signal risk for psychosis transition, repeated measurement of cortisol is needed to determine whether longitudinal increases predict poorer outcomes in this group. Moreover, it should be assumed that at 2-year follow-up there are some false negative cases in the CHR non-converted groups , and thus differences between converters and non-converters may increase with a longer follow-up period.
We predicted that the degree of stressor-cortisol concordance, when pooled across stressors, would increase in parallel with the level of symptom expression at follow-up . Whilst the highest degree of concordance was indeed observed among the CHR converters, the control group was intermediate, and pooled beta coefficients in the three non-converted CHR subgroups were approximately zero . Moreover, confidence intervals for pooled stressor-cortisol concordance estimates for all CHR subgroups showed a high degree of overlap with the control group ; it has been proposed that for many effect sizes, confidence intervals overlapping by greater than 50% suggests that effect sizes are not significantly different . Thus, none of the CHR subgroups showed significant hyper- or hypo-responsivity of the HPA axis in response to psychosocial stressors encountered in the natural environment when compared to controls. Stressor-cortisol concordance was, however, substantially higher among CHR converters compared to all other CHR subgroup . This finding is consistent with the only previous study to examine the relationship between stressor-cortisol concordance and outcome status in CHR individuals: Labad and colleagues similarly reported a moderate-to-strong correlation between salivary cortisol and stressful life events among those who later transitioned to psychosis, but only a weak correlation in the non-transitioned CHR group . Overall, we found few significant associations between individual stressors and basal cortisol across all groups . Whilst this could be due to the HPA axis and/or stress measures employed, previous studies of at-risk youth which have used different measures have likewise found inconsistent associations between stress and cortisol . Similarly, in healthy subjects, correlations between self-reported stress and cortisol have not been observed . Although the exact mechanisms underlying HPA responsivity to stress are unknown , it has been demonstrated that there are individual differences in responsivity that are partially determined by genetic variants that modify the effect of acute and chronic stress/trauma on cortisol levels in healthy adolescents and adults and patients with psychosis . Thus, genetic and other vulnerability factors are likely responsible for the different patterns of association between stressors and cortisol that we observed across both individual stressor types and, when pooled across stressors, CHR subgroups.Despite the large overall sample size, individual CHR subgroups were notably smaller thereby reducing our ability to detect statistically significant associations between psychosocial stressors and cortisol. Conversely, as we did not adjust for multiple comparisons in our primary analyses examining stressor-cortisol concordance, some significant associations may have arisen by chance. However, we tested specific a priori hypotheses and were largely interested in the overall pattern of stressor-cortisol concordance rather than statistical significance. Moreover, we adjusted for a range of potential confounders which, had we not accounted for these variables, would have led to spurious associations. A further limitation is that a small proportion of participants , experienced a long delay between baseline assessment visits, which led to a large lapse-of-time between completion of psychosocial stressor assessments and cortisol collection. Including these participants in the analyses increased statistical power to test the moderating effect of time-lapse on stressor-cortisol concordance. One major limitation is that we examined only three stressor types. There are a range of other stressors relevant to psychosis that might conceivably impact on HPA axis function ; it is possible that examining a wider range of stressors might yield different patterns of stressor-cortisol concordance across CHR subgroups. Similarly, our findings are specific to basal salivary cortisol, other measures may have produced different results.