The first aim of the study was to investigate whether CHR patients responded differently to a social stress paradigm compared with healthy controls. The second aim was to assess whether a 7-day treatment with CBD would attenuate the acute effects of exposure to social stress.

As expected, we found that CHR patients under placebo treatment (CHR-P) had a blunted cortisol response to the Trier Social Stress Test (TSST) compared with healthy controls (HC). Blunted cortisol response to experimental stress in CHR-P compared with HC reported here is consistent with previous reports in CHR individuals (Pruessner et al. 2013) and in those with established psychosis (Jansen et al. 1998). Our report extended these previous results in CHR participants by showing that blunted cortisol response to experimental stress was also associated with a greater psychological response in terms of anxiety and perception of public speaking as stressful in CHR-P compared with HC participants.

In line with our main hypothesis, we also found that CHR-CBD displayed intermediate levels of neuroendocrine (cortisol reactivity) and psychological [anxiety and perception of stress (negative self-statements)] response to experimental stress comparedwith CHR-P and HC. Collectively, these findings suggest that CHR participants under placebo displayed abnormal neuroendocrine and psychological responses to experimental stress compared with HC participants, and that 7-day treatment with CBD may potentially help partially attenuate these altered responses to experimental stress in CHR participants. However, one needs to be cautious in considering this interpretation, as the significant linear relationship across the 3 participant groups was mainly driven by the significant difference between HC and CHR-P on these measures. Therefore, further research is needed to investigate whether CBD may be used to influence the stress response in early psychosis. Differences in cortisol reactivity and anxiety response to stress in pairwise comparisons between CHR-P and CHR-CBD were not statistically significant, although this difference approached significance when change in perception of stress (negative self-statements) was compared between them. Future longitudinal studies in larger samples taking into account confounders are necessary to independently confirm whether CBD treatment can significantly attenuate altered responses to experimental stress in CHR participants relative to placebo treatment.

Nevertheless, our results provide preliminary evidence that CBD may affect the altered neuroendocrine as well as the psychological responses to acute stress in daily life in CHR patients. Therefore, in contrast to atypical antipsychotics, which have been shown to reduce elevated diurnal cortisol levels (Zhang et al. 2005; Cohrs et al. 2006; Mondelli et al. 2010), but not correct the blunted cortisol response observed in psychosis (Mondelli et al. 2010), CBD may potentially attenuate abnormalities in some of the main components of the acute stress response in the CHR state, future research using larger samples would be required to confirm this suggestion. These results are in line with previous research, which suggests that CBD may have anxiolytic (Bergamaschi et al. 2011b) and antipsychotic effects (Bhattacharyya et al. 2018; Leweke et al. 2012; McGuire et al. 2017).

The mechanism underlying the potential anti-stress effect of CBD is currently unclear, with multiple potential mechanisms being posited. Antipsychotic effects of CBD have been linked to its effects on levels of the endogenous cannabinoid anandamide (AEA) potentially by inhibiting its catalytic enzyme fatty acid amide hydrolase (FAAH). Recent preclinical work has also suggested that CBD may block the anxiogenic effects of chronic stress that was associated with a concomitant decrease in the expression of FAAH following CBD treatment (Fogaça et al. 2018). Anxiolytic effects of CBD were blocked by cannabinoid receptor (type 1 and type 2) antagonists but not by the serotonergic 5HT-1A receptor antagonist. However, other preclinical work suggests an effect of CBD on 5-HT_1a receptors (Fogaça et al. 2016; Bih et al. 2015; Russo et al. 2005) may underlie its anxiolytic effects.

To the best of our knowledge, this is the first study to have investigated the effects of short-term treatment with CBD on experimentally induced stress in the context of psychosis risk. We employed a well-validated stress induction task that has a long history (Kudielka et al. 2007a) of use for experimental stress induction. This experimental stress paradigm employed social evaluative stress, which may arguably be considered more similar to the kind of life-stress that has been linked with psychosis. While the use of a laboratory task may undermine ecological validity compared with other approaches such as experience sampling techniques, it does allow for more stringent control and standardised stress exposure. Furthermore, the task of speaking in public can be said to be life-like in that it is formed as part of a mock job interview, which is a situation that most people who experience it perceive as stressful. Further, in the healthy population sample, we recruited participants such that they were matched to the at-risk participants.

However, these strengths and the results presented herein need to be considered in light of certain caveats. In particular, we were only able to investigate a relatively modest sample size, which may have affected our ability to detect significant difference between CHR-P and CHR-CBD in pairwise comparisons as well as potential generalisability of these results. Previous studies of a similar nature have employed between 10 and 35 participants (Jansen et al. 1998; Jansen et al. 2000; van Venrooij et al. 2012). Related to this, it may be argued that the dose of CBD employed by us also affected our ability to detect an anxiolytic effect of CBD. A recent study found an anxiolytic effect of an acute dose of CBD at 300 mg but not at 100 or 900 mg (Zuardi et al. 2017), suggesting an inverted U-shaped dose response. However, a previous study (Bergamaschi et al. 2011b) also reported an anxiolytic effect during an experimental public speaking task in patients with social phobia, following a dose of 600 mg of CBD, a dose that we have employed.

Furthermore, in this study, we used venous blood sampling, which in itself can be quite stressful. However, in order to mitigate the effects of venepuncture, we used an atraumatic needle and employed an intravenous cannula that enabled us to avoid repeated venepuncture. While we cannot completely rule out the possibility that the stress of venepuncture may have added to the stress of participants, we do not believe that this would have confounded our results as such stress would have acted across all participants and only added to the social evaluative stress induced by the public speaking task, thereby contributing to how stressed participants felt overall. Furthermore, there was a gap of approximately 50 min between venepuncture and stress exposure, by which time any effects of stress from the venepuncture would have reduced substantially.

It is also important to note several additional factors may also have affected cortisol response to the Trier Social Stress Test (TSST), which were not controlled for in this study, including body mass index (Bose et al. 2009), night-shift work (Kudielka et al. 2007b), menstrual cycle and the use of contraceptives (Kirschbaum et al. 1999). However, gender distributions between the groups were not significantly different and participants were randomly allocated to the two treatment groups, making it likely that these effects would have affected the study groups to a similar extent. The cortisol levels could also have been affected by individual differences in the diurnal decline of the cortisol slope, but as the CBD and placebo group were randomised, it is reasonable to conclude that these differences would be equal between these groups. Furthermore, in a previous study (Collip et al. 2011), it has been shown that the diurnal slope did not differ between healthy controls and those at above average genetic risk for psychosis. The procedure we used for the TSST may also have been insufficient to distinguish between anticipatory and reactive cortisol. The baseline sample was taken 60 min before the onset of the stressor and the first sample after was taken 20 min after stressor onset. Previous studies have suggested that anticipatory cortisol levels may provide information when considering differences in mental health and protocols have been developed to capture this (Engert et al. 2013). Similarly, we measured cortisol levels up to 20 min after the stressful task concluded. It is agreed that cortisol levels reach a peak within 20–30 min after a stressor onset and that they return back to baseline levels within 90 min of cessation of the stressor (Fink 2000). The first sample after cessation of the stressor represents 20 min after the onset of the stressor; therefore, the + 10-min sample represents 30 min after and + 20 min represents 40 min after the onset of the stressor. It is therefore possible that the sampling is picking up anticipatory cortisol reactions and not reactive cortisol. However, the 0-min cortisol samples immediately after the TSST represents 20 min after onset of the stressor and cortisol levels in the healthy control group peak at this point and then continue to reduce, it is possible that this sample is still picking up reactive cortisol. It should also be noted that it has been found that high schizotypes have been found to have a delayed cortisol response (Walter et al. 2018). However, in this study, the peak of the cortisol response was at 15 min after the cessation of the TSST in both high and low schizotypes, and the delay was in the anticipatory rise. High schizotypes had no rise until 15 min after the cessation of the TSST whereas the low schizotypes had a rise immediately before the TSST commenced. While it is possible due to the sampling procedure, we may have missed the cortisol peak for the CHR participants, this is unlikely given the last sample is taken a full 40 min after the stressor commenced and no peak is seen in our data. Future research should follow the protocol outlined by Engert et al. (2013) and ensure adequate sampling is taken both in anticipation of the TSST and after the TSST to fully assess the increase and decrease of cortisol in response to stress. In any case, the conclusions of this study that CBD may have a potential to affect cortisol in response to stress is still a possibility.

It could also be argued that education level may affect how stressful the participants found the task, while education level was lower in CHR compared with HC, it was not different between the CBD and placebo groups, making it unlikely that our conclusions would be affect by this difference.

Notwithstanding its limitations, the present study provides a strong rationale for future studies to investigate whether CBD may have potential to mitigate the harmful effects of stress in the course of daily life by attenuating the altered neuroendocrine and psychological responses to acute stress in CHR participants.