Basic research has demonstrated that HPA axis activation in the face of overwhelming challenge, direct physical threat, or novel threat exposures occurs via direct connections from peripheral sensory inputs through thalamus to hypothalamus (Herman et al., 2003). Activation is also modulated by complex processing of indirect inputs through limbic-associated channels, which allow activation by psychological ‘anticipatory’ signals that predict exposure to threat (Herman et al., 2003). Learning shapes system responsivity through these indirect inputs, providing pathways for modulation by prior experience, access to safety or coping responses, or social support (Herman et al., 2003; Levine, 2000). Human research has shown that direct homeostatic threat, such as that associated with surgery, is a highly potent HPA activator (Gholami et al., 2004). In the psychosocial realm, lack of control over potential challenges and social evaluative threat appear to be particularly salient to HPA axis activation, whereas level of subjective distress itself is not (Dickerson and Kemeny, 2004).

The psychiatric HPA literature has evolved somewhat independently of this basic work, growing out of the original demonstration of cortisol escape from dexamethasone suppression in patients with endogenous depression (Carroll et al., 1981). This generated hope for a biological test for depression as a disease state and led to on ongoing search for specific, brain-based dysregulations that might contribute to HPA dysfunction and play pathophysiological roles in psychiatric disorders. Increasingly sophisticated biological probes, including corticotropin releasing hormone (CRH) stimulation tests, combined CRH/dexamethasone tests, ACTH stimulation, and metyrapone have been used to more precisely dissect the nature of HPA axis dysregulation in psychiatric disorders and identify abnormalities that might contribute to disease etiology or course (Heuser et al., 1994; Holsboer, 2001). This literature has generally assumed that the biological probes utilized tap into fundamental dysregulations within the HPA axis itself (Heim et al., 2008).

HPA axis activity in healthy humans, however, is sensitive to psychosocial cues, which raises the possibility that unrecognized but salient environmental aspects of experimental paradigms could influence responses to biological probes in studies of psychiatric patients. If so, then reputed HPA axis abnormalities could actually reflect differential sensitivity to contextual cues rather than to the specific probes utilized. It is possible that laboratory measured responses even to direct probes like a CRH stimulation test, could be influenced by individual differences in reactivity to the environment or psychosocial context even in the absence of CRH receptor sensitivity differences. Abnormalities detected in psychiatric disorders could then be due to disorder-related dysregulation in extra-hypothalamic modulatory circuits rather than to abnormalities within primary structures of the HPA axis. We have consistently demonstrated HPA axis abnormalities in patients with panic disorder, for example (Abelson and Curtis, 1996; Abelson et al., 1994; Abelson et al., 1996; Curtis et al., 1997), but our data suggest to us that these abnormalities may be due to reactivity to the testing paradigm rather than to altered basal activity within the HPA axis or altered sensitivity to specific biological probes.

Before addressing implications, it is important to acknowledge a need for replication. The observation of a cognitive intervention effect on ACTH only in subjects receiving CRH on first visit to the laboratory, but not those receiving it on second visit, left us with a small cell size for the key finding. Given potential skepticism that a brief, verbal discussion can significantly alter ACTH responses to a pituitary stimulant as direct and potent as CRH, replication with a larger sample is necessary. However, this effect was robust, despite the small sample, and quite consistent within the first visit CRH group (there were no outliers and variance is fairly small). It is also consistent with the clear and replicated effect of a similar intervention on the HPA response to pentagastrin, which also stimulates ACTH release through direct effect on the pituitary (Abelson et al., 2008). The finding is also consistent with a growing literature documenting the impact of psychological modulation on HPA axis activity across species (Ellenbogen et al., 2006; Gaab et al., 2005; Herman et al., 2003). Existing literature also supports the particular salience of novelty, control and coping to this system (Abelson et al., 2008; Dickerson and Kemeny, 2004; Herman et al., 2003), further supporting the possibility that an intervention focused on these factors might be able to modulate HPA responses even to direct, pharmacological activators. Replication efforts should address additional weaknesses. They should equalize the time spent with subjects, since in the current design the cognitive intervention took slightly more time to administer than the standard instructions. They should also include direct measurement of cortisol binding globulin (CBG), as there is growing recognition that CBG levels vary between sexes and across the menstrual cycle, and can influence results in HPA axis studies (Kumsta et al., 2007). We conducted additional analyses on these data to insure that group differences in CBG levels could not explain our results (Abelson, 2009) (see supplementary materials on-line), but future studies should include direct measurement and analytic control of this variable.

If replicable, this finding has important implications for HPA axis studies of psychiatric disorders. If manipulations of cognitive set can impact results in a CRH stimulation test, then pre-existing individual or group differences in cognitive set might also influence such results. If so, then abnormal responses to CRH in psychiatric patients could be due to disorder-related differences in cognitive set/expectancies rather than to differential sensitivity of CRH receptors. Such effects would likely be mediated via indirect input to the hypothalamus from limbic and prefrontal cortical areas (Jankord and Herman, 2008). Disorder-related dysregulation in such extra-hypothalamic areas might contribute, for example, to attentional or memory biases (Bradley et al., 1995; McNally, 1998) that influence how the experimental context or instructions are experienced or perceived. These in turn may impact modulatory input to the HPA axis and contribute to the detection of HPA axis abnormalities in psychiatric patients by altering reactivity to CRH even without changes in CRH or glucocorticoid receptor sensitivity. In a recent review of HPA data in Panic Disorder (Abelson et al., 2007), we concluded that apparent HPA abnormalities in panic patients, and the considerable inconsistencies in this literature, could well be due to a hypersensitivity to novelty cues in the panic patients, interacting with study variations in the extent of novelty exposure and the kind of preparation provided to subjects. The current data support this possibility by showing that sensitivity of the HPA axis to contextual factors is a normal phenomenon and that altering familiarity with context or preparedness to control/cope can impact HPA reactivity data even in healthy subjects and even when using a very direct probe like CRH.

The HPA axis has been studied more extensively in PTSD and depression than in panic, and evidence supports a more central role for HPA axis dysregulation in those disorders (Holsboer, 2001; Yehuda, 2002). However, even for PTSD and MDD, the notion that a simple defect within the HPA axis might provide a primary, etiologic pathophysiological factor is being questioned. Growing evidence suggests that hypercortisolemia and resistance to feedback inhibition in depression may not be directly linked to the disorder itself, but may at least partially be a consequence of early life trauma (Heim et al., 2008). Inconsistent HPA findings in PTSD may not be surprising if this system is highly context sensitive, since PTSD patients may have a specific deficit in context processing (Liberzon and Sripada, 2008). A recent, non-replication of low basal cortisol and feedback hyper-suppression findings in PTSD led to the conclusion that HPA axis abnormalities reported in PTSD likely reflect other risk factors rather than the presence of PTSD itself (Metzger et al., 2008). HPA axis dysregulation could be a consequence of interacting genetic and developmental factors that create deficits in adaptive capacities, reflected in poor coping in the face of environmental challenge, and detected in laboratory studies that expose subjects to novel and sometimes challenging experiences. Such biobehavioral adaptive deficits could create a general vulnerability to a multitude of psychiatric disorders, contributing to the pervasiveness of HPA abnormalities across disorders. Given this system’s role in shaping responses to environmental challenge, the wide range of environments that psychiatric patients have experienced over their developmental years, and the widely varying environments to which we expose them in experiments, it does not seem surprising that individual differences are seen in HPA studies and that the psychiatric HPA literature is full of contradictions and inconsistencies. The relevance of these results to psychiatric disorders must be tempered by awareness that we studied healthy subjects. Direct examination of HPA context sensitivity across clinical populations – comparing patients to controls in differing laboratory contexts and to themselves in novel versus familiar contexts such as their homes – is needed to better inform interpretation of HPA data from laboratory studies of patients. Heightened attention to relatively ignored aspects of experimental designs – such as prior experience in laboratory settings and the precise nature of instructions/preparation provided – might clarify some existing inconsistencies in the literature. Some documented abnormalities in some patient groups might “disappear” if they are studied in contexts that are highly familiar to them.

Our data cannot identify the psychological mechanisms through which cognitive set might alter responses to CRH. The impact of a verbal manipulation on ACTH response to first visit CRH suggests, however, that there are “top-down” pathways through which higher order cognitive/emotional inputs can modulate pituitary response to direct stimulation. The fact that this effect was only evident on first visit suggests a novelty-related, psychological amplification effect that augments response on first visit, with cognitive preparation able to override or block this amplification. Novelty alone cannot fully explain our results, however, since we did not see a smaller response to CRH when it was given on second visit under standard instructions. Also, the CI did not reduce perceived novelty and perceived novelty did not predict any HPA measure. However, novelty is a known activator of the HPA axis, perceived novelty was higher on first visit than second, and the CI specifically tried to reduce novelty by thoroughly preparing subjects for their laboratory experiences. Alternatively, the CI effect might be mediated by enhancement of sense of control, which may be less salient to subjects when they have had prior experience with the study procedures. The CI did in fact enhance sense of control, though this variable also failed to predict ACTH responses. The best subjective correlate of HPA axis response was perceived dangerousness of CRH, but this measure was not impacted by the CI. Measurement error may have contributed to the failure to find more significant subjective correlates. Our measure of perceived novelty has not been previously used and may not have reliably captured the relevant experience. Future work might include better established measures of novel experience. We suspect that as yet unidentified interactions between preparation, expectancies, uncertainty, and first visit novelty – perhaps involving unmeasured aspects of automatic processing (Ellenbogen et al., 2006) or anticipatory cognitive appraisal (Gaab et al., 2005) – are able to psychologically modulate HPA axis sensitivity when CRH is given on first exposure to the research setting. If replicable, further work will be needed to dissect the intervention components (novelty reduction, control, coping) and identify the relevant factors.

Placebo-day data support the relevance of laboratory novelty to HPA axis activation by showing higher cortisol levels on first compared to second visit. Cortisol levels declined over time on placebo day, regardless of visit, as would be expected on the basis of normal diurnal decline and accommodation to setting novelty. Intriguingly, this decline was disrupted by a late cortisol rise under standard instructions, not seen with cognitive intervention. This rise corresponded in time to a transition to less frequent sampling, which meant that the nurse no longer remained at the bedside between sample collections. This may be an example of the type of “context” change that is usually ignored but may be salient. Though speculative, we wonder whether this context shift was experienced differently by the two groups, enhancing uncertainty and perceived lack of control under SI, but having less effect on the more thoroughly prepared CI subjects. Sense of control in fact declined over time in SI groups on placebo day, but rose in CI groups. Sense of control is known to modulate cortisol release (Dickerson and Kemeny, 2004) and isolated manipulation of control directly modulates cortisol in our pentagastrin model (Abelson et al., 2008). Declining sense of control from baseline to last sampling point on placebo day directly predicted cortisol level at that last sampling point. These findings support the importance of sense of control in shaping HPA axis activity, corroborating that the HPA axis is sensitive to subtle but perhaps fairly specific psychological phenomena in healthy humans and supporting the hypothesis that unattended shifts in experimental contexts might influence HPA data in human laboratory studies.

Failure to see modulation of the cortisol response to CRH in the first visit group was surprising, given significant modulation of ACTH and robust cognitive intervention effects on cortisol in the pentagastrin model (Abelson et al., 2008). Dose and potency issues may be salient. CRH in this study produced a somewhat larger cortisol response (14 to 20 mcg/dL) than was seen in our recent pentagastrin replication study (8 to 14 mcg/dL). Even at the low dose utilized here, CRH may produce a maximal cortisol response that is less amenable to psychological modulation. Replication using a smaller dose is needed. In contrast to cortisol, the ACTH response to this dose of CRH is still quite far from its ceiling (Keller-Wood et al., 1983). We may thus be within a still dynamic segment of the ACTH range with this level of CRH stimulation, but having produced a robust and prolonged ACTH surge using oCRH, we may have pushed cortisol beyond its dynamic range. On placebo day, when cortisol release is not driven by exogenous CRH, we can see effects of novelty and cognitive manipulation. On placebo day, ACTH levels were also in a much lower range, close to the assay detection limit, where variability undermines the robustness of statistical analyses.

01

02