Abnormal long-term rhythm of cortisol secretion

The MDE-total group did not show differences in cortisol secretion levels in comparison with the control group, contradicting the notion that increased cortisol levels are a consistent neurobiological marker in depression as a whole.^{Reference Pariante43} In fact, the opposite was suggested in the largest subtype studied here, in that the NM-MDE group had decreased AUCg in comparison with the control group.

However, the reduced daily cortisol output in saliva was measured over only approximately 12 h, and thus these differences in AUCg do not indicate whether low cortisol levels may be considered a chronic phenomenon in individuals with depression. In fact, the HCC findings suggest that over the long term (3 months), cortisol levels are not altered in either MDE subtype. The presence of decreased AUCg together with normal HCC in the NM-MDE group suggests that there may be episodes of increased cortisol levels on some days or some time points not measured in this study, thereby creating cumulatively normal cortisol levels later detectable in hair. However, those potential episodes may also be sporadic, as otherwise the patients with NM-MDEs would have exhibited increased rather than decreased daily cortisol output (AUCg). It may be that environmental factors trigger sporadic episodes of increased cortisol output, since we found that no other confounding factor had a significant effect on short-term cortisol levels in the NM-MDE group.

A previous study has already shown a heightened response to the Corticotrophin Releasing Hormone (CRH) test in people with NM-MDEs, a response that may mimic the effect of environmental factors.^{Reference O'Keane, Dinan, Scott and Corcoran44} Moreover, we recently found that some environmental factors, specifically daily hassles, were indeed more commonly associated with one type of NM-MDE, atypical depression, in comparison with non-atypical subtypes.^{Reference Herane-Vives, de Angel, Papadopoulos, Wise, Chua and Strawbridge19} In the present study, environmental factors were also strongly associated with NM-MDEs (supplementary Table 1). Thus, it is possible that an increased reactivity to daily hassles might trigger sporadic episodes of increased cortisol output in NM-MDEs.

Alternatively, it is also possible that there are periods of increased cortisol release occurring during the night, which are undetected by the daytime salivary cortisol measures we took. Such periods may play a compensatory role for the lack of secreted cortisol during the day. However, it is also possible that nocturnal hypercortisolaemia could be related to certain depressive symptoms. Indeed, it has been shown that patients with insomnia have increased cortisol levels during the evening and first half of the night.^{Reference Vgontzas, Bixler, Lin, Prolo, Mastorakos and Vela-Bueno45} Furthermore, some authors have also found increased nocturnal cortisol levels in participants with depression, although this was in participants with melancholic features.^{Reference Steiger, von Bardeleben, Herth and Holsboer46} Either way, during the day or during the night, episodes of cortisol hypersecretion would need to be large enough or frequent enough to counter the short-term hyposecretion we found in our short-term measure in order to lead to the accumulation of normal, rather than decreased cortisol levels in hair.

Taken together, the unexpected and contrasting results found using HCC and AUCg may imply an intermittently hyperactive pattern of cortisol secretion in NM-MDEs. This may also mean that the main neuroendocrine feature for the largest subtype of depressive disorder, according to Zisook et al,^{Reference Zisook, Shuchter, Gallagher and Sledge11} Jarrett et al,^{Reference Jarrett, Schaffer, McIntire, Witt-Browder, Kraft and Risser9} and Parker et al,^{Reference Parker, Roy and Mitchell10} is an abnormal cortisol rhythm, rather than an absolute concentration alteration, as Peeters et al,^{Reference Peeters, Nicolson and Berkhof2} had also previously claimed using series of short-term specimens. In fact, in one of our most recent studies we also suggested that this neurobiological alteration could describe the long-term pattern of cortisol secretion in atypical depression,^{Reference Herane-Vives, de Angel, Papadopoulos, Wise, Chua and Strawbridge19} since that depressive subtype showed the same results exhibited by the NM-MDE group here.

CAR

Although an increased CAR could theoretically have also contributed to the accumulation of normal HCC in the NM-MDE group, in the face of the low measured daily output, we did not find a difference in the measured CAR and DELTA values in this study, either between all patients and controls, or between the NM-MDE and M-MDE groups. On a whole, our results suggest that CAR and DELTA should not be used as illness biomarkers for this moderate depressive subtype. In contrast to our findings, some previous studies have indeed found an increased CAR in patients with depression.^{Reference Bhagwagar, Hafizi and Cowen47,Reference Vreeburg, Hoogendijk, van Pelt, DeRijk, Verhagen and van Dyck48} The differences in the number of saliva sampling points used to calculate the CAR could explain differing results. Whereas we used three sampling points (0, 30 and 60 min), Bhagwagar et al^{Reference Bhagwagar, Hafizi and Cowen47} used five (at 0, 15, 30, 45 and 60 min), after awakening, although cortisol levels were only higher at 15, 30 and 45 min post-awakening.

Vreeburg et al ^{Reference Vreeburg, Hoogendijk, van Pelt, DeRijk, Verhagen and van Dyck48} collected four saliva samples (at 0, 30, 45 and 60 min) post-awakening and again did not find cortisol variations at 0 min in comparison with controls. Thus, the normal CAR in our NM-MDE group may be explained by the few morning samples that we took in comparison with the Bhagwagar et al^{Reference Bhagwagar, Hafizi and Cowen47} and Vreeburg et al^{Reference Vreeburg, Hoogendijk, van Pelt, DeRijk, Verhagen and van Dyck48} studies, including, also the 0 min sample, which was the measure that both previous investigations did not find any significant variation in. It may also be possible to argue that if we had included an increased number of saliva morning samples around the 30 min delta we would had found increased, rather than normal CAR in our patients, which could, in theory, have also contributed to the accumulation of normal HCC in NM-MDEs. If that hypothesis was true, it would mean that out-patients with NM-MDEs present an abnormal endogenous cortisol reactivity featuring a shorter and delayed response.

In contrast to the above findings, Stetler & Miller^{Reference Stetler and Miller49} found a decreased rather than increased CAR in patients with mild and moderate MDEs in comparison with controls, using salivary sampling points that were the same in number and time as our study. It is notable that the Chilean patients in our study had significantly lower AUCg in comparison with their UK counterparts and they also had significantly lower HRSD-17 scores in relation to the UK participants with depression, without exhibiting a significant difference in the presence of melancholic symptoms (see supplementary Tables 1 and 2). Different site results confirm our NM-MDE and M-MDE group results.

Finally, it seems more likely that cortisol kinetic alterations related to changes in the secretion, metabolism and/or distribution of cortisol, rather than changes in its glucocorticoid receptor activity, could explain a potential long-term cortisol rhythm in MDEs.^{Reference Herane-Vives, de Angel, Papadopoulos, Wise, Chua and Strawbridge19} This is because a potential glucocorticoid receptor resistance in depression would lead to enduring effects,^{Reference Pariante, Pearce, Pisell, Sanchez, Po and Su50} rather than transient kinetic alterations such as the pattern of acute variations that we are speculating may be occurring.^{Reference Patrick, Mikos and McIntire51} Overall, our results may focus attention on the dysregulation hypothesis of depression,^{Reference Siever and Davis52} which suggests that depression may be better understood as a reflection of a relative failure in its regulation, rather than as simple increases or decreases in its neurobiological compounds.

Cortisol in out-patients with melancholic depression

Several factors may explain why we did not find increased cortisol levels in the M-MDE group on any of the measures studied here, in contrast to previous work suggesting that the melancholic and psychotic subtypes are commonly associated with hypercortisolaemia.^{Reference Carroll, Cassidy and Naftolowitz4–Reference Schatzberg, Garlow, Nemeroff, Davis, Charney, Coyle and Nemeroff6,Reference Schatzberg, Rothschild, Stahl, Bond, Rosenbaum and Lofgren53,Reference Sachar, Hellman, Fukushima and Gallagher54} First, previous studies were conducted in participants with more severe MDEs. It is clearly likely that increased cortisol levels are not a frequent finding with a less severe clinical presentation. We believe that the latter is an important point, since patients with an M-MDE in our study had, on average, depression of moderate severity; thus, our M-MDE group represented a less severe subgroup of patients with an MDE with classical features. This suggests that the severity of the episode, rather than its subtype alone, may be more critical in explaining cortisol level variations within depression, at least in out-patient settings. We believe that the aforementioned point is important, and that results might be different if we had studied a group of in-patients, who have not only more severe depression, but also more commonly present with melancholic symptoms.

Second, the use of hair specimens, as a measure of chronic cortisol concentrations is a relatively new tool and has only been used in a few studies to date. More studies are needed in order to corroborate our findings, as well as the exploration of other biological sources with the potential to measure cortisol levels over a longer-term period, such as fingernails.^{Reference Herane-Vives, Cleare, Chang, de Angel, Papadopoulos and Fischer55,Reference Izawa, Miki and Tsuchiya56} Indeed, by using fingernail specimens, we recently found increased fingernail cortisol concentration in out-patients with moderate MDEs in comparison with controls. This difference was, indeed, mainly driven by melancholic symptoms.^{Reference Herane-Vives, Fischer, de Angel, Wise, Cheung and Chua57} These contrasting results could challenge the validity of hair or fingernails as long-term measures of aggregate cortisol release, or suggest that these specimens accumulate cortisol differently. Finally, as was mentioned earlier, the M-MDE group was smaller than the NM-MDE group; consequently, the power of any comparison with the M-MDE group was also reduced.

Different effects of severity of depressive episode on cortisol levels

Of the two cortisol reactivity measures, only the morning DELTA cortisol showed a significant association with depressive symptom severity, in that increased early morning deltas of cortisol were linked with more severe episodes (supplementary Fig. 3). That association may better describe non-melancholic, rather than the classic form of depression, since the correlation was significant only in the former subtype (supplementary Fig. 3). Our findings contrast with those of Pochigaeva et al^{Reference Pochigaeva, Druzhkova and Yakovlev24} who did not find any association between the HRSD-17 score and a single morning blood cortisol measure, although they took the sample after an average of 60–80 min post-awakening, by which time physiological cortisol reactivity has already passed. The link between severity and DELTA in the NM-MDE group suggest that those early morning deltas of cortisol may be endogenous, rather than reactive to exogenous factors, and imply that the shape of the CAR is also different in more severe NM-MDE.

The positive association between DELTA and HRSD-17 score imply that the CAR in the NM-MDE group may be not only shorter and delayed, but also more pointed in more severe depression. Indeed, whereas the two previous studies^{Reference Bhagwagar, Hafizi and Cowen47,Reference Vreeburg, Hoogendijk, van Pelt, DeRijk, Verhagen and van Dyck48} that found increased CAR in comparison with controls used more saliva samples in their studies in comparison with us, they, at the same time, did not find an association between depression severity scores and CAR. It is quite likely, then, that an increased number of saliva samples better describe the CAR in moderate NM-MDEs, but, with severe non-melancholic depressive episodes the CAR shape becomes more pointed, mainly explained by an increase level of cortisol 30 min after awakening in comparison with the flat level at 0 min. Thus, conversely to moderate NM-MDE, it does not appear that an increased number of post-awakening saliva samples would provide further explanation of the CAR shape alteration in severe NM-MDE.

Furthermore, Adam et al,^{Reference Adam, Doane, Zinbarg, Mineka, Craske and Griffith58} also suggested there is a predictive role of the delta, using a measure of cortisol reactivity similar to our DELTA: wake-up plus 40 min minus wake-up cortisol level. They found that higher baseline DELTA was associated with a significantly increased risk of developing depression at follow-up. Nonetheless, Pruessner et al,^{Reference Pruessner, Hellhammer, Pruessner and Lupien59} found that a higher CAR was associated with more severe depressive symptomatology, although this was using self-reported depressive symptoms in a sample of healthy people; accordingly, this association was only demonstrated in the non-clinical range of depression.

The association between severe NM-MDEs and DELTA suggest that those episodes may be sharp and brief, such as spikes, rather than large and/or intense events, as we suggested for its moderate form. Indeed, the integrated measure of cortisol reactivity (CAR) was not associated with severe NM-MDEs, unlike the integrated measure of cortisol secretion (AUCg) which was associated with moderate NM-MDE. This could be explained by a hyperreactive 30 min level exhausting the physiological reactivity of the HPA axis, leading to a hyporeactive 60 min level, as Bhagwagar et al^{Reference Bhagwagar, Hafizi and Cowen47} found, or by a hypoactive axis at 0 min post-awakening followed by an overreacted 30 min level, as Vreeburg et al^{Reference Vreeburg, Hoogendijk, van Pelt, DeRijk, Verhagen and van Dyck48} found.

Not only DELTA, but also HCC showed significant associations with HRSD-17 scores among patients with severe NM-MDE. However, HCC was linked to reduced depression severity scores. Recently, Pochigaeva et al,^{Reference Pochigaeva, Druzhkova and Yakovlev24} supporting our hair results, indicated that participants with more severe MDEs were associated with decreased long-term cortisol levels. It may be possible that those brief and sharp spikes of cortisol reactivity might also explain why severe NM-MDEs were associated with decreased, rather than normal HCC. In effect, and contrary, again to its moderate form in which the early morning DELTA of cortisol did not differ from controls, in severe NM-MDEs, those potential sharp spikes may occur often enough in order to explain why DELTA was positively associated with HRSD-17 scores. But, at the same time, they do not seem to release enough cortisol to accumulate a normal concentration in hair. By contrast, not only might they exhaust the physiological reactivity response, but they may also exhaust the adrenal gland's ability to secrete enough cortisol during the rest of the day. However, the significant association between decreased HCC and severe M-MDEs is unlikely to be explained by the potential adrenal exhaustion caused by spikes of cortisol reactivity, as there was no corresponding association between DELTA and HRSD-17 in M-MDE. Again, we do not believe that this result is likely to be clinically significant, as the severity of depression in the smaller M-MDE group does not seem to represent the classic form of the disorder.

Limitations

It may be seen as unusual to find a group of patients with M-MDEs who have a HRSD-17 of 17.6. It has been said that melancholic features are less likely to occur in milder than in severe MDEs.^{Reference Taylor, Abrams and Hayman60} However, the DSM does not include depression severity as a diagnostic criterion for an M-MDE. It highlights in its latest version that the severity of depressed mood cannot be considered as a distinct quality for melancholic mood (DSM-5). Furthermore, the author of the Newcastle scale said in 1965 that ‘the distinction between endogenous (melancholic) and neurotic (non-melancholic) forms of depression is a qualitative rather than a quantitative one’. Finally, it has been demonstrated that HRSD fails to distinguish melancholia.^{Reference Demyttenaere and De Fruyt61} Nevertheless, at the same time, in-patients who are melancholic have significantly higher HRSD scores than out-patients who are melancholic.^{Reference Stage, Bech, Gram, Kragh-Sørensen, Rosenberg and Ohrberg62} We were able to confirm that there was a variance of melancholic symptoms in our sample, after finding higher levels of melancholic symptoms as defined by the Newcastle scale within our melancholic depression group, and an extremely significant positive correlation between HRSD-17 and Newcastle scale scores (r = 0.65, P < 0.001). Having said that, the frequently observed association between melancholic symptoms and depression severity may in part occur because depression severity scales, such HRSD, overrate melancholic symptoms, whereas other depressive features, such as the reversed neurovegetative symptoms (hypersomnia and increased appetite) are mainly ignored. It is also important to note that there is a lack of data supporting the unequivocal validity of the DSM-5 specifier for melancholia, and suggestions of other criteria that may be discerning of this group of patients.^{Reference Parker, Hadzi-Pavlovic, Boyce, Parker and Hadzi-Pavlovic63} Thus, we must acknowledge that the results might be different if the Newcastle scale criteria or other methods were used rather than the DSM-5 specifier, and if a group of patients with more severe depression were studied, such as in-patients, who may present not only with more severe episodes, but with more melancholic symptoms.

The effect of age was significant when AUCg values were compared between the different subtypes of MDE studied here, using an ANCOVA test. The M-MDE group was significantly older than the NM-MDE group and we found that AUCg levels were lower in the NM-MDE group in comparison with the control group. Thus, the modulatory effect that age had on short-term cortisol levels did not affect our finding of decreased AUCg levels in NM-MDEs.

Linear regression models showed that severity in the whole MDE group was associated with decreased HCC and a tendency towards increased AUCg (supplementary Fig. 2). This might seem biologically difficult to explain, since both measures were designed to assess the same cortisol profile: the total cortisol output, albeit over different periods. In effect, if those results had reached statistical significance, this would mean that hair may not be very accurate for reflecting long-term systemic levels of cortisol. In fact, it could mean, for instance, that some short-term extraneous acute influences do affect cortisol levels when hair specimens are used, whereas it has been claimed elsewhere that one of main properties of hair when accumulating substances is its immunity to those influences.^{Reference D'Anna-Hernandez, Ross, Natvig and Laudenslager13–Reference Xie, Gao, Li, Qiao, Jin and Deng17} It is true that those potentially controversial results could theoretically be explained if an MDE presented with mild hypercortisolaemia during the day and an extreme hypocortisolaemia during the night, which ended up averaging mild hypocortisolaemic levels in hair. However, we should be wary of speculating too far as, ultimately, inspecting the histograms suggests that the observed differences in AUCg and HCC in severe MDEs are potentially driven by some outliers in our control group with high levels of cortisol.

Future directions

Taken together, our results suggest that severity, rather than different subtypes of MDEs should be used as the basis for MDE classification in out-patients. Future studies might investigate, short- and long-term cortisol levels in mild/moderate and severe MDEs in in-patients. Such a study should be conducted in particular in melancholic forms of MDEs, as our results, contrary to the large body of previous evidence,^{Reference Carroll, Cassidy and Naftolowitz4–Reference Schatzberg, Garlow, Nemeroff, Davis, Charney, Coyle and Nemeroff6} did not find increased cortisol secretion and/or reactivity in M-MDEs. Our study also suggests that a larger number of saliva samples should be taken in order to measure CAR in moderate NM-MDEs. Measuring its delta also seems to provide useful information about different severities of non-melancholic subtypes of depression.

Criteria for non-melancholic depression are less well defined than melancholic clinical features, and indeed, there is no operationalised definition of ‘non-melancholic’ depression per se. The non-melancholic group will be somewhat more heterogeneous and comprise patients with a number of subtypes of depression, including atypical, anxious-distress, peripartum onset and seasonal pattern as well as those with non-specific features. Therefore, the presence of several subtypes of major depression likely contributed to greater heterogeneity in the composition of this non-melancholic group. Future studies should use hair and saliva specimens to investigate cortisol levels in depressive subtypes. As we recently undertook such a study in atypical MDEs,^{Reference Herane-Vives, de Angel, Papadopoulos, Wise, Chua and Strawbridge19} the next step would be to conduct the same study in people with anxious-distressed MDEs. Such a study may reveal the effect of anxiety on cortisol levels in MDEs. Future studies might also investigate whether an abnormal long-term pattern of cortisol secretion is also a specific finding in unipolar or bipolar depression.

It is important to state that the notion of an abnormal long-term rhythm of cortisol secretion is only a hypothetical explanation, as we did not observe any of these potential episodes of increased cortisol release. Yet, based on the specimens that we have for measuring cortisol levels, the recording of a pattern of cortisol secretion in real time seems impossible to achieve. What appears to be more achievable in the near future, however, is the use of multiple specimens that accumulate cortisol over different time period in this disorder. For example, different lengths of hair or fingernail samples may be useful, particularly bearing in mind that recent studies suggest that fingernails may mainly accumulate cortisol levels when the level of this hormone is particularly high, such as during episodes of increased reactivity. This potential property of fingernails needs to be corroborated by associating, for instance, fingernail cortisol levels with the current validated specimens for measuring cortisol reactivity (CAR) and secretion (HCC and AUCg) levels. A longitudinal study design could also contribute by investigating the long-term pattern of cortisol secretion in MDEs.

Implications

In conclusion, our results suggest that severity, rather than different subtypes of MDEs should be used as the basis for MDE classification in out-patient settings. They also strengthen the thesis that the main neurobiological finding for non-melancholic forms of depression is a cortisol rhythm, rather than an absolute concentration, alteration. This alteration may be featured by sporadic large and/or intense episodes of hypercortisolaemia in moderate NM-MDEs and frequent, but brief and sharp early morning deltas in its severe form.