Johan P. Schoeman, BVSc, DSAM, MMedVet, DECVIM-CA
Department of Companion Animal Clinical Studies, Faculty of Veterinary Science, University of Pretoria
Onderstepoort, South Africa
Several studies in human critical illness have shown a positive association between high basal serum cortisol and mortality1. In addition, human critical illness is characterized by multiple and complex changes in the thyroid axis2 where low concentrations of T3 and T4 have been shown to indicate prolonged illness and a poor prognosis3. In a human study testing several endocrine predictors, high basal cortisol and low thyroxine correlated best with outcome4. After many years of debate regarding the incidence and importance of 'relative adrenal insufficiency' (RAI), consensus is emerging. Initial studies originating from the Glasgow Royal Infirmary in 1982 which highlighted the phenomenon of adrenal failure in critically ill patients, seems to have been fundamentally flawed5. It was later shown that most of these patients received the sedative etomidate, which was found to suppress adrenal function6. It was therefore not surprising that this study obtained such a high success rate with corticosteroid replacement therapy, because these patients had been iatrogenically deprived of endogenous cortisol in the first place. Their excellent response to steroid replacement therapy has not been rivaled by any other study in human critical illness.
A landmark study conducted in the 90s, in which a positive response to physiological replacement doses of corticosteroids had been shown, has recently been questioned7. Firstly, the study performed post hoc subgroup analysis and showed a positive response only in a subset of patients. Importantly, these were septic shock patients which had relative adrenal insufficiency (based on a delta cortisol < 9 μg/dl [250 nmol/l]) and remained hypotensive after 1 hour of vasopressor therapy following prior adequate fluid replacement. There is convincing evidence that a subset of patients exists in which physiological glucocorticoid replacement therapy is beneficial--the problem lies in prospectively identifying these patients. After many conflicting studies, the most recent multi-centre Corticus study has shown that the ACTH stimulation test is not predictive of response to corticosteroid therapy. Moreover, this study showed no survival benefit of administering replacement doses of corticosteroids to septic shock patients; on the contrary, it showed a higher incidence of superinfection, including repeated septic shock in treated patients8. In the author's current opinion of the human critical care literature, absolute adrenal insufficiency is a rare event (except when caused by etomidate), while true relative adrenal insufficiency is a nebulous and poorly characterized entity, reserved for a subset of patients with vasopressor-resistant hypotensive septic shock.
Studies performed 60 years ago on adrenalectomized dogs showed the adrenal response to shock and critical illness to be essential for survival9. A study conducted in critically ill dogs with a variety of diseases failed to show any significant difference in basal cortisol concentrations between survivors and non-survivors and also failed to detect adrenal insufficiency10. Non-thyroidal illness in dogs is a well-known cause of euthyroid sick syndrome and the effects of such illness on lowering thyroid hormones have been well-described11. A recently published study performed on 33 critically ill dogs has purported to document RAI in septic patients12. It calculated delta cortisol values for each patient by subtracting basal cortisol from post ACTH cortisol. This delta cortisol value was tested for its association with the following outcome variables: mortality at hospital discharge, 28 day mortality and hypotension. No significant association was demonstrated between delta cortisol and outcome at the time of hospital discharge, but lower delta cortisol was significantly associated with hypotension. When delta cortisol was dichotomized into non-responders--a value below 3 μg/dl (83 nmol/l), and responders--a value above 3 μg/dl, it was shown that non-responders were much more likely to die. Potential critique on this particular study is:
Firstly, the study population had only one ACTH stimulation test performed, which was temporally separated from the outcome variables (i.e., discharge mortality and especially 28 day mortality) by many days.
Secondly, dogs that received a single dose of corticosteroids more than 72 hours before the ACTH stimulation test could potentially have been included in the study.
Thirdly, ACTH was administered intramuscularly. Ironically the lowest delta cortisol concentrations were demonstrated in hypotensive dogs. Both adrenal blood flow and the systemic uptake of ACTH could have been markedly impeded by the systemic hypotension--hence the worse the uptake or the poorer the adrenal blood flow, the poorer the adrenal response, potentially explaining the association between non-response and hypotension that was found.
Finally and most importantly, the basal cortisol concentrations of the survivors in comparison to the non-survivors were not reported. The concept behind this important omission has been addressed in a recent human publication and concerns the definition of relative adrenal insufficiency13. If we accept that the adrenal gland has a limit to the amount of cortisol it can produce, then it follows that dogs with the highest basal cortisol concentrations (the sickest dogs) would have the lowest delta cortisol values--and hence be classified as non-responders. The association of low delta cortisol with increased mortality could therefore be confounded by high basal cortisol.
In two different, comparatively homogenous models of critical illness and sepsis viz. virulent canine babesiosis and canine parvoviral diarrhoea, the author found the following: in 95 dogs with canine babesiosis, admission median basal cortisol concentration was significantly higher in non-survivors versus survivors; 482 nmol/L (IQR 456-562) versus 115 nmol/L (IQR 64-208)14. Within the group of more severely ill, admitted dogs, the median cortisol of the non-survivors was still significantly higher than that of the admitted dogs that survived; 482 nmol/L (IQR 456-562) versus 151 nmol/L (IQR 70-274) (Figure 1). Basal ACTH concentrations were also significantly higher in dogs that died compared to dogs that survived. ACTH stimulation tests performed in these dogs demonstrated lower delta cortisol in the non-survivors than in the survivors15. Although basal T4 and fT4 was significantly lower in non-survivors versus survivors (Figure 2), TSH concentrations were similarly low in both groups.
In 57 dogs with parvoviral diarrhoea, median basal day 2 and day 3 cortisol concentrations were significantly higher in non-survivors versus survivors16 (Figure 3). The above finding was confirmed in a follow-up study on a different group of 63 dogs with parvoviral diarrhoea, on which daily intravenous ACTH stimulation tests were performed17. In keeping with many human studies and the Burkitt canine study, these 63 dogs had significantly lower delta cortisol concentrations in the non-survivors than survivors on day 1 and day 2. However, on day 3 there was no statistical difference in delta cortisol between the survivors and non-survivors, mainly because of a relative reduction in basal cortisol in the non-survivors. In addition, although 44% of surviving dogs were classified as non-responders to the ACTH stimulation test on day 1, only 1% of them were still non-responders on day 3. Furthermore, all these dogs survived their illness without receiving glucocorticoid treatment, providing novel, albeit circumstantial, canine evidence supporting recent human findings of no survival benefit for corticosteroid supplementation in critical care8. There was a significant negative correlation between basal and delta cortisol, in part explaining the apparent adrenal insufficiency in very sick patients with high basal cortisol concentrations on days 1 and 2. These findings demonstrate the limitations of extrapolating and according patients non-responder status based on one-off ACTH stimulation tests. Basal hormone concentrations changed dramatically in these dogs during the progression of their illness. These changes and the different patient populations sampled at different stages of illnesses of varying severity could explain the disparate results in the few canine studies. Very different neuroendocrine paradigms prevail in acute versus chronic human critical illness and indications are that the same might be true for dogs18.
Both groups of parvo puppies also demonstrated significantly lower T4 and fT4 in the non-survivors versus the survivors on all three days19 (Figure 4). Interestingly, the endogenous TSH concentrations were higher in non-survivors on day 1 and 2, before also decreasing to below the limit of detection in the non-survivors. T4 and fT4 are thus more sensitive than TSH in predicting mortality and decrease earlier than TSH in canine critical illness. A study conducted in a mixed population of sick dogs also found significantly lower T4, fT4 and T3 in dogs that died compared to dogs that made a full recovery20. Low T3 concentrations have also been found in 56% of critically ill dogs and non survivors had significantly lower T3 than survivors21.
In conclusion, it would appear that similar to humans with critical illness, dogs with the most severe illness have the highest concentrations of pituitary adrenal axis hormones, while the pituitary thyroid axis hormones are markedly suppressed. We need large prospective studies, containing enough cases with adverse outcomes, so that multivariate analysis can identify the combined value of various hormone parameters in the prediction of mortality.
Click on a chart to see a larger view.
Figure 1. Cortisol boxplots in 3 outcome groups.
Figure 2. TT4 boxplot in 3 outcome groups.
Figure 3. Cortisol day 1 to 3 between groups with controls.
Figure 4. Total T4 days 1 to 3 between groups with controls.
References are available upon request.