Free Cortisol and Accuracy of Total Cortisol Measurements in the Diagnosis of Adrenal Insufficiency in Brain-dead Patients

• Adrenal insufficiency is common in brain-dead patients and contributes to hemodynamic instability and loss of potential organ donation

• Whether total cortisol measurement is accurate to diagnose adrenal insufficiency in these patients, who frequently have low albumin and corticosteroid-binding globulin, is not known

• In 42 patients with brain death, total cortisol correlated strongly with free cortisol and accurately diagnosed adrenal insufficiency in the face of low circulating binding proteins

DESPITE improved management of patients on the transplant waiting list, the mortality during this period in the United States, as in other countries, has remained as high as 15% in 2005, with 42% of patients waiting for more than 2 yr.1Thus, increasing the number of potential donors is a real challenge.

Hemodynamic instability occurs in 80% of the potential donors and is at the origin of 10% of lost potential donors because of refractory hypotension despite adapted fluid infusion and norepinephrine administration.2During brain death, disruption of the hypothalamic pituitary-adrenal axis may be one of the mechanisms that contribute to this hemodynamic instability. Actually, despite lack of evidence of effect on outcome and transplant organ viability and function, high-dose corticosteroid replacement therapy is recommended in case of hemodynamic instability in potential organ donors.3We have previously shown that adrenocorticotropic hormone (ACTH) stimulation followed 1 h later by a supplementary dose of hydrocortisone enhanced the systemic hemodynamic stability in brain-dead patients.4Nevertheless, corticosteroid function in brain-dead patients is still a matter of debate, being found as normal,5decreased,6or increased.7Then, before studying the effect of low or high dose of corticosteroid therapy on transplant outcome, it seems necessary to better characterize adrenal insufficiency (AI) in brain-dead patients. Most often, AI is diagnosed by measurement of serum total cortisol, meaning serum-free cortisol plus the binding-globulin fraction of cortisol. However, only free cortisol is responsible for the physiologic action of the hormone. Because more than 90% of circulating cortisol is protein-bound (corticosteroid-binding globulin [CBG] and albumin), it is reasonable to suggest that alterations in the serum binding protein concentrations could affect measured serum total cortisol concentrations,8and thus the interpretation of tests used to assess adrenal function. Moreover, a misdiagnosis of AI may have been made in some critically ill patients because of a low baseline and tetracosactin-stimulated total cortisol related to hypoalbuminemia, despite normal circulating free cortisol concentrations.8However, the pathophysiologic basis of AI may markedly differ between living critically ill patients and brain-dead patients who may suffer from both hypothalamic-pituitary ischemia and relative adrenal gland insufficiency, and thus these previous results8may not apply to brain-dead patients.

The aims of this study were to assess the exact prevalence of AI when diagnosed with free cortisol measurements and to assess the ability of serum total cortisol concentration measurement to diagnose AI in brain-dead patients.

All brain-dead patients 18 yr or older admitted to our intensive care unit for a potential multiple organ harvesting during a 12-month period were included prospectively. Before inclusion, brain death was diagnosed clinically with usual criteria (unresponsiveness to noxious pain stimuli, abolition of brainstem reflexes and apnea in the absence of hypothermia, metabolic or electrolyte disturbances and depressant drugs) and then was confirmed with cerebral angiography or electroencephalography, as previously described.3The inclusion of the patients in the study was performed before any discussion with families about potential organ donation. Patients were excluded if they had a history of hypothalamic–pituitary or adrenal disease or if they received corticosteroids during the previous 7 days or etomidate during the previous 3 days.9The study was approved by our ethics committee (Comité de Protection des Personnes de l'Hôpital Pitié-Salpêtrière, Paris, France) and was performed in accordance with the Declaration of Helsinki. This study was also conducted in accordance with French laws concerning multiple organ procurement. Waived informed consent was authorized because serum albumin measurement, diagnosis of adrenal insufficiency through serum total cortisol measurement, and hormone resuscitation are routinely performed in our unit for the assessment of brain-dead patients and because interventional procedures to evaluate the potential organ donor are authorized before the presumed consent could be verified. Free cortisol, ACTH, and CBG concentrations measurements were performed on blood withdrawn during routine biologic testing.

Serum ACTH and albumin were measured and the capacity of adrenal gland to respond to tetracosactin was tested. Tetracosactin contains the first 24 amino acids of the natural ACTH sequence and has the same physiologic properties, similar to cosyntropin. We use the standard short tetracosactin stimulation, which measured free cortisol and total cortisol concentrations before (t0, baseline cortisol concentration) and 30 min (t30, stimulated cortisol concentration) after an intravenous injection of 250 μg tetracosactin (Synacthène®; Novartis Pharma SAS, Rueil Malmaison, France). Tetracosactin-stimulated absolute increment was defined as Δ cortisol = tetracosactin-stimulated cortisol concentration − baseline cortisol concentration. Serum samples were centrifugated and stored at −80°C for later analysis.

We defined adrenal insufficiency as baseline free cortisol concentration less than 55 nM⁻¹and/or Δ free cortisol less than 55 nM⁻¹in response to 250 μg tetracosactin, measured at t30.8,10A free cortisol concentration more than 55 nM⁻¹defined “ACTH responder” patients, whereas a free cortisol concentration less than 55 nM⁻¹defined “ACTH nonresponder” patients.11Patients were divided into two groups according to their serum albumin or CBG concentration. Hypoalbuminemia was defined as a serum albumin concentration lower than or equal to 25 g·L⁻¹and a low concentration of CBG was defined as lower than 27 mg·L⁻¹in men and lower than 31 mg·L⁻¹in women.12 

Baseline plasma ACTH concentration was measured with the use of immunoradiometric assay kits (Immulite 2000, Siemens, Erlangen, Germany). When samples are withdrawn at 08:00 AM, values are considered as normal between 10 and 50 pg·mL⁻¹. The range of “normal” values in critically ill patients is unknown. The intraassay and interassay coefficients of variation were less than 9% and less than 10% respectively. Measurements of serum total cortisol were performed with the use of a standard radioimmunoassay. The intraassay and interassay coefficients of variation were less than 2% and less than 3%. When samples are withdrawn at 08:00 AM, values are considered as normal between 170 and 535 nM⁻¹. The range of “normal” values in critically ill patients is unknown. Serum free cortisol concentrations were measured with the use of equilibrium dialysis of undiluted serum samples followed by radioimmunoassay. The intraassay and interassay coefficients of variation were less than 10% and less than 12%, respectively. Values are considered as normal between 2 and 43 nM⁻¹in males and nonestrogen-treated females. The serum CBG concentration was measured by the Biosource radioimmunoassay (RIA; Lifescreen, Watford, Herts, United Kingdom) (normal range: males 27.1–52.3 mg·L⁻¹, females 31.0–53.4 mg·L⁻¹, and estrogen-treated females 64.4–116.0 mg·L⁻¹). Intraassay and interassay coefficients of variation were less than 6% and less than 9%, respectively. Measurements of serum free cortisol and CBG concentrations were performed at the Laboratoire de Radioanalyse, Institut de Physique Biologique, Faculté de Médecine, Strasbourg, France.

There was no a priori  sample size calculation: we based our study on the available patients during the 12-month period of the study. Data are expressed as mean ± SD for parametric variables or median (25–75 interquartiles) for nonparametric variables, or number and percentage and its 95% CI. Comparison of means was performed using the Student t  test for parametric variables and comparison of medians using the Mann–Whitney U test for nonparametric variables. Proportions were compared using the Fisher exact test method. Correlations between two variables were calculated using Spearman rank test method. The area under the receiver operating characteristic curve was used to summarize the accuracy of serum baseline, stimulated and absolute increment after tetracosactin stimulation testing total cortisol in diagnosing AI. Optimal cutoffs for each measurement were determined using the Youden index (sensitivity + specificity − 1) to maximize appropriate classification.13All P  values were two-tailed and a P  value less than 0.05 was considered significant. Statistical analysis was performed using NCSS 2007 software (Statistical Solutions Ltd., Cork, Ireland).

Forty-five consecutive brain-dead patients were screened in this prospective study. Three patients were treated with corticosteroids during the previous 7 days; none had received etomidate during the previous 3 days. Therefore, 42 patients, 18 women (43%) and 24 men (57%), 42 ± 17 yr old, were included in the study. The cause of brain death was trauma in 21 (50%) patients, cerebrovascular disease in 19 (45%) and anoxia in 2 (5%). The interval between accident and brain death was 85 ± 128 h.

The global incidence of AI defined by free cortisol was 83% (95% CI, 69–93%). Patients' patterns in regard to baseline free cortisol and responses to ACTH stimulation test are shown in figure 1. An ACTH concentration less than 10 pg·mL⁻¹was diagnosed in 25 patients (60%) and was not present in 17 (40%). Thirty-four patients (81%) were ACTH responders, whereas 8 (19%) were ACTH nonresponders (fig. 2). The incidence of no response to tetracosactin-stimulation test was similar in patients with low or normal baseline ACTH concentration (29%vs.  39%, respectively, P = 0.50). However, the interval between the causal accident and onset of brain death was shorter in ACTH nonresponders (37 ± 25 vs.  109 ± 152 h, P = 0.02).

Twenty-five patients (60%) had hypoalbuminemia and 17 (40%) did not. A slight but significant correlation between the interval between accident and brain death and albumin concentration was noted (r²= 0.14, P = 0.01). There was no significant difference in the incidence of AI between patients with low and normal albumin (80 vs.  88%, P = 0.68). Sixteen patients (38%) had a low CBG concentration and 26 (62%) did not. There was no significant difference in incidence of AI among patients with low and normal CBG concentration (88%vs.  81%, P = 0.69). Cortisol concentrations, with related albumin and CBG concentrations, are shown in table 1.

Baseline serum total cortisol concentrations were correlated with serum free cortisol concentrations in all patients (r²= 0.83, P < 0.001). There was no influence of albumin concentration on correlation between total and free cortisol (fig. 2). Patients with AI had lower baseline total cortisol concentrations than patients without AI (239 [127–393] nM⁻¹vs.  968 [572–1,245] nM⁻¹, respectively, P < 0.001) and lower tetracosactin-stimulated total cortisol concentration (833 [621–1,090] nM⁻¹vs.  1393 [1,238–1,464] nM⁻¹, respectively, P = 0.003), but there was no significant difference in Δ total cortisol (588 [430–782] nM⁻¹vs.  394 [302–58] nM⁻¹, respectively, P = 0.18). Individual data are shown in figure 3.

Areas under the receiver operating characteristic curves to diagnose AI were 0.94 (95% CI, 0.81–0.98) for baseline total cortisol concentration measurement, 0.83 (95% CI, 0.62–0.93) for tetracosactin-stimulated total cortisol, and 0.67 (95% CI, 0.38–0.83) for Δ total cortisol (fig. 4). The optimal cutoffs determined by the Youden index were 485 nM⁻¹for baseline total cortisol, 1,073 nM⁻¹for tetracosactin-stimulated total cortisol, and 199 nM⁻¹for Δ total cortisol. Baseline total cortisol measurement and tetracosactin-stimulated total cortisol measurement have similar sensitivities (89%vs.  74%, P = 0.22) and similar specificities (100%vs.  86%, P = 1.0). Combining test and Δ total cortisol did not increase the sensitivity or the specificity of the single baseline total cortisol test. The characteristics of each test are detailed in table 2.

Our study confirmed the high prevalence of AI in brain-dead patients when diagnosed with a reference test, i.e. , free cortisol concentration measurement. However, given that this test is not routinely available we have demonstrated that, in brain-dead patients as opposed to living critically ill patients, the diagnosis of AI may be easily and accurately made with baseline total cortisol, whatever serum albumin or CBG concentrations. Moreover, adding a tetracosactin test and then increasing the cost of the diagnosis of AI did not increase the accuracy of the diagnosis.

Because cortisol is approximately 90% bound to CBG, 6% to albumin, and 4% is unbound or free and because only free circulating cortisol is biologically active,14it has been suggested that decrease in binding proteins could lead to misinterpretation of low serum total cortisol concentrations, thereby overestimating the incidence of AI.8In critically ill patients, these alterations in protein concentrations are very common because of hemodilution or loss of proteins (for example in patients with hypercatabolism, malnutrition, burns, hepatic failure, and nephrotic syndrome). In septic shock, serum concentrations of CBG and albumin are decreased to approximately 50% of those observed in healthy control subjects.15In practice, total cortisol measurement has been found to have poor sensitivity and specificity and high variability in critically ill patients.8,16 

In our study, more than one-half of brain-dead patients had hypoalbuminemia and 38% had a low CBG concentration. In contrast with a previous study of various living critically ill patients without cerebral injury,8we did not find any significant difference in CBG concentrations in brain-dead patients with or without hypoalbuminemia. CBG and albumin are saturated more quickly with cortisol when CBG and albumin concentrations decrease, especially in a stress situation when cortisol production is increased and the proportion of free cortisol increases. In his situation, Hamrahian et al.  8have suggested that total cortisol measurement was not an effective way of diagnosing AI in critically ill patients. However, in our study, brain-dead patients with or without hypoalbuminemia had CBG concentrations similar to those in healthy volunteers (31.7 mg·L⁻¹and 26.8 mg·L⁻¹, respectively), and higher than those in critically ill patients (17.7 mg·L⁻¹in hypoalbuminemia group and 21.4 mg·L⁻¹in normal albumin concentration group).8Thus, in brain-dead patients, because of persistence of unsaturated CBG and despite hypoalbuminemia, both bound and unbound cortisol may have increased when cortisol concentration is increased after stress. On the other hand, in the presence of AI, cortisol secretion is decreased and so both forms of cortisol, bound and unbound, may have decreased proportionally. This observation may explain the preserved correlation between total baseline and free cortisol concentrations, independently of albumin or CBG concentration, even if this result may be also underpowered due to modest sample size combined with the high influence of several extreme scores. Moreover, in terms of cortisol concentrations, AI was more severe in brain-dead patients than in living critically ill patients: despite stress, baseline free cortisol concentrations were similar to those of healthy volunteers and decreased by a factor of 6 to 7 compared with that of living critically ill patients.8This difference may be explained by the pathophysiologic basis of AI in brain-dead patients, which results from both an absolute AI due to brain death itself (hypothalamic-pituitary injury) and a relative AI due to incapacity of adrenal gland to response to the stress. Then, they may have a more severe AI than living critically ill patients. In practice, total cortisol measurement was not accurate enough to diagnose relative AI but accurate enough to diagnose severe AI, as in the case of brain-dead patients.

In our patients, baseline total cortisol ranged widely, from 14 to 1,546 nM⁻¹. This high interindividual variability has been reported in most previous studies.7,17,18In normal subjects, Weitzman et al.  19have described the 24-h pattern of the episodic secretion of cortisol where peak total cortisol concentrations occur around the time of waking (10–20 μg/dl⁻¹, 276–552 nM⁻¹), whereas a minimum value (less than 5 mg·dL⁻¹, <138 nM⁻¹) is observed at night. However, in the critically ill patients, circadian rhythm is not maintained. Moreover, in the case of stressed patients, there is no absolute cortisol concentration that distinguishes an adequate from an insufficient adrenal response. Some authors have suggested that a baseline total cortisol concentration lower than 15 mg·dL⁻¹(414 nM⁻¹) was inappropriate for a critically ill situation.20,21Using receiver operating characteristic curves, we determined that the best cutoff of baseline total cortisol to diagnose AI in brain-dead patients was 485 nM⁻¹. Accordingly, at this cut-off, sensitivity and specificity of this assay were excellent, equal to 0.89 and 1.00, respectively. It should be noted that a short tetracosactin test was not useful for the diagnosis of AI because sensitivity of combined measurements was not higher than baseline total cortisol measurement alone (0.89 vs.  0.94, P = 0.67). We chose to use a high-dose tetracosactin stimulation (250 μg) instead of the low-dose stimulation (1 μg), even if the last one has been demonstrated to be more sensitive especially for the evaluation of secondary AI,22which seemed to be the major cause in our patients. However, because the low-dose tetracosactin stimulation is performed by diluting 1 ml tetracosactin in 249 ml normal saline and then administering 1 ml of the mixture, the preparation may be inaccurate. It is noteworthy that a recent meta-analysis had demonstrated that the low-dose stimulation is no more effective than the standard stimulation.23 

Several limitations of our study should be noted. First, we did not find any significant influence of the albumin (or CBG) concentration on AI incidence, but the number of our brain-dead patients was small and the result may be due to a lack of power. To have reached such a power, we should have included a number as large as 402 brain-dead patients (201 with a normal albumin concentration and 201 with a low albumin concentration) to finally demonstrate a clinically irrelevant increase in AI incidence from 88%, as we have previously found,4to 95%. Second, in critically ill patients, the free cortisol cutoff level in defining AI is still a matter of debate. In these patients, because baseline free cortisol concentrations would be expected to exceed tetracosactin-stimulated concentrations in healthy unstressed subjects, Hamrahian et al.  8recommended that a baseline serum free cortisol concentration of 55 nM⁻¹should be considered as a threshold that identifies patients at risk of AI. More recently, Annane et al.  10considered metyrapone stimulation test as the reference test to diagnose AI and found a best cutoff of 22 nM⁻¹for baseline free cortisol and 55 nM⁻¹for Δ free cortisol. However, it should be noted that in that study, free cortisol was not measured but calculated, using the Coolens equation.24Because we did measure free cortisol, we choose to define AI by baseline cortisol levels as did Hamrahian et al .8Because 55 nM⁻¹was the same cutoff for Δ free cortisol calculated by Annane et al. ,10we choose this cutoff to simplify the test. Nevertheless, we have shown that tetracosactin stimulation test was not necessary for the diagnosis. Currently, biologic cutoff levels remain a matter of debate, especially in critically ill patients.

In conclusion, our study demonstrates that total baseline cortisol measurement is accurate and sufficient to diagnose AI in brain-dead patients, even if low albumin or low CGB concentration occurs.

The authors thank the nursing team and the hospital coordinating the organ procurement team (Joelle Florit, nurse, Viviane Justice, nurse, Véronique Kokoszka-Biscompte, nurse, Isabelle Oilic, nurse) for their help with this study. We also thank David Baker, D.M., F.R.C.A. (Staff Anesthesiologist, Department of Anesthesiology, CHU Necker-Enfants Malades, Paris, France), for reviewing the manuscript.