The Absence of Adrenal Gland Enlargement during Septic Shock Predicts Mortality: A Computed Tomography Study of 239 Patients

• The use of low dose steroids in patients with septic shock appears to improve mortality in subgroups of patients, but identification of patients who will benefit is difficult

• Computed tomography identification of adrenal gland volume may be helpful in identifying septic patients who might benefit from steroid therapy

BUNDLE therapy of septic shock, a disease with a mortality rate ranging from 30 to 50%,1–5may include low-dose steroid.6–11Indeed, independent factors of poor outcome include hypothalamic–pituitary–adrenal axis impairment resulting in critical illness-related corticosteroid insufficiency.8,9,11–16Several studies have reported that daily treatment of certain patient subgroups presenting with septic shock with a prolonged (5–7 days) low dose (200–300 mg) of hydrocortisone offers potential benefits of reduced intensive care unit (ICU) length of stay or mortality.6,8,13,14,17However, one multiple center study failed to confirm the effect of such treatment on mortality18and the recent international consensus conference on critical illness-related corticosteroid insufficiency management in the ICU setting recommended considering low-dose steroid therapy only for vasopressor-dependent patients with septic shock.15Key patient selection criteria may include diagnosis of impaired hypothalamic–pituitary–adrenal axis or a poor response to endogenous steroids.10,11However, currently there is no absolute plasma cortisol concentration that distinguishes adequate and inadequate adrenal response.11,12,15More sophisticated tests are not routinely performed and are therefore not recommended to guide treatment.11,15In a pilot study using computed tomography (CT), we previously reported an increased adrenal gland volume, unrelated to adrenal hemorrhage, in patients with septic shock.19,20 

The aim of the current study was to describe adrenal gland volume and its accuracy as a prognostic value in septic shock. We hypothesized that the absence of adrenal gland volume enlargement could be associated with mortality in patients with septic shock.

This study was conducted from January 2005 to January 2009 in a 16-bed medical–surgical ICU. In accordance with French law, informed consent was not mandatory, given that this observational study did not modify current diagnostic or therapeutic strategies. The study design is represented in figure 1. Our study followed the Standards for Reporting Diagnostic Accuracy (STARD) recommendations concerning the report of studies of diagnostic accuracy.21,22This study was approved by the Montpellier University Hospital Institutional Review Board, Montpellier, France.

Consecutive patients were included if they presented with septic shock23and had been explored, within the first 48 h of shock, with an abdominal CT and measurements of basal and response plasma cortisol concentrations, before and after a short cosyntropin test with 250 μg cosyntropin intravenous injection (fig. 1). In our ICU, it is part of the standard of care to order a body CT scan in the early phase of septic shock to rule out any surgical or interventional radiology treatment for source control but also to assess the initial pattern of an acute lung injury. Thus, more than 70% of the CT scans performed combine both lung and abdominal CT. CT may be delayed or not performed when the cause of sepsis is obvious and/or its risks overwhelm its benefits. In this study, patients younger than 18 yr, who are pregnant, with pituitary or adrenal disease, or with a history of steroid use were excluded.

During the study period, septic shock management followed the 2004 Surviving Sepsis Campaign guidelines.4,5All patients were treated with intravenous hydrocortisone, 50 mg four times a day, during 5–7 days regardless of the result of the cosyntropin stimulation test.2 

Patients without sepsis admitted to the ICU (ICU nonseptic group, n = 15) and 40 consecutive ambulatory patients who had a normal virtual colonoscopy served as control subjects (control group, n = 40).

We collected data for sex, age, date, reason for ICU admission, and Simplified Acute Physiology Score II24for all ICU patients (tables 1 and 2).

No laboratory variables were collected in the control group. In the ICU groups, cortisol was assayed at baseline and 60 min after cosyntropin stimulation. Plasma cortisol concentration was measured using immunoradiology assay (SP2100, Beckmancoulter SAS, Roissy, France). Response to cosyntropin stimulation was considered to be positive if T60-T0 (Δ) was above 9 μg/dl.8,9In the septic shock group, we also recorded serum protein and albumin concentrations and the results of the blood cultures as well as plasma creatinine, bilirubin, lactate, platelets, leukocytes, hemoglobin, and procalcitonin (tables 1, 2).

Semiautomatic measurement of the adrenal glands by CT scan was performed as previously described by our group in a pilot study.20The measurement technique was reproducible as we reported in the pilot study a concordance correlation coefficient between the radiologist and the intensivist of 0.87 (95% CI, [0.76–0.93]) for the total gland volume measurement. The adrenal contour was semiautomatically traced by one radiologist (SN), who was unaware of the clinical status of the subjects, every 3-mm section. The radiologist traced the contour twice for each patient in random order. The scanner software automatically calculated the adrenal volume by summing the area on each slice. Adrenal gland volume between control and septic shock groups were compared.

The main endpoint was mortality in the ICU in relation to adrenal gland volume. The secondary endpoints were to analyze total adrenal gland volume in relation to septic shock, and to the cosyntropin stimulation test results.

Assumptions for the sample size calculation were based on previous studies on septic shock performed in our ICU.25Assuming a 40% mortality in the ICU in the group with adrenal gland volume above 10 cm³during septic shock, we calculated that 154 patients would need to be studied to detect a 50% absolute increase in mortality if the adrenal gland volume was less than 10 cm³; with 80% statistical power and a one-sided α value of 0.05.

Data are represented as mean ± SD or median and 95% CI. Continuous data were compared with Student t  test or Mann-Whitney U test when appropriate. Categoric data were analyzed with chi-square and Fisher exact test. The Pearson correlation coefficient was used to assess the correlation between the cosyntropin test results and adrenal gland volume measurement. Sensitivity and specificity of adrenal volume and basal/response plasma cortisol concentration curves to predict mortality were assessed using receiver operating characteristics (ROC) curves and compared with the Mann-Whitney U test. The cutoff point for ROC was predetermined by the statistics software to automatically minimize the mathematic distance between the ROC curve and the ideal point as a method to minimize misclassification of patients.21As previously reported, we also determined two cutoff points to detect volume, which predicts mortality with the best sensitivity and the best specificity.21Survival at day 28 was estimated by the Kaplan–Meier method and compared between groups with the log-rank test. A theory-driven logistic regression model was also developed in three steps. First, we identified all variables previously reported as associated with mortality in septic shock. Then, we assessed the colinearity among the candidate variables and selected the more clinically relevant variable in case of high correlation. Finally, we constructed a full multivariable model forcing all selected predictors (age, sex, Simplified Acute Physiology Score II, comorbidities, plasma creatinine and lactate, basal and random plasma cortisol concentration) and adding the total adrenal gland volume during septic shock. Two-way multiplicative interactions were systematically evaluated. The statistical significance of predictors was assessed using bootstrap resampling with replacement (200 samples). Odds ratio of dying was calculated along with 95% CI. All values were two-tailed and P < 0.05 was considered significant. Statistical analysis was performed with SAS v9 (Cary, NC) and checked by an independent statistician (GM).

During the study period, among 1,462 patients admitted to the ICU, 470 presented with septic shock. Among these 470, 266 either did not have a cosyntropin test or CT scan or had CT but not in the first 48 hours of septic shock. Thus, 184 consecutive patients meeting the inclusion criteria were analyzed (fig. 2). Table 1shows the main characteristics of the patients with septic shock (n = 184), the ICU patients without sepsis (ICU nonseptic group, n = 15) and the control group (n = 40). Table 2shows the comparison between survivors and nonsurvivors in the 184 patients with septic shock.

In the septic shock group, CT was performed 1 day [95% CI, 1–2 days] after the cosyntropin test, which was performed in the first 12 hours [95% CI, 8 to 16 h] after the administration of vasopressors for septic shock.

Total adrenal gland volume of the 184 patients in the septic shock group (12.5 cm³[95% CI, 11.3–13.3])was significantly higher in comparison with the 15 patients in the nonseptic ICU group (8.0 cm³[95% CI, 6.8–10.1]; P < 0.0001 for both comparison), and with the 40 patients in the control group (7.2 cm³[95% CI, 6.3–8.5], P < 0.0001 for both comparison). There was no significant difference in total adrenal gland volume between the nonseptic ICU and control groups (table 1).

Among the 184 patients with septic shock who had a CT scan and cosyntropin test, 100 were also explored before septic shock and had a medical checkup that could be considered as a baseline value (time before septic shock onset: 123 days [95% CI, 26–618 days]). During septic shock, adrenal gland volume increased in comparison with this baseline value (12.4 cm³[95% CI, 11.3–13.3]vs.  8.8 cm³[95% CI, 7.8–9.6]; P < 0.0001).

Furthermore among those 100 patients, 16 survivors were explored with a third CT scan after ICU discharge (time after ICU discharge: 100 days [95% CI, 44–214 days]) at the referent physician's discretion for a systematic follow-up or for any reason other than sepsis. Interestingly, in those 16 survivors adrenal gland volume returned to baseline value (fig. 3).

There was no evidence for a correlation between adrenal gland volume measurement during septic shock and baseline cortisol concentration (r = 0.09; [CI 95%, −0.11–0.29], P = 0.36) and a weak correlation with the Δ cortisol (r = 0.34 [CI 95%, 0.15–0.51], P < 0.001).

In septic shock, total adrenal gland volume during shock was 13.8 cm³[95% CI, 13.0–15.1] in the 125 survivors compared with 9.7 cm³[95% CI, 8.8–9.9] (P < 0.0001) in the 59 nonsurvivors. In the 100 patients who had a CT scan before and during septic shock, adrenal glands volume increased more during shock in the survivors (+33 ± 24%) than in the non survivors (+8 ± 7%) (P < 0.0001) (fig. 4).

In the 184 patients with septic shock, ROC analysis identified a total adrenal gland volume of 10 cm³as the best predictor of death with a sensitivity of 70% [95% CI, 56–81], a specificity of 92% [95% CI, 86–96], a positive predictive value of 82%, an accuracy of 84%, and an area under the ROC curve of 0.84 [95% CI, 0.78–0.89] (fig. 5). Using this threshold, the likelihood ratio of death in the ICU was 8.7 times greater if the total adrenal gland volume was less than 10 cm³. To better describe the ROC analysis, we also determined the cutoff value that resulted in the best sensitivity (15 cm³allowing a sensitivity of 98% [95% CI, 91–99] but a specificity of 41% [95% CI, 32–50]) and in the best specificity (7.3 cm³allowing a specificity of 98% [95% CI, 93–99] but a sensitivity of 25% [95% CI, 15–38]).

The best cutoff value for baseline plasma cortisol concentration was 32 μg/dl with a sensitivity of 33% [95% CI, 20–47], a specificity of 79% [95% CI, 70–86] and an area under the ROC curve of 0.52 [95% CI, 0.44–0.59] and 2.7 μg/dl for response cortisol concentration with a sensitivity of 44% [95% CI, 30–60], a specificity of 73% [95% CI, 63–82] and an area under the ROC curve of 0.55 [95% CI, 0.46–0.63].

The area under the ROC curves was significantly higher for total adrenal gland volume than basal cortisol concentration (P < 0.001) and than peak cortisol concentration (P < 0.001). Figure 6shows the survival curves in the sepsis validation for total adrenal gland volume (≤ or more than 10 cm³) (fig. 6A), basal plasma cortisol concentration (≤ or more than 32 μg/dl) (fig. 6B), and response cortisol concentration (≤ or more than 2.7 μg/dl) (fig. 6C) and Δ cortisol (≤ or more than 9 μg/dl)9(fig. 6D). Total adrenal gland volume was strongly associated with 28-day survival (hazard ratio 5.0 [95% CI, 4.3–14.1]; P < 0.0001) (fig. 6A). Although Simplified Acute Physiology Score II, lactate, platelets, plasma creatinine, and total bilirubin concentration were associated with mortality on univariate analysis (table 2), multivariate analysis showed that adrenal gland volume was the sole independent risk factor of mortality at day 28 (table 3). Interestingly, 68% of the patients with Δ cortisol less than 9 μg/dl and a total adrenal gland volume of less than 10 cm³died before day 28, whereas only 14% of the patients with a total adrenal gland volume higher than 10 cm³and a Δ cortisol higher than 9 μg/dl died before day 28 (P < 0.0001) (table 4). Four representative patients are presented in figures 7 and 8.

In a multiple regression model including sex, age, Simplified Acute Physiology Score upon ICU admission, admission category, sepsis source, and total hydrocortisone dose as independent data and adrenal gland volume as the dependent data; male sex and total hydrocortisone dose were statistically associated with adrenal gland volume (table 5). There were 3 of 15 deaths (20%) at 28 days in the nonseptic ICU patient group. In this group, adrenal gland volume was not associated with 28-day mortality.

This study shows that, in septic shock, total adrenal gland volume measured by CT scan was an independent prognostic factor for 28-day mortality. Adrenal gland volume was nearly doubled in septic shock groups in comparison with the nonseptic ambulatory group, and was increased by 35% in comparison with the nonseptic ICU group. Second, in hydrocortisone–treated septic shock, the cosyntropin test result was poorly correlated with adrenal gland volume. Finally, by multivariate analysis, we report for the first time that total adrenal gland volume less than 10 cm³during septic shock may represent a mortality risk factor.

Adrenal glands volume was evaluated with a semiautomated organ volumetric technique by a radiologist (SN) by CT and dedicated software (Myrian®, Intrasense, Montpellier, France). This software is widely available and is certified for organ volume measurement. The use of semiautomated organ volumetry technique has become part of a routine process in radiology for various situations, such as liver or gastric surgery,26but volumetry has been poorly studied in intensive care medicine27–29and was used only once for adrenal gland volume measurement, in a pilot study.20Adrenal gland volume has been previously shown to increase in stress conditions such as depression30and Cushing disease.31 

In this study, adrenal gland volume was not correlated with basal cortisol but showed some correlation with cortisol response to cosyntropin. These findings suggested that morphologic assessment of the adrenal glands may provide additional information on adrenal function to improve the diagnosis of critical illness-related corticosteroid insufficiency, which remains challenging.11,12,15,32In practice, patients with adrenal gland volume less than 10 cm³had higher risk of death (fig. 5 and 6). In the current study, every patient with septic shock was treated with hydrocortisone. Multiple regression showed that a higher total dose of hydrocortisone was associated with a smaller adrenal gland (table 5). More than a very unlikely direct effect of a very short hydrocortisone treatment on adrenal gland,33this association may be interpreted as the sicker the patients, the longer they were treated with hydrocortisone and the smaller the adrenal gland volume.

The mechanisms leading to the increased volume did not include adrenal gland hemorrhage or necrosis. Furthermore, there was no sign of excess fluid loading between septic shock groups on CT (such as ascitis or bowel edema) and it is then unlikely that adrenal gland enlargement could be related to a gland edema. Interestingly, in 16 patients who had been evaluated before, during, and after recovery of sepsis, we were able to demonstrate that the morphologic changes in adrenal glands were reversible (fig. 3).

This study has some limitations. Although the external validity of this study may be limited because of its single center and observational design, we included all consecutive patients in whom a CT and a cosyntropin test were prescribed within the 48 h of septic shock onset, representing 40% of patients with septic shock (fig. 2). All patients with septic shock received hydrocortisone as part of routine management. Therefore, we could not analyze the effect of this treatment on both adrenal morphology and patient outcome or speculate on the adrenal gland volume in patients with septic shock who were not treated with hydrocortisone. In this first study showing the association between adrenal gland volume and mortality, we did not include patients who did not have septic shock or those acutely treated with steroids, which could represent a limitation to our study. Finally, although we clearly demonstrated a link between adrenal gland volume and mortality, our study was not designed to explain the potential pathophysiologic pathways. However, according to the gland density profile similar to normal parenchyma, the radiologist did not observe any sign of gland edema, hemorrhage or necrosis.

In conclusion, our findings show that in septic shock, CT of the adrenal gland may contribute to assess adrenal function and a total adrenal gland volume less than 10 cm³may identify high-risk patients. Further studies may evaluate the adrenal gland volume in association with the cosyntropin stimulation test in other populations at high risk of critical illness-related corticosteroid insufficiency such as brain dead34or multiple trauma patients35and whether assessment of adrenal gland morphology on CT may identify patients likely to respond to corticosteroids.

The authors thank Christian Bonnel, M.D., Medical Coordinator, Intrasense Company, Montpellier, France, for the technological support of CT volumetry (Myrian® Software, Montpellier, France) and Julie Carr, M.D., Fellow, Anesthesia and Critical Care Department, Saint Eloi University Hospital, Centre Hospitalier Universitaire Montpellier, Montpellier, France, for her English editing.