Overall Study Question
This trial was a prospective, randomized, placebo-controlled study conducted in 19 intensive care units in France from September 1995 to March 1999. The study was designed to assess if endocrine replacement therapy with hydrocortisone and fludrocortisone could improve 28-day survival in patients with septic shock and relative adrenal insufficiency.
Hospitalized adult patients were included if they had clinical characteristics of the systemic inflammatory response syndrome (SIRS) with suspected or documented infection and signs of hypoperfusion and organ dysfunction and met all 7 of the following inclusion criteria:
(1) temperature < 35.6°C or > 38.3°C;
(2) heart rate > 90bpm;
(3) presence of polymorphonuclear leukocytes or a positive gram stain or culture of a normally sterile body fluid, or clinical focus of infection, or a wound with purulent discharge, or other evidence of systemic infection;
(4) systolic blood pressure < 90 mmHg despite 1 hour of adequate fluid replacement, or dopamine > 5 mcg/kg/min, or any treatment with epinephrine or norepinephrine;
(5) urinary output < 0.5 mL/kg/h, or PaO2:FiO2 ratio < 280;
(6) mechanical ventilation; and
(7) arterial lactate > 2 mmol/L.
Patients initially had to be enrolled within 3 hours of the onset of shock, however a protocol amendment on July 18, 1996 removed arterial lactate as a criterion, and the enrolment window was extended to within 8 hours of the onset of shock.
Patients were excluded if they were pregnant, had evidence of acute myocardial infarction or pulmonary embolism, advanced cancer or AIDS infection, and contraindications or formal need for corticosteroids. A protocol amendment on June 19, 1997 also excluded patients receiving etomidate within 6 hours of randomization.
Patients were randomized to hydrocortisone 50mg IV every 6 hours plus 0.5mg of fludrocortisone via a nasogastric (NG) tube or matching placebos for 7 days. At inclusion, patients received a short corticotropin (ACTH) test using a 250mcg IV bolus of tetracosactrin, and were subdivided into “responders” if their 30 or 60 minute cortisol response was > 9 mcg/dL (250 nmol/L), or “nonresponders” if the cortisol response was £ 9 mcg/dL (250 nmol/L).
The primary outcome was stated as the 28-day survival distribution from randomization in the subgroup of patients who were ACTH stimulation test “nonresponders”.
Are the Results of the Study Valid?
1. Was assignment of patients randomized?
Yes. Concealed randomization was centrally performed using a computer-generated random number table, and participants were stratified into blocks of four patients. The randomization process was not stratified based on whether patients were “nonresponders” or “responders” to the ACTH stimulation test because the results of the test were not available until after the patient received the study drug. Thus, although patients did have an equal chance of receiving placebo or active treatment in the overall cohort, this was not the case within the subgroups analyzed as “nonresponders” or “responders”.
2. Were all patients who entered the trial properly accounted for and attributed at its conclusion?
Yes. During a 42-month period, 1,326 patients were screened, and 300 (23%) were enrolled. A screening log was kept and a study flow chart was provided that accounted for all of the patients screened and enrolled in the trial. None of the patients discontinued treatment or were lost to follow up.
3. Were patients, their clinicians, and study personnel ‘blind’ to treatment?
Yes. The randomization schedule was concealed, identical placebos were used, and all patients, medical, nursing, and pharmacy staff were blinded to the treatment allocation throughout the study period.
4. Were the groups similar at the start of the trial?
Yes. Overall, patients in the placebo and active treatment groups had similar baseline demographic, clinical, and prognostic characteristics. Of the entire cohort, 18% were “responders” to a 250 mcg ACTH stimulation test with a good prognosis based on their adrenal response profile. The remaining 82% of patients were nonresponders, with 73% classified as having an intermediate prognosis and 9% classified as having a poor prognosis based on their adrenal response profile.
Since the randomization process was not stratified by response to the ACTH stimulation test, baseline clinical or prognostic factors may not have been equally distributed between the placebo and active treatment groups within the “nonresponders” subgroup. In this subgroup, 5% more patients in the placebo arm h
5. Aside from the experimental intervention, were the groups treated equally?
Yes. There was limited information concerning the co-interventions used and whether they were mandated by the investigators.Since the randomization process was not stratified by response to the ACTH stimulation test, and the primary analysis was of a subgroup, we cannot assume that known or unknown prognostic factors or co-interventions were balanced between the placebo and active treatment groups within the “nonresponders” subgroup.Nonetheless, the amount of fluid administered, dosage of vasopressors, time to vasopressor treatment, and ventilatory support appeared similar between the placebo and active treatment arms within the “nonresponders” subgroup.
What were the Results?
1. How large was the treatment effect?
In the primary analysis of the “nonresponders” subgroup, the 28-d mortality was 53% in the corticosteroid group, and 63% in the placebo group (unadjusted ARR = 10.8% (-1.9% to 23.6%) calculated from a 2×2 table, p=NS; reported RR = 83% (66% to 104%), p=NS). Using a Cox proportional hazards model to adjust for differences between the groups in prognostic factors, the adjusted analyses showed a statistically significant 28-day mortality benefit (OR = 0.54 (0.31 to 0.97), (p=0.04), NNT = 7 (4 to 140; my calculation); HR = 0.67 (0.47 to 0.95), (p=0.02), NNT = 7 (4 to 56)). The reported NNT for the 28-day mortality endpoint was 7 (4 to 49).
In the entire cohort, the 28-day mortality was 55% in the corticosteroid group and 61% in the placebo group (unadjusted ARR = 6.4% (-4.8% to 17.6%) calculated from a 2×2 table, p=NS; adjusted OR = 0.65 (0.39 to 1.07), p=NS). The adjusted hazard ratio was also statistically significant (HR = 0.71 (0.53 to 0.97), p=0.03, NNT = 8 (5 to 90, my calculation)). The reported NNT for the 28-day mortality endpoint was 8 (5 to 81).
The benefit for the entire cohort was driven by the effect in the “nonresponders” subgroup, which comprised the majority of the patients in the trial. There was no mortality benefit in the “responder” group (adjusted OR = 0.97 (0.32 to 2.99)) and the widely skewed confidence intervals failed to rule out a clinically important harm from corticosteroids in this subgroup.
In both the “nonresponder” subgroup and the entire cohort, patients receiving corticosteroids had a shorter median time to withdrawal of vasopressor support. The unadjusted ARR for remaining on vasopressors during the 28-day period was 17.0% (4.3% to 29.8%) (NNT = 6 (3 to 23) in the “nonresponder” subgroup; ARR = 12.4% (1.1% to 23.6%), NNT = 8 (4 to 91) in the entire cohort).
There were no significant differences between the corticosteroid and placebo groups in the incidence of adverse effects possibly related to corticosteroids or vasopressors in either group at 21% vs. 22% and 3% vs. 2% respectively. No patients in either group discontinued treatment due to adverse effects.
2. How precise was the estimate of the treatment effect?
For the primary outcome of 28-day mortality in the “nonresponders” to the ACTH stimulation test, the ARR for the raw analysis crossed zero, and was not statistically significant. The confidence intervals (CI) are definitely skewed towards benefit in the treatment group however, and the lack of statistical significance from the raw analysis may be due to a lack of statistical power. The 95% CI for the adjusted analyses using the OR and HR were statistically significant as they did not cross zero, but were quite wide. Interestingly, the authors report the results of the analysis with the narrowest 95% CI in the abstract. Based on the article the author’s reference for the NNT calculation, it appears that their calculations of the NNT and their 95% CI are incorrect.
Will the Results Help Me in Caring for My Patients?
1. Can the results be applied to my patient care?
Yes. In a very select proportion of critically-ill septic shock patients who are “nonresponders” to a 250 mcg ACTH stimulation test, the combination of hydrocortisone 50 mg IV q6h and fludrocortisone 0.5mg NG daily given within 8 hours of the onset of septic shock may shorten the time to withdrawal of vasopressor support and may reduce 28-day overall mortality.
2. Were all clinically important outcomes considered?
Although this trial was underpowered to show a difference in hospital and 1-year mortality, these endpoints are reported, and there was a strong trend toward improved survival in both the “nonresponders” and the entire cohort. Although adverse effects were prospectively evaluated, it would be very difficult to determine which adverse effects were attributable to the study drug, and the utilization of a validated adverse drug reaction scoring system would have been beneficial. Furthermore, the incidence of hyperglycemia, which can adversely affect outcome, was not reported in the trial.
3. Are the likely treatment benefits worth the potential harms and costs?
Although corticosteroid therapy shortened the time to withdrawal of vasopressor support, this is a surrogate endpoint and may not be of clinical benefit in patients tolerating vasopressor therapy. However, this is the first trial to show a reduction in 28-day overall mortality in a subgroup of ACTH stimulation test “nonresponders”, with no detectable increase in adverse effects or significant drug acquisition costs associated with therapy. The safety of corticosteroid use in ACTH stimulation test “responders” needs to be confirmed in a larger trial.
Although many specific anti-sepsis therapies have been studied over the past four decades, until recently, no therapy has reduced overall mortality in patients with severe sepsis or septic shock. Activated protein C is indicated in the U.S. for the reduction of mortality in adult patients with severe sepsis (i.e. sepsis associated with acute organ dysfunction) who have a high risk of death (as determined by APACHE II score). Although the NNT to prevent one death at 28 days with activated protein C is 15 (9 to 45), there is a strong trend towards increased serious bleeding (3.5% vs. 2.0%, p=0.06), and the acquisition cost based on a 4-day course of therapy for a 70-kg patient is ~$6800 US. More recently, intensive insulin therapy for aggressive control of blood glucose in mechanically-ventilated surgical ICU patients has also shown a reduction in 12-month mortality (4.6% vs. 8.0%), ARR = 3.5% (1.1% to 5.9%), NNT = 29 (17 to 91).
Two meta-analyses of short-course, high-dose corticosteroids failed to show a benefit on overall mortality in septic shock, and a strong trend towards increasing mortality. Two small published randomized, controlled trials (RCT) and one unpublished abstract of a small RCT in patients with septic shock have shown that early or late treatment with hydrocortisone 300mg/d given either as intermittent bolus doses or by continuous infusion for at least 5 days improves the blood pressure response to vasopressors and shortens the time to withdrawal of vasopressor support.
The current trial confirms the hemodynamic benefits of corticosteroids in a very select proportion of critically-ill patients, and is the first corticosteroid trial to show a reduction in 28-day overall mortality. The results of this trial suggest that a 250mcg IV ACTH stimulation test should be performed within 8 hours of the onset of septic shock in all patients meeting the inclusion criteria from this trial. This should be followed immediately by the administration of hydrocortisone and fludrocortisone. The results of the ACTH stimulation test will identify the subgroup of “nonresponders” who may achieve a survival benefit from continuing this combination therapy for 7 days.
There are concerns over the non-stratified randomization strategy in this study, and the primary analysis was of a subgroup of “nonresponders” in which there may have been differences in baseline characteristics, prognostic variables or treatment factors between the patients who received corticosteroids and those who received a placebo. Larger, appropriately randomized, adequately powered studies are required to confirm the efficacy of corticosteroid replacement in all septic shock patients and ACTH stimulation test “nonresponders”, and to rule out a potentially detrimental effect of routine corticosteroid treatment in ACTH stimulation test “responders”.
Annane D, Sebille V, Charpentier C, Bollaert PE, Francois B, Korach JM, Capellier G, Cohen Y, Azoulay E, Troche G, Chaumet-Riffaut P, Bellissant E. Effect of treatment with low doses of hydrocortisone and fludrocortisone on mortality in patients with septic shock. JAMA 2002;288:862-871.