Peter C. Minneci, MD; Katherine J. Deans, MD; Steven M. Banks, PhD; Peter Q. Eichacker, MD; Charles Natanson, MD
Note: Drs. Minneci and Deans contributed equally to the creation of this manuscript; the order of their names is arbitrary.
Potential Financial Conflicts of Interest: None disclosed.
Requests for Single Reprints: Peter C. Minneci, MD, Critical Care Medicine Department, National Institutes of Health, 10 Center Drive, Building 10, Room 7D43, Bethesda, MD 20892; e-mail, email@example.com.
Current Author Addresses: Drs. Minneci, Deans, Banks, Eichacker, and Natanson: Critical Care Medicine Department, National Institutes of Health, 10 Center Drive, Building 10, Room 7D43, Bethesda, MD 20892.
Minneci PC, Deans KJ, Banks SM, Eichacker PQ, Natanson C. Meta-Analysis: The Effect of Steroids on Survival and Shock during Sepsis Depends on the Dose. Ann Intern Med. 2004;141:47-56. doi: 10.7326/0003-4819-141-1-200407060-00014
Download citation file:
Published: Ann Intern Med. 2004;141(1):47-56.
Previous meta-analyses demonstrated that high-dose glucocorticoids were not beneficial in sepsis. Recently, lower-dose glucocorticoids have been studied.
To compare recent trials of glucocorticoids for sepsis with previous glucocorticoid trials.
Systematic MEDLINE search for studies published between 1988 and 2003.
Randomized, controlled trials of sepsis that examined the effects of glucocorticoids on survival or vasopressor requirements.
Two investigators independently collected data on patient and study characteristics, treatment interventions, and outcomes.
The 5 included trials revealed a consistent and beneficial effect of glucocorticoids on survival (I2 = 0%; relative benefit, 1.23, [95% CI, 1.01 to 1.50]; Pâ€‰= 0.036) and shock reversal (I2 = 0%; relative benefit, 1.71 [CI, 1.29 to 2.26]; Pâ€‰< 0.001). These effects were the same regardless of adrenal function. In contrast, 8 trials published before 1989 demonstrated a survival disadvantage with steroid treatment (I2 = 14%; relative benefit, 0.89 [CI, 0.82 to 0.97]; Pâ€‰= 0.008). In comparison with the earlier trials, the more recent trials administered steroids later after patients met enrollment criteria (median, 23 hours vs. <2 hours; Pâ€‰= 0.02), for longer courses (6 days vs. 1 day; Pâ€‰= 0.01), and in lower total dosages (hydrocortisone equivalents, 1209 mg vs. 23â€‰975 mg; Pâ€‰= 0.01) to patients with higher control group mortality rates (mean, 57% vs. 34%; Pâ€‰= 0.06) who were more likely to be vasopressor-dependent (100% vs. 65%; Pâ€‰= 0.03). The relationship between steroid dose and survival was linear, characterized by benefit at low doses and increasing harm at higher doses (Pâ€‰= 0.02).
We could not analyze time-related improvements in medical care and potential bias secondary to nonreporting of negative study results.
Although short courses of high-dose glucocorticoids decreased survival during sepsis, a 5- to 7-day course of physiologic hydrocortisone doses with subsequent tapering increases survival rate and shock reversal in patients with vasopressor-dependent septic shock.
Despite effective antibiotics, septic shock remains the most common cause of death in the intensive care unit, incurring a mortality rate of 30% to 50% (1, 2). Several therapies targeting the upregulated inflammatory pathways of sepsis have been studied to improve survival. However, few therapies have proven beneficial (3-10). In the 1960s, preclinical studies reported that high doses of glucocorticoids in models of Escherichia coli and endotoxic shock improved survival. These studies prompted the initiation of human sepsis trials (11-13). Subsequently, more than 50 human trials have examined the role of high-dose steroid therapy in sepsis. These trials administered doses of methylprednisolone as high as 30 to 120 mg/kg of body weight over 24 hours. Because the reported results of these trials were inconsistent, there was little consensus on the appropriate use of steroids in patients with septic shock.
To clarify the treatment effects of high-dose steroids, 3 meta-analyses performed in the 1990s examined the more rigorously conducted randomized, controlled clinical trials of sepsis (7, 14, 15). The meta-analysis by Lefering and colleagues (14) incorporated 10 trials and found no overall beneficial effect of glucocorticoid therapy on mortality in septic patients (absolute difference in mortality rates between treatment and control groups, −0.2 percentage point [95% CI, −9.2 percentage points to 8.8 percentage points]). A second meta-analysis by Cronin and colleagues (15) examined 9 trials with variable effects (P = 0.02) and reported no evidence of a beneficial effect of high-dose steroids on mortality from sepsis (relative risk for death with treatment, 1.13 [CI, 0.99 to 1.29]). A third meta-analysis, performed by our group (7), examined the trials included in the previous meta-analyses. Nine trials, which were the same as those investigated by Cronin and colleagues (15), met inclusion criteria for that analysis (7). In this group of trials, we identified 1 study (16) as a statistical outlier that accounted for the variability reported by Cronin and colleagues. After exclusion of this outlier, our analysis revealed a homogenous group of 8 studies (P > 0.2) that demonstrated an overall increase in mortality associated with the use of high-dose steroids in septic patients (odds ratio of survival with treatment, 0.70 [CI, 0.55 to 0.91]; P = 0.008) (7). The increased mortality in these studies may have been due to the immunosuppressive effects of steroids, which led to more severe secondary infections (17-19). In response to these overall discouraging results, the use of high-dose glucocorticoids in septic patients decreased in the late 1980s and 1990s.
Recently, interest in examining the role of the adrenal axis in sepsis has been renewed. Briegel and colleagues (20) reported that septic patients have an attenuated response to corticotropin stimulation testing during their acute illness. Furthermore, Annane and colleagues (21) demonstrated that a high cortisol level and an attenuated response to corticotropin stimulation indicate relative adrenal insufficiency during sepsis that may increase mortality. On the basis of these findings, several clinical trials have been performed to determine whether administering glucocorticoids in dosages similar to the amount produced physiologically during a stressful state (that is, 300 mg of cortisol per day) affects outcome in septic patients. We performed the current study to update our previous meta-analysis and compare recent clinical trials with previous clinical trials of steroid use in patients with sepsis (22).
We searched MEDLINE for medical literature published from 1988 to December 2003 by using the following keywords: steroids and sepsis, steroids and septic shock, glucocorticoids and sepsis, glucocorticoids and septic shock, corticosteroids and sepsis, and corticosteroids and septic shock. Studies were included if they met all of the following criteria: randomized, controlled trial design; enrollment of adult patients who met criteria for sepsis or septic shock; and a primary end point, including either the discontinuation of vasopressor therapy or a change in survival comparing glucocorticoid treatment with a control group with or without placebo. Included studies must have administered similar treatments to both the control and steroid groups, with the exception of the administration of a predetermined glucocorticoid regimen. Criteria for sepsis or septic shock needed to be clearly defined in each study and be consistent with the American College of Chest Physicians and Society of Critical Care Medicine Consensus Conference (23) definition for sepsis (including documented site or strong suspicion of infection, temperature > 38 ° C or < 36 ° C, heart rate > 90 beats/min, respiratory rate > 20 breaths/min, and leukocyte count > 12 × 109 cells/L), severe sepsis (sepsis plus organ dysfunction; hypotension or hypoperfusion, including oliguria, altered mental status, or lactate acidosis), and septic shock (hypotension despite fluid resuscitation plus hypoperfusion abnormalities) (23).
Two investigators trained in critical care medicine independently reviewed the included studies by using a standardized protocol and data collection form. A third author trained in critical care medicine evaluated and resolved discrepancies. We collected data on patient characteristics, study characteristics, treatment interventions, and treatment outcomes. Abstracted data included the presence of sepsis, severe sepsis, or septic shock; type, dose, and duration of glucocorticoid administered; incidence and severity of secondary infections; response to corticotropin stimulation testing; the number of patients with shock reversal; and the number of patient deaths. We evaluated the quality of the included trials by assessing the method and adequacy of randomization, blinding protocols, completeness of follow-up, adherence to treatment protocols, and co-interventions or treatments to each group in the studies. Our primary goal was to compare the effect of glucocorticoid administration on survival in the recent studies with the effects reported in the previously analyzed trials (22). Since the glucocorticoid regimen differed among the trials, we converted all dosages to hydrocortisone equivalents (24).
Survival data were analyzed by using a Cochran-Mantel-Haenszel test to estimate the pooled effect of steroids (25). The similarity of the effect across studies was assessed by using a Breslow-Day test and reported with an I2 value (26, 27). When statistically significant heterogeneity of treatment effects was observed, studies were partitioned (for example, early vs. late studies) to decrease the heterogeneity of studies in a particular partition and increase the differences among the partitions, which can be seen when the I2 value is substantially lower in each partition as compared with the overall I2 value (28). One study increased the I2 value substantially in the set of early studies and was removed from all subsequent analyses. Partitioning variables were determined by regressing study characteristics (for example, steroid dose in first 24 hours) on mortality, specifically the log relative survival benefit (29). Regression was performed by using an inverse-variance-weighted restricted maximum likelihood random-effects method. When the regression was performed by using log steroid dose in the first 24 hours as the independent variable, 1 study was observed to be both a statistical outlier and influential. An indicator variable for this study was included in the regression analysis. Similar estimates of the slope associated with the effect of log steroid dose in the first 24 hours were observed when the influential study was removed and for early and late studies separately. A regression analysis that included control group mortality rate as an additional independent variable did not change the relationship between steroid dose in the first 24 hours and relative survival benefit. All pooled relative survival benefits are reported with associated 95% CIs by using a fixed-effects model. Random-effects estimates of survival were also calculated and reported. Statistically significant differences in characteristics between early and late studies were assessed by using analysis of variance (ANOVA) (when a weighted analysis was needed) or a 2-sample Wilcoxon test (when an unweighted analysis was performed). To analyze the different types of severity of illness scores used in the studies, we computed an effect size for each. This effect size was calculated by determining the difference between the mean steroid severity score and the mean control severity score, divided by the control standard deviation in each study.
The Warren G. Magnuson Clinical Center at the National Institutes of Health, Bethesda, Maryland, provided intramural funds for this study. The funding source played no role in the design, conduct, or reporting of the study or decision to submit the manuscript for publication.
Since 1988, more than 1300 articles on steroids and sepsis have been published. Five randomized, controlled trials, all published after 1997, met inclusion criteria and were included in our analysis (30-34)(Figure 1). Four of these studies were published manuscripts, and 1 study was reported in abstract form (33).
The 5 studies published after 1997 were randomized, double-blind, placebo-controlled trials (Table 1). Each study listed specific inclusion and exclusion criteria that were consistent with American College of Chest Physicians and Society of Critical Care Medicine Consensus Conference definitions of sepsis and septic shock (23). Each study used a severity of illness score (Simplified Acute Physiology Score [SAPS], Sequential Organ Failure Assessment [SOFA], or Acute Physiology and Chronic Health Evaluation [APACHE]) to compare treatment and control groups. Four of these studies enrolled only patients with vasopressor-dependent septic shock (30-33)(Table 2). In 4 of the 5 studies, 28-day mortality and treatment-related complications were the end points examined (30-32, 34). Three of these studies reported their results based on the patients' response to a corticotropin stimulation test (30, 31, 34). In the 4 studies that reported mortality data (30-32, 34), follow-up was complete and randomization was adequate, with no statistically significant baseline differences in severity of illness, underlying disease states, and demographic characteristics between treatment and control groups (Table 1).
In comparison with the 5 more recent trials, the 9 steroid sepsis trials published before 1989 used a wider range of inclusion criteria (from “severe infection” to shock) (Tables 1 and 2) (16-19, 35-39). Two trials were open-label studies (18, 37), and 1 study was reported in abstract form (36). Three studies reported statistically significant differences in baseline characteristics between study groups (19, 35, 38). One study enrolled only patients with cancer (17), and 2 studies administered 1 of 2 different steroid regimens to enrolled patients (16, 18). One study included both children and adults, was performed a decade before the previous studies, and had an unusually high percentage of patients (almost 50% of the enrolled patients) with meningitis as the indication for enrollment compared with the other studies (35). Another study was performed by 1 investigator for a longer period (8 years) than any other study and reported the lowest mortality rate with steroid treatment of any study (10%; next lowest mortality rate, 21%) (16). Furthermore, this study, which enrolled “septic shock patients consecutively admitted” at 1 institution over 8 years, reported the results of simultaneous prospective and retrospective trials during the study time period, suggesting unintentional selection bias during enrollment (16).
The more recent trials of steroids in sepsis focused on a more severely ill group of patients. In the glucocorticoid trials published after 1997, the control group mortality rate was higher (P = 0.06) and a higher percentage of patients were in shock (P = 0.15) and were receiving vasopressor therapy (P = 0.03) (Table 2). In comparison with the trials published before 1989, the trials published after 1997 delayed the onset of steroid therapy for a longer period after patients met enrollment criteria (P = 0.02), administered longer courses (P = 0.01) of lower-dose steroid therapy (P = 0.01), and were more likely to taper the steroid dose (P = 0.03) (Table 2).
Combined analysis of the trials published before 1989 (16-19, 35-39) and after 1997 (30-32, 34) demonstrated that the effects of steroids on survival were highly variable (I2= 70%; P = 0.001) with an overall non-statistically significant effect on survival. There was an interaction between the study's publication year and the treatment effects of steroids (P = 0.001). Trials published after 1997 revealed a consistent and overall improvement in survival associated with glucocorticoid use (I2= 0%;P > 0.2) (relative survival benefit, 1.23 [CI, 1.01 to 1.50]; P = 0.036) (Figure 2). The trials published before 1989 (16-19, 35-39) revealed a nonbeneficial treatment effect of steroids (relative survival benefit, 0.97 [CI, 0.89 to 1.04]; P > 0.2) that statistically differed from those of the trials published after 1997 (P = 0.02). Furthermore, as previously demonstrated, the effects of glucocorticoids on survival in the trials published before 1989 varied widely (I2= 75%; P = 0.001). This was secondary to 1 study that was a statistical outlier with methodologic differences in comparison with the other 8 studies (16)(Figure 2). Excluding this study revealed a consistent effect of steroids with an overall decrease in survival (I2= 14%; P > 0.2) (relative survival benefit, 0.89 [CI, 0.82 to 0.97]; P = 0.008) (Figure 2). Study size did not affect the treatment effects of steroids (P > 0.2). Thus, the effects of steroids were opposite in the trials published before 1989 compared with those published after 1997.
The relative survival benefits are shown with fixed-effects model and 95% CIs with glucocorticoid therapy in the sepsis trials. Both the fixed-effects estimate (to compare across studies) and the random-effects estimate (to generalize to other samples) of relative survival benefit are presented . (40). Meta-analysis of all 13 trials demonstrated variability (I2 = 70%) with no overall improvement in relative survival benefit (fixed-effects estimate, 1.01 [95% CI, 0.94 to 1.09]; random-effects estimate, 1.04 [CI, 0.90 to 1.20]). The effect of steroids in the trials published before 1989 compared with those published after 1997 significantly differed ( = 0.02). In the 4 trials published after 1997 (1 study did not report mortality data ), )the effect of steroids on the relative survival benefit was consistently beneficial (I2 = 0%) (fixed-effects estimate, 1.23 [CI, 1.01 to 1.50]; random-effects estimate, 1.19 [CI, 0.99 to 1.43]). The effects of steroids on the relative survival benefit in the 9 sepsis trials published before 1989 varied (I2 = 75%; fixed-effects estimate, 0.97 [CI, 0.89 to 1.04]; random-effects estimate, 0.97 [CI, 0.81 to 1.16]). Excluding 1 trial (16), which was a statistically significant outlier, yields a homogeneous group of 8 trials with a consistent harmful effect of steroids on survival (I2 = 14%; fixed-effects estimate, 0.89 [CI, 0.82 to 0.97]; random-effects estimate, 0.90 [CI, 0.80 to 1.02]). This excluded trial (16) had methodologic differences, including being performed by 1 investigator over an 8-year period and enrolling patients both prospectively and retrospectively. VA = Veterans Administration.
The total dose of glucocorticoids used in the trials published before 1989 was greater than that of studies published after 1997 (P = 0.01) (Table 2). Furthermore, a regression analysis performed on all trials demonstrated that as the dose of steroids increased, the relative survival benefit decreased linearly (P = 0.02), indicating that steroids were beneficial at lower doses and became harmful as the dose increased (Figure 3). One trial that methodologically differed from the other trials was overly influential on this analysis and was removed (35)(Figure 3).
The relationship between the dose of steroids administered in the first 24 hours after enrollment in a sepsis trial and relative survival benefit (black circle) is presented. There is a linear relationship (that is, the relative survival benefit decreases with high-dose steroids but increases with lower doses) ( = 0.02). One study (white circle) was overly influential in our regression analysis, was a statistical outlier ( = 0.001) compared with the other trials, and was therefore excluded (35). This study was performed a decade before all of the other trials, included children, and had a high percentage of patients with meningitis.
In the studies published before 1989, the mean control group mortality rate was lower than that in the studies published after 1997 (34% vs. 57%; P = 0.06) (Table 3). However, the linear relationship between control group mortality rate and the survival effect of steroids in the individual trials was not statistically significant (P > 0.2).
Trials performed after 1997 reported a consistent effect of glucocorticoids on shock reversal (I2= 0%; P > 0.2) (30-33)(Figure 4). Discontinuation of vasopressor therapy statistically significantly improved with the administration of steroids in 3 of the 4 studies. Only 2 of the 8 studies published before 1989 reported the effects of steroids on discontinuation of vasopressor therapy. These studies (18, 19) demonstrated opposite effects of high-dose steroids on shock reversal (P > 0.2 for both).
The relative benefit (95% CI) and the hazard ratio (with 95% CI) of shock reversal for the sepsis trials published after 1997 are presented. Of note, in 3 of the 4 studies, the discontinuation of vasopressor therapy with steroid treatment statistically significantly improved. In the fourth trial, the effect of steroid therapy on vasopressor discontinuation was similar to the effect in the other trials (I2 = 0%; > 0.2).
Three of the studies published after 1997 stratified their patient samples into responders and nonresponders on the basis of the results of a 250-µg corticotropin stimulation test (Table 3) (30, 31, 34). Two of these studies (30, 34) defined a nonresponder to this test as having a change in cortisol level less than 248.3 nmol/L (9 µg/dL) from baseline. The third study (31) defined a nonresponder as having a change in cortisol level less than 165.53 nmol/L (6 µg/dL) from baseline. Three of these trials reported mortality and 2 of these trials reported shock reversal separately for responders and nonresponders (30, 31)(Table 3). The treatment effects of steroids on mortality or shock reversal did not statistically significantly differ on the basis of this division in both the individual trials and when these trials were combined (P > 0.2 for all). The overall effect of steroids in these studies, which delineated responders and nonresponders, was to decrease mortality (61% in the control group vs. 51% in the steroid group; P = 0.04) and increase shock reversal (40% in the control group vs. 54% in the steroid group; P = 0.01) in all patients.
Severity of illness scores were reported in 4 of the 5 trials published after 1997 (30-32, 34). There was no difference in the severity of illness between the control groups and steroid groups in these 4 studies that could explain the beneficial effect of steroids (scoring system, SAPS II [55 in the control group and 55 in the steroid group (32)]; SAPS [14 in the control group and 14 in the steroid group (31)]; APACHE II [18 in the control group and 15 in the steroid group (34)]; SAPS II [57 in the control group and 60 in the steroid group (30)]; P > 0.2 for all studies). Severity of illness scores were not reported in any study published before 1989.
Four of the 5 steroid trials published after 1997 (30-32, 34) and 6 of the 8 steroid trials published before 1989 (17-19, 35, 38, 39) reported the incidence of secondary infections. In these trials, the median overall incidences of secondary infections were similar (trials before 1989: control group, 13% [range, 3% to 21%]; steroid group, 17% [range, 3% to 26%]; and trials after 1997: control group, 27% [range, 5% to 47%]; steroid group, 19% [range, 0% to 50%]). However, the trials published before 1989 reported more associations among steroid treatment and either an increased incidence of secondary infection in subsamples (17, 18) or more severe secondary infections (38) with increased mortality (19)(4 of 6 trials vs. 0 of 4 trials; P = 0.08).
Five randomized, controlled trials published after 1997 reported the effects of steroids in septic shock. Four of the 5 trials reported mortality data and demonstrated a consistent beneficial effect of glucocorticoids on mortality. In addition, 4 of the 5 trials reported shock reversal data and demonstrated a consistent beneficial effect of steroids on shock reversal. In contrast, the 9 trials published before 1989 demonstrated heterogeneous effects of steroids with an overall nonbeneficial effect. Excluding 1 trial that was a statistical outlier in this group (16) produced a homogeneous group of 8 studies that demonstrated a consistent harmful effect of steroids on survival in sepsis. The studies published after 1997 administered lower doses of glucocorticoids compared with those published before 1989. Furthermore, the relationship between steroid dose and survival was linear; this may indicate that the effects of steroids are dose-dependent during sepsis. Therefore, our meta-analysis suggests that lower, more physiologic doses of glucocorticoids reverse shock and confer a survival advantage to patients with established vasopressor-dependent septic shock.
The studies performed after 1997 administered physiologic doses of steroids in an attempt to provide replacement therapy for sepsis-induced relative adrenal insufficiency (30-34). In contrast, the trials published before 1989 used large doses of steroids to block the excessive inflammation of sepsis (17-19, 35-39)(Table 2). The harmful effect of the high-dose steroids administered in the earlier trials may have been caused by a pronounced immunosuppressive effect of the steroids. Although the numbers of secondary infections did not differ between the earlier and later studies, steroid treatment before 1989 was reported to increase the time to resolution of secondary infections (38) and subsequently increased the mortality from these infections (19). Studies published after 1997 that used lower, less immunosuppressive doses of steroids did not report an increase in the severity of secondary infections. These differences may partly explain the contradicting results of the trials published before 1989 compared with those published after 1997.
The timing of steroid initiation, duration of steroid administration, and differences in severity of illness of study samples may also account for the contradictory treatment effects of steroid therapy in these 2 sets of trials. The trials performed before 1989 administered shorter courses of glucocorticoids earlier in the patients' septic episode. However, in the more recent trials, steroid therapy was beneficial when administered to more severely ill patients with higher control group mortality rates who were in vasopressor-dependent shock for at least 2 hours. This therapy remained beneficial when started as late as 72 hours after the initiation of vasopressors. Moreover, these trials suggest that divided doses of steroids equivalent to 200 to 300 mg of hydrocortisone daily should be administered for a minimum of 5 days, followed by tapering over 5 to 7 days. Of note, we have previously shown a relationship between mortality in the control group and treatment effect in the preclinical and clinical trials of mediator-specific anti-inflammatory agents in sepsis (29). In this previous analysis, mediator-specific anti-inflammatory agents, such as anti-tumor necrosis factor antibodies, soluble tumor necrosis factor receptors, and interleukin-1 receptor antagonists, were more beneficial in septic patients as mortality in the control group increased (29). However, in our current analysis, we did not identify such a linear relationship between steroids and mortality in the control group. This may be secondary to an overwhelming influence of drug dose on treatment effect, an insufficient difference in mortality in the control groups among the studies, or a lack of power secondary to only 13 studies available for analysis. Thus, when physiologic doses of steroids are administered to patients with a wide range of control group mortality rates, differing effects on survival may still occur.
Response to corticotropin stimulation testing has been used to determine which septic patients should receive steroid therapy. Annane and colleagues reported that survival improved with steroid administration only in nonresponders to this test (30). However, their analysis revealed no statistically significant difference in the treatment effects of steroids between responders and nonresponders. Our analysis of the 3 recent trials that reported mortality and shock reversal data by corticotropin stimulation test results showed that the beneficial effects of steroid therapy did not statistically differ between responders and nonresponders (30, 31, 34). Therefore, unless further clinical sepsis trials demonstrate that responders do not benefit from therapy, steroids should be considered for all patients with vasopressor-dependent septic shock.
The relative survival benefit demonstrated with physiologic doses of glucocorticoids is similar to that of activated protein C, the only other immunomodulatory agent that improves survival in septic patients (relative survival benefit, 1.23 [CI, 1.08 to 1.43]) (41). However, activated protein C is associated with an increased risk for bleeding, may be harmful in low-risk patients, necessitates the use of a dedicated line (42), and can cost up to $8800 to treat 1 patient. In contrast, the 5- to 7-day course of steroids reported in the more recent glucocorticoid trials was safe, easy to administer, and relatively inexpensive ($50). It is unknown whether the mechanisms of physiologic-dose steroids and activated protein C are similar during sepsis or whether the combination of steroids and activated protein C provides additional benefit than each agent alone. Further trials are needed to address these issues. If giving these 2 drugs together provides no additional benefit, then the physiologic dose of steroids is preferred because they are safer, easier to administer, and less expensive.
The limitations of our analysis are consistent with those of all meta-analyses, including the potential bias secondary to nonreporting of negative study results. The beneficial effects of glucocorticoid therapy in the later trials may be partially due to time-related factors, such as improvements in ventilator management, fluid therapy, and the use of vasopressor and inotropic agents. In addition, our meta-analysis includes some relatively small studies conducted after 1997; however, we did not find any study during this time that was overly influential on the analysis.
Clinical sepsis trials of high-dose steroids published before 1989 revealed that steroids are harmful. However, the more recent trials of physiologic doses of steroids in sepsis demonstrate an overall improvement in survival and shock reversal. These contradictory results may be explained by a linear relationship between the dose of steroids and their effect on survival. At high doses, steroids have marked immunosuppressive effects that may increase the severity of primary or secondary infections and lead to worse outcomes. In contrast, low doses of steroids may be beneficial through either augmentation of adrenal function in the stressed state or limited anti-inflammatory properties that do not cause harmful immunosuppression. The effects of physiologic doses of steroids do not statistically differ between responders and nonresponders to corticotropin stimulation testing. From our analysis, we cannot definitively identify the optimal dose of steroids to administer and the timing of treatment. However, our analysis demonstrates that patients with established vasopressor-dependent septic shock for at least 2 hours and for as long as 72 hours will have improved shock reversal and survival if given a 5- to 7-day course of physiologic doses of hydrocortisone (200 to 300 mg/d) followed by a 5- to 7-day taper. Additional studies are necessary to determine whether physiologic doses of steroids are beneficial if administered to septic patients who do not develop shock or patients who develop shock but have not yet advanced to a vasopressor-dependent state.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
Enrique J. SÃ¡nchez-Delgado,
Hospital Metropolitano Vivian Pellas, Managua, Nicaragua
July 11, 2004
Low dose steroids stabilizes patient in sepsis.
Short after the inauguration of the Hospital Metropolitano Vivian Pellas in Managua, a 75 years old lady was received in the ER presenting chills, myalgias, hypotension and anuria, as well as low oxygen saturation, tachypnea and rales. She had been in treatment for congestive heart failure and atrial fibrillation, but presented with sinus rythmus. She had a plasty repair three years ago in USA for a mitral regurgitation, and the valve was fibrotic and calcified, difficulting the echo diagnosis for endocarditis. There was a marked leucocytosis, with left deviation and toxic granulations of the leucocytes, markedly elevated CRP and elevated creatinine. Both ACS or AMI were ruled out.
She was admitted to the ICU and treated for sepis, probable endocarditis, congestive heart failure and pre-renal renal failure.
Despite restoration of renal function, clearing of lung parenchima, recovering of oxygen saturation and appropiate antibiotic treatment, the hemodynamic pharmacologycal support with dopamine/dobutamine and properly regulated fluids were not enough to keep her blood pressure stable, and she had frecuent hypotensive episodes.
At the third day we decided to start with low dose hydrocortisone, 100 mg i.v. every 8 hours. She stabilized and could be continued with 100 mg a day until her release at the 8th. day with all her organ functions normalized to continue her treatment ambulatory, under supervision of her cardiologist.
David W Noble
Aberdeen Royal Infirmary
July 17, 2004
Corticosteroids for septic shock: - a question of interpretation.
The meta-analysis of Minneci and colleagues with regard to steroids for septic shock  has been interpreted too narrowly and may provide a misleading picture for practicing intensivists. I have three areas of concern. Firstly, the data analysis is retrospective, exploratory and uncorrected for multiple comparisons . Should then conclusions be less definitive than those actually offered? Secondly Bennett's low-dose steroid study is excluded from the low-dose steroid analysis. Although a rationale is provided, it remians the second largest study in this group even after pediatric patients have been excluded. Other quoted meta-analyses have not excluded this study. Thus to provide a more complete picture, sensitivity analyses are necessary and the authors should inform readers how exclusion or inclusion of this study would change results and conclusions. Finally, the trial that has the greatest impact on this aspect of the meta -analysis is fundamentally flawed in that a large percentage of patients were pharmacologically adrenalectomized with etomidate immediately prior to enrolment. With the widespread use of etomidate it is hardly surprising that so many shocked septic patients were unresponsive to corticotropin and benefited from hydrocortisone replacement therapy . Incorporation of this trial into the meta-analysis, without dealing with the confounding effects of etomidate, is problematic. Although surrogate measures promise much, currently there is insufficient scientifically sound outcomes data to mandate use of corticosteroids in septic patients. The meta-analysis of Minneci and colleagues has too many limitations, in that multiple comparisons, lack of sensitivity analyses and insufficient acknowledgement of individual study limitations has lead to narrow and unsustainable conclusions . Corticosteroids have serious adverse effects and our patients deserve decision making based on good and robust evidence. Support for the CORTICUS study must continue in the hope that intensive care practitioners can ultimately exercise truly evidence-based decision making . Corticosteroids for septic shock is not yet an evidence-based standard of care.
References. . Minneci PC, Deans KJ, Banks SM, Eichaker PQ, Natanson C. Meta- analysis: The effect of steroids on survival and shock during sepsis depends on the dose. Ann Intern Med 2004; 141: 47-56. . Pogue J, Yusuf S. Overcoming the limitations of current meta-analysis of randomised controlled trials. Lancet 1998; 351: 46-52. . Corticosteroids for septic shock "“ a standard of care? Bloomfield R, Noble DW. Br J Anaesth 2004; 93:178-80. . Eysenck HJ. Meta-analysis and its problems. BMJ 1994; 309: 789-92. . Luce JM. Physicians should administer low-dose corticosteroids selectively to septic patients until an ongoing trial is completed. Ann Intern Med 2004; 141: 70-82.
Robert G Badgett
University of Texas Health Science Center at San Antonio
July 30, 2004
Controversy due to heterogeneous mortality results among patients with normal adrenal responsiveness
The meta-analysis of corticosteroids for septic shock by Minneci et al is an important synthesis of this difficult topic (1). Could the authors please comment on the following two observations?
First the meta-analysis has led to conflicting opinions between the authors and the editorialist about safety and efficacy of continuing corticosteroids among patients with normal adrenal response to corticotropin stimulation testing ('responders') (2). Close examination of Table 3 reveals significant heterogeneity exists in the mortality data among the responders, I2=58% (3). These three studies should not be pooled, either mathematically or informally. The trail by Annane is the outlier and is the only study that suggests harm from corticosteroids among responders. Although the degree of harm (absolute increase in mortality of 8%) is insignificant, its magnitude is similar to the benefit realized in the nonresponders to the corticotropin stimulation testing and the insignificance may be due to smaller sample size. In addition to being the only study to suggest harm among the responders, this study had the largest difference between the results of the nonresponders and responders. Irregardless of whether the small studies by Yildiz and Boleart should be discarded, one must explain the results of the Annane study. One difference in the Annane study is the higher proportion of nonresponders which suggests sicker patients. A second difference is the cotreatment with fludrocortisone. Until more studies establish the correct explanation, it seems premature to continue corticosteroids for responders when the best study suggests harm. In addition, fludocortisone should be considered for cotreatment.
Second, the recent study by Hamrahian (4) suggests the free cortisol level may better predict adrenal sufficiency and should be used in further research of this topic rather than the response in the total cortisol level.
References: 1. Minneci PC, Deans KJ, Banks SM, Eichacker PQ, Natanson C. Meta-analysis: the effect of steroids on survival and shock during sepsis depends on the dose. Ann Intern Med. 2004;141:47-56. PMID: 15238370 [Full text] 2. Luce JM. Physicians should administer low-dose corticosteroids selectively to septic patients until an ongoing trial is completed. Ann Intern Med. 2004;141:70-2. PMID: 15238374 [Full text] 3. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327:557-60. PMID: 12958120 4. Hamrahian AH, Oseni TS, Arafah BM. Measurements of serum free cortisol in critically ill patients. N Engl J Med. 2004;350:1629-38. PMID: 15084695
Katherine J Deans
Critical Care Medicine Department, National Institutes of Health
September 1, 2004
Corticosteroids in sepsis: should therapy be based on response to ACTH testing?
In Response: Drs. Noble and Badgett raise important concerns about the findings of our meta-analysis (1). In response to Dr. Noble, a properly performed meta- analysis is similar to a clinical trial in that they are both prospective studies with inclusion criteria, endpoints, and analytic strategies defined prior to the beginning of the study. Furthermore, in a meta- analysis with significant heterogeneity, it is essential to partition the dataset into homogeneous groups to explain the heterogeneity of the data (2). This is not considered an analysis of multiple comparisons, but rather a fundamental technique of meta-analysis. In our study, the significantly heterogeneous effects of steroids on survival were explained by partitioning the trials by year of publication. Based on this division, the Bennett study was included in the set of trials published before 1989 (1). Subsequent descriptions of the trials were based on differences between the sets of trials, and sensitivity analyses were performed within each of these sets (1). Finally, the use of etomidate in the trial by Annane et al (3) is concerning. However, low-dose steroids remain significantly beneficial when this trial is removed (relative survival benefit, 1.36 [95% CI, 1.04 to 1.79]; P= 0.03). Moreover, recently available data from the trial by Chawla et al. also demonstrate a consistent beneficial effect of steroids on survival (RSB, 1.82 [95% CI, 0.80 to 4.00]; P>0.2) (4).
In response to Dr. Badgett's concern of heterogeneity in mortality data among "responders", these 3 trials are not different based on traditionally used significance cutoffs (P=0.15). However, even if the studies are not combined because of the concerning I2 value, there was no consistent trend towards harm in the trials and there were no significantly different treatment effects of steroids in "responders" and "nonresponders" within the individual trials. Categorizing septic patients as "responders" and "nonresponders" to corticotropin stimulation testing is based on a classification system that has not been validated. Moreover, the validity of these subgroups needs to be reassessed based on recent data which suggests that total cortisol levels may not accurately reflect adrenal function in critically ill patients (5).
Withholding steroids may potentially harm one out of every nine eligible patients. Low-dose steroids have demonstrated a consistent beneficial effect on survival in five trials and a significant improvement in survival in two meta-analyses (1, 4). This beneficial effect remains when the Annane trial is excluded (3). Furthermore, there was no increase in adverse events with steroid therapy reported in these trials or meta- analyses. The currently available data indicate that low-dose steroids should be considered for all patients with vasopressor-dependent septic shock. Treatment decisions should be based on individual patient risk- benefit profiles and not on a categorization system which has not been validated, which uses total cortisol levels of questionable value, and which has produced inconsistent and nonsignificant results in clinical trials.
References: 1. Minneci PC, Deans KJ, Banks SM, Eichacker PQ, Natanson C. Meta- analysis: the effect of steroids on survival and shock during sepsis depends on the dose. Ann Intern Med. 2004;141(1):47-56. 2. Thompson SG. Why sources of heterogeneity in meta-analysis should be investigated. Bmj. 1994;309(6965):1351-5. 3. Annane D, Sebille V, Charpentier C, et al. Effect of treatment with low doses of hydrocortisone and fludrocortisone on mortality in patients with septic shock. JAMA. 2002;288(7):862-71. 4. Annane D, Bellissant E, Bollaert PE, Briegel J, Keh D, Kupfer Y. Corticosteroids for severe sepsis and septic shock: a systematic review and meta-analysis. Bmj. 2004. 5. Hamrahian AH, Oseni TS, Arafah BM. Measurements of serum free cortisol in critically ill patients. N Engl J Med. 2004;350(16):1629-38.
Infectious Disease, Pulmonary/Critical Care, Multi-Organ Failure and Sepsis.
Results provided by:
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only