Physicians Should Administer Low-Dose Corticosteroids Selectively to Septic Patients until an Ongoing Trial Is Completed
J. Luce
DOI: https://doi.org/10.7326/0003-4819-141-1-200407060-00019
IF: 39.2
2004-07-06
Annals of Internal Medicine
Abstract:Most physicians prefer administering therapies that have a physiologic rationale and proven benefit. We urgently need beneficial therapies for sepsis severe enough to cause shock, which is responsible for approximately 400000 intensive care unit admissions and 200000 deaths each year in the United States (1). In this issue, a meta-analysis by Minneci and colleagues (2) from the National Institutes of Health (NIH) concludes that low but not high doses of corticosteroids reduce mortality from septic shock and that we should administer these agents to all patients who have vasopressor-dependent sepsis. How the investigators reached these conclusions and why I agree with the first but not the second of them are the subjects of this editorial. The adrenal glands release cortisol during various stresses, of which septic shock is one of the most profound. Both high and low plasma levels of the endogenously released hormone are associated with increased mortality, presumably because high levels reflect severe stress and low levels an inability to respond to stress by releasing cortisol (3). The beneficial actions of cortisol are due to its anti-inflammatory effects (including inhibition of cytokine production and prevention of circulating inflammatory cell migration into tissues) and cardiovascular effects (including attenuation of nitric oxide synthase induction and enhancement of the vasoconstrictive response to catecholamines) (4). Between 1959 and 1986, investigators administered short courses of high (also called pharmacologic) doses of corticosteroids (for example, methylprednisolone, 30 mg/kg of body weight, in 4 intravenous doses over 24 hours) to almost 1300 septic patients in 9 clinical trials. One rationale for these trials was to interrupt the inflammatory cascade hypothesized to perpetuate sepsis (5). Unfortunately, no large randomized trial demonstrated that corticosteroids improved mortality. Such improvement was seen only in a single-center study with both prospective and retrospective enrollment. Furthermore, 2 of the large trials reported an increase in secondary infections among patients who received high-dose corticosteroids. In 1995, 2 meta-analyses of these investigations concluded that high-dose corticosteroids were ineffective in treating septic shock and might be harmful (6, 7). Two years later, another meta-analysis conducted by NIH investigators (8) showed that high doses of both corticosteroid and noncorticosteroid anti-inflammatory agents had no benefit in patients with septic shock. These findings cast doubt on the then-prevalent rationale for use of these agents. Although the use of high-dose corticosteroids decreased after these publications, physicians have continued to administer lower doses of exogenous hormone (typically 100 mg of intravenous hydrocortisone 3 times daily, known as a low or replacement dose even though it is supraphysiologic) to septic patients who have meningococcemia, long-term corticosteroid use, and other known causes of adrenal insufficiency. Some physicians have done so without proving that such insufficiency exists. Others have diagnosed adrenal insufficiency by obtaining a low baseline cortisol level during stress (a review published in 1996 [9] suggested a level < 525 nmol/L, especially if the level does not increase 30 or 60 minutes after adrenal stimulation with 250 g of intravenous corticotropin). In recent years, a new rationale for the use of corticosteroids has emerged: Relative adrenal insufficiency may be a complication of septic shock. Early observational studies showed that, despite adequate baseline cortisol levels, up to 40% of patients in septic shock had a blunted response to corticotropin stimulation and many of these patients also manifested impaired vasopressor sensitivity to catecholamines (10, 11). In 2000, Annane and colleagues (12) observed that the mortality from septic shock was greatest in patients whose baseline cortisol levels exceeded 938 mmol/L and did not increase by more than 248 nmol/L after corticotropin stimulation. These findings suggested that impaired adrenal reserve occurred in septic shock and implied that replacement corticosteroids might help to resolve this condition. After this observational study, Annane and colleagues (13) administered either placebo or hydrocortisone (50 mg every 6 hours) and fludrocortisone (50 g by mouth daily) for 7 days to 300 patients with septic shock who had just undergone corticotropin stimulation. Their goal was to determine whether corticosteroids affect survival distribution 28 days after randomization in patients whose cortisol levels did not increase by more than 248 mmol/L in response to corticotropin stimulation. The investigators found that the median time until death was longer in those nonresponders who received corticosteroids than in those who received placebo, but it was not longer in responders. The median time until death was longer overall for patients who received corticosteroids than for those who received placebo, largely because nonresponders outnumbered responders in the study, but 28-day crude mortality did not differ between the 2 groups. The time to withdrawal of vasopressors was shorter in the nonresponders who received corticosteroids than in nonresponders who received placebo. The time to withdrawal among responders was not statistically different, but because the nonresponders outnumbered responders, vasopressors were withdrawn more rapidly from patients who received corticosteroids overall. As a result of these findings, the investigators recommended that all patients with vasopressor-dependent septic shock receive corticosteroids initially but that only those who do not respond to corticotropin should continue to receive these agents for 7 days. The observational and experimental studies by Annane and colleagues (12, 13) have been highly influential. The most recent major review of adrenal insufficiency recommends that 2 groups of patients with septic shock receive low doses of glucocorticoids (but not necessarily of mineralocorticoids). The first group has regular adrenal insufficiency (baseline cortisol levels < 414 mmol/L). The second group has relative adrenal insufficiency (baseline levels between 414 and 938 mmol/L and failure to increase by >248 mmol/L after corticotropin stimulation). The authors of this review also recommend that all patients with suspected adrenal insufficiency undergo corticotropin stimulation and state that patients without absolute or relative adrenal insufficiency need not continue to receive low-dose corticosteroid therapy (14). These recommendations and those of Annane and colleagues (13) differ from those of Minneci and colleagues (2). These NIH investigators reiterate their previous meta-analysis of the 9 trials conducted before 1997 and conclude that high doses of corticosteroids are not beneficial in patients with septic shock. I agree with this conclusion. They also provide a meta-analysis of 5 randomized, double-blind, placebo-controlled trials published after 1997 in which corticosteroids were given in low doses over several days. The trial conducted by Annane and colleagues (13) was by far the largest of these recent trials; the other 4 trials enrolled approximately 40 patients each, and only 4 trials analyzed mortality and shock reversal. Other than the trial of Annane and colleagues (13), no trial was large enough to demonstrate a statistically significant outcome difference in patients given low-dose corticosteroids. Furthermore, in the trial of Annane and colleagues (13), corticosteroids only increased the median time until death and decreased the time until shock reversal but did not reduce crude mortality. Nevertheless, Minneci and colleagues (2) interpreted this trial as demonstrating benefit and, after combining it with the other trials, concluded that low-dose corticosteroids were beneficial overall. Only 3 of the 5 trials stratified patients as responders and nonresponders to corticotropin stimulation. One trial used the same definition of unresponsiveness that Annane and colleagues (13) did; the third study used a smaller increase in cortisol. Only 2 trials, including that of Annane and colleagues (13), reported mortality and shock reversal separately for responders and nonresponders. When comparing time until death and vasopressor withdrawal, Annane and colleages (13) found that these outcomes were different in nonresponders but not in responders. However, Minneci and colleagues (2) found no statistically significant differences comparing crude rates of mortality or vasopressor withdrawal. Because of these latter findings, Minneci and colleagues (2) conclude that response to corticotropin stimulation does not predict benefit from corticosteroid administration. They also found that no recent trial reported an increase in secondary infections from corticosteroids. As a result, they recommend that low-dose corticosteroids be administered to all patients with vasopressor-dependent septic shock until the results of larger studies are available. Physicians should follow this recommendation if they are convinced that low-dose corticosteroids have been proven to benefit and not harm septic patients. Furthermore, physicians need not perform corticotropin stimulation if they believe that corticosteroids are truly effective regardless of responsiveness or nonresponsiveness to this test. Yet I remain concerned that the meta-analysis is based on a small number of low-dose corticosteroid trials, most of which enrolled few patients, and not all of which performed the same analyses. The large trial of Annane and colleagues (13) accounts for three quarters of the patients in the meta-analysis, and these investigators concluded that low-dose corticosteroids benefited primarily patients who did not respond to corticotropin stimulation, a conclusion different from that of Minneci and colleagues (2). Because only 1 other trial