Results
From a total of 391 patients screened for eligibility, 170 patients were enrolled (Figure 1). Regarding the exclusion of patients under chronic corticosteroid intake, it should be underscored that none of the enrolled patients was in any chronic intake of corticosteroids. Taking into consideration the time frame between start of vasopressors and start of hydrocortisone, four quartiles were defined: 1–9 hours with 46 patients; 10–24 hours with 46 patients; 25–72 hours with 39 patients; and more than 72 hours with 39 patients. The first quartile with 46 patients was considered the early initiation group; the other three quartiles with 124 patients in total were considered the late initiation group. There were no patients in whom the time delay for start of hydrocortisone was between 9 and 10 hours. Comparisons of demographic and clinical characteristics of the two groups are shown in Table 1 .
(Enlarge Image)
Figure 1.
Study flowchart. Patient recruitment and classification in groups of early and late intiation of hydrocortisone. PBMCs = peripheral blood mononuclear cells, TNF-α = tumor necrosis factor-α.
ROC analysis confirmed the discriminating ability of early administration of hydrocortisone on overall survival (area under the curve = 0.613; p = 0.014). Furthermore, the clinically intuitive first quartile cutoff value of less than 9 hours for the early initiation of hydrocortisone concurs with the value that ROC analysis yields as the best trade-off between sensitivity and specificity. The proportion of survivors was significantly greater in the early in comparison to the late initiation group (52.2% vs 30.6%; Fisher exact test; p = 0.012), and the OR for favorable outcome for patients of the early initiation group was 0.40 (95% CI, 0.20–0.81). Furthermore, as Kaplan-Meier survival analysis showed (Figure 2), survival was significantly prolonged among patients in the early initiation group compared to the late initiation group (p = 0.018).
(Enlarge Image)
Figure 2.
Impact of early initiation of hydrocortisone on final outcome. Using quartile analysis, two groups of patients were defined: those with early initiation of hydrocortisone treatment (i.e., < 9 hr from start of vasopressors, n = 46) and those with late initiation of hydrocortisone treatment (i.e., > 9 hr from start of vasopressors, n = 124). Statistical comparisons between early- and late-initiation groups are shown.
To elaborate if early initiation of hydrocortisone was an independent risk factor associated with final outcome, univariate analysis was done to identify potential factors linked with unfavorable outcome. It was found that among all tested variables, only the APACHE II and SOFA scores and the delay until start of hydrocortisone were linked with unfavorable outcome (data not shown). ROC analysis confirmed the discriminating ability of APACHE II score (area under the curve = 0.653; p = 0.002) and SOFA score (area under the curve = 0.621; p = 0.016) on overall survival. However, the two scores are very highly correlated (r = 0.625); therefore, the inclusion of only the APACHE II score, as proven by stepwise logistic regression (data not shown) should suffice. The best trade-off providing sensitivity greater than 80% was an APACHE II score greater than or equal to 19. Table 2 shows the results of the stepwise forward Cox regression and logistic regression analysis of high APACHE II score (≥ 19) and early initiation start of hydrocortisone on the final outcome.
As shown in Table 2 , the effects of low or high APACHE II score and early or late hydrocortisone initiation are additive. In patients with high APACHE score (≥ 19), the early initiation of hydrocortisone increased the survival rate from 19.8% to 41.2% (p = 0.021). Likewise, in patients with low APACHE score (< 19), the early initiation of hydrocortisone increased the survival rate from 55.0% to 83.3%.
The time until withdrawal of vasopressors was earlier in the early-initiation group than in the late-initiation group (Figure 3). More precisely, median time until discontinuation of vasopressors was 4 days for patients of the early-initiation group; it was prolonged to 15 days for patients of the late-initiation group (p < 0.0001).
(Enlarge Image)
Figure 3.
Impact of early initiation of hydrocortisone on the total time on vasopressors. Using quartile analysis, two groups of patients were defined: those with early initiation of hydrocortisone treatment (i.e., < 9 hr from start of vasopressors, n = 46) and those with late initiation of hydrocortisone treatment (i.e., > 9 hr from start of vasopressors, n = 124). Statistical comparisons between early- and late-initiation groups are shown.
The demographic characteristics of patients of the early- and late-initiation groups identified a considerable lower rate of infection with Acinetobacter baumannii in the early-initiation group than in the late-initiation group. This may lead to erroneous results because the appropriateness of antimicrobial therapy was 66.7% among the total of A. baumannii–infected patients compared with 90.3% among patients infected by other type of pathogens (p = 0.006). Survival analysis was repeated after excluding patients treated with inappropriate antimicrobials for A. baumannii. Mortality was 33.3% in the early-initiation group and 53.3% in the late-initiation group (log-rank, 4.760; p = 0.029). Cytokine stimulation by PBMCs and monocytes of patients in relation with the quartile of delay from start of hydrocortisone showed that the production of TNF-α was lower from cells coming from patients who were initiated early hydrocortisone (mainly from the first two quartiles). The differences were pronounced when cells were stimulated with LPS, Pam3Cys, PHA, and heat-killed P. aeruginosa (Figure 4).
(Enlarge Image)
Figure 4.
Impact of hydrocortisone on the release of tumor necrosis factor (TNF)-α. Production of TNF-α from peripheral blood mononuclear cells (PBMCs) or monocytes isolated 24 hr after initiation of therapy is shown in relation with the quartiles of delay of start of hydrocortisone from start of vasopressors. The numbers of patients in each quartile are as follows: 0–9 hr, nine patients; 10–24 hr, 10 patients; 24–72 hr, seven patients; and > 72 hr, eight patients. Release from unstimulated cells was below the lower limit of detection. The type of stimulus is shown and the types of cells tested are shown. p values indicate statistically significant comparisons with the 0–9 hr quartile. LPS = lipopolysaccharide, PHA = phytohemagglutin.