Study selection and characterization
A total of 5791 studies were identified, in which the latest study was from 2021. A total of 1587 articles were excluded as duplicates. The screening of the titles and abstracts of the remaining 4204 records revealed that 79 articles were potentially eligible for the assessment of the full text. Finally, 8 articles with a total of 5124 severe sepsis patients and 3482 septic shock patients were included in this meta-analysis. The details of the exclusion are shown in Figure 1. Seventeen studies were review or meta-analysis. Sixteen studies analyzed the same data. Two studies were conference abstracts containing no data on mortality. Thirteen studies were clinical trials but not RCTs. Twenty-three studies did not compare albumin with crystalloid. The included studies are characterized in Table S2. Briefly, 5 studies used 4-5% albumin for fluid resuscitation, whereas 2 studies used 20% albumin treatment. One RCT included both concentrations. Normal saline was used as crystalloid in 4 studies. Ringer’s lactate solution was used as crystalloid in 2 trials. The remaining 2 trials used various type of crystalloid solutions. Data of mortality were extracted and included two time points, 28-day and 90-day mortality rates for both severe sepsis and septic shock. The bias assessment of the relevant studies is described in Table S3.
Efficacy of albumin treatment on mortality
Eight studies contained the data of 28-day mortality in patients with severe sepsis (Figure 2A). Fluid therapy with human albumin did not reduce the mortality compared with crystalloid therapy, as the difference was not statistically significant [OR 0.94 (0.83, 1.07); P= 0.35]. The heterogeneity was not-significant (P= 0.53, I2= 0%).
Information of 90-day mortality in patients with severe sepsis was extracted from 8 trials (Figure 2B). The data of the longest complete follow-up (90-day mortality) were used in the analysis since two studies investigated the mortality rate for more than one time point[8, 19]. The heterogeneity was not-significant (P= 0.64, I2= 0%). Although a trend toward reduced mortality was found in patients who received albumin solution compared with crystalloid, the difference was again not statistically significant [OR 0.91 (0.80, 1.02); P= 0.11].
The effects on 90-day mortality in patients with septic shock were evaluated in 7 trials (Figure 2C). The heterogeneity was not significant (P= 0.60, I2= 0%). The resuscitation using albumin significantly decreased the 90-day mortality in septic shock patients, compared with the use of crystalloid solutions [OR 0.85 (0.74, 0.99); P= 0.04]. Thus, our analysis showed that albumin might be effective in reducing all-cause mortality for septic shock, but not that for severe sepsis.
Impact of different concentrations of albumin on mortality in patients with severe sepsis
The 28-day mortality in severe sepsis patients was reported in all studies involving the direct comparisons between two concentrations of albumin and crystalloid. Five trials compared 4-5% albumin with crystalloid, while two trials estimated the impact of 20% albumin compared with crystalloid. One trial included both concentrations of albumin without defining the mortality rate for each group. No direct comparison between 4-5% albumin and 20% albumin was found in the literature. Thus, finally 7 RCTs were included (Figure 3A). The meta-analysis results showed that albumin of both concentrations displayed no benefical effect on 28-day mortality in severe sepsis patients, compared with crystalloid. [4-5% albumin vs. crystalloid, OR 0.90 (0.74, 1.08); P = 0.25; 20% albumin vs. crystalloid, OR 0.96 (0.81, 1.14); P = 0.67]. Figure 3B shows the probability-based ranking results, indicating that 4-5% albumin ranked the highest and crystalloid ranked the lowest in reduing the mortality rate. Thus, resuscitation with 4-5% albumin might be associated with the lowest 28-day mortality in patients with severe sepsis.
Seven RCTs that considered the 90-day mortality in patients with severe sepsis were included in the meta-analysis. Five RCTs compared 4-5% albumin with crystalloid, and 2 trials compared 20% albumin with crystalloid (Figure 4A). None of them found a significant difference from the meta-analysis [4-5% albumin vs. crystalloid, OR 0.90 (0.74, 1.08); P= 0.25; 20% albumin vs. crystalloid, OR 0.90 (0.77, 1.06); P= 0.21]. The probability-based ranking results are shown in Figure 4B. The dose 4-5% albumin ranked as similar as 20% albumin, while crystalloid ranked the lowest in reduing the mortality rate. Patients with severe sepsis resuscitated with either concentration of albumin may have lower 90-day mortality than those resuscitated with crystalloid.
Impact of different concentrations of albumin on mortality in patients with septic shock
The 90-day mortality in patients with septic shock was reported in 6 studies. Four compared 4-5% albumin with crystalloid in 953 patients, and 2 trials compared 20% albumin with crystalloid in 1913 patients (Figure 5A). The meta-analysis results indicated that 4-5% albumin did not significantly decrease mortality compared with crystalloid [OR 0.89 (0.68, 1.08); P = 1.15]. However, 20% albumin markedly reduced the mortality, in comparison to crystalloid [OR 0.81 (0.67, 0.98); P= 0.03]. Figure 5B shows the probability-based ranking results, which revealed that 20% albumin ranked the highest in reducing the mortality rate. Thus, resuscitation with 20% albumin in patients with septic shock could represent the best choice in reducing the all-cause mortality.