Of the 2326 citations identified from electronic and hand searches, after duplicate checking and screening the titles and abstracts of all searched articles, we selected 52 studies for full text review. Through the full-text screening, this meta-analysis contained 15 studies,[22–36] as the flowchart shown (Figure 1 and details seen in the supplementary Table S2). During searching, both English and non-English articles were considered, thus 11 Chinese articles and 4 English articles were included.
The characteristics of the included studies were shown in Table 1, Table 2. All the RCTs totally reported 2617 participants, in which 1079 patients received UFH as experimental group and 1538 patients received placebo (saline) or usual care as control group. Except for a lack of 28 d mortality data in Zhang et al. and Guo et al' study,[24,36] the remaining 13 studies all performed analyses of 28 d mortality. Zhang et al's study didn't report APACHE II score.
Moreover, after calculation, the result showed that the median value of APACHE II score could be used to estimate its mean in Jaimes et al' trial. In Supplementary Figure S1, the X-axis was RR value and the Y-axis was SE (log [RR]). The outer dashed lines indicate the triangular region within which 95% of studies are expected to lie in the absence of biases and heterogeneity. The solid vertical line corresponds to no intervention effect. The funnel plot was basically symmetrical, indicating a small risk of publication bias. As shown in Table 2, there were various doses of UFH administration in different researches. The route of UFH administration was intravenous for 1 week in most researches. Our group usually use 100 U/kg UFH by continuous intravenous pumping in clinic work. Furthermore, a multicenter RCT (NCT02654561) on the effect of UFH in sepsis is ongoing by our team.
Main reasons for bias for individual studies were shown in Figure 2 and Figure 3, each risk of bias item was presented as percentages. In general, all included studies were usually classified as moderate quality. Random sequence generation, allocation concealment, blinding of outcome assessments, participants and personnel were unclear or seldom reported in these trials. But adequate outcome data was reported in all studies and selective reporting was low for most studies. The strength of evidence assessment was shown in Supplementary Table S3. It showed that the 28 d mortality and PLT were high quality as outcome indicators.
Risk of bias summary. Review author's judgements about each included study. Low, moderate and high risk of bias are represented as green, yellow, and red colors respectively
Risk of bias graph. Review author' judgements about each risk item presented as percentages across all included studies
The forest plot for 28 d mortality was shown in Figure 4. Within 28 days of admission, 236 (23.9%) died in the UFH group, and 429 (30%) died in the control group (RR = 0.84; 95% CI = 0.73 to 0.96; P = 0.009 < 0.05), indicating a statistically significant reduction in 28 d mortality in UFH-treated patients with sepsis. There was no evidence of heterogeneity between studies (I2 = 0.0%).
Mortality in patients randomized to UFH versus placebo or usual care. CI, confidence interval; RR, risk ratio
Subgroup analysis was utilized to identify the effect of UFH on sepsis which is based on the various definitions and diagnostic criteria of sepsis. As Figure 5 showed, except the Deng's and Peng's studies,[22,30] all the other studies were included in the group of Sepsis 1.0, with the RR = 0.85; 95% CI = 0.74 to 0.97; P = 0.02 < 0.05; I2 = 0.0%. And for Sepsis 2.0, with the RR = 0.58; 95% CI = 0.35 to 0.94; P = 0.03 < 0.05; I2 =0.0%. The 28 d mortality of patients treated with UFH was statistically significant in both sepsis 1.0 group and sepsis 2.0 group. There was no evidence of heterogeneity between studies (I2 = 0.0%). There was high heterogeneity among subgroup studies (I2 = 54.1%). The finding from this subgroup analysis further reflects the benefit of UFH for treating septic patients. In clinical practice, APACHE II score was widely used to indicate the disease severity and predict the prognosis of critically ill patients. To date, there is no unified international rules relating to APACHE II cut-off value. In this trial, the mean of APACHE II score ranged from 9 to 23.8 for patients in UFH group, 13, 15, 17 and 19 were selected as APACHE II cut-off values respectively. On the basis of various APACHE II scores, subgroup analysis was performed (Table 3).
Among 13 trials, except Zhang et al' study, the other 12 studies were included in the subgroup analysis. The patients with APACHE II scores between 13 and 17 might benefit from UFH treatment. Subsequently, APACHE II score of 15 was selected as the ultimate cut-off value which could provide the best diagnostic accuracy to predict mortality of critical ill patients.
As Figure 6 showed, five studies[23,25,26,29,34] were comprised in the group of APACHE II ≤ 15, with the RR = 0.84; 95% CI = 0.58 to 1.21; P = 0.35 > 0.05; I2 = 0.0%. And seven studies[22,27,28,30–32,35] were comprised in the group of APACHE II > 15, with the RR = 0.83; 95% CI = 0.72 to 0.96; P = 0.01 < 0.05; I2 = 14.0%. Because the I2 < 30%, the heterogeneity could be ignored. UFH treatment might be beneficial in patients with APACHE II > 15 but not with APACHE II ≤ 15.
Subgroup analysis of mortality in which APACHE II cut-off value is 15. APACHE II, Acute Physiology and Chronic Health of Evaluation II; CI, confidence; RR, risk ratio
There were eight studies[22,25,28,32–36] reported PLT, PT and APTT, respectively. In total, 277 participants were enrolled in the UFH group and 262 participants were enrolled in the control group. As shown in the forest plot (Supplementary Figures S2, S3 and S4), the P values were 0.03, 0.93 and 0.007, respectively. The PLT (MD = 9.18; 95% CI = 0.68 to 17.68; P = 0.03 < 0.05; I2 = 21%) in UFH group was higher than that in control group. The PT results (MD = − 0.05; 95% CI = − 1.34 to 1.23; P = 0.93 > 0.05; I2 = 81%) failed to reach statistical significance, although the considerable heterogeneity, the APTT in UFH group was shorter than that in control group (MD = − 8.01 95% CI = − 13.84 to − 2.18; P = 0.007 < 0.05; I2 = 94%).
UFH reduced the incidence of MODS and the LOS in ICU. Four studies[22,28,30,34] compared the incidence of MODS as Supplementary Figure S5 shown. In summary, 177 participants were enrolled in the UFH group and 169 participants were enrolled in the control group. Comparing to control group, UFH reduced MODS incidence with statistical significance (RR = 0.61 95% CI = 0.45 to 0.84, P = 0.002 < 0.05; I2 = 0%).
Five researches[24,25,28,30,33] evaluated the effect of UFH on LOS in ICU in sepsis patients. Forest plot (Supplementary Figure S6) showed the results. In total, 154 participants were enrolled in the UFH group and 147 participants were enrolled in the control group. Comparing to control group, UFH reduced LOS in ICU with statistical significance (MD = − 4.94 95% CI = − 6.89 to − 2.99, P < 0.00001 < 0.05; I2 = 66%). But the heterogeneity was considerable.
Four studies[24,28,30,34] reported the duration of ventilation (Supplementary Figure S7). Generally, 160 participants were enrolled in the UFH group and 158 participants were enrolled in the control group. Comparing to control group, UFH decreased ventilation time with statistical significance (MD = − 3.01 95%; CI = − 4.00 to − 2.02, P < 0.00001 < 0.05; I = 0%).
Four studies reported the bleeding complications[22,26,27,32] (Supplementary Figure S8). Seven hundred fifty-one participants were enrolled in the UFH group and 1211 participants were enrolled in the control group totally. There were 15 (1.9%) bleeding events in the UFH group and 15 (1.2%) in the control group (RR = 1.10; 95% CI = 0.54 to 2.23; P = 0.80 > 0.05; I2 = 0%). There were no statistically significant differences. UFH had no effect on bleeding events in patients with sepsis.
BMC Anesthesiol. 2022;22(28) © 2022 BioMed Central, Ltd.