Search results and study characteristics
We selected 898 articles from the search and finally included 8 randomized controlled trials into the analysis (Fig. 1)(12, 18-24) . We didn't come across any quasi-randomized trials. The characteristics of the included studies are described in Table 1. Five studies were conducted in North American civilians who visited major trauma centers, while the other three were conducted in China. Two studies only evaluated patients with penetrating injuries, while the remaining six included penetrating injuries combined with blunt injuries. Two studies only studied preoperative resuscitation, two studies only studied intraoperative resuscitation, and three trials studied patients in two periods. In addition, one trial did not mention surgery related resuscitation strategies. All but one study excluded patients with suspected traumatic brain injury.
Assessment of Reporting Bias
The methodological quality of RCTs was assessed by the Cochrane risk-of-bias tool, which consists of six factors: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting (Fig. 2).
The main outcome data of mortality are shown in Table 2. Two studies showed 30-day mortality (18, 21) , while five studies showed in-hospital mortality (12, 19, 20, 22, 23) . One study did not describe mortality(24). The heterogeneity test revealed that there was little heterogeneity between the two groups (χ 2 = 8.9; P = 0.18; I2 = 33%). In these trials, the mortality rates in the PH and CR groups accounted for 133 (18.2%) of 731 and 197 (26.4%) of 745, respectively. The results showed that limited fluid resuscitation could reduce the mortality of patients with hemorrhagic shock (RR = 0.70; 95% CI = 0.58-0.84; P < 0.05) (Fig. 3).
Blood routine index (PLT and Hb)
As shown in Figure 4, three trials (12, 21, 23) compared the platelet values of PH group or CR group. According to the heterogeneity test, a high degree of heterogeneity was exhibited between studies (χ2 =131.35; P < 0.00001; I2 = 98%). Thus sensitivity analysis was conducted to eliminate the heterogeneity. After removing Bickell's data, the two experiments included data of 240 patients (χ2 =0.13; P=0.72; I2 = 0%). It is indicated that the platelet value in pH group was higher than that in CR group (MD = 9.98; 95%CI=8.03-11.93; P < 0.00001). Three trials (12, 20, 23) , involving 922 patients, investigated the hemoglobin values of patients with hemorrhagic shock after treatment in PH group or CR group, revealing high heterogeneity between studies (χ2 =12.44; P=0.002; I2 = 84%). The overall effect showed that the hemoglobin value of PH group was higher than that of CR group (MD = 16.74; 95%CI=10.81-22.68; P < 0.00001).
Blood coagulation function (PT, APTT)
As shown in Figure 5, four trials(12, 20, 21, 23) , including 1001 patients, compared the changes of prothrombin time (PT), while three trials (12, 21, 23) , including 842 patients, compared the changes of activated partial thromboplastin time (APTT) between PH group and CR group. Heterogeneity test showed that the comparison results of PT (χ2 =201.75; P < 0.00001; I2 = 99%) and APTT (χ2 =34.80; P < 0.00001; I2 = 94%) were highly heterogeneous. The overall effect showed that there was no significant difference in PT and APTT time between PH group and CR group after hemorrhagic shock resuscitation (MD = -2.00; 95%CI=-4.83 to 0.84; P =0.17).
Fluid balance and transfusion requirements
As shown in Figure 6, the overall fluid resuscitation and blood transfusion of PH group and CR group were analyzed by subgroup. Heterogeneity test showed that packed red blood cell (PRBC), prehospital crystal, emergency department crystal volume and total inputs were highly heterogeneous, while intra operative crystal (χ2 =0.07; P=0.79; I2 = 0%) and estimated blood loss (χ2 =0.51; P=0.48; I2 = 0%) had no heterogeneity. It is indicated by the overall effect that the total amount of fluid resuscitation in PH group was lower than that in CR group (MD = -564.21; 95%CI=-833.66 to -244.77; P < 0.001), and the estimated blood loss in PH group was less than that in CR group (MD = -721.53; 95%CI=-1326.19 to -116.87; P < 0.05). There was no significant difference in the amount of blood transfusion between the two groups (MD = -144.13; 95%CI=-820.53 to 532.27; P =0.68).
The included studies described a series of complications, including AKI, anemia, infection, thrombocytopenia, ARDS, differentiated intravascular coagulation (DIC), MODS, etc. As shown in Fig.7, three trials (12, 18, 22) compared the incidence of AKI between PH group and CR group. Heterogeneity test showed high heterogeneity (χ2 =17.08; P =0.0002; I2 = 88%) among the studies， which was attempted to be eliminated through sensitivity analysis. After removing the data of Schreiber, two experiments (12, 18) compared the data of 41 patients (χ2 =0.02; P =0.90; I2 = 0%). The incidence of AKI accounted for 13/364 (3.6%) in PH group and 28/375 (37.5%) in CR group. The results showed that limited fluid resuscitation could reduce the incidence of AKI in patients with hemorrhagic shock (RR = 0.43; 95%CI= 0.24-0.79; P < 0.05).
The comparison between ARDS and MODS showed similar results. Four trials (12, 19, 20, 24) involving 1034 patients compared the incidence of ARDS between PH group and CR group. No heterogeneity was shown in the heterogeneity test (χ2 =0.12; P =0.99; I2 = 0%). The incidence of ARDS in PH group and CR group was 24/507 (4.7%) and 58/527 (11.0%) respectively, and the incidence of ARDS was higher (RR = 0.42; 95%CI= 0.27-0.65; P < 0.05) during conventional fluid resuscitation. Four trials compared the incidence of MODS between PH group and CR group. Heterogeneity test showed that there was high heterogeneity between the studies (χ2 =6.48; P =0.09; I2 = 54%). After removing Bickell's data, three trials (19, 20, 23) compared the data of 434 patients (χ 2 = 0.22; P = 0.89; I2 = 0%). The incidence of MODS was 17/217 (7.8%) in PH group and 44/217 (20.3%) in CR group. The results showed that the risk of MODS during conventional fluid resuscitation was also higher than that during limited fluid resuscitation (RR = 0.40; 95%CI= 0.24-0.66; P < 0.05).
Days of ICU or hospital
As shown in Figure 8, two trials (12, 22) involving 651 patients compared days of hospital showed less heterogeneity (χ2 =1.32; P =0.25; I2 = 24%). With 651 patients included, two trials (12, 22) compared days of intensive care unit (ICU). No heterogeneity was exhibited in the test (χ2 =0.66; P =0.42; I2 = 0%). Based on the overall effect, it is shown that there was no significant difference in hospital stay and ICU treatment time between pH group and CR group during hemorrhagic shock resuscitation (MD = -0.65; 95%CI=-2.14 to 0.84; P =0.39).