We identified 2173 articles in the initial literature search (Figure S1). After removing duplicates, we screened the titles and abstracts of 1864 records. 42 articles were retrieved for full-text reviewing. Finally, nine RCTs invovled 1925 episodes of patients that compared BL/BLIs with carbapenems fulfilled the inclusion criteria for this study[16–24]. Three[17, 20, 23] of the 9 studies were conducted in Turkey, two[16, 24] were conducted in USA; and one study was conducted in Spain[18], Germany[21], China[19] and Japan[22]. These studies were conducted between 1996 and 2017. Seven studies[16, 17, 20–24] focused on haematological malignancy and solid malignancy; one study[19] focused on haematological malignancy and one study[18] was not specified. Five studies[16, 19–21, 24] included only adults, two[17, 23] included only children, one study[22] included mainly children, and the other one[18] did not provide the patient age. Among the 9 trials, piperacillin/tazobactam was evaluated in 6 trials[18–23] and cefoperazone/sulbactam in evaluated in 3 trials[16, 17, 24]. Amikacin was added to the treatment groups in one study[20] and vancomycin was used in another study[16] (Table 1).
Table 1
Characteristics of included studies.
Study | Country | Age(median, range) | Episodes of patients | Dose regimen | Outcomes | Timing of evaluation for treatment success without modification | Mortality observation duration |
Carbapenems | BL/BLIs | Carbapenems | BL/BLIs |
Bodey et al. 1996[16] | USA | 50(18–79) | 52(17–84) | 369 | IPM(500 mg/m2 × 4) + VAN(1g × 2) | CFS(3g × 3) + VAN(1g × 2) | Ⅰ,Ⅳ | End of treatment | NA |
Winston et al. 1998[24] | USA | 39(16–81) | 47(16–76) | 203 | IPM(0.5g × 4) | CFS(6g × 2) | Ⅰ,Ⅱ,Ⅳ | End of treatment | NA |
Figuera et al. 2001[18] | Spain | NA | NA | 137 | IPM(0.5g × 4) | TZP(4.5g × 4) | Ⅰ,Ⅱ,Ⅲ,Ⅳ | 72 h after treatment | NA |
Reich et al. 2005[21] | Germany | 44(18–70) | 44(20–63) | 232 | MEM(1g × 3) | TZP(4.5g × 3) | Ⅰ,Ⅱ,Ⅳ | 72 h after treatment | 7 d after treatment |
Oztoprak et al. 2010[20] | Turkey | 52.79(17–83) | 120 | MEM(1g × 3) + AMK(1g × 1) IPM(0.5g × 4) + AMK(1g × 1) | TZP(4.5g × 4) + AMK(1g × 1) | Ⅰ,Ⅱ,Ⅳ | End of treatment | End of treatment |
Vural et al. 2010[23] | Turkey | 4(2–13) | 5(2–14) | 99 | IPM(15 mg/kg × 4) | TZP(90 mg/kg × 4) | Ⅰ,Ⅱ,Ⅳ | End of treatment | During neutropenia |
Demir et al. 2011[17] | Turkey | 7.5(0.58–17.4) | 5.7(0.5-17.16) | 208 | IPM(20 mg/kg × 3) MEM(20 mg/kg × 3) | CFS(60 mg/kg × 3) | Ⅰ,Ⅱ,Ⅳ | 72 h after treatment | During treatment |
Jing et al. 2016[19] | China | 40(25–50) | 41(25–48) | 123 | IPM(0.5g × 4) | TZP(4.5g × 4) | Ⅰ,Ⅱ,Ⅳ | End of treatment | During treatment |
Sano et al. 2017[22] | Japan | 10(0–25) | 9(0–25) | 434 | MEM(40 mg/kg × 3) | TZP(112.5 mg/kg × 3) | Ⅰ,Ⅱ,Ⅲ, Ⅳ | 120 h after treatment | NA |
IPM: imipenem/cilastatin, VAN: vancomycin, MEM: meropenem, CFS: cefoperazone/sulbactam, TZP: piperacillin/tazobactam, AMK: amikacin. |
h: hour, d: day, NA: not available. |
Ⅰ: treatment success without modification, Ⅱ: all-cause mortality, Ⅲ: infection related mortality, Ⅳ: adverse event |
More than half of the studies were assigned unclear risk of bias related to random sequencing, allocation concealment and blinding. One trial was judged as high risk by by lack of blinding. High risk of bias in incomplete outcome data were assigned because of the lack of data related to mortality and major outcomes. Three trials funded by pharmaceutical company were assigned high risk in other bias. (Fig. 1).
Treatment success without modification
Treatment success without modification was evaluated in 9 studies including 1925 episodes, with no difference between carbapenems and BL/BLIs in empiric treatment of febrile neutropenia (RR 1.04, 95% CI 0.93–1.15) (Fig. 2).
In a subgroup analysis including studies only for adults, carbapenems had a similar treatment success rate with BL/BLIs (RR 1.04, 95% CI 0.90–1.20). In a subgroup analysis including studies only for children, no significant difference between carbapenems and BL/BLIs in treatment success rate (RR 0.99, 95% CI 0.86–1.14) (Figure S2).
In a subgroup analysis by BL/BLIs, the pooled analysis of 6 studies showed carbapenems had a treatment success rate similar with piperacillin/tazobactam (RR 1.07, 95% CI 0.92–1.23), and cefoperazone/sulbactam (RR 0.97, 95% CI 0.90–1.04) (Figure S3)..
In the sensitivity analysis after studies were randomly excluded, similar RRs were observed I2 = 43%). After excluding a study that only focused on haematological malignancy, heterogeneity (I2) between studies was less than 50%.
The funnel plot for treatment success without modification was approximately symmetrical (Fig. 3). Egger’s test was conducted to assess publication bias and no evidence of asymmetry was observed in treatment success without modification (P = 0.222 ).
All-cause mortality
All-cause mortality was reported in 7 studies, including 1457 episodes. No significant difference between carbapenems and BL/BLIs in mortality was observed (RR 1.15, 95% CI 0.64–2.06) (Fig. 4).
Infection-related mortality
Only two studies reported on the infection related mortality, including 637 episodes. A difference in favour of carbapenems was observed, but this difference was not statistically significant (RR 0.45, 95% CI 0.07–3.05) (Fig. 5).
Adverse events
Risk of nausea/vomiting was higher in the carbapenems arm compared with BL/BLIs (RR 2.83, 95% CI 1.35–5.92, P = 0.006). Inversely, BL/BLIs were associated with a significant increase in the risk of diarrhea relative to carbapenems (RR 0.47, 95% CI 0.27–0.80, P = 0.006). In addition, no significant difference was observed for rash (RR 0.87, 95% CI 0.39–1.92), hepatotoxicity (RR 1.16, 95% CI 0.31–4.40), and seizure (RR 4.05, 95% CI 0.68–24.31) (Fig. 6).