We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines when searching for articles, selecting studies, evaluating article quality, and analyzing data ( 23 ). We searched PubMed, Web of Science, EBSCO, Cochrane Library, and Ovid MEDLINE for eligible articles published before October 20, 2021. The search terms used were “doripenem,” “doribaxa,” “finabax,” and “pneumonia.” No publication year or language limitations were applied during the literature search. Only clinical studies comparing the efficacies of doripenem and other alternative antibiotics in the treatment of adult patients with HAP/VAP were included in the study. Next, 3 investigators reviewed the full texts of the candidate articles to finalize the experimental and control groups for meta-analysis. They reviewed study methods, sites, durations, populations, treatment regimens, and clinical outcomes reported in the articles. Initially, 2 investigators (Lan and Chang) independently examined the publications to avoid bias, and the third investigator (Lu) was consulted for resolving any disagreements. The reported study designs, durations, sites, populations, antibiotic regimens, and outcomes for doripenem and comparators were recorded, along with studies’ years of publication. The protocol was registered in PROSPERO as CRD42021284915.

The primary outcome was all-cause mortality at 28 to 30 days, and the secondary outcome was the clinical cure rate, which was defined as improvement or lack of progression of baseline radiographic findings resolution of signs and symptoms of pneumonia at follow-up.

The Cochrane Risk-of-Bias tool was used to assess the quality and associated risk of bias of the included RCTs ( 24 ). The Newcastle– Ottawa scale was used to assess the quality of retrospective observational trials ( 25 ). A random-effects model in Review Manager (version 5.3) was used for statistical analyses. Pooled odds ratios (ORs) and 95% CIs were calculated for outcome analyses.

The online database search yielded 421 articles, including 97 from PubMed, 142 from Web of Science, 28 from EBSCO, 37 from the Cochrane Library, and 116 from Ovid MEDLINE (Appendix 1). After screening title and abstracts, we excluded 177 irrelevant articles. Next, after screening the full texts, we excluded 6 irrelevant studies. Finally, five clinical studies14-18 were enrolled for meta-analysis (Figure 1).

In general, 3 RCTs14-16 and 2 retrospective studies were included in the meta-analysis ( 17 , 18 ). The RCTs were multicentered, whereas 2 retrospective studies were single-centered conducted in France and Taiwan each (Table 1). One study ( 15 ) used a 1-g (high) dose of doripenem, whereas the other 4 studies ( 14 – 18 ) used 500-mg dose every 8 hours. Moreover, in the RCT using 1 g of doripenem, doripenem was given daily for 7 days, whereas the comparator imipenem administered at 1 g every 8 hours for 10 days ( 15 ). In other 2 RCTs, both doripenem and the comparators were used for a total of 7 to 14 days. In the 2 retrospective studies, antibiotic treatment was left to the physician’s discretion. The median antibiotic treatment durations used in the study of Liu et al were 10 ( 5 – 17 ) and 11 ( 7 – 15 ) days for doripenem and meropenem, respectively ( 18 ). Three RCTs ( 14 – 16 ) demonstrated had high risk of selection, performance, and detection bias and one retrospective study had deficiency of timing and adequacy of follow-up ( 17 ).

In the pooled analysis of all 5 studies, all-cause mortality did not differ between doripenem and the comparators (14.8% and 16.9%,respectively; OR 1.08, 95%CI 0.78-1.5, I2=0%, Figure 2). A similar trend was seen in the pooled analysis of the 3 RCTs (13.6% and 12.3%, respectively; OR 1.12, 95%CI 0.79-1.6, I2=0%) and 2 retrospective studies (22.5% and 23.1%, respectively; OR 0.9, 95%CI 0.31-2.58, I2=71%). In patients with VAP alone, no differences were noted between doripenem and the comparators in the pooled analysis of 4 studies ( 14 – 18 ) (13.8% and 17.9%, respectively; OR 0.98, 95%CI 0.56-1.71, I2=33%). Only 2 studies ( 15 , 17 ) recorded mortality of patients with P.aeruginosa, the and subgroup analysis demonstrated that mortality did not differ between doripenem and the comparators (26.5% and 33.3%, respectively; OR 0.81, 95%CI 0.37-1.73; I2=76%).

In the subgroup analysis of the 4 studies ( 14 , 15 , 17 , 18 ) that used carbapenem as the comparator, the all-cause mortality of the doripenem and carbapenem groups did not differ (18.4% and 15.4%, respectively; OR 1.3, 95%CI 0.58-1.81, I2=44%). In the subgroup analysis of the 3 studies ( 14 , 15 , 18 ) comparing doripenem and imipenem, no differences were observed (14.2% and 14.0%, respectively; OR 0.92,95%CI 0.42-2.01, I2=65%). In the subgroup analysis of the 2 studies ( 17 , 18 ) that compared doripenem and meropenem, no differences were observed (22.5% and 16.7%, respectively; OR 1.45, 95%CI 0.83-2.52, I2=0%).

In the pooled analysis of the four studies ( 14 – 17 ) that reported clinical responses, the clinical cure rates were 60.7% and 57.3% for doripenem and the comparators, respectively (OR 1.06, 95%CI 0.80-1.39, I2=0%; Figure 3). A similar trend was observed in the subgroup analyses for VAP (56.6% and 54.4%; OR 1.01, 95%CI 0.73-1.39, I2=15%). Moreover, the clinical cure rate for doripenem and the comparators in the treatment of HAP/VAP did not differ for Enterobacteriaceae bacteria (62.2% and 62.1%, respectively; OR 1.00, 95% CI 0.68-1.47, I2 = 0%) or P. aeruginosa (69.1% and 58.1%, respectively; OR 1.35, 95%CI 0.2-9.03, I2=75%) or for the presence (65.5% and 64.5%, respectively; OR 1.05, 95%CI 0.36-3.05, I2=0%) or absence (73% and 69%, respectively; OR 1.22, 95%CI 0.85- 1.73, I2=0%) of bacteremia.