1. Literature Search
A total of 940 articles were confirmed using our electronic database search. After removing duplicate articles and screening the studies titles and abstracts, 56 articles meeting the inclusion criteria underwent full‐text assessment resulting in 10 relevant studies [13-22]. Flowchart of the selection procedure (Fig. 1).
2. Study Characteristics
In 2 studies treatment arms were evaluated separately: PLD plus carbo vs PAC plus carbo (four trials [13-16]: 851 PLD plus carbo and 916 PAC plus carbo). PLD vs other monotherapy (eight trials [17-22] : 963 PLD and 1,017 other monotherapy), Vergote2009  was categorized into both trials: PLD vs Topotecan and PLD vs Canfosfamide; Kaye2012  was categorized into both trials: PLD vs 200mg Olaparib and PLD vs 400mg Olaparib. All features of the included studies are demonstrated in Table.1. We assessed the study quality based on The Cochrane Collaboration's tool and the criteria specified in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions （Table.2. Each study was evaluated for potential bias and quality by two independent authors experienced. Disagreements were resolved by consensus.
3. Extraction of Data
3.1 Doublet Regimens overall analysis: PLD plus carbo vs PAC plus carbo
PLD plus carbo was associated with a significant improvement in PFS (HR, 0.85; 95% CI, 0.74-0.97; I2=28%; p =0.02). And OS was similar to standard chemotherapy regimen PAC plus carbo (HR, 1.00; 95% CI, 0.88-1.14; I2=0%; p=0.99). (Fig.2)
Grade 3-4 toxicities: PLD plus carbo was associated with a decreased risk of an allergic reaction (RR,0.38; 95% CI,0.19-0.78; I2=0%; p<0.01), Arthralgia/myalgia (RR, 0.19; 95% CI, 0.05-0.68; I2=0%; p=0.01), and neutropenia (RR, 0.76; 95% CI, 0.67-0.86; I2=0%; p<0.01). PLD plus carbo was associated with an increased risk of anemia (RR,1.82; 95% CI,1.22-2.71; I2=0%; p<0.01) and thrombocytopenia (RR,2.67; 95% CI,1.94-3.67; I2=0%; p<0.01). There was no difference risk of fatigue/asthenia (RR, 1.10; 95% CI, 0.78-1.56; I2=0%; p=0.57), mucositis/stomatitis (RR,2.04; 95% CI,0.90-4.66; I2=0%; p=0.09), hand-foot syndrome (RR,2.76; 95% CI,0.50-15.16; I2=0%; p=0.24) and vomiting (RR, 1.38; 95% CI, 0.72-2.66; I2=44%; p=0.33). (Fig.4)
3.2 Monotherapy Regimens overall analysis: PLD vs Single agent
PLD was similar in PFS (HR,1.02; 95% CI,0.90–1.16; I2=0%; p=0.72) and OS (HR,0.88; 95% CI, 0.77–1.01; I2=0%; p =0.07) to other single agent. (Fig.3)
Grade 3-4 toxicities: PLD was associated with an significant increased risk of mucositis/stomatitis (RR, 0.10; 95% CI, 0.04–0.23; I2=0%; p<0.01), and Hand-foot syndrome (RR, 0.03; 95% CI, 0.01–0.09; I2=0%; p <0.01) compared with the other monotherapies. There were no difference risk of anemia(RR, 1.26; 95% CI, 0.86–1.83; I2=0%; p=0.23), vomiting (RR, 0.97; 95% CI, 0.57–1.66; I2=38%; p=0.91), fatigue/asthenia (RR, 1.09; 95% CI, 0.73–1.64; I2=19%; p=0.66), thrombocytopenia (RR, 1.73; 95% CI, 0.93–3.24; I2=4%; p=0.08) and neutropenia (RR, 1.32; 95% CI, 0.59–2.96; I2=86%; p =0.50).( Fig.5)
3.3 Subgroup analysis
We perform neutropenia side effect subgroup analyze based on different drugs in Monotherapy Regimens (I2=86%), one subgroup [19, 20] showed that Canfosfamide and Patupilone were lower risk than PLD (RR, 0.39; 95% CI, 0.21–0.72; I2=33%; p<0.01), the other subgroup [17-19, 21] showed that Gemcitabine, Topotecan, LIFA and Olaparib were higher risk than PLD (RR, 2.26; 95% CI, 1.61–3.17; I2=0%; p<0.01). And perform subgroup for the differences of bad toxic and side effects based on the different doses of PLD: In doublet regimens: anemia, 30mg/m2 vs 45mg/m2 PLD (I2=0%); thrombocytopenia, 30mg/m2 vs 45mg/m2 PLD (I2=0%). There were no difference in the incidence of adverse reactions at different doses of PLD. In monotherapy regimens: mucositis/stomatitis, 40mg/m2 vs 50mg/m2 PLD (I2=60.5%); Hand-foot syndrome, 40mg/m2 vs 50mg/m2 PLD (I2=30.2%). There were difference in the incidence of adverse reactions at different doses of PLD.
3.4 Publication bias
To assess all studies PFS potential publication bias, we used Egger's linear regression test (p=0.635), as well as Begg's funnel plot (p=0.592). The tests results showed that this updated meta-analysis was no significant publication bias. (Supplementary Material 2.)