Studies included in meta-analysis
The results of literature search are shown in Fig. 1. Ten retrospective cohort studies published between 1998–2016 were included in the meta-analysis [6,7,9-14,19,20]. Three of the studies matched samples by PSM to minimize the effect of potential confounders. Table 1 summarizes the Main characteristics of the included studies. The recruitment period ranged from 1977 to 2013. Among 10 included studies, 1 was multicenter and 9 single-center. All studies were conducted in East Asia. Five of these ten studies were from Japan, four from China and one from Korea. Ten studies reported the data of postoperative overall survival and seven studies showed the data of postoperative disease-free survival. Each of the studies was independently assessed using the NOS and 7 studies (70%) were high quality with scores ranging from 7 to 8.
Characteristics of the studied patients
The total number of patients across all the studies was 3222, of which 377 patients with ruptured HCC and 2845 patients with non-ruptured HCC underwent hepatectomy. Six studies recorded liver cirrhosis. In a fixed model, there was no significant difference between the ruptured and non-ruptured HCC group (I2 = 7%, OR=0.84, 95% CI 0.66-1.08, P=0.12). Five studies recorded liver function. There was no significant difference in the percentage of Child-Pugh A between the two group (I2 = 6%, OR=1.05, 95% CI 0.68-1.62, P=0.84). The percentage of the patients with hepatitis B infection (OR=1.07, 95% CI 0.78-1.46, P=0.67) and with multiple nodules (OR=0.93, 95%CI 0.68-1.28, P = 0.67) was similar between the two group. Patients in the ruptured HCC group (33.4%) had a higher prevalence of macrovascular invasion in comparison with patients in the non-ruptured HCC group (23.1%), resulting in an OR of 1.51 (I2 = 40%, 95%CI 1.16-1.96, P = 0.002).
Overall survival
Data on OS was available in nine out of the ten studies. The meta-analysis revealed a statistically significant poorer OS for the patients with ruptured HCC compared to the patients with non-ruptured HCC. The pooled HR for OS calculated was 2.02 (95%CI 1.61-2.54, P < 0.00001) (Fig.2A). No evidence of publication bias was detected by visual examination of funnel plot (Fig.2B) and with Egger’s test (P = 0.466).
There was moderate, borderline significant heterogeneity for this analysis of OS (P = 0.05; I2 = 48%). Stratified meta-analysis showed that heterogeneity was largely attributable to the quality of the studies assessed by NOS (test for subgroups difference: P = 0.007) (Fig.3A). Both studies with high and low quality showed that ruptured HCC was associated to the poorer OS after liver resection, being pooled HR, respectively, of 1.65 (95% CI: 1.40-1.95; P < 0.00001) and 2.50 (95% CI: 1.94-3.23; P < 0.00001).
Stratified meta-analysis according to PSM analysis showed stable results and no heterogeneity among studies with PSM analysis versus non-PSM (test for subgroups difference: P = 0.48) (Fig.4A). In the pooled analysis, both the 6 non-PSM studies and 3 PSM studies demonstrated patients in the non-ruptured HCC group had significantly higher OS than the ruptured HCC group (P < 0.00001 and P=0.02, respectively). Of note, the studies with PSM analysis provided a lower estimate of the postoperative death risk associated with ruptured HCC, as compared to studies with non-PSM studies (P = 0.02 vs P < 0.00001)
Sensitivity analysis conducted by excluding each article once per time did not change the main summary estimate.
Disease free survival
DFS was reported by six studies. The meta-analysis revealed that patients in the ruptured HCC group had a significantly poorer DFS than patients in the non-ruptured HCC group, with a pooled HR of 1.92 (I2 = 22%, 95% CI 1.56-2.35, P <0.00001) (Fig.5A). There was not significant heterogeneity in this analysis (P=0.27; I2 =22%). No evidence of publication bias was found (Fig.5B).
Stratified meta-analysis according to the quality of the studies showed that ruptured HCC was associated with a poorer DFS after liver resection in both studies with high and low quality. The pooled HR of the different quality studies was 1.91 (95% CI: 1.52-2.40; P < 0.00001) and 1.94 (95% CI: 1.22-3.11; P = 0.005), respectively (Fig.3B).
Stratified meta-analysis according to PSM analysis showed DFS in the ruptured HCC group was significantly poorer than the non-ruptured HCC group in the non-PSM subgroup, with a HR of 2.13 (I2 = 0%, 95% CI 1.70-2.67, P <0.00001). On the other hand, meta-analysis of two PSM studies showed that there was not significant difference in the DFS between the two groups. The pooled HR was 1.18 (I2 = 0%, 95% CI 0.73-1.92, P=0.50) (Fig.4B).
Sensitivity analysis performed by elimination of each article once per time did not show significant deviations in overall summary estimate.
Morbidity, mortality and recurrence rate
Yang T et al [6] and Liu CL et al [7] reported similar postoperative morbidity between the two groups. Pooled analysis of these two studies demonstrated the morbidity to be 29.5% (n=52/176) in the ruptured HCC group and 29% (n=422/1454) in the non-ruptured HCC group (OR=1.07, I2 = 18%, 95%CI 0.76-1.51, P = 0.70) (Fig.6A).
Three studies reported hospital mortality. Patients in the ruptured HCC group (5.5%) had a higher mortality than the non-ruptured HCC group (3.2%), but the difference was not statistically significant (OR=1.89, I2 = 0%, 95%CI 1.01-3.52, P = 0.05) (Fig.6B).
Six studies reported postoperative recurrence rate. Among them, Liu CL et al [7] only reported extrahepatic recurrence rate. They found that there was significantly more extrahepatic recurrence in the ruptured group than the non-ruptured group (45.5% vs 25.8%, P=0.015). Pooled analysis of the other five studies demonstrated patients in the ruptured HCC group had a higher recurrence rate than the non-ruptured HCC group (75.8% vs 60.0%), however, the difference was not significant (OR= 1.76, I2 = 55%, 95% CI 0.97-3.2, P=0.06) (Fig.6C).