Characteristics of included studies
A total of 1344 records were screened in a primary search for possible related between HSPs expression and the clinicopathological features or prognosis in HCC. Fig. 1 illustrates the process of literature research and selection. Ultimately, 19 eligible studies were included in this meta-analysis [27-45]. As shown in Table 1, these studies were published from 2000 to 2018, and included a total of 1809 HCC patients. The sample size ranged from 40 to 392 patients, with 12 of the studies including > 60 patients and 7 including ≤60 patients. Fourteen studies evaluated patients from China, three from Korea, two from Japan. The NOS score of all these included studies were ≥6, which implied that they were of high quality.
Association between HSPs expression and clinicopathological characteristics in HCC patients
In this study, we evaluated the correlation between HSPs expression and clinicopathological features of patients with HCC, and the results were shown in Table 2 and Figures 2 to 3. Fourteen studies, including a total of 1653 HCC patients, were used to evaluate the differential expression of HSPs in HCC and non-HCC tissues. The results revealed that the expression of HSPs in HCC was significantly higher than that in the non-HCC tissues (OR =2.21, 95%CI = 1.50-3.25, P = 0.000, Fig. 2 A). The ORs for tumor differentiation were reported in 16 studies that included 1108 cases. The results evaluation of these data indicated that HSPs expression was associated with tumor differentiation (OR = 1.34, 95%CI = 1.08-1.66, P = 0.008, Fig. 2 B). Eleven studies containing 744 patients showed a significant relationship between HSPs expression and vascular invasion in HCC (OR = 1.29, 95%CI = 1.01-1.65, P = 0.039, Fig. 2 C). A greater proportion of patients with high HSPs expression had vascular invasion than those who had low HSPs expression. The ORs for lymphatic metastasis were reported in 5 studies on 282 patients, and the pooled OR with 95%CI was 1.63 (95%CI = 1.08-2.46, P = 0.021, Fig 2. D), indicating the high HSPs expression was associated with lymphatic metastasis. However, the HSPs expression were not significantly associated with any of the following parameters: gender (OR = 1.09, 95%CI = 0.85-1.41, P = 0.486, Fig. 3 A), HBsAg (OR = 1.14, 95%CI = 0.87-1.48 P = 0.347, Fig. 3 B), TNM stage (OR = 1.10, 95%CI = 0.90-1.35, P = 0.370, Fig. 3 C), tumor size (OR = 1.04, 95%CI = 0.75-1.45, P = 0.811, Fig. 3 D), tumor number (OR = 0.84, 95%CI = 0.48-1.47, P = 0.543, Fig. 4 A), AFP (OR = 0.99, 95%CI = 0.61-1.62, P = 0.983, Fig. 4 B), PVTT (OR = 1.19, 95%CI = 0.64-2.22, P = 0.591, Fig. 4 C).
Meta-analysis of OS associated with the expression of HSPs in patients with HCC
The HRs for the OS were reported in nine studies (Harimoto et al has two set of data), which included 412 and 344 HCC patients with and without HSPs expression, respectively. The meta-analysis revealed that HSPs expression was not significantly associated with OS (HR = 1.34, 95%CI = 0.84-2.12, P=0.216, I2= 67.0%, Fig. 5 A). However, the expression of HSP27 was significantly associated with the poor OS of HCC (HR = 1.69, 95%CI = 1.24-2.31, P=0.001, I2= 0.0%, Fig. 5 B). As seen in Fig 5 C and D, the pooled HRs of HSP70 and HSP90 showed that the expression of these HSPs could not act as an effective prognostic indicator in patients with HCC.
Meta-regression
To further explore the possible sources of heterogeneity, we conducted a meta-regression based on sample size, NOS score, country, and cut-off values (>10%, ≤10% or Not reported). The results of meta-regression were summarized in Table 3, and none of the above covariates were found to be a significant source of heterogeneity.
Publish bias and sensitivity analysis
No publish bias was found for expression, gender, tumor differentiation, HBsAg, TNM stage, tumor size, vascular invasion, tumor number, AFP, PVTT, and OS, according to the Begg’s and Egger’s test. However, publish bias existed for lymphatic metastasis. The funnel plots obtained from the Begg’s test are shown in Fig. 6, and the corresponding P values from Egger’s test are presented in Table 4. A sensitivity analysis was also calculated to determine whether individual studies influence the pooled ORs or HRs, and there were no significant impacts on ORs or HRs after excluding any one study, indicating the results were reliability.