Page 218 - Read Online

Tang et al. Hepatoma Res 2019;5:19 I http://dx.doi.org/10.20517/2394-5079.2019.07 Page 3 of 9

Study selection
Two investigators (Tang WN, Deng Y) independently screened the titles and abstracts of retrieved articles
to choose potential relevant articles. Disagreement about particular studies were discussed and resolved by
consensus.

Data extraction
Data extraction was carried out independently by the two reviewers (Tang WN, Deng Y). The following
information was extracted from the eligible studies: information of the articles (author, published year,
and study design), patient characteristics (number, age, area or nationality, race, and gender), liver disease
condition (hepatitis virus infection, Eastern Cooperative Oncology Group (ECOG) performance scale,
Child-Pugh stage, prior therapy), intervention in patients (agent, target, dosage, duration of dosing), outcome
of efficacy (ORR).

Statistical analysis
The data analysis process was initially conducted by the third author (Ma LT). The pooled estimated ORRs
and their 95%CI were derived. Meta and Metafor Packages in R were utilized to accomplish meta proportion
analysis. Logit transformation of raw proportion was performed before further analyses to increase validity.
2
The ratio of between-study heterogeneity to total heterogeneity was quantified by I and P value. The
2
assumption of homogeneity was considered invalid for I > 25% and P < 0.10. A chi-square test (Q-test) was
performed to test whether the heterogeneity between studies existed or not. If between-study heterogeneity
were not significant, a fixed model would be applied to get a summarized proportion; otherwise a
DerSimonian and Laired random effects model would be adopted. Two side P < 0.05 was considered
statistically significant.

RESULTS
Eligible studies
A total of 545 related articles were identified by the initial search strategy. After screening titles and
abstracts, 525 studies were excluded because of irrelevant topics, review articles, molecular mechanism
studies, experiments on animals, and clinical trials on recruiting. We then carefully reviewed the full texts
of the remaining 20 potentially eligible papers. And then, 13 articles were excluded, because one of them was
not able to obtain full test, two of them did not have the outcome of ORR, and ten papers were case-reports.
Finally, seven studies were chosen for the following analysis. Figure 1 shows the study selection flowchart.
Data from all eligible studies were obtained from published manuscripts.

Study characteristics
A total of 462 populations from seven studies were included in this meta-analysis. There was one paper
published in 2013, five papers in 2017 and one paper in 2018. Six studies were carried out with multi-center
clinical trials and one study was carried out with mono-center in China. Most of studies were in phase 1
or phase 2 of clinical trials. There were two studies conducting with CTLA-4 inhibitor (Tremelimumab),
four studies with PD1/PD-L1 inhibitor (Pembrolizumab, Nivolumab, Durvalumab) and one study with
combination of PD1 inhibitor and CTLA inhibitor (durvalumab and tremelimumab). Table 1 shows the
main characteristics of the eligible trials.

Efficacy of immune checkpoint inhibitors in HCC
Overall, the pooled estimated ORR of patients treated with ICIs was 19.8% (95%CI: 16.4%-23.7%). No
substantial heterogeneity was observed among single-study (Q = 2.0427, P = 0.92, I = 0.0%). Study of El-
2
[18]
Khoueiry et al. weighed most with estimated proportion of 19.6% (95%CI: 14.8%-25.5%). And the second
[17]
weighted article was Zhu et al. with estimated proportion of 17.3% (95%CI: 11.2%-25.8%). See Figure 2.