longest diameter should be limited to non-necrosis part in mRECIST. As a result, bias was unneglectable in comparing trials using different criteria. 5 Besides, time to progression (TTP) was set as the second end point in 4 trials (SHARP [NCT00105443], Asia-Pacific SHARP [NCT00492752], BRISK-FL [NCT00858871], and BRISK-PS [NCT00825955]). The TTP was defined as the time from randomization to radiologic disease progression. Because the numbers of patients who died without tumor progression were not available in these trials, it was difficult to evaluate the difference between TTP and PFS. It is best to disclose the criteria used in each clinical trial and which was evaluated in the original trial, TTP or PFS.Second, approximately 27.2% and 12.0% of patients with 2 previous systemic treatments were enrolled in the CELESTIAL (NCT01908426) and REACH (NCT01140347) trials, respectively, resulting in bias. In detail, the hazard ratio (HR) for death between 130 patients (28.0%) in the cabozantinib group and 62 patients (26.3%) in the placebo group with 2 previous systemic regimens was 0.90 (95% CI, 0.63-1.29), while the HR between patients with only sorafenib as the previous systemic regimen was 0.74 (95% CI, 0.59-0.92). Analysis of PFS revealed superiority of cabozantinib over placebo in both second-line and third-line cohorts. Unfortunately, the REACH trial did not supply a subgroup analysis like this. We believe that it is more reasonable to use HRs of subgroup analysis in network meta-analysis than HRs of the whole cohort, although the baseline characteristics of this subgroup were unavailable.In addition to the trials analyzed in this research, 1 trials of several immune checkpoint inhibitor-based therapies were ongoing, such as toripalimab plus lenvatinib (NCT04523493) and pembrolizumab plus lenvatinib and transcatheter arterial chemoembolization (NCT04246177). The update of this network meta-analysis 1 when these trials are completed will provide more evidence for clinical practice.