Letter: the effect of sirolimus on recurrence and survival in liver transplant recipients with hepatocellular carcinoma
Abstract
Linked Content
This article is linked to Menon et al papers. To view these articles visit https://doi.org/10.1111/apt.12185 and https://doi.org/10.1111/apt.15038.
As hepatocellular carcinoma (HCC) increases as an indication for liver transplantation, and as some centres push to include patients with more advanced tumours, it is crucial that we have a good understanding of strategies used to mitigate the risk of recurrence.
We therefore read with interest the systematic review and meta-analysis by Menon et al,1 which investigated the effect of sirolimus vs calcineurin inhibitor-based immunosuppression on tumour recurrence and survival in patients undergoing liver transplantation for HCC.
Although this was published in 2012, it continues to be influential and relevant, and has been cited 11 times since the beginning of 2017, including in the most recent International Liver Transplantation Society consensus statement.2 Unfortunately, their review contains data extraction and methodological errors, which we believe compromise the reliability of the results. These are listed below:
Data extraction:
- In their table, Menon et al correctly reported the total recurrence in Zimmerman et al3 as 3/45 (6.7%) and 9/52 (17.3%) in the sirolimus and calcineurin inhibitor-based groups, respectively. However, in their pooled meta-analysis, these values inexplicably change to 7/45 (16%) and 15/52 (29%).
- Recurrence-related mortality with sirolimus was recorded from Zimmerman et al3 as 4/45, even though the number with tumour recurrence was extracted as 3/45. These figures were incongruent.
- Overall mortality from Chinnakotla et al4 was reported as 12% (15/121) and 32% (34/106) with sirolimus and calcineurin inhibitor immunosuppression. The correct values in the primary study were 16% (19/121) and 50% (53/106), respectively.
- One-, 3- and 5-year recurrence-free survival from Chinnakotla et al4 were extracted as 94%, 85% and 80%, respectively, which was identical to the figures they state for overall survival. This is illogical, and in fact, in the primary study and its appendix, there is no mention of recurrence-free survival data.
Methodological errors:
- The authors inappropriately switched to the less conservative fixed effect model if I2 was below 50%. This practice is strongly discouraged, as the decision about which model to use should be based on an understanding of the primary studies, and not the outcome of heterogeneity statistics.5 As the maximum amount of pooled studies was three, this test will be notably under-powered. Since there were clear differences in tumour characteristics and immunosuppression protocols, the random-effects model would have been more suitable.
- Regardless, there is an error in their heterogeneity statistics for the meta-analysis of recurrence-related mortality. Menon et al calculated I2 = 0%. Visual inspection of the forest plot suggests that this was incorrect. Using the same data, we re-calculated I2 as 67%, which under their methods, indicates that a random-effects model should have been used. It appears Menon et al inadvertently copied the same heterogeneity statistics from the overall mortality meta-analysis, which were identical. When we re-performed the recurrence-related mortality pooled analysis using a random-effects model, the summary effect no longer reached significance.
Peer review of meta-analyses poses some difficulties for journals because reviewers do not have the resources to check all the input data for themselves. In fact, a meta-analysis on the same topic published this year also contains extensive extraction mistakes that compromise the results.6 However, it is encouraging that AP&T has led the way in recent years by including a specialist technical reviewer for such papers.
Ultimately, this review by Menon et al prominently influenced the collective understanding of the role of sirolimus in HCC transplant patients. Regrettably, we believe it contains important errors that should be elucidated.
ACKNOWLEDGEMENT
Declaration of personal interests: None.
FUNDING INFORMATION
None.
