Sirolimus (SRL) and liver transplantation have had a complicated relationship. The 2002 FDA black box warning of de novo SRL use in liver transplantation warned of increased hepatic artery thrombosis, graft loss and death, contradicting the experience of many clinicians. This warning has served to taint the use of SRL for liver transplantation for most clinicians ever since. And yet, the source data upon which the warning was based was never published. In 2009, the FDA issued a second warning, this time regarding SRL conversion, after a registration trial found a higher overall treatment failure rate and “increased mortality in patients converted from a calcineurin inhibitor (CNI) to SRL”(1). Unfortunately until now, all of these warnings have come without any published reports, so clinicians have been unable to critically review these warnings to determine why they contradict the many largely favorable single-center reports of de novo (2,3) and conversion use of SRL in liver transplantation (4,5).
In this context, the study by Abdelmalek et al. (6) is extremely important, because it is the first publication of the data upon which the 2009 FDA warning against SRL conversion in liver transplant recipients was based. The study aim was to address the safety and efficacy of converting from a CNI-based regimen to a SRL based regimen in patients >6 months posttransplant. Although it intended to show that SRL improved renal function, the study failed to demonstrate either superiority with SRL conversion in terms of improved glomerular filtration rate or noninferiority with respect to the composite endpoint of graft loss and death. Although the conversion trial has carried the scientific and regulatory weight usually accorded a randomized, controlled multicenter registration trial, there are sufficient problems in its methodology that it casts doubt on its safety findings; in effect, the study inadvertently set up SRL to fail. This then brings into question the validity of the 2009 FDA warning that stemmed from this study.
The primary issue with the study methodology is that the SRL dose far exceeded accepted practice, with a loading dose of 10–15 mg and supratherapeutic serum levels of 10–20 ng/mL that were nearly three times higher than required. In our experience, a SRL dose of 2 mg daily, with no loading dose, with serum levels of 4–10 ng/mL allows successful SRL conversion; neither author of this editorial (with a collective experience with SRL approaching 2000 liver patients) has administered a 15-mg dose. The excessive dosing was unnecessary for effective immunosuppression and invariably contributed to the high rate of adverse events as well as the high rate of early drug discontinuation. Ultimately, with so many early discontinuations of SRL in this intent-to-treat trial, it is difficult to truly assess the drug's clinical impact on renal function.
Second, documentation of critical adverse events was problematic. “Herpes infection” was surprisingly the most commonly identified and highest statistically significant “infection” in the SRL group, yet it was “investigator-reported and … may not have been proven diagnostically”. Infections were cited by the FDA as an important adverse event, so the differentiation of serious herpetic infections (which are quite rare) from unverified and uncultured mouth sores (which are quite common with SRL and never dangerous) is crucial for ascertaining a drug's true risk.
Third, many important study endpoints were never documented. In table 2, in the SRL group, 50% of death (13 of 26 pts) and 46% of biopsy-proven rejections (21 of 46 pts) were recorded as such despite missing data. Even more troubling is that 4 of the remaining 13 deaths within the SRL cohort occurred in patients who “in violation of protocol, had either clinical documentation of preexisting malignancy at enrollment or in whom a malignancy was strongly suspected to have existed preenrollment”. Some of these problems may have reflected inherent difficulties in codifying study conduct at 82 participating centers worldwide. Although all data must be strictly registered from a regulatory standpoint, the omission of corroborative data and misclassification of these clinical outcomes cast doubt on the study findings.
Usually a randomized, controlled, multicenter registration trial provides the definitive answers with regards to safety and efficacy of a drug. Unfortunately, there are enough issues with this study's methodology—with the excessive SRL dosing and uncertainty in properly identifying adverse events and trial endpoints—that this is not the case. When a study drug carries preexisting negative biases as in the case of SRL, it is imperative the study methodology be sound.
More importantly, this shows that for some FDA warnings, the “devil is in the details”. Despite the flaws in this study's methodology, the FDA was obligated to release a warning related to these outcomes. However, with the publication of Abdelmalek et al.'s work in this issue of AJT, there is now transparency, and with that, we question the validity of the 2009 FDA warning. The investigators and the corporate sponsor should be commended for publishing this important study in full. To this effect, there should also now be a need to publish the data from the original de novo registration trials, in a manner similar to how Abdelmalek and his group have done. Only with a similar transparency of the details, will we be able to fully clarify the true risks of SRL in liver transplantation.