When Less Becomes More: Life and Losses without the ‘Roids’?

Despite short-term successes and a marked reduction in the overall incidence of acute rejection, long-term allograft survival after kidney transplantation has remained largely unchanged for the past three decades.¹ Less than 50% of our transplant recipients have survived with a functioning graft 15 years after transplantation. The most important outcome of transplantation from the patient’s perspective is the survival of their graft.² Strategies to improve longer-term graft outcomes remain a key priority for patients, caregivers, and health professionals. Balancing the allograft response and reducing toxicity associated with immunosuppression is essential to achieving optimal outcomes. Current immunosuppression usually includes a combination of induction therapies, such as T cell–depleting agents or IL-2 receptor antibodies, followed by a typical maintenance regimen of calcineurin inhibitors, antiproliferative agents, and corticosteroids.

Corticosteroids are known to have anti-inflammatory and immunosuppressive effects and are useful to prevent acute rejection post-transplant.³ However, they are not without adverse effects. Chronic steroid use is associated with a wide range of complications including weight gain, hyperglycemia, osteoporosis, cataract, cardiovascular disease, and infections, and is a risk factor for premature death.⁴ In view of these complications, there had been a concerted effort to minimize long-term corticosteroids use. Many centers have advocated for steroid minimization or early withdrawal, often within days after transplantation. Others have been concerned regarding the short-term risks of steroid withdrawal, including acute rejection, leading to early allograft loss. Globally, the patterns of long-term steroid use vary considerably between regions, and within regions according to high-risk recipient or donor characteristics. Notably, most trials of early steroid withdrawal have excluded patients at increased risk of adverse allograft outcome, such as presensitized patients and those who have developed delayed graft function (DGF) post-transplant; both are known correlates with acute rejection and allograft loss.⁵ In the absence of trial-based evidence to inform policy and clinical practices in the high-risk settings, the next best option is to rely on observational data, often collected for completely different purposes, and multiple analyses (including subgroups) are performed to explain why some hypotheses hold or why they may be wrong.

In this issue of JASN, using data from the Scientific Registry of Transplant Recipients, Bae et al.⁶ retrospectively examine the effect of DGF on the association between early steroid withdrawal and transplantation outcomes including acute rejection, allograft loss, and mortality in all adult recipients transplanted between 2005 and 2017. Early steroid withdrawal was defined as complete steroid withdrawal by the time of discharge and DGF was characterized by the need for dialysis within the first week after transplantation. The authors excluded those who experienced early graft loss or death (within 7 days post-transplant) and recipients who were not discharged within 30 days after transplantation from the analyses, because the exact dates of steroid withdrawal were not recorded in the registry. Not surprisingly, there were substantial differences between the exposure groups. The authors attempted to account for confounding and indication biases using inverse probability of treatment weights and the influence of center effects on the choice of early steroid withdrawal versus continued steroid maintenance using a multilevel logistic model. Inverse probability of treatment weights is a statistical method used to create groups that are otherwise similar when examining the effect of an exposure. The individuals within the cohort were then assigned a weight on the basis of the likelihood that they would be exposed to the treatment effect under investigation. Applying this weight when conducting the Cox regression models may potentially reduce the effect of confounding.⁷ The analyses were robust, but key biases inherent to observational studies may still have occurred.

Despite the recent development of some sophisticated statistical approaches such as the high dimensional propensity score, evaluation of intended effects of therapies using observational studies is notoriously difficult because they suffer from the unavoidable problem of confounding and selection biases. This is even more challenging with very large health care databases where the data may be inaccurate, imprecise, and incomplete, including measurement error, resulting in misclassification of the exposure and outcomes with uncertain effects on the magnitude and direction of any observed association. In this study, the causes of death were extracted from multiple sources, including follow-up reports from transplant centers, the Centers for Medicare & Medicaid Services ESKD Death Notification Form, and the Social Security Death Master File. With this method of ascertainment, we might expect that death could be classified incorrectly. Reasons for misclassification may include cases with multiple causes of death and ill-defined causes of death on the death certificate. Similar concerns also apply to the acute rejection data where granular details from protocol and indication biopsy samples were unavailable.

In this analysis, the authors found that the association between early steroid withdrawal and transplantation outcome (death-censored graft loss and overall mortality but not acute rejection) was modified by the DGF status. Recipients who experienced DGF and had early steroid withdrawal experienced an excess risk of allograft loss of 16% compared with those on continued steroid maintenance. Although the magnitude of the increased risk of allograft loss was lower among those with immediate graft function, an increased risk of approximately 8% was observed among recipients with early steroid withdrawal. On the contrary, early steroid withdrawal appeared to be protective against all-cause and cardiovascular deaths in recipients with immediate graft function but not in those with DGF. Interestingly and unexpectedly, this interactive effect was not observed for acute rejection. That is, regardless of DGF status, recipients who experienced early steroid withdrawal reported a higher incidence of acute rejection. This is contrary to the hypothesis that the increased risk of graft loss from DGF was mediated by acute rejection.⁸

Where do we go from here? This study has been designed to answer an important and relatively unanswered research question and sought to address it using a sophisticated approach applied to a large observational dataset. There is clear potential to change clinical practice. Collaborative and concerted efforts within the international transplant community are needed to design and test the efficacy of steroid minimization and/or withdrawal among the ethnically diverse high-risk groups using large-scale, well powered, low-risk-of-bias intervention trials. The intervention must consider the timing and strategies for how steroids should be tapered, and their interaction with the other elements of the immunosuppressive protocols. The outcomes must be patient-relevant and should consider the relative values and preferences placed on the benefits (avoidance of steroid side effects) against the harms (acute rejection and allograft loss) of the intervention. Until then, steroids should be considered and maintained in high-risk recipients, including those with DGF.

DISCLOSURES

None.