We retrospectively reviewed data from AA transplant recipients at two Chicago-area academic medical centers from 2003 to 2011. During this period, The University of Chicago followed a continued steroid therapy (CST) protocol, and the University of Illinois employed a rapid steroid withdrawal protocol (RSW). Practice patterns at each center post-transplant are summarized in Table 1. Inclusion criteria for the study were AA transplant recipients at least 18 years of age who received either a deceased donor (including ECD and DCD) or living donor kidney, and were induced with anti-thymocyte globulin. Exclusion criteria included: 1) patients requiring corticosteroids prior to transplantation that were continued after transplant; 2) re-transplants or multi-organ transplants; and 3) positive cross-match and ABO incompatible transplants that required maintenance steroid therapy. The Institutional Review Board at both the University of Chicago and University of Illinois at Chicago approved this study.

Primary end-points included patient, graft, and death-censored graft survival. Secondary end-points included the estimated glomerular filtration rate (eGFR) at 1 and 5 years as determined by the Modification of Diet in Renal Diseases (MDRD) equation, the 1 and 5 year incidence of acute cellular and humoral rejection, and cumulative incidence of post-transplant diabetes mellitus (PTDM) defined as the a fasting glucose >126 mg/dL or random glucose >200 mg/dL requiring the initiation of oral anti-hyperglycemic or insulin based agents after transplant.

Patients in the CST group were induced with 4 doses of anti-thymocyte globulin (maximum dose 100 mg/day). Either mycophenolate mofetil 1000 mg twice a day or mycophenolate sodium 720 mg twice was used as an anti-proliferative agent. Corticosteroid treatment included intravenous methylprednisolone followed by a taper to maintenance steroid dosing of 5 mg per day at 1 month post-transplant. Patients were maintained on tacrolimus with target 12-hr trough level ranging 6 - 9 ng/ml for the first six months and then 4 - 7 ng/ml thereafter (Table 1).

In the RSW group, patients were induced with 5 doses of 1.5 mg/kg/day anti-thymocyte globulin based on ideal body weight. Mycophenolate mofetil 1000 mg twice a day or mycophenolate sodium 720 mg twice a day

was also used as an anti-proliferative agent, and prednisone was tapered quickly from 1 mg/kg/day to 0.25 mg/ kg/day and off by post-operative day 6. Patients were maintained on tacrolimus with target trough of 8 - 12 ng/mL in the first 2 months, followed 5 - 10 ng/mL thereafter.

The CST group received valganciclovir cytomegalovirus (CMV) prophylaxis for 3 months in the intermediate risk (donor CMV positive/negative and recipient CMV positive) and high risk group (donor CMV positive and recipient CMV negative). All recipients also received trimethoprim/sulfamethoxazole single-strength daily for Pneumocystis prophylaxis in the first 6 months and then three times weekly indefinitely post-transplantation. Fluconazole fungal prophylaxis was provided for 1 month immediately post-transplantation.

The recipients in the RSW group that were at intermediate or high risk for CMV infection received valganciclovir for 6 months and all recipients received trimethoprim/sulfamethoxazole single-strength daily for the first 12 months post-transplant. No fungal prophylaxis is provided after discontinuation of steroids. For both the CST and RSW groups, acyclovir HSV prophylaxis was used for 1 month if recipients were low risk for CMV (donor and recipient CMV negative).

The incidence of rejection was determined by either biopsy-proven rejection or empiric treatment for rejection as described below. The types of rejection were determined using the Banff ’05 criteria when possible. In cases where C4d staining was not done, the Banff ’97 criteria were used. Cases that had both an acute cellular component and antibody-mediated component were categorized as antibody-mediated rejection.

The diagnosis and management of rejection varied slightly between the two groups. In the CST group, all clinically suspected rejections had an ultrasound-guided renal allograft biopsy performed. For Banff 1A or 1B acute cellular rejection (ACR), patients were treated with methylprednisolone 500 mg daily for 3 days followed by a quick taper. Banff 2A or greater ACR were treated with anti-thymocyte globulin. For the RSW group, treatment for clinically suspected rejection without biopsy or a borderline/Banff 1A or 1B ACR included methylprednisolone 500 mg daily for 3 days without an oral taper. Patients with Banff 2A or greater ACR were treated with anti-thymocyte globulin, dosed according to ideal body weight. As per protocol, subjects with first time rejections in the RSW group were not started on oral steroid therapy.

To assess the baseline differences in patients from the two centers, we examined the difference in means and standardized difference in means (Cohen’s D statistic) for a set of demographic and clinical characteristics determined prior to the transplant. The two samples were out of balance with respect several important risk factors. To reduce bias from these baseline differences, we estimated propensity scores using a logistic regression of treatment group membership (RSW/CST) on a list of donor, recipient, and transplant factors including recipient age, gender, body mass index, history of diabetes, pre-transplant dialysis, time on dialysis, primary renal disease, PRA ≥ 30%, Hepatitis C; donor age, gender, race, body mass index (BMI), type; HLA matches, CMV risk, and transplant era. Next, we used inverse probability of treatment weights (IPTW) to assign a weight to each member of the CST group (control). To avoid instability from very large weights, the weights in the CST group were normalized and individuals were trimmed from the sample if their weight represented more than 5% of the sum of the weights. To maintain symmetry, we also excluded individuals in the RSW group (treatment) if they had propensity scores greater than the minimum propensity score among the individuals who were trimmed from the control group. We examined several candidate specifications for the propensity score model and selected the specification that achieved the most balanced sample. Categorical and continuous outcomes were estimated using IPTW for proportions and means, respectively, and with 95% confidence intervals.

We used Kaplan-Meier analysis with IPTW to compare graft and patient survival between the two groups. A Cox proportional hazards model using IPTW (unadjusted model) was used to estimate the effect of the maintenance regimen on survival with adjustments for recipient (model 1), donor, and transplant covariates (model 2). The addition of covariates into the Cox proportional model with IPTW provides a doubly robust estimation that increases the chances of an accurate estimation of the outcome [25] .

From 2003 to 2011, 194 patients from the CST group and 212 from the RSW group were initially included in the study. Table 2 shows that, prior to IPTW and trimming, the CST group had more deceased donors, especially DCD, more time on dialysis, fewer black donors, and higher proportion of individuals with a PRA > 30%. After applying IPTW and trimming, there were 150 individuals in the analytic CST group and 157 in the analytic RSW group.

Prior to outcome analysis, a basic question is whether the analytic sample is sufficiently balanced for sound causal inference. There is no universal standard to apply, of course. However, from a clinical perspective, the analytic sample appears well balanced with respect to key risk factors. In addition, the standardized mean difference in covariates (Cohen’s D) is well below 0.25 standard deviations for each of the baseline covariates so that the remaining imbalances is not “too large” from the perspective of one common rule of thumb in the statistical matching literature [26] [27] . In fact, after IPTW and trimming, most baseline covariates have Cohen’s D statistics less than 0.10 standard deviations and the most imbalanced variable has a Cohen’s D of 0.16 standard deviations.

Graft, death-censored graft, and patient survival were assessed at 1, 3, and 5 years after transplant. In the CST analytic group, there were 25 graft failures and 24 deaths during the follow up period while the RSW analytic group had 24 graft failures and 18 deaths (Table 3). The graft survival between the two centers was similar during the follow up period. The 1, 3, and 5 year graft survival was 93%, 79%, and 59% in the CST analytic group and 94%, 84%, and 72% in the RSW analytic group (Figure 1). Death-censored graft survival showed similar results with 1, 3, and 5 year survival at 99%, 87%, and 74% in the CST analytic group and 97%, 92%, and 83% in the RSW analytic group. Patient survival was 94%, 91%, and 79% at 1, 3, and 5 years for the CST analytic group and 97%, 92%, and 87% in the RSW analytic group. The Cox proportional hazard models for graft and patient survival did not show any significant difference in survival between the two analytic groups (Table 4).

The causes of graft failure and patient death with a functioning graft are shown in Table 3. Overall, the causes

Reported as either mean (standard deviation) or proportions. ^aWeighted total, rounded to the nearest whole number.

of graft failure were similar between the two groups with interstitial fibrosis/tubular atrophy being the most common followed by acute rejection. The causes of death were difficult to determine for a number of cases, particularly in the RSW group. However, it appeared that the most common causes were either cardiovascular, infectious, or due to malignancy.

^aweighted total, rounded to the nearest whole number.

^aUnadjusted model with Inverse Probability Treatment Weight (IPTW) adjustment only; ^bmodel 1 includes IPTW plus recipient age, gender, BMI, diabetes, pre-transplant dialysis, dialysis time, cause of ESRD (hypertension, lupus, or other vs diabetes), PRA > 30%, and Hepatitis C status; ^cmodel 2 includes model 1 plus donor type (cadaver vs living), expanded criteria donor, donation after circulatory death, HLA matches, donor age, donor gender, donor race (black vs other), donor BMI, CMV risk (intermediate, high vs low), and transplant era (2006-2008, 2009-2012 vs 2003-2005).

The eGFR at 1 and 5 years after transplant are shown in Table 5. Overall, the eGFR were not statistically different between the two groups at 1 and 5 years after transplant. However, when we stratified subjects into type of kidney transplant, recipients of deceased donor kidneys in the CST analytic group had a higher eGFR compared to deceased donors in the RSW analytic group at 1 year (64.1 vs 52.7 mL/min/1.73 m², p = 0.047) that was no longer significant in those that survived out to 5 years (CST 55.3 vs RSW 51.3 mL/min/1.73 m², p = 0.59). Living donor kidney transplants were not different at either 1 or 5 years after transplant.

The incidence of rejection and types of rejection are shown in Table 5. Compared to the CST analytic group, the RSW analytic group had a higher rate of rejection at 1 (32% vs 10%, p < 0.001) and 5 years (44% vs 15%, p < 0.001) compared to the CST analytic group. When limited to biopsy-proven rejections in the first year, the rates

^aweighted total, rounded to the nearest whole number.

were similar (18% vs 10%, p = 0.10). The types of biopsy-proven rejection varied by group with more grade 1 and borderline rejections in the RSW analytic group while grade 2 rejections were more common in the CST analytic group.

The incidence of delayed graft function was much higher among the CST analytic group compared to the RSW analytic group (32% vs 10%, p < 0.001) (Table 5). The incidence of post-transplant diabetes mellitus was similar between the two groups (25% vs 22%, p = 0.66). In regards to viral infections, CMV viremia was more commonly found in the RSW analytic group compared to the CST analytic group (24% vs 11%, p = 0.036) while BK viremia was similar between two groups (18% vs 10%, p = 0.15).

Among the RSW group, 20 of the 157 individuals were started on prednisone after their transplant (Table 6).

The most common cause of starting prednisone was rejection (35%). The next most common cause was leukopenia at 25% while BK nephropathy and GI intolerance to mycophenolate mofetil or mycophenolate sodium were each 10%. Of the 20 individuals, 13 (65%) were started in the first year of transplant.