Immunosuppression withdrawal in renal transplantation.

SINCE MUCH of the morbidity and mortality late after renal transplantation is directly or indirectly linked to complications of immunosuppression, a number of studies have examined whether one or more agents can be safely withdrawn. This strategy is designed to capture the early posttransplant benefit of reduced acute rejection without paying the price of long-term toxicity. A metaanalysis of randomized controlled trials of prednisone withdrawal and avoidance was published in 1993. However, since then several additional prednisone withdrawal trials have been conducted, including two very large, multicenter, randomized, controlled trials. Cyclosporine A (CsA) may not only contribute to morbidity and mortality after renal transplantation, but may also cause nephrotoxicity that is indistinguishable from chronic allograft nephropathy. A number of randomized controlled trials have been conducted to examine whether CsA can be safely withdrawn after renal transplantation, and we published a metaanalysis in 1993. However, the length of follow-up was relatively short in these studies. Since then several groups have reported the results of long-term follow-up, and additional CsA withdrawal trials have been conducted. Withdrawal of either prednisone or CsA remains controversial. Therefore, we conducted a metaanalysis using a fixed-effects model to combine randomized controlled trials examining the effects of prednisone or CsA withdrawal on acute rejection and graft failure. In addition to pooling studies, we also examined possible reasons for differences in their results. There were 10 randomized controlled trials that examined the effects of prednisone withdrawal (n 5 1984 patients); follow-up was 31 6 21 months. The pooled mean difference in the proportion of patients with acute rejection after prednisone withdrawal versus control was 0.14 (95% confidence interval 5 0.09–0.18, P , .0001), indicating that the proportion of patients with acute rejection following prednisone withdrawal was significantly greater compared to control. A test for homogeneity (x 5 13.7, P . .05) indicated that the studies were relatively homogeneous. The relative risk (relative risk of 1.0 indicates no risk) for graft failure was 1.38 (1.03–1.72, P , .05), that is, significantly more patients lost their grafts in the prednisone withdrawal group compared to controls. The test for homogeneity (x 5 12.3, P . .05) indicated that the individual study results were relatively homogeneous. Ten studies examined CsA and provided adequate data on acute rejection (n 5 1049); follow-up 50.7 6 33.9 months. The pooled mean difference for acute rejection between treatment and control was 0.11 (0.06–0.66, P , .0001), that is, the proportion of patients with acute rejection following CsA withdrawal was significantly greater compared to control. The test for homogeneity (x 5 64.9, P , .001) indicated that study results were heterogeneous. To explore possible reasons for this heterogeneity, univariate regression analysis was carried out to determine which, if any, patient or study characteristics correlated with the effect of CsA withdrawal on acute rejection. A study quality index (r 5 20.70, P 5 .024) and whether the study was published in a peer-reviewed journal (r 5 20.73, P 5 .17) correlated inversely with the difference in the rate of acute rejection (withdrawal vs. control). This suggested that CsA withdrawal was associated with more acute rejection in studies of lower quality. Twelve studies provided adequate data on graft survival following CsA withdrawal (n 5 1151); follow-up 45.1 6 32.9 months. In these studies the relative risk of graft failure after CsA withdrawal was 1.06 (0.79–1.32, P . .05), but a x test indicated there was considerable heterogeneity between studies. In a regression analysis, a study quality index was inversely proportional to the relative risk for graft failure (r 5 20.65, P 5 .024), suggesting that CsA withdrawal had a more adverse effect on graft survival in studies of lower quality. The pooled relative risk for graft failure among five studies (n 5 189 patients) that were not published in peer-reviewed journals was 2.97 (0.99–3.99, P . .05), x test for homogeneity 5 39.4, P , .001, indicating that these study results were heterogeneous. In contrast, the pooled relative risk for graft failure among the seven studies (n 5 962 patients) published in peer-reviewed journals was much less, 0.95 (0.65–1.25, P . .05), x test for