IV.13 Analysis of patient and graft survival. Guidelines. Commentary on Guidelines IV.13: Analysis of patient and graft survival

Transcription

1 60 SECTION IV: Long-term management of the transplant recipient IV.13 Analysis of patient and graft survival Guidelines A. It is important for a transplant unit to follow-up on the results of their transplant activities. In order to achieve correct reports on graft and patient outcome in all patients, it is necessary to have sufficient resources, such as a computerized database, and continuous updates of patient information. All data collected should be subjected to validation procedures to ensure completeness and accuracy. B. Improved outcomes following implementation of new protocols, based on evaluation of clinical multi-centre trials, should be verified at local transplant centres since centres often include a range of patients different from those selected for the trial. C. The most widely accepted descriptor of outcome is the Kaplan Meier probability estimate of patient and graft survival. Survival estimates should be calculated at intervals of time after transplantation and should always be expressed with their 95% confidence intervals. D. Kaplan Meier survival estimates may be calculated in three ways. (i) Patient survival should be calculated from the date of transplantation to the date of death or the date of the last follow-up. (ii) Graft survival (non-censored for death) should be calculated from the date of transplantation to the date of irreversible graft failure signified by return to long-term dialysis (or retransplantation) or the date of the last follow-up during the period when the transplant was still functioning or to the date of death. Here, death with graft function is treated as graft failure. (iii) Graft survival censored for death with a functioning graft (death-censored graft survival) should be calculated from the date of transplantation to the date of irreversible graft failure signified by return to long-term dialysis (or retransplantation) or the date of last follow-up during the period when the transplant was still functioning. In the event of death with a functioning graft, the follow-up period is censored at the date of death. (Evidence level B) E. The outcome of transplants carried out at a centre should be compared with those achieved across a range of data from centres collated by national and international multi-centre registries. Interpretation of a centre s performance should take into account the number of transplants performed and the prevalence of major risk factors. F. Major risk factors that influence transplant outcome are identifiable by applying multivariate analytical methods to large multi-centre follow-up databases. Although these major risk factors may not be identifiable in individual centre data, they should nonetheless be taken into account in patient management. G. When designing a clinical trial or evaluating data from a recent trial, the expected improvement in graft survival resulting from a reduction in acute rejection may be estimated from a knowledge of the rejection and graft survival rates that existed prior to the introduction of the new therapeutic regimen. (Evidence level B) H. When designing or evaluating a clinical trial, it is important to analyse the power of the study to verify statistically the difference (in graft survival) that might be expected and its statistical significance. A study resulting in absence of statistically significant differences between two treatment groups with insufficient statistical power to verify a difference at the expected level should not be taken as evidence of absence of a true difference. (Evidence level A) Commentary on Guidelines IV.13: Analysis of patient and graft survival Guidelines A and B. The outcome of renal transplantation, in terms of patient and graft survivals as well as graft function, has improved to a great extent over the last decade. There are many reasons for this, primarily more efficient immunosuppression, improved diagnosis and treatment of acute rejection, prophylaxis of infection and better management of hypertension and hyperlipidaemia. Clinicians may be guided by the results of studies of new drugs, drug combinations or other methods of treatment, performed at single centres or in multi-centre trials. Often, trials have clear-cut inclusion and exclusion criteria that do not exactly mirror the wider range of patients or donors in clinical practice. Therefore, the implementation of new protocols may give different results at a single transplant centre compared with those of the trial. This is one argument,

2 SECTION IV: Long-term management of the transplant recipient among several others, for transplant centres to allocate resources to follow-up the outcomes of their transplant activities, especially after protocol changes. In this respect, it is important to have procedures in place for receiving updated clinical data from the nephrologists who follow the patients in the long term. It is also important that data on all patients are included and all data verified for accuracy. This applies particularly to dates. For example, when calculating graft survival, the time of observation should be from the date of transplant to the date when the patient was last seen by the medical team. It should not be assumed that a patient s graft is functioning beyond this date since this will tend to increase the estimated graft survival artificially. The definition of patient survival is not controversial. However, it must be stressed that fatal events occurring after graft loss should also be included in the data collection. This adds an extra burden on the system for data collection, but it is especially important for the evaluation of pre-operative risk factors for acceptance of a patient on to the waiting list. For example, factors such as cardiovascular disease, diabetes mellitus or old age may imply limited expectations of patient survival. Guideline C. Kaplan Meier (K M) or actuarial survival techniques are simply a way of graphically presenting outcome data and are widely applied as methods of summarizing transplant outcome data [1]. The K M technique may also be used to estimate the probability of other definable events that occur after transplantation (e.g. first acute rejection episode). Notwithstanding the apparent simplicity of this method, there are some features that may trap the unwary. One is a tendency for the curve to assume a long horizontal line at times more distant from the day of transplantation. This may be misinterpreted as stable graft function or even as a manifestation of transplant tolerance. However, it usually arises because too few patients remain under observation at these later times and none of them have been observed to fail. This pitfall can be avoided by presenting the estimated survival curve with 95% confidence intervals (95% CI) at each time point. The 95% CI can be estimated roughly by calculating the formula: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi p ð1 pþ=n where p is the estimated probability of survival at a particular time and n is the number of transplant recipients still under observation at that time [2]. When two survival curves are compared, the log-rank or Mantel Haenszel test should be used [3]. Guideline D. The most appropriate routine is to make separate K M survival estimates for patient survival, graft survival and death-censored graft survival as defined in Guideline D. Patient survival is a K M estimate of the probability of a transplant recipient being alive at a finite time after transplantation. Graft survival is an estimate of the probability of the transplant functioning at a finite time after transplantation. If the patient dies and has not returned to long-term dialysis, the date of death is assumed to be the date of graft failure. Therefore, failure is graft loss or patient death, whichever comes first. Graft survival illustrates the overall rate of success in terms of graft and patient survival. It assumes that all deaths are associated primarily with the transplantation. This holds, for example, when evaluating variables such as increased risk of infection, malignancy or aggravation of cardiovascular disease. Graft survival, censored for patient death with a functioning graft, estimates the probability of graft loss only. In the event of death with a functioning graft, the date of death is the date of last follow-up. The fatal event is thereby handled as a case lost to follow-up, not as a graft failure. Death-censored graft survival illustrates the rate of success in terms of graft survival. It assumes that all deaths are associated primarily with causes other than the transplantation. Death-censored survival estimates are particularly important when analysing the influence of factors such as recipient age, which may have other natural causes, and therefore is accompanied by an increased death rate in recipients whose transplants are still functioning [4]. It should be noted that the assumption that death with a functioning transplant equates to graft failure leads to a lower estimate of graft survival than deathcensored graft survival and therefore the method used should be clearly specified. It is rarely straightforward to assume that all or none of the deaths are caused by the transplantation per se. Hence, neither estimate gives the complete picture and preferably both should be used in parallel. Guideline E. Several multi-centre kidney transplant registries are maintained by different institutions. The size, completeness and accuracy of the data registered determine their usefulness in providing standards. Table IV.11 gives the numbers of patients receiving their first kidney transplant together with their K M survival estimates. The outcome data are derived from three large databases, Eurotransplant (ET), United Network for Organ Sharing (UNOS) and the Collaborative Transplant Study (CTS). ET collect data from Austria, Belgium, Germany, Luxembourg, The Netherlands and Slovenia. UNOS is a nonvoluntary registry that collects data from centres in the USA who receive government funding for transplant costs. CTS includes both ET and UNOS data but also collects from many other countries worldwide. The era covered by all three registries is from the mid-1980s up to 2000, after the introduction of cyclosporine. (ET also includes a number of transplants dating back to 1967.) Outcome data at 1, 5 and 10 years are divided into patient and graft 61

3 62 SECTION IV: Long-term management of the transplant recipient Table IV.11. Outcome of first renal transplants Outcome measure Living donor zero mismatches for HLA Living donor one or two HLA haplotypes mismatched Cadaver donor all primary transplants ET data: (ns497) (ns3107) (ns47 015) No. of transplanted (73%) 1877 (60%) (76%) patients observed (48%) 968 (31%) (48%) at years (28%) 498 (16%) 9731 (21%) 20 4 (1%) 9 (-1%) 241 (-1%) K M % 95% CI K M % 95% CI K M % 95% CI Patient survival at years Graft survival* at years UNOS data: (ns5077) (ns24 177) (ns87 535) No. of transplanted (88%) (83%) (78%) patients observed (40%) 7122 (21%) (29%) at years (6%) 786 (3%) 2782 (3%) K M % 95% CI K M % 95% CI K M % 95% CI Patient survival at years Graft survival* at years CTS data: (ns6698) (ns37 208) (ns ) No. of transplanted (89%) (84%) (78%) patients observed (51%) (37%) (39%) at years (18%) 3608 (10%) (12%) (4%) 507 (1%) 3149 (2%) K M % 95% CI K M % 95% CI K M % 95% CI Patient survival at years Graft survival* at years *Patient death included as graft failure. survival. No registry has sufficient data to estimate outcome at 20 years. One exception is the UCLA Registry who published 20-year survival estimates based on transplants dating back to 1967 [5]. Outcomes are given for recipients of kidneys from live sibling donors who were mismatched for zero HLA haplotypes (LD-0), recipients of kidneys from live donors mismatched for one or two HLA haplotypes (LD-1u2) and recipients of kidneys from cadaveric donors (CAD). The largest number of patients analysed is in the CTS registry, next is UNOS and then ET. Of interest is the variation between registries in the proportion of the originally registered patients who are included in the subsequent analyses. Although dwindling numbers are partly due to death and graft loss, they are also due to incomplete follow-up, selective reporting or non-reporting. The proportion of CAD recipients included in the analysis (see patients observed in Table IV.11) at 1 year was equally high in all three registries (CTS, 78%; UNOS, 78%; and ET, 76%). However, the proportion of LD- 0 and LD-1u2 recipients included at 1 year was lower

4 SECTION IV: Long-term management of the transplant recipient at ET (CTS, 89 and 84%; UNOS 88 and 83%; and ET, 73 and 60% respectively). However, recipients included in the analysis at 10 years were highest at ET (28 16%) and lowest at UNOS (6 3%). These inter-registry differences contribute to inequalities in long-term K M survival estimates. Thus, patient survival for live donor transplants at 10 years was highest in ET and lowest in CTS. Ten-year patient survival was 89, 85 and 84% for the LD-0 recipients and 85, 81 and 78% for LD-1u2 recipients analysed at ET, UNOS and CTS, respectively. Patient survival for CAD recipients at 10 years was 70, 69 and 66% at ET, CTS and UNOS, respectively. Graft survival was more comparable between registries with the one exception of a higher survival at 10 years for the LD-0 group at ET. Thus, graft survival at 10 years was 78, 69 and 68% for LD-0, and 56, 55 and 53% for LD-1u2 at ET, UNOS and CTS, respectively. Finally, graft survival for CAD recipients was concordant at 45, 42 and 41% at CTS, ET and UNOS, respectively. The estimated half-life was first proposed as a way of predicting long-term graft survival in 1977 [2]. The calculation assumes that survival follows an exponential distribution, in other words a constant hazard of graft failure after a certain time after transplantation. Although this assumption is clearly invalid during the first post-transplant year, it appears that graft survival curves plotted on a semi-logarithmic scale assume linearity at 3 years post-transplant [6]. More recently, half-life analyses have been calculated for transplants surviving beyond 1 year [7,8]. Some examples of estimated half-lives from a recent study [9] include the following: HLA identical sibling to sibling transplants, 39 years; one HLA-haplotype sibling to sibling transplants, 16 years; unrelated living donor transplants, 17 years; and cadaver donor transplants, 10 years. Two centres A and B are less likely to share identical survival results than they are to have different survival results. Thus, A may have higher results than B, and we have an apparent centre variation. However, the extent to which the order is attributable to randomness or to truly significant differences needs to be determined before any judgement is made. Every league table purporting to list centres by outcome should carry an explanation of how the limits of significance have been set. How is this done? A simple method described by Gilks et al. [10] used the combined data from all centres to calculate the global survival estimate (S) and its standard error (SE) for different numbers of patients under observation. When S was plotted against the SE, the graphic illustrated in Figure IV.1 was obtained. K M survival estimates for individual centres (S c ) were then superimposed on this plot according to the number of patients under observation. Some centres fell outside the limits set by S"SE, whilst others were within the limits. The S c values falling outside were classified as low or high, and S c values falling within S"SE were considered indeterminate. This illustrates how discrete categories of outcome (low, indeterminate and high) are definable by a non-arbitrary method that takes account of the number of transplants performed at an individual centre. More importantly, this approach illustrates that it is possible to conclude that there is no significant centre variation, despite the fact that centre results are distributed over a spectrum. This occurs when all the S c values are indeterminate and fall within the limits set by S"SE and the rank order of the centres is merely a reflection of randomness. Guideline F. Single-centre analyses have the advantage of being relatively uncomplicated by interpatient variation in therapeutic protocols, but they rarely accumulate sufficient numbers of transplants to estimate reliably anything other than the major variables influencing graft outcome. Univariate analyses produce qualitative or at best semi-quantitative outcome statistics. Consequently, they are less sensitive than analyses of multi-centre databases. Furthermore, since many factors influence transplant survival, there is a need to identify and compare the strength of many risk factors simultaneously. One approach is to apply K M univariate methods to dissect very large multi-centre databases such as those held by UNOS or CTS [9,11]. This is analogous to slicing up a large cake into small segments. However, there remain three major problems with univariate analyses. First, they cannot account for correlations between variables. Secondly, they may falsely identify variables whose influence in reality is attributable to surrogate effects. Thirdly, they cannot measure the proportion of the overall risk associated with each individual factor. For these reasons, multivariate methods were introduced, and the Cox s proportional hazards regression method exemplifies this approach [12,13]. This type of analysis progresses in several stages. (i) A decision is made on a distinct period of postoperative time during which the analysis is to be conducted (e.g days or days post-transplant). (ii) The outcome measure (e.g. graft failure) is selected. (iii) Those variables that are likely to be associated with outcome in a qualitative sense (donor age, recipient age, HLA mismatch, etc.) are identified. This is achieved by examining the K M survival estimates for the variable at several different levels and testing the significance of the difference between them. (iv) The statistical model is built by listing all the variables and their chosen levels and subjecting them to multiple regression analysis, the computational details of which can be found elsewhere. The method makes three major assumptions. First, the risk associated with a particular variable does not change over the period of time studied. Secondly, the effect of one variable is independent of the effect of 63

5 64 SECTION IV: Long-term management of the transplant recipient Fig. IV.1. Actuarial survival and simulated standard errors for survival for each of 29 transplant centres showing the division into low, indeterminate and high survival categories. others. Thirdly, in the analysis of multi-centre databases, the effects do not differ among transplant centres. Results are usually expressed as a ratio of the hazard associated with two levels of each variable. One level is selected as reference and termed the baseline (it does not matter which), and divided into the hazard value for each other level for that factor. This ratio or relative risk (RR) for each level of each variable is accompanied by its 95% CI. If the 95% CIs for a given level span unity, this strata of the data is not significantly different from the baseline, allowing the model to be simplified by reducing the number of levels. This, in turn, reduces the number of degrees of freedom, increases the goodness of fit of the model and increases the power of the analysis. An illustrative example of this analytical approach as applied to a single-centre database is given by Roodnat et al. [14]. In this analysis, graft survival in 722 transplanted patients was modelled successfully with five independent variables that had a significant influence on outcome. An example of how a larger multi-centre database increases the power of multivariate analysis is given by Gjertson and Cecka [15]. In this analysis, graft survival in 8422 transplanted patients was modelled with 26 independent variables, each of which was shown to have a significant influence on graft outcome. Guideline G. A large number of clinical trials on new immunosuppressive drugs have shown improved graft survival of only a few percent at 6 or 12 months, and the differences are rarely statistically significant. Many clinicians draw the conclusion from such results that graft survival is not improved by these novel drugs, despite a significant reduction in acute rejection. This raises the question of the relationship between acute rejection and graft survival and the maximum level of improvement of graft survival that might be expected from reducing acute rejection. In the literature, there are several reports stating that acute rejection is associated with impaired graft survival [16,17]. Let us look at the experience from Malmö University Hospital in Sweden (Figure IV.2). The difference in death-censored graft survival between patients who had at least one episode of acute rejection and those who did not was 12% at 1 year, 17% at 3 years and 20% at 5 years. These levels represent the maximum levels that could be achieved through a reduction in the incidence of acute rejection, since this is a comparison between groups with acute rejection rates of 100 and 0%. However, in reality, the reduction of acute rejection is much less. In a clinical trial of a new drug vs an old drug, the incidence of acute rejection may be reduced from, for example, 50 to 25%. Not all patients with the new drug would then have a change of prognosis that would affect the overall outcome of the group only 25% of them (Figure IV.3). Therefore, differences in graft survival at levels as high as 12% cannot be expected when comparing the effect of the new and the old drug, but rather only approximately a quarter of this difference, i.e. 3%. More specifically, the expected improvement in graft survival may be calculated as a weighted average of the outcome of rejectors and non-rejectors and

6 SECTION IV: Long-term management of the transplant recipient 65 Fig. IV2. Graft survival (censored for patient death with a functioning graft) for renal transplants performed in at the Department of Nephrology and Transplantation in Malmö, Sweden. Patients without acute rejection (ns323) had 12% improved 1-year graft survival compared with patients with one or more acute rejection episodes (ns363). The difference at 3 years was 17%, and at 5 years was 20%. Fig. IV.3. The difference in graft survival at 1 year was 12% when patients with and without acute rejection were compared, i.e. 100% of the patients in the latter group may have a different prognosis. In a clinical trial, rejection rates might be 50% with the old drug and 25% with the new drug. Consequently, 25% of the patients may improve their prognosis. The overall improvement at 1 year may therefore be approximately a quarter of 12%, i.e. 3%. the proportion of rejection and non-rejection (Figure IV.4). The proportions of graft survival in rejectors and non-rejectors should preferably be those from the same centre or centres, and based on the same inclusion criteria as in the trial to be evaluated. Here, let us use the overall Malmö experience of 1-year (death-censored) graft survival in rejectors (being 82.0%) and in non-rejecters (93.8%) (CI"2.0 and "1.4%, respectively). What levels of improvement in graft survival could be expected through reduction of acute rejection, for example, in the phase III trials with MMF or daclizumab (DZB)? In the European MMF trial [18], biopsy-proven acute rejection rates in the MMF Ess 1 s 0 where s 1 sq 1 3p 1 q(1 q 1 ) 3p 0 s 0 sq 0 3p 1 q(1 q 0 ) 3p 0 and s 1 sgraft survival with the new drug s 0 sgraft survival with the old drug q 1 sacute rejection with the new drug q 0 sacute rejection with the old drug p 1 sgraft survival with acute rejection p 0 sgraft survival without acute rejection. Fig. IV.4. Estimated difference (E ) in graft survival in treatment groups of a new vs old drug.

7 66 SECTION IV: Long-term management of the transplant recipient Table IV.12. Expected levels of improvement in graft survival and actual results of clinical trials Acute rejection (%) Improvement in graft survival censored for death with functioning graft (%) New drug (MMF or DZB) Old drug (control group) At 1 year At 3 years Expected* Trial Expected* Trial Pooled MMF trials [20] q European MMF trial [19] q Pooled DZB trial [22,23] y 4.8y 2.4y 4.0y MMF, mycophenolate mofetil; DZB, daclizumab. *Calculated from the formula in Figure IV.4. qstatistically significant improvement in the trial; P ygraft survival including patient death as graft failure. Table IV.13. Sample size and power to verify true differences in graft survival of 4 or 5% Graft survival in the two treatment groups Difference in Sample size Power at sample graft survival at 80% power size of 150 patients 86% 90% 4% % 82% 86% 4% % 78% 82% 4% % 85% 90% 5% % 80% 85% 5% % 75% 80% 5% % Sample size is the number of patients in each treatment arm. Power is the chance of verifying a true difference with statistical significance (P-0.05). (2 guday) and control groups were 17 and 46.4%, respectively. The expected difference in graft survival at 1 year would therefore be 3.5%, whereas in the trial the result was 3.6% [19]. This illustrates that the results obtained were those that we could have expected (Table IV.12). The results of the pooled MMF trials at 1 year [20] and the European MMF trial at 3 years [19] in terms of censored graft survival were superior to estimated expectations (4.0% vs 2.1%, and 7.3% vs 5.0%, respectively). This suggests that there was an effect of the drug independent of the impact on acute rejection. This notion was supported further by analyses of data from the US Renal Transplant Scientific Registry [21]. It should be noted that such analyses may be performed for non-censored graft survival, deathcensored graft survival and patient survival. For example, non-censored graft survival was 91.4 and 86.6% for DZB and placebo, respectively, in the pooled analysis of the phase III multi-centre trials [22,23]. This difference of 4.8% was not statistically significant (Ps0.065). Using the Malmö experience of noncensored 1-year graft survival in rejectors (79.2%; CI"1.8%) and in non-rejectors (89.3%; CI"1.7%), the expected difference is estimated to be only 1.6%. The result of the trial therefore exceeded expectations and the question arises of whether the trial was powered to verify differences at these levels. Guideline H. Although several trials have shown differences in graft survival at similar or higher levels than would have been expected, the results were not statistically significant. It is, however, premature to draw the conclusion that there is no true difference or improvement in graft survival achieved by using the new drugs. The pivotal question is whether or not the trial was statistically powered to verify a true difference at the expected level. If the expected level is 4%, it is necessary to include patients in each treatment arm to reach a power of 80% (i.e. to have an 80% chance of verifying a true difference and, conversely, a 20% risk of failing to verify the true difference). The number of patients necessary varies depending on the overall proportion of treatment failures (graft losses) in the two groups (Table IV.13). In the European MMF trial, there were 160 patients in each treatment arm. With an expected difference in graft survival of 3.5%, the power of the study to verify such a difference was 14%. Consequently, the risk of not verifying a true difference at this level was 86%. By pooling data at 1 year from three MMF trials, the power was increased and the difference of 4.0% was statistically significant (Table IV.13). At 3 years, however, the difference in death-censored graft survival was larger (7.3%) and statistically significant, without the need to pool data to achieve an increased power.

8 SECTION IV: Long-term management of the transplant recipient Since huge trials are necessary to verify small differences, it is not to be expected that improvements at these levels will be verified statistically. There is also the question of whether or not a small improvement, such as 3 5% at 1 year, is clinically relevant. However, a reduction in graft loss of a few percent each year has a great impact on the demands for dialysis and re-transplantation. Furthermore, with additional strategies for reducing risk factors other than acute rejection, such as reperfusion damage, chronic nephrotoxicity, hypertension and hyperlipidaemia, it is not unlikely that further improvements in graft function and graft survival may be achieved. The verification of these improvements will most probably be just as complex to verify as with reduction of acute rejection, and large clinical trials with sufficient statistical power will be required. Acknowledgement. The EPBG experts are grateful to Guido Persijn and the Research Department of the Eurotransplant International Foundation who provided data from ET, Gerhard Opelz who provided data from CTS, and Steve Takemoto, J. Michael Cecka and David Gjertson who provided data on behalf of UNOS. References 1. Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Statist Assoc 1958; 53: Peto R, Pike M, Armitage P et al. Design and analysis of randomised clinical trials requiring prolonged observation of each patient. II Analysis and examples. Br J Cancer 1977; 35: Parmar M, Machin D. Survival Analysis: A Practical Approach. John Wiley, Chichester, Tesi RJ, Elkhammas EA, Davies EA, Henry ML, Ferguson RM. Renal transplantation in older people. Lancet 1994; 343: Terasaki PI, Cecka JM, Gjertson DW, Takemoto S, Cho YW, Yuge J. Risk rate and long-term kidney transplant survival. Clinical Transplants 1996: Opelz G, Mickey MR, Terasaki PI. Influence of race on kidney transplant survival. Transplant Proc 1997; 9: Takiff H, Cook D, Himaya N, Mickey M, Terasaki P. Dominant effect of histocompatibility on ten-year kidney transplant survival. Transplantation 1988; 45: Gjertson DW. Look-up survival tables for renal transplantation. Clin Transplants 1997: Cecka JM. The UNOS Scientific Renal Transplant Registry. Clin Transplants 1999: Gilks WR, Selwood N, Bradley BA. The variation among transplant center results in the United Kingdom and Ireland from 1977 to Transplantation 1984; 38: Opelz G. Factors influencing long-term graft loss. The Collaborative Transplant Study. Transplant Proc 2000; 32: Gore SM. Graft survival after renal transplantation: agenda for analysis. Kidney Int 1983; 24: Gilks WR, Gore SM, Bradley BA. Analyzing transplant survival data. Transplantation 1986; 42: Roodnat JI, Mulder PG, Rischen Vos J et al. Proteinuria after renal transplantation affects not only graft survival but also patient survival. Transplantation 2001; 72: Gjertson DW, Cecka JM. Determinants of long-term survival of pediatric kidney grafts reported to the United Network for Organ Sharing kidney transplant registry. Pediatr Transplant 2001; 5: Matas AJ, Humar A, Payne WD, Gillingham KJ, Dunn DL, Sutherland DE, Najarian JS. Decreased acute rejection in kidney transplant recipients is associated with decreased chronic rejection. Ann Surg 1999; 230: Humar A, Hassoun A, Kandaswamy R, Payne WD, Sutherland DE, Matas AJ. Immunologic factors: the major risk for decreased long-term renal allograft survival. Transplantation 1999; 68: European Mycophenolate Mofetil Cooperative Study Group. Placebo-controlled study of mycophenolate mofetil combined with cyclosporin and steroids for prevention of acute rejection. Lancet 1995; 345: European Mycophenoalte Mofetil Study Group. Mycophenolate mofetil in renal transplantation: 3-year results from the placebocontrolled trial. Transplantation 1999; 68: Halloran P, Mathew T, Tomlanovich S, Groth C, Hooftman L, Barker C. Mycophenolate mofetil in renal allograft recipients: a pooled efficacy analysis of three randomized, double-blind, clinical studies in prevention of rejection. The International Mycophenolate Mofetil Renal Transplant Study Groups. Transplantation 1997; 63: Ojo AO, Meier-Kriesche HU, Hanson JA et al. Mycophenolate mofetil reduces late renal allograft loss independent of acute rejection. Transplantation 2001; 69: Ekberg H, Bäckman L, Tufveson G, Tydén G, Nashan B, Vincenti F. Daclizumab prevents acute rejection and improves patient survival posttransplantation: one year pooled analysis. Transplant Int 2000; 13: Vincenti F, Nashan B, Bumgardner G, Hardie I, Pescovitz M, Johnson RWG. Three year outcome of the phase III clinical trials with daclizumab. Transplantation 2001; 69: S261 67