Random allocation

The method of random allocation,^71,86,128 including the method of sequence generation, was clearly stated and adequate in only two induction studies^71,128 and 18 maintenance studies,^{86,103,110,112,119,122,124,126,127,129–136,150} whereas 65 studies (nine induction studies^{72–74,87,95,98,137} and 54 maintenance studies^{51,58,59,75–85,88,89,91–94,99–102,104–109,111,113–118,120,121,125,138–147,152–155}) and both of the studies of induction and maintenance treatment^123,148 did not clearly specify the method used. The remaining maintenance study¹⁴⁹ used a minimisation technique that included a random element.

Concealment of allocation

The method of concealment of allocation was clearly reported in 12 trials (two induction studies,^97,128 nine maintenance studies,^{58,114,129,130,133,140,147,150,152} and one study¹³⁶ of both induction and maintenance treatment). Fifty-four trials^{51,72–74,76–79,81–85,87–89,91–93,95,96,98–100,102–106,108–113,115–120,124,127,131,134,135,139,141,143–145,153–155} did not report any information on allocation concealment, whereas 20 trials^{71,75,80,86,94,101,107,121–123,125,126,132,136–138,142,146,149,156} provided some information pertaining to allocation concealment but lacked sufficient detail or clarity to demonstrate that allocation was adequately concealed.

Baseline characteristics

Fifty-seven trials (48 maintenance studies,^{51,58,77,80–82,84,86,94,99,100,102,104–109,113–117,119–121,124–127,131,132,134,138,139,141–147,149,150,152,154,156,157} eight induction studies^{71,72,74,87,97,128,137} and one study¹²³ for induction and maintenance) fully reported baseline characteristics. Nine trials (eight maintenance studies^{88,89,92,110–112,122,148} and one study¹⁴⁸ of both induction and maintenance) reported significant baseline between-group differences for key factors, including PRA grade, number of previous transplants, patient age, pretransplant diabetes mellitus, HLA mismatches and ECD donor kidneys. A further six maintenance studies^{91,101,130,133,140,155} were rated as ‘partial’ because they reported a baseline difference in patient sex.

The remaining trials (four induction studies,^71,95,96,98 26 maintenance studies^{59,75–80,83–85,93,94,103,107,114,115,118,126,127,129,131,132,142,150,152,153} and one study¹²³ of both induction and maintenance) did not provide sufficient information for a judgement to be made about baseline similarity of groups, either by omitting to report sufficient statistical information, by reporting on a very limited range of patient baseline characteristics or by not reporting any patient baseline characteristics.

Treatment allocation masked from participants

Five induction studies,^{87,96,98,128,137} 47 maintenance studies^{51,59,76,78–80,82–84,86,88,92–94,103,105–108,111,113,116,118,125,126,129–135,138–142,144–149,151–153,155} and both of the studies of induction and maintenance treatment^123,148 did not blind participants to treatment allocation.

Only two maintenance studies^89,124 and four induction studies^71–74 made clear that the participants were blinded to treatment allocation. A further four maintenance studies^{58,77,143,150} were rated as ‘partial’ because it was reported that participants were blinded for a limited period of time only (until 24 weeks for one study⁵⁸ and until 12 months for the other three studies.^77,143,150

One further induction study⁹⁵ was rated as ‘unclear’ because, despite being PBO controlled, no further details were reported about blinding. The remaining trials (one induction study⁹⁷ and 20 maintenance studies^{75,81,85,91,99–102,104,109,110,112,114,115,117,119–122,127}) did not report any information about blinding participants to treatment allocation.

Treatment allocation masked from clinicians

All of the trials that did not blind participants from treatment allocation also failed to mask treatment allocation from clinicians.^{51,59,76,78–80,82–84,86–88,92–94,96,98,103,105–108,111,113,116,118,123,125,126,128–135,137–142,144–149,151–153,155} An additional induction study⁹⁷ also stated that treatment allocation was not masked from clinicians (participant blinding was not reported). Similarly, the four induction studies^71–74 and two maintenance studies^89,124 that reported blinding participants to treatment allocation also masked treatment allocation from clinicians. Again, four maintenance studies^{58,77,143,150} were rated as ‘partial’ for clinician blinding because blinding occurred for only a limited time, and one induction study⁹⁵ was rated as ‘unclear’ because, although it was a PBO-controlled trial, no further details were reported about blinding. The other 20 maintenance studies^{75,81,85,91,99–102,104,109,110,112,114,115,117,119–122,127} did not report any details about clinician blinding.

Treatment allocation masked from outcome assessors

The majority of trials (52 maintenance studies,^{51,75–77,79–84,86,89,91–94,99–102,104–106,108,109,111–114,116–122,130,131,133,138,140,141,144–149,151–153,155} nine induction studies,^{71–73,87,95–98,128} and both of the studies^123,148 of induction and maintenance treatment) did not report whether outcome assessors were blind to treatment allocation.

One induction study¹³⁷ and five maintenance studies^{78,132,134,135,139} made it clear that the outcome assessors were not blinded to treatment allocation. For fifteen trials^{58,59,74,85,88,103,107,110,115,124–127,129,142} (one induction study⁷⁴ and 14 maintenance studies^{58,59,85,88,103,107,110,115,124–127,129,142}) it was clear that outcome assessors were blinded for at least one outcome, and a further two maintenance studies^143,150 were given a ‘partial’ rating because the outcome assessors were blinded for the first 12 months of the study.

Reporting of all a priori outcomes

All trials were rated as ‘unclear’ with regard to reporting of a priori outcomes.^{51,58,59,71–89,91–135,137–153,155} This was because the trial reports failed to explicitly state whether or not all outcomes defined in the study protocol were reported.

Reporting of loss to follow-up, withdrawals and dropouts

Fifty-four trials adequately reported loss to follow-up, withdrawals and dropouts (by providing numbers and reasons by treatment group). Of these, 45 were maintenance studies,^{51,58,59,80,81,83,84,88,102,104,106–108,111–114,116,118–120,124–127,130–135,138,139,141,142,144–152,155} eight were induction studies,^{71–74,87,98,128,137} and one¹⁴⁸ was a study of both induction and maintenance treatment. In 22 trials (20 maintenance studies^{76,85,86,91–94,99–101,103,105,109,110,115,121,122,129,140,143} and two induction studies^95,96), the reporting of loss to follow-up, withdrawals and dropouts was inadequate, with key information omitted. A further four trials^75,79,97,123 (one induction study,⁹⁷ two maintenance studies^75,79 and one study of both induction and maintenance treatment¹²³) were rated as ‘unclear’. For the study of both induction and maintenance, this was because, despite all of the relevant information being provided, the numbers did not appear to tally. For the other three trials,^75,79,97 this was because of the fact that all participants appeared to complete the study but this was not explicitly stated. For the remaining six maintenance studies,^{77,78,82,89,117,153} information regarding loss to follow-up, withdrawals and dropouts was not reported.

Intention-to-treat analysis

Primarily, a strict definition of ITT was used (all randomised and transplanted participants). According to this definition, 48 trials (seven induction studies^{71–74,87,98,137} and 41 maintenance studies^{51,59,77,79,80,84,86,88,89,100–102,104,106–108,110,113,115,117,120,121,124–127,129–131,134,135,139,141–143,146,149–153}) were rated as adequately performing an ITT analysis, with 19 trials (three induction studies,^128,158,159 14 maintenance studies,^{58,83,91,114,119,132,133,138,140,144145,147,148,155} and both studies^123,148 of induction and maintenance treatment) not performing an adequate ITT analysis. In 16 cases (two induction studies^96,97 and 14 maintenance studies^{75,76,81,82,92–94,99,103,105,109,111,112,116}) there was a lack of clarity regarding whether or not an ITT analysis had been conducted. The other five trials (one induction⁹⁵ and four maintenance studies^{78,85,118,122}) did not report any relevant information regarding whether or not an ITT analysis had been conducted.

A secondary definition of ITT analysis was also used (all randomised and transplanted participants or < 10% excluded). When this definition was applied, 13 of the trials previously rated as inadequate were instead rated as adequate (11 maintenance studies^{58,83,114,119,132,133,138,140,147,148,155} and both of the studies^123,148 of induction and maintenance treatment). Thus, only four trials^{91,128,144,145} did not perform an adequate ITT analysis. The number of trials rated as ‘unclear’ or ‘not reported’ did not change when this definition of ITT was used.

Applicability to the current NHS in England

Only 11 trials (one induction study,⁷⁴ nine maintenance studies^{51,58,86,114,124,125,132,133,155} and one study¹²³ of both induction and maintenance) were adequately applicable to the current NHS in England. The majority of trials (seven induction studies,^{71,87,95,97,98,128,137} 41 maintenance studies,^{59,75,77–82,84,85,88,89,93,94,99,101,109,112,115–118,120,129–131,134,135,138,139,141,142,144–152} and one study¹⁴⁸ of both induction and maintenance) were limited in some way with regard to applicability to the current NHS in England. In all except one of these trials this was primarily as a result of the fact that patients, donors or organ characteristics were not representative of the current NHS in England (e.g. > 90% deceased donors or ‘suboptimal transplants’ or ‘high risk of rejection population’). In the other trial¹³⁵ this was primarily owing to a lack of statistical power.

The remaining three induction studies^72,73,96 and 23 maintenance studies^{76,83,91,92,100,102–108,110,111,113,119,121,122,126,127,140,143,153} were rated as ‘unclear’ regarding applicability to the current NHS in England. The primary reason for this was as follows: the study lacked clarity regarding key demographic or patient–donor characteristics (two induction studies^73,96 and 10 maintenance studies^{76,83,91,92,102–104,107,113,140}); the study was based on a non-EU population (two induction studies^72,159 and 13 maintenance studies^{100,105,106,108,110,111,119,121,122,126,127,143,153}).

Mortality

Participant mortality was recorded at 6 months by three studies.^73,74,123 Six studies^{71–74,97,128} report mortality at 1 year.

As displayed in Table 13 and Figure 8, the OR at 0.5 years for the studies by Ponticelli et al.,⁷³ Albano et al.¹²³ and Lawen et al.⁷⁴ indicates that BAS is associated with lower odds of mortality, although the results are not statistically significant (OR 0.36, 95% CI 0.13 to 1.01).

TABLE 13

Mortality for BAS vs. PBO/no induction

FIGURE 8

Forest plot: mortality for BAS vs. PBO/no induction.

Pooled results at 1 year for the studies by Lawen et al.⁷⁴ and Sheashaa et al.⁹⁷ also display no statistically significant difference between BAS and PBO/no induction up to 1 year, which is in agreement with the previous HTA⁶⁵ (OR 0.95, 95% CI 0.49 to 1.87). The effect estimate for the Sheashaa et al.⁹⁷ study at 3, 5, 7 and 10 years also shows no difference between arms.

Graft loss

Of the seven studies in this group,^{71–74,97,123,128} three studies^73,74,123 recorded graft loss at 6 months and six studies^{71–74,97,128} at 1 year (Table 14 and Figure 9).

TABLE 14

Graft loss for BAS vs. PBO

FIGURE 9

Forest plot: graft loss for BAS vs. PBO/no induction.

At both time points the OR may indicate some benefit of BAS compared with PBO or no induction in reducing graft loss (0.5 years: OR 0.78, 95% CI 0.50 to 1.22; 1 year: OR 0.82, 95% CI 0.56 to 1.21). However, this estimate must be treated with caution because of the wide CIs indicating a lack of statistical significance.

The one study⁹⁷ reporting results at 3, 5, 7 and 10 years showed no statistically significant difference between arms (see Table 14).

Graft function

Pooled analysis for GRF measured as CRC (Table 15 and Figure 10) implies no beneficial effect of BAS compared with PBO [0.5 years: weighted mean difference (WMD) –1.38 ml/minute/1.73 m², 95% CI –5.96 to 3.20 ml/minute/1.73 m²; 1 year: WMD 1.93 ml/minute/1.73 m², 95% CI –0.97 to 4.83 ml/minute/1.73 m²].^{71–73,97,123} In particular, results for 0.5 years must be treated with caution because of the substantial heterogeneity across studies (I² = 83.4%). It should also be noted that, at 1 year, the study reported by Kahan et al.,⁷² which indicates an improved GRF for participants on BAS, had a higher percentage of African American participants (34% and 27%) who generally exhibit poor long-term graft survival compared with other ethnic groups.⁷²

TABLE 15

Pooled analysis for BAS vs. PBO/no induction: GRF

FIGURE 10

Forest plot: GRF for BAS vs. PBO/no induction. SD, standard deviation; WMD, weighted mean difference.

Data up to 10 years reported by Sheashaa et al.⁹⁷ (Table 16) indicate no statistically significant difference between BAS and no induction.

TABLE 16

Graft function for BAS vs. no induction (unpooled)

Biopsy-proven acute rejection

The results of BPAR at 0.5 years are inconclusive because of the substantial heterogeneity across studies (I² = 80.7%).^71,73,74,123 In contrast, at 1 year, BAS statistically significantly reduced BPAR compared with PBO/no induction (OR 0.53, 95% CI 0.40 to 0.70, I² = 0.0%) (Table 17 and Figure 11).^{72–74,97,128} Furthermore, the report by Sheashaa et al.⁹⁷ indicates this effect is maintained up to 10 years (OR 0.41, 95% CI 0.18 to 0.96).⁹⁷

TABLE 17

Pooled analysis for BAS vs. PBO: BPAR

FIGURE 11

Forest plot: BPAR for BAS vs. PBO.

Severity of biopsy-proven acute rejection

Six studies^{71–74,97,123} report severity of BPAR (Table 18). Overall, Table 18 indicates that BAS may be associated with less severe exacerbations of BPAR.

TABLE 18

Severity of BPAR for BAS vs. PBO

Time to biopsy-proven acute rejection

Only one study¹²⁸ reported time to BPAR (Table 19). In general, the results seem similar between arms, although no induction has a broader range (BAS 35–267 days, no induction 10–364 days).

TABLE 19

Time to BPAR for BAS vs. no induction

Summary of results for BAS compared with PBO/no induction

Pooled results indicate no statistically significant difference between BAS and PBO/no induction for mortality up to 1 year (six studies^{71–74,97,128}) (OR 0.95, 95% CI 0.49 to 1.87).

The effect estimate for the Sheashaa et al.⁹⁷ study at 3, 5, 7 and 10 years also shows no difference between arms.⁹⁷

No statistically significant difference is found between BAS and PBO/no induction for graft loss (six studies^{71–74,97,128}) (0.5 years OR 0.78, 95% CI 0.50 to 1.22; 1 year OR 0.82, 95% CI 0.56 to 1.21). This is also the case for the single study when follow-up continues up to 10 years.⁹⁷

Pooled analysis for GRF measured as CRC implies no beneficial effect of BAS compared with PBO (0.5 years, WMD –1.38 ml/minute/1.73 m², 95% CI –5.96 to 3.20 ml/minute/1.73 m²; 1 year, 1.93, 95% CI –0.97 to 4.83 ml/minute/1.73 m²).^{71–73,97,123}

The results of BPAR at 0.5 years are inconclusive because of the substantial heterogeneity across studies^71,73,74,123 (I² = 80.7%). In contrast, at 1 year, BAS statistically significantly reduced BPAR compared with PBO/no induction (OR 0.53, 95% CI 0.40 to 0.70; I² = 0.0%).^{72–74,97,128} Furthermore, the report by Sheashaa et al.⁹⁷ indicates that this effect is maintained up to 10 years (OR 0.41, 95% CI 0.18 to 0.96).⁹⁷ In general, severity of BPAR appeared reduced with BAS.

Mortality

Two trials^96,148 provided data on mortality for rATG vs. no induction (Table 20). Follow-up data are provided to only 1 year.⁹⁶ No clear evidence of a difference between arms is visible, as the OR is close to ‘1’ and the CIs are wide.

TABLE 20

Mortality for rATG vs. no induction

Graft loss

Two trials^96,148 provide graft loss data for rATG vs. no induction (Table 21). For both studies,^96,148 CIs are extremely wide, crossing an OR of 1, indicating no statistical difference between arms.

TABLE 21

Graft loss for rATG vs. no induction

Graft function

No studies reported GRF.

Biopsy-proven acute rejection

Two studies^96,148 report on BPAR for rATG vs. no induction for 0.5 years and 1 year (Table 22). The data at 1 year suggest a statistically significant beneficial effect for rATG (OR 0.41, 95% CI 0.24 to 0.52).⁹⁶

TABLE 22

Biopsy-proven acute rejection for rATG vs. no induction

Severity of biopsy-proven acute rejection

One study¹⁴⁸ reports severity of BPAR at 0.5 years (Table 23). For people identified with BPAR, the occurrence of the most severe classification was 10.7% for rATG and 6.4% for no induction. For Banff classification II, there is a greater association with no induction (rATG 25%, no induction 36.2%).

TABLE 23

Biopsy-proven acute rejection for rATG vs. no induction

Time to biopsy-proven acute rejection

Time to BPAR is reported by one study⁹⁶ (Table 24), in which more participants experience BPAR at 7–10 days with no induction than with rATG.

TABLE 24

Time to BPAR for rATG vs. no induction

Summary of results for rATG vs. no induction

Only two studies^96,148 report rATG vs. no induction. No statistically significant difference was seen for mortality, graft loss or GRF. For BPAR, the data at 1 year suggest a statistically significant beneficial effect for rATG (OR 0.41, 95% CI 0.24 to 0.52) and for severity of BPAR; at Banff classification II, there are greater odds of association with no induction (1 year: OR 0.09, 95% CI 0.01 to 0.73).

Mortality

The comparison between BAS and rATG for mortality is reported by three studies^87,98,137 (Table 25 and Figure 12). Two studies are pooled with 1-year results where no statistically significant effect is seen between arms (OR 1.03, 95% CI 0.35 to 3.00).^98,137

TABLE 25

Mortality for BAS vs. rATG

FIGURE 12

Forest plot: mortality for BAS vs. rATG.

Graft loss

Data from three trials^87,98,137 were pooled at the 1-year time point (Table 26 and Figure 13). Although the OR indicates lower odds of graft loss associated with rATG, the effect is not statistically significant (OR 1.36, 95% CI 0.61 to 3.03). There was no evidence of heterogeneity across studies. For the individual study⁸⁷ at 0.5 years there was no statistically significant effect for BAS or rATG.

TABLE 26

Graft loss for BAS vs. rATG

FIGURE 13

Forest plot: graft loss for BAS vs. rATG.

Graft function

Only Lebranchu et al.⁸⁷ report GRF, with results at 0.5 years and 1 year (Table 27). The MD for CRC of 6.10 ml/minute/1.73 m² at 1 year in favour of BAS is not statistically significant (p = 0.1103).

TABLE 27

Graft function for BAS vs. rATG

Biopsy-proven acute rejection

A total of three studies^87,98,137 report on BPAR for BAS vs. rATG (Table 28 and Figure 14). At both 0.5 years and 1 year, the 95% CIs imply a lack of statistically significant difference between treatments (0.5 years, OR 1.00, 95% CI 0.24 to 4.24; 1 year, OR 1.57, 95% CI 0.95 to 2.61). For Brennan et al.,¹³⁷ as a much larger study with narrower CIs, rATG appears to reduce BPAR, although this effect is lost when pooled with the smaller studies.

TABLE 28

Biopsy-proven acute rejection for BAS vs. rATG

FIGURE 14

Forest plot: BPAR for BAS vs. rATG.

Severity of biopsy-proven acute rejection

Two studies^87,98 report on severity of BPAR, although results are not provided for all Banff classifications (Table 29). No difference is seen between treatments.

TABLE 29

Severity of BPAR for BAS vs. rATG

Time to biopsy-proven acute rejection

Time to BPAR is reported by two studies^87,98 (Table 30). Neither of the studies^87,98 revealed a statistically significant difference between BAS and rATG, despite the study by Mourad et al.⁹⁸ reporting a mean time for BAS of 155 days (SD 196.27 days) and for rATG of 35 days (SD 30.19 days).

TABLE 30

Time to BPAR for BAS vs. no rATG

Summary of results for BAS vs. rATG

Three RCTs were identified.^87,98,137 No statistically significant difference was seen for any of the outcomes.

Mortality

Ten studies^{76,79,80,83,84,88,99,100,104,148} report mortality, with meta-analysis possible at the 0.5- and 1-year time points (Table 31 and Figure 15). All studies^{76,79,80,83,84,88,99,100,104,148} are presented graphically on the forest plot to provide a visual overview (see Figure 15). At 0.5 years, pooled results of only two studies^{84,148,164,165} generate an OR of 0.54 (95% CI 0.18 to 1.62), indicating lower odds of mortality for TAC; however, the large CIs indicate a low level of precision, and, as they all overlap, the null value (OR = 1) there is unlikely to be a significant difference between treatments. Although the OR at 1 year, which includes eight studies,^{76,80,83,84,88,99,100,104} has shifted to 1.51, indicating reduced odds of mortality in the CSA arm, the 95% CI of 0.75 to 3.06 also suggests no significant difference between treatments. Heterogeneity across studies for the 1-year time point is low and may not be important at this level according to the Cochrane Handbook²⁰¹ (I² = 14.8%). Mayer et al.⁸⁸ report mortality up to 5 years; however, the results are consistent with earlier time points and indicate no difference between arms (OR 1.20, 95% CI 0.69 to 2.07).

TABLE 31

Mortality for TAC + AZA vs. CSA + AZA

FIGURE 15

Forest plot: mortality for TAC + AZA vs. CSA + AZA.

Graft loss check

Graft loss is reported for 10 trials^{76,79,80,83,84,88,99,100,104,148} (Table 32 and Figure 16). Results were pooled for the 0.5-, 1- and 2-year time points. The pooling of trials reported by Margreiter et al.⁸⁴ and Charpentier et al.¹⁴⁸ at 0.5 years gives an OR of 0.45 (95% CI 0.24 to 0.84), which is statistically significant in favour of TAC.^84,148 The 1-year time point is more reliable, at which seven studies are pooled (see Table 32), generating an OR of 1.18 (95% CI 0.72 to 1.93). However, as with mortality, the results for graft loss suggest no difference between TAC and CSA. This lack of statistical significance for either treatment remains at 5 years (OR 0.92, 95% CI 0.61 to 1.40).

TABLE 32

Graft loss for TAC + AZA vs. CSA + AZA

FIGURE 16

Forest plot: graft loss for TAC + AZA vs. CSA + AZA.

Graft function

Graft function was measured and reported by four studies,^75,76,79,84 with effects measured from 0.08 to 3 years. No meta-analysis is provided for GRF, as the results are presented in a number of ways and are not appropriate for pooling. In general, Table 33 shows some variation between arms with large SDs; for example, results presented by Margreiter et al.⁸⁴ at 1 year imply an improved GRF for TAC as opposed to CSA [68.9 (SD 23.2) ml/minute/1.73 m² and 61.8 (SD 23.2) ml/minute/1.73 m², respectively], which is in contrast with the study of Van Duijnhoven et al.,⁷⁵ who report 60.2 ml/minute/1.73 m² (range 11.5–86.2 ml/minute/1.73 m²) and 64.9 ml/minute/1.73 m² (range 29.5–84.5 ml/minute/1.73 m²), respectively. This conflict between studies is seen at all time points.

TABLE 33

Graft function for TAC + AZA vs. CSA + AZA

Biopsy-proven acute rejection

All time points from 0.08 to 4 years reveal ORs of < 1 for BPAR, indicating that TAC is more effective than CSA in reducing this outcome (Table 34 and Figure 17).^{76,79–84,88,99,100,104,148} BPAR outcomes were reported by nine studies^{76,81–84,88,99,100,104} at 1 year, where pooled analysis gives an OR of 0.50 and 95% CI 0.39 to 0.64. Minimal heterogeneity is indicated across the studies at year 1 (I² = 8.1%). Mayer et al.⁸⁸ report BPAR at 4 years, when the beneficial effect of TAC appears to be maintained (OR 0.38, 95% CI 0.25 to 0.57).

TABLE 34

Biopsy-proven acute rejection for TAC + AZA vs. CSA + AZA

FIGURE 17

Forest plot: BPAR for TAC + AZA vs. CSA + AZA. Note: only the low-dose TAC arm is used for Laskow et al., as this is closest to the dose used in practice.

Severity of biopsy-proven acute rejection

Four trials^{82,84,100,148} report on severity of BPAR from 6 months to 2 years (Table 35). For the studies by Baboolal et al.⁸² and Hardinger et al.,¹⁰⁰ at 1 year, no participants with BPAR experienced Banff grade III for either arm.^82,100 At 6 months, Charpentier et al.¹⁴⁸ report the proportion of people with BPAR classified as Banff III as 10.7% for TAC and 15.4% for CSA and by 2 years Margreiter et al.⁸⁴ report 6.4% and 16.8% of people with BPAR experiencing Banff III, for TAC and CSA, respectively.

TABLE 35

Severity of BPAR at 6 months for TAC + AZA vs. CSA + AZA

Time to biopsy-proven acute rejection

Time to first BPAR is reported by only two studies,^82,83 with contrasting results (Table 36). The results reported by Baboolal et al.⁸² indicate that BPAR is achieved more quickly for participants receiving TAC (35 days, SD 13 days) rather than CSA (59 days, SD 38 days).

TABLE 36

Time to BPAR for TAC + AZA vs. CSA + AZA

Summary of results for TAC + AZA vs. CSA + AZA

• Ten studies^{76,79,80,83,84,88,99,100,104,148} report mortality, with meta-analysis possible at the 0.5- and 1-year time points. At 0.5 years, pooled results of only two studies^84,148 generates an OR of 0.54, 95% CI 0.18 to 1.62, indicating lower odds of mortality for TAC; however, the large CIs overlap the null value (OR = 1) therefore there is unlikely to be a significant difference between treatments. Although the OR at 1 year, which includes eight studies,^{76,80,83,84,88,99,100,104} has shifted to 1.51, indicating reduced odds of mortality in the CSA arm, the 95% CI of 0.75 to 3.06 also suggest no significant difference between treatments. Heterogeneity across studies for the 1-year time point is low and may not be important at this level according to the Cochrane Handbook²⁰¹ (I² = 14.8%).
• Graft loss is reported for 10 trials.^{76,79,80,83,84,88,99,100,104,148} Results were pooled for the 0.5-, 1- and 2-year time points. The pooling of trials reported by Margreiter et al.⁸⁴ and Charpentier et al.¹⁴⁸ at 0.5 years give an OR of 0.45 (95% CI 0.24 to 0.84), which is statistically significant in favour of TAC.^84,148 The 1-year time point, where seven studies^{76,83,84,88,99,100,104} are pooled, generates an OR of 1.18 and a 95% CI 0.72 to 1.93, which is not statistically significant, and this remains the case at 5 years.
• Graft function was measured and reported by four studies,^75,80,84,98 with effects measured from 0.08 to 3 years. No meta-analysis is possible, as the results are presented in a number of ways and are not appropriate for pooling. In general, there is some variation between arms with large SDs, for example results presented by Margreiter et al.⁸⁴ at 1 year imply an improved GRF for TAC as opposed to CSA [68.9 ml/minute/1.73 m² (SD 23.2 ml/minute/1.73 m²) and 61.8 ml/minute/1.73 m² (SD 23.2 ml/minute/1.73 m²), respectively], which is in contrast with Van Duijnhoven et al. 2002, who report 60.2 ml/minute/1.73 m² (range 11.5–86.2 ml/minute/1.73 m²) and 64.9 ml/minute/1.73 m² (range 29.5–84.5 ml/minute/1.73 m²), respectively. This conflict between studies is seen at all time points.
• All time points from 0.08 to 4 years reveal ORs of < 1 for BPAR, indicating that TAC is more effective than CSA in reducing this outcome. BPAR outcomes were reported by nine studies^{76,81–84,88,99,100,104} at 1 year, where pooled analysis gives an OR of 0.50 and a 95% CI of 0.39 to 0.64. Low heterogeneity is indicated across the studies at year 1 (I² = 8.1%). Mayer et al.⁸⁸ report BPAR at 4 years, where the beneficial effect of TAC appears to be maintained (OR 0.38, 95% CI 0.25 to 0.57).
• Four trials^{82,84,100,148} report on severity of BPAR from 6 months to 2 years. For the studies by Baboolal et al.⁸² and Hardinger et al.,¹⁰⁰ at 1 year, no participants with BPAR experienced Banff grade III for either arm.^82,100 At 6 months, Charpentier et al.¹⁴⁸ report the proportion of people with BPAR classified as Banff III as 10.7% for TAC and 15.4% for CSA and, by 2 years, Margreiter et al.⁸⁴ report 6.4% and 16.8% of people with BPAR experiencing Banff III, for TAC and CSA, respectively.
• Time to first BPAR is reported by only two studies, with contrasting results. However, the difference between arms for Campos et al.⁸³ is not statistically significant (p = 0.6631). The results reported by Baboolal et al.⁸² indicate that BPAR is achieved more quickly for participants receiving TAC (35 days, SD 13 days) rather than CSA (59 days, SD 38 days).

Mortality

Seven studies^{77,78,86,89,101,104,138} report on mortality for CSA + MMF compared with CSA + AZA. Pooling results of five studies^{78,86,101,104,138} for this combination imply no difference between arms at 1 year, with no evidence of heterogeneity across studies (Table 37 and Figure 18). The ORs switch from > 1 to < 1 for the pooled results at 1 and 3 years; however, the CIs cross ‘OR = 1’ in both cases, suggesting that there may be no difference between MMF and AZA (OR 1.19, 95% CI 0.47 to 3.02 and OR 0.56, 95% CI 0.26 to 1.23, respectively). The study reported by Tuncer et al.⁷⁸ provides data at 5 years, which also indicates no preference for either MMF or AZA (OR 0.73, 95% CI 0.15 to 3.50).

TABLE 37

Mortality for CSA + MMF vs. CSA + AZA

FIGURE 18

Forest plot: mortality for CSA + MMF vs. CSA + AZA.

Graft loss

Five studies^{77,86,89,104,138} report on graft loss, with results pooled at 0.5- and 1-year time points (Table 38 and Figure 19). However, the 0.5-year time point has only two studies^77,89 and a substantial level of heterogeneity (I² = 72.2%), therefore the OR of 0.58 (95% CI 0.04 to 0.59), which indicates that MMF is more effective at reducing graft loss, must be treated with caution.²⁰¹ The results for 1 year suggest no difference between arms (OR 0.76, 95% CI 0.38 to 1.50). Merville et al.¹³⁸ appear to show more of an effect in favour of MMF; however, the population is much smaller than that for the Tricontinental study⁸⁹ and Sadek et al.⁸⁶ and Weimer et al.¹⁰⁴ found no evidence of graft loss in either arm.

TABLE 38

Pooled results of graft loss for CSA + MMF vs. CSA + AZA

FIGURE 19

Forest plot: graft loss for CSA + MMF vs. CSA + AZA.

Graft function

Only Merville et al.¹³⁸ reported on this outcome. At both 6 months and 1 year there was no statistically significant difference in mean GRF (0.5 years; p = 0.7236 and 1 year; p = 0.6584) (Table 39).

TABLE 39

Graft function for CSA + MMF vs. CSA + AZA

Biopsy-proven acute rejection

Six studies^{77,86,89,101,104,138} report on BPAR. Unlike mortality and graft loss, BPAR analysis reveals that MMF is more beneficial than AZA at 0.5 and 1 year (0.5 years OR 0.50, 95% CI 0.35 to 0.72; 1 year OR 0.47, 95% CI 0.36 to 0.62) (Table 40 and Figure 20).¹⁰⁴

TABLE 40

Pooled results of BPAR for CSA + MMF vs. CSA + AZA

FIGURE 20

Forest plot: BPAR for CSA + MMF vs. CSA + AZA.

Severity of biopsy-proven acute rejection

Two studies were available for 0.5 years^77,89 and one study¹³⁸ for 1 year, although sample numbers are low for this study (Table 41). Overall, at 0.5 years the more severe classification of Banff III appears to be more likely in the AZA arm for people with BPAR (CSA 9.1%, AZA 15.9% for Sollinger et al.;⁷⁷ CSA 5.9%, AZA 11.9% for the Tricontinental group 1996⁸⁹).

TABLE 41

Severity of BPAR at 6 months for CSA + MMF vs. CSA + AZA

Time to biopsy-proven acute rejection

Insufficient data are provided for analysis on this outcome. Merville et al.¹³⁸ report 48.5 days for MMF and 43.7 days for AZA.¹³⁸

Summary of results for CSA + MMF vs. CSA + AZA

• Seven studies^{77,78,86,101,104,138,202} report on mortality for CSA + MMF vs. CSA + AZA. Pooling results of five studies^{78,86,101,104,138} for this combination imply no difference between arms at 1 year, with no evidence of heterogeneity across studies. The ORs switch from > 1 to < 1, for the pooled results at 1 and 3 years; however, the CIs cross ‘OR = 1’ in both cases, suggesting that there may be no difference between MMF and AZA (OR 1.19, 95% CI 0.47 to 3.02; and 0.56, 95% CI 0.26 to 1.23). The study reported by Tuncer et al.⁷⁸ provides data at 5 years, which also indicate no preference for either MMF or AZA (OR 0.73, 95% CI 0.15 to 3.50).
• Five studies^{77,86,104,138,202} report on graft loss, with results pooled at 0.5- and 1-year time points. However, the 0.5-year time point has only two studies^77,89 and a substantial level of heterogeneity (I² = 72.2%); therefore, the OR of 0.58 (95% CI 0.04 to 0.59), which indicates that MMF is more effective at reducing graft loss, must be treated with caution.²⁰¹ The results for 1 year suggest no difference between arms (OR 0.76, 95% CI 0.38 to 1.50). The study by Merville et al.¹³⁸ appears to show more of an effect in favour of MMF; however, the population is much smaller than that for the Tricontinental study⁸⁹ and Sadek et al.⁸⁶
• Only Merville et al.¹³⁸ reported on this outcome: at 6 months the mean GRF was greater for the MMF arm; however, this was reversed at 1 year, when AZA had greater GRF.¹³⁸ There is no significant difference between arms (0.5 years, p = 0.7236; 1 year, p = 0.6584).
• Six studies report on BPAR.^{77,86,101,104,138,202} Unlike mortality and graft loss, BPAR analysis reveals that MMF is more beneficial than AZA at 0.5 and 1 year [0.5 years, OR 0.50 (95% CI 0.35 to 0.72); 1 year, OR 0.47 (95% CI 0.36 to 0.62)].
• Two studies were available for 0.5 years^77,89 and one study¹³⁸ for 1 year. Overall, at 0.5 years the more severe classification of Banff III appears to be more likely in the AZA arm for people with BPAR (CSA 9.1%, AZA 15.9% for Sollinger et al.;⁷⁷ CSA 5.9%, AZA 11.9% for the Tricontinental group⁸⁹). Insufficient data are provided for analysis on time to BPAR. Merville et al.¹³⁸ report a slightly more rapid rate of 48.5 days for MMF and 43.7 days for AZA.

Mortality

Wlodarczyk et al.¹³⁹ report mortality at 0.5 years and Vacher-Caponat et al.¹²⁹ report at 1 year (Table 42). At 1 year, there are twice as many deaths for TAC + MMF as for CSA + AZA; however, although the OR is > 1, the wide 95% CIs imply no statistically significant difference between arms.

TABLE 42

Mortality for TAC + MMF vs. CSA + AZA

Graft loss

As with mortality, there is only one study for each time point of 0.5 years¹³⁹ and 1 year¹²⁹ (Table 43). The wide CIs highlight the low precision and indicate no difference between arms.

TABLE 43

Graft loss for TAC + MMF vs. CSA + AZA

Biopsy-proven acute rejection for TAC + MMF vs. CSA + AZA

Only two studies^129,139 have reported BPAR: one study at 0.5 years¹³⁹ and one study at the 1-year time point¹²⁹ (Table 44). In both cases the OR is < 1, indicating that TAC + MMF is associated with lower odds of BPAR (OR 0.64, 95% CI 0.42 to 0.98; OR 0.35, 95% CI 0.15 to 0.83, respectively).

TABLE 44

Biopsy-proven acute rejection for TAC + MMF vs. CSA + AZA

Severity of biopsy-proven acute rejection

This outcome is reported only by Vacher-Caponat et al.,¹²⁹ with no participants experiencing Banff II and III in the TAC + MMF arm, but with 14.3% and 4.8% reported in the CSA + AZA arm, respectively (Table 45).

TABLE 45

Severity of BPAR at 1 year for TAC + MMF vs. CSA + AZA

Summary for TAC + MMF vs. CSA + AZA

• Wlodarczyk et al.¹³⁹ report mortality at 0.5 years and Vacher-Caponat et al.¹²⁹ report mortality at 1 year. In both cases the OR is > 1, indicating that TAC + MMF is associated with greater odds of mortality; however, the 95% CIs cross ‘OR = 1’, implying no statistically significant difference between arms.
• Only one study reporting on graft loss at 0.5 years¹³⁹ and 1 year.¹²⁹ No significant difference is evident between treatments.
• Only two studies have reported BPAR: one study at 0.5 years¹³⁹ and one study at the 1-year time point.¹²⁹ In both cases the OR is < 1, indicating that TAC + MMF is associated with lower odds of BPAR (OR 0.64, 95% CI 0.42 to 0.98; OR 0.35, 95% CI 0.15 to 0.83, respectively).
• Severity of BPAR is reported by only one study,¹²⁹ with the greater proportion of people experiencing Banff II and III in the CSA + AZA arm.

Mortality

The effect estimate of five pooled studies^{51,102,103,130,203} at 1 year suggests that TAC + MMF is associated with higher odds of mortality (OR 1.62, 95% CI 0.77 to 3.44; Table 46 and Figure 21). However, although there is no evidence of heterogeneity across studies (I² = 0.0%), the CIs are wide and cross ‘OR = 1’, indicating low precision and a lack of statistical significance. Results for 2 years and 5 years also demonstrate no statistically significant difference between treatments.

TABLE 46

Mortality for TAC + MMF vs. CSA + MMF

FIGURE 21

Forest plot: mortality for TAC + MMF vs. CSA + MMF.

Graft loss

Graft loss is reported for five studies.^{51,102,103,130,153,203} The OR for pooled results at 1 year and 2 years (1.43 and 1.63, respectively) implies greater odds of graft loss for TAC + MMF; however, the CIs cross ‘OR = 1’, indicating no difference between arms (Table 47 and Figure 22).

TABLE 47

Graft loss for TAC + MMF vs. CSA + MMF

FIGURE 22

Forest plot: graft loss for TAC + MMF vs. CSA + MMF.

Kumar et al.²⁰³ report graft loss up to 5 years, with similar results of no difference between arms.

Graft function

Graft function as CRC is reported by three studies^51,103,130 up to 3 years (Table 48 and Figure 23). Pooling of results for 1- and 2-year data demonstrated a statistically significant difference in GRF in favour of TAC (WMD 4.22 ml/minute/1.73 m², 95% CI 1.23 to 7.20 ml/minute/1.73 m²; WMD 5.75, 95% CI 2.76 to 8.74 ml/minute/1.73 m², respectively). There is low evidence of heterogeneity across the 1-year studies (I² = 9.8%).

TABLE 48

Graft function for TAC + MMF vs. CSA + MMF

FIGURE 23

Forest plot: GRF for TAC + MMF vs. CSA + MMF.

Biopsy-proven acute rejection for TAC + MMF vs. CSA + MMF

Biopsy-proven acute rejection was reported by five studies.^{51,102,103,153,203} One-year outcomes provided by four of these studies^{51,103,153,203} were pooled (Table 49 and Figure 24). The study at 0.5 years by Kumar et al.²⁰³ indicates that lower odds of BPAR are associated with TAC. This is in agreement with the pooled results at 1 year, although some heterogeneity is noted across studies (OR 0.59, 95% CI 0.37 to 0.94; I² = 19.3%). The study reported by Hernández et al.¹³⁰ at 2 years does not demonstrate a statistical difference between arms (OR 1.22, 95% CI 0.51 to 2.91).

TABLE 49

Biopsy-proven acute rejection for TAC + MMF vs. CSA + MMF

FIGURE 24

Forest plot: BPAR for TAC + MMF vs. CSA + MMF.

Severity of biopsy-proven acute rejection

Two studies^51,130 report severity of BPAR separately at 1 and 2 years (Table 50). For year 1, results indicate that for people with BPAR, TAC + MMF and CSA + MMF have a similar proportion experiencing Banff III (TAC + MMF 7.8%; CSA + MMF 7.1%).⁵¹ The study by Hernández et al.¹³⁰ indicates no clear difference for between arms for all three classifications.¹³⁰

TABLE 50

Severity of BPAR at 1 year for TAC + MMF vs. CSA + MMF

Time to biopsy-proven acute rejection

Mean time to BPAR was reported by Ulsh et al.¹⁵³ in favour of TAC (Table 51).

TABLE 51

Time to BPAR for TAC + MMF vs. CSA + MMF

Summary of results for TAC + MMF vs. CSA + MMF

• The effect estimate of five pooled studies^{51,102,103,130,203} at 1 year suggests that TAC + MMF is associated with higher odds of mortality (OR 1.62, 95% CI 0.77 to 3.44). However, although there is no evidence of heterogeneity across studies (I² = 0.0%), the CIs are wide and cross ‘OR = 1’, indicating low precision and a lack of statistical significance. Results for 2 years and 5 years also demonstrate no statistically significant difference between treatments.
• Graft loss is reported for five studies.^{51,102,103,122,153} The OR for pooled results at 1 and 2 years (1.43 and 1.63, respectively) implies greater odds of graft loss for TAC + MMF; however, the CIs cross ‘OR = 1’, indicating no statistically significant difference between arms. The lack of difference remains at 5 years for the study reported by Kumar et al.²⁰³
• GRF is reported by three studies up to 3 years.^51,103,130 Pooling of results for 1- and 2-year data demonstrated a statistically significant difference in GRF in favour of TAC (WMD 4.22 ml/minute/1.73 m², 95% CI 1.23 to 7.20 ml/minute/1.73 m²; WMD 5.75 ml/minute/1.73 m², 95% CI 2.76 to 8.74 ml/minute/1.73 m², respectively).
• BPAR was reported by five studies,^{51,102,103,122,153} with four studies^{51,103,122,153} reporting at 1 year as being suitable for meta-analysis. The study at 0.5 years by Kumar et al.²⁰³ indicates that lower odds of BPAR are associated with TAC. This is in agreement with the pooled results at 1 year, although some heterogeneity is noted across studies (OR 0.59, 95% CI 0.37 to 0.94; I² = 19.3%). Two studies^51,130 report severity of BPAR separately at 1 year and 2 years with no clear difference in proportion of people with Banff III grading.
• Time to BPAR was reported by Ulsh et al.,¹⁵³ with a difference in favour of TAC of 88.7 days (p = 0.0001).

Mortality

Four studies^{105,123,141,204} report on mortality: two studies report at 0.5 years^105,123 and two studies report at 1 year^141,204 (Table 52 and Figure 25). At each time point, one of the studies had no deaths in either arm and both ORs indicate no statistical difference (0.5 years, OR 0.65, 95% CI 0.23 to 1.84; 1 year, OR 0.78, 95% CI 0.31 to 2.01).

TABLE 52

Mortality for TAC + MMF vs. TAC-PR + MMF

FIGURE 25

Forest plot: mortality for TAC + MMF vs. TAC-PR + MMF.

Graft loss

Four studies^{105,123,141,204} report on graft loss: two studies report at 0.5 years^105,123 and two studies report at 1 year.^141,204 As illustrated by the forest plot (Table 53 and Figure 26), no clear benefit is seen for either immediate-release or TAC-PR with regard to graft loss at 6 months and 1 year. ORs for both are identical and < 1; however, CIs cross ‘OR = 1’, indicating no statistical difference between arms (OR 0.83, 95% CIs 0.30 to 2.30 and 0.47 to 1.47).

TABLE 53

Graft loss for TAC + MMF vs. TAC-PR + MMF

FIGURE 26

Forest plot: graft loss for TAC + MMF vs. TAC-PR + MMF.

Graft function

Graft function is reported by three studies:^123,141,204 one study¹²³ for 0.5 years and two studies^141,204 for 1 year (Table 54 and Figure 27). Pooling of results at 1 year demonstrated no statistically significant difference in GRF (WMD 0.21 ml/minute/1.73 m², 95% CI –2.10 to 2.53 ml/minute/1.73 m²); however, the single study by Albano et al.¹²³ suggests immediate-release TAC to be more effective than TAC-PR for GRF (WMD 1.90 ml/minute/1.73 m², 95% CI to 5.40 ml/minute/1.73 m²).

TABLE 54

Graft function for TAC + MMF vs. TAC-PR + MMF

FIGURE 27

Forest plot: GRF for TAC + MMF vs. TAC-PR + MMF.

Biopsy-proven acute rejection for TAC + MMF vs. TAC-PR + MMF

Three studies^105,123,204 report BPAR at 0.5 years and two studies^141,204 report at 1 year (Table 55 and Figure 28). Pooling of results at both time points shows no significant difference between arms (OR 1.37 95% CI 1.00 to 1.87; OR 1.03, 95% CI 0.48 to 2.17). Furthermore, moderate heterogeneity exists across studies (I² = 34.8% and 44.4%).²⁰¹

TABLE 55

Biopsy-proven acute rejection for TAC + MMF vs. TAC-PR + MMF

FIGURE 28

Forest plot: BPAR for TAC + MMF vs. TAC-PR + MMF.

Severity of biopsy-proven acute rejection

Two studies^123,204 report severity of BPAR, both of which indicate that, for people with BPAR, the severity may be reduced with immediate TAC (Table 56).

TABLE 56

Severity of BPAR for TAC + MMF vs. TAC-PR + MMF

Summary for TAC + MMF vs. TAC-PR + MMF

• Four studies^{105,123,141,204} report on mortality: two studies report at 0.5 years^123,204 and two studies report at 1 year.^105,141 At each time point, one of the studies had no deaths in either arm and both ORs indicate no statistical difference (0.5 years, OR 0.65, 95% CI 0.23 to 1.84; 1 year, OR 0.78, 95% CI 0.31 to 2.01).
• Four studies^{105,123,141,204} report on graft loss: two studies report at 0.5 years^105,123 and two studies report at 1 year.^141,204 No clear benefit is seen for either immediate-release TAC or TAC-PR with regard to graft loss at 6 months and 1 year. ORs for both are identical and < 1; however, CIs cross ‘OR = 1’, indicating no statistical difference between arms (OR 0.83, 95% CI 0.30 to 2.30; and 95% CI 0.47 to 1.47). GRF is reported by three studies:^123,141,205 one study for 0.5 years¹²³ and two studies for 1 year.^141,205 Pooling of results at 1 year demonstrated no statistically significant difference in GRF (WMD 0.21 ml/minute/1.73 m², 95% CI –2.10 to 2.53 ml/minute/1.73 m²); however, the single study by Albano et al.¹²³ suggests that TAC is more effective than TAC-PR for GRF (WMD 1.90 ml/minute/1.73 m², 95% CI 1.70 to 2.10 ml/minute/1.73 m²).
• Three studies^105,123,204 report BPAR at 0.5 years and two studies report^141,204 at 1 year. Pooling of results at both time points shows no significant difference between arms (OR 1.37, 95% CI 1.00 to 1.87; OR 1.03, 95% CI 0.48 to 2.17).
• Two studies^123,204 report severity of BPAR, both of which indicate that, for people with BPAR, the severity may be reduced with immediate.

Summary for MMF + CSA vs. MPS + TAC

Only one study¹⁰⁶ was identified for this combination. No difference was identified between interventions, other than for GRF, where a statistically significant difference in favour of MPS at 0.5 years and 1 year (p < 0.0001) was noted. This effect is lost at later time points.

Summary for MMF + CSA vs. MPS + CSA

• Only one trial reported by Salvadori et al.¹²⁴ uses this combination. GRF and time to BPAR are not reported. All other results indicate no significant difference between MMF and MPS.

Mortality

Three studies^60,125,206 report 1-year outcomes, with the Belatacept Evaluation of Nephroprotection and Efficacy as First-line Immunosuppression Trial (BENEFIT)⁶⁰ and the BENEFIT–Extended Criteria Donors (BENEFIT-EXT)¹⁴² providing data for up to 5 years. The ORs generally fall at < 1 for all time points, indicating that BEL has a lower association with mortality than CSA (Table 60 and Figure 29). However, the CIs indicate that this is not statistically significant.

TABLE 60

Mortality for BEL + MMF vs. CSA + MMF

FIGURE 29

Forest plot: mortality for BEL + MMF vs. CSA + MMF.

Graft loss

The OR for graft loss is also reported by three studies^60,125,206 up to 5 years. Pooled results indicate that BEL may be preferable to CSA, although the results are not statistically significant (1 year, OR 0.74, 95% CI 0.42 to 1.31) (Table 61 and Figure 30). However, at 5 years, there may be more confidence that this effect is true (OR 0.40, 95% CI 0.19 to 0.87).

TABLE 61

Graft loss for BEL + MMF vs. CSA + MMF

FIGURE 30

Forest plot: graft loss for BEL + MMF vs. CSA + MMF.

Graft function

Graft function is reported by three studies^60,125,206 up to 5 years (Table 62 and Figure 31). The results must be treated with caution because of substantial heterogeneity across studies, which may be caused by variations in methods of calculation and measurement of GRF (I² = 73.6–91.2%). Pooling of results for 1- and 3-year data demonstrated a statistically significant difference for GRF in favour of BEL (WMD 7.83 ml/minute/1.73 m², 95% CI 1.57 to 4.10 ml/minute/1.73 m², and WMD 16.08 ml/minute/1.73 m², 95% CI 5.59 to 26.56 ml/minute/1.73 m², respectively).

TABLE 62

Graft function for BEL + MMF vs. CSA + MMF

FIGURE 31

Forest plot: GRF for BEL + MMF vs. CSA + MMF.

Biopsy-proven acute rejection

The results for BPAR indicate substantial heterogeneity for the 1-, 2- and 3-year time points (I² = 58.7%, 38.4% and 62.2%, respectively) (Table 63).^60,206,207 Overall, participants in the CSA arm appear to be less likely to experience BPAR at between 1 and 5 years, as opposed to those in the BEL arm (1 year, OR 1.53, 95% CI 0.78 to 3.02).

TABLE 63

Biopsy-proven acute rejection for BEL + MMF vs. CSA + MMF

Severity of biopsy-proven acute rejection

Three studies^60,125,142 report severity of BPAR at 1 year (Table 64). Overall, there is no clear difference between arms in the proportion of people with BPAR experiencing Banff II or III classification.^201,206

TABLE 64

Severity of BPAR for BEL + MMF vs. CSA + MMF

Summary for BEL + MMF vs. CSA + MMF

• Three studies^60,125,142 report 1-year outcomes, with two studies^60,142 providing data up to 5 years. The ORs generally fall to < 1 for all time points, indicating that BEL has a lower association with mortality than CSA. However, the CIs indicate that this is not statistically significant.
• The OR for graft loss up is also reported by three studies^60,125,142 up to 4 years. Pooled results indicate that BEL may be preferable to CSA, although the results are not statistically significant (1 year, OR 0.74, 95% CI 0.42 to 1.31). However, at 5 years, there may be more confidence that this effect is true (OR 0.40, 95% CI 0.19 to 0.87).
• GRF is reported by three studies^60,125,142 up to 5 years. The results must be treated with caution because of substantial heterogeneity across studies, which may be caused by variations in methods of calculation and measurement of GRF (I² = 73.6–91.2%). Pooling of results for 1- and 3-year data demonstrated a statistically significant difference for GRF in favour of BEL (WMD 7.83 ml/minute/1.73 m², 95% CI 1.57 to 14.10 ml/minute/1.73 m² and WMD 16.08 ml/minute/1.73 m², 95% CI 5.59 to 26.56 ml/minute/1.73 m², respectively).
• In contrast with previous outcomes, results for BPAR are more clear for the three studies.^60,125,142 However, there is substantial heterogeneity across studies at the 1-, 2- and 3-year time points (I² = 58.7%, 38.4% and 62.2%, respectively).^60,125,142 Overall, participants in the CSA arm appear to be less likely to experience BPAR at between 1 and 5 years, as opposed to those in the BEL arm (1 year, OR 1.53, 95% CI 0.78 to 3.02). Three studies^60,125,142 report severity of BPAR at 1 year. Overall, there is no clear difference between arms in the proportion of people with BPAR experiencing Banff II or III classification.^60,125,142

Mortality

Mortality is reported at 0.5,¹⁵⁰ 1¹³¹ and 3¹⁴³ years (Table 66 and Figure 32). Results are pooled for the 1- and 2-year time points, where the OR is > 1, indicating a preference in favour of MMF; however, this is not statistically significant (OR 1.83, 95% CI 0.80 to 4.20; OR 1.06, 95% CI 0.60 to 1.85, respectively). This trend is reflected at 0.5 years and 3 years.

TABLE 66

Mortality for EVL + CSA vs. MMF + CSA

FIGURE 32

Forest plot: mortality for EVL + CSA vs. MMF + CSA.

Graft loss

Three RCTs^131,143,150 report graft loss for this combination (Table 67 and Figure 33). There is considerable heterogeneity across studies for 1 and 3 years (I² = 80.0% and 74.3%, respectively) therefore results must be treated with caution. The study reported by Lorber et al.,¹⁴³ which favours MMF, appears to be in contrast with the ATLAS study;¹⁵⁰ however, there is no statistically significant difference between arms for either trial.

TABLE 67

Graft loss for EVL + CSA vs. MMF + CSA

FIGURE 33

Forest plot: graft loss for EVL + CSA vs. MMF + CSA.

Graft function

Lorber et al.¹⁴³ provide a median and range for GRF rather than a SD; therefore, results could not be pooled with the ATLAS study¹⁵⁰ (Table 68). Overall, there is no significant difference in GRF between EVL + CSA and MMF + CSA (p = 0.1989 to 0.3703).

TABLE 68

Graft function for EVL + CSA vs. MMF + CSA

Biopsy-proven acute rejection

The pooled and unpooled ORs of < 1 for this outcome all suggest that EVL is associated with lower odds of BPAR; however, the CIs indicate a lack of statistical significance (Table 69 and Figure 34).^131,143,150 There is no evidence of heterogeneity across studies.

TABLE 69

Biopsy-proven acute rejection for EVL + CSA vs. MMF + CSA

FIGURE 34

Forest plot: BPAR for EVL + CSA vs. MMF + CSA.

Severity of biopsy-proven acute rejection

Severity of BPAR is reported by only Takahashi et al.¹³¹ at 1 year (Table 70). No occurrences of Banff II or III classification were reported.

TABLE 70

Severity of BPAR for EVL vs. MMF

Summary for EVL + CSA vs. MMF + CSA

• Results for mortality are pooled for three studies^131,143,150 at the 1-year time point. The OR is > 1, indicating a preference in favour of MMF; however, this is not statistically significant (OR 1.83, 95% CI 0.80 to 4.20). This trend is reflected at 0.5 years and 3 years.
• Three RCTs^131,143,150 report graft loss for this combination; however, there is significant heterogeneity across studies for 1 and 3 years (I² = 80.0% and 74.3%, respectively). The study reported by Lorber et al.,¹⁴³ which favours MMF, appears to be in contrast with the ATLAS study,¹⁵⁰ which favours EVL; however, there is no statistical difference between arms for either trial.
• Lorber et al.¹⁴³ provide a median and range for GRF, rather than a SD; therefore, results could not be pooled with the ATLAS study.¹⁵⁰ Overall, there is no significant difference in GRF between EVL + CSA and MMF + CSA (p = 0.1989 to 0.3703).
• The pooled and unpooled ORs of < 1 for BPAR all suggest that EVL is associated with lower odds; however, the CIs indicate a lack of statistical significance.^131,143,150 There is no evidence of heterogeneity across studies. Severity of BPAR is reported only by Takahashi et al.¹³¹ at 1 year, when no occurrences of Banff II or III classifications are reported.¹³¹

Mortality

Pooled analysis of three studies^107,144,152 at 1 year for mortality indicates no significant difference between EVL + CSA and MPS + CSA (OR 1.02, 95% CI 0.42 to 2.45; Table 71 and Figure 35). No heterogeneity was evident across studies.

TABLE 71

Mortality for EVL + CSA vs. MPS + CSA

FIGURE 35

Forest plot: mortality for EVL + CSA vs. MPS + CSA.

Graft loss

The OR for graft loss is generated from three pooled studies,^107,144,152 which indicates that EVL may be preferable in reducing graft loss; however, this result is not statistically significant (OR 0.65, 95% CI 0.15 to 2.87) (Table 72 and Figure 36). Furthermore, moderate heterogeneity is noted across studies.

TABLE 72

Graft loss for EVL + CSA vs. MPS + CSA

FIGURE 36

Forest plot: graft loss for EVL + CSA vs. MPS + CSA.

Graft function

Two studies^107,144 report GRF; however, although results are pooled, the heterogeneity between them is extremely high (I² = 91.2%) (Table 73 and Figure 37). As such, the evidence is unclear as to which treatment may be beneficial.

TABLE 73

Graft function for EVL + CSA vs. MPS + CSA

FIGURE 37

Forest plot: GRF for EVL + CSA vs. CSA + MPS.

Biopsy-proven acute rejection

Biopsy-proven acute rejection is reported by three studies^107,144,152 at 1 year. Pooling of results indicates no statistically significant difference between EVL + CSA vs. MPS + CSA (OR 1.01, 95% CI 0.68 to 1.48) (Table 74 and Figure 38).

TABLE 74

BPAR for EVL + CSA vs. MPS + CSA

FIGURE 38

Forest plot: BPAR for EVL + CSA vs. MPS + CSA.

Severity of biopsy-proven acute rejection

The study reported by Tedesco-Silva et al.¹⁰⁷ suggests that more people with BPAR receiving MPS experienced Banff II classification; however, there was no difference for Banff III (Table 75). There were no Banff II or III episodes reported in the EVL treatment for Chadban et al.,¹⁵² with only one episode among those receiving MPS treatment; however, the sample size is small.

TABLE 75

Severity of BPAR for EVL + CSA vs. MPS + CSA

Summary for EVL + CSA vs. MPS + CSA

• Pooled analysis of three studies^107,144,152 at 1 year indicates no significant difference between EVL + CSA and MPS + CSA (OR 1.02, 95% CI 0.42 to 2.45). No heterogeneity was evident across studies.
• The OR for graft loss is generated from three pooled studies,^107,144,152 which indicates that EVL may be preferable in reducing graft loss; however, this result is not statistically significant (OR 0.648, 95% CI 0.146 to 2.870). Furthermore, moderate heterogeneity is noted across studies.
• BPAR is reported by three studies^107,144,152 at 1 year. Pooling of results indicates no statistically significant difference between EVL + CSA compared with MPS + CSA (OR 1.01, 95% CI 0.68 to 1.48).
• The study reported by Tedesco-Silva et al.¹⁰⁷ suggests that more people with BPAR receiving MPS experienced Banff II grading; however, there was no difference for Banff III (see Table 78). There were no Banff II or III episodes reported in the EVL treatment for Chadban et al.,¹⁵² with only one episode among those receiving MPS treatment; however, the sample size is small.¹⁰⁷

TABLE 78

Graft loss for SRL + CSA vs. MMF + CSA

Mortality

Two studies^109,122 were available for pooling at 1 year; however, one of the studies¹⁰⁹ had no deaths in either arm (Table 77 and Figure 39). The ORs appear to indicate lower odds associated with mortality for SRL; however, this is not statistically significant (1 year, OR 0.49, 95% CI 0.04 to 5.59). The 2- and 5-year time points also show no statistically significant difference (2 years, OR 0.31, 95% CI 0.05 to 1.92; 5 years, OR 1.0, 95% CI 0.36 to 2.77).

TABLE 77

Mortality for SRL + CSA vs. MMF + CSA

FIGURE 39

Forest plot: mortality for SRL + CSA vs. MMF + CSA.

Graft loss

Three studies^108,109,122 report on graft loss for SRL + CSA vs. MMF + CSA from 1 to 5 years (Table 78 and Figure 40).^108,109,122 The ORs up to 4 years slightly favour MMF; however, there is no statistically significant effect overall. At 5 years, the OR becomes ‘1’, indicating no benefit for either treatment.

FIGURE 40

Forest plot: graft loss for SRL + CSA vs. MMF + CSA.

Graft function

Graft function is monitored by one study¹⁰⁹ at 1 year (Table 79). No statistical difference is apparent between SRL and MMF (WMD 0.11 ml/minute/1.73 m²; p = 0.5708).

TABLE 79

Graft function for SRL + CSA vs. MMF + CSA

Biopsy-proven acute rejection

The study by Anil Kumar et al.¹²² reporting on BPAR at 1 year a similar percentage of events in both arms and therefore no difference between treatments (Table 80).¹²² At 2 years, Barsoum et al.¹⁰⁸ report more favourable outcomes for SRL; however, this is not statistically significant (OR 0.65, 95% CI 0.22 to 1.87).

TABLE 80

Biopsy-proven acute rejection for SRL + CSA vs. MMF + CSA

Summary for SRL + CSA vs. MMF + CSA

• Two studies^109,122 were available for pooling at 1 year; however, one of the studies¹⁰⁹ had no deaths in either arm. The ORs appear to indicate lower odds associated with mortality for SRL; however, this is not statistically significant (1 year, OR 0.49, 95% CI 0.04 to 5.59). The 2- and 5-year time points also show no statistically significant difference (2 years, OR 0.31, 95% CI 0.05 to 1.92; 5 years, OR 1.0, 95% CI 0.36 to 2.77).
• Three studies^108,122,208 report on graft loss for SRL + CSA compared with MMF + CSA from 1 to 5 years. ORs slightly favour MMF, but the effect is not statistically significant (1 year, OR 1.53, 95% CI 0.24 to 9.59).
• GRF is monitored by one study¹⁰⁹ at 1 year. No statistical difference is apparent between SRL and MMF (WMD 0.11 ml/minute/1.73 m²; p = 0.5708).
• The study by Anil Kumar et al.¹²² reporting on BPAR at 1 year had eight events in both arms and therefore no difference between treatment. At 2 years, Barsoum et al.¹⁰⁸ report more favourable outcomes for SRL; however, this is not statistically significant (OR 0.65, 95% CI 0.22 to 1.87).

Mortality

Eight RCTs^{94,110,112,114,122,145,155,180} report mortality from 0.08 years to 3 years (Table 81 and Figure 41). The ORs vary from < 1 at 0.08 years to > 1 at 3 years; however, the CIs are wide and cross ‘OR = 1’, indicating no statistical significance at any time point.

TABLE 81

Mortality for SRL + TAC vs. MMF + TAC

FIGURE 41

Forest plot: mortality for SRL + TAC vs. MMF + TAC.

Graft loss

Five RCTs^{112,114,122,145,180} were identified reporting graft loss (Table 82 and Figure 42). Four RCTs^{112,122,145,180} are pooled at 1 year, at which increased odds of graft loss are associated with SRL. However, the effect is not statistically significant (OR 1.43, 95% CI 0.44 to 4.66). There may also be moderate heterogeneity across studies following pooling (I² = 38.8%). The study by Anil Kumar et al.¹²² provides follow-up to 5 years, with the OR of < 1 favouring SRL; however, the results are not statistically significant.

TABLE 82

Graft loss for SRL + TAC vs. MMF + TAC

FIGURE 42

Forest plot: graft loss SRL + TAC vs. MMF + TAC.

Graft function

Three RCTs^111,114,145 were identified reporting GRF; however, because of the different time points, only two RCTs^111,114 could be pooled at 0.5 years (Tables 83 and 84; Figure 43). The results indicate no statistical difference between arms (WMD –1.875 ml/minute/1.73 m², 95% CI –8.425 to 4.675 ml/minute/1.73 m²). Furthermore, substantial heterogeneity across studies is evident (I² = 81.6%).

TABLE 83

Graft function for SRL + TAC vs. MMF + TAC (pooled results)

TABLE 84

Graft function for SRL + TAC vs. MMF + TAC (unpooled results)

FIGURE 43

Forest plot: GRF for SRL + TAC vs. MMF + TAC.

Biopsy-proven acute rejection

Biopsy-proven acute rejection is reported in four studies,^{112,122,145,180} with three studies^112,122,145 pooled at 1 year (Table 85 and Figure 44). The ORs for 0.5 years and 1 year suggest that MMF + TAC has lower odds of BPAR; however, the effect is not statistically significant (1 year, OR 1.16, 95% CI 0.56 to 2.60). There is also a low level of heterogeneity (I² = 27.8%).

TABLE 85

Biopsy-proven acute rejection for SRL + TAC vs. MMF + TAC

FIGURE 44

Forest plot: BPAR for SRL + TAC vs. MMF + TAC.

Severity of biopsy-proven acute rejection

Four studies^{94,112,114,155} report severity of BPAR (Table 86). No clear difference is apparent at either time point for Banff II or III classification between SRL and MMF.

TABLE 86

Severity of BPAR for SRL + TAC vs. MMF + TAC

Time to biopsy-proven acute rejection

Time to BPAR is reported by Sampaio et al.,¹¹² which appears to favour MMF (Table 87).

TABLE 87

Time to BPAR for SRL + TAC vs. MMF + TAC

Summary for SRL + TAC vs. MMF + TAC

• Eight RCTs^{94,110,112,114,122,145,155,180} report mortality from 0.08 years to 3 years. The ORs vary from < 1 at 0.08 years to > 1 at 3 years; however, the CIs are wide and cross ‘OR = 1’, indicating no statistical significance at any time point.
• Five RCTs were identified reporting graft loss.^{112,114,122,145,180} Four RCTs are pooled at 1 year where increased odds of graft loss are associated with SRL; however, the effect is not statistically significant (OR 1.43, 95% CI 0.44 to 4.66). There may also be moderate heterogeneity across studies following pooling (I² = 38.8%). The study by Anil Kumar et al.¹²² provides follow-up to 5 years, with the OR of < 1 favouring SRL; however, the results are not statistically significant.
• Three RCTs^111,114,145 were identified reporting GRF; however, because of the different time points, only two RCTs^111,114 could be pooled at 0.5 years. The results indicate no statistical difference between arms (WMD –1.875 ml/minute/1.73 m², 95% CI –8.425 to 4.675 ml/minute/1.73 m²). Furthermore, substantial heterogeneity across studies is evident (I² = 81.6%).
• BPAR is reported in four studies,^{112,122,145,180} with three studies^112,122,145 pooled at 1 year. The ORs for 0.5 years and 1 year suggest that MMF + TAC has lower odds of BPAR; however, the effect is not statistically significant (1 year, OR 1.16, 95% CI 0.56 to 2.60). There is also a low level of heterogeneity (I² = 27.8%).
• Four studies^{94,112,114,155} report severity of BPAR. No clear difference is apparent for Banff II or III classification between SRL and MMF. Time to BPAR is reported by Sampaio et al.,¹¹² with a statistically significant difference demonstrated in favour of MMF (MD 48.6 days; p = 0.0017).

Mortality

Eight studies^{115–117,127,134,146,147,149} report on mortality, with seven pooled at 1 year (Table 88 and Figure 45). No statistically significant difference was evident at this time point (1 year, OR 0.98, 95% CI 0.28 to 3.42). Data are available up to 5 years; however, the effect is also not statistically significant (5 years, OR 1.15, 95% CI 0.42 to 3.13).^115,209

TABLE 88

Mortality for SRL + MMF vs. CSA + MMF

FIGURE 45

Forest plot: mortality for SRL + MMF vs. CSA + MMF.

Graft loss

Eight studies^{115–117,127,134,146,147,149} report on graft loss from 0.5 years to 5 years (Table 89 and Figure 46). Seven studies^{115–117,127,134,147,149} are pooled at 1 year; however, there is no statistically significant difference between SRL + MMF and CSA + MMF (1 year, OR 1.06, 95% CI 0.44 to 2.56). Flechner et al.¹²⁷ and Büchler et al.¹³⁴ report graft loss at 5 years; however, again, there is no statistically significant difference and heterogeneity across studies is substantial (5 years, OR 0.57, 95% CI 0.05 to 7.25, I² = 76.6%).

TABLE 89

Graft loss for SRL + MMF vs. CSA + MMF

FIGURE 46

Forest plot: graft loss for SRL + MMF vs. CSA + MMF.

Graft function

Six studies^{117,118,127,134,146,149} report GRF (note, this includes Lebranchu et al.,⁶⁷ with 68.9 ml/minute/1.73 m² for SRL and 64.4 ml/minute for CSA; however, a SD is not provided). Pooled analysis for 0.5 years and 1 year suggests that improved GRF is associated with CSA, although this effect is not statistically significant (0.5 year, WMD 6.99 ml/minute/1.73 m², 95% CI 0.45 to 13.53 ml/minute/1.73 m²; 1 year, WMD 9.41 ml/minute/1.73 m², 95% CI –1.28 to 20.09 ml/minute/1.73 m²) (Table 90 and Figure 47). The individual studies for 2, 3, 4 and 5 years all have OR of < 1 and are statistically significant, therefore CSA appears beneficial in terms of GRF.

TABLE 90

Graft function for SRL + MMF vs. CSA + MMF

FIGURE 47

Forest plot: GRF for SRL + MMF vs. CSA + MMF.

Biopsy-proven acute rejection

Eight studies^{115–117,127,134,146,147,149} report on BPAR from 0.5 years to 5 years (Table 91 and Figure 48). Seven studies^{115–117,127,134,147,149} are pooled at 1 year; however, there is no statistically significant difference between arms, although the OR falls in favour of CSA + MMF (1 year, OR 1.29, 95% CI 0.81 to 2.04). Flechner et al.¹²⁷ and Büchler et al.¹³⁴ report BPAR at 5 years; however, again, there is no statistically significant difference and heterogeneity across studies is substantial (5 years, OR 0.77, 95% CI 0.37 to 1.63).

TABLE 91

Biopsy-proven acute rejection for SRL + MMF vs. CSA + MMF

FIGURE 48

Forest plot: BPAR for SRL + MMF vs. CSA + MMF.

Severity of biopsy-proven acute rejection

Severity of BPAR is reported by three studies^116,127,134 at 1 year (Table 92). Flechner et al.¹²⁷ also report results for 5 years. Sample sizes are relatively low, with similar proportions of people with BPAR experiencing Banff II and III classification.

TABLE 92

Severity of BPAR: SRL + MMF vs. CSA + MMF

Time to biopsy-proven acute rejection

Time to BPAR is reported by three studies^127,134,146 (Table 93). A statistically significant difference is seen by Durrbach et al.¹⁴⁶ (SRL 56 days, SD 57 days; CSA 94 days, SD 47 days; p = 0.0035).¹⁴⁶ The studies reported by Büchler et al.¹³⁴ and Flechner et al.¹²⁷ show no statistical difference between treatments (p = 0.3858 and p = 0.982, respectively).

TABLE 93

Time to BPAR: SRL + MMF vs. CSA + MMF

Summary of results for SRL + MMF vs. CSA + MMF

• Eight studies^{115–117,127,134,146,147,149} report on mortality, with seven studies^{115–117,127,134,147,149} pooled at 1 year. No statistically significant difference was evident at this time point (1 year, OR 0.98, 95% CI 0.28 to 3.42) or up to 5 years (5 years, OR 1.15, 95% CI 0.42 to 3.13).^115,209
• Eight studies^{115–117,127,134,146,147,149} report on graft loss from 0.5 years to 5 years. Seven studies^{115–117,127,134,147,149} are pooled at 1 year; however, there is no statistically significant difference between SRL + MMF and CSA + MMF (1 year, OR 1.06, 95% CI 0.44 to 2.56). Flechner et al.¹²⁷ and Büchler et al.¹³⁴ report graft loss at 5 years; however, again, there is no statistically significant difference and heterogeneity across studies is substantial (5 years, OR 0.57, 95% CI 0.05 to 7.25, I² = 76.6%).
• Six studies^{117,118,127,134,146,149} report GRF (note, this includes Lebranchu et al.,¹⁴⁹ with 68.9 ml/minute/1.73 m² for SRL and 64.4 ml/minute/1.73 m² for CSA; however, a SD is not provided). Pooled analysis for 0.5 years and 1 year suggests that improved GRF is associated with TAC, although this effect is not statistically significant (0.5 year, WMD 6.99 ml/minute/1.73 m² 95% CI 0.45 to 13.53 ml/minute/1.73 m²; 1 year, WMD 9.41 ml/minute/1.73 m², 95% CI –1.28 to 20.09 ml/minute/1.73 m²). The individual studies for 2, 3, 4 and 5 years all have OR of < 1 and are statistically significant, therefore TAC appears to be beneficial in terms of GRF.
• Eight studies^{115–117,127,134,146,147,149} report on BPAR from 0.5 years to 5 years. Seven studies^{115–117,127,134,147,149} are pooled at 1 year; however, there is no statistically significant difference between arms, although the OR falls in favour of CSA + MMF (1 year, OR 1.29, 95% CI 0.81 to 2.04). Flechner et al.¹²⁷ and Büchler et al.¹³⁴ report BPAR at 5 years; however, again, there is no statistically significant difference and heterogeneity across studies is substantial (5 years, OR 0.77, 95% CI 0.37 to 1.63).
• Severity of BPAR is reported by three studies^116,127,134 at 1 year. Sample sizes are relatively low, with similar proportions of people with BPAR experiencing Banff II and III classification. Time to BPAR is reported by three studies.^127,134,146 A statistically significant difference is seen by Durrbach et al.¹⁴⁶ (SRL 56 days, SD 57 days; CSA 94 days, SD 47 days; p = 0.0035). The studies reported by Büchler et al.¹³⁴ and Flechner et al.¹²⁷ show no statistical difference between treatments (p = 0.3858 and p = 0.982, respectively).

Mortality

Four studies^{92,93,135,154} are pooled with 1-year results for mortality; however, two of these studies^93,135 had no deaths for either TAC + MMF or SRL + MMF (Table 94 and Figure 49). Furthermore, analysis suggests no statistically significant difference between TAC + MMF and SRL + MMF (OR 0.80, 95% CI 0.13 to 4.99). Heilman et al.¹³⁵ also present results at 2 years (see Table 99). Again, results are not statistically significant (OR 2.10, 95% 0.19 to 23.83).

TABLE 94

Mortality for TAC + MMF vs. SRL + MMF

FIGURE 49

Forest plot: mortality for TAC + MMF vs. SRL + MMF.

TABLE 99

Summary of outcomes for TAC + MPS vs. SRL + MPS

Graft loss

Four studies^{92,93,135,154} are pooled with 1-year results for graft loss (Table 95 and Figure 50). Again, two of these studies^93,135 had no graft loss in either arm. Although the OR implies that reduced graft loss is associated with TAC, this is not statistically significant (OR 0.68, 95% CI 0.18 to 2.58).

TABLE 95

Graft loss for TAC + MMF vs. SRL + MMF

FIGURE 50

Forest plot: graft loss for TAC + MMF vs. SRL + MMF.

Graft function

Two studies^135,154 report GRF at 1 year and 2 years (Table 96 and Figure 51). The pooled ORs for both time points indicate no statistically significant difference between TAC + MMF and SRL + MMF (1 year, WMD –2.50 ml/minute/1.73 m², 95% CI –6.85 to 1.85 ml/minute/1.73 m²).

TABLE 96

Graft function for TAC + MMF vs. SRL + MMF

FIGURE 51

Forest plot: GRF for TAC + MMF vs. SRL + MMF.

Biopsy-proven acute rejection

Biopsy-proven acute rejection is reported by three studies^92,93,135 (Table 97 and Figure 52). Pooled results indicate that there are lower odds of BPAR associated with TAC at 1 year (OR 0.32, 95% CI 0.12 to 0.87). There does not appear to be any evidence of heterogeneity across studies (I² = 0.0%).

TABLE 97

Pooled results for BPAR – TAC + MMF vs. SRL + MMF

FIGURE 52

Forest plot: BPAR for TAC + MMF vs. SRL + MMF

Severity of biopsy-proven acute rejection

Only one study⁹³ reports on severity of BPAR (Table 98). For Banff classification II, there is no difference at 1 year between TAC + MMF and SRL + MMF; however, the sample size is very small.

TABLE 98

Severity of BPAR for TAC + MMF vs. SRL + MMF

Summary of results for TAC + MMF vs. SRL + MMF

• Four studies^{92,93,135,154} are pooled with 1-year results for mortality; however, two of these studies had no deaths for either TAC + MMF or SRL + MMF. Furthermore, analysis suggests no significant difference between TAC + MMF and SRL + MMF (OR 0.80, 95% CI 0.13 to 4.99) Heilman et al.¹³⁵ also present results at 2 years of mortality for TAC + MMF vs. SRL + MMF. Again, results are not statistically significant (OR 2.10, 95% 0.19 to 23.83).
• Four studies^{92,93,135,154} are pooled with 1-year results for graft loss. Again, two of these studies^93,135 had no graft loss in either arm. Although the OR implies reduced graft loss associated with TAC, this is not statistically significant (OR 0.68, 95% CI 0.18 to 2.58).
• Two studies^135,154 report GRF at 1 year and 2 years. The pooled ORs for both time points indicate no statistically significant difference between TAC + MMF and SRL + MMF (1 year, WMD –2.50 ml/minute/1.73 m², 95% CI –6.85 to 1.85 ml/minute/1.73 m²).
• BPAR is reported by three studies.^92,93,135 Pooled results indicate that there are lower odds of BPAR associated with TAC at 1 year (OR 0.32, 95% CI 0.12 to 0.87). There does not appear to be any evidence of heterogeneity across studies (I² = 0.0%). Only one study⁹³ reports on severity of BPAR. Banff classification I and II demonstrate no difference at 1 year between TAC + MMF and SRL + MMF; however, the sample size is very small.

Mortality

At both 1 year and 5 years there is no statistically significant difference between TAC + SRL and CSA + SRL for mortality (Table 102).^121,122 Notably, for Anil Kumar et al.¹²² there are no deaths in either arm at 1 year.

TABLE 102

Mortality for TAC + SRL vs. CSA + SRL

Graft loss

Two studies^121,122 report graft loss, with pooled result at 1 year and individual results up to 5 years (Table 103 and Figure 53). Results are consistent across all time points for lower odds being associated with graft loss for TAC + SRL; however, the effect is not statistically significant (1 year, OR 0.68, 95% CI 0.16 to 2.90).

TABLE 103

Graft loss for TAC + SRL vs. CSA + SRL

FIGURE 53

Forest plot: graft loss for TAC + SRL vs. CSA + SRL.

Graft function

Chen et al.¹²¹ report GRF at 0.5 years and 1 year (Table 104), which appears to be statistically significantly greater for TAC + SRL at 0.5 years and 1 year (p < 0.0001 and p = 0.0004, respectively).

TABLE 104

Graft function for TAC + SRL vs. CSA + SRL

Biopsy-proven acute rejection

This is reported only by Anil Kumar et al.¹²² at 1 year (Table 105). The OR implies BPAR to be more likely for CSA + SRL; however, this is not statistically significant (OR 0.48, 95% CI 0.08 to 2.74).

TABLE 105

Biopsy-proven acute rejection for TAC + SRL vs. CSA + SRL

Summary of results for TAC + SRL vs. CSA + SRL

• Owing to the same number events in either arm at both time points, there is no difference between TAC + SRL and CSA + SRL for mortality.^121,122
• Two studies^121,122 report graft loss, with pooled result at 1 year and individual results up to 5 years. Results are consistent across all time points in showing that lower odds are associated with graft loss for TAC + SRL; however, the effect is not statistically significant (1 year, OR 0.68, 95% CI 0.16 to 2.90).
• Chen et al.¹²¹ report GRF at 0.5 years and 1 year, which appears to be statistically significantly greater for TAC + SRL at 0.5 years and 1 year (p < 0.0001 and p = 0.0004, respectively).
• BPAR is reported only by Anil Kumar et al.¹²² at 1 year. The OR implies BPAR to be more likely for CSA + SRL; however, this is not statistically significant (OR 0.48, 95% CI 0.08 to 2.74).

Induction therapy results

Network meta-analysis was performed for all induction studies reporting graft loss, mortality, BPAR and eGFR at 1-year follow-up. Figure 54 displays the network for included induction studies.

FIGURE 54

Network diagram for all included induction studies. Note: circles denote number of studies.

Graft loss

Ten RCTs^{71–74,87,95–98,137} informing the effectiveness of three treatments (no induction/PBO, BAS and rATG) were included in the network for graft loss (Figure 55).

FIGURE 55

Network diagram for induction studies reporting graft loss. Note: circles denote number of studies.

The DIC suggested little difference between the fit of the fixed- and random-effects models, with the fixed effects being the slightly better fit; thus, only the results of the fixed-effects model are shown in Table 106.

TABLE 106

Odds ratios for induction therapy from a fixed-effects model: posterior mean (95% CrI)

From these analyses there is little evidence to suggest that BAS and rATG are more effective than no induction/PBO in reducing graft loss, as the 95% CrIs include an OR of ‘1’. Furthermore, there is little evidence to suggest that rATG is more effective than BAS. Of the three treatments analysed in this network, rATG was estimated as having a 57% probability of being the most effective treatment, with BAS having a 38% probability of being the most effective treatment. Analyses suggested that there was little evidence of inconsistency within this network.

Mortality

Ten RCTs^{71–74,87,95–98,137} informing the effectiveness of three treatments (no induction/PBO, BAS and rATG) were included in the network for mortality (Figure 56).

FIGURE 56

Network diagram for induction studies reporting mortality. Note: circles denote number of studies.

The DIC suggested little difference between the fit of the fixed- and random-effects models, with the fixed effects being the slightly better fit, thus only the results of the fixed-effects model are shown in Table 106.

From these analyses there is little evidence to suggest that BAS and rATG are more effective than no induction/PBO in reducing mortality, as the 95% CrIs include an OR of ‘1’ (see Table 106), and there is little evidence to suggest that rATG is more effective than BAS. Of the three treatments analysed in this network, rATG was estimated as having a 54% probability of being the most effective treatment, with BAS having a 22% probability of being the most effective treatment. Analyses suggested that there was little evidence of inconsistency within this network.

Biopsy-proven acute rejection

Nine RCTs^{71–74,87,96–98,137} informing the effectiveness of three treatments (no induction/PBO, BAS and rATG) were included in the network for mortality (Figure 57).

FIGURE 57

Network diagram for induction studies reporting BPAR. Note: circles denote number of studies.

The DIC suggested little difference between the fit of the fixed- and random-effects models, with the fixed effects being the slightly better fit, and so only the results of the fixed-effects model are shown in Table 106.

From these analyses, evidence suggests that BAS and rATG are more effective than no induction/PBO in reducing BPAR and that rATG is more effective than BAS. Of the three treatments analysed in this network, rATG was estimated as having a 96% probability of being the most effective treatment, with BAS having a 3% probability of being the most effective treatment. Analyses suggested that there was little evidence of inconsistency within this network.

Graft function

Five RCTs^{71–73,87,97} informing the effectiveness of three treatments (no induction/PBO, BAS and rATG) were included in the network for GRF (Figure 58).

FIGURE 58

Network diagram for induction studies reporting GRF. Note: circles denote number of studies.

The DIC suggested very little difference between the fit of the fixed- and random-effects models. For comparison with the above outcomes, the results of the fixed-effects model are shown in Table 107.

TABLE 107

Mean effects for induction therapy for the outcome GRF from a fixed-effects model: posterior mean (95% CrI)

There is no evidence to suggest that BAS or rATG is more effective than PBO/no induction, and no evidence to suggest that one treatment is more effective than the other. BAS has a 89% probability of being the most effective treatment, whereas rATG has a 5% probability of being the most effective treatment. Analyses suggested that there was little evidence of inconsistency within this network.

Maintenance therapy results

Network meta-analysis was performed for all maintenance studies reporting graft loss, mortality, BPAR and eGFR at 1-year follow-up. Figure 59 displays the network for included induction studies.

FIGURE 59

Network diagram for all included maintenance studies reporting graft loss. Note: circles denote number of studies.

Data on 13 treatments from 49 studies^{51,59,76,80,82–84,86,88–90,92,93,100,102–104,107–112,115,117,118,121,122,125–127,129,131,133–136,138,142–145,147,149–152,155,210} were potentially includable in the NMA (Figure 60). However, 11 studies had zero events in all treatment arms, so would not contribute information to the NMA; therefore, they were excluded from the NMA. Owing to the exclusion of these studies, the treatment EVL + MPS could not be included in the network. Therefore, data from 40 studies^{51,59,76,80,82–84,86,88–90,92,100,103,104,107,108,110–112,117,118,122,125–127,129,134,136,138,142–145,147,149–152,155} (including five three-arm studies^{51,104,126,152,155} and one four-arm study¹²²) on the effectiveness of 12 treatments to reduce graft loss informed the NMA. Thirteen^{82,90,100,104,112,126,127,138,144,145,147,149,152} of the 40 studies had at least one treatment arm with no graft loss events; therefore, 0.5 was added to each cell.

FIGURE 60

Network diagram for maintenance studies reporting graft loss. Note: circles denote number of studies.

The DIC indicated that the random-effects model was a slightly better fit to the data than the fixed-effects model (154.4 vs. 157.5), and so results from only the random-effects models are presented here. The results of the fixed-effects models are given in Appendix 6. The probabilities that each treatment was the most effective in reducing graft loss compared with all other treatments are shown in Table 108.

TABLE 108

Probability that each treatment is the most effective treatment for reducing graft loss

Although the results suggest that EVL has a 60% probability of being the most effective treatment for reducing graft loss compared with all other treatments (with SRL + AZA having a 29% probability), there is little evidence to suggest that treatment with EVL reduces graft loss compared with other treatments. The posterior median ORs for EVL compared with all of the other treatments are < 1, indicating a reduction in the odds of having a graft loss; however, the upper 95% CrIs limits are > 1, suggesting that EVL could increase the odds of a graft loss compared with all other treatments (Table 109). In fact, there is little evidence from the NMA to suggest that any treatment is more effective at reducing graft loss than any other treatment.

TABLE 109

Odds ratios (intervention vs. comparator treatment) for the outcome graft loss from a random-effects NMA: posterior median (95% CrI)

There is no evidence to suggest that this network is affected by inconsistencies between the direct and indirect evidence (see Appendix 6). The DICs were very similar between the consistency and inconsistency models (154.4 vs. 153.7) and the 95% CrIs based on the direct evidence overlapped those based on the direct and indirect evidence.

Mortality

Data on 13 treatments from 52 studies^{51,59,76,78,80,83,84,86,88–90,92,93,100,102–104,107–112,115–118,120–122,125–127,129–131,133–136,138,142,145,147,149–151,155,210} were potentially includable in the NMA (Figure 61). However, 10 trials^{93,102,104,109,115,117,121,131,135,149} had zero events in all arms and were excluded from the NMA, resulting in 42 trials^{51,59,76,78,80,83,84,86,88–90,92,100,103,107,108,110–112,116,118,120,122,125–127,129,130,133,134,136,138,142–145,147,150–152,155,210} contributing to the NMA (including four three-arm trials^{51,104,122,152,155} and one four-arm trial¹²²). Twelve^{59,76,78,90,92,100,120,126,127,136,145,152} of the 42 included trials had zero events in at least one treatment arm and so 0.5 was added to all cells in those trials.

FIGURE 61

Network diagram for maintenance studies reporting mortality. Note: circles denote number of studies.

Although the DIC indicated that the fixed-effects model was a slightly better fit to the data than the random-effects model (137.7 vs. 139.5), the random-effects results are presented here and used in the economic model for consistency, as the remaining maintenance treatment analyses indicated the random-effects model to be the best-fitting model. The results of the fixed-effects models are given in Appendix 6. The probabilities that each treatment was the most effective in reducing graft loss compared with all other treatments are shown in Table 110.

TABLE 110

Probability that each treatment is the most effective treatment for reducing mortality

The regimens SRL + AZA (34%), EVL (30%) and BEL + SRL (27%) were estimated to have the greatest probabilities of being the most effective treatments to reduce mortality compared with all others, with the remaining treatments having a very low probability of being the best treatment. This reflects the findings presented below (see Table 120), which show that SRL + AZA, EVL and BEL + SRL are consistently estimated to have posterior median ORs of < 1 compared with all treatments, but, as the upper 95% CrI limits are > 1, there is the possibility that these treatments could increase mortality compared with other treatments.

TABLE 120

Infections: induction therapies

The NMA suggests that BEL + MMF is more effective than TAC + MMF and SRL + MMF at reducing mortality. However, there is a great deal of uncertainty associated with many of the results presented (Table 111), especially for BEL + SRL.

TABLE 111

Odds ratios (intervention vs. comparator treatment) for the outcome mortality from a random-effects NMA: posterior median (95% CrI)

There is no evidence to suggest that this network is affected by inconsistencies between the direct and indirect evidence (see Appendix 6). The DICs were slightly lower for the consistency model than for the inconsistency model (139.5 vs. 143.9) and the 95% CrIs that were based on the direct evidence overlapped those based on the direct and indirect evidence.

Biopsy-proven acute rejection

Thirteen treatments and 42 studies^{51,59,76,81–83,86,88–90,92,93,100,103,107,110,112,115–118,120,121,125–127,129,131,133–136,142,144,145,147,149,150,152,210} (including three three-arm studies^51,126,152 and one four-arm study¹²²) contribute to this NMA (Figure 62).