Comparing Outcomes between Antibody Induction Therapies in Kidney Transplantation

Abstract

Kidney transplant recipients often receive antibody induction. Previous studies of induction therapy were often limited by short follow-up and/or absence of information about complications. After linking Organ Procurement and Transplantation Network data with Medicare claims, we compared outcomes between three induction therapies for kidney recipients. Using novel matching techniques developed on the basis of 15 clinical and demographic characteristics, we generated 1:1 pairs of alemtuzumab–rabbit antithymocyte globulin (rATG) (5330 pairs) and basiliximab-rATG (9378 pairs) recipients. We used paired Cox regression to analyze the primary outcomes of death and death or allograft failure. Secondary outcomes included death or sepsis, death or lymphoma, death or melanoma, and healthcare resource utilization within 1 year. Compared with rATG recipients, alemtuzumab recipients had higher risk of death (hazard ratio [HR], 1.14; 95% confidence interval [95% CI], 1.03 to 1.26; P<0.01) and death or allograft failure (HR, 1.18; 95% CI, 1.09 to 1.28; P<0.001). Results for death as well as death or allograft failure were generally consistent among elderly and nonelderly subgroups and among pairs receiving oral prednisone. Compared with rATG recipients, basiliximab recipients had higher risk of death (HR, 1.08; 95% CI, 1.01 to 1.16; P=0.03) and death or lymphoma (HR, 1.12; 95% CI, 1.01 to 1.23; P=0.03), although these differences were not confirmed in subgroup analyses. One-year resource utilization was slightly lower among alemtuzumab recipients than among rATG recipients, but did not differ between basiliximab and rATG recipients. This observational evidence indicates that, compared with alemtuzumab and basiliximab, rATG associates with lower risk of adverse outcomes, including mortality.

Antibody induction is a common and costly therapy during kidney transplantation.¹ As of 2014, approximately 90% of adult kidney transplant recipients (KTRs) received some form of antibody induction, more than double the percentage in 1999.^2–5 Induction therapy offers several potential benefits in kidney transplantation, including reduced risk of acute rejection and, potentially, longer graft survival.^6–9 Antibody induction agents may have particular value for patients with high immunologic reactivity and elevated probability of acute T cell rejection.^9,10

Less is known about long-term consequences of the choice of induction therapy on death, allograft failure, sepsis, and cancers, particularly in real-world practice. Infection is the second-highest cause of death among kidney transplant patients,^11,12 and a particular concern for elderly recipients.¹³ KTRs also have an elevated risk of cancer relative to the general population.¹⁴ For example, immunosuppression intensity is an important determinant of the risk of post-transplant lymphoproliferative disease after kidney transplantation.^15–19 Studies assessing the relative risks of these agents have often been hampered by short follow-up (in the case of randomized trials), limited power (with single-center studies), or the use of registry data, which have incomplete ascertainment of malignancy outcomes and sepsis. Furthermore, healthcare resource utilization of patients following various induction therapies is poorly understood.

In 2014, the most commonly used induction therapies in kidney transplantation were rabbit antithymocyte globulin (rATG) (approximately 50%), basiliximab (approximately 20%), and alemtuzumab (approximately 15%).^4,20,21 rATG is a poly-clonal T cell–depleting antibody manufactured in rabbits.²⁰ Alemtuzumab is a humanized mAb that depletes B and T cells by targeting the CD-52 glycoprotein cell surfaces.²² Some proponents of alemtuzumab endorsed the idea that this therapy could facilitate steroid-free transplants.^20,23,24 Basiliximab is a nondepleting mAb that prevents T cell activation by blocking the IL-2 receptor on cell surfaces. A randomized trial suggests basiliximab is associated with fewer infections overall than rATG, but is less effective at preventing acute allograft rejection.^11,20 Thus, patients with high risk of acute rejection typically receive a more potent agent such as alemtuzumab or rATG.^25–27

To compare a broad range of outcomes by induction strategy, we linked data from the Organ Procurement and Transplantation Network (OPTN) and the Centers for Medicare and Medicaid Services (CMS). To generate transparent results and minimize confounding, we leveraged recent advances in multivariable matching^28–32 and created precisely matched pairs of KTRs exposed to rATG versus alemtuzumab, and rATG versus basiliximab.

Results

Figure 1 shows cohort generation. Among the 21,168 rATG recipients, we generated matches to 5330 alemtuzumab recipients and 9378 basiliximab recipients (>99% of alemtuzumab and basiliximab recipients meeting study criteria were matched).

Figure 1.

Generation of two cohorts of kidney transplant recipients matched on induction therapy. ¹Adults defined as 18-90 years of age. ²Evidence of any of the following drugs: alg, anti-cam-1, anti-IL-6, anti-LFA-1, anti-tnf, atg, IL-1-receptor antagonist, okt3, okt4, rituximab, t10b9, xomazymecd5, zenapax, soluble il-1 receptor, nratgnrats. ³Evidence of multiple cell cycle targets, cytosolic protein inhibitors, or azathioprine. ⁴Rabbit antithymocyte globulin. UNOS, United Network for Organ Sharing; HMO, health maintenance organization; ALG, antilymphocyte globulin; anti-cam-1, anti-intercellular adhesion molecule 1; anti-LFA-1, anti-lymphocyte function-associated antigen 1; anti-tnf, anti-tumor necrosis factor; atg, equine antithymocyte globulin; nratgnrats, Nashville rabbit antithymocyte globulin/Nashville rabbit antithymocyte serum; okt3, Muromonab-CD3; okt4, Anti-CD4 Antibody-Cytokine Fusion Protein; t10b9, T10B9 monoclonal antibody; xomazymedcd5, Xoma Zyme - CD5+; zenapax, Zenapax - Daclizumab.

Table 1 shows the characteristics of the initial rATG cohort and the matched pairs. Before matching, a higher percentage of rATG relative to basiliximab recipients were black (30% versus 19%), had a previous kidney transplant (15% versus 7%), had peak panel reactive antibody (PRA) between 80% and 100% (12% versus 4%), and received deceased donor organs (58% versus 54%). These findings suggest that before matching, the basiliximab group was at lower immunologic risk than the rATG group. Before matching, a higher percentage of patients who received rATG relative to alemtuzumab recipients were black (30% versus 26%), had a previous kidney transplant (15% versus 11%), had peak PRA between 80% and 100% (12% versus 9%), and received deceased donor organs (58% versus 51%).

Table 1.

Subject characteristics

Characteristics	Match 1: Alemtuzumab and rATG				All rATG (n=21,168) n (%)	Match 2: Basiliximab and rATG
	Matched Alemtuzumab (n=5330), n (%)	Matched rATG (n=5330), n (%)	Stand. Diff. after Match	P Value after Match		Matched Basiliximab (n=9378), n (%)	Matched rATG (n=9378), n (%)	Stand. Diff. after Match	P Value after Match
Mean age	50.5	50.6	0.00	0.85	50.2	51.3	51.3	0.00	0.88
Age 65–90 yra	857 (16)	857 (16)	0.00	1.00	3324 (16)	1755 (19)	1755 (19)	0.00	1.00
Men	3185 (60)	3201 (60)	−0.01	0.77	12,489 (59)	5993 (64)	5928 (63)	0.01	0.33
Black racea	1382 (26)	1382 (26)	0.00	1.00	6431 (30)	1781 (19)	1781 (19)	0.00	1.00
Median neighborhood income, $	42,766	42,103	0.04	0.46	42,519	42,798	43,198	−0.03	0.10
Prior kidney transplanta	603 (11)	603 (11)	0.00	1.00	3228 (15)	640 (7)	640 (7)	0.00	1.00
Peak PRAa
0%–20%	3489 (65)	3489 (65)	0.00	1.00	13,417 (63)	6931 (74)	6931 (74)	0.00	1.00
20%–80%	608 (11)	608 (11)	0.00	1.00	2915 (14)	879 (9)	879 (9)	0.00	1.00
80%–100%	458 (9)	458 (9)	0.00	1.00	2452 (12)	400 (4)	400 (4)	0.00	1.00
Missing	775 (15)	775 (15)	0.00	1.00	2384 (11)	1168 (12)	1168 (12)	0.00	1.00
HLA mismatcha	4772 (90)	4772 (90)	0.00	1.00	19,134 (90)	8330 (89)	8330 (89)	0.00	1.00
Transplant year	2006.5	2006.5	0.01	0.78	2006.2	2005.9	2006.0	−0.02	0.23
Recipient BMI	27.7	27.6	0.02	0.90	27.6	27.4	27.4	0.00	0.54
Weight, kg	81.0	80.8	0.01	0.87	80.4	80.5	80.5	0.00	1.00
Missing BMIa	236 (4)	236 (4)	0.00	1.00	1944 (9)	573 (6)	573 (6)	0.00	1.00
Missing weight	115 (2)	112 (2)	0.00	0.89	969 (5)	210 (2)	212 (2)	0.00	0.96
Missing BMI and weighta	112 (2)	112 (2)	0.00	1.00	965 (5)	210 (2)	210 (2)	0.00	1.00
Time on dialysis
0–1 yr	1083 (20)	1118 (21)	−0.02	0.42	3573 (17)	2010 (21)	2043 (22)	−0.01	0.57
1–3 yr	1426 (27)	1388 (26)	0.02	0.42	5181 (24)	2618 (28)	2650 (28)	−0.01	0.62
3–6 yr	1300 (24)	1306 (25)	0.00	0.91	5531 (26)	2143 (23)	2180 (23)	−0.01	0.53
6–10 yr	451 (8)	464 (9)	−0.01	0.68	1825 (9)	572 (6)	577 (6)	0.00	0.90
>10 yr	98 (2)	81 (2)	0.02	0.23	534 (3)	126 (1)	126 (1)	0.00	1.00
Missing	972 (18)	973 (18)	0.00	1.00	4524 (21)	1909 (20)	1802 (19)	0.03	0.05
Cause of ESRD
GN	1040 (20)	1035 (19)	0.00	0.92	4034 (19)	1841 (20)	1844 (20)	0.00	0.97
HTN	1078 (20)	1041 (20)	0.02	0.38	4365 (21)	1764 (19)	1785 (19)	−0.01	0.71
PKD	410 (8)	382 (7)	0.02	0.32	1463 (7)	776 (8)	769 (8)	0.00	0.87
Congenital	77 (1)	65 (1)	0.02	0.35	336 (2)	166 (2)	156 (2)	0.01	0.61
Other	1440 (27)	1494 (28)	−0.02	0.25	6506 (31)	2673 (29)	2644 (28)	0.01	0.65
Diabetes	1768 (33)	1771 (33)	−0.00	0.97	6549 (31)	3032 (32)	3042 (32)	0.00	0.89
Hepatitis C+	172 (3)	162 (3)	0.01	0.62	1230 (6)	494 (5)	486 (5)	0.00	0.82
Donor type
Living	1856 (35)	1880 (35)	−0.01	0.64	6192 (29)	3364 (36)	3336 (36)	0.01	0.68
Deceased	2703 (51)	2734 (51)	−0.01	0.56	12,186 (58)	5042 (54)	5018 (54)	0.01	0.74
Extended criteria	771 (15)	716 (13)	0.03	0.13	2790 (13)	972 (10)	1024 (11)	−0.02	0.23
Propensity score	0.221	0.218	0.04	0.37	0.196	0.348	0.346	0.02	0.16
Immunosuppression (not used for match)
MMF or mycophenolate sodium	4530 (85)	5070 (95)	−0.34	<0.001	20,135 (95)	8836 (94)	8932 (95)	−0.05	0.002
Prednisone use	1543 (29)	3529 (66)	−0.82	<0.001	14,685 (69)	8143 (87)	6177 (66)	0.52	<0.001
Calcineurin or mTOR use
Tacrolimus	4514 (85)	4590 (86)	−0.04	0.04	17,929 (85)	6482 (69)	7792 (83)	−0.34	<0.001
Cyclosporine	345 (6)	411 (8)	−0.05	0.01	1876 (9)	1972 (21)	970 (10)	0.30	<0.001
Sirolimus	74 (1)	122 (2)	−0.07	<0.001	499 (2)	414 (4)	235 (3)	0.11	<0.001
None	397 (7)	207 (4)	0.15	<0.001	864 (4)	510 (5)	381 (4)	0.06	<0.001

Stand. Diff., standardized difference; HTN, hypertension; PKD, polycystic kidney disease; MMF, mycophenolate mofetil; mTOR, mammalian target of rapamycin.

^aVariable matched exactly.

After matching, the demographic and clinical characteristics were extremely similar in both matches, with differences in means that were <5% of the SD and P values >0.05 for all matched characteristics. Among the alemtuzumab-rATG pairs, 78 centers used alemtuzumab and 188 centers used rATG. Among the basiliximab-rATG pairs, 196 centers used basiliximab and 201 centers used rATG.

As noted in the Concise Methods, we purposefully did not match on oral immunosuppression in the primary analyses. Over 75% of matched patients received tacrolimus. Prednisone was much more commonly prescribed to basiliximab (87%) and rATG (66%) versus alemtuzumab (29%) recipients (Supplemental Table 1, Table 1). Secondary analyses were conducted to compare outcomes among matched pairs who received tacrolimus and mycophenolate mofetil or mycophenolate sodium; these analyses separated pairs in which either both received maintenance prednisone or neither received prednisone.

Primary Outcomes

Relative to matched rATG recipients, alemtuzumab and basiliximab recipients had higher mortality risk (alemtuzumab hazard ratio [HR], 1.14; 95% confidence interval [95% CI], 1.03 to 1.26; P<0.01; basiliximab HR, 1.08; 95% CI, 1.01 to 1.16; P=0.03). Alemtuzumab recipients also had higher risk of death or allograft failure (HR, 1.18; 95% CI, 1.09 to 1.28; P<0.001) versus rATG recipients (Table 2). Figures 2 and 3 show Kaplan–Meier survival estimates for the outcomes, with scaled versions in the Supplemental Appendix (Supplemental Figure 1, A and B). Additional analyses of outcomes are presented in Supplemental Tables 2–7.

Table 2.

Outcomes after kidney transplantation among matched pairs

Clinical Outcomes	HR	95% CI Lower	95% CI Upper	Paired Cox Model P Value
Alemtuzumab-rATG Matches
Death	1.14	1.03	1.26	<0.01
Death or allograft failure	1.18	1.09	1.28	<0.001
Death or sepsis	0.94	0.86	1.02	0.13
Death or lymphoma	1.15	1.01	1.32	0.04a
Death or melanoma	1.17	1.02	1.34	0.03
Basiliximab-rATG Matches
Death	1.08	1.01	1.16	0.03
Death or allograft failure	1.04	0.98	1.10	0.25a
Death or sepsis	1.05	0.99	1.12	0.13a
Death or lymphoma	1.12	1.01	1.23	0.03
Death or melanoma	1.10	1.00	1.21	0.05

rATG is the reference group. HRs, 95% CIs, and P values computed using a paired Cox proportional hazards model. Outcomes that include those derived from Medicare claims (those using sepsis, lymphoma, melanoma) are censored at 3 years post-transplant.

^aSubsequent analyses suggested the possibility that hazards were nonproportional over time (see Kaplan–Meier figures). For alemtuzumab-rATG pairs, the 3-year differences across groups for death or lymphoma were 1.3% (95% CI, −0.1% to 2.7%). For basiliximab-rATG pairs, the 5-year difference in death or allograft failure was 1.1% (95% CI, −0.1% to 2.4%), and 3-year differences in death or sepsis were 1.9% (95% CI, 0.4% to 3.2%). 95% CIs for differences in survival probabilities were generated using bootstrapping.

Figure 2.

Kaplan–Meier survival estimates for alemtuzumab-rATG pairs with five outcomes. (A) Survival. (B) Survival without allograft failure. (C) Survival without sepsis. (D) Survival without lymphoma. (E) Survival without melanoma.

Figure 3.

Kaplan–Meier survival estimates for basiliximab-rATG pairs with five outcomes. (A) Survival. (B) Survival without allograft failure. (C) Survival without sepsis. (D) Survival without lymphoma. (E) Survival without melanoma.

Secondary Outcomes

Patients who received alemtuzumab experienced modestly elevated risk of death or melanoma (HR, 1.17; 95% CI, 1.02 to 1.34; P=0.03) relative to patients who received rATG. We found that alemtuzumab recipients had a higher risk of death or lymphoma (HR, 1.15; 95% CI, 1.01 to 1.32; P=0.04), although subsequent analyses and inspection of the Kaplan–Meier figure revealed the possibility that the risk associated with alemtuzumab was nonproportional over time. At 3 years, alemtuzumab recipients had a 1.3% higher rate of death or lymphoma (95% CI, −0.1% to 2.7%) than rATG recipients (Supplemental Table 7).

Patients who received basiliximab had higher risk of death or lymphoma (basiliximab HR, 1.12; 95% CI, 1.01 to 1.23; P=0.03) versus patients who received rATG (Table 2).

Secondary Analyses of Pairs with Identical Maintenance Immunosuppression Medications

Among the subset of pairs in which both patients received prednisone (513 pairs), alemtuzumab recipients had higher risk of death or allograft failure relative to rATG recipients (HR, 1.35; 95% CI, 1.04 to 1.77; P=0.03). Matched alemtuzumab-rATG pairs without prednisone did not differ significantly in post-transplant outcomes. Basiliximab-rATG pairs with equivalent oral immunosuppression medications also did not differ in post-transplant outcomes (Table 3).

Table 3.

Outcomes after kidney transplantation among maintenance immunosuppression matched pairs; rATG is the reference group

Clinical Outcomes	Prednisone		No Prednisone
	HR (95% CI)	Paired Cox Model P Value	HR (95% CI)	Paired Cox Model P Value
Alemtuzumab-rATG matches	n=513 Matched Pairs		n=973 Matched Pairs
Death	1.37 (0.97 to 1.93)	0.07	1.08 (0.82 to 1.42)	0.58
Death or allograft failure	1.35 (1.04 to 1.77)	0.03	1.16 (0.93 to 1.44)	0.18
Death or sepsis	1.10 (0.85 to 1.44)	0.46	0.98 (0.79 to 1.22)	0.87
Death or lymphoma	1.05 (0.68 to 1.64)	0.82	1.29 (0.90 to 1.86)	0.17
Death or melanoma	1.09 (0.68 to 1.74)	0.72	1.26 (0.86 to 1.84)	0.25
Basiliximab-rATG matches	n=2882 Matched Pairs		n=299 Matched Pairs
Death	0.96 (0.85 to 1.09)	0.56	1.15 (0.72 to 1.84)	0.55
Death or allograft failure	0.97 (0.87 to 1.08)	0.62	1.21 (0.81 to 1.81)	0.36
Death or sepsis	1.03 (0.92 to 1.15)	0.58	0.98 (0.65 to 1.48)	0.92
Death or lymphoma	0.97 (0.82 to 1.16)	0.76	1.06 (0.55 to 2.01)	0.87
Death or melanoma	0.98 (0.81 to 1.17)	0.78a	0.82 (0.44 to 1.53)	0.53

rATG is the reference group. HRs, 95% CIs, and P values computed using a paired Cox proportional hazards model. Outcomes that include those derived from Medicare claims (those using sepsis, lymphoma, melanoma) are censored at 3 years post-transplant.

^aSubsequent analyses suggested the possibility that hazards were nonproportional over time (see Kaplan–Meier figure). The 3-year difference with a bootstrapped 95% CI was −0.1% (−1.9% to 1.8%).

Health Care Utilization in the Year after Transplantation

After 1 year, the maximum likelihood type estimate (m-estimate) of total cost for alemtuzumab recipients was $3083 lower than for rATG recipients (P<0.001), although this difference was only 4% of the total cost for alemtuzumab recipients ($84,291). Costs were not significantly different between basiliximab and rATG recipients (P=0.13) (Table 4).

Table 4.

Comparison of costs and payments after 1 year, relative to rATG

Cost and Payment Outcomes	Alemtuzumab Match (n=5330 Matched Pairs)				Basiliximab Match (n=9378 Matched Pairs)
	Alemtuzumab	Matched rATG	Paired Difference (95% CI)	P Value	Basiliximab	Matched rATG	Paired Difference (95% CI)	P Value
1-yr total cost, $
Mean	86,290	88,948	−2658 (−4648 to −668)	<0.01	87,926	88,679	−753 (−2324 to 819)	0.35
M-estimate	84,291	87,597	−3083 (−4858 to −1303)	<0.001	85,990	87,134	−1058 (−2427 to 313)	0.13
1-yr total payment, $
Mean	75,382	79,112	−3729 (−6614 to −844)	0.01	79,589	76,814	2775 (356 to 5195)	0.03
M-estimate	73,500	76,346	−2847 (−5345 to −355)	0.03	76,102	74,183	1782 (−118 to 3685)	0.07

P values were calculated using the test of the weighted m-statistic for continuous m-estimates and the permutational t test for continuous means. M-estimate, maximum likelihood type estimate.

Effect Modification by Age

We examined whether the effect of induction therapy was modified by elderly status (age 65–90 years). For the outcome of death or sepsis, alemtuzumab (versus reference rATG) provided a benefit to nonelderly recipients (HR, 0.89; 95% CI, 0.81 to 0.98; P=0.02) that was absent for elderly recipients (HR, 1.14; 95% CI, 0.95 to 1.37; P=0.17; elderly pairs versus nonelderly interaction HR, 1.28; 95% CI, 1.04 to 1.57; P=0.02). However, an inspection of all of the outcomes did not suggest consistent advantages for alemtuzumab versus rATG in the nonelderly subgroup (Supplemental Table 2).

Analyses of effect modification were nonsignificant among the basiliximab-rATG pairs (Supplemental Table 2).

Subcohort Analysis of Nonblack Patients

Nonblack alemtuzumab and basiliximab patients did not differ from patients who received rATG in risk of death or melanoma (Supplemental Table 4).

Exploratory Analyses of Rejection

Compared with rATG matched pairs, alemtuzumab and basiliximab patients showed statistically different risk of the composite outcome of death, allograft failure, or acute rejection (Supplemental Table 4; Stuart–Maxwell P<0.01 for alemtuzumab; P=0.04 for basiliximab). Relative to rATG recipients, alemtuzumab and basiliximab patients had higher odds of acute rejection by 1 year (alemtuzumab odds ratio, 1.31; 95% CI, 1.14 to 1.51; P<0.001; basiliximab odds ratio, 1.16; 95% CI, 1.04 to 1.30; P<0.01). Overall, 84.3% of patients who received alemtuzumab were alive at 1 year without allograft failure or rejection versus 86.6% of patients who received rATG (P<0.001).

Discussion

This matched-cohort study suggests that matched KTRs who receive rATG experience better or similar post-transplant outcomes relative to patients who receive alemtuzumab or basiliximab. Compared with rATG matched pairs, patients who received alemtuzumab had a 14% higher hazard of death and an 18% higher hazard of either death or allograft failure. Patients who received basiliximab had slightly greater risk of death versus matched patients who received rATG. These findings should motivate clinicians to consider the advantages of rATG when attempting to reduce complications after kidney transplantation.

Our study advances the knowledge about induction therapy in several ways. First, by linking CMS administrative claims to OPTN registry data, we report a wider range of adverse outcomes than many prior studies. Second, the analysis leverages major advances in multivariable matching methods, which enable transparent reporting of outcomes across highly similar pairs of KTRs. We generated 1:1 matches for >99% of basiliximab and alemtuzumab recipients meeting inclusion criteria. Matching comprised multiple demographic and clinical characteristics, including an estimate of neighborhood poverty. Third, the study is unique in its presentation of 1-year healthcare resource utilization at >200 centers.

In our primary approach, we did not adjust for oral immunosuppression because induction therapies are often intentionally prescribed with particular combinations of oral immunosuppression. Rather, we compared induction therapies in conjunction with the other immunosuppression with which they are commonly combined. Most notably, alemtuzumab has been promoted as an agent that facilitates corticosteroid withdrawal. In practice, prednisone was much more commonly prescribed to basiliximab (87%) and rATG (66%) versus alemtuzumab (29%) recipients. As a result, the underlying reasons and threshold for prescribing prednisone to patients may differ by induction strategy. However, we recognize that many clinicians may attribute differences in outcomes to oral immunosuppression. We addressed this concern with subgroup analyses of pairs who received tacrolimus, mycophenolate mofetil, or mycophenolate sodium, and who both either received prednisone or no prednisone. They demonstrated advantages for rATG versus alemtuzumab for death and death or allograft failure among prednisone pairs, but differences were not significant among pairs without prednisone. Differences between the basiliximab-rATG pairs were not statistically significant. These subgroup analyses had less power than the main analyses.

Our results showing superior outcomes for rATG are supported by many^{19,21,26,33,34} but not all prior observational studies and clinical trials. In a retrospective analysis of United States registry data (n=14,336), Schold et al. observed less allograft failure associated with rATG versus alemtuzumab among retransplant patients specifically.²¹ In a literature review of both observational studies and trials, Gaber et al. concluded that rATG provides better allograft survival versus basiliximab among patients with higher immunologic risk.²⁶ Additionally, a recent trial suggested inferior death-censored graft survival with alemtuzumab induction compared with rATG.³⁴

In particular, our results should be contrasted with the Induction with Tacrolimus (INTAC) trial. INTAC participants at high risk of rejection were randomized to alemtuzumab versus rATG whereas low-risk participants were randomized to alemtuzumab versus basiliximab; all 501 participants received tacrolimus, mycophenolate mofetil, and a 5-day glucocorticoid taper. INTAC reported no difference in 3-year mortality or graft failure across the agents.²³ In analyses of patients with equal maintenance immunosuppression, we still find lower mortality or graft failure for rATG patients versus alemtuzumab patients when the pairs received prednisone. However, when patients did not receive prednisone, no difference in outcomes was observed, consistent with INTAC. The INTAC trial had lower rates of death or allograft failure at 3 years post-transplant than our study, perhaps explained by the trial’s exclusion of expanded criteria donors (ECD), donation after cardiac death (DCD), and hepatitis C–positive organs. INTAC participants may also have been healthier and better monitored under the trial infrastructure than individuals in our cohort. Therefore, our study adds important new information from real-world practice, suggesting meaningful risks associated with alemtuzumab compared with rATG. We estimate that the observed rates of death or allograft failure are 3 percentage points higher for alemtuzumab versus rATG recipients by 5 years after transplantation, a clinically important difference.

In our study, differences in outcomes between the basiliximab-rATG pairs were smaller in magnitude and less robust in subgroup analyses. The hazard of death was only 8% higher among basiliximab versus rATG recipients, and death or allograft failure did not differ between groups. A clinical trial by Brennan et al. (n=288 recipients) reported that mortality between rATG and basiliximab was not significantly different, although the trial was not powered to detect differences in mortality alone.¹¹ Despite concerns that induction might confer greater risks among elderly patients, our analyses detected no evidence to suggest that basiliximab has substantial advantages among elderly kidney transplant patients. Some observational studies and randomized controlled trials saw no difference in mortality,^21,35,36 allograft failure,^35,36 and/or acute rejection^{21,23,34–36} for alemtuzumab and basiliximab recipients relative to rATG recipients.

In an exploratory analysis, the composite outcome of death or allograft failure or acute rejection was lower among rATG recipients versus matched pairs who received alemtuzumab. In particular, 12-month rejection rates were higher for alemtuzumab (9.9% versus 7.7%; P<0.001) and basiliximab (8.1% versus 7.1%; P<0.01) recipients relative to patients who received rATG. This finding of lower rejection with rATG adds important new information to the conclusions from the randomized trial of rATG and basiliximab by Brennan et al.¹¹ because our study was not limited to patients at high risk of early adverse outcomes. In a study of elderly KTRs using OPTN data, Gill et al. similarly observed lower risk of acute rejection for rATG versus alemtuzumab and basiliximab patients.³³ The reduced rejection associated with rATG is plausible, because this agent causes persistent depletion of T cell populations and likely targets several pathways of alloimmune activation.²³

We acknowledge some emerging evidence that alemtuzumab is associated with lower risk of acute rejection for low-risk patients versus basiliximab or rATG, but similar acute rejection risk to rATG among high-risk patients.^{7,21,23,34–38} Specifically, one randomized controlled trial found that patients who received alemtuzumab had significantly lower 6-month acute rejection risk than patients who received basiliximab.³⁸ We considered our analysis of acute rejection to be exploratory because of an earlier validation study by our group that acute rejection is reported with variable sensitivity across centers.³⁹ However, we have no reason to believe that reporting rates would correlate with induction therapy choice.

Existing literature comparing sepsis, lymphoma, and skin cancer across induction therapies is limited. The elevated risk of death or lymphoma associated with basiliximab versus rATG in our study is novel. By contrast, in a study of KTRs from 2000 to 2004, Kirk et al. used the OPTN database to suggest higher risk of post-transplant lymphoproliferative disease among rATG relative to basiliximab patients.¹ This difference could be explained in part by differences in outcome ascertainment. First, OPTN data have limited sensitivity (52%) related to malignancy outcomes, whereas we used CMS claims.⁴⁰ Second, we combined the outcomes of death or lymphoma; an examination of lymphoma alone would be problematic given higher mortality rates associated with basiliximab. From a mechanistic perspective, we hypothesize that the higher risk of rejection with basiliximab could have exposed these groups to higher long-term doses of maintenance immunosuppression, which could promote lymphoma. However, unmeasured confounding is also possible. For example, frail KTRs are at risk for several adverse outcomes including delayed graft function and mortality. It is plausible that clinicians may have preferentially selected basiliximab for recipients with more frailty or worse functional status.^11,20 Although we matched on over a dozen clinical and demographic factors, we did not have access to frailty data.^41,42

We also observed a higher risk of death or melanoma among alemtuzumab recipients relative to rATG recipients. Given our limited data on these outcomes, further studies should explore comparative post-transplant outcomes of sepsis, lymphoma, and melanoma by induction therapy.

We compared 1-year health care utilization by induction therapy and measured significantly greater 1-year costs among rATG versus alemtuzumab recipients, although the difference ($3083) was a small percentage of total costs. One-year costs were not different between basiliximab and rATG recipients. In CMS claims, we cannot disentangle payments to hospitals for induction therapy from the lump sum payments that CMS pays for the transplant hospitalization. Additionally, differences in mortality rates may influence the accumulation of costs. Yet, despite these limitations, the minor differences in costs suggest that it is reasonable to focus on clinical outcomes when selecting induction therapy.

These results should be considered in the context of their limitations. Although we matched on multiple important characteristics, unmeasured confounding remains an important possibility. While our matching techniques were robust, they can only adjust for differences among pairs that were measurable in the data. For example, as noted above, we did not have information on important characteristics that could affect outcomes such as health literacy or prior cancers.^{11,20,33,41,42} We also did not have information on induction therapy dose, which may vary considerably for rATG. Although recipients in each induction therapy group received their transplants at dozens of centers, outcomes might be confounded by variation and overall quality of care by transplant center.

In conclusion, antibody induction therapy has potent effects on the immune system with important implications for outcomes after kidney transplantation. In a large cohort of KTRs matched on diverse characteristics, rATG recipients enjoyed longer survival and generally similar or better outcomes compared with alemtuzumab and basiliximab recipients. Clinically, alemtuzumab recipients had 18% higher risk of death or allograft failure post-transplant compared with rATG recipients (estimated rates of 25.9% versus 22.9% at 5 years for alemtuzumab-rATG pairs). Basiliximab recipients had 8% higher risk of death compared with rATG recipients (estimated rates of 16.3% versus 15.2% at 5 years for basiliximab-rATG pairs). These results should help guide clinicians’ selection of induction therapy to prolong survival and avoid important complications including allograft failure after kidney transplantation.

Concise Methods

Overview

This retrospective cohort study using registry data analyzes outcomes after the administration of alemtuzumab, rATG, and basiliximab (labeled as “Campath - Alemtuzumab,” “Thymoglobulin,” and “Simulect - Basiliximab” in OPTN data) to KTRs in the United States. The study was approved by the University of Pennsylvania Institutional Review Board (protocol no. 821200).

Data Sources

We used data from the OPTN, including immunosuppression files reported from the transplant hospitalization. The OPTN data system includes data on all donors, wait-listed candidates, and transplant recipients in the United States, submitted by the members of the OPTN. The Health Resources and Services Administration, US Department of Health and Human Services, provides oversight to the activities of the OPTN contractor. These OPTN data were linked to data from the CMS. Using zip codes, we categorized participants’ neighborhoods into four groups on the basis of the percentage of residents living in poverty.⁴³

Population

The cohort comprised adults (aged 18–90 years) who received a kidney transplant between January 1, 2003, through December 31, 2008, and had CMS insurance coverage at the time of transplantation and a Medicare claim related to the transplant procedure on or near the OPTN transplant date. Exclusion criteria included multiorgan transplantation; induction therapy other than alemtuzumab, rATG, and basiliximab, or multiple induction agents; discharge on azathioprine (because of the enhanced risk of skin cancer^44,45 as well as the very rare use of azathioprine at discharge during this period [<1% of recipients]); discharge on a combination of tacrolimus, cyclosporine, and/or sirolimus (because such regimens are uncommon and might drive higher risk of sepsis and allograft failure^46–48); missing zip code (because of matching on neighborhood poverty); and body mass index (BMI) <12 or >50 kg/m².

Outcomes

The primary outcomes were death and death or allograft failure. Secondary outcomes were death or sepsis, death or lymphoma, and death or melanoma. We did not examine post-transplant complications such as sepsis alone because, from a patient’s perspective, an improvement in sepsis rates would not be important unless the combined outcome of death and sepsis were also improved. We also analyzed a composite outcome of death or allograft failure or acute rejection by 1 year post-transplant, using OPTN data. We did not use Cox regression to analyze this composite outcome because the OPTN data do not provide specific dates for acute rejection events; these events are only reported on standard forms as having occurred at the time of transplant and defined intervals afterward (e.g., 6 months and 1 year). This composite outcome was exploratory because of evidence that acute rejection is under-reported at many centers.³⁹ Finally, we prespecified subgroup analyses among elderly (≥65 years) and nonelderly patients, because of the elevated risk of infection^13,49 and reduced rejection risk^50,51 in elderly KTRs.

Death and allograft failure were determined through the OPTN dataset. Death dates in the OPTN dataset were reported by centers and via linkage to the Social Security Death Master File. The last date of follow-up for analyses of death and death or allograft failure was January of 2012. Sepsis, lymphoma, and melanoma were ascertained through CMS. The last date of follow-up for outcomes involving CMS claims was December 31, 2009, such that every patient had the opportunity for at least 1 year of follow-up for these outcomes. The Supplemental Appendix lists administrative codes used for outcomes ascertained with CMS claims.

One-year post-transplant costs were determined on the basis of resource utilization found in Medicare inpatient, outpatient, and Part B claims.^52–54 For the transplant hospitalization, the cost accounted for days in the hospital and level of care (intensive care unit versus floor),⁵³ all procedures for which a bill was charged to CMS (including operating room cost⁵⁵ and anesthesia units), and total revenue value units determined from all bills. Follow-up costs up to 1 year after the admission date of the transplant hospitalization included all costs from any subsequent hospitalization, as well as any outpatient, physician, or emergency-department bills⁵⁶ (see Supplemental Appendix). The goal was to compare healthcare resource utilization in matched pairs.^57–59 Therefore, the same cost was assigned for specific resources, such as a hospitalization day or procedure, across all areas of the United States regardless of variation in price. Using the Consumer Price Index, all 1-year costs were indexed to 2008, our last year of transplant cases. Notably, the cost of the transplant hospitalization itself is delivered by CMS as a bundled payment to the transplant program, which includes induction therapy given with transplant. Transplant hospitals are also reimbursed with a standard acquisition cost for the organ (included in our analyses). Finally, we calculated CMS payments to hospitals and physicians.

Matching on Recipient and Donor Characteristics

Two independent matches without replacement were conducted to create highly similar alemtuzumab-rATG and basiliximab-rATG pairs. We chose rATG as the reference group, because it is the most commonly used agent in kidney transplantation. We implemented a 1:1 matched-pair design because of advantages over multivariable regression approaches that were used in a number of retrospective analyses of induction therapy.^28,60,61 Multivariable regression relies upon construction of a model with a valid functional form, including sufficient attention to complex interactions between recipient and donor characteristics. In contrast, matching enables a transparent comparison of very similar groups of patients receiving different therapies and does not have the same distributional assumptions required in most regression. We completed matches before examining any outcome analyses.²⁹ We used both traditional propensity score matching and other more recently developed matching techniques to create very similar pairs. Matches were created using R MIPMatch with exact matching, fine balance,^30,32 and a distance matrix defined through the Mahalanobis distance.³¹

Recipients were matched exactly for elderly status (<65 years versus ≥65–90 years), race (black versus nonblack), previous kidney transplant (binary), HLA mismatch (zero versus nonzero), PRA category (missing, 0%–20%, 20%–80%, 80%–100%), and availability of BMI (missing versus nonmissing) and weight (missing versus nonmissing) data. We matched age in years, transplant year, neighborhood median income, BMI, and weight (if BMI was missing) as continuous variables by minimizing the total penalized Mahalanobis distance between cases and matched controls. We used near-fine balance^30,32 to ensure that the marginal distributions of the following categorical covariates were extremely similar between cases and matched controls: sex (men versus women), hepatitis C serostatus (positive versus not positive), categories of dialysis vintage (0–1, 1–3, 3–6, 6–10, and >10 years, and missing), diabetes (present or absent), cause of ESRD (specified as individual categories of GN, hypertension, polycystic kidney disease, and congenital), and donor type (live, deceased standard criteria, or deceased extended criteria [defined conventionally during the period of study as donor age ≥60 years, or age 50–59 years plus the presence of at least two of the following characteristics: donor hypertension, donor cause of death cerebrovascular accident, and donor terminal creatinine ≥1.5 mg/dl]). All of these were coded as binary covariates and included in the Mahalanobis distance. Finally, in each match we added to the distance a propensity score for receiving the focal agent (alemtuzumab or basiliximab) with scaled penalties added to minimize differences in the propensity score inside pairs.^61,62 Notably, diabetes was considered present if coded either as a comorbidity or as cause of ESRD in the OPTN data.

After careful deliberation, we decided not to match on oral immunosuppression at discharge in the primary analyses. The rationale was that induction antibody therapy has commonly been prescribed as a “bundle” of therapies, and oral immunosuppression was sometimes customized on the basis of induction therapy. For example, alemtuzumab is often used in “steroid-free” maintenance regimens.^63–65 Adjusting for prednisone use would therefore adjust away the manner in which these agents were combined in real-world practice. However, we performed supplementary analyses restricted to pairs who received tacrolimus and either mycophenolate mofetil or mycophenolate sodium, the most common oral immunosuppression regimen for KTRs. We then divided these pairs into subsets where both patients received prednisone or neither patient received prednisone and compared outcomes between induction therapy groups.

Balance for all matches between treatment and comparison groups was assessed using the Wilcoxon rank sum test for each continuous covariate and the Fisher exact test for binary covariates and using the standardized difference in means after matching.

Statistical Analyses

To compare differences between the treatment and comparison group pairs, HRs with 95% CIs were generated using the paired version of the Cox proportional hazards model.⁶⁶ Survival curves were estimated using the method of Kaplan and Meier. Pairs were considered censored after 3 years for the Medicare outcomes of sepsis, lymphoma, and melanoma because nonelderly patients commonly lose Medicare coverage after 3 years. Analyses of death and allograft failure were not censored at 3 years because these outcomes were derived using OPTN data. Person-time follow-up is included in Supplemental Table 5 of the Supplemental Appendix.

We tested for violations of the proportional hazards assumption in the Cox models by fitting additional models that included an induction therapy by time interaction. In this large study, there were small violations of proportional hazards that were statistically significant, although perhaps of negligible clinical interest. When this interaction was statistically significant, we presented both Cox regression results and also tested (see Supplemental Tables 6 and 7) the difference between the curves using the Prentice–Wilcoxon statistic⁶⁷ and reported the difference between the curves at 3 years (when using Medicare data) and 5 years (when using OPTN data only). We obtained standard errors for paired differences in Kaplan–Meier survival curves at 3 or 5 years using the bootstrap method.⁶⁸

We examined for effect modification of therapy by elderly age category using paired Cox regression.⁶⁶ We also performed subcohort analyses among nonblack patients for the skin cancer outcomes, because of higher risks of skin malignancy in this group.

The composite outcome of 1-year death, allograft failure, or acute rejection was compared using the Stuart–Maxwell test.^69,70 The McNemar test was used to compare differences in overall acute rejection rates and specific components of the composite 1-year death, allograft failure, or rejection outcome. When examining the continuous outcomes of CMS payments and 1-year costs, we used robust tests, specifically m-statistics^71–73 (Maximum Likelihood Type Estimates) and the permutational t test.^71–73 All analyses were conducted using R 2.13.1 (R Foundation) and SAS 9.3 (SAS Institute Inc).

Disclosures

J. Schold is on the Speakers Bureau for Sanofi and Novartis.