Journal Logo

Original Clinical Science—General

Immunosuppression and Graft Rejection in Living-related HLA-identical Renal Transplantation: The RADOVFULL Study

Ossman, Rim MD1; Jamme, Matthieu MD1; Moulin, Bruno MD, PhD2; Legendre, Christophe MD3; Morelon, Emmanuel MD, PhD4; Frimat, Luc MD, PhD5; Hourmant, Maryvonne MD, PhD6; Durrbach, Antoine MD, PhD7; Malvezzi, Paolo MD8; Rostaing, Lionel MD, PhD9; Luc Taupin, Jean MD, PhD10; Mesnard, Laurent MD, PhD1,11; Rondeau, Eric MD, PhD1,11; for the RADOVFULL Study Group*

Author Information
doi: 10.1097/TP.0000000000002937

Abstract

INTRODUCTION

The first kidney transplant between 2 identical twins took place on December 23, 1956, in Boston, United States. No HLA typing existed at that time. The transplanted kidney survived without maintenance immunosuppression until the death of the recipient 9 years later.1 Kidney transplants from identical twins and between siblings have better survival rates, indicating that genetics play an essential role.2 Living-related HLA-identical (LR HLAid) renal transplantation represents about 6% of all living donor renal transplantations in France. Little data are available about the immunosuppressive regimens and graft rejection among this population. Guidelines do not provide clear recommendations, and minimization of immunosuppression is the usual practice. The addition of calcineurin inhibitors (CNIs) to steroid inclusion or steroid avoidance immunosuppressive regimens in LR HLAid has shown no additional benefit for graft or patient survival in these recipients.3 However, rejection of renal transplant occurs in LR HLAid as well, with 10-year survival of 68%.4 The aim of this study is to describe the immunosuppression management and to assess the incidence of rejection and patient and graft survival, in LR HLAid transplantations in France.

MATERIAL AND METHODS

Study Design

In collaboration with the French “Agence de la Biomédecine,” we conducted a retrospective analysis of the Cristal database for adult LR HLAid renal transplantations performed at 28 centers between January 2002 and December 2012. LR HLAid renal transplantation was defined as living donor matched with HLA A, B, DR, and DQ antigens by intermediate resolution DNA typing by a Luminex Flow Analyzer with sibling recipient. A demographic database was used to study all recipients of LR HLA grafts. Authorization was obtained to extract data from patients' files at each center. Data were collected using checklists completed from patients' hospital records. Patients were categorized as having graft rejection or not.

Study Variables

Demographic characteristics at transplantation, immunosuppressive regimens, rejection rates, complications, and graft and patient survivals were analyzed.

Survival times were censored at the study end in November 2015. The main outcome was graft rejection defined by renal graft biopsy. Graft loss was defined as time from initial transplantation to dialysis, preemptive retransplantation, or recipient death at 1, 5, and 10 years.

Statistical Analysis

Continuous variables were presented as median (interquartile range) and categorical variables as counts (percentages). Univariate screening for predictors of primary outcome used the Fisher test or Student t test. Variables significantly associated with outcome in the univariate analysis were investigated in the multivariate analysis. A multivariate logistic regression model was constructed with both backward and forward variable selection procedures based on the likelihood ratio test. Goodness of fit was assessed with the Hosmer–Lemeshow test. The P values were 2-sided. The type I error was set at 0.05. Dealing with missing data was as follows: complete case analysis if <5%, multiple imputation by equation chained if between 5% and 30%, and variable excluded if >30%. In the case of multiple imputation by equation chained, data were imputed using an imputation model repeated 10 times. An analysis model was fit in each of the 50 imputed datasets separately, and these 50 datasets were, therefore, pooled and gave overall sets of estimates and corresponding standard errors. Analyses were undertaken with R 3.3.3 statistical software using the packages stepwise, mice, and rms (R foundation for Statistical Computing, Vienna, Austria).

RESULTS

Baseline and Graft Characteristics

From 2002 to 2012, 27 218 renal transplantations were performed in France, including 2475 transplantations from living donors. Among these, 163 patients had a LR HLAid donor sibling. Biopsy-proven acute rejection was diagnosed in 21 of the 163 patients (12.9%) (Figure 1).

FIGURE 1.
FIGURE 1.:
Flow chart of the study.

When compared with the 142 other patients with no rejection, the patients with rejection were younger on univariate analysis (median age, 32 y [24–45 y] versus 42 y [34–52 y]; P = 0.03). There was no significant difference regarding sex ratio (male gender, 62% versus 64.8%; P = 0.81), body mass index (BMI) (23 kg/m2 [19–25 kg/m2] versus 23 kg/m2 [21–26 kg/m2]; P = 0.5), previous transplant history (14% versus 16%; P > 0.9), initial renal disease (P = 0.83), length of dialysis (10 mo [0–17 mo] versus 13 mo [3–25 mo]; P = 0.24), or rate of preemptive transplantation (28.5% versus 24%; P = 0.59). Initial graft function was also identical in both groups (P = 0.82) (Table 1).

TABLE 1.
TABLE 1.:
Baseline patient characteristics

Immunization was defined by detection of anti-HLA antibodies by panel reactive antibodies (PRAs), ELISA, or LUMINEX since the first month of transplantation. Anti-HLA antibodies were of both I and II classes (A, B, C, DR, DQ, and DP) but were not directed against HLA alleles of the donor (donor-specific antibodies). The patients with rejection had more frequently detectable anti-HLA antibodies after transplantation (38% versus 19.7%; P = 0.03), while the rate of immunization before transplantation was similar (19% and 16.2%, respectively; P = 1.0). Immunization was not related to the timing of rejection episodes. There was no difference in either group for immunosuppressive regimen at induction or long term (Table 1). Minimization of immunosuppression (mainly CNI withdrawal, with or without mammalian target of rapamycin inhibitor [mTORi] addition) was applied in 20 patients; 9 of them were in the rejecting group and 11 in the nonrejecting group.

Concerning induction treatment, 19.6% of the whole cohort received polyclonal induction therapy, and about 38% had monoclonal induction therapy. Thirty-one percentage of the patients received no induction treatment other than high-dose steroids, and 11% received no induction therapy at all. Over the long term, 29% of patients in both groups received no CNI, and 35%–43% did not receive steroids (Table 1).

When the donors were compared, they were younger (median age, 37 y [24–43 y] versus 43 y [33–53 y]; P = 0.02) and had higher BMI (25 kg/m2 [22–30 kg/m2] versus 24 kg/m2 [22–27 kg/m2]; P = 0.02) in the rejecting group than in the nonrejecting group. Sex ratio (male gender, 33% versus 48.6%; P = 0.24) and estimated glomerular filtration rate (88 mL/min/1.73 m2 [80–99 mL/min/1.73 m2] versus 95 mL/min/1.73 m2 [78–114 mL/min/1.73 m2]; P = 0.34) were similar in both groups (Table 2).

TABLE 2.
TABLE 2.:
Donor characteristics

Acute Rejection

Rejection occurred at an average of 24 months (4–36 mo) after transplantation, in 9 of the 21 patients (42.8%) after minimization of immunosuppression (mainly CNI withdrawal, with or without mTORi addition, before rejection occurred). Minimization was applied to only 11 patients (6%) in the nonrejecting group (P < 0.001) (Table S1, SDC, http://links.lww.com/TP/B805). The pathologic pattern was primarily of a cellular type (43%; n = 9). Borderline lesions were found in 19% of patients (n = 4). Mixed cellular and antibody-mediated rejection were found in 2 patients (9.5%). Only 1 patient (4.7%) developed antibody-mediated rejection. Treatment mostly consisted of corticosteroids (CS) in 66% of cases while there was an increase in the immunosuppression regimen in 38% of patients. Plasmapheresis was performed in 2 patients (9.5%) with humoral rejection. Monoclonal anti-CD20 antibody was given to 1 patient (4.7%) with borderline lesions, and 1 patient (4.7%) with mixed cellular and humoral rejection received intravenous immunoglobulin therapy (Table 3).

TABLE 3.
TABLE 3.:
Acute rejection

Complications

The rates of cardiovascular (11%), metabolic (5%), neoplastic (5.5%), and infectious (25%) complications were also similar between the 2 groups (P = 0.28) (Table 4).

TABLE 4.
TABLE 4.:
Univariate analysis of complications after transplantation

Graft and Patient Survival

There was no difference in patient survival rates after 1 and 5 years of transplantation between the 2 groups (P > 0.9, respectively, for 1- and 5-y follow-up). No death was noted in the rejection group compared with 5 deaths in the group without rejection after 10 years of follow-up (P = 0.32). Graft loss occurred, respectively, in none, 2 (10%), and 3 (23%) patients in the rejection group and 3 (2.1%), 5 (4.6%), and 6 (14.6%) in the group without rejection, respectively, at 1, 5, and 10 years of follow-up. The differences for graft survival were not statistically significant between either group of patients (P > 0.9, P = 0.24, and P = 0.36, respectively). Graft function was evaluated according to the Modification of Diet in Renal Disease Equation. Graft function was 79, 44, and 40 mL/min/1.73 m2 for the rejection group and 67, 58, and 49 mL/min/1.73 m2 for the group without rejection (P = 0.11, P = 0.27, P = 0.39, respectively, at 1, 5, and 10 y of follow-up) (Table 5).

TABLE 5.
TABLE 5.:
Univariate analysis of overall survival and graft survival during follow-up

Multivariate Analysis of Graft Rejection

After adjustment for confounding factors, the only significant factors which remained associated with the risk of graft rejection in LR HLAid recipients were the recipient age (odds ratio [OR], 0.91 [0.84–0.96]; P = 0.003), the secondary minimization, for example, mainly CNI withdrawal, with or without mTORi addition (OR, 26.2 [5.48–166.6]; P < 0.001), and donor BMI (OR, 1.22 [1.04–1.46]; P = 0.01), with a higher BMI representing a higher risk. There was a trend for a protective effect of polyclonal antibodies at induction, but it did not reach statistical significance (Table 6). The same factors remained associated with rejection after exclusion of retransplant cases (Table 7) and patients with detectable anti-HLA antibodies before transplantation (Table 8).

TABLE 6.
TABLE 6.:
Multivariate analysis of graft rejection
TABLE 7.
TABLE 7.:
Multivariate analysis of graft rejection in patients without previous transplantation (n = 134 patients)
TABLE 8.
TABLE 8.:
Multivariate analysis of graft rejection in patients without anti-HLA antibodies before transplantation (n = 103 patients)

DISCUSSION

During the past 60 years, outcomes from kidney transplantations have improved, with graft survival being now >90% after 1 year of transplantation and 50% after 10 years of transplantation for deceased donor grafts. These advances are due to a better understanding of HLA pairing, progression in immunosuppressive therapies, and better management of cardiovascular and infectious complications.

LR HLAid renal transplantations represent a particular model in which rejection risk factors that are not linked to the HLA system may be individualized. Because of the high rate of success, little attention has been paid to the fact that not all such transplantations are successful. In fact—and despite immunosuppressive therapies—graft rejection still occurs, with rates varying between 3.2% and 55% depending on the immunosuppression regimens (azathioprine [AZA]-CS, mycophenolate mofetil-CS, CNI-AZA, CNI-CS, CNI monotherapy, and mTORi conversion protocols).5

Choice of immunosuppression has always been influenced by CNI toxicity. Reasons for withdrawing or at least minimizing CNI therapy are obvious and result from the widely recognized side effects of these drugs, such as predisposition to infection, carcinogenic properties, diabetes, hyperlipidemia, hypertension, hematologic alterations, and bone diseases.6 Gascó et al7 have demonstrated, in 6 recipients of a first kidney transplant HLAid and without anti-HLA antibodies (negative PRA by cytotoxicity testing and negative Luminex testing), that a stepwise reduction of the immunosuppressive regimen to achieve mycophenolate monotherapy might be a practical and reasonably safe approach compatible with a sufficient level of immunosuppression. A comparison of AZA- and cyclosporine-treated recipients of renal transplants from HLA-identical siblings has demonstrated no difference in outcomes.8 Walker et al5 treated 20 LR HLAid recipients without steroids and tapered immunosuppression to achieve mycophenolate monotherapy after 1 year. No safety concerns were raised after 18 months of follow-up. A French study reported excellent results in 7 cases of LR HLAid receiving mycophenolate (6 patients) or rapamycin (1 patient) monotherapy after antithymocyte globulin induction.9 Another study from the University of Minnesota reported high rates of chronic allograft nephropathy in patients receiving CNI but no difference in graft and patient survival in patients without CNI. In this study, patients from the era before the introduction of cyclosporine were compared with those who received CNI.3 However, few studies on induction avoidance in these patients were published. The largest study performed in the United States showed excellent graft and patient outcomes even with induction avoidance and CNI withdrawal, with rejection rates of 4%.10

Dziewanowski et al11 showed that, even in cases of homozygous twin transplantations, selection of the immunosuppressive therapy needs to be carefully individualized to take into account transplantation history, primary disease of the recipient, and the level of immunization of the patient. In one center that included 108 HLAid siblings, a total of 66 acute rejection episodes were noted in 50 recipients. The incidence of acute rejection was not significantly increased in the sensitized compared with the nonsensitized group. The authors noted that graft survival was significantly better in the group of patients without a history of rejection. Immunosuppressive treatment consisted of a double therapy of prednisone and AZA, with only 10 patients treated with cyclosporine.12 Keitel et al13 revealed an increased incidence of acute rejection episodes among the AZA plus prednisone therapy versus cyclosporine plus prednisone therapy patients in a retrospective analysis of 67 cases. An everolimus-based regimen with planned withdrawal of CNI therapy was not effective for LR HLAid because of a high incidence of acute rejection (15.4%), with all episodes occurring after discontinuation of CNI treatment.14 In another study, LR HLAid recipients had a high incidence of acute rejection (32%) when there was no initial CNI immunosuppression.15

Based on this review of the literature, the clinician cannot be certain that the decision to withdraw immunosuppressive therapy will not result in transplanted kidney rejection. These reference data remain inconclusive with respect to a practical management of such patients. Our study is the first multicenter national study to describe 10 years of transplantations follow-ups in LR HLAid recipients in France. It presents more concrete information about the demographics, characteristics, overall survival of patients and grafts, and the incidence of rejection among this particular group of patients.

We showed that 10-year graft survival was as high as 77% in the rejection group and 85.4% in the nonrejecting group. These data are concordant with the available literature. Acute rejection did not affect graft survival when both groups were compared, unlike the results of de Mattos et al12 in their study on LR HLAid patients. Unexpectedly, the incidence of rejection was high (12.9%) and was independent of prior or later immunization. Overall, patients received induction less frequently and were treated less than usually with intensive immunosuppressive regimens. In fact, 59.6% of the patients received induction therapy other than high-dose steroids and 11% received no induction therapy. Only 72% received CNIs and 64.4% long-term CS. In 9 of the 21 patients (42.8%), secondary immunosuppression was minimized (CNI withdrawal mainly, with or without mTORi addition) before the occurrence of rejection. There was no recurrence of original disease despite less intense immunosuppression.

The multivariate regression analysis revealed that recipient age is a factor associated with rejection. This result was described in an Australian study that aimed to compare the rate and severity of rejection between patients receiving kidney transplants from living versus deceased donors. It showed that live-donor kidney transplant recipients had a higher rate of, and more severe, rejection. Recipient age proved to be the unique demographic factor that differed between both groups since live-donor kidney transplant recipients were younger,16 suggesting a role of patient noncompliance. However, the recipient age was found to be linked inversely in other studies.17

Donor BMI was also associated with rejection. Donor BMI correlated with the incidence of delayed graft function18 and with higher rejection rates in recipients with a lower BMI category than their donors.19

Our data showed the occurrence of a variety of complications (cardiovascular, metabolic, neoplastic, and infectious). We found that, although a majority of these events were related to immunosuppression, their incidence was relatively lower than commonly known and described elsewhere.20 We believe this is primarily due to the lighter immunosuppression regimen applied to the patients of our study.

In our study, HLA typing was performed at a low-resolution level and in most cases did not include C and DP loci. However, giving that donors and recipients were siblings, we assume that HLA identity will also occur at the high-resolution level for A, B, DR, and DQ, which were always typed. For C, the strong linkage disequilibrium with B will also lead to identity at the low- and high-resolution levels.21 For DP, which is less strongly associated with the other class II genes, low- and high-resolution mismatches will seldomly occur. However, overall, the epitopes mismatch load will remain very low, being supported only by the DP locus, whose role as an antigenic target is not consensual. More than the role of epitope load, donor-specific antibody development is most probably more correlated to qualitative characteristics of given more immunogenic epitopes, to date not retrieved on DP molecules.22

In our study, the patients with rejection had more frequently detectable anti-HLA antibodies after transplantation. Detection of anti-HLA antibodies by cytotoxic PRA was not systematic because it was not reproducible between centers, as it depended on local cytotoxic cell panels of quite heterogenous phenotype distribution. The degree of immunization was progressively replaced with the identification of positive antigens with Luminex. The presence of antigens other than HLA capable of inducing rejection was first described in 1978. The authors insisted on immunosuppressive therapies in this type of transplantation because “the responses to non-HLA antigens could be significant in untreated cases.”23 Non-HLA immunity has probably a stronger impact in clinical transplantation than previously thought and is associated with chronic graft loss.24 The effect of minor H antigen mismatching which was studied in 444 patients did not seem to be correlated with the risk of rejection and graft loss in HLA-identical renal transplantations.25 Numerous other non-HLA antibodies have been identified in renal organ transplantation, directed against a heterogeneous subset of both allo- and autoantigens, including the HLA class-I–related chain A MICA alloproteins and the angiotensin II type 1 receptor.26 A recent genome-wide analysis was done in a prospective cohort of 477 pairs of deceased donors and first kidney transplant recipients. The authors of this study concluded that genetic mismatch of non-HLA haplotypes is associated with an increased risk of functional graft loss and that donor-specific antibodies could be identified.27 Steers et al28 showed that a genomic collision at LIM and senescent cell antigen-like domains 1 locus affected the risk of rejection of the kidney allografts with production of anti-LIM and senescent cell antigen-like domains 1 IgG2 and IgG3. We believe that in LR HLAid patients with low-immunologic phenotype risk, a genomic analysis with whole exome sequencing may help to identify the non-HLA antigens that mediate the rejection.29

In the particular cases of twins, and because of intrauterine fetus growth and DNA mutation which results in phenotypic and genotypic divergence even in homozygous twins, zygosity testing has been proposed to decide which immunosuppression protocol the patient should receive.30

Our study has several limitations. First, due to the retrospective design, we are not able to confirm the causality of the association we observed. Second, we have not been able to assess the incidence of biopsy-proven chronic graft nephropathy due to the lack of protocol biopsies available in many centers. The causes of death and graft loss could also not be assessed. Third, in cohort with time to event data, survival analysis methods are more adapted than logistic regression but were impossible to apply in our cohort due to the absence of dates of event.

CONCLUSIONS

Despite full HLA matching, the physician should know that his or her decision to use less intensive treatment in LR HLAid may result in acute rejection. We suggest that minimization of immunosuppression should be done with caution, especially in young recipients. Genomic analysis may help to identify the tissue antigens, other than HLA, that drive the rejection process in these cases.

ACKNOWLEDGMENTS

The authors sincerely thank Felicity Kay for her help in preparing the article. The authors performed this study with the collaboration of the French “Agence de la Biomédecine.”

REFERENCES

1. Merrill JP, Murray JE, Harrison JH, et al. Successful homotransplantation of the human kidney between identical twins. J Am Med Assoc. 1956; 160:277–282
2. Sayegh MH, Carpenter CB. Transplantation 50 years later–progress, challenges, and promises. N Engl J Med. 2004; 351:2761–2766
3. Verghese PS, Dunn TB, Chinnakotla S, et al. Calcineurin inhibitors in HLA-identical living related donor kidney transplantation. Nephrol Dial Transplant. 2014; 29:209–218
4. Cecka JM. The OPTN/UNOS Renal Transplant Registry. Clin Transpl. 20051–16
5. Walker JK, Alloway RR, Roy-Chaudhury P, et al. A prospective trial of a steroid-free/calcineurin inhibitor minimization regimen in human leukocyte antigen (HLA)-identical live donor renal transplantation. Transplantation. 2009; 87:408–414
6. Shimmura H, Tanabe K, Ishida H, et al. Long-term results of living kidney transplantation from HLA-identical sibling donors under calcineurin inhibitor immunosuppression. Int J Urol. 2006; 13:502–508
7. Gascó B, Revuelta I, Sánchez-Escuredo A, et al. Long-term mycophenolate monotherapy in human leukocyte antigen (HLA)-identical living-donor kidney transplantation. Transplant Res. 2014; 3:4
8. Gill IS, Hodge EE, Novick AC, et al. Azathioprine vs cyclosporine in recipients of HLA-identical renal allografts. Cleve Clin J Med. 1994; 61:206–210
9. Venot M, Abboud I, Duboust A, et al. Calcineurin inhibitor-free monotherapy in human leukocyte antigen–identical live donor renal transplantation. Transplantation. 2011; 91:330–333
10. Brifkani Z, Brennan DC, Lentine KL, et al. The privilege of induction avoidance and calcineurin inhibitors withdrawal in 2 haplotype HLA matched white kidney transplantation. Transplant Direct. 2017; 3:e133
11. Dziewanowski K, Drozd R, Chojnowska A, et al. Kidney transplantation among identical twins: therapeutic dilemmas. BMJ Case Rep. 2011; 2011:bcr0120113752
12. de Mattos AM, Bennett WM, Barry JM, et al. HLA-identical sibling renal transplantation–a 21-yr single-center experience. Clin Transplant. 1999; 13:158–167
13. Keitel E, Santos AF, Alves MA, et al. Immunosuppression protocols for HLA identical renal transplant recipients. Transplant Proc. 2003; 35:1074–1075
14. Cristelli MP, Ferreira A, Hannun P, et al. De novo everolimus for recipients of kidney transplants from HLA identical donors. J Bras Nefrol. 2016; 38:225–233
15. Peddi VR, Weiskittel P, Alexander JW, et al. HLA-identical renal transplant recipients: immunosuppression, long-term complications, and survival. Transplant Proc. 2001; 33:3411–3413
16. Campbell SB, Hothersall E, Preston J, et al. Frequency and severity of acute rejection in live- versus cadaveric-donor renal transplants. Transplantation. 2003; 76:1452–1457
17. Boran M, Boran M, Boran E. HLA-identical sibling renal transplantation: influence of donor and recipient gender mismatch on long-term outcomes. Transplant Proc. 2014; 46:3423–3425
18. Weissenbacher A, Jara M, Ulmer H, et al. Recipient and donor body mass index as important risk factors for delayed kidney graft function. Transplantation. 2012; 93:524–529
19. Wang HH, Lin KJ, Liu KL, et al. Size does matter-donor-to-recipient body mass index difference may affect renal graft outcome. Transplant Proc. 2012; 44:267–269
20. Legendre C. La Transplantation Rénale. 2011Cachan, FranceLavoisier
21. Sanchez-Mazas A, Djoulah S, Busson M, et al. A linkage disequilibrium map of the MHC region based on the analysis of 14 loci haplotypes in 50 French families. Eur J Hum Genet. 2000; 8:33–41
22. Snanoudj R, Kamar N, Cassuto E, et al. Epitope load identifies kidney transplant recipients at risk of allosensitization following minimization of immunosuppression. Kidney Int. 2019; 95:1471–1485
23. Carpenter CB. Transplant rejection in HLA-identical recipients. Kidney Int. 1978; 14:283–291
24. Opelz G; Collaborative Transplant StudyNon-HLA transplantation immunity revealed by lymphocytotoxic antibodies. Lancet. 2005; 365:1570–1576
25. Dierselhuis MP, Spierings E, Drabbels J, et al. Minor H antigen matches and mismatches are equally distributed among recipients with or without complications after HLA identical sibling renal transplantation. Tissue Antigens. 2013; 82:312–316
26. Michielsen LA, van Zuilen AD, Krebber MM, et al. Clinical value of non-HLA antibodies in kidney transplantation: still an enigma? Transplant Rev (Orlando). 2016; 30:195–202
27. Reindl-Schwaighofer R, Heinzel A, Kainz A, et al.; iGeneTRAiN ConsortiumContribution of non-HLA incompatibility between donor and recipient to kidney allograft survival: genome-wide analysis in a prospective cohort. Lancet. 2019; 393:910–917
28. Steers NJ, Li Y, Drace Z, et al. Genomic mismatch at LIMS1 locus and kidney allograft rejection. N Engl J Med. 2019; 380:1918–1928
29. Mesnard L, Muthukumar T, Burbach M, et al. Exome sequencing and prediction of long-term kidney allograft function. Plos Comput Biol. 2016; 12:e1005088
30. Kessaris N, Mukherjee D, Chandak P, et al. Renal transplantation in identical twins in United States and United Kingdom. Transplantation. 2008; 86:1572–1577

Supplemental Digital Content

Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.