Impact of polyclonal antithymocyte globulins on kidney transplant outcomes in real-life: a propensity score analysis

Background: Randomized studies reported a marginal superiority of polyclonal antithymocyte globulins (ATG, Thymoglobulin®, Sano, Gentilly, France, or Fresenius®, Bad Homburg, Germany) to prevent acute rejection compared to monoclonal anti-CD25 antibodies (IL2Ra). Nevertheless, the representativeness and the generalizability of these studies are questionable. Methods: We studied the impact of ATG use in real-life conditions in a multicenter study. Propensity score analysis was performed to address potential confounding by indication. Results: 817 patients were included. Logistic regression revealed that age, male gender, a pre-transplant history of cancer, presence of anti-HLA antibodies, previous kidney transplantation, and transplant center were associated with ATG use. The area under the curve of the propensity score was 0.84 + 0.02. ATG use was not associated with a lower rate of acute rejection (18.2% in ATG-treated patients vs 15.8% in non-ATG-treated patients, p =0.356). Adjustment for propensity score slightly modied the relationship between ATG and acute rejection towards a more neutral effect ( p =0.913). Score match analysis recapitulated the previous result. ATG use was associated with the occurrence of opportunistic infection ( p =0.034). There was no difference in graft loss or death between the two groups. Conclusions: In real-life conditions, ATG does not substantially reduce the risk of acute rejection after kidney transplantation. A better discrimination of patients who may benet from ATG is required.


Background
Even when randomized studies are the gold-standard for the evaluation of therapeutic strategies, they are not exempt of potential bias. Indeed, their interpretation has to take into account whether the population is representative of the one receiving the treatment in real life conditions. Moreover, a priori selection of patients by investigators could exclude a fraction of the population meeting inclusion criteria but judged not subject to randomization. Those estimated having the greatest bene t or the highest risks with any treatment arm are likely to be excluded. Additionally, the strict and controlled conditions in which they are conducted, leads to low generalizability because they are performed in very different conditions from real life usual care. Such pitfalls have been identi ed in different studies (1). This bias is likely concerning evidence-based data concerning ATG. First, ATG use is highly variable from a center to center both as regards the indication and the dose. Second, randomized studies are likely to miss patients with very high immunological risk as well as those carrying severe infectious or neoplastic risks. Another point to consider is time-change modi cations of patients' phenotypes and associated treatments that may affect the relevance of past studies. Thus, even when randomized trials are the gold standard, they do have some inconveniences that can be clari ed by other methodologic approaches using real-life data.
Two randomized studies reported similar results suggesting that ATG signi cantly reduced acute rejection risks in high risk patients (2)(3)(4). By contrast, the superiority of ATG in low-to-moderate immunological risk patients is more questionable. A large meta-analysis found no bene t of ATG as compared to IL2Ra regarding clinically diagnosed acute rejection and a marginal one concerning biopsyproven acute rejection (5). Moreover, whereas the rate of acute rejection is not reduced by ATG use in those patients, infections are more frequent leading to a worse bene t/risk ratio (6)(7)(8). Even when these studies have highlighted preferential indications for ATG, its use is highly center-dependent. As a consequence, the global bene t of ATG in real life is more speculative.
We studied the impact of ATG use in real-life conditions in a multicenter study. Propensity score analysis was performed to address potential confounding by indication.

Study design and populations
This study has been conducted in the rst 817 consecutive renal transplant recipients (RTR) from the ORLY-EST trial. ORLY-EST is a prospective cohort study including incident RTR in seven French transplant centers (Besançon, Clermont-Ferrand, Dijon, Kremlin-Bicêtre, Nancy, Reims, Strasbourg). The main objective of this study is to assess associations between immune status and post-transplant complications. The ethic committee of Franche-Comté study has approved the study (2008). Patients enrolled in the ORLY-EST study gave their written informed consent.
Each center uses the anti-Cytomegalovirus (CMV) prevention of their choice. A majority of patients with previous CMV exposition received valganciclovir for 3 months. CMV-naïve patients who received a CMV positive kidney were given valganciclovir for 3 or 6 months. All patients received Pneumocystis antimicrobial prophylaxis with trimethoprim-sulfamethoxazole for at least 6 months.

Confounding factors
Demographic parameters and medical history were assessed. Dialysis modality and its duration were also recorded. HLA mismatches were registered as well as other relevant immunological parameters such as pre-transplant panel reactive antibodies (PRA) (0 vs. positive PRA at any level) and transplant rank ( rst vs. second or more). Delayed graft function de ned as the need for dialysis in the rst week posttransplant was recorded. Methods of assessment and de nitions of these variables have been previously described in detail (9).

Acute rejection
Only biopsy-proven acute rejections were retained. Acute rejection was de ned according to the Banff classi cation (10).

Severe infections
Methods to assess infections have been previously described (9).
Brie y, severe bacterial infections or opportunistic infections where only considered if subsequent hospitalization was required. BK infections were not recorded.
CMV disease CMV disease was de ned by the need of treatment in a patient with viral replication.

Statistical analysis
Median, mean values with standard error, and frequency were provided for continuous and categorical variables, respectively. Medians, means, and proportions were compared using Student's t test and chisquare test (or Fisher's exact test, if appropriate), respectively. Acute rejection free survival was estimated using the Kaplan-Meier method. Follow-up duration was calculated using a reverse Kaplan-Meier estimation. Cox proportional hazard models were performed to estimate hazard ratio (HR) and 95%CI for factors associated with acute rejection.
The association of baseline parameters with acute rejection was rst assessed using univariate Cox analyses. Then parameters with p values of less than 0.20 were entered into a nal multivariable Cox regression model, after considering collinearity among variables with a correlation matrix. Similar analyses were performed for infections.
In a nonrandomized study, a bias of treatment attribution is always present. Differences between the two groups due to ATG indication may greatly in uence the outcomes. To face these constraints, we calculated propensity scores to correctly adjust for confounding factors and to address potential confounding by indication (11).
We assessed the propensity score of receiving ATG through multivariable logistic regression model with ATG use as the dependent variable. Data known to in uence the choice of using or not ATG were used to calculate the propensity score and included age, pretransplant history of malignancy, duration of dialysis before transplant, pre-transplant anti-HLA antibodies, previous history of transplantation, Other factors such as sex, previous cardiovascular disease, hypercholesterolemia, hypertension, body mass index, smoking status, and diabetes, were also tested in the model. We matched one ATG-treated patient to one non-ATG-treated patient by their propensity score + 0.02 (0.10SD of the propensity score) to generate a sub-cohort of 154 patients who were treated with ATG and 154 patients who were not.

Study population
Characteristics of the study population were depicted in table 1.
Propensity score matching successfully balanced the distribution of characteristics across patients who were treated with ATG compared with those who were not (

Death and graft loss
Twenty-four patients (3%) died in the rst year post-transplant. There was no difference in the two groups (3% in ATG-treated patients and 2.9% in those having received IL2Ra; p=0.924).
Thirty-eight patients (4.7%) lost their graft in the rst year post-transplant. There was no difference in graft loss between the two groups (4.5% in ATG-treated patients and 4.7% in those having received IL2Ra; p=0.908).

Discussion
T cell depletion by polyclonal anti-thymocyte globulins (ATG) has been used for many years for immunosuppression in organ transplantation. These products are composed of polyclonal antibodies mainly directed against T lymphocytes, but also against other types of immune and non-immune cells, such as stromal thymic cells (12, 13). Although they have been widely used for over 30 years, few evidence-based guidelines for ATG use are available and prescription remains largely empirical. A substantial bene t of ATG has been described only in immunological high-risk patients. However, a signi cant part of ATG-treated patients included in this study could not be classi ed as immunological high-risk patients according to traditional de nitions. One in ve patients having received ATG was the recipient of a rst transplant and had no HLA sensitization. However, 12% of HLA-sensitized recipients of a second transplant received IL2Ra induction.
We found some factors, usually not used as inclusion or exclusion criteria in randomized studies comparing ATG and a-CD25, to signi cantly in uence the choice of induction treatment. A pre-transplant history of cancer was strongly associated with avoidance of ATG. A large meta-analysis reported an increased incidence of cancer in ATG-treated patients as compared to those having received IL2Ra (4). Our group also emphasized the impact of ATG-induced persistent CD4 lymphopenia on post-transplant cancer occurrence (14,15). It is likely that, not only an immunological risk, but also a putative enhanced risk of cancer participates in the nal choice of the type of induction. While only 7% of patients were 70 years old or over, older age signi cantly prevents ATG use. However, cut off was 70 or 75 years old in the different randomized studies comparing ATG to IL2Ra (5). A supposed frailty of immune responses in elderly probably explains this association. Finally, while no data supports a speci c bene t of ATG in female, male were less prone to receive ATG than female. Apart from these parameters, an unexplainable proportion of the decision is center-dependent and re ects individual practice. Furthermore, the use of ATG was not associated with any different strategy of maintenance immunosuppression. Scheduled corticosteroid withdrawal was similar in the two treatments groups as well as the use of Tacrolimus or Mycophenolic acid. This suggests that the choice to use ATG is based on both evidence-and belief-based criteria or that evidence-based data are not su ciently convincing to bring together physicians' opinions.
Our study suggests that, in real life conditions, ATG use is not associated with reduced rates of acute rejection. This result was observed in all sub-groups of patients, including those who have been identi ed as having greater bene ts in randomized studies (2, 3). Because the majority of patients included in this study were at low-to-intermediate immunological risk, a lack of statistical power may explain this result.
Nevertheless, no trends were observed suggesting that enhancing power would not change the results. Some explanations are possible. Randomized studies have been published almost ten years ago and more. Better determination of HLA antibodies and generalization of ow cytometry crossmatch may have improved donor-recipient compatibility and reduce the risk of acute rejection. As a consequence, the bene cial effects of ATG may be reduced. According to this hypothesis, HLA sensitization was not associated with acute rejection even in patients with PRA>30%. The two only predictive factors of acute rejection were a previous history of kidney transplantation and delayed graft function. We observed a non-signi cant reduction of 19% in the risk of acute rejection after delayed graft function in ATG-treated patients. As a result, a randomized study should include 778 patients to demonstrate such a signi cant difference with a power of 80% and an alpha risk of 5%. Our study highly suggests that ATG did not reduce the rate of acute rejection in current conditions of utilization.
Even when CMV disease was more frequent in ATG-treated patients, the association did not persist after adjustment for confounding factors. Furthermore, due to different center practices, CMV prophylaxis was surprisingly less frequent in ATG-treated patients. According to propensity score analysis, this con rms that CMV status did not in uence the choice to use ATG. Under systematic CMV prophylaxis, the rate of CMV disease are relatively similar in ATG-treated and non-ATG-treated patients (2,8). By contrast, in the absence of CMV prophylaxis, CMV disease was probably much more frequent in patients having received ATG (5,16). No difference in the incidence of severe bacterial infection was observed between the two arm groups. Such a result is consistent with previous reports (3). However, we observed more frequent opportunistic infections in ATG-treated patients. This result con rms preliminary analyses performed earlier in this study (17). Overall, ATG is not associated with a substantial increase in early infectious complications except for opportunistic infections. Furthermore, both patient and graft survivals were similar in the two treatment groups.
Our study has some limitations. A bias of indication is inherent to non-randomized comparisons of treatment. We used propensity score analysis to overcome this limitation. Patients who received ATG differed in many ways from those who did not. Propensity score analysis takes these differences into account and minimizes potential confounding by indication. After adjustment for propensity score, we did not observe any difference in most post-transplant outcomes between patients having or not having received ATG. Registry studies have the potential to include a larger number of patients, but often lack accurate confounding factors analysis. Our study is likely to be a fair compromise between less precise observational registers and unfeasible randomized trials. Due to the prospective design of the study, missing data are low (<5%).

Conclusion
Our study suggests that, in real-life conditions, ATG does not substantially reduce the risk of acute rejection after kidney transplantation. By contrast, ATG-treated patients exhibited an increased risk of opportunistic infections. A better discrimination of patients who may bene t of ATG is required.

Declaration
Ethics approval and consent to participate Sample collection was performed after regulatory approval by the French ministry of health (agreement number # DC-2008-713, June 11 th 2009). The ethic committee of Franche-Comté study has approved the study (2008). Patients enrolled in the ORLY-EST study gave their written informed consent. Clinical data were prospectively collected.

Not applicable
Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interest
The authors have no competing interest Funding organizations had no access on the design of the study, and collection, analysis, and interpretation of data    Figure 1