Causal inference in case of near-violation of positivity: comparison of methods.

In causal studies, the near-violation of the positivity may occur by chance, because of sample-to-sample fluctuation despite the theoretical veracity of the positivity assumption in the population. It may mostly happen when the exposure prevalence is low or when the sample size is small. We aimed to compare the robustness of g-computation (GC), inverse probability weighting (IPW), truncated IPW, targeted maximum likelihood estimation (TMLE), and truncated TMLE in this situation, using simulations and one real application. We also tested different extrapolation situations for the sub-group with a positivity violation. The results illustrated that the near-violation of the positivity impacted all methods. We demonstrated the robustness of GC and TMLE-based methods. Truncation helped in limiting the bias in near-violation situations, but at the cost of bias in normal conditions. The application illustrated the variability of the results between the methods and the importance of choosing the most appropriate one. In conclusion, compared to propensity score-based methods, methods based on outcome regression should be preferred when suspecting near-violation of the positivity assumption.

Lack of impact of pre-emptive deceased-donor kidney transplantation on graft outcomes: a propensity score-based study.

BACKGROUND: A significant number of studies have compared graft outcomes between patients with Pre-emptive Kidney Transplantation (PreKT) and patients on Dialysis before their Kidney Transplantation (DiaKT). These studies have suffered from the limitation that the DiaKT group is composed of all the dialysed patients, including those placed on a waiting list at the time of their first dialysis session. This seriously questions the comparability of these patients with those placed on the waiting list a long time before the need for renal replacement therapy. The aim of this study was to precisely evaluate the causal effect of PreKT from deceased donors. METHODS: Data were extracted from the multicentric French DIVAT (Données Informatisées et VAlidées en Transplantation) cohort. The DiaKT group was composed of patients placed on the waiting list with an initial intention of pre-emptive transplantation. Cause-specific Cox models with propensity scores (inverse probability weighting) were used to study the patient and graft outcomes. RESULTS: Among the 1138 included patients, 554 patients were in the PreKT group. The outcomes of the PreKT group were similar compared with the DiaKT group. In particular, the life expectancy with a functioning graft was 8.51 years [95% confidence interval (CI) 8.20-8.81] for the PreKT recipients versus 8.49 years (95% CI 8.15-8.84) for the DiaKT recipients. CONCLUSIONS: Our results challenge the utility of PreKTs from deceased donors, especially with regard to the consequential increase in the waiting list.

Comparison of graft and patient survival according to the transplantation centre policy for 1-year screening biopsy among stable kidney recipients: a propensity score-based study.

BACKGROUND: The clinical utility of screening biopsies (SBs) at 1 year post-transplantation is still debated, especially for stable kidney graft recipients. Given the heterogeneity in practices between transplantation centres, the objective of this study was to compare graft and patient survival of stable patients according to whether they were followed up in a transplantation centre with or without a policy for having an SB at 1 year post-transplantation. MATERIALS: From a French multicentre cohort, we studied 1573 kidney recipients who were alive with stable graft function at 1 year post-transplantation, with no acute rejection in their first year post-transplantation. RESULTS: Using propensity score-based analyses, we did not observe any significant difference in the relative risk for graft failure between patients from centres with a 1-year SB policy and those from other centres [hazard ratio = 1.15, 95% confidence interval (CI) 0.86-1.53]. The corresponding adjusted survival probability at 8 years post-transplantation was 69% (95% CI 61-74%) for patients from centres with a 1-year SB policy versus 74% (95% CI 67-79%) for those from other centres. CONCLUSION: A 1-year SB policy for stable patients may not lead to therapeutical benefits for improved graft and patient survival. Further studies examining the benefits versus the risks of a 1-year SB policy are warranted to demonstrate the long-term utility of this intervention.

Inverse probability weighting to control confounding in an illness-death model for interval-censored data.

Multistate models with interval-censored data, such as the illness-death model, are still not used to any considerable extent in medical research regardless of the significant literature demonstrating their advantages compared to usual survival models. Possible explanations are their uncommon availability in classical statistical software or, when they are available, by the limitations related to multivariable modelling to take confounding into consideration. In this paper, we propose a strategy based on propensity scores that allows population causal effects to be estimated: the inverse probability weighting in the illness semi-Markov model with interval-censored data. Using simulated data, we validated the performances of the proposed approach. We also illustrated the usefulness of the method by an application aiming to evaluate the relationship between the inadequate size of an aortic bioprosthesis and its degeneration or/and patient death. We have updated the R package multistate to facilitate the future use of this method.

Is pre-transplant sensitization against angiotensin II type 1 receptor still a risk factor of graft and patient outcome in kidney transplantation in the anti-HLA Luminex era? A retrospective study.

We aimed to assess the correlation of anti-angiotensin II type 1 receptor antibodies (anti-AT1R-Abs) before transplantation on a multicentric cohort of kidney transplant recipients (2008-2012), under tacrolimus and mycophenolate mofetil (MMF), screened by Luminex technology for anti-HLA immunization. Anti-AT1R antibody levels were measured by ELISA in pretransplantation sera of 940 kidney recipients from three French centers of the DIVAT cohort. Multivariable Cox models estimated the association between pretransplant anti-angiotensin II type 1 receptor antibodies and time to acute rejection episodes (ARE) or time to graft failure. Within our cohort, 387 patients (41.2%) had pretransplant AT1R-Abs higher than 10 U/ml and only 8% (72/970) greater than 17 U/ml. The cumulative probability of clinically relevant (cr)-ARE was 22.5% at 1 year post-transplantation [95% CI (19.9-25.4%)]. The cumulative probability of graft failure and patient death were 10.6% [95% CI (8.4-13.3%)] and 5.7% [95% CI (4.0-8.1%)] at 3 years post-transplantation, respectively. Multivariate Cox models indicated that pretransplant anti-AT1R antibody levels higher than 10 U/ml were not significantly independently associated with higher risks of acute rejection episodes [HR = 1.04, 95% CI (0.80-1.35)] nor with risk of graft failure [HR = 0.86, 95% CI (0.56-1.33)]. Our study did not confirm an association between pretransplant anti-AT1R antibody levels and kidney transplant outcomes.

Comparisons of the performance of different statistical tests for time-to-event analysis with confounding factors: practical illustrations in kidney transplantation.

Confounding factors are commonly encountered in observational studies. Several confounder-adjusted tests to compare survival between differently exposed subjects were proposed. However, only few studies have compared their performances regarding type I error rates, and no study exists evaluating their type II error rates. In this paper, we performed a comparative simulation study based on two different applications in kidney transplantation research. Our results showed that the propensity score-based inverse probability weighting (IPW) log-rank test proposed by Xie and Liu (2005) can be recommended as a first descriptive approach as it provides adjusted survival curves and has acceptable type I and II error rates. Even better performance was observed for the Wald test of the parameter corresponding to the exposure variable in a multivariable-adjusted Cox model. This last result is of primary interest regarding the exponentially increasing use of propensity score-based methods in the literature.

A useful scoring system for the prediction and management of delayed graft function following kidney transplantation from cadaveric donors.

Delayed graft function (DGF) is a common complication in kidney transplantation and is known to be correlated with short- and long-term graft outcomes. Here we explored the possibility of developing a simple tool that could predict with good confidence the occurrence of DGF and could be helpful in current clinical practice. We built a score, tentatively called DGFS, from a French multicenter and prospective cohort of 1844 adult recipients of deceased donor kidneys collected since 2007, and computerized in the Données Informatisées et VAlidées en Transplantation databank. Only five explicative variables (cold ischemia time, donor age, donor serum creatinine, recipient body mass index, and induction therapy) contributed significantly to the DGF prediction. These were associated with a good predictive capacity (area under the ROC curve at 0.73). The DGFS calculation is facilitated by an application available on smartphones, tablets, or computers at www.divat.fr/en/online-calculators/dgfs. The DGFS should allow the simple classification of patients according to their DGF risk at the time of transplantation, and thus allow tailored-specific management or therapeutic strategies.

The role of donor hypertension and angiotensin II in the occurrence of early pancreas allograft thrombosis.

Background: About 10-20% of pancreas allografts are still lost in the early postoperative period despite the identification of numerous detrimental risk factors that correlate with graft thrombosis. Methods: We conducted a multicenter study including 899 pancreas transplant recipients between 2000 and 2018. Early pancreas failure due to complete thrombosis, long-term pancreas, kidney and patient survivals were analyzed and adjusted to donor, recipient and perioperative variables using a multivariate cause-specific Cox model stratified to transplant centers. Results: Pancreas from donors with history of hypertension (6.7%), as well as with high body mass index (BMI), were independently associated with an increased risk of pancreas failure within the first 30 post-operative days (respectively, HR= 2.57, 95% CI from 1.35 to 4.89 and HR= 1.11, 95% CI from 1.04 to 1.19). Interaction term between hypertension and BMI was negative. Donor hypertension also impacted long-term pancreas survival (HR= 1.88, 95% CI from 1.13 to 3.12). However, when pancreas survival was calculated after the postoperative day 30, donor hypertension was no longer a significant risk factor (HR= 1.22, 95% CI from 0.47 to 3.15). A lower pancreas survival was observed in patients receiving a pancreas from a hypertensive donor without RAAS (Renin Angiotensin Aldosterone System) blockers compared to others (50% vs 14%, p < 0.001). Pancreas survival was similar among non-hypertensive donors and hypertensive ones under RAAS blockers. Conclusion: Donor hypertension was a significant and independent risk factor of pancreas failure. The well-known pathogenic role of renin-angiotensin-aldosterone system seems to be involved in the genesis of this immediate graft failure.

Impact of Late and Recurrent Acute Graft Pyelonephritis on Long-Term Kidney Graft Outcomes.

Background: While Urinary tract infections are the most common infections in kidney transplant recipients, the impact of late acute graft pyelonephritis (AGPN) on graft outcomes remains unknown. Our study was performed to more precisely evaluate the long-term impact of AGPN. Methods: We included 9052 kidney and combined kidney-pancreas recipients who underwent transplantation between 2008 and 2018 from a French multicenter cohort. The relationships between AGPN and patient and graft survival were analyzed with a time-dependent multivariate Cox model. Results: The cumulative incidence of AGPN was 20.9%. A first episode of early AGPN is associated with a non-significant increase in the risk of graft failure (hazard ratio [HR], 1.27; 95% confidence interval [95% CI], 0.90 to 1.79). Though, cumulative number of AGPN episodes (HR = 1.51; 95% CI, 0.89 to 2.57 for two episodes and HR = 2.08; 95% CI, 1.17 to 3.69 for three or more episodes) is associated with an increased risk of graft failure. In contrast, when the first episode of AGPN occurred late (i.e., 6 months post transplantation), the risk of graft failure is significantly increased (HR = 2.25; 95% CI, 1.65 to 3.07), and this risk remains relatively stable with the recurrence of late AGPN episodes. The onset of late AGPN were also associated with a higher risk of patient death. Conclusion: This analysis shows that late AGPN and recurrent AGPN are both risk factors for a poor long-term graft outcome and mortality. Late AGPN should not be considered benign infections in post-transplantation follow-up.

Antibody-mediated allograft rejection is associated with an increase in peripheral differentiated CD28-CD8+ T cells - Analyses of a cohort of 1032 kidney transplant recipients.

BACKGROUND: CD28-CD8+ T cells represent a differentiated CD8+ T cell subset that is found to be increased in various conditions associated with chronic antigenic stimulation such as aging, chronic viral infections, autoimmune diseases, cancers, and allotransplantation. METHODS: Using multivariate models, we analyzed a large cohort of 1032 kidney transplant patients in whom 1495 kidney graft biopsies were performed concomitant with a peripheral blood leukocyte phenotyping by flow cytometry. We investigated the association between the level of CD28-CD8+ T cells in the blood and the diagnosis of graft rejection according to the recent Banff classification of renal allograft pathology. FINDINGS: We found that antibody-mediated rejection (ABMR) was associated with a significant increase in the percentage as well as the absolute number of CD28-CD8+ T cells in the peripheral blood of kidney transplant patients at the time of biopsy. The confounder-adjusted mean difference of log percentage and log absolute value between the ABMR group and the normal/subnormal histology group were 0.29 (p=0.0004) and 0.38 (p=0.0004), respectively. Moreover, we showed that CD28-CD8+ T cells from the patients diagnosed with ABMR responded more rigorously to TCR and FcγRIIIA (CD16) engagement compared to their CD28+ counterparts as evidenced by an increase in the expression of IFNγ, TNFα, and CD107a. INTERPRETATION: Collectively, our data suggest that differentiated CD28-CD8+ T cells, with increased frequency, number, and function, may participate in the pathobiology of ABMR. Further studies are warranted to clarify the immunological role of this T cell subset in kidney graft rejection. FUNDING: Agence nationale de la recherche (France).

Impact of the Kidney Transplantation Moratorium in France Because of the COVID-19 Pandemic: A Cohort-based Study.

BACKGROUND: The COVID-19 pandemic has resulted in worldwide kidney transplantation (KT) moratoriums. The impacts of these moratoriums on the life expectancy of KT candidates remain unclear. METHODS: We simulated the evolution of several French candidate populations for KT using a multistate semi-Markovian approach and according to moratorium durations ranging from 0 to 24 mo. The transition rates were modeled from the 63 927 French patients who began dialysis or were registered on the waiting list for KT between 2011 and 2019. RESULTS: Among the 8350 patients active on the waiting list at the time of the French KT moratorium decided on March 16, 2020, for 2.5 mo, we predicted 4.0 additional months (confidence interval [CI], 2.8-5.0) on the waiting list and 42 additional deaths (CI, -70 to 150) up to March 16, 2030, compared with the scenario without moratorium. In this population, we reported a significant impact for a 9-mo moratorium duration: 135 attributable deaths (CI, 31-257) up to March 16, 2030. Patients who became active on the list after March 2020 were less impacted; there was a significant impact for an 18-mo moratorium (175 additional deaths [CI, 21-359]) in the 10 862 prevalent end-stage renal disease patients on March 16, 2020 and for a 24-mo moratorium (189 additional deaths [CI, 10-367]) in the 16 355 incident end-stage renal disease patients after this date. CONCLUSION: The temporary moratorium of KT during a COVID-19 peak represents a sustainable decision to free up hospitals' resources if the moratorium does not exceed a prolonged period.

Early use of barbiturates is associated with increased mortality in traumatic brain injury patients from a propensity score-based analysis of a prospective cohort.

Barbiturates are proposed as a second/third line treatment for intracranial hypertension in traumatic brain injury (TBI) patients, but the literature remains uncertain regarding their benefit/risk balance. We aimed to evaluate the impact of barbiturates therapy in TBI patients with early intracranial hypertension on the intensive care unit (ICU) survival, the occurrence of ventilator-associated pneumonia (VAP), and the patient's functional status at three months. We used the French AtlanREA prospective cohort of trauma patients. Using a propensity score-based methodology (inverse probability of treatment weighting), we compared patients having received barbiturates within the first 24 hours of admission (barbiturates group) and those who did not (control group). We used cause-specific Cox models for ICU survival and risk of VAP, and logistic regression for the 3-month Glasgow Outcome Scale (GOS) evaluation. Among the 1396 patients with severe trauma, 383 had intracranial hypertension on admission and were analyzed. Among them, 96 (25.1%) received barbiturates. The early use of barbiturates was significantly associated with increased ICU mortality (HR = 1.85, 95%CI 1.03-3.33). However, barbiturates treatment was not significantly associated with VAP (HR = 1.02, 95%CI 0.75-1.41) or 3-month GOS (OR = 1.67, 95%CI 0.84-3.33). Regarding the absence of relevant clinical trials, our results suggest that each early prescription of barbiturates requires a careful assessment of the benefit/risk ratio.

G-computation and machine learning for estimating the causal effects of binary exposure statuses on binary outcomes.

In clinical research, there is a growing interest in the use of propensity score-based methods to estimate causal effects. G-computation is an alternative because of its high statistical power. Machine learning is also increasingly used because of its possible robustness to model misspecification. In this paper, we aimed to propose an approach that combines machine learning and G-computation when both the outcome and the exposure status are binary and is able to deal with small samples. We evaluated the performances of several methods, including penalized logistic regressions, a neural network, a support vector machine, boosted classification and regression trees, and a super learner through simulations. We proposed six different scenarios characterised by various sample sizes, numbers of covariates and relationships between covariates, exposure statuses, and outcomes. We have also illustrated the application of these methods, in which they were used to estimate the efficacy of barbiturates prescribed during the first 24 h of an episode of intracranial hypertension. In the context of GC, for estimating the individual outcome probabilities in two counterfactual worlds, we reported that the super learner tended to outperform the other approaches in terms of both bias and variance, especially for small sample sizes. The support vector machine performed well, but its mean bias was slightly higher than that of the super learner. In the investigated scenarios, G-computation associated with the super learner was a performant method for drawing causal inferences, even from small sample sizes.

Covariates adjustment questioned conclusions of predictive analyses: an illustration with the Kidney Donor Risk Index.

OBJECTIVES: We aimed to illustrate that considering covariates can lead to meaningful interpretation of the discriminative capacities of a prognostic marker. For this, we evaluated the ability of the Kidney Donor Risk Index (KDRI) to discriminate kidney graft failure risk. STUDY DESIGN AND SETTING: From 4114 French patients, we estimated the adjusted area under the time-dependent ROC curve by standardizing the marker and weighting the observations. By weighting the contributions, we also studied the impact of KDRI-based transplantations on the patient and graft survival. RESULTS: The covariate-adjusted AUC varied from 55% (95% confidence interval [CI]: 51-60%) for a prognostic up to 1 year post-transplantation to 56% (95% CI: 52-59%) up to 7 years. The Restricted Mean Survival Time (RMST) was 6.44 years for high-quality graft recipients (95% CI: 6.30-6.56) and would have been 6.31 years (95% CI: 6.13-6.46) if they had medium-quality transplants. The RMST was 5.10 years for low-quality graft recipients (95% CI: 4.90-5.31) and would have been 5.52 years (95% CI: 5.17-5.83) if they had medium-quality transplants. CONCLUSION: We demonstrated that the KDRI discriminative capacities were mainly explained by the recipient characteristics. We also showed that counterfactual estimations, often used in causal studies, are also interesting in predictive studies, especially regarding the new available methods.

G-computation, propensity score-based methods, and targeted maximum likelihood estimator for causal inference with different covariates sets: a comparative simulation study.

Controlling for confounding bias is crucial in causal inference. Distinct methods are currently employed to mitigate the effects of confounding bias. Each requires the introduction of a set of covariates, which remains difficult to choose, especially regarding the different methods. We conduct a simulation study to compare the relative performance results obtained by using four different sets of covariates (those causing the outcome, those causing the treatment allocation, those causing both the outcome and the treatment allocation, and all the covariates) and four methods: g-computation, inverse probability of treatment weighting, full matching and targeted maximum likelihood estimator. Our simulations are in the context of a binary treatment, a binary outcome and baseline confounders. The simulations suggest that considering all the covariates causing the outcome led to the lowest bias and variance, particularly for g-computation. The consideration of all the covariates did not decrease the bias but significantly reduced the power. We apply these methods to two real-world examples that have clinical relevance, thereby illustrating the real-world importance of using these methods. We propose an R package RISCA to encourage the use of g-computation in causal inference.

Standardized and weighted time-dependent receiver operating characteristic curves to evaluate the intrinsic prognostic capacities of a marker by taking into account confounding factors.

Time-dependent receiver operating characteristic curves allow to evaluate the capacity of a marker to discriminate between subjects who experience the event up to a given prognostic time from those who are free of this event. In this article, we propose an inverse probability weighting estimator of a standardized and weighted time-dependent receiver operating characteristic curve. This estimator provides a measure of the prognostic capacities by taking into account potential confounding factors. We illustrate the robustness of the estimator by a simulation-based study and its usefulness by two applications in kidney transplantation.

Increased soluble Flt-1 correlates with delayed graft function and early loss of peritubular capillaries in the kidney graft.

BACKGROUND: Ischemia-reperfusion induces tubular and endothelial damage in the renal graft and leads to delayed graft function (DGF) and to an early loss of peritubular capillaries (PTC). Few, if any, clinical studies have assessed the impact of proangiogenic and antiangiogenic factors on endothelial repair during renal transplantation (RT)-related ischemia-reperfusion. METHODS: We prospectively assessed the kinetics of the antiangiogenic factor soluble Fms-like tyrosine kinase-1 (sFlt-1) in 136 consecutive RT patients and analyzed sFlt-1 impact on DGF and PTC loss. RESULTS: sFlt-1 plasma levels increased by twofold to threefold throughout the first week after RT. This increase was more marked in recipients of grafts from deceased donors compared with living donors. Patients with DGF had higher sFlt-1 levels at all time points during the first 7 days after RT and a higher peak sFlt-1 compared with those without DGF. In multivariate analysis, a peak plasma sFlt-1 of 250 pg/mL or higher was associated with 2.5-fold increase in the risk of DGF (P=0.04). Similarly, patients with a peak plasma sFlt-1 of 250 pg/mL or higher had a more pronounced early decrease in PTC compared with those with a peak sFlt-1 less than 250 pg/mL. CONCLUSIONS: sFlt-1 is a new nonimmunologic independent risk factor for DGF and PTC loss. Its inhibition may help improve the outcome of RT.