Nomograms for Predicting Graft Function and Survival in Living Donor Kidney Transplantation Based on the UNOS Registry

Shear wave elastography parameters adds prognostic value to adverse outcome in kidney transplantation recipients

INTRODUCTION

The tissue stiffness of donor kidneys in transplantation may increase due to pathological changes such as glomerulosclerosis and interstitial fibrosis, and those changes associate worse outcomes in kidney transplantation recipients. Ultrasound elastography is a noninvasive imaging examination with the ability to quantitatively reflect tissue stiffness. Aim of this study was to evaluate the prognostic value of ultrasound elastography for adverse kidney outcome in kidney transplantation recipients.

METHODS

Shear wave elastography (SWE) examinations were performed by two independent operators in kidney transplantation recipients. The primary outcome was a composite of kidney graft deterioration, all-cause re-hospitalization, and all-cause mortality. Survival analysis was calculated by Kaplan-Meier curves with the log-rank test and Cox regression analysis.

RESULTS

A total of 161 patients (mean age 46 years, 63.4% men) were followed for a median of 20.1 months. 27 patients (16.77%) reached the primary endpoint. The mean and median tissue stiffness at the medulla (hazard ratio: 1.265 and 1.229, respectively), estimated glomerular filtration rate (eGFR), and serum albumin level were associated with the primary outcome in univariate Cox regression. Adding mean or median medulla SWE to a baseline model containing eGFR and albumin significantly improved its discrimination (C-statistics: 0.736 for the baseline, 0.766 and 0.772 for the model added mean and median medulla SWE, respectively).

CONCLUSION

The medullary tissue stiffness of kidney allograft measured by shear wave elastography may provide incremental prognostic value to adverse outcomes in kidney transplantation recipients. Including SWE parameters in kidney transplantation recipients management could be considered to improve risk stratification.

Prediction models for the recipients' ideal perioperative estimated glomerular filtration rates for predicting graft survival after adult living-donor kidney transplantation

Introduction

The impact of the perioperative estimated glomerular filtration rate (eGFR) on graft survival in kidney transplant recipients is yet to be evaluated. In this study, we developed prediction models for the ideal perioperative eGFRs in recipients.

Methods

We evaluated the impact of perioperative predicted ideal and actual eGFRs on graft survival by including 1,174 consecutive adult patients who underwent living-donor kidney transplantation (LDKT) between January 2008 and December 2020. Prediction models for the ideal perioperative eGFR were developed for 676 recipients who were randomly assigned to the training and validation sets (ratio: 7:3). The prediction models for the ideal best eGFR within 3 weeks and those at 1, 2, and 3 weeks after LDKT in 474 recipients were developed using 10-fold validation and stepwise multiple regression model analyzes. The developed prediction models were validated in 202 recipients. Finally, the impact of perioperative predicted ideal eGFRs/actual eGFRs on graft survival was investigated using Fine-Gray regression analysis.

Results

The correlation coefficients of the predicted ideal best eGFR within 3 weeks and the predicted ideal eGFRs at 1, 2, and 3 weeks after LDKT were 0.651, 0.600, 0.598, and 0.617, respectively. Multivariate analyzes for graft loss demonstrated significant differences in the predicted ideal best eGFR/actual best eGFR within 3 weeks and the predicted ideal eGFRs/actual eGFRs at 1, 2, and 3 weeks after LDKT.

Discussion

The predicted ideal best eGFR/actual best eGFR within 3 weeks and the predicted ideal eGFRs/actual eGFRs at 1, 2, and 3 weeks after LDKT were independent prognostic factors for graft loss. Therefore, the perioperative predicted ideal eGFR/actual eGFR may be useful for predicting graft survival after adult LDKT.

Pancreas Rejection in the Artificial Intelligence Era: New Tool for Signal Patients at Risk

INTRODUCTION

Pancreas transplantation is currently the only treatment that can re-establish normal endocrine pancreatic function. Despite all efforts, pancreas allograft survival and rejection remain major clinical problems. The purpose of this study was to identify features that could signal patients at risk of pancreas allograft rejection.

METHODS

We collected 74 features from 79 patients who underwent simultaneous pancreas-kidney transplantation (SPK) and used two widely-applicable classification methods, the Naive Bayesian Classifier and Support Vector Machine, to build predictive models. We used the area under the receiver operating characteristic curve and classification accuracy to evaluate the predictive performance via leave-one-out cross-validation.

RESULTS

Rejection events were identified in 13 SPK patients (17.8%). In feature selection approach, it was possible to identify 10 features, namely: previous treatment for diabetes mellitus with long-term Insulin (U/I/day), type of dialysis (peritoneal dialysis, hemodialysis, or pre-emptive), de novo DSA, vPRA_Pre-Transplant (%), donor blood glucose, pancreas donor risk index (pDRI), recipient height, dialysis time (days), warm ischemia (minutes), recipient of intensive care (days). The results showed that the Naive Bayes and Support Vector Machine classifiers prediction performed very well, with an AUROC and classification accuracy of 0.97 and 0.87, respectively, in the first model and 0.96 and 0.94 in the second model.

CONCLUSION

Our results indicated that it is feasible to develop successful classifiers for the prediction of graft rejection. The Naive Bayesian generated nomogram can be used for rejection probability prediction, thus supporting clinical decision making.

Development of a Nomogram to Predict Graft Survival After Descemet Stripping Endothelial Keratoplasty

BACKGROUND/PURPOSE

The purpose of this study was to analyze Descemet stripping endothelial keratoplasty (DSEK) outcomes and develop a nomogram to compute the probability of 3- and 5-year DSEK graft survival based on risk factors.

STUDY DESIGN/METHODS

The medical records of 794 DSEK procedures between January 1, 2008, and August 1, 2019, were retrospectively reviewed to identify 37 variables. We also evaluated for the presence of corneal graft failure, defined as irreversible and visually significant graft edema, haze, or scarring. Variables were assessed by multivariable Cox models, and a nomogram was created to predict the probability of 3- and 5-year graft survival.

RESULTS

Graft failure occurred in 80 transplants (10.1%). The strongest risk factors for graft failure included graft detachment [hazard ratio (HR) = 4.46; P < 0.001], prior glaucoma surgery (HR = 3.14; P = 0.001), and glaucoma (HR = 2.23; P = 0.018). A preoperative diagnosis of Fuchs dystrophy was associated with a decreased risk of graft failure (HR = 0.47; P = 0.005) compared with secondary corneal edema. Our nomogram has a concordance index of 0.75 (95% confidence interval, 0.69 to 0.81), which indicates that it may predict the probability of graft survival at 3 and 5 years with reasonable accuracy. We also analyzed graft rejection, which occurred in 39 cases (4.9%). The single risk factor found to be significantly associated with graft rejection was prior glaucoma surgery (HR = 2.87; P = 0.008).

CONCLUSIONS

Our nomogram may accurately predict DSEK graft survival after 3 and 5 years based on 4 variables. This nomogram will empower surgeons to share useful data with patients and improve collective clinical decision-making.

Individual dynamic prediction and prognostic analysis for long-term allograft survival after kidney transplantation

Background

Predicting allograft survival is vital for efficient transplant success. With dynamic changes in patient conditions, clinical indicators may change longitudinally, and doctors’ judgments may be highly variable. It is necessary to establish a dynamic model to precisely predict the individual risk/survival of new allografts.

Methods

The follow-up data of 407 patients were obtained from a renal allograft failure study. We introduced a landmarking-based dynamic Cox model that incorporated baseline values (age at transplantation, sex, weight) and longitudinal changes (glomerular filtration rate, proteinuria, hematocrit). Model performance was evaluated using Harrell’s C-index and the Brier score.

Results

Six predictors were included in our analysis. The Kaplan–Meier estimates of survival at baseline showed an overall 5-year survival rate of 87.2%. The dynamic Cox model showed the individual survival prediction with more accuracy at different time points (for the 5-year survival prediction, the C-index = 0.789 and Brier score = 0.065 for the average of all time points) than the static Cox model at baseline (C-index = 0.558, Brier score = 0.095). Longitudinal covariate prognostic analysis (with time-varying effects) was performed.

Conclusions

The dynamic Cox model can utilize clinical follow-up data, including longitudinal patient information. Dynamic prediction and prognostic analysis can be used to provide evidence and a reference to better guide clinical decision-making for applying early treatment to patients at high risk.

Using Information Available at the Time of Donor Offer to Predict Kidney Transplant Survival Outcomes: A Systematic Review of Prediction Models

Statistical models that can predict graft and patient survival outcomes following kidney transplantation could be of great clinical utility. We sought to appraise existing clinical prediction models for kidney transplant survival outcomes that could guide kidney donor acceptance decision-making. We searched for clinical prediction models for survival outcomes in adult recipients with single kidney-only transplants. Models that require information anticipated to become available only after the time of transplantation were excluded as, by that time, the kidney donor acceptance decision would have already been made. The outcomes of interest were all-cause and death-censored graft failure, and death. We summarised the methodological characteristics of the prediction models, predictive performance and risk of bias. We retrieved 4,026 citations from which 23 articles describing 74 models met the inclusion criteria. Discrimination was moderate for all-cause graft failure (C-statistic: 0.570–0.652; Harrell’s C: 0.580–0.660; AUC: 0.530–0.742), death-censored graft failure (C-statistic: 0.540–0.660; Harrell’s C: 0.590–0.700; AUC: 0.450–0.810) and death (C-statistic: 0.637–0.770; Harrell’s C: 0.570–0.735). Calibration was seldom reported. Risk of bias was high in 49 of the 74 models, primarily due to methods for handling missing data. The currently available prediction models using pre-transplantation information show moderate discrimination and varied calibration. Further model development is needed to improve predictions for the purpose of clinical decision-making.

Systematic Review Registration:https://osf.io/c3ehp/l.

Predicting Kidney Graft Survival Using Machine Learning Methods: Prediction Model Development and Feature Significance Analysis Study

BACKGROUND

Kidney transplantation is the optimal treatment for patients with end-stage renal disease. Short- and long-term kidney graft survival is influenced by a number of donor and recipient factors. Predicting the success of kidney transplantation is important for optimizing kidney allocation.

OBJECTIVE

The aim of this study was to predict the risk of kidney graft failure across three temporal cohorts (within 1 year, within 5 years, and after 5 years following a transplant) based on donor and recipient characteristics. We analyzed a large data set comprising over 50,000 kidney transplants covering an approximate 20-year period.

METHODS

We applied machine learning-based classification algorithms to develop prediction models for the risk of graft failure for three different temporal cohorts. Deep learning-based autoencoders were applied for data dimensionality reduction, which improved the prediction performance. The influence of features on graft survival for each cohort was studied by investigating a new nonoverlapping patient stratification approach.

RESULTS

Our models predicted graft survival with area under the curve scores of 82% within 1 year, 69% within 5 years, and 81% within 17 years. The feature importance analysis elucidated the varying influence of clinical features on graft survival across the three different temporal cohorts.

CONCLUSIONS

In this study, we applied machine learning to develop risk prediction models for graft failure that demonstrated a high level of prediction performance. Acknowledging that these models performed better than those reported in the literature for existing risk prediction tools, future studies will focus on how best to incorporate these prediction models into clinical care algorithms to optimize the long-term health of kidney recipients.

Dynamic prediction models for graft failure in paediatric kidney transplantation

BACKGROUND

Several models have been proposed to predict kidney graft failure in adult recipients but none in younger recipients. Our objective was to propose a dynamic prediction model for graft failure in young kidney transplant recipients.

METHODS

We included 793 kidney transplant recipients waitlisted before the age of 18 years who received a first kidney transplantation before the age of 21 years in France in 2002-13 and survived >90 days with a functioning graft. We used a Cox model including baseline predictors only (sex, age at transplant, primary kidney disease, dialysis duration, donor type and age, human leucocyte antigen matching, cytomegalovirus serostatus, cold ischaemia time and delayed graft function) and two joint models also accounting for post-transplant estimated glomerular filtration rate (eGFR) trajectory. Predictive performances were evaluated using a cross-validated area under the curve (AUC) and R2 curves.

RESULTS

When predicting the risk of graft failure from any time within the first 7 years after paediatric kidney transplantation, the predictions for the following 3 or 5 years were accurate and much better with the joint models than with the Cox model (AUC ranged from 0.83 to 0.91 for the joint models versus 0.56 to 0.64 for the Cox model).

CONCLUSION

Accounting for post-transplant eGFR trajectory strongly increased the accuracy of graft failure prediction in young kidney transplant recipients.

Development and validation of routine clinical laboratory data derived marker-based nomograms for the prediction of 5-year graft survival in kidney transplant recipients

Background: To develop and validate predictive nomograms for 5-year graft survival in kidney transplant recipients (KTRs) with easily-available laboratory data derived markers and clinical variables within the first year post-transplant.

Methods: The clinical and routine laboratory data from within the first year post-transplant of 1289 KTRs was collected to generate candidate predictors. Univariate and multivariate Cox analyses and LASSO were conducted to select final predictors. X-tile analysis was applied to identify optimal cutoff values to transform potential continuous factors into category variables and stratify patients. C-index, calibration curve, dynamic time-dependent AUC, decision curve analysis, and Kaplan-Meier curves were used to evaluate models’ predictive accuracy and clinical utility.

Results: Two predictive nomograms were constructed by using 0–6- and 0–12- month laboratory data, and showed good predictive performance with C-indexes of 0.78 and 0.85, respectively, in the training cohort. Calibration curves showed that the prediction probabilities of 5-year graft survival were in concordance with actual observations. Additionally, KTRs could be successfully stratified into three risk groups by nomograms.

Conclusions: These predictive nomograms combining demographic and 0–6- or 0–12- month markers derived from post-transplant laboratory data could serve as useful tools for early identification of 5-year graft survival probability in individual KTRs.

Combined kidney and hematopoeitic cell transplantation to induce mixed chimerism and tolerance

Based on preclinical studies, combined kidney and hematopoietic cell transplantation was performed on fully HLA matched and haplotype matched patients at the Stanford University Medical Center. The object of the studies was to induce mixed chimerism, immune tolerance, and complete immunosuppressive drug withdrawal. Tolerance, persistent mixed chimerism, and complete withdrawal was achieved in the majority of fully matched patients. Persistent mixed chimerism and partial withdrawal has been achieved in the haplotype matched patients at present.

Use of the Living Kidney Donor Profile Index in the Canadian Kidney Transplant Recipient Population: A Validation Study

Background:

The Living Kidney Donor Profile Index (LKDPI) was derived in a cohort of kidney transplant recipients (KTR) from the United States to predict the risk of total graft failure. There are important differences in patient demographics, listing practices, access to transplantation, delivery of care, and posttransplant mortality in Canada as compared with the United States, and the generalizability of the LKDPI in the Canadian context is unknown.

Objective:

The purpose of this study was to externally validate the LKDPI in a large contemporary cohort of Canadian KTR.

Design:

Retrospective cohort validation study.

Setting:

Toronto General Hospital, University Health Network, Toronto, Ontario, Canada

Patients:

A total of 645 adult (≥18 years old) living donor KTR between January 1, 2006 and December 31, 2016 with follow-up until December 31, 2017 were included in the study.

Measurements:

The predictive performance of the LKDPI was evaluated. The outcome of interest was total graft failure, defined as the need for chronic dialysis, retransplantation, or death with graft function.

Methods:

The Cox proportional hazards model was used to examine the relation between the LKDPI and total graft failure. The Cox proportional hazards model was also used for external validation and performance assessment of the model. Discrimination and calibration were used to assess model performance. Discrimination was assessed using Harrell’s C statistic and calibration was assessed graphically, comparing observed versus predicted probabilities of total graft failure.

Results:

A total of 645 living donor KTR were included in the study. The median LKDPI score was 13 (interquartile range [IQR] = 1.1, 29.9). Higher LKDPI scores were associated with an increased risk of total graft failure (hazard ratio = 1.01; 95% confidence interval [CI] = 1.0-1.02; P = .02). Discrimination was poor (C statistic = 0.55; 95% CI = 0.48-0.61). Calibration was as good at 1-year posttransplant but suboptimal at 3- and 5-years posttransplant.

Limitations:

Limitations include a relatively small sample size, predicted probabilities for assessment of calibration only available for scores of 0 to 100, and some missing data handled by imputation.

Conclusions:

In this external validation study, the predictive ability of the LKDPI was modest in a cohort of Canadian KTR. Validation of prediction models is an important step to assess performance in external populations. Potential recalibration of the LKDPI may be useful prior to clinical use in external cohorts.

Predicting donor, recipient and graft survival in living donor kidney transplantation to inform pretransplant counselling: the donor and recipient linked iPREDICTLIVING tool - a retrospective study

Although separate prediction models for donors and recipients were previously published, we identified a need to predict outcomes of donor/recipient simultaneously, as they are clearly not independent of each other. We used characteristics from transplantations performed at the Oslo University Hospital from 1854 live donors and from 837 recipients of a live donor kidney transplant to derive Cox models for predicting donor mortality up to 20 years, and recipient death, and graft loss up to 10 years. The models were developed using the multivariable fractional polynomials algorithm optimizing Akaike’s information criterion, and optimism‐corrected performance was assessed. Age, year of donation, smoking status, cholesterol and creatinine were selected to predict donor mortality (C‐statistic of 0.81). Linear predictors for donor mortality served as summary of donor prognosis in recipient models. Age, sex, year of transplantation, dialysis vintage, primary renal disease, cerebrovascular disease, peripheral vascular disease and HLA mismatch were selected to predict recipient mortality (C‐statistic of 0.77). Age, dialysis vintage, linear predictor of donor mortality, HLA mismatch, peripheral vascular disease and heart disease were selected to predict graft loss (C‐statistic of 0.66). Our prediction models inform decision‐making at the time of transplant counselling and are implemented as online calculators.

Using machine learning techniques to develop risk prediction models to predict graft failure following kidney transplantation: protocol for a retrospective cohort study

Background: A mechanism to predict graft failure before the actual kidney transplantation occurs is crucial to clinical management of chronic kidney disease patients.  Several kidney graft outcome prediction models, developed using machine learning methods, are available in the literature.  However, most of those models used small datasets and none of the machine learning-based prediction models available in the medical literature modelled time-to-event (survival) information, but instead used the binary outcome of failure or not. The objective of this study is to develop two separate machine learning-based predictive models to predict graft failure following live and deceased donor kidney transplant, using time-to-event data in a large national dataset from Australia.   Methods: The dataset provided by the Australia and New Zealand Dialysis and Transplant Registry will be used for the analysis. This retrospective dataset contains the cohort of patients who underwent a kidney transplant in Australia from January 1 st, 2007, to December 31 st, 2017.  This included 3,758 live donor transplants and 7,365 deceased donor transplants.  Three machine learning methods (survival tree, random survival forest and survival support vector machine) and one traditional regression method, Cox proportional regression, will be used to develop the two predictive models.  The best predictive model will be selected based on the model's performance. Discussion: This protocol describes the development of two separate machine learning-based predictive models to predict graft failure following live and deceased donor kidney transplant, using a large national dataset from Australia.   Furthermore, these two models will be the most comprehensive kidney graft failure predictive models that have used survival data to model using machine learning techniques.  Thus, these models are expected to provide valuable insight into the complex interactions between graft failure and donor and recipient characteristics.

Using machine learning techniques to develop risk prediction models to predict graft failure following kidney transplantation: protocol for a retrospective cohort study

Background: A mechanism to predict graft failure before the actual kidney transplantation occurs is crucial to clinical management of chronic kidney disease patients.  Several kidney graft outcome prediction models, developed using machine learning methods, are available in the literature.  However, most of those models used small datasets and none of the machine learning-based prediction models available in the medical literature modelled time-to-event (survival) information, but instead used the binary outcome of failure or not. The objective of this study is to develop two separate machine learning-based predictive models to predict graft failure following live and deceased donor kidney transplant, using time-to-event data in a large national dataset from Australia.   Methods: The dataset provided by the Australia and New Zealand Dialysis and Transplant Registry will be used for the analysis. This retrospective dataset contains the cohort of patients who underwent a kidney transplant in Australia from January 1 st, 2007, to December 31 st, 2017. This included 3,758 live donor transplants and 7,365 deceased donor transplants. Three machine learning methods (survival tree, random survival forest and survival support vector machine) and one traditional regression method, Cox proportional regression, will be used to develop the two predictive models (for live donor and deceased donor transplants). The best predictive model will be selected based on the model's performance. Discussion: This protocol describes the development of two separate machine learning-based predictive models to predict graft failure following live and deceased donor kidney transplant, using a large national dataset from Australia. Furthermore, these two models will be the most comprehensive kidney graft failure predictive models that have used survival data to model using machine learning techniques. Thus, these models are expected to provide valuable insight into the complex interactions between graft failure and donor and recipient characteristics.

Development and validation of a new prediction model for graft function using preoperative marginal factors in living-donor kidney transplantation

BACKGROUND

Recently, living-donor kidney transplantation from marginal donors has been increasing. However, a simple prediction model for graft function including preoperative marginal factors is limited. Here, we developed and validated a new prediction model for graft function using preoperative marginal factors in living-donor kidney transplantation.

METHODS

We retrospectively investigated 343 patients who underwent living-donor kidney transplantation at Kyushu University Hospital (derivation cohort). Low graft function was defined as an estimated glomerular filtration rate of < 45 mL/min/1.73 m² at 1 year. A prediction model was developed using a multivariable logistic regression model, and verified using data from 232 patients who underwent living-donor kidney transplantation at Tokyo Women's Medical University Hospital (validation cohort).

RESULTS

In the derivation cohort, 89 patients (25.9%) had low graft function at 1 year. Donor age, donor-estimated glomerular filtration rate, donor hypertension, and donor/recipient body weight ratio were selected as predictive factors. This model demonstrated modest discrimination (c-statistic = 0.77) and calibration (Hosmer-Lemeshow test, P = 0.83). Furthermore, this model demonstrated good discrimination (c-statistic = 0.76) and calibration (Hosmer-Lemeshow test, P = 0.54) in the validation cohort. Furthermore, donor age, donor-estimated glomerular filtration rate, and donor hypertension were strongly associated with glomerulosclerosis and atherosclerotic vascular changes in the "zero-time" biopsy.

CONCLUSIONS

This model using four pre-operative variables will be a simple, but useful guide to estimate graft function at 1 year after kidney transplantation, especially in marginal donors, in the clinical setting.

Machine learning in predicting graft failure following kidney transplantation: A systematic review of published predictive models

INTRODUCTION

Machine learning has been increasingly used to develop predictive models to diagnose different disease conditions. The heterogeneity of the kidney transplant population makes predicting graft outcomes extremely challenging. Several kidney graft outcome prediction models have been developed using machine learning, and are available in the literature. However, a systematic review of machine learning based prediction methods applied to kidney transplant has not been done to date. The main aim of our study was to perform an in-depth systematic analysis of different machine learning methods used to predict graft outcomes among kidney transplant patients, and assess their usefulness as an aid to decision-making.

METHODS

A systemic review of machine learning methods used to predict graft outcomes among kidney transplant patients was carried out using a search of the Medline, the Cumulative Index to Nursing and Allied Health Literature, EMBASE, PsycINFO and Cochrane databases.

RESULTS

A total of 295 articles were identified and extracted. Of these, 18 met the inclusion criteria. Most of the studies were published in the United States after 2010. The population size used to develop the models varied from 80 to 92,844, and the number of features in the models ranged from 6 to 71. The most common machine learning methods used were artificial neural networks, decision trees and Bayesian belief networks. Most of the machine learning based predictive models predicted graft failure with high sensitivity and specificity. Only one machine learning based prediction model had modelled time-to-event (survival) information. Seven studies compared the predictive performance of machine learning models with traditional regression methods and the performance of machine learning methods was found to be mixed, when compared with traditional regression methods.

CONCLUSION

There was a wide variation in the size of the study population and the input variables used. However, the prediction accuracy provided mixed results when machine learning and traditional predictive methods are compared. Based on reported gains in predictive performance, machine learning has the potential to improve kidney transplant outcome prediction and aid medical decision making.

Predicting urine output after kidney transplantation: development and internal validation of a nomogram for clinical use

Purpose:

To analyze pre-transplantation and early postoperative factors affecting post-transplantation urine output and develop a predictive nomogram.

Patients and Methods:

Retrospective analysis of non-preemptive first transplanted adult patients between 2001-2016. The outcomes were hourly diuresis in mL/Kg in the 1^st (UO₁) and 8^th (UO₈) postoperative days (POD). Predictors for both UO₁ and UO₈ were cold ischemia time (CIT), patient and donor age and sex, HLA I and II compatibility, pre-transplantation duration of renal replacement therapy (RRT), cause of ESRD (ESRD) and immunosuppressive regimen. UO₈ predictors also included UO₁, 1^st/0^th POD plasma creatinine concentration ratio (Cr_1/0), and occurrence of acute cellular rejection (AR). Multivariable linear regression was employed to produce nomograms for UO₁ and UO₈.

Results:

Four hundred and seventy-three patients were included, mostly deceased donor kidneys’ recipients (361, 70.4%). CIT inversely correlated with UO₁ and UO₈ (Spearman's p=-0.43 and −0.37). CR_1/0 inversely correlated with UO₈ (p=-0.47). On multivariable analysis UO₁ was mainly influenced by CIT, with additional influences of donor age and sex, HLA II matching and ESRD. UO₁ was the strongest predictor of UO₈, with significant influences of AR and ESRD.

Conclusions:

The predominant influence of CIT on UO₁ rapidly wanes and is replaced by indicators of functional recovery (mainly UO₁) and allograft's immunologic acceptance (AR absence). Mean absolute errors for nomograms were 0.08 mL/Kg h (UO₁) and 0.05 mL/Kg h (UO₈).

Predicting the function of transplanted kidney in long-term care processes: Application of a hybrid model

BACKGROUND

A tool that can predict the estimated glomerular filtration rate (eGFR) in routine daily care can help clinicians to make better decisions for kidney transplant patients and to improve transplantation outcome. In this paper, we proposed a hybrid prediction model for predicting a future value for eGFR during long-term care processes.

METHODS

Longitudinal, historical data of 942 transplant patients who received a kidney between 2001 and 2016 at Urmia kidney transplant center was used to develop a hybrid model. The model was based on three primary models: multi-layer perceptron (MLP), linear regression (LR), and a model that predicted a smoothed value of eGFR. The hybrid model used at-hand, longitudinal data of physical examinations and laboratory test values available at each visit. Two different datasets, a generalized dataset (GData) and a personalized dataset (PData), were created. Then, in both datasets, two data subsets of development and validation were created. For prediction, all records related to the fourth to tenth previous visits of patients in time order from the target date, i.e., window size (WS) = 4-10, were used. The performance of the models was evaluated using Mean Square Error (MSE) and Mean Absolute Error (MAE). The differences between the models were evaluated with the F-test and the Akaike Information Criterion (AIC).

RESULTS

The datasets contained 35,066 records, totally. The GData contained 26,210 and 8856 records and the PData had 24,079 and 9103 records in the development and validation datasets, respectively. In the hybrid model, the MSE and MAE were 153 and 8.9 in the GData, and 113 and 7.5 in the PData, respectively. The model performance improved using a wider WS of historical records (from 4 to 10). When the WS of ten was used the MSE and MAE declined to 141 and 8.5 in the GData and to 91 and 6.9 in the PData, respectively. In both datasets, the F-test showed that the hybrid model was significantly different from other models. The AIC showed that the hybrid model had a better performance than that of others.

CONCLUSIONS

The hybrid model can predict a reliable future value for eGFR. Our results showed that longitudinal covariates help the models to produce better results. Smoothing eGFR values and using a personalized dataset to develop the models also improved the models' performances. They can be considered as a step forward towards personalized medicine.

A dynamic model for predicting graft function in kidney recipients' upcoming follow up visits: A clinical application of artificial neural network

BACKGROUND

Predicting the function of transplanted kidneys would help clinicians in individualized medical interventions. We aimed to develop and validate a predictive tool for a future value of estimated glomerular filtration rate (eGFR) at upcoming visits.

METHODS

We used static and time-dependent covariates as inputs of artificial neural network based prediction models for predicting an eGFR value for an upcoming visit. We included 675 kidney recipients, who received transplant in the Urmia kidney transplant center in 2001-2013 and were longitudinally cared for in 2001-2017. The first 75% of records of longitudinal data of each patient were used to develop the prediction models and the remaining last 25% for evaluating its performance. Models' performances were evaluated by Mean Square Error (MSE) and Mean Absolute Error (MAE).

RESULTS

The development and validation datasets included 18,773 and 7038 records of historical data, respectively. The most accurate model included 3 static covariates of recipients' gender and donors' age and gender as well as 11 dynamic covariates of recipients including current age, time since transplant, serum creatinine, fasting blood sugar, weight and blood pressures available at each visit time. The performance of prediction models in the validation cohort was improved when history window of time dependent variables' recent values was increased from 1 to 10 (an MSE decline from 161 to 99).

CONCLUSIONS

Our best performed model is able to dynamically predict a future eGFR value for kidney recipients' upcoming visits. Integrating such a clinical tool into daily workflow of outpatient clinics can potentially support clinicians in optimal and individualized decision makings.

Predictive Score for Posttransplantation Outcomes

BACKGROUND

Most current scoring tools to predict allograft and patient survival upon kidney transplantion are based on variables collected posttransplantation. We developed a novel score to predict posttransplant outcomes using pretransplant information including routine laboratory data available before or at the time of transplantation.

METHODS

Linking the 5-year patient data of a large dialysis organization to the Scientific Registry of Transplant Recipients, we identified 15 125 hemodialysis patients who underwent first deceased transplantion. Prediction models were developed using Cox models for (a) mortality, (b) allograft loss (death censored), and (c) combined death or transplant failure. The cohort was randomly divided into a two thirds set (Nd = 10 083) for model development and a one third set (Nv = 5042) for validation. Model predictive discrimination was assessed using the index of concordance, or C statistic, which accounts for censoring in time-to-event models (a-c). We used the bootstrap method to assess model overfitting and calibration using the development dataset.

RESULTS

Patients were 50 ± 13 years of age and included 39% women, 15% African Americans, and 36% persons with diabetes. For prediction of posttransplant mortality and graft loss, 10 predictors were used (recipients' age, cause and length of end-stage renal disease, hemoglobin, albumin, selected comorbidities, race and type of insurance as well as donor age, diabetes status, extended criterion donor kidney, and number of HLA mismatches). The new model (www.TransplantScore.com) showed the overall best discrimination (C-statistics, 0.70; 95% confidence interval [95% CI], 0.67-0.73 for mortality; 0.63; 95% CI, 0.60-0.66 for graft failure; 0.63; 95% CI, 0.61-0.66 for combined outcome).

CONCLUSIONS

The new prediction tool, using data available before the time of transplantation, predicts relevant clinical outcomes and may perform better to predict patients' graft survival than currently used tools.

Cross-trial prediction of treatment outcome in depression: a machine learning approach

BACKGROUND

Antidepressant treatment efficacy is low, but might be improved by matching patients to interventions. At present, clinicians have no empirically validated mechanisms to assess whether a patient with depression will respond to a specific antidepressant. We aimed to develop an algorithm to assess whether patients will achieve symptomatic remission from a 12-week course of citalopram.

METHODS

We used patient-reported data from patients with depression (n=4041, with 1949 completers) from level 1 of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D; ClinicalTrials.gov, number NCT00021528) to identify variables that were most predictive of treatment outcome, and used these variables to train a machine-learning model to predict clinical remission. We externally validated the model in the escitalopram treatment group (n=151) of an independent clinical trial (Combining Medications to Enhance Depression Outcomes [COMED]; ClinicalTrials.gov, number NCT00590863).

FINDINGS

We identified 25 variables that were most predictive of treatment outcome from 164 patient-reportable variables, and used these to train the model. The model was internally cross-validated, and predicted outcomes in the STAR*D cohort with accuracy significantly above chance (64·6% [SD 3·2]; p<0·0001). The model was externally validated in the escitalopram treatment group (N=151) of COMED (accuracy 59·6%, p=0.043). The model also performed significantly above chance in a combined escitalopram-buproprion treatment group in COMED (n=134; accuracy 59·7%, p=0·023), but not in a combined venlafaxine-mirtazapine group (n=140; accuracy 51·4%, p=0·53), suggesting specificity of the model to underlying mechanisms.

INTERPRETATION

Building statistical models by mining existing clinical trial data can enable prospective identification of patients who are likely to respond to a specific antidepressant.

FUNDING

Yale University.

Risk Factors of Graft Survival After Diagnosis of Post-kidney Transplant Malignancy: Using Cox Proportional Hazard Model

BACKGROUND

All recipients of kidney transplantation, especially those with posttransplant malignancy, are at risk of long-term graft failure.

OBJECTIVES

The purpose of our study was to evaluate the risk factors associated with graft survival after diagnosis of malignancy.

PATIENTS AND METHODS

To reach this purpose, we conducted a historical cohort study in Iran and 266 cases with posttransplant malignancy were followed up from diagnosis of malignancy until long-term graft loss or the date of last visit. These patients were taken as a census from 16 Transplant Centers in Iran during 22 years follow-up period since October 1984 to December 2008. A Cox proportional hazards model was performed to determine the important independent predictors of graft survival after malignancy.

RESULTS

At the end of the study, long-term graft failure was seen in 27 (10.2%) cases. One-year and 2-year graft survival after diagnosis of cancer were 93.6% and 91.7%, respectively. The univariate analysis showed that the incidence of chronic graft loss was significantly higher in male patients with solid cancers, withdrawal of immunosuppressant regimen, no response to treatment, and tumor metastasis. In continuation, the Cox model indicated that the significant risk factors associated with graft survival were type of cancer (P < 0.0001), response to treatment (P < 0.0001, HR = 0.14, 95% CI: 0.06 - 0.32), metastasis (P < 0.0001, HR = 5.68, 95% CI: 2.24 - 14.42), and treatment modality (P = 0.0001).

CONCLUSIONS

By controlling the modifiable risk factors and modality of treatment in our study, physicians can reach more effective treatment.

The Kidney Transplant Morbidity Index (KTMI): A Simple Prognostic Tool to Help Determine Outcome Risk in Kidney Transplant Candidates

Background

The Kidney Transplant Morbidity Index (KTMI) is a novel prognostic morbidity index to help determine the impact that pretransplant comorbid conditions have on transplant outcome.

Objective

To use national data to validate the KTMI.

Design

Retrospective analysis of the Organ Procurement and Transplant Network/United Network for Organ Sharing database.

Setting and Participants

The study sample consisted of 100 261 adult patients who received a kidney transplant between 2000 and 2008.

Main Outcome Measure

Kaplan-Meier survival curves were used to demonstrate 3-year graft and patient survival for each KTMI score. Cox proportional hazards regression models were created to determine hazards for 3-year graft failure and patient mortality for each KTMI score.

Results

A sequential decrease in graft survival (0 = 91.2%, 1 = 88.2%, 2 = 85.4%, 3 = 81.7%, 4 = 77.8%, 5 = 74.0%, 6 = 69.8%, and ≥7 = 68.7) and patient survival (0 = 98.2%, 1 = 96.6%, 2 = 93.7%, 3 = 89.7%, 4 = 84.8%, 5 = 80.8%, 6 = 76.0%, and ≥7 = 74.7%) is seen as KTMI scores increase. The differences in graft and patient survival between KTMI scores are all significant ( P < .001) except between 6 and ≥7. Multivariate regression analysis reveals that KTMI is an independent predictor of higher graft failure and patient mortality rates and that risk increases as KTMI scores increase.

Conclusion

The KTMI strongly predicts graft and patient survival by using pretransplant comorbid conditions; therefore, this easy-to-use tool can aid in determining outcome risk and transplant candidacy before listing, particularly in candidates with multiple comorbid conditions.

Chimerism, graft survival, and withdrawal of immunosuppressive drugs in HLA matched and mismatched patients after living donor kidney and hematopoietic cell transplantation

Thirty-eight HLA matched and mismatched patients given combined living donor kidney and enriched CD34(+) hematopoietic cell transplants were enrolled in tolerance protocols using posttransplant conditioning with total lymphoid irradiation and anti-thymocyte globulin. Persistent chimerism for at least 6 months was associated with successful complete withdrawal of immunosuppressive drugs in 16 of 22 matched patients without rejection episodes or kidney disease recurrence with up to 5 years follow up thereafter. One patient is in the midst of withdrawal and five are on maintenance drugs. Persistent mixed chimerism was achieved in some haplotype matched patients for at least 12 months by increasing the dose of T cells and CD34(+) cells infused as compared to matched recipients in a dose escalation study. Success of drug withdrawal in chimeric mismatched patients remains to be determined. None of the 38 patients had kidney graft loss or graft versus host disease with up to 14 years of observation. In conclusion, complete immunosuppressive drug withdrawal could be achieved thus far with the tolerance induction regimen in HLA matched patients with uniform long-term graft survival in all patients.

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.

Chronic graft loss and death in patients with post-transplant malignancy in living kidney transplantation: a competing risk analysis

Background:

Malignancy is a common complication after renal transplantation. Death with functioning graft and chronic graft loss are two competing outcomes in patients with post-transplant malignancies.

Objectives:

The purpose of our study was to evaluate the risk factors associated with cumulative incidence of these two outcomes.

Patients and Methods:

Fine-Gray model was used for 266 cases with post-transplant malignancy in Iran. These patients were followed-up from the diagnosis until the date of last visit, chronic graft loss, or death, subsequently.

Results:

At the end of the study, as competing events, chronic graft loss and death with functioning graft were seen in 27 (10.2%) and 53 cases (19.9%), respectively, while 186 cases (69.9%) were accounted as censored. The incidence rate of death was approximately two-time of the incidence rate of chronic graft loss (8.6 vs. 4.4 per 100 person-years). In multivariate analysis, significant risk factors associated with cumulative incidence of death included age (P < 0.007, subhazard ratio (SHR) = 1.03), type of cancer (P < 0.0001), and response to treatment (P < 0.0001, SHR = 0.027). The significant risk factors associated with cumulative incidence of chronic graft loss were gender (P = 0.05, SHR = 0.37), treatment modality (P < 0.0001), and response to treatment (P = 0.048, SHR = 0.47).

Conclusions:

Using these factors, nephrologists may predict the occurrence of graft loss or death. If the probability of graft loss was higher, physicians can decrease the immunosuppressive medications dosage to decrease the incidence of graft loss.

Adding propensity scores to pure prediction models fails to improve predictive performance

Background. Propensity score usage seems to be growing in popularity leading researchers to question the possible role of propensity scores in prediction modeling, despite the lack of a theoretical rationale. It is suspected that such requests are due to the lack of differentiation regarding the goals of predictive modeling versus causal inference modeling. Therefore, the purpose of this study is to formally examine the effect of propensity scores on predictive performance. Our hypothesis is that a multivariable regression model that adjusts for all covariates will perform as well as or better than those models utilizing propensity scores with respect to model discrimination and calibration. Methods. The most commonly encountered statistical scenarios for medical prediction (logistic and proportional hazards regression) were used to investigate this research question. Random cross-validation was performed 500 times to correct for optimism. The multivariable regression models adjusting for all covariates were compared with models that included adjustment for or weighting with the propensity scores. The methods were compared based on three predictive performance measures: (1) concordance indices; (2) Brier scores; and (3) calibration curves. Results. Multivariable models adjusting for all covariates had the highest average concordance index, the lowest average Brier score, and the best calibration. Propensity score adjustment and inverse probability weighting models without adjustment for all covariates performed worse than full models and failed to improve predictive performance with full covariate adjustment. Conclusion. Propensity score techniques did not improve prediction performance measures beyond multivariable adjustment. Propensity scores are not recommended if the analytical goal is pure prediction modeling.

Bayesian modeling of pretransplant variables accurately predicts kidney graft survival

INTRODUCTION

Machine learning can enable the development of predictive models that incorporate multiple variables for a systems approach to organ allocation. We explored the principle of Bayesian Belief Network (BBN) to determine whether a predictive model of graft survival can be derived using pretransplant variables. Our hypothesis was that pretransplant donor and recipient variables, when considered together as a network, add incremental value to the classification of graft survival.

METHODS

We performed a retrospective analysis of 5,144 randomly selected patients (age ≥18, deceased donor kidney only, first-time recipients) from the United States Renal Data System database between 2000 and 2001. Using this dataset, we developed a machine-learned BBN that functions as a pretransplant organ-matching tool.

RESULTS

A network of 48 clinical variables was constructed and externally validated using an additional 2,204 patients of matching demographic characteristics. This model was able to predict graft failure within the first year or within 3 years (sensitivity 40%; specificity 80%; area under the curve, AUC, 0.63). Recipient BMI, gender, race, and donor age were amongst the pretransplant variables with strongest association to outcome. A 10-fold internal cross-validation showed similar results for 1-year (sensitivity 24%; specificity 80%; AUC 0.59) and 3-year (sensitivity 31%; specificity 80%; AUC 0.60) graft failure.

CONCLUSION

We found recipient BMI, gender, race, and donor age to be influential predictors of outcome, while wait time and human leukocyte antigen matching were much less associated with outcome. BBN enabled us to examine variables from a large database to develop a robust predictive model.

Influence of Donor's Renal Function on the Outcome of Living Kidney Transplantation: 10-Year Follow-up

Purpose

With the improved surgical techniques and immunosuppression available today, conventional prognostic factors have taken on less significance. Accordingly, the native renal function of the donor is thought to be more important. Thus, we analyzed the prognostic significance of the donor's renal function as assessed by 24-hour urine creatinine clearance on kidney graft survival for 10 years after living kidney transplantation.

Materials and Methods

From January 1998 to July 2000, 71 living kidney transplantations were performed at a single institution. From among these, 68 recipients were followed for more than 6 months and were included in the present analysis. We analyzed kidney graft survival according to clinical parameters of the donor and the recipient.

Results

Mean follow-up duration of recipients after living kidney transplantation was 115.0±39.4 months (range, 10 to 157 months), and 31 recipients (45.6%) experienced kidney graft loss during this time period. Estimated mean kidney graft survival time was 131.8±6.2 months, and 5-year and 10-year kidney graft survival rates were estimated as 88.2% and 61.0%, respectively. Donor's mean 24-hour urine creatinine clearance (Ccr) before kidney transplantation was 122.8±21.2 ml/min/1.73 m² (range, 70.1 to 186.6 ml/min/1.73 m²). The 10-year kidney graft survival rates for cases stratified by a donor's Ccr lower and higher than 120 ml/min/1.73 m² were 39.0% and 67.2%, respectively (p=0.005). In univariate and multivariate analysis, donor's Ccr was retained as an independent prognostic factor of kidney graft survival (p=0.001 and 0.005, respectively).

Conclusions

Donor's 24-hour urine Ccr before living kidney transplantation was an independent prognostic factor of kidney graft survival. Therefore, it should be considered before living kidney transplantation.

An empirical approach to model selection through validation for censored survival data

Medical prognostic models can be designed to predict the future course or outcome of disease progression after diagnosis or treatment. The existing variable selection methods may be precluded by full model advocates when we build a prediction model owing to their estimation bias and selection bias in right-censored time-to-event data. If our objective is to optimize predictive performance by some criterion, we can often achieve a reduced model that has a little bias with low variance, but whose overall performance is enhanced. To accomplish this goal, we propose a new variable selection approach that combines Stepwise Tuning in the Maximum Concordance Index (STMC) with Forward Nested Subset Selection (FNSS) in two stages. In the first stage, the proposed variable selection is employed to identify the best subset of risk factors optimized with the concordance index using inner cross-validation for optimism correction in the outer loop of cross-validation, yielding potentially different final models for each of the folds. We then feed the intermediate results of the prior stage into another selection method in the second stage to resolve the overfitting problem and to select a final model from the variation of predictors in the selected models. Two case studies on relatively different sized survival data sets as well as a simulation study demonstrate that the proposed approach is able to select an improved and reduced average model under a sufficient sample and event size compared with other selection methods such as stepwise selection using the likelihood ratio test, Akaike Information Criterion (AIC), and lasso. Finally, we achieve better final models in each dataset than their full models by most measures. These results of the model selection models and the final models are assessed in a systematic scheme through validation for the independent performance.