Dynamic prediction models for graft failure in paediatric kidney transplantation

Clinical prediction models in children that use repeated measurements with time-varying covariates: a scoping review

BACKGROUND

Emerging evidence suggests that clinical prediction models that use repeated (time-varying) measurements within each patient may have higher predictive accuracy than models that use patient information from a single measurement.

OBJECTIVE

To determine the breadth of the published literature reporting the development of clinical prediction models in children that use time-varying predictors.

DATA SOURCES

MEDLINE, EMBASE and Cochrane databases.

ELIGIBILITY CRITERIA

We included studies reporting the development of a multivariable clinical prediction model in children, with or without validation, to predict a repeatedly measured binary or time-to-event outcome and utilizing at least one repeatedly measured predictor.

SYNTHESIS METHODS

We categorized included studies by the method used to model time-varying predictors.

RESULTS

Of 99 clinical prediction model studies that had a repeated measurements data structure, only 27 (27%) used methods that incorporated the repeated measurements as time-varying predictors in a single model. Among these 27 time-varying prediction model studies, we grouped model types into nine categories: time-dependent Cox regression, generalized estimating equations, random effects model, landmark model, joint model, neural network, K-nearest neighbor, support vector machine and tree-based algorithms. Where there was comparison of time-varying models to single measurement models, using time-varying predictors improved predictive accuracy.

CONCLUSIONS

Various methods have been used to develop time-varying prediction models in children, but there is a paucity of pediatric time-varying models in the literature. Incorporating time-varying covariates in pediatric prediction models may improve predictive accuracy. Future research in pediatric prediction model development should further investigate whether incorporation of time-varying covariates improves predictive accuracy.

Trajectories of eGFR after kidney transplantation according to trajectories of eGFR prior to kidney replacement therapies in children with chronic kidney disease

BACKGROUND

In children with chronic kidney disease (CKD), certain risk factors are associated with faster eGFR decline and earlier kidney failure. Whether these factors have lingering effects on post-transplant eGFR trajectory remains unclear. We characterized pre- and post-transplant eGFR trajectories in pediatric kidney transplant recipients by their pre-kidney replacement therapy (KRT) risk factors.

METHODS

We studied eGFR trajectories before KRT initiation and after transplantation among Chronic Kidney Disease in Children (CKiD) Study participants. We used mixed-effects models to compare pre-KRT versus post-transplant eGFR trajectories within individual participants by 7 pre-KRT risk factors: glomerular/non-glomerular etiology, race, preemptive transplant, proteinuria, albuminuria, and systolic/diastolic blood pressure (SBP/DBP).

RESULTS

We analyzed 1602 pre-KRT and 592 post-transplant eGFR measurements from 246 transplant recipients. Mean annual eGFR decline was decreased from 18.0% pre-KRT (95%CI, 16.1-19.8) to 5.0% post-transplant (95%CI, 3.3-6.7). All 7 pre-KRT risk factors showed strong associations with faster pre-KRT eGFR decline, but not with post-transplant eGFR decline; only albuminuria, high SBP, and high DBP reached statistical significance with notably attenuated associations. In our multivariable model of the pre-KRT risk factors, post-transplant eGFR decline was more rapid only when albuminuria and high SBP were both present.

CONCLUSIONS

eGFR decline substantially slows down after transplant even among children with rapidly progressing forms of CKD. Nonetheless, those who had albuminuria and high SBP before KRT might continue to show faster eGFR decline after transplant, specifically when both risk factors were present. This subgroup might benefit from intensive pre-transplant management for at least one of the two risk factors. A higher resolution version of the Graphical abstract is available as Supplementary information.

Is Non-Adherence Associated with Adverse Outcomes in Kidney Transplant Recipients? The Role of Non-Adherence as a Risk and Predictor Factor for Graft Loss and Death

Introduction

 Non-adherence in kidney transplants is diversely defined. Immunosuppression non-adherence (INA) is the most used definition and has been associated with graft loss and acute rejection. But INA assesses only one fraction of adherence. Therefore, we analyzed the association of a holistic non-adherence definition with transplant outcomes and compared its prediction performance with other definitions.

Methods

 We retrospectively included 739 kidney recipients between 2019 and 2021. We evaluated holistic non-adherence (HNA), suboptimal-immunosuppressor levels (SIL), appointment non-adherence (ANA), procedure non-adherence (PNA) and INA. The main outcomes were graft loss, graft rejection, and mortality. A backward logistic regression was performed estimating adjusted and un-adjusted odds ratio (OR) for each outcome. Finally, we compared the non-adherence definitions' prediction for the main outcomes using the area under the curve.

Results

HNA was present in 28.7% of patients. Non-adherent patients had an adjusted OR of 2.66 (1.37-5.15) for mortality, 6.44 for graft loss (2.71-16.6), and 2.28 (1.15-4.47) for graft rejection. INA and PNA presented a moderate discrimination for graft loss and HNA and ANA mild-to-moderate discrimination for graft loss and death.

Conclusion

Holistic non-adherence was associated with worst outcomes in kidney recipients and had a significant prediction performance for graft loss and mortality.

[REIN: a tool at the service of the paediatric patients]

On the occasion of the 20th anniversary of the REIN (French Renal Epidemiology and Information Network), a summary work on the contributions of the national French ESKD register was carried out. On the issue of paediatric CKD patients, the following key messages were retained. Paediatric stage 5 chronic kidney disease (CKD) has particularities that require to be analysed and taken into account because the mortality of these patients remains 30 times higher than that of children of the same age. The REIN registry enables illustrating the specificities of stage 5 CKD in the paediatric age-group in France and providing a set of indicators making it possible to describe the future of these patients as well as the choices made concerning the modalities of replacement therapy. As compared to other European countries, the incidence and prevalence of stage 5 CKD in France is in the middle range for children under 15 and 20 years old. Renal transplant is by far the leading treatment for stage 5 CKD in children and adolescents under 18 years of age in France, allowing to offer these patients the best possible life expectancy. Owing to the small volume of patients, only a nationwide registry can provide an unbiased view and enables analysing this population requiring a hyperspecialised treatment. The participation of French paediatric nephrologists in the REIN French registry also enables providing input to the European registry (ESPN/ERA www.espn-reg.org/index.jsp) and the international registry (IPNA https://ipna-registry.org) (Consulted on September 15th 2022) and thus the possibility of international studies, which are vital to be in line with an approach to improving practices.

A pre-transplantation risk assessment tool for graft survival in Dutch pediatric kidney recipients

Background

A prediction model for graft survival including donor and recipient characteristics could help clinical decision-making and optimize outcomes. The aim of this study was to develop a risk assessment tool for graft survival based on essential pre-transplantation parameters.

Methods

The data originated from the national Dutch registry (NOTR; Nederlandse OrgaanTransplantatie Registratie). A multivariable binary logistic model was used to predict graft survival, corrected for the transplantation era and time after transplantation. Subsequently, a prediction score was calculated from the β-coefficients. For internal validation, derivation (80%) and validation (20%) cohorts were defined. Model performance was assessed with the area under the curve (AUC) of the receiver operating characteristics curve, Hosmer-Lemeshow test and calibration plots.

Results

In total, 1428 transplantations were performed. Ten-year graft survival was 42% for transplantations before 1990, which has improved to the current value of 92%. Over time, significantly more living and pre-emptive transplantations have been performed and overall donor age has increased (P < .05).The prediction model included 71 829 observations of 554 transplantations between 1990 and 2021. Other variables incorporated in the model were recipient age, re-transplantation, number of human leucocyte antigen (HLA) mismatches and cause of kidney failure. The predictive capacity of this model had AUCs of 0.89, 0.79, 0.76 and 0.74 after 1, 5, 10 and 20 years, respectively (P < .01). Calibration plots showed an excellent fit.

Conclusions

This pediatric pre-transplantation risk assessment tool exhibits good performance for predicting graft survival within the Dutch pediatric population. This model might support decision-making regarding donor selection to optimize graft outcomes.

Trial registration

ClinicalTrials.gov Identifier: NCT05388955.

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.

A Statistical Prediction Model for Survival After Kidney Transplantation from Deceased Donors

Background

In an environment of limited kidney donation resources, patient recovery and survival after kidney transplantation (KT) are highly important. We used pre-operative data of kidney recipients to build a statistical model for predicting survivability after kidney transplantation.

Material/Methods

A dataset was constructed from a pool of patients who received a first KT in our hospital. For allogeneic transplantation, all donated kidneys were collected from deceased donors. Logistic regression analysis was used to change continuous variables into dichotomous ones through the creation of appropriate cut-off values. A regression model based on the least absolute shrinkage and selection operator (LASSO) algorithm was used for dimensionality reduction, feature selection, and survivability prediction. We used receiver operating characteristic (ROC) analysis, calibration, and decision curve analysis (DCA) to evaluate the performance and clinical impact of the proposed model. Finally, a 10-fold cross-validation scheme was implemented to verify the model robustness.

Results

We identified 22 potential variables from which 30 features were selected as survivability predictors. The model established based on the LASSO regression algorithm had shown discrimination with an area under curve (AUC) value of 0.690 (95% confidence interval: 0.557–0.823) and good calibration result. DCA demonstrated clinical applicability of the prognostic model when the intervention progressed to the possibility threshold of 2%. An average AUC value of 0.691 was obtained on the validation data.

Conclusions

Our results suggest that the proposed model can predict the mortality risk for patients after kidney transplants and could help kidney specialists choose kidney recipients with better prognosis.