Trajectories of Systolic Blood Pressure Decline in Kidney Donors After Circulatory Death and Delayed Graft Function

Yingxin Lin The University of Sydney Armando Teixeira-Pinto The University of Sydney Jonathan Craig Flinders University of South Australia: Flinders University Helen Opdam Austin Hospital Jeremy C Chapman Westmead Hospital Henry Pleass The University of Sydney Sydney Medical School: The University of Sydney School of Medicine Angus Carter Cairns Hospital Natasha Rogers The University of Sydney Sydney Medical School: The University of Sydney School of Medicine Chris Davies ANZDATA Registry Stephen McDonald ANZDATA Registry Jean Yang The University of Sydney Wai Lim University of Western Australia Germaine Wong (  germaine.wong@health.nsw.gov.au ) University of Sydney https://orcid.org/0000-0001-8422-7269


Introduction
The persistent shortfall in the availability of donor organ kidneys for transplantation has prompted the use of donation after circulatory death (DCD) donor kidneys. Since the introduction of the DCD program in 2006 in Australia, DCD donors represent the fastest growing form of organ donation, now constituting over 30% of all deceased kidney donors 1  Unlike donation after the neurological determination of death process, where the donor kidneys are welloxygenated until the time of cold perfusion, DCD kidneys suffer a period of warm ischemia following withdrawal of circulatory support, which inevitably leads to early ischemic injury 4 . Despite being a valuable source of donor kidneys for transplantation with promising allograft survival rates (at least 75% graft survival rate at ve years after transplantation), DCD kidneys remain under-utilised compared to donation after brain death donors (DBD). One of the key factors that determines organ procurement from DCD donors is the agonal time, de ned as the period between withdrawal of life-sustaining treatment and circulatory arrest. For Maastricht category 3 kidney donors (controlled DCD donors), there are emerging data showing the duration of the agonal phase, combined with the haemodynamic changes and hypoxia during this critical phase are key predictors for adverse outcomes such as primary non-function and delayed graft function (DGF). A single study indicated an increase from 7 to 20 minutes with systolic blood pressure (BP) less than 80 mmHg doubled the risk of DGF, independent of the duration of hypoxia 5 . Others have found that greater area under the curve of the systolic BP decline prior to circulatory arrest were predictive of DGF and early graft loss 6 . Many centres, therefore, have concerns of the potential ischaemic injury during the agonal phase, and donation is deemed unsuitable if cardiorespiratory arrest has not occurred within a certain timeframe of controlled withdrawal of life-supporting treatment. However, the evidence underpinning this speci ed duration is uncertain.
More importantly, progressive hemodynamic changes during the agonal phase may further in uence the risk of DGF 7 . Continuous hemodynamic measures during the pre-donation period allows meticulous assessment of the relationship between dynamic changes of the systolic BP of the potential donors and DGF in transplant recipients. This study aimed to determine the association between trajectories of various patterns of systolic BP decline and DGF in a cohort of kidney transplant recipients of Maastricht category 3 kidney donors. Quanti cation of the variability of blood pressure changes during the agonal time on short-term graft outcomes will provide the necessary evidence to inform decision-making about donor suitability for transplantation.

Methods
This study was reported according to the STROBE guidelines for observational studies 8 .

Study population (derivation cohort)
We included all actual DCD kidney donors that have been transplanted, and their corresponding recipients transplanted in Australia between 9th April 2014 and 2nd January 2018. The Electronic Donor Record (EDR) system was introduced in 2014, therefore details of the pre-donation hemodynamic records were not available prior to 2014. The EDR contains pre-, peri-and post-mortem information of all donors from each state/territory in Australia who were consented for organ donation in Australia and is managed and held by Australian Organ and Tissue Authority in Canberra, Australia. Multi-organ DCD transplants were excluded. Other data were sourced from Australia and New Zealand Dialysis and Transplant (ANZDATA) Registry (recipient data and outcome measures). The institutional board and the human research ethics committee of the University of Western Australia approved the conduct of the study (ethics reference: RA/4/20/4743). Approval for data extraction from the EDR was granted by state and territory health departments and donor agencies. We have assigned this cohort as the derivation cohort because this dataset was provided to the investigator team during the rst data request.

Data collection and linkage
Clinical data from the EDR includes hemodynamic parameters after WCRS such as systolic/diastolic BP (documented every 1-5 minutes using intra-arterial blood pressure monitoring, or in the absence of an arterial line, non-invasive oscillometric method (upper-arm cuff) was used), heart rate (documented every 10 minutes), prior requirement for inotropic/vasopressor support, the time to death, urine output over the 12 hours prior to death, cause of death and terminal serum creatinine. The EDR also contains the baseline characteristics of all actual deceased donors including age, sex, ethnicity, smoking history, personal history of diabetes, hypertension and body mass index (BMI). These data were entered into the EDR as part of routine data capture during the donation process by Donate Life Agency donor coordinator staff in each of the states and territories. The average (standard deviation [SD]) number of repeated hemodynamic measures per donor across the cohort was 9.5 (5.6).
In Australia, absence of circulation is evidenced by absent arterial pulsatility for a minimum of 3 minutes and a maximum of 5 minutes, using intra-arterial pressure monitoring and con rmed by clinical examination (absence of heart sounds and/or central pulses). In cases without an arterial line, electrical asystole is observed for a minimum of 3 minutes and a maximum of 5 minutes on the electrocardiogram and con rmed by clinical examination.
The ANZDATA Registry contains recipient data and includes records of all recipient baseline characteristics of age, sex, ethnicity, smoking status, diabetes, body mass index (BMI), transplanting state, primary cause of kidney failure, comorbidities such as chronic lung disease, history of hepatitis, prior cancer, vascular disease and time on dialysis before transplantation. Other transplant characteristics such as the transplant date, number of human leukocyte antigen (HLA) mismatches and total ischemic time were also recorded in the registry.
Data on DCD donors and the corresponding recipients were collated to allow for the alignment of the donor pre-, peri-and post-mortem events with other donor and recipient characteristics and outcomes measures using unique donor identi er. The de-identi ed dataset was prepared by ANZDATA registry and all linked data between the ANZDATA registry and the EDR (donors and recipients) were not re-identi able. During the timeframe of this study period, machine perfusion was not used routinely across all states and jurisdictions in Australia.
Outcome measures DGF was de ned as recipients who required dialysis within the rst seven days of transplantation.

Statistical analyses
Continuous variables were described using means (SD) and medians (interquartile range [IQR]). Categorial variables were summarised with counts and percentages.

Latent class mixed-effect model
We tted a latent class mixed model to identify classes of systolic BP trajectories 9 . This approach assumed that the individual patterns of systolic BP decline during the agonal phase could be grouped into several patterns of systolic BP decline (latent classes), sharing similar tendency. Latent class analysis (LCA) is a modelling technique based on structural equation model, which aims to identify subgroups of individuals with "unmeasured" class memberships of similar clinical characteristics or outcomes that are not directly observable from the data 10 . The individual was assigned to a particular latent class membership with the highest probability of the outcome measure. When the measurement or outcome of interest is longitudinal, the formulation of the problem will become the identi cation of the subgroups of the developmental trajectories.
In our model, a two stage-approach was applied to link the latent class with the risk of DGF: in the rst stage, the latent classes are estimated based on clinical characteristics; in the second stage, the estimated latent classes were used as predictors for clinical outcomes. The latent class model was tted by considering time as a xed effect and donor characteristics (including sex, age, donor body mass index, donor terminal creatinine, warm ischemia, donor causes of death) as random effects, with systolic BP as the response variable, and took the form of a Gaussian distribution. In addition, the model was tted with class-speci c variance-covariance of the random-effects, with quadratic I-splines with 5 knots as link function to model the non-linear trend. Each trajectory was assigned to a certain latent class with the largest posterior likelihood. The number of latent classes was selected based on the Akaike information criterion (AIC), an estimator of in-sample prediction error, the entropy and by visual inspection of the patterns of trajectory decline in systolic BP 11 .

Characterisation of latent classes and the association with DGF
The comparison of the donor pro les between the latent classes was conducted using one-way analysis of variance (ANOVA) and chi-square tests for continuous and categorical variables, respectively. The association between classes of trajectories of systolic BP decline and DGF was examined using a multivariable logistic regression and adjusted for selected donor and recipient characteristics, and the transplanting states. These variables were selected using two different approaches: Least absolute shrinkage and selection operator (LASSO) 12 and random forest 13 .
Variable selection using LASSO and random forest For LASSO, the penalisation parameter was chosen by cross-validation and the covariates with non-zero coe cient were included in the logistic regression, together with class, and tted using maximum likelihood (that is, LASSO was used solely for variable selection). For the random forest model, we trained the model to predict DGF with 500 trees, using ve randomly chosen variables at each split. We examined the rankings of variables importance in random forest model and selected the variables with the importance scores of greater than 10 to be included in the nal logistic models together with systolic BP trajectory classes.

Effects of the slope and intercept of systolic BP decline on the risk of DGF
We tted a two-stage model to assess the association between donor-derived systolic BP trajectories and its association with DGF.
Details of the two-stage model was denoted as the systolic BP (mmHg) for transplant at time point . First stage was written as , where and were the xed effect parameters, and and were the donor speci c random effect intercept and slope, respectively using lmer function in R package lme4 14 .
In the rst stage, we used a linear mixed-effects model for systolic BP, with a random intercept and random effect for measurement time for each recipient. This longitudinal model allowed characterisation of individual systolic BP trajectories by the donor-speci c intercept and rate of systolic BP decline (slope). We also tested the interaction between classes of systolic BP decline, donor age and DGF, and found donor age was not an effect modi er between the speci c systolic BP latent classes and DGF.
In the next stage, the donor-speci c intercept and rate of systolic BP decline were used in a multivariable logistic model for DGF, adjusted for other donor and recipient characteristics, and the transplanting states. These covariates were selected using similar methods to those described for the latent class approach (LASSO and random forest). The second stage model used the slope of systolic BP estimated from Stage 1. This was written as: , where was the coe cient for the estimated slope of systolic BP; and included the other variables in the multivariate logistic regression model. These variables were similar to those integrated in the latent class analyses and were the vectors of the logistic regression coe cients.

Classi cation DGF and model evaluation
We compared the additional performance gained from the latent class trajectories, slope and intercept of systolic BP in addition to traditional clinical variables using random forest modelling. Each model was evaluated using 5fold cross validation and repeated by 50 times. The accuracy rate and the Area Under the Receiving Operator Curve (AUC-ROC) were used to evaluate the performance of the models.

Validation cohort
Validation of the ndings from the derivation cohort (the 8 latent classes, slopes and intercept of systolic BP) was conducted using data from all Australian DCD donors between 6th January 2018 to 24th December 2019.
This was made available to the investigators during the second data request.

Sensitivity analyses
Using similar modelling strategies of latent class modelling, variable selections using Random Forest and LASSO, logistic regression and model evaluation; we also assessed the association between trajectories of diastolic BP and DGF. Given the AIC and entropy with 8 and 9 latent classes were similar, we then evaluated the association between 9 different classes of systolic BP trajectories and DGF in the sensitivity analyses.

Derivation cohort
Characteristics of the donors A total of 472 actual DCD kidney donors between January 2014 and January 2018 was available for analysis. We included 462 in the latent class analyses, after excluding 10 donors with one haemodynamic entry ( Figure 1).
Most donors were male (62.8%), with a mean (SD) age of 47.6 (16.4) years. The mean (SD) terminal serum creatinine was 101 (107.6) µmol/L, and the average (SD) duration of withdrawal of cardio-respiratory support (WCRS) was 22.3 (15.7) minutes. Approximately 50% required inotropic support prior to WCRS and only 5.7% experienced oliguria in the preceding 24 hours prior to the declaration of death. The most common causes of death were cerebral hypoxia (39.2%), followed by intracranial bleed (24.5%) and traumatic brain injury (16.9%) ( Table 1).
Characteristics of kidney transplant recipients ( Table 2) A total of 696 transplant recipients (from 379 donors) were included, and 380 (54.6%) had DGF. Approximately 60% of the recipients were men, most (70%) were European Australians. The major primary causes of kidney failure were glomerulonephritis, followed by diabetes mellitus and polycystic kidney disease. Around 20% had prevalent vascular disease at the time of transplantation. The maximum number of patterns examined was eleven and the number of patterns that best captured the trajectories of systolic BP decline was identi ed as eight (Supplementary Figure 1). The characteristics of the eight patterns are shown in Table 1 and Figure 2. There were signi cant differences in the WCRS times, intercept and slope of systolic BP decline, donor age, use of inotropic support during the pre-terminal phases, presence of oliguria in the last 12 hours preceding death, total ischemic time, and the causes of death between these classes of systolic BP decline. Donors that belonged to trajectory class 8 had the longest WCRS time (mean (SD): 60 (23) mins) and the slowest decline in systolic BP (-1.6 (0.9) mmHg/min), whereas those that were assigned to class one had the shortest WCRS duration (12 (3)

Performance of the classi cation models for DGF
The model that included the slope of systolic BP decline had the highest accuracy rate of the cross validation with an average of 65.2% (for Random Forest selected variables) and 64.7% (for LASSO selected variables) compared to the models that included only clinical variables (65.0% and 64.1%) (Figure 7). The AUC-ROC for the models that included the slope was 0.72 (0.68-0.76), while AUC-ROC for the models that included only the clinical variables was 0.7 (0.66-0.74).

Sensitivity analyses
A total of 6 different classes of diastolic BP trajectories were identi ed using latent class modelling (Supplementary Figure 3). The factors for de ning these separate classes were intercept of the diastolic BP (initial diastolic BP at the time of WCRS), slope of diastolic BP decline and duration of withdrawal of cardiorespiratory support. There was no association between the various trajectories of diastolic BP decline and DGF, adjusted for variables selected by both LASSO and Random Forest (Supplementary Figure 4).  Tables 4 and 5). The accuracy rate for the validation cohort of the classi cation models trained on derivation cohort is 59.6% (for Random Forest selected variables) and 60.7% (for LASSO selected variables) for the model that included the slope of systolic BP decline, indicating moderate discriminative ability, compared to the models that included only clinical variables (60.4% and 59.6%).

Main ndings
Using large development and validation datasets of recipients who have received DCD kidneys in Australia, we have shown that the slope of systolic BP decline and random intercept (initial systolic BP at the time of WCRS) are key factors that de ne the patterns of systolic BP decline during the agonal phase in our derivation and validation cohorts. For every 1 mmHg per minute reduction in the rate of decline of systolic BP, there was approximately a 1-5% reduction in the risk of DGF, but improvement in model performance was only modest with inclusion of the slope of systolic BP decline compared to using clinical variables alone. A longer agonal time was not associated with increased risk of DGF compared to donors with shorter agonal time, in both the derivation and validation cohorts. Our research extends current knowledge by evaluating not only the duration of the agonal phase and blood pressure thresholds in isolation, but also the overall pattern and archetype of the rates of blood pressure decline over time in DCD donors using advanced statistical modelling techniques. This nding has provided the necessary evidence to safely extend and standardise the accepted agonal time to at least 90 minutes for DCD donors.

Comparison To Previous Studies
This novel nding differs from previous investigations, which reported a longer duration of functional warm ischemia, de ned as the rst recorded time of systolic BP less than 50 mmHg to cold perfusion, was associated with slow graft function, and premature graft loss in DCD kidneys 15 . Prior work have indicated that a longer agonal time was associated with a signi cantly increased risk of DGF and primary non-function by 50% 16 , particularly in donors where the duration of systolic BP less than 80 mmHg exceeded 20 minutes before cold perfusion, with this effect more apparent in younger donors (aged < 40 years) 5 . However, there was no consistent relationship between the duration of hypotension in the agonal phase and allograft failure.
The explanation behind the discordance with our ndings may be explained by rapidity of the blood pressure decline on the severity of ischemic-reperfusion injury (IRI) of the donor graft 17 . Ischemia, resulting from the extremes of hemodynamic changes in the deceased donor during the pre-donation phase, may lead to injury of the renal tubular, epithelial and endothelial cells 18 . This process can then initiate cytokine/chemokine release and recruits immune cells, including neutrophils, monocytes/macrophages and natural killer T cells, contributing to in ammatory responses and subsequent damage to the allograft 18 . The relationship between hypotension and the risk of acute kidney injury is also well documented in animal studies. Experimental work has shown that under normal physiological conditions, glomerular ltration rate (GFR) and renal blood ow are adequately autoregulated over a wide range of systolic BP. However, if systolic BP falls below the lower thresholds of 60-70 mmHg, renal blood ow auto-regulation will be lost, causing stimulation of the renin-angiotensin-aldosterone system, promoting renin release from granular cells of the renal juxtaglomerular apparatus to maintain perfusion pressure 19 . In the context of rapid blood pressure lowering, severe dysfunction of autoregulation of the afferent renal arterioles will occur, contributing to signi cant glomerular injury through reduction of renal blood ow to the compromised kidneys 20 . It is important to emphasize that even though autoregulation of renal blood ow is maintained during periods of systemic hypotension, intra-glomerular pressures of DCD donors are unpredictable during prolonged periods of severe hypotension, particularly among those that required inotropic support 21 .
Other methodological differences which may explain the discrepant results between our study and prior research include dissimilar modelling approaches, characteristics of our study cohort and our sample size was very large in comparison to previous research.

Implications For Practice
Our study ndings have important clinical implications. Currently, the duration of functional warm ischemic time for the kidneys is highly variable between countries, without robust evidence to support guideline recommendations. In the UK, organ retrieval will not occur if the time interval from the onset of functional warm ischemia is greater than 120 minutes 22 . In Spain and Belgium, the donation sector would accept a withdrawal of cardio-respiratory support time of up to 2 hours before abandoning the process of donation 23 . In Australia, the time varies between 60-90 minutes, depending on the state jurisdiction 22 . Many of the DCD donors (approximately 40%) do not proceed to donation because they do not die within an acceptable timeframe after withdrawal of life support 24 . We have challenged the current belief and have shown that an extended duration of warm ischemia (at least up to 90 minutes) may not necessarily be harmful to the donor grafts, provided the decline in the blood pressure parameters is gradual over time. These data are important because it has unmasked potential opportunities to increase the donor pool of DCD kidneys by reducing discard rates, which are approximately 5-8% in Australia, but up to 15% in the US.

Strengths And Limitations
This study has some strengths. It is one of the largest cohorts of DCD donors with near complete repeated measures of hemodynamic parameters during the pre-donation phase. Unlike other registries, reporting and recording of details of the blood pressure decline of the donor organs are mandatory for all states and territory in Australia. Therefore, the current ndings are re ective and generalizable to a diverse range of DCD kidneys. Using innovative, rigorous machine learning techniques and statistical modelling, we were able to de ne, for the rst time, the various trajectories of systolic BP decline during the agonal phase of a national cohort of DCD donors. We have conducted extensive sensitivity analyses using different machine learning strategies for variable selections and our results are robust against a range of potential confounders, different classes of systolic BP trajectories, and the ndings are consistent also for diastolic BP trajectories, This study has some potential limitations. Longitudinal data on oxygen saturation concentrations in our donors was not available, so we were unable to examine the in uence of hypoxia on DGF in the modelling. Confounding from unmeasured factors may also exist. We have only included those kidney donors that have progressed to donation and transplantation in the analyses and were unable to perform a direct comparison to donors that have not progressed, or non-utilised. We also do not have details of the second warm ischemia and duration of the anastomotic time from the ANZDATA Registry. Machine perfusion, unlike other countries, is not routine practice in Australia. Given the relatively short follow-up times, we did not have su cient power and events to assess for medium and long-term graft survival. Future studies to examine the causal effects of pre-donation hemodynamic changes, DGF and graft survival are warranted.

Conclusions
In conclusion, we have de ned the distinct and unique classes of systolic BP trajectories of DCD kidneys during the agonal phase, and the group membership, determined by the initial and rate of systolic BP decline from the time of withdrawal of circulatory support in both the derivation and validation cohorts. Notwithstanding potential injurious effects of prolonged functional warm ischemia prior to perfusion, we have shown that a slower decline in systolic BP during the agonal phase may not be harmful to the allograft in the short-term. It is therefore conceivable that the maximum allowable duration of agonal phase could be extended safely in the event of gradual donor systolic BP decline without necessarily compromising short-term allograft outcome. The institutional board and the human research ethics committee of the University of Western Australia approved the conduct of the study (ethics reference: RA/4/20/4743). Approval for data extraction from the EDR was granted by state and territory health departments and donor agencies.   Figure 1 Cohort ow of DCD donors of the derivation cohort   Association between various classes of the trajectories of systolic BP decline and DGF using Random Forest and Lasso selected variables (Derivation cohort)

Figure 6
Association between the slope of decline of systolic blood pressure and DGF (Random Forest and Lasso selected variables in the derivation and validation cohorts) Performance of the derivation model using random forest and LASSO selected features

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.