The interval between brainstem death and cardiac assessment influences the retrieval of hearts for transplantation

The interval between brainstem death and cardiac assessment influences the retrieval of hearts... Abstract OBJECTIVES The optimum time after brainstem death (BSD) at which to assess the function of donor hearts is unknown. We hypothesized that a longer interval may be associated with a higher transplantation rate due to improved function. METHODS Data were obtained from the UK Transplant Registry for the period between April 2010 and March 2015. The time when fixed dilated pupils were first noted in the donor was considered as the time of BSD. Retrieval was defined as the time when the abdominal organs were surgically perfused. RESULTS BSD to retrieval duration was available for 1947 donors, of which 458 (24%) donated their heart. In the univariable analysis (not adjusting other donor risk factors), evidence was available to suggest that the BSD to cardiac assessment duration had a non-linear association with heart utilization (P < 0.0001). Adjusting for donor risk factors, the relationship remained with longer intervals being associated with increased transplantation (P = 0.0056). The modelled probability of heart utilization had a similar pattern to the observed rate of heart utilization. However, the probability of heart donation began to plateau after approximately 48 h. The analysis of the subset of donors attended by a cardiothoracic retrieval team showed a similar pattern. CONCLUSIONS These data suggest that time interval from BSD to organ retrieval influences the heart retrieval rate. When the sole reason for declining a donor heart is poor function, a period of further observation and optimization up to 2 days should be considered. Brain death, Heart transplantation/mortality, Tissue and organ harvesting, Risk assessment INTRODUCTION Heart transplantation remains the definitive long-term treatment for advanced cardiac failure irrespective of cause. A discrepancy between required and available organs persists internationally with donation after brain death being the main source of donor hearts. The catecholamine storm associated with brainstem death (BSD) may have a detrimental effect on donor heart function via complex cardiovascular changes, endocrine and metabolic disturbances and release of proinflammatory signals [1]. There is a window of opportunity after BSD for organs to be optimized and retrieved for transplantation before irreversible damage occurs. Early active donor management has been shown to exert a positive influence on this process without impacting on post-transplant recipient outcome [2]. Given a lack of evidence about the optimal interval after BSD that allows recovery from the deleterious effects of the catecholamine storm, the first objective of this study was to assess whether a longer interval may be associated with a higher transplantation rate in a large cohort of donors. The second objective was to investigate the relationship of the BSD interval with the post-transplant recipient outcome. MATERIALS AND METHODS Data collection and analysis In the UK, a National Organ Retrieval Service was introduced in April 2010, and since then, the UK Transplant Registry (UKTR) data have been collected on the retrieval process. Follow-up data on 30-day survival within the cohort are 100% complete, and the accuracy and consistency of the data are maintained by regular computer-based and case–record validations. Data are stored on the UKTR that is maintained by the NHS Blood and Transplant. The data analysis was carried out using the SAS version 9.4 software. In this article, BSD is defined as the time when fixed dilated pupils were first noted in the donors, and the BSD interval is defined as the time from BSD to cardiac assessment at organ retrieval. Outcomes We investigated the impact of the BSD interval on 2 primary outcomes: (i) heart utilization and (ii) post-heart transplant survival. Heart utilization Heart utilization is defined as donor hearts retrieved with the intention of transplantation. Data on adult organ donors after brain death—also described on the literature as donors after neurological determination of death—were obtained retrospectively from the UKTR for the 5-year period from 1 April 2010 to 31 March 2015. Donors older than 65 years with a history of cardiothoracic disease, donors with no consent for heart donation and donors whose cause of death was myocardial infarction were excluded. Post-heart transplant survival To further assess the impact of the BSD interval on recipient survival, we analysed adult patients who received a heart transplant at any of the 6 designated UK centres where the heart donors were part of the heart utilization cohort previously described. We analysed short-term survival (30 and 90 days) and medium-term survival (1 and 3 years). To ensure a homogeneous transplant cohort, we excluded multiorgan transplants, retransplantations, paediatric patient transplants and donor hearts that had used the Organ Care System (OCS; TransMedics, Inc., Andover, MA, USA) for perfusion. Missing data Missing data for risk factors (maximum 18%: Tables 1 and 2) identified in the analysis were imputed with multiple imputation fully conditional specification method. Missing BSD intervals were not imputed and so the final cohort used 84% of the full cohort. Table 1: Donor risk factors and categories used in the risk-adjusted heart utilization model n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with heart utilization by the multivariable stepwise logistic regression model selection. BMI: body mass index; IQR: interquartile range; LRT: likelihood ratio test. Table 1: Donor risk factors and categories used in the risk-adjusted heart utilization model n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with heart utilization by the multivariable stepwise logistic regression model selection. BMI: body mass index; IQR: interquartile range; LRT: likelihood ratio test. Table 2: Donor, transplant and recipient risk factors and categories used in the risk-adjusted 30-day, 90-day, 1-year and 3-year survival models, with hazard ratios and P-values from the 30-day model n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with 30-day post-heart transplant survival by the multivariable stepwise Cox proportional hazards regression model selection. BMI: body mass index; ECMO: extracorporeal membrane oxygenation; IQR: interquartile range; LRT: likelihood ratio test; MCS: mechanical circulatory support. Table 2: Donor, transplant and recipient risk factors and categories used in the risk-adjusted 30-day, 90-day, 1-year and 3-year survival models, with hazard ratios and P-values from the 30-day model n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with 30-day post-heart transplant survival by the multivariable stepwise Cox proportional hazards regression model selection. BMI: body mass index; ECMO: extracorporeal membrane oxygenation; IQR: interquartile range; LRT: likelihood ratio test; MCS: mechanical circulatory support. Statistical analysis Heart utilization We used the multivariable logistic regression to assess whether the BSD interval is associated with the retrieval of hearts for transplantation, after adjusting for known risk factors. A stepwise variable selection method was used where candidate risk factors were retained in the model if they reduced the model deviance significantly (P < 0.1) according to the likelihood ratio test. Donor risk factors included were blood group, sex, history of hypertension, use of prednisolone, cardiac arrest, age, body mass index (BMI), heart rate, history of diabetes, history of alcohol abuse and smoking history. The BSD interval was tested in the model as both a linear term and a non-linear term. A linear term assumes a relationship between the BSD interval and the probability of utilization that is constant as the BSD interval increases, whereas a non-linear term allows the relationship to vary depending on the BSD interval. We used a natural cubic spline method to explore non-linearity, allowing cubic expressions between intervals defined by ‘knots’ at the 5%, 35%, 65% and 95% percentiles. The model was internally validated using the Hosmer and Lemeshow measure of concordance. The concordance statistic of the heart utilization model was 0.758, indicating acceptable predictive ability. Post-heart transplant survival We used multivariable Cox proportional hazards regression to assess whether the BSD intervals associated with patient survival time post-transplant, after adjusting for known risk factors. Known risk factors were identified in the adult heart risk-adjusted models that were developed for the Annual Report on Cardiothoracic Transplantation, published by the NHSBT in July 2016. These include donor cause of death, donor BMI, donor age, donor respiratory arrest, recipient BMI, recipient creatinine at transplant, if recipient had extracorporeal membrane oxygenation at transplant (for short-term survival models only), if recipient had ventricular assist device at transplant (for medium-term survival models only), if recipient was in-hospital before the transplant, recipient primary disease, recipient and donor sex mismatch and ischaemia time. To check the proportional hazards assumption on which the Cox regression model is based, log cumulative hazard plots and Schoenfeld residual plots were used. There was no evidence to suggest violation of the proportional hazard assumption in the model of interest. The BSD interval was considered as both a linear term and a non-linear term in the model using the natural cubic spline method previously described. Combining the heart utilization and post-heart transplant survival model To explore the results of the 2 models in combination, we interrogated how the utilization/survival probabilities changed as the BSD interval varied, while retaining all other variables in the models fixed. This was achieved by selecting common or average donor and recipient characteristics. RESULTS From the donor inclusion criteria, there were 1947 records. Of these, 1354 (70%) did not proceed to heart donation. Out of 593 heart donors, only 458 subjects fulfilled all inclusion criteria and were further analysed in the post-transplant analysis (Fig. 1). The observed median BSD interval for the donor hearts retrieved and not retrieved was 39 and 35 h, respectively (Fig. 2). Figure 1: View largeDownload slide Data and analysis flow chart. DBD: donors after brain death. Figure 1: View largeDownload slide Data and analysis flow chart. DBD: donors after brain death. Figure 2: View largeDownload slide Box plots of the BSD interval for non-heart donors and heart donors. BSD: brainstem death. Figure 2: View largeDownload slide Box plots of the BSD interval for non-heart donors and heart donors. BSD: brainstem death. Heart utilization In the univariable analysis (not adjusting any donor risk factors), evidence was available to suggest that the BSD interval had a non-linear association with heart utilization; longer intervals being associated with increased utilization (overall effect P < 0.0001, non-linearity P = 0.009). Adjusting for donor risk factors listed in Table 1, the duration still had this non-linear association with heart utilization (overall effect P = 0.0056, non-linearity P = 0.009). The modelled probability of heart utilization has a similar trend to the observed rate of heart utilization within the deciles of the BSD interval. Specifically, longer durations were associated with a higher probability of donation. However, the probability of heart donation began to plateau after approximately 48 h indicating that durations beyond this were not associated with the chance of heart utilization (Fig. 3). Figure 3: View largeDownload slide (A) The risk-adjusted probability of heart utilization from the logistic regression model plotted against the BSD interval. Each grey dot is the estimated probability of heart utilization for a donor by the actual BSD interval for that donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of heart utilization and the BSD interval. (B) The observed rate of heart utilization within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of heart utilization and the BSD interval. BSD: brainstem death. Figure 3: View largeDownload slide (A) The risk-adjusted probability of heart utilization from the logistic regression model plotted against the BSD interval. Each grey dot is the estimated probability of heart utilization for a donor by the actual BSD interval for that donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of heart utilization and the BSD interval. (B) The observed rate of heart utilization within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of heart utilization and the BSD interval. BSD: brainstem death. As a sensitivity analysis, the model was refitted using the donors with complete data for all risk factors (n = 1883). The BSD interval also had a significant non-linear association with heart utilization in this cohort. To explore the impact of donor optimization, we excluded donors who were not attended by a cardiothoracic retrieval team and performed an analysis on a subset of donors who were attended by the cardiothoracic retrieval team only (n = 1306). Analysing this with the same methodology, the BSD interval has a significant association with heart utilization (P = 0.007). The effect of the BSD interval had a similar pattern. Post-heart transplant survival In the univariable post-transplant analysis, evidence was available to suggest that the BSD interval had a non-linear association with 30-day patient survival only (overall effect P = 0.013, non-linearity P = 0.005). Adjusting for the donor, recipient and transplant risk factors in Table 2, the duration still had a non-linear association with 30-day patient survival (overall effect P = 0.04, non-linearity P = 0.017) but not longer term survival (P ≥ 0.19). The modelled 30-day patient survival probabilities had a similar trend to the observed patient survival rates within the deciles of the BSD interval obtained using the Kaplan–Meier method. For patients with transplanted hearts retrieved less than 36 h after BSD, the estimated 30-day survival has a decreasing trend with increasing duration to approximately 90% survival (Fig. 4). For patients in whom transplanted hearts were retrieved between 36 and 72 h after BSD, the estimated 30-day survival increased from approximately 90% to 95%. For patients in whom transplanted hearts were retrieved after 72 h, the estimated 30-day survival decreased again; however, the confidence interval of the LOESS trend line was a lot wider, as there were fewer transplants with a heart retrieved more than 72 h after BSD. Figure 4: View largeDownload slide (A) The risk-adjusted probability of 30-day survival from the Cox proportional hazards regression model plotted against the BSD interval. Each grey dot is the estimated probability of 30-day survival for a patient by the actual BSD interval for their donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of 30-day survival and the BSD interval. (B) The observed (Kaplan–Meier) 30-day survival rate within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of 30-day survival and the BSD interval. BSD: brainstem death. Figure 4: View largeDownload slide (A) The risk-adjusted probability of 30-day survival from the Cox proportional hazards regression model plotted against the BSD interval. Each grey dot is the estimated probability of 30-day survival for a patient by the actual BSD interval for their donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of 30-day survival and the BSD interval. (B) The observed (Kaplan–Meier) 30-day survival rate within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of 30-day survival and the BSD interval. BSD: brainstem death. As a sensitivity analysis, the model was refitted using the transplants with complete data for all risk factors (n = 374). The BSD interval was also found to be significantly associated with 30-day patient survival times but not with 90-day, 1-year or 3-year patient survival times. Combining the heart utilization and 30-day post-heart transplant survival model The association between the BSD interval and both heart utilization and the 30-day post-heart transplant patient survival is significant. Therefore, it is of interest to obtain an optimum window for the start of the retrieval operation after BSD, using the results of the 2 statistical models. For a male donor, aged 44 years, with BMI 26 kg/m2, blood group A, no history of hypertension, no cardiac arrest, heart rate 91 bpm, no history of diabetes, no history of alcohol abuse and was a past smoker, the probability of heart utilization is optimized from approximately 50 h of BSD interval (Fig. 5). For a male patient, with no respiratory arrest, no extracorporeal membrane oxygenation required, was in-hospital before transplant and had coronary heart disease, receiving a heart from a male donor, aged 44 years, with BMI 26 kg/m2, who died due to vascular disease, the probability of 30-day survival is very high from approximately 70 h of BSD interval onwards. Figure 5: View largeDownload slide The change in estimated probability of heart utilization/30-day survival as BSD interval increases for a typical donor/patient. The solid line shows how the probability of the heart being utilized varies for a male donor, aged 44 years, with body mass index 26 kg/m2, blood group A, no history of hypertension, no cardiac arrest, heart rate 91 bpm, no history of diabetes, no history of alcohol abuse and was a past smoker, with its 95% CL. The dashed line shows how the probability of 30-day survival varies for a male patient, with no respiratory arrest, no extracorporeal membrane oxygenation required, was in-hospital before transplant and had coronary heart disease, receiving a heart from a male donor, aged 44 years, with body mass index 26 kg/m2, who died from vascular disease, with its 95% CLs. BSD: brainstem death; CL: confidence limit. Figure 5: View largeDownload slide The change in estimated probability of heart utilization/30-day survival as BSD interval increases for a typical donor/patient. The solid line shows how the probability of the heart being utilized varies for a male donor, aged 44 years, with body mass index 26 kg/m2, blood group A, no history of hypertension, no cardiac arrest, heart rate 91 bpm, no history of diabetes, no history of alcohol abuse and was a past smoker, with its 95% CL. The dashed line shows how the probability of 30-day survival varies for a male patient, with no respiratory arrest, no extracorporeal membrane oxygenation required, was in-hospital before transplant and had coronary heart disease, receiving a heart from a male donor, aged 44 years, with body mass index 26 kg/m2, who died from vascular disease, with its 95% CLs. BSD: brainstem death; CL: confidence limit. DISCUSSION Heart transplantation is a successful treatment for advanced heart failure but is limited by the scarcity of hearts judged suitable for transplantation. Only a minority of hearts offered for transplantation is actually used. The selection of hearts is based partly on cardiac function, but this is known to change with time after BSD. Here, we demonstrate for the first time, in a large cohort of cardiac donors, that an increased interval between BSD and organ retrieval increases the probability of heart utilization up to an interval of 48 h. Furthermore, an increased interval did not affect overall post-transplant survival. The time of procurement can be utilized as the distinct time point separating pre- from post-organ retrieval events. It is of essence to further break down the variable preretrieval period into specific time periods and study their impact on retrieval and long-term outcomes. As all relevant time points could not be obtained from our database, calculated indirect estimations were used (Fig. 6). Time from injury to BSD is known as the brain injury interval, and time from BSD to procurement is known as BSD interval. As exact time of injury is difficult to obtain in majority of cases, many groups use the time of intubation. Similarly, the exact time of BSD cannot be used as an objective hard time point with time of the second assessment of BSD being used as an approximation. We chose to use the documented time of coning (detection of fixed dilated pupils) as a more realistic time point indicative of BSD. The interval from procurement to implantation is subdivided into cold ischaemia and warm ischaemia times and has been extensively studied [3, 4]. Figure 6: View largeDownload slide Diagram of the BSD interval analysed and justification for using fixed pupil dilation. BSD: brainstem death; IQR: interquartile range. Figure 6: View largeDownload slide Diagram of the BSD interval analysed and justification for using fixed pupil dilation. BSD: brainstem death; IQR: interquartile range. In the UKTR, coning time was reported only within 37% of the donor cohort, whereas time of fixed and dilated pupils were noted in 86%. To assess whether time of fixed and dilated pupils is appropriate to represent BSD/coning, the difference between the 2 time points were compared with the cases where both are reported. The median difference between the time of coning and fixed and dilated pupil first noted was only 1 h (interquartile range 0–10 h) (n = 780). On the other hand, brainstem test time was reported in 97% of cases; however, the median of the difference between coning and the first brainstem test performed was 12 h (interquartile range: 6–20 h) (n = 866). The difference in time of coning and the first brainstem test was 1-10 h in 41%, and 11–30 h in 46%. Hence, the time of fixed and dilated pupils was used as a time point indicative of BSD. In the trauma setting, increasing brain death intervals were not associated with decreased organ procurement rates or an increased number of non-salvageable organs due to poor function [5]. Data sub-analysis actually showed an increased rate of procurement for heart and pancreas with increasing brain death intervals. In a retrospective study of 215 heart transplants, Marasco et al. [6] did not find an association between brain death interval and mortality. The most important finding of this study was the strong correlation between brain death interval and hypoxia inducible factor 1a activity in donor atrial tissue with the authors suggesting a potential role in ameliorating the detrimental effects of the catecholamine storm. A few investigators have assessed the influence of the brain death interval on outcomes after heart transplantation. The Stanford groups has reported on adverse survival trends with management times greater than 66 h, defined as the sum of the brain injury and brain death intervals [7]. No difference in survival was observed in relation to brain injury or brain death intervals when assessed individually. Solomon et al. [8] did not demonstrate any statistical significance in outcomes in relation to brain death interval, although all intervals were <12 h. In a paediatric setting, Odim et al. [9] demonstrated a trend towards improved rejection-free survival with longer brain death intervals (35–90 h) that did not achieve statistical significance. An increased risk of mortality for heart transplantation patients was observed with prolonged brain death intervals in a limited study from Manchester [10]. BSD pathophysiology involves a cascade of cardiovascular, respiratory, endocrine, metabolic and stress responses that may jeopardize potentially transplantable organs [11]. Catecholamine levels are elevated several folds following brain death in humans with colocalized apoptotic and necrotic damage observed in 40% of donors [12]. In more detail, investigators reported histological evidence of damaged heart tissue in 84%, with the affected area never corresponding to >5% of the overall tissue and thus unlikely to have had any major detrimental effect on post-transplant heart function according to the authors. In the context of marginal donor hearts, load-independent indices of cardiac function should be assessed as haemodynamic instability alone is not a reliable indicator of subsequent primary organ dysfunction [13]. The explosive increase in intracranial pressure also may have a detrimental effect on the myocardium [14]. If the sympathetic storm builds up in a gradual nature, haemodynamic changes and structural myocardial damage are less pronounced. In terms of post-transplantation outcomes, there is evidence that brain death is also associated with the activation of pro-inflammatory mediators or even alteration of myocardial gene expression [15, 16]. Identifying molecular pathways regulating this process may provide therapeutic targets to modify the aftermath of the severe stress response in brain death [17]. Donor management has been the focus of many groups to increase the yield of transplantable organs [2, 18, 19]. Singbartl et al. [20] reported an increase in organ recovery for transplantation via implementation of intensivist-led management of brain-dead donors, although these findings did not achieve statistical significance regarding the number of hearts transplanted. Recommendations for optimal medical management of the heart-beating brain-dead donor to optimize organ potential have been published [11, 21]. During the study period, the probability of heart donation began to plateau after approximately 48 h with optimal 30-day survival rate noted at 72 h. Although interpretation of these findings remains descriptive with no exact knowledge of underlying pathophysiological mechanisms, it seems that ‘riding out’ the storm has a dual clinical benefit. First, potential donors are allowed to self-select as organs irreversibly damaged will manifest clinical deterioration. Second, during this period, there seems to be a window of donor optimization that may be amplified by donor management that has a prognostic impact on long-term outcomes. Although not assessed, we believe that there may be financial benefits associated with such a strategy. Limitations A number of limitations are inherent to this study. First, the data reflect a national clinical registry. Even though these data provide an insight into real-world practice, it is conceivable that there are unmeasured confounding variables. Second, although we attempted to resolve missing data and transcriptional errors using reproducible algorithms and statistical imputation methodology, we have potentially underestimated the standard errors, and wrongly assumed data are missing at random. However, not imputing the variable of interest and conducting sensitivity analyses have mitigated against the risk of misrepresentation. Third, the LOESS method was used to visualize the trends obtained from our models, but the interpretation is limited due to substantial variation of the model estimates around the LOESS curves. Finally, the impact of donor management-specific strategies has not been assessed in this study. Conflict of interest: none declared. REFERENCES 1 Mascia L , Mastromauro I , Viberti S , Vincenzi M , Zanello M. Management to optimize organ procurement in brain dead donors . Minerva Anestesiol 2009 ; 75 : 125 – 33 . Google Scholar PubMed 2 Abuanzeh R , Hashmi F , Dimarakis I , Khasati N , Machaal A , Yonan N et al. Early donor management increases the retrieval rate of hearts for transplantation in marginal donors . Eur J Cardiothorac Surg 2015 ; 47 : 72 – 7 ; discussion 77. Google Scholar CrossRef Search ADS PubMed 3 Banner NR , Thomas HL , Curnow E , Hussey JC , Rogers CA , Bonser RS et al. The importance of cold and warm cardiac ischemia for survival after heart transplantation . Transplantation 2008 ; 86 : 542 – 7 . Google Scholar CrossRef Search ADS PubMed 4 Marasco SF , Kras A , Schulberg E , Vale M , Lee GA. Impact of warm ischemia time on survival after heart transplantation . Transplant Proc 2012 ; 44 : 1385 – 9 . Google Scholar CrossRef Search ADS PubMed 5 Inaba K , Branco BC , Lam L , Salim A , Talving P , Plurad D et al. Organ donation and time to procurement: late is not too late . J Trauma 2010 ; 68 : 1362 – 6 . Google Scholar CrossRef Search ADS PubMed 6 Marasco S , Kras A , Schulberg E , Vale M , Chan P , Lee GA et al. Donor brain death time and impact on outcomes in heart transplantation . Transplant Proc 2013 ; 45 : 33 – 7 . Google Scholar CrossRef Search ADS PubMed 7 Cantin B , Kwok BW , Chan MC , Valantine HA , Oyer PE , Robbins RC et al. The impact of brain death on survival after heart transplantation: time is of the essence . Transplantation 2003 ; 76 : 1275 – 9 . Google Scholar CrossRef Search ADS PubMed 8 Solomon NA , McGiven JR , Alison PM , Ruygrok PN , Haydock DA , Coverdale HA et al. Changing donor and recipient demographics in a heart transplantation program: influence on early outcome . Ann Thorac Surg 2004 ; 77 : 2096 – 102 . Google Scholar CrossRef Search ADS PubMed 9 Odim J , Laks H , Banerji A , Mukherjee K , Vincent C , Murphy C et al. Does duration of donor brain injury affect outcome after orthotopic pediatric heart transplantation? J Thorac Cardiovasc Surg 2005 ; 130 : 187 – 93 . Google Scholar CrossRef Search ADS PubMed 10 Ramjug S , Hussain N , Yonan N. Prolonged time between donor brain death and organ retrieval results in an increased risk of mortality in cardiac transplant recipients . Interact CardioVasc Thorac Surg 2011 ; 12 : 938 – 42 . Google Scholar CrossRef Search ADS PubMed 11 McKeown DW , Bonser RS , Kellum JA. Management of the heartbeating brain-dead organ donor . Br J Anaesth 2012 ; 108(Suppl 1) : i96 – 107 . Google Scholar CrossRef Search ADS PubMed 12 Pérez López S , Otero Hernández J , Vázquez Moreno N , Escudero Augusto D , Álvarez Menéndez F , Astudillo González A. Brain death effects on catecholamine levels and subsequent cardiac damage assessed in organ donors . J Heart Lung Transplant 2009 ; 28 : 815 – 20 . Google Scholar CrossRef Search ADS PubMed 13 Szabo G. Physiologic changes after brain death . J Heart Lung Transplant 2004 ; 23 : S223 – 6 . Google Scholar CrossRef Search ADS PubMed 14 Shivalkar B , Van Loon J , Wieland W , Tjandra-Maga TB , Borgers M , Plets C et al. Variable effects of explosive or gradual increase of intracranial pressure on myocardial structure and function . Circulation 1993 ; 87 : 230 – 9 . Google Scholar CrossRef Search ADS PubMed 15 Wilhelm MJ , Pratschke J , Beato F , Taal M , Laskowski IA , Paz DM et al. Activation of proinflammatory mediators in heart transplants from brain-dead donors: evidence from a model of chronic rat cardiac allograft rejection . Transplant Proc 2002 ; 34 : 2359 – 60 . Google Scholar CrossRef Search ADS PubMed 16 Yeh T Jr , Wechsler AS , Graham LJ , Loesser KE , Sica DA , Wolfe L et al. Acute brain death alters left ventricular myocardial gene expression . J Thorac Cardiovasc Surg 1999 ; 117 : 365 – 74 . Google Scholar CrossRef Search ADS PubMed 17 Zhai W , Feng R , Huo L , Li J , Zhang S. Mechanism of the protective effects of N-acetylcysteine on the heart of brain-dead Ba-Ma miniature pigs . J Heart Lung Transplant 2009 ; 28 : 944 – 9 . Google Scholar CrossRef Search ADS PubMed 18 Wheeldon DR , Potter CD , Oduro A , Wallwork J , Large SR. Transforming the “unacceptable” donor: outcomes from the adoption of a standardized donor management technique . J Heart Lung Transplant 1995 ; 14 : 734 – 42 . Google Scholar PubMed 19 Rosendale JD , Kauffman HM , McBride MA , Chabalewski FL , Zaroff JG , Garrity ER et al. Hormonal resuscitation yields more transplanted hearts, with improved early function . Transplantation 2003 ; 75 : 1336 – 41 . Google Scholar CrossRef Search ADS PubMed 20 Singbartl K , Murugan R , Kaynar AM , Crippen DW , Tisherman SA , Shutterly K et al. Intensivist-led management of brain-dead donors is associated with an increase in organ recovery for transplantation . Am J Transplant 2011 ; 11 : 1517 – 21 . Google Scholar CrossRef Search ADS PubMed 21 Shemie SD , Ross H , Pagliarello J , Baker AJ , Greig PD , Brand T et al. Organ donor management in Canada: recommendations of the forum on Medical Management to Optimize Donor Organ Potential . CMAJ 2006 ; 174 : S13 – 32 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png European Journal of Cardio-Thoracic Surgery Oxford University Press

The interval between brainstem death and cardiac assessment influences the retrieval of hearts for transplantation

Loading next page...
 
/lp/ou_press/the-interval-between-brainstem-death-and-cardiac-assessment-influences-X8av0dp76e
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.
ISSN
1010-7940
eISSN
1873-734X
D.O.I.
10.1093/ejcts/ezx513
Publisher site
See Article on Publisher Site

Abstract

Abstract OBJECTIVES The optimum time after brainstem death (BSD) at which to assess the function of donor hearts is unknown. We hypothesized that a longer interval may be associated with a higher transplantation rate due to improved function. METHODS Data were obtained from the UK Transplant Registry for the period between April 2010 and March 2015. The time when fixed dilated pupils were first noted in the donor was considered as the time of BSD. Retrieval was defined as the time when the abdominal organs were surgically perfused. RESULTS BSD to retrieval duration was available for 1947 donors, of which 458 (24%) donated their heart. In the univariable analysis (not adjusting other donor risk factors), evidence was available to suggest that the BSD to cardiac assessment duration had a non-linear association with heart utilization (P < 0.0001). Adjusting for donor risk factors, the relationship remained with longer intervals being associated with increased transplantation (P = 0.0056). The modelled probability of heart utilization had a similar pattern to the observed rate of heart utilization. However, the probability of heart donation began to plateau after approximately 48 h. The analysis of the subset of donors attended by a cardiothoracic retrieval team showed a similar pattern. CONCLUSIONS These data suggest that time interval from BSD to organ retrieval influences the heart retrieval rate. When the sole reason for declining a donor heart is poor function, a period of further observation and optimization up to 2 days should be considered. Brain death, Heart transplantation/mortality, Tissue and organ harvesting, Risk assessment INTRODUCTION Heart transplantation remains the definitive long-term treatment for advanced cardiac failure irrespective of cause. A discrepancy between required and available organs persists internationally with donation after brain death being the main source of donor hearts. The catecholamine storm associated with brainstem death (BSD) may have a detrimental effect on donor heart function via complex cardiovascular changes, endocrine and metabolic disturbances and release of proinflammatory signals [1]. There is a window of opportunity after BSD for organs to be optimized and retrieved for transplantation before irreversible damage occurs. Early active donor management has been shown to exert a positive influence on this process without impacting on post-transplant recipient outcome [2]. Given a lack of evidence about the optimal interval after BSD that allows recovery from the deleterious effects of the catecholamine storm, the first objective of this study was to assess whether a longer interval may be associated with a higher transplantation rate in a large cohort of donors. The second objective was to investigate the relationship of the BSD interval with the post-transplant recipient outcome. MATERIALS AND METHODS Data collection and analysis In the UK, a National Organ Retrieval Service was introduced in April 2010, and since then, the UK Transplant Registry (UKTR) data have been collected on the retrieval process. Follow-up data on 30-day survival within the cohort are 100% complete, and the accuracy and consistency of the data are maintained by regular computer-based and case–record validations. Data are stored on the UKTR that is maintained by the NHS Blood and Transplant. The data analysis was carried out using the SAS version 9.4 software. In this article, BSD is defined as the time when fixed dilated pupils were first noted in the donors, and the BSD interval is defined as the time from BSD to cardiac assessment at organ retrieval. Outcomes We investigated the impact of the BSD interval on 2 primary outcomes: (i) heart utilization and (ii) post-heart transplant survival. Heart utilization Heart utilization is defined as donor hearts retrieved with the intention of transplantation. Data on adult organ donors after brain death—also described on the literature as donors after neurological determination of death—were obtained retrospectively from the UKTR for the 5-year period from 1 April 2010 to 31 March 2015. Donors older than 65 years with a history of cardiothoracic disease, donors with no consent for heart donation and donors whose cause of death was myocardial infarction were excluded. Post-heart transplant survival To further assess the impact of the BSD interval on recipient survival, we analysed adult patients who received a heart transplant at any of the 6 designated UK centres where the heart donors were part of the heart utilization cohort previously described. We analysed short-term survival (30 and 90 days) and medium-term survival (1 and 3 years). To ensure a homogeneous transplant cohort, we excluded multiorgan transplants, retransplantations, paediatric patient transplants and donor hearts that had used the Organ Care System (OCS; TransMedics, Inc., Andover, MA, USA) for perfusion. Missing data Missing data for risk factors (maximum 18%: Tables 1 and 2) identified in the analysis were imputed with multiple imputation fully conditional specification method. Missing BSD intervals were not imputed and so the final cohort used 84% of the full cohort. Table 1: Donor risk factors and categories used in the risk-adjusted heart utilization model n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with heart utilization by the multivariable stepwise logistic regression model selection. BMI: body mass index; IQR: interquartile range; LRT: likelihood ratio test. Table 1: Donor risk factors and categories used in the risk-adjusted heart utilization model n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) n (%) or median (IQR) derived from preimputed cohort Odds ratio derived from imputed cohort P-valuea Donor BMI  Non-linear spline with knots at 19.7, 24.0, 27.4 and 36.3 26 (23–29) Not available as variable is modelled non-linearly 0.08  Not reported 0 (0) Donor age  Non-linear spline with knots at 20, 41, 51 and 63 47 (36–55) Not available as variable is modelled non-linearly <0.001  Not reported 0 (0) Donor blood group  O 883 (45) 1.87 (1.32–2.65) 0.001  A 789 (41) 1.69 (1.19–2.4)  B and AB 275 (14) 1.00 (reference group)  Not reported 0 (0) Donor gender  Male 971 (50) 1.80 (1.44–2.24) <0.001  Female 976 (50) 1.00 (reference group)  Not reported 0 (0) Donor heart rate  Non-linear spline with knots at 64, 84, 100 and 130 91 (79–107) Not available as variable is modelled non-linearly <0.001  Not reported 64 (3) History of hypertension  Yes 402 (21) 1.00 (reference group) 0.01  No 1514 (21) 1.47 (1.08–2.02)  Not reported 31 (1) Donor cardiac arrest  Yes 577 (30) 1.00 (reference group) <0.001  No 1353 (69) 1.80 (1.41–2.3)  Not reported 17 (1) Past diabetes  Yes 86 (4) 1.00 (reference group) 0.03  No 1845 (95) 2.07 (1.06–4.04)  Not reported 16 (1) History of alcohol abuse  Yes 294 (15) 1.00 (reference group) 0.02  No 1623 (83) 1.46 (1.05–2.03)  Not reported 30 (2) Past smoker  Yes 999 (51) 1.00 (reference group) 0.02  No 932 (48) 1.32 (1.06–1.64)  Not reported 16 (1) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with heart utilization by the multivariable stepwise logistic regression model selection. BMI: body mass index; IQR: interquartile range; LRT: likelihood ratio test. Table 2: Donor, transplant and recipient risk factors and categories used in the risk-adjusted 30-day, 90-day, 1-year and 3-year survival models, with hazard ratios and P-values from the 30-day model n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with 30-day post-heart transplant survival by the multivariable stepwise Cox proportional hazards regression model selection. BMI: body mass index; ECMO: extracorporeal membrane oxygenation; IQR: interquartile range; LRT: likelihood ratio test; MCS: mechanical circulatory support. Table 2: Donor, transplant and recipient risk factors and categories used in the risk-adjusted 30-day, 90-day, 1-year and 3-year survival models, with hazard ratios and P-values from the 30-day model n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) n (%) or median (IQR) derived from preimputed cohort Hazard ratio from 30-day model derived from imputed cohort P-value from 30-day modela Donor cause of death  Vascular 293 (64) 1.00 (reference group) 0.06  Trauma 56 (12) 1.12 (0.46–2.72)  Hypoxic 61 (13) 0.92 (0.37–2.26)  Other 46 (10) 0.22 (0.05–1.06)  Not reported 2 (<1) Imputed Donor BMI  Modelled as continuous variable 25 (23–28) 1.03 (0.98–1.08) 0.34  Not reported 0 (0) Imputed Donor age  Modelled as continuous variable 41 (30–49) 0.99 (0.97–1.02) 0.93  Not reported 0 (0) Respiratory arrest  Yes 107 (23) 1.00 (reference group) 0.56  No 335 (73) 0.85 (0.45–1.59)  Not reported 16 (4) Recipient BMI  Modelled as continuous variable 25 (23–29) 1.04 (0.97–1.11) 0.17  Not reported 1 (<1) Recipient creatinine at transplant  Non-linear spline with knots at 60, 95, 119, and 181 101 (80–127) Not available as variable is modelled non-linearly 0.27  Not reported 9 (2) ECMO at transplant (30-day model only)  Yes 10 (2) 1.62 (0.29–9.08) 0.90  No 368 (80) 1.00 (reference group)  Not reported 80 (18) MCS at transplant (1- and 3-year models only)  Short-term  Long-term (including total artificial hearts)  ECMO  None  Not reported Hospital status at transplant  In-hospital 299 (65) 1.00 (reference group) 0.28  Not in-hospital 157 (35) 1.54 (0.77–3.07)  Not reported 2 (<1) Primary disease  Dilated cardiomyopathy 257 (56) 1.43 (0.73–2.78) 0.21  Coronary heart disease 68 (15) 0.71 (0.26–1.96)  Congenital heart disease 43 (9) 1.71 (0.67–4.34)  Other 90 (20) 1.00 (reference group)  Not reported 0 (0) Sex mismatch  RM: DM 264 (58) 0.41 (0.19–0.87) 0.03  RM: DF 84 (18) 0.47 (0.2–1.11)  RF: DM 70 (15) 0.93 (0.39–2.24)  RF: DF 40 (9) 1.00 (reference group)  Not reported 0 (0) Ischaemic time (h)  Modelled as continuous variable 3 (2–4) 1.4 (1.07–1.83) 0.007  Not reported 63 (14) a The LRT P-values represent the reduced model deviance with the inclusion of each of the variables associated with 30-day post-heart transplant survival by the multivariable stepwise Cox proportional hazards regression model selection. BMI: body mass index; ECMO: extracorporeal membrane oxygenation; IQR: interquartile range; LRT: likelihood ratio test; MCS: mechanical circulatory support. Statistical analysis Heart utilization We used the multivariable logistic regression to assess whether the BSD interval is associated with the retrieval of hearts for transplantation, after adjusting for known risk factors. A stepwise variable selection method was used where candidate risk factors were retained in the model if they reduced the model deviance significantly (P < 0.1) according to the likelihood ratio test. Donor risk factors included were blood group, sex, history of hypertension, use of prednisolone, cardiac arrest, age, body mass index (BMI), heart rate, history of diabetes, history of alcohol abuse and smoking history. The BSD interval was tested in the model as both a linear term and a non-linear term. A linear term assumes a relationship between the BSD interval and the probability of utilization that is constant as the BSD interval increases, whereas a non-linear term allows the relationship to vary depending on the BSD interval. We used a natural cubic spline method to explore non-linearity, allowing cubic expressions between intervals defined by ‘knots’ at the 5%, 35%, 65% and 95% percentiles. The model was internally validated using the Hosmer and Lemeshow measure of concordance. The concordance statistic of the heart utilization model was 0.758, indicating acceptable predictive ability. Post-heart transplant survival We used multivariable Cox proportional hazards regression to assess whether the BSD intervals associated with patient survival time post-transplant, after adjusting for known risk factors. Known risk factors were identified in the adult heart risk-adjusted models that were developed for the Annual Report on Cardiothoracic Transplantation, published by the NHSBT in July 2016. These include donor cause of death, donor BMI, donor age, donor respiratory arrest, recipient BMI, recipient creatinine at transplant, if recipient had extracorporeal membrane oxygenation at transplant (for short-term survival models only), if recipient had ventricular assist device at transplant (for medium-term survival models only), if recipient was in-hospital before the transplant, recipient primary disease, recipient and donor sex mismatch and ischaemia time. To check the proportional hazards assumption on which the Cox regression model is based, log cumulative hazard plots and Schoenfeld residual plots were used. There was no evidence to suggest violation of the proportional hazard assumption in the model of interest. The BSD interval was considered as both a linear term and a non-linear term in the model using the natural cubic spline method previously described. Combining the heart utilization and post-heart transplant survival model To explore the results of the 2 models in combination, we interrogated how the utilization/survival probabilities changed as the BSD interval varied, while retaining all other variables in the models fixed. This was achieved by selecting common or average donor and recipient characteristics. RESULTS From the donor inclusion criteria, there were 1947 records. Of these, 1354 (70%) did not proceed to heart donation. Out of 593 heart donors, only 458 subjects fulfilled all inclusion criteria and were further analysed in the post-transplant analysis (Fig. 1). The observed median BSD interval for the donor hearts retrieved and not retrieved was 39 and 35 h, respectively (Fig. 2). Figure 1: View largeDownload slide Data and analysis flow chart. DBD: donors after brain death. Figure 1: View largeDownload slide Data and analysis flow chart. DBD: donors after brain death. Figure 2: View largeDownload slide Box plots of the BSD interval for non-heart donors and heart donors. BSD: brainstem death. Figure 2: View largeDownload slide Box plots of the BSD interval for non-heart donors and heart donors. BSD: brainstem death. Heart utilization In the univariable analysis (not adjusting any donor risk factors), evidence was available to suggest that the BSD interval had a non-linear association with heart utilization; longer intervals being associated with increased utilization (overall effect P < 0.0001, non-linearity P = 0.009). Adjusting for donor risk factors listed in Table 1, the duration still had this non-linear association with heart utilization (overall effect P = 0.0056, non-linearity P = 0.009). The modelled probability of heart utilization has a similar trend to the observed rate of heart utilization within the deciles of the BSD interval. Specifically, longer durations were associated with a higher probability of donation. However, the probability of heart donation began to plateau after approximately 48 h indicating that durations beyond this were not associated with the chance of heart utilization (Fig. 3). Figure 3: View largeDownload slide (A) The risk-adjusted probability of heart utilization from the logistic regression model plotted against the BSD interval. Each grey dot is the estimated probability of heart utilization for a donor by the actual BSD interval for that donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of heart utilization and the BSD interval. (B) The observed rate of heart utilization within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of heart utilization and the BSD interval. BSD: brainstem death. Figure 3: View largeDownload slide (A) The risk-adjusted probability of heart utilization from the logistic regression model plotted against the BSD interval. Each grey dot is the estimated probability of heart utilization for a donor by the actual BSD interval for that donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of heart utilization and the BSD interval. (B) The observed rate of heart utilization within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of heart utilization and the BSD interval. BSD: brainstem death. As a sensitivity analysis, the model was refitted using the donors with complete data for all risk factors (n = 1883). The BSD interval also had a significant non-linear association with heart utilization in this cohort. To explore the impact of donor optimization, we excluded donors who were not attended by a cardiothoracic retrieval team and performed an analysis on a subset of donors who were attended by the cardiothoracic retrieval team only (n = 1306). Analysing this with the same methodology, the BSD interval has a significant association with heart utilization (P = 0.007). The effect of the BSD interval had a similar pattern. Post-heart transplant survival In the univariable post-transplant analysis, evidence was available to suggest that the BSD interval had a non-linear association with 30-day patient survival only (overall effect P = 0.013, non-linearity P = 0.005). Adjusting for the donor, recipient and transplant risk factors in Table 2, the duration still had a non-linear association with 30-day patient survival (overall effect P = 0.04, non-linearity P = 0.017) but not longer term survival (P ≥ 0.19). The modelled 30-day patient survival probabilities had a similar trend to the observed patient survival rates within the deciles of the BSD interval obtained using the Kaplan–Meier method. For patients with transplanted hearts retrieved less than 36 h after BSD, the estimated 30-day survival has a decreasing trend with increasing duration to approximately 90% survival (Fig. 4). For patients in whom transplanted hearts were retrieved between 36 and 72 h after BSD, the estimated 30-day survival increased from approximately 90% to 95%. For patients in whom transplanted hearts were retrieved after 72 h, the estimated 30-day survival decreased again; however, the confidence interval of the LOESS trend line was a lot wider, as there were fewer transplants with a heart retrieved more than 72 h after BSD. Figure 4: View largeDownload slide (A) The risk-adjusted probability of 30-day survival from the Cox proportional hazards regression model plotted against the BSD interval. Each grey dot is the estimated probability of 30-day survival for a patient by the actual BSD interval for their donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of 30-day survival and the BSD interval. (B) The observed (Kaplan–Meier) 30-day survival rate within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of 30-day survival and the BSD interval. BSD: brainstem death. Figure 4: View largeDownload slide (A) The risk-adjusted probability of 30-day survival from the Cox proportional hazards regression model plotted against the BSD interval. Each grey dot is the estimated probability of 30-day survival for a patient by the actual BSD interval for their donor. The line is the LOESS curve through the dots, with its 95% confidence limits, representing the general trend of the relationship between the probability of 30-day survival and the BSD interval. (B) The observed (Kaplan–Meier) 30-day survival rate within the deciles of the BSD interval, plotted at the mid-point, compared with the LOESS curve from (A). This plot shows positive agreement between the pattern of the observed and modelled relationships between the probability of 30-day survival and the BSD interval. BSD: brainstem death. As a sensitivity analysis, the model was refitted using the transplants with complete data for all risk factors (n = 374). The BSD interval was also found to be significantly associated with 30-day patient survival times but not with 90-day, 1-year or 3-year patient survival times. Combining the heart utilization and 30-day post-heart transplant survival model The association between the BSD interval and both heart utilization and the 30-day post-heart transplant patient survival is significant. Therefore, it is of interest to obtain an optimum window for the start of the retrieval operation after BSD, using the results of the 2 statistical models. For a male donor, aged 44 years, with BMI 26 kg/m2, blood group A, no history of hypertension, no cardiac arrest, heart rate 91 bpm, no history of diabetes, no history of alcohol abuse and was a past smoker, the probability of heart utilization is optimized from approximately 50 h of BSD interval (Fig. 5). For a male patient, with no respiratory arrest, no extracorporeal membrane oxygenation required, was in-hospital before transplant and had coronary heart disease, receiving a heart from a male donor, aged 44 years, with BMI 26 kg/m2, who died due to vascular disease, the probability of 30-day survival is very high from approximately 70 h of BSD interval onwards. Figure 5: View largeDownload slide The change in estimated probability of heart utilization/30-day survival as BSD interval increases for a typical donor/patient. The solid line shows how the probability of the heart being utilized varies for a male donor, aged 44 years, with body mass index 26 kg/m2, blood group A, no history of hypertension, no cardiac arrest, heart rate 91 bpm, no history of diabetes, no history of alcohol abuse and was a past smoker, with its 95% CL. The dashed line shows how the probability of 30-day survival varies for a male patient, with no respiratory arrest, no extracorporeal membrane oxygenation required, was in-hospital before transplant and had coronary heart disease, receiving a heart from a male donor, aged 44 years, with body mass index 26 kg/m2, who died from vascular disease, with its 95% CLs. BSD: brainstem death; CL: confidence limit. Figure 5: View largeDownload slide The change in estimated probability of heart utilization/30-day survival as BSD interval increases for a typical donor/patient. The solid line shows how the probability of the heart being utilized varies for a male donor, aged 44 years, with body mass index 26 kg/m2, blood group A, no history of hypertension, no cardiac arrest, heart rate 91 bpm, no history of diabetes, no history of alcohol abuse and was a past smoker, with its 95% CL. The dashed line shows how the probability of 30-day survival varies for a male patient, with no respiratory arrest, no extracorporeal membrane oxygenation required, was in-hospital before transplant and had coronary heart disease, receiving a heart from a male donor, aged 44 years, with body mass index 26 kg/m2, who died from vascular disease, with its 95% CLs. BSD: brainstem death; CL: confidence limit. DISCUSSION Heart transplantation is a successful treatment for advanced heart failure but is limited by the scarcity of hearts judged suitable for transplantation. Only a minority of hearts offered for transplantation is actually used. The selection of hearts is based partly on cardiac function, but this is known to change with time after BSD. Here, we demonstrate for the first time, in a large cohort of cardiac donors, that an increased interval between BSD and organ retrieval increases the probability of heart utilization up to an interval of 48 h. Furthermore, an increased interval did not affect overall post-transplant survival. The time of procurement can be utilized as the distinct time point separating pre- from post-organ retrieval events. It is of essence to further break down the variable preretrieval period into specific time periods and study their impact on retrieval and long-term outcomes. As all relevant time points could not be obtained from our database, calculated indirect estimations were used (Fig. 6). Time from injury to BSD is known as the brain injury interval, and time from BSD to procurement is known as BSD interval. As exact time of injury is difficult to obtain in majority of cases, many groups use the time of intubation. Similarly, the exact time of BSD cannot be used as an objective hard time point with time of the second assessment of BSD being used as an approximation. We chose to use the documented time of coning (detection of fixed dilated pupils) as a more realistic time point indicative of BSD. The interval from procurement to implantation is subdivided into cold ischaemia and warm ischaemia times and has been extensively studied [3, 4]. Figure 6: View largeDownload slide Diagram of the BSD interval analysed and justification for using fixed pupil dilation. BSD: brainstem death; IQR: interquartile range. Figure 6: View largeDownload slide Diagram of the BSD interval analysed and justification for using fixed pupil dilation. BSD: brainstem death; IQR: interquartile range. In the UKTR, coning time was reported only within 37% of the donor cohort, whereas time of fixed and dilated pupils were noted in 86%. To assess whether time of fixed and dilated pupils is appropriate to represent BSD/coning, the difference between the 2 time points were compared with the cases where both are reported. The median difference between the time of coning and fixed and dilated pupil first noted was only 1 h (interquartile range 0–10 h) (n = 780). On the other hand, brainstem test time was reported in 97% of cases; however, the median of the difference between coning and the first brainstem test performed was 12 h (interquartile range: 6–20 h) (n = 866). The difference in time of coning and the first brainstem test was 1-10 h in 41%, and 11–30 h in 46%. Hence, the time of fixed and dilated pupils was used as a time point indicative of BSD. In the trauma setting, increasing brain death intervals were not associated with decreased organ procurement rates or an increased number of non-salvageable organs due to poor function [5]. Data sub-analysis actually showed an increased rate of procurement for heart and pancreas with increasing brain death intervals. In a retrospective study of 215 heart transplants, Marasco et al. [6] did not find an association between brain death interval and mortality. The most important finding of this study was the strong correlation between brain death interval and hypoxia inducible factor 1a activity in donor atrial tissue with the authors suggesting a potential role in ameliorating the detrimental effects of the catecholamine storm. A few investigators have assessed the influence of the brain death interval on outcomes after heart transplantation. The Stanford groups has reported on adverse survival trends with management times greater than 66 h, defined as the sum of the brain injury and brain death intervals [7]. No difference in survival was observed in relation to brain injury or brain death intervals when assessed individually. Solomon et al. [8] did not demonstrate any statistical significance in outcomes in relation to brain death interval, although all intervals were <12 h. In a paediatric setting, Odim et al. [9] demonstrated a trend towards improved rejection-free survival with longer brain death intervals (35–90 h) that did not achieve statistical significance. An increased risk of mortality for heart transplantation patients was observed with prolonged brain death intervals in a limited study from Manchester [10]. BSD pathophysiology involves a cascade of cardiovascular, respiratory, endocrine, metabolic and stress responses that may jeopardize potentially transplantable organs [11]. Catecholamine levels are elevated several folds following brain death in humans with colocalized apoptotic and necrotic damage observed in 40% of donors [12]. In more detail, investigators reported histological evidence of damaged heart tissue in 84%, with the affected area never corresponding to >5% of the overall tissue and thus unlikely to have had any major detrimental effect on post-transplant heart function according to the authors. In the context of marginal donor hearts, load-independent indices of cardiac function should be assessed as haemodynamic instability alone is not a reliable indicator of subsequent primary organ dysfunction [13]. The explosive increase in intracranial pressure also may have a detrimental effect on the myocardium [14]. If the sympathetic storm builds up in a gradual nature, haemodynamic changes and structural myocardial damage are less pronounced. In terms of post-transplantation outcomes, there is evidence that brain death is also associated with the activation of pro-inflammatory mediators or even alteration of myocardial gene expression [15, 16]. Identifying molecular pathways regulating this process may provide therapeutic targets to modify the aftermath of the severe stress response in brain death [17]. Donor management has been the focus of many groups to increase the yield of transplantable organs [2, 18, 19]. Singbartl et al. [20] reported an increase in organ recovery for transplantation via implementation of intensivist-led management of brain-dead donors, although these findings did not achieve statistical significance regarding the number of hearts transplanted. Recommendations for optimal medical management of the heart-beating brain-dead donor to optimize organ potential have been published [11, 21]. During the study period, the probability of heart donation began to plateau after approximately 48 h with optimal 30-day survival rate noted at 72 h. Although interpretation of these findings remains descriptive with no exact knowledge of underlying pathophysiological mechanisms, it seems that ‘riding out’ the storm has a dual clinical benefit. First, potential donors are allowed to self-select as organs irreversibly damaged will manifest clinical deterioration. Second, during this period, there seems to be a window of donor optimization that may be amplified by donor management that has a prognostic impact on long-term outcomes. Although not assessed, we believe that there may be financial benefits associated with such a strategy. Limitations A number of limitations are inherent to this study. First, the data reflect a national clinical registry. Even though these data provide an insight into real-world practice, it is conceivable that there are unmeasured confounding variables. Second, although we attempted to resolve missing data and transcriptional errors using reproducible algorithms and statistical imputation methodology, we have potentially underestimated the standard errors, and wrongly assumed data are missing at random. However, not imputing the variable of interest and conducting sensitivity analyses have mitigated against the risk of misrepresentation. Third, the LOESS method was used to visualize the trends obtained from our models, but the interpretation is limited due to substantial variation of the model estimates around the LOESS curves. Finally, the impact of donor management-specific strategies has not been assessed in this study. Conflict of interest: none declared. REFERENCES 1 Mascia L , Mastromauro I , Viberti S , Vincenzi M , Zanello M. Management to optimize organ procurement in brain dead donors . Minerva Anestesiol 2009 ; 75 : 125 – 33 . Google Scholar PubMed 2 Abuanzeh R , Hashmi F , Dimarakis I , Khasati N , Machaal A , Yonan N et al. Early donor management increases the retrieval rate of hearts for transplantation in marginal donors . Eur J Cardiothorac Surg 2015 ; 47 : 72 – 7 ; discussion 77. Google Scholar CrossRef Search ADS PubMed 3 Banner NR , Thomas HL , Curnow E , Hussey JC , Rogers CA , Bonser RS et al. The importance of cold and warm cardiac ischemia for survival after heart transplantation . Transplantation 2008 ; 86 : 542 – 7 . Google Scholar CrossRef Search ADS PubMed 4 Marasco SF , Kras A , Schulberg E , Vale M , Lee GA. Impact of warm ischemia time on survival after heart transplantation . Transplant Proc 2012 ; 44 : 1385 – 9 . Google Scholar CrossRef Search ADS PubMed 5 Inaba K , Branco BC , Lam L , Salim A , Talving P , Plurad D et al. Organ donation and time to procurement: late is not too late . J Trauma 2010 ; 68 : 1362 – 6 . Google Scholar CrossRef Search ADS PubMed 6 Marasco S , Kras A , Schulberg E , Vale M , Chan P , Lee GA et al. Donor brain death time and impact on outcomes in heart transplantation . Transplant Proc 2013 ; 45 : 33 – 7 . Google Scholar CrossRef Search ADS PubMed 7 Cantin B , Kwok BW , Chan MC , Valantine HA , Oyer PE , Robbins RC et al. The impact of brain death on survival after heart transplantation: time is of the essence . Transplantation 2003 ; 76 : 1275 – 9 . Google Scholar CrossRef Search ADS PubMed 8 Solomon NA , McGiven JR , Alison PM , Ruygrok PN , Haydock DA , Coverdale HA et al. Changing donor and recipient demographics in a heart transplantation program: influence on early outcome . Ann Thorac Surg 2004 ; 77 : 2096 – 102 . Google Scholar CrossRef Search ADS PubMed 9 Odim J , Laks H , Banerji A , Mukherjee K , Vincent C , Murphy C et al. Does duration of donor brain injury affect outcome after orthotopic pediatric heart transplantation? J Thorac Cardiovasc Surg 2005 ; 130 : 187 – 93 . Google Scholar CrossRef Search ADS PubMed 10 Ramjug S , Hussain N , Yonan N. Prolonged time between donor brain death and organ retrieval results in an increased risk of mortality in cardiac transplant recipients . Interact CardioVasc Thorac Surg 2011 ; 12 : 938 – 42 . Google Scholar CrossRef Search ADS PubMed 11 McKeown DW , Bonser RS , Kellum JA. Management of the heartbeating brain-dead organ donor . Br J Anaesth 2012 ; 108(Suppl 1) : i96 – 107 . Google Scholar CrossRef Search ADS PubMed 12 Pérez López S , Otero Hernández J , Vázquez Moreno N , Escudero Augusto D , Álvarez Menéndez F , Astudillo González A. Brain death effects on catecholamine levels and subsequent cardiac damage assessed in organ donors . J Heart Lung Transplant 2009 ; 28 : 815 – 20 . Google Scholar CrossRef Search ADS PubMed 13 Szabo G. Physiologic changes after brain death . J Heart Lung Transplant 2004 ; 23 : S223 – 6 . Google Scholar CrossRef Search ADS PubMed 14 Shivalkar B , Van Loon J , Wieland W , Tjandra-Maga TB , Borgers M , Plets C et al. Variable effects of explosive or gradual increase of intracranial pressure on myocardial structure and function . Circulation 1993 ; 87 : 230 – 9 . Google Scholar CrossRef Search ADS PubMed 15 Wilhelm MJ , Pratschke J , Beato F , Taal M , Laskowski IA , Paz DM et al. Activation of proinflammatory mediators in heart transplants from brain-dead donors: evidence from a model of chronic rat cardiac allograft rejection . Transplant Proc 2002 ; 34 : 2359 – 60 . Google Scholar CrossRef Search ADS PubMed 16 Yeh T Jr , Wechsler AS , Graham LJ , Loesser KE , Sica DA , Wolfe L et al. Acute brain death alters left ventricular myocardial gene expression . J Thorac Cardiovasc Surg 1999 ; 117 : 365 – 74 . Google Scholar CrossRef Search ADS PubMed 17 Zhai W , Feng R , Huo L , Li J , Zhang S. Mechanism of the protective effects of N-acetylcysteine on the heart of brain-dead Ba-Ma miniature pigs . J Heart Lung Transplant 2009 ; 28 : 944 – 9 . Google Scholar CrossRef Search ADS PubMed 18 Wheeldon DR , Potter CD , Oduro A , Wallwork J , Large SR. Transforming the “unacceptable” donor: outcomes from the adoption of a standardized donor management technique . J Heart Lung Transplant 1995 ; 14 : 734 – 42 . Google Scholar PubMed 19 Rosendale JD , Kauffman HM , McBride MA , Chabalewski FL , Zaroff JG , Garrity ER et al. Hormonal resuscitation yields more transplanted hearts, with improved early function . Transplantation 2003 ; 75 : 1336 – 41 . Google Scholar CrossRef Search ADS PubMed 20 Singbartl K , Murugan R , Kaynar AM , Crippen DW , Tisherman SA , Shutterly K et al. Intensivist-led management of brain-dead donors is associated with an increase in organ recovery for transplantation . Am J Transplant 2011 ; 11 : 1517 – 21 . Google Scholar CrossRef Search ADS PubMed 21 Shemie SD , Ross H , Pagliarello J , Baker AJ , Greig PD , Brand T et al. Organ donor management in Canada: recommendations of the forum on Medical Management to Optimize Donor Organ Potential . CMAJ 2006 ; 174 : S13 – 32 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

European Journal of Cardio-Thoracic SurgeryOxford University Press

Published: Jan 22, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off