Transplant benefit-based offering of deceased donor livers in the United Kingdom.

BACKGROUND & AIMS: The National Liver Offering Scheme (NLOS) was introduced in the UK in 2018 to offer livers from deceased donors to patients on the national waiting list based, for most patients, on calculated transplant benefit. Before NLOS, livers were offered to transplant centres by geographic donor zones and, within centres, by estimated recipient need for a transplant. METHODS: UK Transplant Registry data on patient registrations and transplants were analysed to build statistical models for survival on the list (M1) and survival post-transplantation (M2). A separate cohort of registrations - not seen by the models before - was analysed to simulate what liver allocation would have been under M1, M2 and a transplant benefit score (TBS) model (combining both M1 and M2), and to compare these allocations to what had been recorded in the UK Transplant Registry. The number of deaths on the waiting list and patient life years were used to compare the different simulation scenarios and to select the optimal allocation model. Registry data were monitored, pre- and post-NLOS, to understand the performance of the scheme. RESULTS: The TBS was identified as the optimal model to offer donation after brain death (DBD) livers to adult and large paediatric elective recipients. In the first 2 years of NLOS, 68% of DBD livers were offered using the TBS to this type of recipient. Monitoring data indicate that mortality on the waiting list post-NLOS significantly decreased compared with pre-NLOS (p <0.0001), and that patient survival post-listing was significantly greater post-compared to pre-NLOS (p = 0.005). CONCLUSIONS: In the first two years of NLOS offering, waiting list mortality fell while post-transplant survival was not negatively impacted, delivering on the scheme's objectives. IMPACT AND IMPLICATIONS: The National Liver Offering Scheme (NLOS) was introduced in the UK in 2018 to increase transparency of the deceased donor liver offering process, maximise the overall survival of the waiting list population, and improve equity of access to liver transplantation. To our knowledge, it is the first scheme that offers organs based on statistical prediction of transplant benefit: the transplant benefit score. The results are important to the transplant community - from healthcare practitioners to patients - and demonstrate that, in the first two years of NLOS offering, waiting list mortality fell while post-transplant survival was not negatively impacted, thus delivering on the scheme's objectives. The scheme continues to be monitored to ensure that the transplant benefit score remains up-to-date and that signals that suggest the possible disadvantage of some patients are investigated.

Early Ultrasound Surveillance of Newly-Created Hemodialysis Arteriovenous Fistula.

Introduction: We assess if ultrasound surveillance of newly-created arteriovenous fistulas (AVFs) can predict nonmaturation sufficiently reliably to justify randomized controlled trial (RCT) evaluation of ultrasound-directed salvage intervention. Methods: Consenting adults underwent blinded fortnightly ultrasound scanning of their AVF after creation, with scan characteristics that predicted AVF nonmaturation identified by logistic regression modeling. Results: Of 333 AVFs created, 65.8% matured by 10 weeks. Serial scanning revealed that maturation occurred rapidly, whereas consistently lower fistula flow rates and venous diameters were observed in those that did not mature. Wrist and elbow AVF nonmaturation could be optimally modeled from week 4 ultrasound parameters alone, but with only moderate positive predictive values (PPVs) (wrist, 60.6% [95% confidence interval, CI: 43.9-77.3]; elbow, 66.7% [48.9-84.4]). Moreover, 40 (70.2%) of the 57 AVFs that thrombosed by week 10 had already failed by the week 4 scan, thus limiting the potential of salvage procedures initiated by that scan's findings to alter overall maturation rates. Modeling of the early ultrasound characteristics could also predict primary patency failure at 6 months; however, that model performed poorly at predicting assisted primary failure (those AVFs that failed despite a salvage attempt), partly because patency of at-risk AVFs was maintained by successful salvage performed without recourse to the early scan data. Conclusion: Early ultrasound surveillance may predict fistula maturation, but is likely, at best, to result in only very modest improvements in fistula patency. Power calculations suggest that an impractically large number of participants (>1700) would be required for formal RCT evaluation.

Doppler ultrasound surveillance of recently formed haemodialysis arteriovenous fistula: the SONAR observational cohort study.

Background: Arteriovenous fistulas are considered the best option for haemodialysis provision, but as many as 30% fail to mature or suffer early failure. Objective: To assess the feasibility of performing a randomised controlled trial that examines whether, by informing early and effective salvage intervention of fistulas that would otherwise fail, Doppler ultrasound surveillance of developing arteriovenous fistulas improves longer-term arteriovenous fistula patency. Design: A prospective multicentre observational cohort study (the 'SONAR' study). Setting: Seventeen haemodialysis centres in the UK. Participants: Consenting adults with end-stage renal disease who were scheduled to have an arteriovenous fistula created. Intervention: Participants underwent Doppler ultrasound surveillance of their arteriovenous fistulas at 2, 4, 6 and 10 weeks after creation, with clinical teams blinded to the ultrasound surveillance findings. Main outcome measures: Fistula maturation at week 10 defined according to ultrasound surveillance parameters of representative venous diameter and blood flow (wrist arteriovenous fistulas: ≥ 4 mm and > 400 ml/minute; elbow arteriovenous fistulas: ≥ 5 mm and > 500 ml/minute). Mixed multivariable logistic regression modelling of the early ultrasound scan data was used to predict arteriovenous fistula non-maturation by 10 weeks and fistula failure at 6 months. Results: A total of 333 arteriovenous fistulas were created during the study window (47.7% wrist, 52.3% elbow). By 2 weeks, 37 (11.1%) arteriovenous fistulas had failed (thrombosed), but by 10 weeks, 219 of 333 (65.8%) of created arteriovenous fistulas had reached maturity (60.4% wrist, 67.2% elbow). Persistently lower flow rates and venous diameters were observed in those fistulas that did not mature. Models for arteriovenous fistulas' non-maturation could be optimally constructed using the week 4 scan data, with fistula venous diameter and flow rate the most significant variables in explaining wrist fistula maturity failure (positive predictive value 60.6%, 95% confidence interval 43.9% to 77.3%), whereas resistance index and flow rate were most significant for elbow arteriovenous fistulas (positive predictive value 66.7%, 95% confidence interval 48.9% to 84.4%). In contrast to non-maturation, both models predicted fistula maturation much more reliably [negative predictive values of 95.4% (95% confidence interval 91.0% to 99.8%) and 95.6% (95% confidence interval 91.8% to 99.4%) for wrist and elbow, respectively]. Additional follow-up and modelling on a subset (n = 192) of the original SONAR cohort (the SONAR-12M study) revealed the rates of primary, assisted primary and secondary patency arteriovenous fistulas at 6 months were 76.5, 80.7 and 83.3, respectively. Fistula vein size, flow rate and resistance index could identify primary patency failure at 6 months, with similar predictive power as for 10-week arteriovenous fistula maturity failure, but with wide confidence intervals for wrist (positive predictive value 72.7%, 95% confidence interval 46.4% to 99.0%) and elbow (positive predictive value 57.1%, 95% confidence interval 20.5% to 93.8%). These models, moreover, performed poorly at identifying assisted primary and secondary patency failure, likely because a subset of those arteriovenous fistulas identified on ultrasound surveillance as at risk underwent subsequent successful salvage intervention without recourse to early ultrasound data. Conclusions: Although early ultrasound can predict fistula maturation and longer-term patency very effectively, it was only moderately good at identifying those fistulas likely to remain immature or to fail within 6 months. Allied to the better- than-expected fistula patency rates achieved (that are further improved by successful salvage), we estimate that a randomised controlled trial comparing early ultrasound-guided intervention against standard care would require at least 1300 fistulas and would achieve only minimal patient benefit. Trial Registration: This trial is registered as ISRCTN36033877 and ISRCTN17399438. Funding: This award was funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme (NIHR award ref: NIHR135572) and is published in full in Health Technology Assessment; Vol. 28, No. 24. See the NIHR Funding and Awards website for further award information.

For people with advanced kidney disease, haemodialysis is best provided by an 'arteriovenous fistula', which is created surgically by joining a vein onto an artery at the wrist or elbow. However, these take about 2 months to develop fully ('mature'), and as many as 3 out of 10 fail to do so. We asked whether we could use early ultrasound scanning of the fistula to identify those that are unlikely to mature. This would allow us to decide whether it would be practical to run a large, randomised trial to find out if using early ultrasound allows us to 'rescue' fistulas that would otherwise fail. We invited adults to undergo serial ultrasound scanning of their fistula in the first few weeks after it was created. We then analysed whether we could use the data from the early scans to identify those fistulas that were not going to mature by week 10. Of the 333 fistulas that were created, about two-thirds reached maturity by week 10. We found that an ultrasound scan 4 weeks after fistula creation could reliably identify those fistulas that were going to mature. However, of those fistulas predicted to fail, about one-third did eventually mature without further intervention, and even without knowing what the early scans showed, another third were successfully rescued by surgery or X-ray-guided treatment at a later stage. Performing an early ultrasound scan on a fistula can provide reassurance that it will mature and deliver trouble-free dialysis. However, because scans are poor at identifying fistulas that are unlikely to mature, we would not recommend their use to justify early surgery or X-ray-guided treatment in the expectation that this will improve outcomes.

Prognostic models: What the statistician wants the clinician to know.

Prognostic model building is a process that begins much earlier than data analysis and ends later than when a model is reached. It requires careful delineation of a clinical question, methodical planning of the approach and attentive exploration of the data before attempting model building. Once following these important initial steps, the researcher may postulate a model to describe the process of interest and build such model. Once built, the model will need to be checked, validated and the exercise may take the researcher back a few steps - for instance, to adapt the model to fit a variable that displays a 'curved' pattern - to then return to check and validate the model again. To interpret and report the results it is vital to relate the output to the original question, to be transparent in the methodology followed and to understand the limitations of the data and the approach.