Segmented relationships to model erosion of regression effect in Cox regression

Determining optimal donor heart ischemic times in adult cardiac transplantation

Objectives

Unsupervised statistical determination of optimal allograft ischemic time (IT) on heart transplant outcomes among ABO donor heart types.

Methods

We identified 36,145 heart transplants (2000–2018) from the United Network for Organ Sharing database. Continuous and categorical variables were analyzed with parametric and nonparametric testing. Determination of IT cutoffs for survival analysis was performed using Contal and O'Quigley univariable method and Vito Muggeo multivariable segmented modeling.

Results

Univariable and multivariable IT threshold determination revealed a cutoff at about 3 h. The hourly increase in survival risk with ≥3 h IT is asymmetrically experienced at the early 90 days (hazard ratio [HR] = 1.29, p < .001) and up to 1‐year time point (HR = 1.16, p < .001). Beyond 1 year the risk of prolonged IT is less impactful (HR = 1.04, p = .022). Longer IT was associated with more postoperative complications such as stroke (2.7% vs. 2.3, p = .042), dialysis (11.6% vs. 9.1%, p < .001) and death from primary graft dysfunction (1.8% vs. 1.2%, p < .001). O blood type donor hearts with IT ≥ 3 h has significantly increased hourly mortality risk at 90 days (HR = 1.27, p < .001), 90 days to 1 year (HR = 1.22, p < .001) and >1 year (HR = 1.05, p = .041). For non‐O blood types with ≥3 h IT hourly mortality risk was increased at 90 days (HR = 1.33, p < .001), but not at 90 days to 1 year (HR = 1.09, p = .146) nor ≥1 year (HR = 1.08, p = .237).

Conclusions

The donor heart IT threshold for survival determined from unbiased statistical modeling occurs at 3 h. With longer preservation times, transplantation with O donor hearts was associated with worse survival.