The study was approved by the Swedish National Research Ethics Committee, following the Helsinki Declaration (Diary number 728-12, amendment 2020-04281). Being a retrospective study, patient consent was not deemed necessary and waived by the ethical committee. Inclusion criteria were all patients >18 years having had a HTx procedure between June 1984 and December 2021 at Sahlgrenska University Hospital, Sweden. Patients with incomplete medical records preventing assessment of early graft function were excluded. For further details regarding study cohort, see Supplementary Figure SA1.

Primary graft dysfunction (PGD) is defined in the ISHLT consensus document as acute onset of heart failure affecting the right, left, or both ventricles within the first 24 h after HTx. Additionally, to be classified as PGD, the graft dysfunction may not be attributed to other causes, such as uncontrolled bleeding, surgical complications, pulmonary hypertension or acute rejection. In these circumstances the graft dysfunction is labeled secondary. PGD is categorized into two entities: PGD-Right ventricle (PGD-RV) and PGD-Left ventricle (PGD-LV). The subclassification of PGD-LV includes mild, moderate, and severe, providing a detailed spectrum of severity. For a more comprehensive understanding, see Supplementary Figure SA1.

Data was collected retrospectively for all patients included in the study. We are maintaining an updated database of the HTx patients including main patient characteristics as well as follow-up data at our center. However, specific information on PGD has not continuously been collected. Therefore, all anesthesia records and intensive care unit (ICU) charts were retrieved and analyzed (including use of inotropic agents). ICU charts records included hourly measurements of central venous pressure (CVP), continuous cardiac output (CCO), mean arterial pressure (MAP). Pulmonary capillary wedge pressure (PCWP) and mean pulmonary arterial pressure (MPAP) via a Swan-Ganz were recorded at regular intervals. From CCO the cardiac index (CI) was calculated using the preoperative body surface area of the patient. Trans pulmonary gradient (TPG) was calculated under the assumption that PCWP is equal to the left atrial pressure (LAP) and was deducted from MPAP to receive the TPG. All hemodynamic parameters must have met the criteria in Figure 1 for at least >1 h for the diagnosis of PGD to be set. To determine the inotropic score, the highest dosage using the concentration in µg/kg/min per inotrope during the first 24 h was selected [12].

By analyzing all the perioperative and postoperative data during the first 24 h in the ICU we were able to identify a graft dysfunction (GD) and non-GD cohort. The data was meticulously reviewed to be able to determine whether PGD had occurred. The operative notes as well as the anesthesia reports were analyzed to exclude secondary graft dysfunction. The PGD group was then divided into four sub-groups using the ISHLT definition (Supplementary Figure SA1).

Continuous variables are presented by mean, standard deviation, median, minimum and maximum, and categorical variables by frequency and percentage.

For tests between PGD and no PGD group Fisher’s exact test was used for dichotomous variables, Mantel-Haenszel Chi-square trend test for ordered categorical variables, Chi-square test for non-ordered categorical variables and Mann-Whitney U-test for continuous variables. In the analyses on HTx level the observations were assumed to be independent given the small number of hearts with previous re-HTx (n = 21). All completed transplants were divided into three time-era periods (1984–1999, 2000–2010 and 2011-2022). Associations between pre-HTx and intraoperative clinically relevant a-priori selected variables and incidence of PGD were investigated using logistic regression, adjusted for age and sex. A multivariable model was obtained by running a stepwise selection (forward and backward) and keeping only statistically significant variables beside adjustment for age and sex. From these analyses odds-ratios with 95% confidence intervals (CI), p-values, and area under receiver operating characteristic curve (AUC or c-statistics) were presented. In these analyses only the first HTx was studied. Results from stepwise models should be interpreted as exploratory. No evidence of significant collinearity among the included variables was observed. Missing data were limited and managed using a complete-case approach in the multivariable models, with negligible expected impact on the results.

Associations between pre-HTx and intraoperative variables and the primary outcome all-cause death or re-HTx during first year were performed using Cox regression, adjusted for age and sex. A multivariable model was obtained by running a stepwise selection (forward and backward) and keeping only statistically significant variables beside adjustment for age and sex. From these analyses hazard ratios (HR) and adjusted HR (aHR) with 95% CI and p-values were presented. In the selected multivariable model, the assumption of proportional hazards was tested by introducing interaction terms between variables and log(time) which were non-significant. Hence, proportional hazards could be assumed. Descriptively, survival data was presented by Kaplan-Meier plots. In these analyses only first HTx was studied.

All tests were two-tailed and conducted at 0.05 significance level. All analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).

Among 830 subjects that underwent HTx between June 1984 and December 2021, 103 (12.4%) were excluded due to age <18 years and 60 (7.2%) due to incomplete medical records. A total of 667 (80.4%) HTx, in 646 unique patients, met the inclusion criteria and were included in the study (Supplementary Figure SA1). Of these, 504 (75.6%) were men and 163 (24.4%) women.

In total, 16 patients had secondary graft dysfunction due to, e.g., uncontrolled bleeding, stenosis of the vena cava anastomoses or hyperacute rejection. Our primary endpoint PGD was evident in 70 HTx patients and the overall incidence of PGD was 10.5%. This group was now compared to the non-PGD cohort that consisted of 597 HTx in 580 unique patients.

Other secondary endpoints, such as PGD-LV (n = 41,58.6%)) was more common than PGD-RV (n = 29) of the PGD cases. Among patients with PGD-LV, 5 had (12%) moderate and 36 (88%) the severe form. Patients who developed PGD were on average younger than those in the non-PGD group (45.5 vs. 49.7, p = 0.03). They were also shorter and had a lower body weight (173 cm vs. 176 cm; p = 0.0063 and 75 kg vs. 80kg; p = 0.0067, respectively). Patients with and without PGD had different diagnoses leading to HTx (p = 0.02). The PGD group more frequently had dialysis preoperatively (7.7% vs. 1.7%; p = 0.012), a history of cardiac surgery (60.3% vs. 37.3%, p = 0.0004) and pre-HTx ventricular assist device (VAD) (32.8% vs. 18.8%, p = 0.0098) compared to the non-PGD group. For further details see Table 1.

Other secondary outcomes, donor age for the whole cohort was 41.6 ± 14.9 years. In the donor population, men were overrepresented with 431 individuals (64.6%) compared to 236 women (35.4%). Intracerebral hemorrhage emerged as the predominant cause of death (49.0%), followed by trauma (26.7%). There were no differences in donor characteristics between recipients with or without PGD. Donor-related risk factors are presented in Table 2.

A trend towards longer graft ischemic time in patients with PGD was observed (195 min vs. 181 min, p = 0.06) (Table 3). When re-transplantations were excluded from the analysis, this difference was significant (196 min vs. 181 min, p = 0.035). Patients with PGD had longer extra corporeal circulation (ECC) times (median 193.0 min vs. 145.0 min, p-value <0.0001) indicating more complex procedures.

Postoperative secondary outcome data is presented in Table 4. The number of days in the ICU was longer in PGD patients with an average difference of more than 2 weeks (22.1 ± 23.1 vs. 7.8 ± 12.9, p < 0.0001). Postoperative mechanical circulatory support (MCS) was seen in 75.4% of patients with PGD and in 0.9% in the non-PGD group (p < 0.0001). PGD subjects required re-operation to a greater extent (70.1% vs. 27.2%, p < 0.0001). Post-operative dialysis was more frequent in the PGD cohort (50.0% vs. 19.0%, p < 0.0001) and the time with mechanical ventilation was longer (326 ± 458 h vs. 63 ± 167h, p < 0.0001). The proportion of patients who experienced post-operative sepsis was also higher in the PGD group (28.1% vs. 8.9%, p < 0.0001).

The 30-day mortality or re-transplant rate in the PGD group was 45.5% compared with 3.3% in the non-PGD group (p < 0.0001). Death during hospitalization was 47% in the PGD group and 4.1% in the non-PGD group (p < 0.0001). 1-year mortality or re-HTx was 51.5% in the PGD group and 7.6% in the non-PGD group (Supplementary Table SA1; Figure 1). The median survival for the PGD group was less than 1 year compared to more than 15 years for the non-PGD group (Figure 2).

Predictors of PGD in the age- and sex-adjusted logistic regression model were: age (adjusted (a)OR 0.80 (CI 0.66–0.96), p = 0.017; weight (aOR 0.90 (CI 0.82–0.98), p = 0.021); inverse height risk association (aOR 0.73 (CI 0.62–0.87), p = 0.0005); diagnosis (aOR 0.39 (CI 0.19–0.77), p = 0.0069); pre-HTx dialysis (aOR 7.03 (CI 1.79–27.5), p = 0.0051); pre-HTx sternotomy (aOR 2.51 (CI 1.46–4.29), p = 0.0008); ECC (aOR 1.75 (CI 1.45–2.11), p < 0.0001); arrhythmias pre-HTx (aOR 1.97 (CI 1.16–3.37), p = 0.013); and pre-HTx VAD (aOR 1.85 (CI 1.03–3.33), p = 0.039).

Independent predictors for PGD by multivariable logistic regression were: pre-HTx dialysis (aOR 6.73 (CI 1.46–31.6), p = 0.015); arrhythmias pre-HTx (aOR 2.08 (CI 1.16–3.75), p = 0.014); inverse height risk association (aOR 0.78 (CI 0.64–0.95), p = 0.012); longer ECC time (aOR 1.75 (CI 1.44–2.14, p < 0.0001), per 60 min increase) (Supplementary Table SA2).

A stepwise Cox regression analysis of 1-year outcomes of death from any cause and re-HTx (Supplementary Table SA3) identified 11 predictors in the univariable analysis, and three were independently significant after adjustment for age and sex in the multivariable analysis. In the multivariable analysis, PGD had an almost 7-fold higher risk of death vs. non-PGD (aHR of 6.89 [CI 4.01–11.83], p < 0.0001, Supplementary Table SA3). Individuals transplanted 2011–2022 had a lower risk of death or re-HTx compared to 1984–1999, (aHR 0.27 [CI 0.14–0.52], p = 0.0001). For every 60 min of increased ECC time, the risk of death or re-HTx was increased (aHR 1.38 [CI 1.17–1.63], p = 0.0002).

The limitations include the retrospective nature, some missing data and limited information from echocardiography that was simply not used routinely in the early eras of HTx. Furthermore, in the current study moderate PGD were diagnosed primarily based on hemodynamic criteria as echocardiography within 24 h of transplantation was usually not available. Thus, the incidence of moderate PGD could be underestimated leading to a PGD population consisting of more severe forms of PGD which in turn may have overestimated the mortality [14]. Furthermore, the lack of relevant donor data prevents us from investigating mismatch issues. In addition, the study spans many decades without detailing such things as preservation techniques and perioperative management. Finally the study has been worked up over several years but does not take into account the new and modified PGD definition.

However, the strength of this study is the consecutively transplanted and followed population, with all charts manually checked for PGD retrospectively, in a system where all patients can be tracked according to social security numbers. There are therefore no patients lost to follow-up.

In conclusion, primary graft dysfunction, using the ISHLT 2014 definitions, was a frequent and severe complication following HTx, characterized by poor prognosis in terms of both 30-day and 1-year survival rates. Four independent predictors for PGD were pre-HTx dialysis, pre-transplant arrythmias, inverse height association and longer ECC time. However, there were no donor variables identified as predictors for PGD. However, caution in interpretations is encouraged since cohort is relatively small.