Results

Patient demographics and baseline characteristics

Patient demographics and baseline characteristics were shown in Table 1. The mean follow-up duration was 82 months, ranging from 70 to 175 months. None were lost to follow-up. There were 27 men (44.3%) and 34 women (55.7%). Age ranged between 42 and 77 years, with a mean of 58.6 ± 11.0 years. 24.6% had dyspnea with mild exertion or minimal activity, corresponding to New York Heart Association (NYHA) functional class III or IV status. 57.4% had a prior hospitalization for acute decompensated heart failure. 73.8% (n=45) had normal left ventricular ejection fraction. One in four patients had moderate-to-severe pulmonary hypertension, as defined by mean pulmonary arterial pressures.
Data regarding right ventricular systolic function was missing in 15 patients (21.7%) due to institutional database purge of aged data. For the patients with available data, one in five had right ventricular systolic dysfunction.
8.2% of the operations were performed on an urgent basis due to refractory heart failure or repeated hospitalizations for heart failure. 6.6% had prior cardiac surgery.

Operative details

Operative details were shown in Table 2. 86.9% (n=53) had mechanical aortic and mitral valve replacement. All tricuspid valves were successfully repaired. The aetiology of the aortic and mitral valve dysfunction was predominantly rheumatic (70.5%) and degenerative (26.2%). One patient had active aortic and mitral valve endocarditis at the time of surgery. Concomitant CABG was performed in six cases. The crossclamp and cardiopulmonary bypass times were 186 ± 30 and 132 ± 24 minutes, respectively.

Postoperative complications

Postoperative complications were shown in Table 3. 6.6% of patients required resternotomy for haemostasis. Perioperative cerebrovascular events were not uncommon. Four patients had an ischaemic stroke, and one became debilitated as a result; two patients suffered from intracranial haemorrhage which required neurosurgical intervention. None required permanent pacemaker implantation. 13.1% required renal replacement therapy due to acute renal failure, another 13.1% required prolonged ventilation.

Operative mortality and its association with pre-existing hepatorenal dysfunction

The prevalence and severity of hepatorenal dysfunction was shown in Table 4. 27.9% of all patients had at least one biochemical abnormality, namely hyponatremia (defined as plasma sodium less than 135mEq/L), hyperbilirubinemia (total bilirubin ≥ 2.0mg/dL), hypoalbuminemia (albumin <3.5g/dL) and elevated serum creatinine (over 200 ųmol/L or dialysis). One in ten patients had two or more deranged biochemical parameters.
MELD-Na was used as a surrogate measure of the severity of hepatorenal dysfunction. Most patients had a MELD-Na score less than 9 (68.9%), 26.2% had a moderately elevated score of 9 to 15, and 4.9% had a severely elevated score of greater than 15.
There were six operative deaths. One patient died from a perforated peptic ulcer and refractory sepsis. Five deaths can be attributed to multiorgan failure. The most common cause of demise was exacerbation of pre-existing hepatorenal dysfunction by cardiopulmonary bypass, systemic venous congestion, and post-bypass right ventricular dysfunction, culminating in frank multiorgan failure. Despite instititution of intra-aortic counter-pulsation and dialysis to reduce right ventricular afterload and right-sided filling pressures, mortality remained high. Patients who died exhibited more profound pre-existing hepatorenal dysfunction than those who survived, as reflected by a higher MELD-Na,t (61) = 8.91, p < 0.001.
Unadjusted operative mortality increased with increasing MELD-Na score, χ2(2) = 23.8, p = 0.000007. Kendall’s tau (τb) was 0.551, indicating a moderate association. Patients with a MELD-Na greater than 9 also suffered from more complications, including postoperative mechanical circulatory support, prolonged ventilation, need for dialysis and acute liver failure after TVS (Table 5).
A scatterplot of the MELD-Na and EuroSCORE values of the cohort segregated into survivors and non-survivors were shown in Figure 1 to facilitate visual comparison. The diagnostic accuracy of MELD-Na was compared with EuroSCORE using area under the receiver operating characteristic (ROC) curves. For MELD-Na score, the AUC was an impressive 0.992 [95% CI 0.925-1.000], indicating high diagnostic accuracy. The Youden Index was 0.982 at a cut-off of 13. For EuroSCORE, the AUC was 0.611 [95% CI 0.477-0.733], indicating low diagnostic accuracy.

Risk-adjusted association of MELD-Na with operative mortality and major morbidities

A hierarchical logistic regression model incorporating MELD-Na was constructed with the logistic EuroSCORE as the base model. EuroSCORE alone is weakly associated with operative mortality with an odds ratio of 1.135 [95% CI 1.016 – 1.268] (p = 0.025). After adjustment for baseline EuroSCORE, the regression model demonstrated continued associations between MELD-Na and operative mortality and morbidities (Table 6). Of note, each point of MELD-Na score increase was associated with 1.405 times increase in the odds of operative mortality (p = 0.015).
The regression analysis was repeated by incorporating individual components of the MELD-Na score, including bilirubin, sodium, and albumin. All three biochemical parameters were significantly associated with operative mortality (Table 7). Creatinine was excluded because it already is a component of the logistic EuroSCORE.

Long-term survival and cardiac events

Using the Kaplan-Meier method, the 5-year overall and event-free survival was 78.6% and 72.1% respectively (figure 2 panel A and B). 12.7% of patients had at least one subsequent hospitalization for congestive heart failure. There were two incidences of late prosthetic valve endocarditis and one patient died from septic shock. One patient suffered from structural valve degeneration of the bioprosthetic aortic valve requiring reoperation.
A total of 20 major bleeding events occurred over 424 person-years of follow-up. The unadjusted rate of bleeding with warfarin was 4.72% per person-year. The most common source of bleeding was from the gastrointestinal tract accounting for half of all bleeding events.