Data collection

This study involved a retrospective analysis of data from a prospectively collected cardiac surgery database (PATS; Dendrite Clinical Systems, Ltd, Oxford, UK). The PATS database contains comprehensive information on a wide range of variables related to cardiac surgery, including preoperative, intraoperative, and postoperative details, as well as complications and mortality outcomes, for all patients treated at our institution. The data undergo regular validation and are submitted annually to the National Institute for Cardiovascular Outcomes Research (NICOR) as part of the National Adult Cardiac Surgery Audit registry. Information about death was obtained from the institutional database and the NHS Spine for all patients. Data regarding postoperative complications and survival were available for all patients in the study. This study adhered to the ethical standards set forth in the Declaration of Helsinki, and it was approved by the local audit committee. Given the retrospective nature of the study, the requirement for individual patient consent was waived.

Study population

We collected data on all ARR procedures performed at our institution between July 2003 and December 2023. All valve-sparing techniques were excluded from this analysis, focusing solely on patients who required complete valve and root replacement. The final sample consisted of 594 patients, of whom 346 received a stentless aortic root (including 97 homografts and 249 xenografts), and 248 received a CVG (comprising 100 biological and 148 mechanical valves).

Selection of surgical strategy

The choice of surgical strategy was based on the surgeon’s preference as well as dictated by patient characteristics. All interventions were performed under general anaesthesia via a midline sternotomy. We have previously described in detail the techniques of homograft replacement and xenograft root replacement [4]. Briefly, a full ARR with coronary artery reimplantation was performed using the largest implantable prosthesis. There was no subcoronary implantation procedure, and no reinforcement techniques or synthetic material were used to support the surgical anastomoses in homografts. The proximal anastomosis between the stentless (homograft/xenograft) root and the native aortic annulus was performed using multiple interrupted 3/0 or 4/0 braided polyester nonabsorbable sutures. The left and then the right coronary ostia were reimplanted into the respective ostia of the bioprosthesis using a running 5/0 polypropylene suture. In cases where a greater segment of the ascending aorta required resection, a straight polyethylene terephthalate tube (Gelweave Graft, Terumo Aortic, Vascutek, Renfrewshire, UK), chosen intraoperatively on the basis of the outer diameter of the distal end of the stentless root and of the proximal aortic arch, was interposed using a running 4/0 polypropylene suture. Because the height and angle between the native porcine coronary ostia in xenograft prostheses may be different from that in normal human anatomy, new ostia were fashioned to avoid tension, torsion, or kinking of the proximal coronary arteries.

The modified Bentall was performed using CVG. The mechanical composite graft of choice was the CarboSeal^TM composite conduit (LivaNova, Saluggia, Italy) formed by assembling a standard Carbomedics bileaflet mechanical prosthesis into a 10 cm straight vascular graft. The bio-Bentall was created by incorporating a bioprosthetic valve within a straight Dacron tube. The valve was secured to the conduit with interrupted or running 4/0 non-absorbable, monofilament, synthetic polypropylene suture. The choice of valve model was influenced by surgeon’s preference. The valve size was selected according to the patient’s body surface area, securing adequate effective orifice area. A graft 3–5 mm larger than the valve was used. Interrupted everting pledgeted 2/0 polyester sutures or semi-continuous 3/0 non-absorbable, monofilament, synthetic polypropylene suture were placed on the aortic annulus and passed through the Valsalva graft collar as well as the cuff of the valve prosthesis simultaneously to complete the proximal anastomosis. The coronary ostia were trimmed with a full-thickness 7–8 mm cuff of the aortic wall to create buttons and anastomosed with a 5/0 polypropylene running suture to corresponding holes in the Dacron graft.

Statistical analysis

Data description and analysis was performed in SPSS (SPSS Statistics 29.0.2.0, IBM Corp, Armonk, NY). Continuous variables are expressed as mean ± standard deviation if normally distributed and as median and interquartile range for the numerical variables not normally distributed. Categorical variables are shown as count and percentages. The Lilliefors (Kolmogorov-Smirnov) test was used for normality assessment. Comparison between continuous variables has been conducted using unpaired Student’s t-test if normally distributed and Mann-Whitney U-test if not normally distributed. Categorical variables have been compared using Pearson Chi-square test or Fisher’s exact test as appropriate.

To control for potential confounders within the dataset, we generated a propensity score for each patient using a multivariable logistic regression model. All 27 preoperative clinical variables listed in Table 1 were included as covariates in the logistic regression model to estimate the propensity score, with treatment type (SARR vs. CVG) as the binary dependent variable. Pairs of patients receiving either SARR or CVG were matched using a greedy 1:1 matching algorithm with a caliper of width equal to 0.2 standard deviations of the logit of the propensity score. The quality of the match was assessed by comparing selected pretreatment variables between the propensity score-matched groups using the standardized mean difference, where an absolute standardized difference of greater than 10% was considered indicative of meaningful covariate imbalance.

A multiple logistic regression model was employed to identify predictors of in-hospital mortality and long-term survival rates. After performing univariable analysis, the final multivariable models were constructed using a stepwise approach, with P values guiding the selection of variables. The variables with a P value < 0.05 were included in the multivariable logistic regression model. Kaplan-Meier curves were generated to visualize survival outcomes across the study groups. The survival functions of the groups were compared using the log-rank test, with survival probabilities and 95% confidence intervals (CIs) reported at 5, 10, 15, and 20 years. Additionally, a Cox proportional hazards model was utilized to identify factors influencing long-term survival with all tests set to a significance level of 0.05.

Patient demographics

The preoperative demographics of the study population revealed some significant differences between patients who underwent SARR and those who received a CVG before matching. Patients in the stentless aortic root group were generally older, with an average age of 60.2 years compared to 56.5 years in the CVG group (P < 0.001). The stentless group also had a higher proportion of females (28% vs. 20.6%, P = 0.019) and a slightly lower mean body mass index (BMI) (27.4 vs. 28.2, P = 0.037). Additionally, there were more patients with past or current heart failure and previous cardiac surgeries in the stentless group. After propensity score matching (PSM), these demographic differences were effectively balanced between the two groups, resulting in well-matched cohorts with no significant differences in key variables such as age, sex, BMI, New York Heart Association (NYHA) class, or previous cardiac history. The preoperative clinical characteristics for the patients are presented in Table 1.

Intraoperative characteristics

The intraoperative characteristics of the study population revealed several key differences between the patients who underwent SARR and those who received a CVG. Before matching, the mean cross-clamp time was significantly longer in the SARR group compared to the CVG group (159.6 ± 113.3 minutes vs. 134.1 ± 44.1 minutes, P < 0.001). Additionally, the stentless group had a shorter mean circulatory arrest time (2.1 ± 12 minutes vs. 6.6 ± 15.7 minutes, P < 0.001) and a higher lowest temperature during surgery (31.1° ± 4.2°C vs. 29.5° ± 6.0°C, P < 0.001). The use of coronary artery bypass grafting (CABG) also showed some variation between the groups.

After PSM, most of these differences were minimized, with the two groups becoming more comparable. The total duration of operation, cardiopulmonary bypass (CPB) time, and cross-clamp time were similar between the matched groups. However, the stentless aortic root group continued to have a slightly shorter mean circulatory arrest time (2.37 ± 14.1 minutes vs. 6.46 ± 15.7 minutes, P = 0.003) and a higher lowest temperature during surgery (31.4° ± 4.3°C vs. 29.7° ± 5.8°C, P < 0.001). The distribution of cardioplegia types and infusion modes, as well as the number of valves replaced and use of CABG, were well balanced between the groups post-matching. The intraoperative characteristics for the patients are presented in Table 2.

In-hospital outcomes

The analysis of in-hospital outcomes revealed no significant differences between the SARR group and the CVG group, both in the unmatched and matched cohorts, in terms of key outcomes such as low cardiac output syndrome, reoperation rates, and postoperative renal replacement therapy. Additionally, both groups had similar lengths of postoperative hospital stay and in-hospital mortality rates. However, neurological complications were consistently higher in the CVG group, with the unmatched analysis showing 8.4% of patients in the CVG group experiencing neurological events compared to 4.9% in the SARR group (P = 0.014). This trend persisted in the matched analysis, where neurological complications occurred in 8.5% of the CVG group compared to 4.3% in the SARR group (P = 0.022). These detailed in-hospital outcomes are presented in Table 3.

The univariable analysis (Table 4) identified several predictors of in-hospital mortality. Key factors that were significantly associated with higher in-hospital mortality in both the unmatched and matched cohorts included older age, higher NYHA classification, previous cardiac surgery, diabetes, hypertension, renal disease, extensive coronary artery disease (CAD), aortic regurgitation (AR), preoperative cardiogenic shock, the need for preoperative inotropes or intubation, and aortic dissection as the underlying pathology. Notably, the incidence of cardiogenic shock and AR demonstrated a particularly strong association with in-hospital mortality, as evidenced by high odds ratios (ORs) in both analyses.

In the multivariable analysis (Table 5), several of these factors remained significant independent predictors of in-hospital mortality. In the matched cohort, advanced age (OR 1.085, 95% CI 1.030–1.142, P = 0.002), AR (OR 7.963, 95% CI 2.448–25.902, P < 0.001), operative urgency (OR 2.425, 95% CI 1.090–5.395, P = 0.030), and longer CPB time (OR 1.015, 95% CI 1.009–1.022, P < 0.001) were all significant predictors of mortality.

Long-term survival

The Kaplan-Meier survival analysis for the unmatched groups (Figure 1, Table 6) revealed no significant difference in long-term survival between the CVG and SAAR groups. Although the SARR group showed a slight trend toward better survival, these differences were not statistically significant.

Display full size

Figure 1. Unmatched Kaplan-Meier survival estimates for composite valve graft and stentless aortic root replacement

In the propensity score-matched cohorts (Figure 2, Table 7), the SARR group demonstrated a statistically significant survival advantage, with a 20-year survival rate of 54% compared to 47% for the CVG group (log-rank P value of 0.047), suggesting a potential benefit for SARR in long-term outcomes when patient characteristics are well-matched.

Display full size

Figure 2. Propensity score matched Kaplan-Meier survival estimates for composite valve graft and stentless aortic root replacement

Further analysis by specific graft type within the unmatched groups (Figure 3, Table 8) showed no significant differences in long-term survival across the different graft types, with the log-rank test yielding a P value of 0.069. While xenografts exhibited the highest survival rates, and biological CVGs the lowest, these trends did not reach statistical significance.

Display full size

Figure 3. Unmatched Kaplan-Meier survival estimates for biological CVG, homograft, mechanical CVG, and xenograft

In the propensity score-matched analysis by graft type (Figure 4, Table 9), xenografts continued to show superior outcomes, with a statistically significant survival advantage (log-rank P value of 0.009). Mechanical CVGs also performed well, although long-term data beyond 10 years were limited.

Display full size

Figure 4. Matched Kaplan-Meier survival estimates for biological CVG, homograft, mechanical CVG, and xenograft

The Cox regression analysis identified several key variables associated with long-term mortality (Table 10). In the unmatched analysis, older age, female gender, AR, longer CPB time, and postoperative complications such as deep sternal wound infection (DSWI) and postoperative ischemic bowel disease were significant predictors of increased long-term mortality.

In the propensity score-matched cohort, age [hazard ratio (HR) 1.052, 95% CI 1.025–1.079, P < 0.001], AR (HR 3.961, 95% CI 2.189–7.166, P < 0.001), longer CPB time (HR 1.009, 95% CI 1.005–1.013, P < 0.001), and postoperative complications such as DSWI (HR 21.131, 95% CI 1.091–409.111, P = 0.044), postoperative ischemic bowel disease (HR 36.975, 95% CI 4.654–293.770, P < 0.001), and postoperative multi-organ failure (HR 11.772, 95% CI 3.268–42.403, P < 0.001) were significant predictors of mortality. Additionally, previous cardiac surgery (HR 2.095, 95% CI 1.029–4.266, P = 0.042) emerged as a significant factor in the matched cohort.