Patients’ mean age was 83.0±2.1 years and 39 (36.8%) were male (Table 1). The surgical procedure was performed electively in 96 patients (90.1%) and as an urgent intervention in the other 10 (9.9%). Twenty-four patients (22.6%) had left ventricular ejection fraction (LVEF) <50%, 37 (34.9%) were in NYHA class III or IV, 39 (36.8%) had angina, 16 (15.1%) had extracardiac arteriopathy and 18 (17.0%) presented creatinine clearance <30 ml/min/m2.

A bioprosthetic valve was implanted in all patients (Table 2); Mitroflow™ was the most widely used model (66%). The most used ring size was 21 mm, with a median of 21 (19 mm: 14 cases; 21 mm: 49 cases; 23 mm: 40 cases; 25 mm: 3 cases).

Median mechanical ventilation time was 10 h (IQR 8–14 h), median intensive care unit stay was 24 h (IQR 14–60 h) and median hospital stay was 9 days (IQR 9–12 days).

Operative mortality occurred in six patients (5.7%). Causes of death were heart failure in three cases (pump failure in two patients, patient-prosthesis mismatch in one), suspected aortic rupture in one, sepsis in one and pneumonia in one.

Mean logistic EuroSCORE I, EuroSCORE II and STS score were 14.6±11%, 4.4±3.1% and 4.0±2.4%, respectively (Table 3 and Figure 1). Calibration-in-the-large showed overestimation of operative mortality with logistic EuroSCORE I (14.6 vs. 5.6%, p=0.036), whereas EuroSCORE II (4.4 vs. 5.6%, p=1.0) and STS score (4.0 vs. 5.7%, p=1.0) exhibited similar mean values, closer to the observed mortality.

Figure 1.

Operative risk score calibration-in-the-large.

(0.07MB).

ROC curve analysis (Figure 2) yielded significant C-statistic values of 0.877 (95% CI 0.800–0.933) for logistic EuroSCORE I, 0.792 (95% CI 0.702–0.864) for EuroSCORE II and 0.702 (95% CI 0.605–0.787) for STS score. Comparison of ROC curves using the DeLong method was not significant for differences in discriminative power (Table 3).

Figure 2.

Distribution and receiver operating characteristic curves for logistic EuroSCORE I (a), EuroSCORE II (b) and STS score (c).

(0.19MB).

Non-fatal complications occurred in fifteen patients (14.2%). Two (1.9%) suffered an ischemic stroke (one with full recovery, one with mild residual hemiparesis), three (2.8%) required temporary hemodialysis and five (4.7%) had a permanent pacemaker implanted due to complete atrioventricular block. No MI or sternal wound infection was observed. Five (4.7%) patients required repeat thoracotomy due to hemodynamic instability or suspected ongoing blood loss.

In this series of octogenarians undergoing isolated AVR for severe symptomatic AS in a single center, operative mortality was 5.6%. In a population generally regarded as high-risk, we consider this figure acceptable.

Elayda et al. published one of the first surgical series, reporting a mortality of 5.3% in 77 patients aged 80 or over undergoing AVR, between 1975 and 1991.23 The authors found LVEF <45% to be an independent predictor of mortality, together with hypertension and associated CABG. Since then, the number of published surgical series and observational studies has increased modestly, but not without some discrepancies (Table 4). Gehlot et al. described an operative mortality of 13% in 322 octogenarians, of whom 43% had concomitant CABG and 56% were urgent or emergent procedures, both of them independent predictors of operative mortality.14 The presence of coronary artery disease requiring CABG and clinical instability are two recognized factors that can contribute to increased operative mortality, by 2- and 4–13-fold, respectively.24 Taken together, these two studies seem to suggest a lower risk associated with isolated elective AVR surgery in octogenarians. However, two other surgical series show higher than average mortality rates. Bloomstein et al. reported mortality of 16.7% in 180 septuagenarians and octogenarians undergoing isolated AVR surgery. In this series, a valve size of 19 mm was associated with significantly increased mortality (26.1% in 66 patients with 19 mm valve vs. 20% in 55 patients with 21 mm valve vs. 23% in 64 patients with 23–29 mm valve, p=0.017). Moreover, there was a strong correlation between small body surface area, small valve size, and female gender.25 In a series by Leontyev et al., operative mortality was also distinctly higher (10.8%) than in other studies. These authors included patients with low (logistic EuroSCORE I ≤10%), intermediate (10–20%) and high (≥20%) operative risk. However, in the 102 low-risk patients, operative mortality was 7.2%.26 More recently, Di Eusanio et al. reported 4.5% operative mortality in a large regional series of 638 octogenarians undergoing isolated AVR surgery, with a mean logistic EuroSCORE I of 13.0%±7.9, very similar to our results.12 In another large series, Saxena et al. reported a 4.0% 30-day mortality rate in 531 octogenarians undergoing isolated AVR surgery. Although no formal surgical risk score calculation was mentioned, precluding a more accurate comparison, these results are also in agreement with the present series.15

Understandably, naturally diminished global physiological and organ reserve makes this population more vulnerable to operative complications. Nevertheless, in the present series we observed a low rate of complications in isolated AVR, similar to those reported in the recent series by Di Eusanio et al. (1.3% for stroke, 0.3% for MI, 4.4% for complete atrioventricular block and 3.7% for redo thoracotomy).12

All scores showed good discriminative power in our series, but the logistic EuroSCORE I overestimated mortality by almost three times the observed value. Previous studies have demonstrated a systematic exaggeration of operative mortality by a factor of two to three.19,20,27–30 As EuroSCORE was originally devised in 1995, a large part of this discrepancy can probably be attributed to natural improvements in surgical techniques and post-operative patient management, and changes in population case mix and outcomes in recent years.12,31 By contrast, both EuroSCORE II and STS score showed excellent calibration.

Nevertheless, the logistic EuroSCORE I showed good discriminative power, which was not statistically different from EuroSCORE II or STS score, so ultimately no particular score was favored. Barili et al. also found this lack of added value when using ROC curves to compare the two versions of EuroSCORE in general cardiac surgery.18 In another study by the same authors, the STS score showed a slightly higher C-statistic, but again without statistically significant difference.29

Still, theoretical and methodological concerns have been voiced regarding calibration of EuroSCORE II, particularly regarding the validity of the Hosmer-Lemeshow test for this purpose. Barili et al. set out to answer such concerns in a recent study including more than 12000 patients undergoing cardiac surgery.18 The authors employed four different methods to assess calibration (including the use of calibration curves) and demonstrated that EuroSCORE II has an optimal calibration performance until 30% actual mortality, beyond which it progressively overestimates mortality. On the basis of these findings, EuroSCORE II could be considered an appropriate tool to assess operative risk in the general octogenarian population, including our study cohort. However, in a truly high-risk population, such lack of calibration can be crucial, particularly when considering transcatheter therapy.

Our findings appear to suggest that EuroSCORE II should be the preferred tool for estimation of operative risk mortality in octogenarians undergoing isolated AVR surgery. The STS score, although equally well calibrated, has the potential disadvantages of being derived from an American cohort and being more cumbersome and time-consuming due to a larger set of variables.

More than a decade after the introduction of TAVI for the treatment of severe AS, should age be a decisive factor in the decision whether to deny AVR surgery? Naturally, age per se, or as a surrogate of an increased prevalence of operative risk factors, is always present in all risk scores. An analysis by Iung et al. showed that low LVEF and greater age were significantly associated with the decision to deny surgery to 72 out of 216 patients with severe AS aged ≥75.32 Neurological dysfunction was the only comorbidity significantly linked with the decision not to operate, which seems to suggest that age itself, rather than associated comorbidities, is used as a true criterion more than a surrogate marker to assess a patient's pre-operative physiological status. Still, issues related to quality of life in the very old, which are inherently subjective and difficult to evaluate, should also be considered when deciding on the potential benefit of AVR.

Our patient cohort, although limited in size and inevitably suffering the limitations and biases of a single-center analysis, contributes to the growing evidence that isolated AVR surgery can indeed be performed with acceptable results in selected octogenarians. Furthermore, current age boundaries are progressively being challenged. Yerebakan et al. reported 7.6% overall operative mortality in a cohort of 119 nonagenarian patients, of whom 66 underwent isolated AVR.33

Proper patient selection has a decisive role in achieving good operative results. A number of conditions such as LVEF <30%, severe renal failure, severe lung disease, emergent surgery and simultaneous CABG have been identified as markers of poor outcome. Each of these variables, when added to age as the sole variable contributing to increased operative risk, may advise against AVR, not only because of the expected high operative mortality but also due to the reduced long-term benefit of the intervention.6,24

In an increasingly aging population, the number of patients who need a therapeutic intervention will continue to grow. From these results, and given the advent of TAVI and the data from the PARTNER trial, the question arises of what is the best treatment option.8,34 The answer can be found by an objective assessment of each case, and for the most difficult and doubtful cases, a multidisciplinary evaluation by the “heart team” would be desirable to find the best option for the patient. In this regard, the development of a new, integrated score, perhaps including variables such as frailty, could help to define the best treatment strategy for each individual case, similar to what the SYNTAX II score did for revascularization in coronary artery disease. Further studies are needed to confirm the feasibility of this strategy.

The authors declare that no experiments were performed on humans or animals for this study.

The authors declare that no patient data appear in this article.

The authors declare that no patient data appear in this article.