Surgical aortic valve replacement (SAVR) is associated with an increased mortality risk in elderly or high-risk patients. Transcatheter aortic valve implantation (TAVI) is an alternative to surgery in patients with symptomatic severe aortic stenosis who are inoperable or at high operative risk. The impact of diabetes mellitus (DM) on patients referred to TAVI merits further investigation. The aim of our study was to evaluate the clinical characteristics and the impact of DM status on the updated Valve Academic Research Consortium 2–defined outcomes of TAVI and to stratify patient outcomes according to their initial glycated hemoglobin (HbA1c) levels. We enrolled and stratified patients who underwent TAVI at our institution according to DM status. A total of 586 patients were enrolled: 348 (59%) without DM and 238 (41%) with DM. There were no significant differences in 30-day mortality patients with diabetes compared to patients without diabetes (3.3% vs 2.9%, p = 0.974). Insulin-treated DM was not associated with adverse outcome in comparison to orally treated DM. To delineate the prognostic power of HbA1C in these patients, the cohort was divided into 3 groups according to HbA1C levels (<5.7%, 5.7% to 6.49%, and ≥6.5%). Patients with HbA1C ≥6.5% were at increased risk for mortality during follow-up (hazard ratio 2.571, 95% confidence interval 1.077 to 6.136, p = 0.033) compared to patients with HbA1C <5.7%. In conclusion, unlike SAVR, DM is not associated with an increased mortality risk after TAVI, nor is it associated with increased complications rates. A more poorly controlled disease, as manifested by elevated HbA1c levels, may be associated with increased mortality during long-term follow-up.
Transcatheter aortic valve implantation (TAVI) is an alternative to surgery for patients with symptomatic severe aortic stenosis who are inoperable or at high operative risk. Diabetes mellitus (DM) has been considered a marker of poor prognosis after cardiac surgery in the Society of Thoracic Surgeons (STS) and the EuroSCORE II risk scores. The Placement of Aortic Transcatheter Valves (PARTNER) trial is a randomized study that compared TAVI with the current standard of treatment, conventional SAVR. A post hoc analysis of the PARTNER trial examined outcomes stratified according to DM status of patients randomly assigned to undergo TAVI or SAVR. Results of that analysis suggested that there was a survival benefit, no increase in stroke, and less renal failure in the patients with DM who underwent TAVI compared with those who underwent SAVR. Conrotto et al recently reported that DM did not significantly affect complication rates in patients who underwent TAVI. Nonetheless, they found that insulin-treated DM, but not orally treated DM, was independently associated with death and myocardial infarction (MI) during a median follow-up of 400 days. To the best of our knowledge, no previous TAVI cohort study had used periprocedural glycated hemoglobin A1c (HbA1c) levels to report undiagnosed DM. The aim of our study was to evaluate the impact of DM status on the updated Valve Academic Research Consortium 2–defined outcomes of TAVI and to stratify patient outcome according to initial HbA1c levels.
Methods
We retrospectively analyzed the medical records of 586 consecutive patients, who underwent TAVI at the Tel Aviv Sourasky Medical Center from March 2009 to December 2014. Data were collected prospectively as part of the Tel-Aviv Prospective Angiography Study, as previously described, after obtaining informed consent from each patient as approved by the institutional ethics committee. As previously reported, severe aortic stenosis was diagnosed by a combination of clinical, echocardiographic, and hemodynamic criteria. A dedicated heart team, consisting of an interventional cardiologist, a cardiac surgeon, and a senior echocardiographist, decided on patient eligibility for TAVI based on predetermined criteria.
The TAVI procedure at our institute has been previously described in depth. In brief, 2 types of aortic valve prostheses were implanted: the CoreValve prosthesis (Medtronic, Minneapolis, Minnesota) and the Edwards Sapien XT prosthesis (Edwards Lifesciences, Irvine, California). Three senior interventional cardiologists performed the peripheral aspects of all the TAVI procedures (i.e., introduction of the sheaths, Prostar deployment, and the suturing of the entry ports). Valve type and size were planned before the procedure according to clinical, echocardiographic, angiographic, and computerized tomography parameters and at the discretion of the senior interventional cardiologist. The available valve sizes for the Edwards Sapien XT prosthesis were 23 and 26 mm, and the valve sizes for the CoreValve prosthesis were 26, 29, and 31 mm. After the valve was implanted, the patients were monitored at the cardiac intensive care unit, where vital signs were recorded continuously during the first 24 hours after the procedure and at least 3 times daily thereafter. All complications following the procedure were defined as such according to the Valve Academic Research Consortium.
The diagnosis of previous DM was documented according to patient history using a prespecified registry questionnaire, current admission, previous medical records, and baseline medications. The patients were further divided into insulin-treated DM or orally treated DM. Blood samples for HbA1C levels were obtained within the first 24 hours after the procedure. In the present analysis, we used previously described HbA1c cutoffs for defining no DM, pre-DM, and DM (<5.7%, 5.7% to 6.49%, and ≥6.5%, respectively) to stratify outcome. As recommended in those guidelines, patients without a known history of DM or under DM treatment with a HbA1c level of ≥6.5% were defined as having DM. This led to the classification of 26 patients (4%) with no reported history of DM or lacking DM treatment as having DM. Because prior studies relied on patient report for the definition of DM, all analyses were repeated with and without the guideline-derived definition of DM to compare our results to previous ones.
All data are displayed as mean (±standard deviation) for continuous variables and as the number (percentage) of patients in each group for categorical variables. The Student t test and the chi-square test were used to evaluate the statistical significance of differences between continuous and categorical variables, respectively. Multivariate adjusted Cox proportional hazard models were fitted for all-cause mortality as the dependent variable and adjusted to variables previously associated with mortality after TAVI. To avoid overfitting, we decided on a more restrictive variable selection and included only variables that have a p value <0.1 on univariate analysis and/or that are significantly different between the HbA1C groups and were not suspected for co-linearity. All the analyses were considered significant at a 2-tailed p value of <0.05. The SPSS statistical package was used to perform all statistical evaluation (version 21; SPSS, Chicago, Illinois).
Results
A total of 586 consecutive patients (57% women) who underwent TAVI at our institution from March 2009 to December 2014 were included in the study. HbA1c levels could be obtained for 531 (91%) of all patients included in the study. From the 586 patients, 212 had been previously diagnosed as having DM. An additional 26 patients (4%) had HbA1c levels of ≥6.5 g/dl and were newly defined as having DM, yielding a total of 238 patients with DM (41%). Demographic data, clinical characteristics, and baseline echocardiographic characteristics of the study population stratified by DM status are presented in Table 1 . The patients with DM were younger and had a higher prevalence of hypertension and dyslipidemia than the nondiabetic patients. The patients with DM had lower mean peak and mean aortic valve gradients, as measured by echocardiography at baseline. The DM and non-DM groups did not show any significant differences in other characteristics. The mean follow-up time was 641 and 729 days for the DM and non-DM groups, respectively.
| Variable | Total (n=586) | Diabetes Mellitus | P value | |
|---|---|---|---|---|
| Yes (n =238) | No (n= 348) | |||
| Age (mean ± SD) (years) | 82.58± 5.910 | 81.54± 6.215 | 83.3 ± 5.7589 | <0.001 |
| Women | 338 (57.7%) | 129 (54.2%) | 209 (60.1%) | 0.159 |
| Hypertension | 512 (87.4%) | 218 (91.6%) | 294 (84.5%) | 0.011 |
| Dyslipidemia | 475 (81.1%) | 212(89.1%) | 263 (75.6%) | <0.001 |
| Ever smoker | 166 (28.3%) | 68 (28.6%) | 98 (28.2%) | 0.931 |
| Peripheral Vascular Disease | 38 (6.5%) | 13 (5.5%) | 25 (7.2%) | 0.406 |
| Ejection Fraction < 50% | 98 (16.7%) | 38 (16%) | 60 (17.2%) | 0.685 |
| Coronary Artery Disease before evaluation | 357(60.9%) | 150 (63%) | 207 (59.5%) | 0.388 |
| Previous Myocardial Infarction | 111 (18.9%) | 51 (21.4%) | 60 (17.2%) | 0.204 |
| Prior coronary artery bypass graft surgery | 104 (17.7%) | 51 (21.4%) | 53 (15.2%) | 0.054 |
| Prior valve surgery | 12 (2%) | 7 (2.9%) | 5 (1.4%) | 0.207 |
| Prior Stroke | 67 (11.4%) | 26 (10.9%) | 41 (11.8%) | 0.749 |
| Atrial Fibrillation | 184 (31.4%) | 73 (30.7%) | 111 (31.9%) | 0.754 |
| Chronic Obstructive Pulmonary Disease | 98(16.7%) | 40 (16.8%) | 58 (16.7%) | 0.964 |
| Creatinine Clearance Test | 51.42±19.1 | 51.80±21.21 | 51.16±17.54 | 0.705 |
| Dialysis | 12 (2%) | 6 (2.5%) | 6 (1.7%) | 0.504 |
| Echocardiography parameters (Baseline) | ||||
| Echo Ejection Fraction (mean ± SD) | 55.74±8.070 | 55.73±8.481 | 55.75±7.790 | 0.975 |
| Echo peak gradient | 76.64 ±22.369 | 72.47±19.435 | 79.49±23.781 | <0.001 |
| Echo mean gradient | 46.86±14.607 | 44.07±12.873 | 48.74±15.414 | <0.001 |
| Echo Aortic Valve Area | 0.725±0.1795 | 0.745±0.1714 | 0.711±0.1837 | 0.023 |
| Echo EPASP | 38.67±16.574 | 40.06±17.057 | 37.75±16.207 | 0.120 |
| Risk scores to predict operative mortality | ||||
| STS Score | 4.18±2.5 | 4.7±2.6 | 3.8±2.4 | <0.001 |
| Euroscore 2 | 6.232±5.00 | 6.644±5.37 | 5.951±4.72 | 0.100 |
| Logistic Euroscore 1 | 16.38±10.38 | 16.3±10.78 | 16.431±10.09 | 0.879 |
There were no significant differences between the patients with and without DM regarding the incidence of all-cause mortality during hospitalization, during the first 30 days after undergoing TAVI, at 1 year, or during the entire post-TAVI follow-up ( Table 2 ). Similarly, no differences were found regarding MI, stroke, and reinterventions at follow-up. Kaplan–Meier curves for all-cause mortality up to 4 years of follow-up are presented in Figure 1 : there were no differences in mortality between patients with and without DM (p = 0.381 log-rank test).
| Variable | Total (n=586) | Diabetes Mellitus | P value | |
|---|---|---|---|---|
| Yes (n =238) | No (n= 348) | |||
| In Hospital Outcomes | ||||
| New Atrial Fibrillation | 33 (5.6%) | 10 (4.2%) | 23 (6.3%) | 0.214 |
| Conduction defect | 191 (32.6%) | 81 (34%) | 110 (31.6%) | 0.539 |
| New Pacemaker | 103 (17.6%) | 46 (19.3%) | 57(16.4%) | 0.357 |
| Venous thromboembolism | 4 (0.68%) | 0 | 4 (1.06%) | 0.252 |
| Sepsis | 12 (2) | 5 (2.1) | 7 (2) | 0.940 |
| In hospital mortality | 17 (2.9%) | 7(2.9%) | 10 (2.9%) | 0.962 |
| Hospitalization days | 8.09±5.27 | 8.17±6.09 | 8.03 ±4.63 | 0.742 |
| 30 Days Outcome | ||||
| Hospitalization for exacerbation/Pulmonary congestion | 26 (4.4%) | 12 (5%) | 14 (4%) | 0.556 |
| Myocardial Infarction | 0 | 0 | 0 | |
| Cardiogenic Shock | 13 (2.2%) | 6 (2.5%) | 7 (2%) | 0.681 |
| Respiratory Failure | 33 (5.6%) | 14 (5.9%) | 19 (5.4%) | 0.827 |
| Ventricular Tachycardia | 1 (0.2%) | 1 (0.4%) | 0 | 0.226 |
| Ventricular Fibrillation | 2 (0.3%) | 2 (0.8%) | 0 | 0.087 |
| Acute Kidney Injury Stage 2 or 3 | 12 (2%) | 7 (3%) | 5 (1.5%) | 0.207 |
| Vascular complications – Major | 52(8.9%) | 19 (8%) | 33 (9.5%) | 0.531 |
| Stroke | 9 (1.5%) | 3 (1.3%) | 6 (1.7%) | 0.654 |
| Conversion to open Surgery | 3 (0.5%) | 1 (0.4%) | 2 (0.6%) | 0.797 |
| Unplanned use of cardiopulmonary bypass | 0 | 0 | 0 | |
| Coronary Obstruction | 2 (0.3%) | 0 | 2 (0.6%) | 0.241 |
| Ventricular septal perforation | 0 | 0 | 0 | |
| Mitral valve damage | 2 (0.3%) | 1 (0.4%) | 1 (0.3%) | 0.787 |
| Tamponade | 7 (1.2%) | 1 (0.4%) | 6 (1.7%) | 0.154 |
| Endocarditis | 2 (0.3%) | 1 (0.4%) | 1 (0.3%) | 0.787 |
| Valve thrombosis | 0 | 0 | 0 | |
| Valve migration | 2 (0.3%) | 0 | 2 (0.6%) | 0.241 |
| Valve embolization | 5 (0.9%) | 2 (0.8%) | 3 (0.9%) | 0.978 |
| TAV-in-TAV | 3 (0.5%) | 1 (0.4%) | 2 (0.6%) | 0.797 |
| Mortality | 17 (2.9) | 7 (3% ) | 10 (2.9%) | 0.974 |
| Extended follow-up | ||||
| 1 year mortality ∗ | 65 (11.1%) | 31 (17.6%) | 34 (12.3%) | 0.114 |
| All time mortality | 122 (21%) | 51 (21.4%) | 71 (20.4%) | 0.764 |
Fifty-seven of the 238 patients with DM were untreated. In a subanalysis, the remaining 181 patients with DM were divided into insulin-treated versus orally treated groups ( Supplementary Table 1 ). Insulin-treated patients were younger than orally treated patients (78.3 vs 81.5, p = 0.002), and the incidence of a previous MI, a previous coronary artery bypass grafting, and a reduced ejection fraction was higher in the insulin-treated patients. Nevertheless, complications were not significantly higher for the insulin-treated patients with DM compared to the orally treated ones ( Supplementary Table 2 ).
Patients with HbA1c level ≥6.5 were younger with a higher incidence of dyslipidemia. They had larger aortic valve area and lower mean peak and mean aortic valve gradients, as measured by echocardiography at baseline ( Supplementary Table 3 ). There was a trend toward higher incidence of all-cause mortality during hospitalization with rising HbA1c levels, during the first 30 days after TAVI, at 1 year, or during the entire post-TAVI follow-up in univariate models ( Supplementary Table 4 ). However, when analyzing results between the HbA1c >6.5 patients versus HbA1c <5.7 patients, we found a significant correlation with all-cause mortality (p = 0.048)
Furthermore, in multivariate Cox regression analysis, HbA1c level ≥6.5 was independently correlated with all-cause mortality compared with HbA1c of <5.7% (hazard ratio [HR] 2.8, 95% confidence interval [CI] 1.18 to 6.64, p = 0.019), whereas an HbA1c level from 5.7 to 6.49 was found to be only marginally significant (multivariate Cox proportional hazard model: Table 3 , Figure 2 ). In a similar Cox proportional hazard model for all-cause mortality with extensive restriction on variable selection (variables with a p value <0.1 in the baseline characteristics according to HbA1C category, Supplementary Table 3 ), we found that HbA1c level ≥6.5 was independently correlated with all-cause mortality compared with HbA1c of <5.7% (HR 1.861, 95% CI 1.055 to 3.283, p = 0.032), whereas an HbA1c level from 5.7 to 6.49 was not (HR 1.470, 95% CI 0.9 to 2.401, p = 0.124; Supplementary Table 5 ). As mentioned previously, all these analyses were repeated with and without the guideline-derived definition of DM, with similar results.
