The aim of this study was to identify predictors of 30-day and midterm mortality after transcatheter aortic valve implantation (TAVI) by means of a systemic review. TAVI was demonstrated to be safe and efficacious in patients with severe aortic stenosis. An accurate estimation of procedural risk of these patients represents an actual challenge. The PubMed and Cochrane Collaboration databases were systematically searched for studies reporting on the incidence and independent predictors of 30-day and midterm mortality. Adverse events were pooled with random effect, whereas independent predictors are reported as odds ratios (ORs) with 95% confidence intervals (CIs). A total of 25 studies with 8,874 patients were included (median age 82.5 ± 1.5 years, 54.6% women). At 30 days, 7.5% of patients (n = 663) died. At midterm follow-up (median 365 days, interquartile range 267 to 365 days), the cumulative mortality rate was 21.6% (n = 1,917). Acute kidney injury (AKI) stage ≥2 (OR 18.0, 95% CI 6.3 to 52), preprocedural hospitalization for heart failure (OR 9.4, 95% CI 2.6 to 35), periprocedural acute myocardial infarction (OR 8.5, 95% CI 2.6 to 33.5), and increased pro–brain natriuretic peptide (pro-BNP) levels (OR 5.4, 95% CI 1.7 to 16.5) were the most important independent predictors of 30-day mortality. Increased pro-BNP levels (OR 11, 95% CI 1.5 to 81), AKI stage 3 (OR 6.8, 95% CI 2.6 to 15.7), left ventricular ejection fraction <30% (OR 6.7, 95% CI 3.5 to 12.7), and periprocedural acute myocardial infarction (OR 6.5, 95% CI 2.3 to 18.1) represented the predictors of midterm mortality. In conclusion, in this large meta-analysis of patients undergoing TAVI, we found that high pro-BNP levels and postprocedural AKI were the strongest independent predictors of both 30-day and 1-year mortality. These findings may contribute to a better understanding of the risk assessment process of patients undergoing TAVI.
Highlights
- •
This review reports on the independent predictors of mortality after transcatheter aortic valve implantation.
- •
High pro–brain natriuretic peptide levels and postprocedural acute kidney injury predict 30-day and 1-year mortality.
- •
These findings may contribute to a better risk assessment of patients undergoing transcatheter aortic valve implantation.
Despite being one of the most important aspects of transcatheter aortic valve implantation (TAVI) practice, adequate patient selection represents a clinical issue not fully resolved. Allocating transcatheter treatment to the correct patients, reasonably expected to benefit in terms of functionality and survival, is essential to avoid unnecessary high-risk procedures with accompanying costs. However, assessment of patient eligibility is often complex because of the extensive co-morbidity encountered in candidates for TAVI, rendering it difficult to estimate whether a beneficial treatment effect can be expected in individual patients, especially because surgical risk scores (EuroSCORE and Society of Thoracic Surgery score) are less reliable in patients undergoing TAVI. The primary challenge to improve patient selection and counseling is hindered by the search for accurate and validated risk factors for mortality after TAVI. The investigation of predictors of mortality may yield helpful tools that could facilitate decision making of the heart team. Contemporary literature falls short in addressing this issue, despite the availability of numerous publications, as the knowledge on predictors of mortality is fragmentary. Meta-analyses have currently been performed only for reporting the incidence of TAVI-related complications or identifying predictors of specific adverse events, for example, paravalvular leak. However, a comprehensive analysis of predictors of 30-day and midterm mortality is still lacking. The aim of this study was thus to identify predictors of post-TAVI 30-day and midterm mortality by means of a systemic review of the currently available literature.
Methods
This study was conducted in accordance with the current guidelines, including the recent Preferred Reporting Items for Systematic reviews and Meta-Analyses amendment to the Quality of Reporting of Meta-analyses statement, and recommendations from The Cochrane Collaboration and Meta-analysis Of Observational Studies in Epidemiology (MOOSE).
PubMed and Cochrane databases were searched for reports published in English from January 2002 to June 2013 according to the following highly sensitive strategy, in compliance with established methods and incorporating wildcards (identified by *): (TAVI* OR TAVR OR “transcatheter aortic”) AND (“adverse events” OR “30 days events” OR “mid-term prognosis”) AND english[lang] AND (“c”[pdat]:“3000”[pdat]) NOT (review[pt] OR editorial[pt] OR letter[pt]).
Retrieved citations were first screened for relevance independently by 2 reviewers (Francesca Giordana and FDA) at the title and/or abstract level. The reports remaining after the initial screening were appraised in the full text with respect to the following inclusion criteria: (1) human study, (2) investigating patients undergoing TAVI, (3) reporting predictors of mortality at 30 days or at midterm follow-up, and (4) online full-text available publication. All criteria had to be met for inclusion. Exclusion criteria were (1) duplicate reporting and (2) study population <50 patients. The presence of even one of the exclusion criteria sufficed for exclusion. Duplicate reporting was handled by selecting the study reporting on the largest sample of patients undergoing TAVI, or if equal, the study with the largest number of overall patients.
The quality and validity of the included studies were independently assessed by 2 reviewers (FG and FDA). After modifying the MOOSE item list, to ensure compatibility with the included studies, each study was summarized and critically appraised with respect to study design, data source, the statistical methods used for multivariate analysis, and the risk of bias (graded as low, moderate, high, or unknown because incomplete reporting may render it impossible to ascertain the risk of bias). Encountered bias was further subdivided into analytical, selection, adjudication, detection, and attrition bias.
Relevant study data were extracted independently by the reviewers (FG and FDA). The extracted data comprised authors and journal names, year of publication, location of the study group, baseline patient characteristics, procedural features, and multivariate predictors of all-cause mortality (estimator, point summary estimate of risk, 95% confidence intervals [CIs]). End points of interest for the present review were the incidence of TAVI-related complications and 30-day and midterm all-cause mortality. Multivariate predictors with an odds ratio (OR) >5 reported in at least 3 studies were reported. Meta-regression analysis was performed for midterm all-cause mortality for baseline clinical variables.
Result analysis of the studies was registered on dedicated electronic forms. The forms were piloted over the first 5 cases for consistency and discrimination. Any divergence in opinion between the reviewers at any stage of the review process was resolved by consensus discussion.
Continuous variables are reported as mean ± SD or median and range. Categorical variables are expressed as counts and percentages. Independent predictors of TAVI-related complications and 30-day and midterm all-cause mortality were appraised and ordered according to their measure of risk (either OR or hazard ratio) with Review Manager (RevMan), version 5.2, freeware package (The Cochrane Collaboration, The Nordic Cochrane Center, Copenhagen, Denmark) and Comprehensive Meta-Analysis. Small study bias was appraised by graphical inspection of funnel plots. Hypothesis testing for superiority was set at the 2-tailed 0.05 level. Hypothesis testing for statistical homogeneity was set at the 2-tailed 0.10 level and based on the Cochran Q test, with I 2 values of 25%, 50%, and 75% representing, respectively, mild, moderate, and extensive statistical inconsistency.
Results
The search strategy yielded 1,088 reports. Four reports derived from congresses were added, resulting in a total of 1,092 citations. All citations were first screened at title and abstract level; the 31 remaining reports were subsequently screened full text. Of the full-text analyzed citations, 4 were excluded because of reporting duplicate data and 2 because of the absence of data on independent predictors for adverse events. Twenty-five studies were finally included in this systemic review ( Figure 1 ).
The 25 included studies reported on a total of 8,874 patients treated with TAVI for symptomatic severe aortic stenosis (AS). Approximately 1/2 of the patients were women (54.6%); median age was 82.5 ± 1.5 years. Mean logistic EuroSCORE-I was 23.4 ± 3.1% and was above 20% in all studies, except for one (19 ± 13%). In those studies (n = 5) in which reporting results was exclusively done using the Society of Thoracic Surgery score, the mean risk of perioperative mortality was >5%. Diabetes mellitus was present in 24% (n = 2,153) of patients, renal insufficiency in 48% (n = 4,264), and a history of previous revascularization (either surgical or percutaneous) in 40% (n = 3,541). Left ventricular ejection fraction (LVEF) was in general preserved, except for 3 studies in which mean LVEF was mildly compromised (range of mean LVEF 47% to 60%). The most important baseline characteristics are reported in Table 1 .
| First Author | Recruitment Years | Number | Age | Women | DM | NYHA ≥III | CAD ∗ | Previous AMI | Previous Revascularization (Surgical or Percutaneous) | Previous Cardiac Surgery | COPD | GFR <60 ml/min/m 2 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Thomas (2011) | 2007–2009 | 1038 | 81 | 576 (56%) | — | 146 (14%) | 537 (52%) | — | 495 (48%) | 236 (23%) | 286 (28%) | 305 (29%) |
| Kempfert (2011) | 2006–2010 | 299 | 82 | 209 (70%) | 115 (52%) | 252 (84%) | 159 (53%) | 8 (3%) | — | 86 (29%) | 129 (43%) | 26 (9%) |
| Parenica (2012) | — | 29 | 82 | 21 (72%) | 3 (10%) | 26 (90%) | — | — | — | 1 (3%) | 5 (17%) | — |
| Halliday (2012) | 2008–2010 | 101 | 83 ± 1 | 52 (51%) | 22 (22%) | — | — | — | — | — | — | — |
| Yong (2012) | 2007–2011 | 119 | 81 ± 8 | 72 (61%) | 31 (26%) | — | 24 (20%) | 22 (18%) | 41 (34%) | 15 (13%) | 40 (34%) | — |
| Barbash (2012) | 2007–2011 | 165 | 85 ± 6 | 96 (58%) | 53 (32%) | — | 92 (56%) | — | 81 (49%) | 54 (33%) | 45 (69%) | 128 (78%) |
| Sinning (2012) | 2009–2010 | 1315 | 82 ± 6 | 765 (58%) | 462 (35%) | 1166 (89%) | 791 (60%) | 209 (16%) | 693 (53%) | 289 (22%) | 314 (24%) | 786 (60%) |
| Buellesfeld (2012) | 2007–2010 | 353 | 83 | 203 (58%) | 92 (26%) | 263 (75%) | 203 (58%) | 61 (17%) | 151 (43%) | 82 (23%) | — | 77 (22%) |
| Humphries (2012) | 2005–2011 | 641 | 83 | 329 (5%1) | 197 (31%) | 554 (86%) | 467 (73%) | 259 (40%) | 328 (51%) | — | 170 (27%) | 413 (64%) |
| Tchetche (2012) | 2010–2011 | 134 | 82 ± 5 | 53 (40%) | 30 (22%) | 111 (83%) | 56 (42%) | — | 56 (42%) | 16 (12%) | 34 (25%) | 37 (28%) |
| Akin (2012) | 2007–2008 | 45 | 82 ± 7 | 27 (60%) | 17 (38%) | 43 (91%) | — | — | — | 3 (6%) | 8 (2%) | 25 (56%) |
| Rodes-Cabau (2012) | 2005–2009 | 339 | 81 ± 8 | 187 (55%) | 79 (23%) | 308 (90%) | 234 (69%) | 173 (51%) | 215 (63%) | 116 (34%) | 100 (30%) | — |
| Latib (2012) | 2003–2011 | 111 | 81 ± 7 | 49 (44%) | 21 (19%) | 75 (67.6%) | 44 (40%) | 16 (14%) | .. | 0 | 29 (26%) | — |
| Pilgrim (2012) | 2007–2011 | 389 | 82 ± 6 | 224 (58%) | 105 (27%) | — | 238 (61%) | 64 (17%) | 166 (43%) | 72 (18%) | 72 (19%) | 268 (69%) |
| Hayashida (2012) | 2006–2011 | 400 | 83 ± 6 | 206 (52%) | 92 (23%) | 347 (87%) | 237 (59%) | — | — | 63 (16%) | 124 (31%) | 249 (63%) |
| Généreux (2013) | 2008–2011 | 218 | 85 ± 8 | 106 (48%) | 62 (29%) | — | — | — | 177 (80%) | 89 (40%) | 64 (29%) | 18 (8%) |
| Amabile (2013) | 2008–2012 | 173 | 84 ± 6 | 91 (53%) | 43 (25%) | — | 107 (63%) | — | — | — | 46 (27%) | 121 (71%) |
| Van der Boon (2013) | — | 940 | 81 ± 7 | 434 (46%) | 268 (29%) | 761 (81%) | 425 (45%) | 158 (17%) | 484 (52%) | 207 (22%) | 323 (34%) | 591 (63%) |
| Toggweiler (2013) | 2005–2007 | 88 | 87 ± 7 | 41 (47%) | 22 (25%) | — | 63 (72%) | 69 (78%) | — | 34 (39%) | 23 (26%) | 47 (53%) |
| Stortecky (2013) | 2007–2011 | 389 | 83 ± 6 | 224 (58%) | 105 (27%) | 255 (66%) | 238 (61%) | 64 (16%) | 166 (43%) | 72 (19%) | — | 268 (69%) |
| Codner (2013) | 2008–2012 | 153 | 82 ± 6 | 95 (62%) | 45 (29%) | 149 (87%) | — | 11 (7%) | 82 (54%) | 38 (25%) | 43 (28%) | 60 (39%) |
| Borz (2013) | 2006–2011 | 250 | 83 ± 7 | 135 (54%) | 64 (26%) | — | 87 (35%) | — | — | 49 (20%) | — | — |
| D’Onofrio (2013) | 2008–2012 | 774 | 81 ± 7 | 446 (58%) | 205 (27%) | 621 (80%) | 168 (22%) | 23 (3%) | 226 (29%) | 167 (33%) | 247 (32%) | 80 (10%) |
| López-Otero (2013) | 2008–2011 | 85 | 83 ± 5 | 31 (37%) | 20 (24%) | 70 (82%) | 9 (9%) | — | — | — | 20 (24%) | — |
| Seiffert (2013) | 2008–2011 | 326 | 81 ± 1 | 181 (56%) | — | 266 (82%) | 201 (62%) | 65 (20%) | 180 (55%) | 65 (20%) | 87 (27%) | 29 (9%) |
A small majority of patients were treated by the transfemoral approach (51.1%), followed by the transapical approach (33.7%), and only few patients received a direct aortic or trans-subclavian procedure (1.5% and 0.2%, respectively). In 60.8% (n = 5,392) of patients, an Edwards SAPIEN or SAPIEN XT balloon-expandable valve (both Edwards Lifesciences, Irvine, California) was implanted, and in 21.4% (n = 1,899) of patients, the self-expandable third-generation Medtronic CoreValve (Medtronic Inc., Minneapolis, Minnesota) prostheses were used, whereas 3 studies (comprising 15.8% of patients, n = 1,583) did not specify the type of valve implanted. Because of the infrequent reporting of conversion to surgery and intraprocedural death, no reliable incidence could be reported for these events. Table 2 summarizes procedural TAVI characteristics in the various studies.
| First Author | LVEF | Aortic Valve MG | MR >2+ | Pulmonary Hypertension | TF | TA | TS | TAo | EDW | CV | Log ES | STS Score |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Thomas (2011) | — | — | 257 (25%) | — | 463 (45%) | 575 (55%) | 0 | 0 | 1038 (100%) | 0 | 28 | — |
| Kempfert (2011) | 55 ± 13 | — | 3 (1%) | 81 (27%) | 0 | 299 (100%) | — | — | 299 (100%) | 0 | 31 ± 16 | 13 ± 8 |
| Parenica (2012) | 57 | 50 | — | 11 (38%) | 14 (25%) | 15 (26%) | — | — | 29 (100%) | 0 | 24 | — |
| Halliday (2012) | — | — | — | — | 45 (45%) | 56 (55%) | 0 | 0 | — | — | 22 ± 0.9 | — |
| Yong (2012) | — | 49 ± 16 | 10 (8%) | — | 119 (100%) | 0 | 0 | 0 | 0 | 119 (100%) | 19 ± 13 | 6 ± 5 |
| Barbash (2012) | 51 ± 15 | — | — | — | 117 (71%) | 48 (29%) | 0 | — | — | — | — | 12 ± 4 |
| Sinning (2012) | 52 ± 15 | — | — | — | 1143 (87%) | 121 (9%) | 41 (3%) | 10 (1%) | — | — | 21 ± 14 | — |
| Buellesfeld (2012) | 50 | 44 | — | 88 (25%) | 353 (100%) | 0 | 0 | 0 | 34 (10%) | 319 (90%) | 26 | — |
| Humphries (2012) | — | 41 | 166 (26%) | — | 351 (55%) | 290 (45%) | 0 | 0 | 622 (97%) | 19 (3%) | — | 8 |
| Tchetche (2012) | 47 ± 13 | 48 ± 15 | 0 | — | 125 (93%) | 0 | 9 (7%) | 0 | 0 | 134 (100%) | 24 ± 10 | — |
| Akin (2012) | 48 | 57 | 41 (85%) | — | 45 (100%) | — | — | — | 45 (100%) | — | 21 | — |
| Rodes-Cabau (2012) | 55 ± 14 | — | 27 (8%) | 84 (25%) | 163 (48%) | 176 (52%) | 0 | 0 | 339 (100%) | 0 | — | 10 ± 6 |
| Latib (2012) | 54 ± 13 | — | — | — | 111 (100%) | 0 | 0 | 0 | 70 (63%) | 41 (37%) | 23 ± 15 | 4.6 ± 2.3 |
| Pilgrim (2012) | 52 ± 15 | 44 ± 17 | — | 111 (31%) | 308 (80%) | 76 (20%) | 5 (1%) | 0 | 164 (42%) | 225 (58%) | 24 ± 14 | 6.8 ± 5.3 |
| Hayashida (2012) | — | — | — | — | — | — | — | — | 347 (87%) | 53 (13%) | 22 | 8 |
| Généreux (2013) | 48 ± 16 | 45 ± 15 | — | — | 140 (64%) | 78 (36%) | 0 | 0 | 218 (100%) | 0 | — | 12 ± 5 |
| Amabile (2013) | 55 ± 15 | 49 ± 18 | — | — | 139 (81%) | 32 (19%) | 0 | 0 | 132 (77%) | 39 (23%) | 22 ± 12 | — |
| Van der Boon (2013) | — | — | — | — | 79 (84%) | 89 (10%) | 57 (6%) | 4 (0.4%) | 435 (46%) | 505 (54%) | 21 | — |
| Toggweiler (2013) | 60 | 46 ± 18 | — | 17 (19%) | 64 (73%) | 24 (27%) | 0 | 0 | 88 (100%) | 0 | — | 9 |
| Stortecky (2013) | 52 ± 15 | 44 ± 17 | — | — | 308 (79%) | 76 (20%) | 5 (1%) | 0 | 165 (42%) | 224 (58%) | 24 ± 14 | 7 ± 5 |
| Codner (2013) | — | 51 ± 15 | 52 (33%) | — | 112 (73%) | 27 (18%) | 13 (9%) | 1 (1%) | 62 (41%) | 91 (60%) | 23 ± 13 | 9 ± 5 |
| Borz (2013) | — | 45 ± 17 | — | — | 190 (76%) | 60 (24%) | 0 | 0 | 250 (100%) | 0 | 23 ± 13 | — |
| D’Onofrio (2013) | 53 ± 13 | 50 ± 15 | 173 (22%) | 87 (11%) | 0 | 774 (100%) | 0 | 0 | 774 (100%) | 0 | 26 ± 16 | 10 ± 8 |
| López-Otero (2013) | — | — | — | — | — | 0 | — | 0 | 0 | 85 (100%) | 29 ± 8 | — |
| Seiffert (2013) | — | 37 ± 2 | 30 (9%) | 48 (19%) | 149 (45%) | 177 (51%) | 0 | 0 | 281 (81%) | 45 (13%) | 23 ± 2 | 8 ± 1 |
At 30 days, 7.5% (n = 663) of patients died. Acute kidney injury (AKI) occurred in 8.02% (n = 712), life-threatening and major bleeding in 13.8% (n = 1,224), major vascular complications in 8.8% (n = 782), periprocedural acute myocardial infarction (AMI) in 0.6% (n = 51), and pacemaker implantation in 12.5% (n = 1,106) of patients. At midterm follow-up (median 365 days, interquartile range 267 to 365), 21.6% (n = 1,917) of patients had died. The strongest predictors of 30-day mortality were AKI stage ≥2 (OR 18.0, 95% CI 6.25 to 52), preprocedural hospitalization for at least 1 week (OR 9.36, 95% CI 2.55 to 35), periprocedural AMI (OR 8.54, 95% CI 2.57 to 33.52), and preprocedural increased pro–brain natriuretic peptide (pro-BNP) levels (OR 5.35, 95% CI 1.74 to 16.5; Figure 2 ). The most important predictors for cumulative midterm mortality were increased pro-BNP levels (OR 11, 95% CI 1.51 to 81), AKI stage 3 (OR 6.80, 95% CI 2.55 to 15.66), LVEF <30% (OR 6.67, 95% CI 3.5 to 12.76), and periprocedural AMI (OR 6.52, 95% CI 2.34 to 18.14; Figure 3 ).
