Abstract
Background
Transcatheter aortic valve implantation (TAVI) is a safe and effective alternative to surgical aortic valve replacement (SAVR) for the treatment of severe aortic valve stenosis (AS). The impact of concomitant baseline elevated pulmonary artery pressures on outcomes after TAVI has not been established, since different studies used different definitions of pulmonary hypertension (PH).
Objective
To determine the association of PH with early and late cardiac and all-cause mortality after TAVI.
Methods
We performed a meta-analysis of studies comparing patients with elevated pulmonary artery pressures (defined as pulmonary hypertension or not) versus patients without elevated pulmonary artery pressures undergoing TAVI. We first performed stratified analyses based on the different PH cut-off values utilized by the included studies and subsequently pooled the studies irrespective of their cut-off values. We used a random effects model for the meta-analysis and assessed heterogeneity with I-square. Separate meta-analyses were performed for studies reporting outcomes as hazards ratios (HRs) and relative risks (RRs). Subgroup analyses were performed for studies published before and after 2013. Meta-regression analysis in order to assess the effect of chronic obstructive pulmonary disease and mitral regurgitation were performed.
Results
In total 22 studies were included in this systematic review. Among studies presenting results as HR, PH was associated with increased late cardiac mortality (HR: 1.8. 95% CI: 1.3–2.3) and late all-cause mortality (HR: 1.56; 95% CI: 1.1–2). The PH cut-off value that was most likely to be associated with worst outcomes among the different endpoints was pulmonary artery systolic pressure of 60 mm Hg (HR: 1.8; 95% CI: 1.3–2.3; I 2 = 0, for late cardiac mortality and HR: 1.52; 95% CI: 1–2.1; I 2 = 85% for late all-cause mortality).
Conclusion
This systematic review and meta-analysis emphasizes the importance of baseline PH in predicting mortality outcomes after TAVI. Additional studies are needed to clarify the association between elevated baseline pulmonary artery pressures and outcomes after TAVI.
Highlights
- •
22 studies were included in our meta-analysis
- •
baseline PH was associated with 30-day cardiac mortality
- •
baseline PH association with 30-day all-cause mortality remains unclear
- •
baseline PH was associated with late cardiac mortality
- •
baseline PH was associated with late all-cause mortality.
1
Introduction
Transcatheter aortic valve implantation (TAVI) is an effective and safe alternative to surgical aortic valve replacement (SAVR) for patients with severe aortic stenosis (AS) with comorbidities and a high or intermediate surgical risk [ ]. PH prevalence ranges from 20% to 71% across different TAVI series, indicating the existing heterogeneity in PH definition and patient selection criteria [ ]. Pulmonary hypertension (PH) has been reported by some studies to be an independent risk factor for cardiovascular and overall mortality after TAVI, in addition to its established association with worse outcomes after surgical aortic valve replacement (SAVR) [ ]. Moreover, given the higher perioperative risk of PH patients, the benefits of valve replacement in these patients merit further investigation. In addition, there is an absence of a clear PH cut-off point that could predict which patients should not undergo TAVI or have a higher risk for post-TAVI mortality and morbidity. For example, PH (defined as pulmonary artery systolic pressure (PASP) > 60 mm Hg) is included in the EuroSCORE criteria but not in the Society of Thoracic Surgeons score [ ]. Due to the heterogeneity in PH definitions used by different studies (elevated PASP or elevated mean pulmonary artery pressure), there is not a clear consensus on the values that affect the mortality after TAVI. Thus, our aim in this study is to systematically review the literature and qualitatively and quantitatively synthesize the literature on the association of baseline PH and early and late cardiovascular and overall mortality.
2
Methods
This review protocol has been registered in the PROSPERO International Prospective Register of systematic reviews ( https://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42017059929 ). This study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [ ]. Medline, Scopus and Cochrane databases were reviewed for prospective or retrospective studies published until April 10, 2017 reporting comparative outcomes for early and late cardiovascular or overall mortality after TAVI for patients with baseline PH vs. patients without PH. More details on the search algorithm can be found on the PROSPERO protocol.
2.1
Study selection and data extraction
Two reviewers (DGK, CAP) independently assessed the eligibility of the included studies. A third reviewer (EKO) was involved to reach a consensus when necessary. In order for a study to be eligible for this review it had to fulfill the following inclusion criteria: i) original research studies published in any language and conducted in patients with severe aortic stenosis who undergo TAVI; ii) studies presenting comparative mortality (cardiac or overall) outcomes between patients with PH and without PH. When we encountered studies that described the same or overlapping series of patients with the same outcomes and endpoints, only one of the studies was included for every endpoint. However, when two studies had included common patients/were conducted in the same centers but reported outcomes in a different way (PH definitions, endpoints, hazard ratios or relative risks/odds ratios), they were both included in our systematic review but were never pooled together in order to avoid counting patients twice in the meta-analysis (See Fig. 1 ). Two reviewers (DGK, CAT) independently extracted the data and discrepancies were resolved by the involvement of a third reviewer (EKO).
2.2
Outcomes
This study had 4 different outcomes: i) 30-day cardiac mortality; ii) 30-day all-cause mortality; iii) late (>6-month) cardiac mortality; iv) late (>6-month) all-cause mortality. Cardiovascular mortality instead of cardiac mortality was used when studies reported only cardiovascular mortality.
2.3
Data analysis
The primary analysis for all the outcomes was performed with HRs. Additional analysis was performed among studies presenting outcomes as RRs and can be found in the Appendix ( Supplementary Figs. 1, 2, 3 and Supplementary Table B ). The authors of the original studies were contacted for additional data as needed. When the original studies reported only relative frequencies of the events, the absolute events numbers in each arm were calculated. The included studies separately for ORs and HRs were combined according to their cut-off values for PH definition initially and were next pooled independently of their PH definition. For studies that reported comparisons for two different PH potential definitions (i.e., PASP = 40 mm Hg and PASP = 60 mm Hg), we used the highest value for the pooled meta-analysis. Studies with potentially duplicated populations were not included in the pooled analysis, so that every population/center would be counted once. A random effects model was selected a priori because the included studies had heterogeneous study design and baseline patients’ characteristics [ ]. A forest plot was used to illustrate the individual study findings and the random effects meta-analysis results. The I-square statistic (I 2 ) was used to assess for heterogeneity among the studies. Values <25% indicated low, 25% to 70% indicated moderate, and >70% indicated severe heterogeneity [ ]. Contoured enhanced funnel plots were used to visually assess for publication bias and the Egger’s test to quantitatively assess it [ ]. Meta-regression analysis was used in order to assess for the effect of comorbidities including chronic obstructive pulmonary disease (COPD) and mitral regurgitation (MR). Subgroup analyses were performed among studies published before and after 2013 to examine if the results would change overtime. Continuous variables are presented as mean values ± SD. Categorical variables are presented as absolute and relative frequencies. The estimated incidence rates were expressed as percent and 95% confidence interval (CI)·The results were regarded as statistically significant at a two-sided p < 0.05. We used STATA 14.1 (StataCorp, College Station, Texas) as the statistical software for all analyses.
2
Methods
This review protocol has been registered in the PROSPERO International Prospective Register of systematic reviews ( https://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42017059929 ). This study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [ ]. Medline, Scopus and Cochrane databases were reviewed for prospective or retrospective studies published until April 10, 2017 reporting comparative outcomes for early and late cardiovascular or overall mortality after TAVI for patients with baseline PH vs. patients without PH. More details on the search algorithm can be found on the PROSPERO protocol.
2.1
Study selection and data extraction
Two reviewers (DGK, CAP) independently assessed the eligibility of the included studies. A third reviewer (EKO) was involved to reach a consensus when necessary. In order for a study to be eligible for this review it had to fulfill the following inclusion criteria: i) original research studies published in any language and conducted in patients with severe aortic stenosis who undergo TAVI; ii) studies presenting comparative mortality (cardiac or overall) outcomes between patients with PH and without PH. When we encountered studies that described the same or overlapping series of patients with the same outcomes and endpoints, only one of the studies was included for every endpoint. However, when two studies had included common patients/were conducted in the same centers but reported outcomes in a different way (PH definitions, endpoints, hazard ratios or relative risks/odds ratios), they were both included in our systematic review but were never pooled together in order to avoid counting patients twice in the meta-analysis (See Fig. 1 ). Two reviewers (DGK, CAT) independently extracted the data and discrepancies were resolved by the involvement of a third reviewer (EKO).
2.2
Outcomes
This study had 4 different outcomes: i) 30-day cardiac mortality; ii) 30-day all-cause mortality; iii) late (>6-month) cardiac mortality; iv) late (>6-month) all-cause mortality. Cardiovascular mortality instead of cardiac mortality was used when studies reported only cardiovascular mortality.
2.3
Data analysis
The primary analysis for all the outcomes was performed with HRs. Additional analysis was performed among studies presenting outcomes as RRs and can be found in the Appendix ( Supplementary Figs. 1, 2, 3 and Supplementary Table B ). The authors of the original studies were contacted for additional data as needed. When the original studies reported only relative frequencies of the events, the absolute events numbers in each arm were calculated. The included studies separately for ORs and HRs were combined according to their cut-off values for PH definition initially and were next pooled independently of their PH definition. For studies that reported comparisons for two different PH potential definitions (i.e., PASP = 40 mm Hg and PASP = 60 mm Hg), we used the highest value for the pooled meta-analysis. Studies with potentially duplicated populations were not included in the pooled analysis, so that every population/center would be counted once. A random effects model was selected a priori because the included studies had heterogeneous study design and baseline patients’ characteristics [ ]. A forest plot was used to illustrate the individual study findings and the random effects meta-analysis results. The I-square statistic (I 2 ) was used to assess for heterogeneity among the studies. Values <25% indicated low, 25% to 70% indicated moderate, and >70% indicated severe heterogeneity [ ]. Contoured enhanced funnel plots were used to visually assess for publication bias and the Egger’s test to quantitatively assess it [ ]. Meta-regression analysis was used in order to assess for the effect of comorbidities including chronic obstructive pulmonary disease (COPD) and mitral regurgitation (MR). Subgroup analyses were performed among studies published before and after 2013 to examine if the results would change overtime. Continuous variables are presented as mean values ± SD. Categorical variables are presented as absolute and relative frequencies. The estimated incidence rates were expressed as percent and 95% confidence interval (CI)·The results were regarded as statistically significant at a two-sided p < 0.05. We used STATA 14.1 (StataCorp, College Station, Texas) as the statistical software for all analyses.
3
Results
3.1
Study selection process and studies characteristics
In total 956 studies were screened and 28 full-text articles were reviewed for eligibility. A PRISMA flow diagram with the selection process is shown in the Data Supplement ( Supplementary Fig. 1 ). There were six studies that despite meeting the initial inclusion criteria were excluded from the analysis since they included potentially duplicated populations without reporting different cut-off values or different outcomes/endpoints. Finally, 22 studies were included in our meta-analysis [ , , ]. In total, 16 of the included studies were conducted in Europe. The mean age of patients in the included studies ranged from 78 ± 6 to 85 ± 7 years, while the percentage of female patients ranged from 34.1% to 65.2%. The total number of enrolled patients in the individual studies ranged from 122 to 3195. Fifteen out of 22 studies reported the diagnostic method used for PH. Transthoracic echocardiogram was used by 10 studies, right heart catheterization by three studies, and a combination of the two strategies by two studies. All the patients in this meta-analysis had severe baseline AS. The most commonly used valves were Edwards Sapien and Sapien XT, and the Medtronic CoreValve. Transfemoral, transapical and the subclavian approach were used in almost all of the included cases. Details on baseline clinical and procedural characteristics of the included studies are presented in Table 1 and in the Supplementary Table A . The results of the Egger’s test and visual inspection of the contour funnel plots did not show publication bias in the included studies. A quantitative synthesis was performed for four different clinical outcomes (30-day cardiac mortality, 30-day all-cause mortality, late cardiac mortality, late all-cause mortality) with two different methods (HRs and ORs) for each outcome. Across the different cut-offs for all the outcomes, MPAP = 25 and PASP = 60 were more strongly associated with worse survival ( Table 2 ). No difference was found in the subgroup analysis between studies published before and after 2013 (P for interaction was 0.54 for late cardiac mortality and 0.38 for late all-cause mortality).
3.2
30-day cardiac mortality
Cardiac mortality after 30-days as an outcome was used in total by six studies (one with HR and five with RRs). In the pooled analysis, PH independently of the cut-off value used was associated with 30-day cardiac mortality (pooled RR [pRR]: 1.41; 95% CI: 1.04–1.92, without important heterogeneity, I 2 = 28%). The pRRs for different PH cut-offs used by the studies are presented in Supplementary Table B . Only one study used HR for 30-day cardiac mortality, thus we did not proceed with a HR meta-analysis for this outcome [ ].
3.3
30-day all-cause mortality
All-cause mortality after 30-days as an outcome was reported in total by 16 studies (three reported HRs only, 11 RRs only and two both RRs and HRs). The pooled HR among five studies was 0.95; 95% CI: 0.33–1.56, with significant heterogeneity (I 2 = 79.2%, Fig. 2 ) while the pooled RR among 13 studies was 1.77 (1.39–2.24), with moderate heterogeneity (I 2 = 47.8%, Supplementary Fig. 1 ). Table 2 presents the subgroup analyses for the different PH cut-offs. PASP > 60 was not associated with 30-day all-cause mortality (HR: 1.45; 95% CI: 0.4–2.5; I 2 = 40%). There was no association between COPD (P = 0.82) or MR (P = 0.53) and 30-day all-cause mortality.
3.4
Late cardiac mortality
Late cardiac mortality as an outcome was used in total by six studies. The pooled HR among the three studies was 1.3; 95% CI: 1.3–2.3; I 2 = 0 ( Fig. 3 ) while the pooled RR was 1.48; 95% CI: 1.25–1.76; I 2 = 20.3 ( Supplementary Fig. 2 ). PASP > 60 was associated with increased later cardiac mortality (HR: 1.8; 95% CI: 1.3–2.3; I 2 = 0). Thus baseline PH was associated with increased late cardiac mortality in both of our pooled analyses. There was no association between COPD (P = 0.61) or MR (P = 0.82) and late cardiac mortality.
3.5
Late all-cause mortality
Late all-cause mortality as an outcome was used in total by 15 studies. Baseline PH was associated with late all-cause mortality in both the HR meta-analysis (12 studies, HR: 1.56; 95% CI: 1.13–1.98; I 2 = 79.7, Fig. 4 ) and in the meta-analysis of RRs (12 studies, RR: 1.66; 95% CI: 1.35–2.03; I 2 = 79.8, Supplementary Fig. 3 ). PASP > 60 was associated with an increased risk of late all-cause mortality (HR: 1.52; 95% CI: 1–2.1; I 2 = 0). MPAP > 25 was associated with increased late all-cause mortality (HR: 1.47; 95% CI: 1.2–1.8; I 2 = 0). There was no association between baseline COPD (P = 0.39) and MR (P = 0.62) and the final outcome.
| Author, year | Country | Total N (PH vs. no PH) | PH cut-off values | PH Dx | Valves | Access route | COPD | Severe MR | STS score | Logistic euroSCORE | Exclusion criteria |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Lindman 2015 | USA | 2180 (1395 vs. 785) | MPAP = 25, MPAP = 35 | RHC | SAPIEN | TF (59%), TA (41%) | 46.2% | 21.8% | 11.1% (4.2%)/11.7% (3.8%) | NA | NA |
| Lucon 2014 | France | 2435 (1590 vs. 845) | SPAP = 40, SPAP = 60 | TTE | SAPIEN (67.3%), CoreValve (32.5%) | TF, TA, TSC | 22.9% | 2.1% | 9.8% (6.0–19.6)/10.1% (5.2–20) | 15.6% (10.0–23.9)/28.3% (20.3–40.7) | NA |
| O’Sullivan 2015 | Switzerland | 433 (325 vs. 108) | MPAP = 25 | RHC | CoreValve (54.1%), SAPIEN (44.1%), Symetis (1.9%) | TF (80.6%), TA (18%), TSC (1.4%) | 17.6% | NA | 4.7% (3.5–7.9)/6.8% (4.3–9.8) | 15.4% (10.0–25)/29% (17.3–43.9) | No RHC, no MPAP recorded, AI, valve-in-valve procedures, > 9 m from RHC to TAVI |
| Testa 2016 | Italy | 990 (644 vs. 346) | SPAP = 40, SPAP = 60 | TTE | NA | TF (90.4%), TSC (9.6%) | 23.3% | 2.6% | 7.7% (6.1–11.1)/8.0% (6.4–11.5) | 12.8% (4%)/26.2% (4%) | Failed aortic bioprosthesis, prior mitral valve surgery |
| Rodés-Cabau 2010 | Canada | 339 (84 vs. 255) | SPAP = 60 | RHC/TTE | SAPIEN | TF (49.6%), TA (50.7%) | 29.5% | 8% | 9.8% (6.4%) | NA | NA |
| Barbash 2015 | USA | 415 (243 vs. 172) | SPAP = 50 | TTE | SAPIEN (66.7%), SAPIEN XT (17.3%), CoreValve (13.7%) | TF (75.4%), TA (23.6%) | 29.7% | 11.6% | 10% (5%) | 30% (24%) | NA |
| Bishu 2014 | USA | 251 (171 vs. 80) | SPAP = 36, SPAP = 48 | TTE | NA | NA | NA | NA | NA | NA | Inadequate TVR signals for SPAP estimation |
| Souza 2016 | Brazil | 136 (33 vs. 103) | SPAP = 60 | NA | CoreValve (97%), SAPIEN (3%) | TF (94.9%), TSC (4.4%), TAo (0.8%) | 9.6% | 68.4% | 9.3% (4.8–22.3) | NA | NA |
| Schewel 2014 | Germany | 439 (231 vs. 208) | MPAP = 25 | RHC | CoreValve (58.3%), SAPIEN (41.7%) | TF (81.4%), TA (10.9%), TSC (7.2%), TAo (0.4%) | 15.7% | 51% | NA | 24% (17%) | NA |
| Medvedofsky 2015 | Israel | 122 (49 vs. 73) | SPAP = 50 | TTE | CoreValve (80.3%), SAPIEN (19.7%) | TF | NA | 30.5% | NA | PH: 33% (13%)/no PH: 26% (15%) | NA |
| Sinning 2014 | Germany, UK | 353 (251 vs. 102) | SPAP = 30, SPAP = 60 | TTE | CoreValve (89%), SAPIEN (11%) | TF (96%), TSC (4%) | 31.6% | NA | 8.7% (6.1%) | 26.6% (16.5%) | NA |
| Lindsay 2015 | UK | 279 (49 vs. 230) | SPAP = 60 | NA | CoreValve, SAPIEN | NA | NA | NA | NA | NA | NA |
| Zahn 2013 | Germany | 1318 (865 vs. 453) | SPAP = 60 | NA | CoreValve (81.6%), SAPIEN (17.9%) | TF (88%), TA (8.6%), TSC (2.7%), TAo (0.8%) | 28.1% | NA | NA | NA | NA |
| D’Ascenzo 2014 Derivation Cohort | Italy | 1069 (221 vs. 848) | SPAP = 50 | NA | SAPIEN (41.5%), CoreValve (41.1%) | TF (64.9%), TA (23.3%), TSC (11.7%) | 26.9% | NA | 9.4% (7.9%)/11.1% (10%) | 18.3% (14.1%)/19.3% (2%) | NA |
| D’Ascenzo 2014 Validation Cohort | 177 (35 vs. 142) | SAPIEN (52.5%), CoreValve (44.1%) | NA | NA | 8.4% (7.1%)/10.4% (5.4%) | NA | |||||
| D’Ascenzo 2015 | Italy, Netherlands | 674 (319 vs. 355) | SPAP = 40 | TTE | SAPIEN (35%), CoreValve (31.9%) | TF (71.5%), TA (16.9%), TSC (11.3%) | 32.3% | NA | PH: 8.7% (6.7%)/no PH: 9.4% (8.4%) | PH: 18.5% (19%), no PH: 17.2% (13.1%) | NA |
| Halkin 2016 | Israel | 1291 (142 vs. 1149) | SPAP = 60 | NA | CoreValve (67.3%), SAPIEN (30.1%) | TF (87.4%), TA (7.6%) | 18.8% | 15.5% | 5.3% (3.6%)/8.9% (7.3%) | 17.6% (11.9%)/28.8% (19%) | NA |
| Muñoz-García 2013 | Spain, Portugal, IberoAmerica | 1218 (229 vs. 989) | SPAP = 60 | TTE | CoreValve | TF (94.7%), TSC (5.3%) | 26.5% | 18.5% | NA | 17.8% (13%) | Missing data |
| Faruta 2015 | France | 3195 (295 vs. 2900) | SPAP = 60 | TTE | SAPIEN (66.4%), CoreValve (32.7%) | TF, TA, TSC | 24.7% | NA | NA | 21.8% (14.3%) | NA |
| Seiffert 2014 | Germany | 1052 (273 vs. 779) | NA | NA | SAPIEN, SAPIEN XT, CoreValve, ACURATE TA, JenaValve | NA | 28.8% | NA | 6.4% (4.2–10.5)/8.6% (5.3–12.7) | 20.4% (12.7–31.7)/25.8% (15.8–40.6) | NA |
| Koifman 2016 | USA | 648 (99 vs. 549) | SPAP = 60 | NA | CoreValve, SAPIEN, SAPIEN XT, Portico | TF (79.7%), TA (20.3%) | 32.6% | 9.9% | 8.7% (4.5%)/10.4% (4.6%) | NA | NA |
| Tamburino 2011 | Italy | 663 (76 vs. 587) | SPAP = 60 | TTE | CoreValve | TF (90.3%), TA (9.7%) | 21.3% | 6.3% | NA | 23% (13.7%) | NA |
| Nijenhuis 2016 | Netherlands | 227* (113 vs. 114) | MPAP = 25 | Cath/TTE | SAPIEN XT, CoreValve, DirectFlow, Lotus, JenaValve, Engager | TF, TA | 24.2% | 26.2% | 5.9% (3.8%)/6.8% (3.8%) | 26.2% (16.3%) | TAVI for AI |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
