Innovations: Technology & Techniques in Cardiothoracic & Vascular Surgery:
Stentless Versus Stented Bioprosthetic Aortic Valves: A Systematic Review and Meta-Analysis of Controlled Trials
Cheng, Davy FRCPC*; Pepper, John FRCS†; Martin, Janet PharmD*‡; Stanbridge, Rex FRCS§; Ferdinand, Francis D. MD¶; Jamieson, W R. Eric FRCSC∥; Stelzer, Paul MD**; Berg, Geoffrey FRCS††; Sani, Guido MD‡‡
From the *Department of Anesthesia and Perioperative Medicine, Evidence-Based Perioperative Clinical Outcomes Research Group (EPiCOR), London Health Sciences Centre, The University of Western Ontario, London, ON, Canada; †Department of Cardiothoracic Surgery, Imperial College, Royal Brompton Hospital, London, UK; ‡High Impact Technology Evaluation Centre, London Health Sciences Centre, London, ON, Canada; §Department of Cardiothoracic Surgery, St. Mary’s Hospital, London, UK; ¶Division of Thoracic and Cardiovascular Surgery, The Lankenau Hospital, Wynnewood, PA USA; ∥Division of Cardiovascular Surgery, St. Paul’s Hospital, University of British Columbia, Vancouver, Canada; **Department of Cardiothoracic Surgery, Mount Sinai Medical Center/Mount Sinai School of Medicine, NY USA; and ††Golden Jubilee National Hospital, Clydebank, UK; and ‡‡Department of Surgery, Siena University School of Medicine, Siena, Italy.
Accepted for publication February 15, 2009.
Supported by The International Society for Minimally Invasive Cardiothoracic Surgery (ISMICS), which has received unrestricted educational grants from industries that produce surgical technologies.
Address correspondence and reprint requests to Davy C. H. Cheng, FRCPC, London Health Sciences Centre-University Hospital, 339 Windermere Road, C3-172, London, ON, Canada N6A 5A5. E-mail: email@example.com.
Disclosure: Geoffrey A. Berg, FRCS, Consultant for Vascutek Ltd.
Francis D. Ferdinand, MD, Investigator for Sorin Group, St. Jude Medical, Edwards Lifesciences.
WR Eric Jamieson, FRCSC, Consultant for Sorin, St. Jude Medical, Medtronic; Speakers Bureau/Honoraria for Edwards Lifesciences, Sorin Group, On-X Life Technologies, St. Jude Medical; Principal Investigator for St. Jude Medical, Edwards Lifesciences, On-X Life Technologies, Medicure Inc, Rib-X Pharmaceuticals, Sorin Group, Medtronic, CIHR; Randomized Controlled Trials for On-X, St. Jude Medical.
Objective: This meta-analysis sought to determine whether stentless bioprosthetic valves improve clinical and resource outcomes compared with stented valves in patients undergoing aortic valve replacement.
Methods: A comprehensive search was undertaken to identify all randomized and nonrandomized controlled trials comparing stentless to stented bioprosthetic valves in patients undergoing aortic valve replacement available up to March 2008. The primary outcomes were clinical and resource outcomes in randomized controlled trial (RCT). Secondary outcomes clinical and resource outcomes in nonrandomized controlled trial (non-RCT). Odds ratios (OR), weighted mean differences (WMD), or standardized mean differences and their 95% confidence intervals (CI) were analyzed as appropriate.
Results: Seventeen RCTs published in 23 articles involving 1317 patients, and 14 non-RCTs published in 18 articles involving 2485 patients were included in the meta-analysis. For the primary analysis of randomized trials, mortality for stentless versus stented valve groups did not differ at 30 days (OR 1.36, 95% CI 0.68–2.72), 1 year (OR 1.01, 95% CI 0.55–1.85), or 2 to 10 years follow-up (OR 0.82, 95% CI 0.50–1.33). Aggregate event rates for all-cause mortality at 30 days were 3.7% versus 2.9%, at 1 year were 5.5% versus 5.9% and at 2 to 10 years were 17% versus 19% for stentless versus stented valve groups, respectively. Stroke or neurologic complications did not differ between stentless (3.6%) and stented (4.0%) valve groups. Risk of prosthesis-patient mismatch was numerically lower in the stentless group (11.0% vs. 31.3%, OR 0.30, 95% CI 0.05–1.66), but this parameter was reported in few trials and did not reach statistical significance. Effective orifice area index was significantly greater for stentless aortic valve compared with stented valves at 30 days (WMD 0.12 cm2/m2), at 2 to 6 months (WMD 0.15 cm2/m2), and at 1 year (WMD 0.26 cm2/m2). Mean gradient at 1 month was significantly lower in the stentless valve group (WMD −6 mm Hg), at 2 to 6 month follow-up (WMD −4 mm Hg,), at 1 year follow-up (WMD −3 mm Hg) and up to 3 year follow-up (WMD −3 mm Hg) compared with the stented valve group. Although the left ventricular mass index was generally lower in the stentless group versus the stented valve group, the aggregate estimates of mean difference did not reach significance during any time period of follow-up (1 month, 2–6 months, 1 year, and 8 years).
Conclusions: Evidence from randomized trials shows that subcoronary stentless aortic valves improve hemodynamic parameters of effective orifice area index, mean gradient, and peak gradient over the short and long term. These improvements have not led to proven impact on patient morbidity, mortality, and resource-related outcomes; however, few trials reported on clinical outcomes beyond 1 year and definitive conclusions are not possible until sufficient evidence addresses longer-term effects.
Aortic stenosis remains the most prevalent acquired heart valve pathology and is typically associated with calcification and degenerative changes of the native aortic valve annulus and leaflets. The resulting increased pressure load leads to progressive left ventricular hypertrophy, and the presence of left ventricular hypertrophy in patients with aortic stenosis is independently associated with higher risk of adverse cardiovascular events and death.1 The ideal aortic valve substitute should be simple to implant, provide a hemodynamic profile identical to a normal native valve with unlimited durability, and have a low thrombogenic potential to preempt the need for ongoing anticoagulants. Unfortunately, no such device yet exists, and currently available valve substitutes are typically xenograft valves with an orifice area that offers at least some degree of obstruction which may adversely impact on long-term valve performance and patient morbidity.
Stentless biologic aortic valves with increased effective orifice area have been developed in response to the need to overcome the obstructive limitations associated with stented biologic aortic valves. Cohort studies of patients implanted with stentless aortic valves have shown significant regression of ventricular hypertrophy; however, randomized trials have shown inconsistent results. Because a number of these trials have been underpowered individually to adequately rule out the presence of significant differences in clinical outcomes and valve performance over time, we sought to perform an updated systematic review and meta-analysis of all existing comparative trials to summarize the best available evidence and maximize the power to measure potential differences in clinically relevant outcomes. Previous meta-analyses in this area have not included the most recent trials and evaluated only selected outcomes.2–5
The purpose of this meta-analysis was to determine whether stentless bioprosthetic valves improve clinical and resource outcomes compared with stented valves in patients undergoing aortic valve replacement. This systematic review and meta-analysis was planned in accordance with current state-of-the art guidelines for performing meta-analyses, including the QUOROM guidelines and the MOOSE guidelines.6,7
We searched the Cochrane Central Register of Controlled Trials on The Cochrane Library, MEDLINE, EMBASE from 1990 to March 2008, as well as selected surgical meeting abstracts (American Association of Thoracic Surgery, Society of Thoracic Surgery, and the European Association of Cardiothoracic Surgery). Search terms included combinations and derivatives of the following as textwords and MESH terms: [(valv* AND aort*) AND surg* OR operat* OR repair* OR replac*]. After the initial searches were completed and potentially relevant studies identified, additional tangential searches were conducted using “related articles” links within MEDLINE. In addition, individual searches for specific names of stentless valves (Table 1) were performed to identify further studies. Reference lists of included studies and recent overviews were hand searched for additional studies. Selected authors of clinical studies and experts were contacted to identify unpublished studies of stentless versus stented aortic valve repair.
Retrieval of Studies
All trials that seemed relevant on the basis of title and abstract were selected for full review. From the potentially relevant articles, two reviewers independently selected trials (based on the full text format) for inclusion in this review. Disagreement was resolved by consensus with third party adjudication. Independent reviewers documented the content of each included study.
For the primary analyses of clinical and resource outcomes, all randomized controlled trials (RCTs) that compared stentless valves versus stented valves for aortic valve replacement were eligible for inclusion if they reported at least one relevant clinical or resource-related outcome. For the secondary analyses of clinical and resource outcomes, nonrandomized trials comparing stentless versus stented aortic bioprosthetic valves were eligible for inclusions if they reported at least one relevant outcome. Trials comparing only stentless versus mechanical heart valves, homograft aortic root replacement, and pulmonary autograft (Ross procedure) were excluded. When trials had multiple arms (ie, stentless vs. stented vs. mechanical vs. homograft aortic valves), only the stentless valve versus stented valve arms were included in the meta-analysis. Studies in any language were accepted, whether published or unpublished.
To address multiple publications of overlapping patient populations, we classified all studies by the center and the years of patient enrollment, and selected the most recent series from each center to extract as many relevant outcomes as possible. When the more recent series failed to report all outcomes of interest, we consecutively consulted the next most recent series from the same center, and extracted the remaining outcomes as far as possible.
Data Collection and Outcomes
Two reviewers independently assessed studies for inclusion eligibility criteria, and data were extracted independently from included studies by two or more reviewers. Discrepancies were resolved by consensus with a third reviewer. Data were extracted using standard forms and included patient demographics such as age, sex, concomitant procedures, baseline cardiac function, duration of follow-up, and loss to follow-up. Information was also extracted regarding study authors, study centers, cities, years of patient enrollment, whether patients were consecutively enrolled, whether loss to follow-up was reported, and whether the study was randomized or nonrandomized (prospective or retrospective). Details of the procedure extracted included type of valve used and duration of procedure.
All clinically relevant outcomes and resource-related outcomes were considered at all timepoints. Outcomes with discrete events included all-cause mortality, prosthesis patient mismatch, valve regurgitation, valvular dysfunction, reoperation for valve complications, stroke or neurologic complications, thromboembolic events, atrial fibrillation, myocardial infarction, atrioventricular block, permanent pacemaker insertion, intra-aortic balloon pump insertion, heart failure, renal dysfunction, ventricular arrhythmias, wound infection, bleeding complications, reoperation for bleeding, reoperation for any cause, early endocarditis (within 2 months), late endocarditis (beyond 2 months), patients asymptomatic or achieving New York Heart Association (NYHA) class I or II status during follow-up, and patients experiencing any complication (when reported by the study authors as a composite of any patient experiencing at least one complication). Other outcomes with continuous measures included effective orifice area index (EOAI), left ventricular mass index (LVMI), mean gradient, peak gradient, ejection fraction (EF), cross clamp time, operating room time, ventilation time, intensive care unit length of stay (days postprocedurally), and hospital length of stay (days postprocedurally). Definitions for clinical events according to recent guidelines were preferred.8
Data were combined for meta-analysis using Comprehensive Meta-Analysis Software version 2.0. For dichotomous variables, individual and pooled statistics were calculated as odds ratios (OR) with 95% confidence intervals (95% CI). For continuous outcomes, individual and pooled statistics were calculated as weighted mean differences (WMD) with 95% CI.
Heterogeneity across trials was explored for each outcome by calculating I2, which indicates the percent of heterogeneity across trials which cannot be explained by chance variation alone. I2 <50% was considered to indicate low heterogeneity, and I2 = 50% to 75% was considered to indicate moderate heterogeneity and >75% indicated high heterogeneity. When heterogeneity was detected, we aimed to find potential explanations for the variation across clinical trials and reported these in the discussion. Potential reasons for heterogeneity hypothesized a priori included: differences in study quality (randomized vs. nonrandomized, loss to follow-up or cross-over between arms postrandomization), study center and surgeon experience, concomitant procedures [ie, coronary artery bypass graft (CABG)], comorbidities and baseline characteristics, valve types, and differences in outcome measure definitions or duration of follow-up.
The random effects model was used for all calculations to provide a conservative analysis because heterogeneity was anticipated across the trials. For outcomes without significant heterogeneity, the random effects analysis produces the same result as a fixed effect analysis. For outcomes with significant heterogeneity the random effects analysis generally provides a more conservative estimate as the CI are generally wider using the random effects approach.
For the primary analysis, only randomized trials were included in the analysis and reported separately for all relevant outcomes reported. For the secondary analysis, nonrandomized trials were analyzed for each outcome and presented separately for comparison with randomized trial results. The aggregate results of randomized data were compared with the aggregate results of nonrandomized data using the test for interaction to determine whether effect sizes differed significantly across the study design subgroups. The data were not combined across randomized and nonrandomized datasets. Subgroup analysis by type of valve used was also planned. Intention-to-treat analysis was used whenever possible, as allowed by the data provided in the included trials.
Figure 1 outlines the flowchart for included articles. The search strategy identified 793 articles. Hand-searching reference lists and searching conference databases identified four further articles. Contacting experts identified one further article. After reviewing the abstracts of the above articles, 614 were eliminated as they were clearly not relevant to the topic of stentless aortic valve replacement. A total of 184 potentially relevant full text articles were retrieved for closer review. Of these, the majority were excluded because they were reviews only, or because they were noncomparative studies.
Seventeen randomized studies (RCT) published in 23 articles9–31 involving 1317 patients were included in the meta-analysis. Table 2 outlines the included trial characteristics.
Fourteen nonrandomized trials (non-RCT) published in 18 articles32–49 involving 2485 patients were included in the meta-analysis. One registry was identified, but not included in the analysis because it was unknown whether the patients in the registry would overlap with the published studies.50 All identified studies were published studies, available in English.
Baseline Patient Characteristics
Table 3 outlines the patient characteristics at baseline. In the randomized trials, baseline mean age was 72 versus 73 years for the stentless and stented valve groups, respectively, and this 1 year difference was significant statistically (WMD −1 year, 95% CI −2 to −0.2 years, P = 0.01). Also, in nonrandomized trials the baseline mean age was significantly lower in the stentless valve group (68 years) versus the stented valve group (71 years; WMD −3 years, 95% CI −4 to −0.4 years; P = 0.02). Whether these differences in age translate to clinically important differences in outcomes is unknown because insufficient data were provided in the trials to examine the relationship between age differences and effect sizes.
Similar proportions of each group were male in randomized (48% vs. 50%) and nonrandomized trials (54% vs. 51%) for stentless versus stented groups, respectively. Similar proportions of patients had undergone concomitant CABG in randomized trials, although in nonrandomized trials fewer patients in the stentless valve group underwent concomitant CABG compared with the stented valve group (38% vs. 53%, Table 3). Advanced NYHA class (NYHA class III or IV) at baseline was similar between stented and stentless valve groups in randomized trials (63% vs. 60%) and nonrandomized groups (39% vs. 35%), but was notably more prevalent in both groups entering randomized trials than nonrandomized trials.
Annulus size was slightly higher in the stentless valve group compared with the stented valve group for randomized studies (WMD 0.5 mm, 95% CI 0.2–0.8, P = 0.002) and for nonrandomized studies (WMD 0.8 mm, 95% CI 0.8, 95% CI −0.3 to 1.9 mm, P = 0.17). Valve size was larger for the stentless group compared with stented group in randomized studies (WMD 0.15 mm, 95% CI 0.9–2.0 mm, P < 0.0001), and for nonrandomized studies (WMD 1.2 mm, 95% CI −1.2 to +3.5 mm, P = 0.33), although the latter did not reach significance (only two nonrandomized trials provided this data). Baseline EF, mean gradient, peak gradient, EOAI, and LVMI were numerically and statistically non significant between stentless and stented groups for randomized trials and nonrandomized trials. All randomized trials inserted the stentless bioprosthetic valves in the subcoronary configuration.
Clinical Outcomes for RCT
Clinical outcomes and associated heterogeneity between trials (I2) are summarized in Table 4 for randomized and nonrandomized trials. Subanalysis by type of valve was not possible due to the small number of trials reporting each type of valve. High heterogeneity was found for a number of continuous measures, such as EOAI, mean and peak gradients, LVMI, and EF.
Mortality for stentless versus stented valve groups did not differ at any time point, including at 30 days (OR 1.36, 95% CI 0.68–2.72), 1 year (OR 1.01, 95% CI 0.55–1.85), and at 2 to 10 years follow-up (OR 0.82, 95% CI 0.50–1.33; Fig. 2A–C) Aggregate event rates for all-cause mortality at 30 days were 3.7% versus 2.9%, at 1 year were 5.5% versus 5.9%, and at 2 to 10 years were 17% versus 19% for stentless versus stented valve groups, respectively.
Valve complications were generally poorly reported in the randomized trials. Valve regurgitation occurred in 4.4% and 6.4% of patients (OR 0.36, 95% CI 0.11–1.17), valvular dysfunction occurred in 1.4% and 2.6% of patients (OR 0.55, 95% CI 0.13–2.35), and reoperation for valvular complications occurred in 0.7% versus 1.9% of patients (OR 0.47, 95% CI 0.11–1.95) were not significantly different between stentless and stented valve groups, respectively.
Stroke, Thromboembolic Events
Stroke or neurologic complications did not differ between stentless (3.6%) and stented (4.0%) valve groups (OR 0.73, 95% CI 0.37–1.44). The risk of thromboembolic events was reported in only one randomized trial (1.6% stentless vs. 7.3% stented), and it suggested possible reduction in risk (OR 0.20, 95% CI 0.04–1.00).
No difference was found for overall composite estimates of complications for stentless versus stented groups (13.0% vs. 13.8%; OR 0.97, 95% CI 0.52–1.83). Early endocarditis did not differ significantly (0.8% vs. 2.2%; OR 0.52, 95% CI 0.07–4.06) and late endocarditis did not differ significantly (0.8% vs. 1.2%; OR 0.85, 95% CI 0.21–3.41) between stentless and stented groups, but was generally poorly reported in randomized trials. Similarly, no difference was found for risk of atrial fibrillation, myocardial infarction, atrioventricular block, permanent pacemaker insertion, intra-aortic balloon pump, heart failure, renal dysfunction, ventricular arrhythmia, wound infection, bleeding complications, reoperation for bleeding, and reoperation for any reason (Table 4).
Prosthesis Patient Mismatch and EOAI
Prosthesis-patient mismatch was numerically lower in the stentless group (11.0% vs. 31.3%, OR 0.30, 95% CI 0.05–1.66), but this parameter was reported in only four randomized trials and did not reach statistical significance. EOAI was significantly greater for patients receiving stentless aortic valve compared with stented valves at 30 days (WMD 0.12 cm2/m2, 95% CI 0.03–0.21 cm2/m2), at 2 to 6 months (WMD 0.15 cm2/m2, 95% CI 0.02–0.28 cm2/m2), and at 1 year (WMD 0.26 cm2/m2, 95% CI 0.10–0.41 cm2/m2). Randomized studies did not provide EOAI for follow-up beyond 1 year.
Mean and Peak Gradient
Mean gradient at 1 month was significantly lower by over 6 mm Hg in the stentless valve group compared with the stented valve group (WMD −6 mm Hg, 95% CI −10 to −2 mm Hg). The mean gradient remained significantly lower at 2 to 6 month follow-up (WMD −4 mm Hg, 95% CI −7 to −1 mm Hg), at 1 year follow-up (WMD −3 mm Hg, 95% CI −6 to −1 mm Hg), and up to 3 year follow-up (WMD −3 mm Hg, 95% CI −3 to −2 mm Hg; Fig. 2F). Similarly, peak gradient was lower at 1 month by 8 mm Hg (WMD −8 mm Hg, 95% CI −14 to −3 mm Hg), and the significant difference was maintained at 2 to 6 month follow-up (WMD −8 mm Hg, 95% CI −13 to −3 mm Hg), 1 year follow-up (WMD −8 mm Hg, 95% CI −14 to −3 mm Hg), and at 3 to 8 years’ follow-up (WMD −10 mm Hg, 95% CI −16 to −5 mm Hg).
LVMI and EF
Although the LVMI was generally lower in the stentless group versus the stented valve group, the aggregate estimates of mean difference did not reach significance during any time period of follow-up (1 month, 2 to 6 months, 1 year, and 8 years) as shown in Table 4. However, there was significant heterogeneity among the trials for this outcome with some studies suggesting large and significant reductions while others suggested no reduction in LVMI. Percent EF was not significantly different between stentless and stented groups at any time point (1 month, 6 months, 1 year, and 3 years). The proportion of patients achieving NYHA class I or II at follow-up did not show a significant difference between stentless and stented valve groups (Table 4).
Resource- and Procedure-Related Outcomes
Cross-clamp time was significantly greater by 23 minutes (WMD 23 minutes, 95% CI 18–27 minutes) and bypass time was significantly greater by 24 minutes (WMD 24, 95% CI 19–30 minutes) in the stentless group compared with the stented valve group. Time in the operating room was not significantly different, but was reported in only one randomized trial. Other resource-related outcomes, including time on the ventilator (WMD −11 hours, 95% CI −42 to +22 hours), intensive care unit length of stay (WMD −0.1 day, 95% CI −1 to +1 day), and total hospital length of stay (WMD −0.2 day, 95% CI −2 to +2 days) did not reach statistical differences between stentless and stented groups.
Clinical Outcomes for Non-RCT
Outcome estimates did not differ significantly between randomized and nonrandomized trials for many outcomes including death at most time points and most clinical complications including overall complications, stroke, atrial fibrillation, atrioventricular block, permanent pacemaker insertion, intra-aortic balloon pump, heart failure, renal dysfunction, ventricular arrhythmia, wound infection, bleeding complications, reoperation for any reason, early endocarditis, and late endocarditis.
There were a number of exceptions, where nonrandomized trials suggested notably different effect sizes than randomized trials (as indicated by a significant test of interaction between randomized and nonrandomized data). However, most of the differences noted between nonrandomized and randomized data were highly susceptible to bias because few trials had reported on the outcomes and definitive conclusions were not possible. For example, the estimate for myocardial infarction from nonrandomized trials suggested lower rates of myocardial infarction overall and in the opposite direction as with the randomized trials (P value for interaction = 0.06); however, this was based on a single nonrandomized trial and only three randomized trials and may not reflect a balanced view of effect size. A number of continuous outcome data differed significantly between randomized and nonrandomized trials (ie, mean gradient, peak gradient, EOAI) and in a number of these cases the heterogeneity occurred because of the differences in magnitude of reduction over time (as opposed to differences in the direction of effect over time). This heterogeneity between randomized and nonrandomized trials is also highly susceptible to bias, given that these outcomes were rarely reported in nonrandomized trials (ie, 1 or 2 only), and the more complete reporting from randomized trials should be relied on preferentially given this information. In addition, differences in loss to follow-up in the nonrandomized trials may have resulted in heterogeneous results.
Importantly, the outcome of death at 30 days was found to be heterogeneous between randomized and nonrandomized trials. The non-RCT studies tended to report lower incidences of 30-day mortality in the stentless versus stented groups (4.0% vs. 6.8%, respectively; OR 0.60, 95% CI 0.39–0.91, P = 0.02). This result was discrepant with randomized trials which showed no significant difference 30-day mortality between groups, with a direction of effect in the opposite direction than for nonrandomized trials (3.7% vs. 2.9% for stentless and stented valve groups, respectively; OR 1.36, 95% CI 0.68–2.72). The P value for interaction between the randomized and nonrandomized trials was 0.05, suggesting that they are not compatible with each other. Given that some of the nonrandomized trials were retrospective, with potential for loss to follow-up in both groups, the prospective randomized trials with adequate follow-up should be considered to provide the most representative estimates of relative effects.
Most differences between randomized and nonrandomized data were due to differences in magnitude of effect size rather than direction of effect size (cross clamp time, bypass time, EOAI, mean gradient, and hospital length of stay). In no case did the results of nonrandomized data provide compelling data that improved on the estimates provided by randomized trials.
To date, RCT do not show significant reductions in all-cause mortality during early or late follow-up for aortic valve replacement with stentless valves when compared with stented valves. Similarly, significant differences in stroke, myocardial infarction, arrhythmias, organ failure, and endocarditis, and risk of any complication were not found. Additionally, structural valve dysfunction, regurgitation, and reoperation for valvular complications were not measurably reduced with stentless versus stented aortic valves. It is noteworthy that for each of these outcomes there was insufficient power to rule out the possibility that a real and important difference might exist. Further RCT are required to definitively rule in or rule out whether differences exist. At this time, however, the totality of the evidence shows that no differences in clinical outcomes have been proven.
Although none of these important clinical outcomes were significantly impacted by the insertion of stentless instead of stented valves, there were a number of intermediate hemodynamic and cardiac parameters that were improved with stentless valves. The EOAI was increased at all time periods of follow-up beyond 30 days (Table 4). Mean and peak gradients were significantly reduced by 6 and 8 mm Hg, respectively, at 30 days, and this difference was maintained at maximal follow-up (3–10 years). Nevertheless, despite these reductions in effective orifice area and peak gradients, LVMI reductions did not reach significance for stentless valves versus stented valves in the randomized trials. Longer term data from nonrandomized trials suggests there is a potential for significant reduction in LVMI in survivors. However, there was heterogeneity between the single randomized trial reporting data up to 8 years (WMD −2, 95% CI −17 to +13 mm Hg) when compared with the aggregate estimate from three nonrandomized trials reporting this information at 2 to 10 years follow-up (WMD −19 mm Hg, 95% CI −35 to −2 mm Hg). This prevents definitive conclusions about whether the difference in the nonrandomized trials is due to real effects attributable to stentless valves or whether other biases played a role in the differences measured. Nonrandomized trials are particularly susceptible to selection bias (ie, differences in prognostically significant baseline characteristics due to differences in clinicians’ decisions to implant stentless valves in patients with different risk characteristics than for stented valves) and differential loss to follow-up between groups (ie, only survivors who are healthy enough to attend clinics will be followed up for hemodynamic measures over time, and retrospective trials are particularly susceptible to bias due to lost follow-up over time).
Unfortunately, the clinical relevance of reductions in EOAI, LVMI, and EF over time is unknown. Whether small or large reductions in any one or a combination of these parameters is sufficient to improve clinically important outcomes such as reduced need for reoperation, reduced adverse cardiac events, improved survival, and quality of life remains unknown. Insufficient data were provided in the trials regarding NYHA functional status, which would provide a more clinically relevant estimate of cardiac functionality than the intermediate outcomes of EOAI, LVMI, and EF. Quality of life was not measured in any of the trials.
For each clinical outcome reported, the CI were wide, suggesting that the available evidence to date provides insufficient power to rule out the possibility of clinically important differences, even when the data is aggregated through meta-analysis to maximize the power across existing studies. Much larger randomized trials will be necessary to determine whether true differences exist in the clinically relevant outcomes of death, stroke, myocardial infarction, endocarditis, patient function, and need for surgical reintervention.
Randomized Versus Nonrandomized Trials
It was encouraging to note that, for this meta-analysis which ascertained all comparative trials of stentless versus stented aortic valve replacement, there were more randomized trials that contributed data than nonrandomized trials. Examining the graphs for each of the outcomes reveals that the nonrandomized analysis provides little information beyond the randomized trials alone. This is encouraging, because it is more common for nonrandomized trials to outnumber randomized trials for many surgical interventions, particularly early in the development of the technology and learning curve associated with newer interventions, and surgeon investigators are to be congratulated for having conducted randomized trials before wide-spread dissemination of this new technology. Randomized trials are generally preferred to nonrandomized trials to circumvent the shortcomings that are inherent in nonrandomized data, including biases which result in prognostic differences due to differential selection and follow-up between intervention and control groups in nonrandomized reports. Furthermore, many nonrandomized trials compare noncontemporaneous groups of patients (either retrospectively or prospectively), with the risk of outcome differences being driven by variable historical practices and loss to follow-up.
Although it was planned a priori in the protocol for this meta-analysis to report both randomized and nonrandomized data, it became apparent after the analysis was complete that the nonrandomized studies provided little strength to the analysis.
Although randomized trials also had important shortcomings related to generalizability (patients were excluded due to difficult morphology, and aortic root configuration not amenable to stentless valves) and lack of power to adequately rule out the possibility of important difference between stentless and stented valves. Each included study was relatively small, and even after maximizing power through meta-analysis, the total amount of information remained insufficient to definitively rule out the possibility of clinically significant differences. The finding of no significant differences does not prove equivalence because there was insufficient power and subsequently wide CI for a number of the clinically important outcomes.
Strengths and Limitations
This comprehensive meta-analysis attempts to assess all relevant clinical outcomes in addition to intermediate hemodynamic and cardiac function outcomes between subcoronary stentless aortic valves and stented aortic valves. The analysis followed methodologically rigorous guidelines as recommended by QUOROM6 for creating a protocol for the analysis, ascertaining all relevant comparative trials, and reporting all clinically relevant outcomes. This analysis provides the best available estimate of the impact of stentless versus stented aortic valve replacement to date, and also highlights the limitations of the existing evidence which should be addressed by further randomized trials with sufficient power to rule in or rule out whether important differences exist between survival and patient function.
A number of limitations exist in this analysis, as a direct result of the shortcomings in the existing trials which provided the data for this analysis. The most important limitation is that very few trials reported late outcomes at greater than 1 year. As a result, no definitive conclusion regarding the impact (or lack thereof) of stentless valves on late morbidity and mortality can be given. If stentless valves improve hemodynamics, these differences may not translate into measurable differences in clinically relevant morbidities or survival until longer-term follow-up. The overall impact of stentless versus stented valves on survival differences, quality of life, valve durability, NYHA class, and patient functional differences over time remains unknown because studies have not addressed these outcomes over sufficient duration to allow potential differences to be assessed.
There are inherent limitations in patient generalizability, because patients who were thought to be unsuitable for stentless surgery were excluded from randomized trials (usually before randomization, but some after randomization if difficult morphology became known during surgery). In particular, patients were excluded if stentless valve implantation was not possible due to severe root wall calcifications that could not be removed surgically, or if the aortic root was dilated. Many studies did not detail exclusions explicitly, and it was not possible to determine whether exclusion criteria (ie, for difficult or borderline morphology or other high-risk patient characteristics) were similar across trials.
In a number of trials with follow-up beyond hospital discharge, the completeness of follow-up was not transparently reported. This may particularly affect the estimates of hemodynamics and survival beyond 1 year, if differential follow-up occurred between groups and if loss to follow-up was significant. Sensitivity analysis to explore the potential impact of loss to follow-up was not possible because insufficient data were provided in the studies.
There were insufficient trials comparing the same stentless valves with the same stented valves to adequately perform subanalyses among all of the different comparisons. As a result, the meta-analysis does not provide information that is specific for each valve type. Rather, it compares a variety of stentless valves with a variety of stented valves, in aggregate. Some studies have suggested that mortality seems to vary across different stented biologic valve substitutes, which are not explained by differences in structural deterioration and hemodynamic performance alone, and this raises the possibility that some of the heterogeneity measured within this meta-analysis might be due to differences in valve performance. This meta-analysis could not assess noncomparative observational longitudinal studies.
Evidence from randomized trials shows that subcoronary stentless aortic valves improve hemodynamic parameters of EOAI, mean gradient, and peak gradient over the short and long term. These improvements have not led to significant improvements in patient morbidity, mortality, and resource-related outcomes; however, few trials reported on clinical outcomes beyond 1 year and definitive conclusions are not possible until sufficient evidence addresses longer-term effects.
The authors acknowledge Ms. Aurelie Alger and Ms. Elizabeth Chouinard for their professional assistance in organizing the consensus conference; Dr. Guyan Wang, Dr. Kathy Fang, Dr. Myoung Kim, and Dr. Ling Pei for their expert data extraction; and Ms. Jennifer Podeszwa-deOliviera and Ms. Karla VanKessel for their services in facilitating the literature searches and retrieval.
1.Levy D, Garrison RJ, Savage DD, et al. Prognostic implications of echocardiographically determine left ventricular mass in the Framingham heart study. N Engl J Med.
2.Sharma UC, Barenbrug P, Pokharel S, et al. Systematic review of the outcome of aortic valve replacement in patients with aortic stenosis. Ann Thorac Surg.
3.Kunadian B, Vijayalakshmi K, Thornley AR, et al. Meta-analysis of valve hemodynamics and left ventricular mass regression for stentless versus stented aortic valves. Ann Thorac Surg.
4.Raja SG, MacArthur KJ, Pollock JC. Impact of stentless aortic valves on left ventricular function and hypertrophy: current best available evidence. J Card Surg.
5.Kallikourdis A, Jacob S. Is a stentless aortic valve superior to conventional bioprosthetic valves for aortic valve replacement? Interact Cardiovasc Thorac Surg.
6.Moher D, Cook DJ, Eastwood S, et al. Improving the quality of reports of meta-analyses of randomised controlled trials: the QUOROM statement. Quality of reporting of meta-analyses. Lancet.
7.Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA.
8.Edmunds LH Jr, Clark RE, Cohn LH, et al. Guidelines for reporting morbidity and mortality after cardiac valvular operations. Ad Hoc Liaison Committee for Standardizing Definitions of Prosthetic Heart Valve Morbidity of the American Association for Thoracic Surgery and the Society of Thoracic Surgeons. J Thorac Cardiovasc Surg.
9.Ali A, Galstead JC, Cafferty F, et al. Early clinical and hemodynamic outcomes after stented and stentless aortic valve replacement: results from a randomized controlled trial. Ann Thorac Surg.
10.Ali A, Halstead JC, Cafferty F, et al. Are stentless valves superior to modern stented valves? A prospective randomized trial. Circulation.
11.Perez de Arenaza D, Lees B, Flather M, et al; on behalf of the ASSERT (Aortic Stentless versus Stented valve assessed by Echocardiography Randomized Trial). Randomized comparison of stentless versus stented valves for aortic stenosis: effects on left ventricular mass. Circulation.
12.Bakhtiary F, Abolmaali N, Dzemali O, et al. Impact of metchanical and biological aortic valve replacement on coronary perfusion: a prospective, randomized study. J Heart Valve Dis.
13.Bakhtiary F, Schiemann M, Dzemali O, et al. Stentless bioprostheses improve postoperative coronary flow more than stented prostheses after valve replacement for aortic stensosis. J Thorac Cardiovasc Surg.
14.Berg GA, McLaughlin E, Akar R, et al. A three year experience with the Toronto stentless porcine valve. Ann Thorac Cardiovasc Surg.
15.Chambers JB, Rimington HM, Hodson F, et al. The subcoronary Toronto stentless versus supra-annular perimount stented replacement aortic valve: early clinical and hemodynamic results of a randomized comparison in 160 patients. J Thorac Cardiovasc Surg.
16.Cohen G, Christakis GT, Joyner CD, et al. Are stentless valves hemodynamically superior to stented valves? A prospective randomized trial. Ann Thorac Surg.
17.Donatello F, Triggiani M, Mariani MA, et al. Rest and exercise hemodynamics of stentless porcine bioprostheses in aortic position. Cardiologia.
18.Doss M, Martens S, Wood JP, et al. Performance of stentless versus stented aortic valve bioprostheses in the elderly patients: a prospective randomized trial. Eur J Cardiothorac Surg.
19.Dunning J, Graham RJ, Thambyrajah J, et al. Stentless vs stented aortic valve bioprostheses: a prospective randomized controlled trial. Eur Heart J.
20.Graham R, Thambyrajah J, Stewart M, et al. Improved haemodynamic profile and left ventricular function following aortic valve replacement with a stentless rather than stented bioprosthesis: a randomized controlled trial. Heart.
2005; 91(Suppl I):A18.
21.Jasinski MJ, Ulbrych P, Kolowca M, et al. Early regional assessment of LV mass regression and function after stentless valve replacement: comparative randomized study. Heart Surg Forum.
22.Lehman S, Walther T, Kempfert J, et al. Stentless versus conventional xenograft aortic valve replacement: midterm results of a prospectively randomized trial. Ann Thorac Surg.
23.Maselli D, Pizio R, Bruno LP, et al. Left ventricular mass reduction after aortic valve replacement: homografts, stentless and stented valves. Ann Thorac Surg.
24.Miraldi F, Spagnesi L, Tallarico D, et al. Sorin stentless pericardial valve versus Carpentier-Edwards perimount pericardial bioprosthesis: is it worthwhile to struggle? Int J cardiol.
25.Santini F, Bertolini P, Montlbano G, et al. Hancock versus stentless bioprostheses for aortic valve replacement in patients older than 75 year. Ann Thorac Surg.
26.Sensky PR, Loubani M, Keal RP, et al. Does the type of prosthesis influence early left ventricular mass regression after aortic valve replacement? Assessment with magnetic resonance imaging. Am Heart J.
27.Szafranek A, Jasinski M, Kolowca M, et al. Plasma ANP and rennin-angiotensin-aldosterone system as new parameters describing the hemodynamics of the circulatory system after implantation of stented or stentless aortic valves. J Heart Valve Dis.
28.Totaro P, Degno N, Zaidi A, et al. Carpentier-Edwards perimount magna bioprosthesis: a stented valve with stentless performance? J Thorac Cardiovasc Surg.
29.Walther T, Falk V, Langebartels G, et al. Prospectively randomized evaluation of stentless versus conventional biological aortic valves. Impact on early regression of left ventricular hypertrophy. Circulation.
30.Walther T, Falk V, Langebartels G, et al. Regression of left ventricular hypertrophy after stentless versus conventional aortic valve replacement. Semin Thorac Cardiovasc Surg.
31.Williams RJ, Muir DF, Pathi V, et al. Randomized controlled trial of stented and stentless aortic bioprostheses: hemodynamic performance at 3 years. Semin Thorac Cardiovasc Surg.
32.Achtelik M, Pethig K, Schafers HJ, Borst HG. Stentless aortic valve replacement with the biocor prosthesis: indication and outcome. J Heart Valve Dis.
33.Bleiziffer S, Eichinger WB, Wagner I, et al. The Toronto root stentless valve in the subcoronary position is hemodynamically superior to the mosaic stented completely supra-annular bioprosthesis. J Heart Valve Dis.
34.Borger MA, Carson SM, Ivanov J, et al. Stentless aortic valves are hemodynamically superior to stented valves during mid-term follow-up: a large retrospective study. Ann Thorac Surg.
35.Bove T, Van Belleghem Y, Francois K, et al. Stentless and stented aortic valve replacement in elderly patients: factors affecting midterm clinical and hemodynamical outcome. Eur J Cardiothorac Surg.
36.Casabona R, D Paulis R, Zattera GF, et al. Stentless porcine and pericardial valve in aortic position. Ann Thorac Surg.
37.Casali G, Auremma S, Santini F, et al. Survival after stentless and stented xenograft aortic valve replacement: a concurrent, case-match trial. Ital Heart J.
38.Chatel D, Mica C, Blanchard D, et al. Short-term hemodynamic advantages of stentless CryoLife-O-Brien valve over stented bioprostheses from aortic valve replacement. Interact Cardiovasc Thorac Surg.
39.Collinson J, Flather M, Pepper JR, Henein M. Effects of valve replacement on left ventricular function in patients with aortic regurgitation and severe ventricular disease. J Heart Valve Dis.
40.Gelsomino S, Morocutti G, Frassani R, et al. Early recovery of left ventricular function after stentless versus stented aortic valve replacement for pure aortic stenosis and severe cardiac dysfunction. Semin Thorac Cardiovasc Surg.
41.Jasinski MJ, Hayton J, Kadziola Z, et al. Hemodynamic performance after stented vs stentless aortic valve replacement. J Cardiovasc Surg.
42.Jin XY, Zhang ZM, Gibson DG, et al. Effects of valve substitute on changes in left ventricular function and hypertrophy after aortic valve replacement. Ann Thorac Surg.
43.Luciani GB, Casali G, Auriemma S, et al. Survival after stentless and stented xenograft aortic valve replacement: a concurrent, controlled trial. Ann Thorac Surg.
44.Pibarot P, Dumesnil JG, Jobin J, et al. Hemodynamic and physical performance during maximal exercise in patients with an aortic bioprosthetic valve. J Am Coll Cardiol.
45.Rao V, Christakis GT, Sever J, et al. A novel comparison of stentless versus stented valves in the small aortic root. J Thorac Cardiovasc Surg.
46.Tamim M, Bove T, van Belleghem Y, et al. Stentless vs stented aortic valve replacement: left ventricular mass regressión. Asian Cardiovasc Thorac Ann.
47.Tsialtas D, Bolognesi R, Beghi C, et al. Stented versus stentless bioprostheses in aortic valve stenosis: effect on left ventricular remodeling. Heart Surg Forum.
48.van Nooten G, Caes F, Francois K, et al. Stentless or stented aortic valve implants in elderly patients? Eur J Cardiothorac Surg.
49.Westaby S, Horton M, Jin XU, et al. Survival advantage of stentless aortic bioprostheses. Ann Thorac Surg.
50.Westaby S, Jonson A, Payne N, et al. Does the use of a stentless bioprosthesis increase surgical risk? Semin Thorac Cardiovasc Surg.
This is a beautifully written meta-analysis comparing stentless to stented bioprosthetic aortic valves. It is an important contribution to the literature and accompanies the consensus statement that appears in this issue. The evidence from the randomized trials showed that subcoronary stentless aortic valves improved the hemodynamic parameters of effective orifice area, mean gradient and peak gradient over the short and long term. However, these improvements did not have any measurable impact on patient morbidity, late mortality and resource-related outcomes. The principal weakness of the conclusions of this report is that few trials reported late outcomes of greater than 1 year. It may be that the improved hemodynamics of the stentless valves would not begin to show any differences in morbidity and mortality until late follow-up was available. Moreover, most of these studies only reported late mortality, and few looked at late functional outcomes in these patients, which may have been improved in the stentless population. Whether there were late improvements in patient morbidity or resource-related outcomes is unclear because of the infrequency of these variables being reported. This represents an inadequacy of our surgical literature, which tends to focus only on early mortality and may mask the benefits of new technology and techniques.
Meta-analysis; Aortic valve surgery; Stentless valve surgery
© 2009 Lippincott Williams & Wilkins, Inc.
Highlight selected keywords in the article text.