Skip to main content
Journal of the American Heart Association: Cardiovascular and Cerebrovascular Disease logoLink to Journal of the American Heart Association: Cardiovascular and Cerebrovascular Disease
. 2026 Sep 9;15(19):e050284. doi: 10.1161/JAHA.126.050284

Minimally Invasive Versus Sternotomy in Redo Mitral Valve Surgery: Short‐ and Long‐Term Outcomes with Meta‐Regression and Time‐to‐Event Analysis

Kristine Santos 1, Joo Y B Kim Kim 2, Neel Patel 1, Halima Şah 3, Sriharsha Koduru 4, Federico Napoli 5, Nicolas A Brozzi 5, Jose L Navia 5, Antonio Polanco 5, Michel Pompeu Sá 5,✉
PMCID: PMC13745103  PMID: 42844757

Abstract

Background

Redo mitral valve surgery carries substantial risk, and the relative benefits of minimally invasive (MI‐MVS) versus sternotomy MVS, particularly regarding long‐term survival, remain uncertain.

Methods

Electronic databases were searched for studies comparing MI‐MVS with sternotomy MVS in adult redo MVS. Kaplan–Meier curves were digitized to reconstruct individual patient data for time‐to‐event analysis, with study‐level hazard ratios (HR) generated using a Cox frailty model. Perioperative outcomes were pooled using random‐effects models. Subgroup analyses evaluated propensity‐matched cohorts, endoscopic/video‐assisted minithoracotomy, and direct‐vision minithoracotomy. Meta‐regression explored the impact of study‐level covariates on mortality.

Results

Fifteen studies including 3457 patients were analyzed. MI‐MVS was associated with lower perioperative mortality compared with sternotomy MVS (odds ratio [OR], 0.33 [95% CI, 0.23–0.48]; P<0.001), reduced incidence of reoperation (OR, 0.65 [95% CI, 0.44–0.98]; P=0.04), and acute kidney injury (OR, 0.58 [95% CI, 0.39–0.87]; P=0.008); rates of stroke and arrhythmia were similar. The direction and magnitude of the mortality effect consistently favored MI‐MVS in the subgroup analyses. Time‐to‐event analysis demonstrated improved survival with MI‐MVS (HR, 0.74 [95% CI, 0.62–0.89]; P=0.001), but this was attenuated in the propensity‐matched analysis. Meta‐regression showed that sex distribution, mean age, and baseline left ventricular ejection fraction did not significantly moderate mortality effects.

Conclusions

In redo MVS, MI‐MVS is associated with improved perioperative outcomes without compromising long‐term survival; however, attenuation of the survival difference after propensity matching suggests that this long‐term advantage may reflect baseline risk rather than procedural effect.

Keywords: minimally invasive cardiac surgery, mitral valve, mitral valve surgery, right minithoracotomy

Subject Categories: Cardiovascular Surgery, Mortality/Survival, Complications


Nonstandard Abbreviations and Acronyms

AKI

acute kidney injury

DV‐MT

direct‐vision minithoracotomy

endo‐MT

endoscopic/video‐assisted minithoracotomy

IPD

individual patient data

KM

Kaplan–Meier

MI‐MVS

minimally invasive mitral valve surgery

re‐MVS

redo mitral valve surgery

STERN‐MVS

sternotomy mitral valve surgery

Clinical Perspective.

What Is New?

  • This study uses reconstructed individual patient time‐to‐event data from published Kaplan–Meier curves to provide the most comprehensive long‐term survival comparison of minimally invasive versus sternotomy approaches in redo mitral valve surgery.

  • Minimally invasive mitral valve surgery was associated with improved perioperative outcomes and a survival advantage in unadjusted analyses, although this effect was attenuated in propensity‐matched cohorts.

What Are the Clinical Implications?

  • Minimally invasive approaches may represent a reasonable alternative to redo sternotomy in appropriately selected patients treated at experienced centers, with perioperative benefits and no evidence of compromised long‐term survival.

The number of patients requiring redo mitral valve surgery (re‐MVS) has increased over the past decades due to more frequent implantation of bioprosthetic valves and the steady increase in prosthetic valve endocarditis. 1 Re‐MVS is associated with increased operative complexity and higher perioperative risk compared with primary MVS, largely due to adhesions, repeat cardiopulmonary bypass, and bleeding risk. 1 Conventional redo sternotomy (STERN‐MVS) has traditionally been the standard approach; however, minimally invasive mitral valve surgery (MI‐MVS) via right minithoracotomy has been increasingly adopted in reoperative settings. Several observational studies and institutional series have reported favorable early outcomes with MI‐MVS, including lower perioperative mortality and reduced morbidity compared with redo STERN‐MVS, supporting its technical feasibility and safety in selected patients. 2 , 3 , 4 , 5 , 6

Despite these favorable perioperative results, the impact of MI‐MVS on long‐term survival after re‐MVS remains unclear. The existing literature is dominated by small, single‐center observational studies, limiting statistical power and generalizability. Moreover, substantial heterogeneity in MI techniques ranging from endoscopic/video‐assisted minithoracotomy (endo‐MT) to direct‐vision minithoracotomy (DV‐MT). Consequently, prior systematic reviews have largely been restricted to perioperative end points and have not adequately determined whether observed advantages of MI approaches reflect durable survival benefit or underlying differences in patient risk profiles.

To address these limitations, we performed a systematic review and meta‐analysis comparing MI‐MVS with redo STERN‐MVS that integrates perioperative outcomes with long‐term survival assessment using reconstructed individual patient data (IPD) derived from published Kaplan–Meier (KM) curves. Prespecified subgroup analyses evaluated surgical technique and propensity‐matched cohorts to isolate higher‐quality comparative data, and meta‐regression was used to examine the influence of baseline patient characteristics on mortality. This comprehensive approach enables a more robust assessment of both short‐ and long‐term outcomes of MI‐MVS versus STERN‐MVS approaches in re‐MVS.

METHODS

This systematic review and meta‐analysis was performed following recognized methodological standards for evidence synthesis, guided by recommendations from the Cochrane Handbook for Systematic Reviews of Interventions. 7 Reporting adheres to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses statement and the completed checklist. The study protocol was registered prospectively in the International Prospective Register of Systematic Reviews (CRD420251243803). All data supporting the findings of this study are available within the article and its Supplementary Material. The analytic methods are described in sufficient detail to allow replication. All data were derived from previously published studies.

Eligibility Criteria

Studies were eligible for inclusion if they met the following criteria:

  1. Included adult patients (≥18 years) undergoing re‐MVS, defined as mitral valve repair or mitral valve replacement;

  2. Compared MI‐MVS performed via right minithoracotomy, including endo‐MT or DV‐MT techniques, with conventional STERN‐MVS;

  3. Reported at least 1 perioperative outcome of interest or long‐term survival outcomes, either as a hazard ratio (HR) or a KM curve; and

  4. Employed a prospective or retrospective comparative study design, including single‐ or multicenter studies with matched or unmatched cohorts.

Studies were excluded if they were noncomparative single‐arm series, case reports, editorials, reviews, or conference abstracts; involved pediatric populations; evaluated transcatheter mitral interventions or hybrid procedures; included partial or mini‐sternotomy comparators; or reported combined cardiac procedures without separately extractable outcomes for re‐MVS. When overlapping patient cohorts were identified, the most comprehensive or most recent data set was retained.

A systematic literature search was conducted to identify relevant studies published between January 2000 and October 2025. Electronic databases searched included MEDLINE, Scopus, and the Cochrane Library. Reference lists of included studies and relevant reviews were manually screened for additional eligible articles. The search strategy combined controlled vocabulary and free‐text terms related to re‐MVS, MI surgery, right minithoracotomy, port‐access or video‐assisted techniques, and sternotomy. The full search strategy is provided in the Supplementary Material (Data S1).

Study selection followed a 2‐stage process consisting of title and abstract screening followed by full‐text review of potentially eligible studies. Duplicate records were removed before screening. Two reviewers independently assessed eligibility, with disagreements resolved by consensus. As the analysis was based exclusively on previously published data, institutional ethics approval was not required.

Data Extraction and Outcomes

Data were independently extracted by 2 reviewers from each included study using a standardized data collection form. Extracted variables included study characteristics, patient demographics, and operative details, as well as the surgical approach used for re‐MVS. Discrepancies in data extraction were resolved by consensus.

Outcomes were categorized as perioperative (short‐term) and long‐term time‐to‐event outcomes for both MI‐MVS and conventional STERN‐MVS. Perioperative outcomes included perioperative mortality, reoperation, acute kidney injury (AKI), stroke, postoperative arrhythmia, and cardiopulmonary bypass time. Detailed outcome definitions are provided in the Supplementary Material (Data S2).

Statistical Analysis

For perioperative outcomes, odds ratios (ORs) and mean differences with corresponding 95% CIs were calculated for dichotomous and continuous variables, respectively. Pooled estimates were derived using a random‐effects model to account for clinical and methodological heterogeneity across studies. Statistical heterogeneity was quantified using the I2 statistic, with values >50% considered indicative of substantial heterogeneity. 8 Analyses were performed using Review Manager version 10.5.0.

For long‐term time‐to‐event outcomes, KM curves comparing MI‐MVS and STERN‐MVS were digitized using WebPlotDigitizer (v4.5) 9 and reconstructed into IPD using the 2‐stage algorithm described by Liu et al. 10 This method incorporates reported numbers at risk, event counts, and censoring patterns to approximate the underlying survival data. Reconstructed IPD were pooled using Cox proportional hazards models with shared frailty terms to account for study‐level clustering. This random‐effects IPD meta‐analysis approach allows for more granular modeling of censored survival data. As a validation step, parallel study‐level meta‐analyses were conducted using the DerSimonian–Laird random‐effects model with reported or reconstructed HRs. Concordance between reconstructed and published KM curves was assessed visually and by comparing numbers at risk and survival estimates where available. All survival analyses were performed in R (v4.3.0) using the survival and meta packages.

Meta Regression, Subgroup, and Leave‐One‐Out Sensitivity Analyses

To explore potential sources of heterogeneity, meta‐regression analyses were performed using study‐level covariates, including mean or median age, proportion of male patients, and baseline left ventricular ejection fraction (LVEF). Restricted maximum likelihood estimation was used for meta‐regression analyses. Given the limited number of studies and sparse event structure for several outcomes, meta‐regression analyses were considered exploratory and hypothesis‐generating. Sensitivity analyses were performed to evaluate the robustness of the primary findings. For perioperative mortality, study‐adjusted Firth penalized logistic regression was applied using reconstructed binary outcome data to account for sparse events and zero‐cell counts. For long‐term survival, reconstructed individual patient time‐to‐event data were reanalyzed using study‐adjusted Firth penalized Cox regression. Results from both analyses were compared with the corresponding primary models to assess robustness. 11

Prespecified subgroup analyses were additionally conducted according to MI surgical technique, including endo‐MT and DV‐MT, as well as propensity‐matched cohorts. Additional sensitivity analyses, including leave‐one‐out testing, were conducted to evaluate the robustness of pooled estimates and to determine whether any single study disproportionately influenced the results.

Quality Assessment—Risk of Bias, Publication Bias, and Certainty of Evidence

The risk of bias of nonrandomized studies was independently assessed using the Risk of Bias in Non‐Randomized Studies of Interventions tool. 12 Publication bias was evaluated by visual inspection of funnel plots for the primary outcomes, assessing the symmetry of study effect estimates around the pooled effect. 13 The overall certainty of evidence for each outcome was appraised using the Grading of Recommendations, Assessment, Development and Evaluation. 14 All assessments were performed independently by 2 reviewers, with discrepancies resolved through discussion and consensus.

RESULTS

Study Selection and Baseline Characteristics

The Preferred Reporting Items for Systematic Reviews and Meta‐Analyses flow diagram outlining study screening and selection is shown in Figure 1. The systematic search identified 672 records from electronic databases. After removal of duplicates, 354 records were screened by title and abstract, of which 334 were excluded. Full‐text review was undertaken for 20 reports, of which 5 were excluded, resulting in 15 studies included in the qualitative and quantitative synthesis. 2 , 3 , 4 , 5 , 6 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 A complete list of the excluded full‐text articles with reasons for exclusion is provided in the Supplementary Material (Data S3).

Figure 1. Preferred Reporting Items for Systematic Reviews and Meta‐Analyses flow diagram demonstrating study selection for the systematic review and meta‐analysis.

Figure 1

A total of 3457 patients, of whom 41% underwent MI‐MVS, were included across the 15 studies. Baseline characteristics were broadly comparable between MI‐MVS and STERN‐MVS cohorts, although reporting was heterogeneous (Table 1). Mean or median age ranged from the early 50s to early 70s years across studies, with similar age distributions between approaches in most series. The proportion of male patients varied widely but was generally comparable between MI‐MVS and STERN‐MVS groups. The prevalence of cardiovascular comorbidities, including hypertension, diabetes, renal dysfunction, and preoperative arrhythmia, was inconsistently reported, though no systematic imbalance favoring either approach was evident. Where reported, baseline LVEF was largely preserved and similar between groups, with several studies demonstrating modestly lower LVEF in STERN‐MVS cohorts.

Table 1.

Baseline Characteristics of Patients in the Included Studies

Study Number of patients MI‐MVS/STERN‐MVS Age MI‐MVS/STERN‐MVS, y Male sex % MI‐MVS/STERN‐MVS Hypertension MI‐MVS/STERN‐MVS Renal failure MI‐MVS/STERN‐MVS Arrhythmia MI‐MVS/STERN‐MVS LVEF % MI‐MVS/STERN‐MVS Diabetes MI‐MVS/STERN‐MVS
Bolotin 2004 23 38/33 67.9*/62.9* NR NR NR NR 46/45 NR
Burfeind 2002 3 60/155 60*/58* 55/27 NR NR NR 45/54 NR
Ghoneim 2016 21 12/6 67+/68.5+ 92/83 12/6 2/3 NR/NR NR/NR 1/3
Hamandi 2021 16 199/106 71+/66+ 54.3/44.3 168/81 6/5 128/62 42.6*/49.2* 62/26
Hiraoka 2013 24 10/27 68*/62.8* 50/66.7 NR/NR NR/NR 3/11 50.5*/64.4* NR/NR
Kwon 2022 6 162/218 51.5+/55.1+ 36.4/42.7 35/63 4/14 81/120 59.5/57.9 19/28
Losenno 2016 5 40/92 68*/62* 70/41 NR/NR 3/4 9/16 NR/NR 11/14
Monsefi 2022 20 27/53 66*/65* 48/47 15/34 6/15 NR/NR 53*/50* 6/14
Olsthoorn 2022 15 85/205 70+/66+ 67/51.2 NR/NR NR/NR NR/NR 55+/55+ 7/33
Onnasch 2002 2 39/25 56*/NR 41/NR NR/NR NR/NR NR/NR NR/NR NR/NR
Patel 2019 18 90/166 70.4*/62.2* 64/11 73/123 4/13 NR/NR 50.2*/52.9* 37/45
Sharony 2006 4 161/337 63.4*/64.9* 64/56.4 NR/NR 4/34 NR/NR NR/NR 23/46
Vallabhajosyula 2015 22 67/220 64*/61* 40/45 40/121 NR/NR 28/99 54*/54* 9/38
Zhang 2020 19 30/30 60.7*/60.2* 60/66 13/25 NR/NR 13/21 NR/NR 2/6
Zwischenberger 2024 17 503/247 63*/58* 48/42 NR/NR 13/10 247/136 71/37 NR/NR

* mean; +, median. LVEF indicates left ventricular ejection fraction; MI‐MVS, minimally invasive mitral valve surgery; NR, not reported; and STERN‐MVS, sternotomy mitral valve surgery.

Across included studies, MI‐MVS was performed via right minithoracotomy using either endo‐MT or DV‐MT techniques. Most MI‐MVS procedures employed peripheral cannulation for cardiopulmonary bypass depending on surgeon preference and patient anatomy. In contrast, STERN‐MVS was uniformly performed via repeat median sternotomy, typically using central cannulation and cardioplegic arrest. Surgical approach selection was nonrandomized and commonly based on surgeon experience, anatomical considerations, and perceived reentry risk, with several studies incorporating propensity matching to account for baseline differences between treatment groups.

Analysis of Perioperative Outcomes in the Total Cohort

Perioperative mortality was significantly lower with MI‐MVS compared with STERN‐MVS (OR, 0.33 [95% CI, 0.23–0.48]; P<0.001; I2=0%) (Figure 2). A sensitivity analysis using study‐adjusted Firth penalized logistic regression to account for sparse‐event data and zero‐cell counts demonstrated consistent results (OR, 0.30 [95% CI, 0.20–0.43]; P<0.001) (Table S1; Figures S1 through S5). In addition, reoperation rates (OR, 0.65 [95% CI, 0.44–0.98]; P=0.040; I2=0%) and incidence of AKI were lower after MI‐MVS compared with STERN‐MVS (OR, 0.58 [95% CI, 0.39–0.87]; P=0.008; I2=7%). There was no significant difference cardiopulmonary bypass time between MI‐MVS and STERN‐MVS, although heterogeneity across studies was substantial (mean difference, 1.17 [95% CI, −11.18 to 13.52]; P=0.85; I2=98%). Whereas incidence of strokes (OR, 0.91 [95% CI, 0.57–1.48]; P=0.710; I2=13%) and arrhythmias (OR, 0.75 [95%, CI 0.49–1.15]; P=0.190; I2=37%) did not differ significantly between MI‐MVS and STERN‐MVS. A summary of these outcomes is presented in Table 2, with corresponding forest plots provided in the Supplementary Material (Table S1).

Figure 2. Forest plot comparing peri‐operative mortality between MI‐MVS and STERN‐MVS in redo mitral valve surgery.

Figure 2

Burfeind 23 ; Monsefi 20 ; Ghoneim 21 ; Zwischenberger 17 ; Patel 18 ; Hamandi 16 ; Kwon 6 ; Losenno 5 ; Zhang 19 ; Sharony 4 ; Vallabhajosyula 22 ; Olsthoorn 15 ; Hiraoka 24 . MI‐MVS was associated with significantly lower peri‐operative mortality compared with STERN‐MVS (OR, 0.33 [95% CI, 0.23–0.48]; P<0.001; I2=0%). M‐H indicates Mantel–Haenszel; MI‐MVS, minimally invasive mitral valve surgery; and STERN‐MVS, sternotomy mitral valve surgery.

Table 2.

Summary of Outcomes for the Total Cohort

Outcome Number of studies Number of patients Effect estimate (95% CI, P value)
In‐hospital mortality 13 3307 OR, 0.33 [95% CI, 0.23 to 0.48]; P<0.001; I2=0%
Reoperation 11 2712 OR, 0.65 [95% CI, 0.44 to 0.98]; P=0.040; I2=0%
Acute kidney injury 7 1836 OR, 0.58 [95% CI, 0.39 to 0.87]; P=0.008; I2=7%
Stroke 12 2906 OR, 0.91 [95% CI, 0.57 to 1.48]; P=0.710; I2=13%
Arrhythmia 8 1665 OR, 0.75 [95% CI, 0.49 to 1.15]; P=0.190; I2=37%
Myocardial infarction 3 434 OR, 0.71 [95% CI, 0.06 to 8.85]; P=0.790; I2=35%
Cardiopulmonary bypass time 13 3109 Mean difference, 1.17 [95% CI, −11.18 to 13.52]; P=0.85; I2=98%

OR indicates odds ratio.

Assessment of Perioperative Mortality in the Subgroups

In the DV‐MT subgroup, 5 studies with 1162 patients contributed to the outcome of perioperative mortality. Perioperative mortality was lower with MI‐MVS compared with STERN‐MVS (OR, 0.39 [95% CI, 0.23–0.65]; P<0.001; I2=0%). A similar reduction in mortality was observed in the endo‐MT subgroup, including 5 studies with 886 patients (OR, 0.36 [95% CI, 0.15–0.88]; P=0.020; I2=0%). This mortality benefit remained statistically significant in the propensity‐matched cohorts, comprising 4 studies and 850 patients (OR, 0.33 [95% CI, 0.11–0.98]; P=0.050; I2=32%). Summaries of these outcomes are presented in Tables 3, 4, and 5, with corresponding forest plots provided in Supplementary Material (Figures S6 through S8).

Table 3.

Summary of Endoscopic/Video Assisted Minithoracotomy

Outcome Number of studies Number of patients Effect estimate (95% CI, P value)
In‐hospital mortality 5 751 OR, 0.36 [95% CI, 0.15–0.88]; P=0.020; I2=0%
Reoperation 5 751 OR, 0.47 [95% CI, 0.13–1.17]; P=0.090; I2=0%
Acute kidney injury 2 185 OR, 0.20 [95% CI, 0.04–1.17]; P=0.070; I2=0%
Stroke 6 724 OR, 1.37 [95% CI, 0.60–3.11]; P=0.450; I2=0%
Arrhythmia 3 222 OR, 0.22 [95% CI, 0.05–0.99]; P=0.050; I2=0%

OR indicates odds ratio.

Table 4.

Summary of Direct Vision Minithoracotomy

Outcome Number of studies Number of patients Effect estimate (95% CI, P value)
In‐hospital mortality 5 1162 OR, 0.39 [95% CI, 0.23–0.65]; P<0.001; I2=0%
Reoperation 5 1162 OR, 0.88 [95% CI, 0.52–1.49]; P=0.630; I2=0%
Acute kidney injury 3 596 OR, 0.82 [95% CI, 0.42–1.58]; P=0.550; I2=0%
Stroke 5 1162 OR, 0.48 [95% CI, 0.27–0.86]; P=0.010; I2=0%
Arrhythmia 2 98 OR, 0.97 [95% CI, 0.41–2.29]; P=0.940; I2=0%
Myocardial infarction 1 18 OR, 0.15 [95% CI, 0.01–4.20]; P=0.260; I2=not reported

OR indicates odds ratio.

Table 5.

Summary of Outcomes for the Propensity Matched Cohorts Only

Outcome Number of studies Number of patients Effect estimate (95% CI, P value)
In‐hospital mortality 4 850 OR, 0.33 [95% CI, 0.11–0.98]; P=0.050; I2=32%
Reoperation 2 514 OR, 0.68 [95% CI, 0.28–1.63]; P=0.390; I2=0%
Acute kidney injury 3 670 OR, 0.57 [95% CI, 0.19–1.66]; P=0.300; I2=59%
Stroke 4 850 OR, 1.07 [95% CI, 0.41–2.79]; P=0.880; I2=0%
Arrhythmia 3 670 OR, 0.64 [95% CI, 0.36–1.55]; P=0.140; I2=52%
Myocardial infarction 2 340 OR, 2.03 [95% CI, 0.18–22.80]; P=0.570; I2=not reported

OR indicates odds ratio.

Analysis of Long‐Term Survival

Five studies contributed to the pooled long‐term survival analysis, comprising 1880 patients, of whom 47% underwent MI‐MVS. Reconstructed IPD time‐to‐event analysis demonstrated significantly improved long‐term survival with MI‐MVS compared with STERN‐MVS (HR, 0.74 [95% CI, 0.62–0.89]; P=0.001), with separation of KM curves maintained to ∼8 years of follow‐up (Figure 3). A sensitivity analysis using study‐adjusted Firth penalized Cox regression demonstrated consistent results (HR, 0.75 [95% CI, 0.62–0.91]; P=0.003) (Table S2). In contrast, the complementary study‐level random‐effects meta‐analysis of reconstructed or published HR showed a consistent direction of effect favoring MI‐MVS but did not reach statistical significance (HR, 0.77 [95% CI, 0.58–1.02]; P=0.06; I2=49%) (Figure S9). In the propensity‐matched cohort, comprising 3 studies with 670 patients, reconstructed IPD analysis showed no significant difference in long‐term survival between MI‐MVS and STERN‐MVS (HR, 0.77 [95% CI, 0.57–1.03]; P=0.080), with largely overlapping survival curves throughout follow‐up (Figure 4). This was corroborated by the study‐level random‐effects meta‐analysis of propensity‐matched studies, which also demonstrated no survival advantage associated with MI‐MVS (HR, 0.89 [95% CI, 0.47–1.65]; P=0.700; I2=74%) (Figure S10).

Figure 3. Kaplan–Meier survival curves reconstructed from published time‐to‐event data comparing long‐term survival following MI‐MVS versus STERN‐MVS in the overall cohort.

Figure 3

MI‐MVS was associated with a survival advantage (HR, 0.74 [95% CI, 0.62–0.89]; P=0.001), with early separation of survival curves persisting during follow‐up. Shaded regions represent 95% CIs. HR indicates hazard ratio; MI‐MVS, minimally invasive mitral valve surgery; and STERN‐MVS, sternotomy mitral valve surgery.

Figure 4. Kaplan–Meier survival curves reconstructed from propensity‐matched cohorts comparing MI‐MVS and STERN‐MVS.

Figure 4

The survival advantage associated with MI‐MVS was attenuated in propensity‐matched analyses (HR, 0.77 [95% CI, 0.57–1.03]; P=0.08), with overlapping CIs and survival curves. Shaded regions represent 95% CIs. MI‐MVS indicates minimally invasive mitral valve surgery; and STERN‐MVS, sternotomy mitral valve surgery.

Meta Regression and Leave‐One‐Out Sensitivity Analysis

Meta‐regression analyses demonstrated no significant effect modification of perioperative mortality by study‐level covariates, including proportion of male patients (β=0.020; P=0.315), mean age (β=0.006; P=0.878), or baseline LVEF (β=0.013; P=0.878) (Figures S11 through S13). Heterogeneity across analyses of long‐term survival and all perioperative outcomes was low to negligible. Leave‐one‐out sensitivity analyses confirmed the robustness of the pooled estimates, with no single study exerting a disproportionate influence on perioperative mortality or secondary outcomes. Given the limited number of studies and sparse event structure, these analyses should be interpreted as exploratory and hypothesis‐generating.

Analysis of Risk of Bias, Publication Bias, and Certainty of Evidence

Using the Risk of Bias in Non‐randomized Studies of Interventions tool (Figure 5), most included studies were judged to be at moderate to serious risk of bias, driven primarily by confounding and selection bias, reflecting the nonrandomized observational design of the evidence base. Overall, 4 studies were assessed as having moderate risk of bias, and the remainder were judged to be at serious risk. According to the Grading of Recommendations, Assessment, Development and Evaluation assessment (Table 6), the certainty of evidence was low for perioperative mortality, supported by a large and consistent effect favoring MI‐MVS despite residual confounding. Certainty was very low for reoperation and AKI, mainly due to risk of bias and imprecision, although effect estimates consistently favored MI‐MVS. Evidence certainty was very low for stroke and postoperative arrhythmia, reflecting serious risk of bias and wide CIs crossing the line of no effect.

Figure 5. Risk‐of‐bias assessment of included studies using the ROBINS‐I tool.

Figure 5

Bolotin 23 ; Burfeind 3 ; Ghoneim 21 ; Hamandi 16 ; Hiroaka 24 ; Kwon 6 ; Losenno 5 ; Monsefi 20 ; Olsthoorn 15 ; Onnasch 2 ; Patel 18 ; Sharony 4 ; Vallabhajosyula 22 ; Zhang 19 ; Zwischenberger 17 . ROBINS‐I indicates Risk of Bias in Non‐randomized Studies of Interventions.

Table 6.

Grading of Recommendations, Assessment, Development and Evaluation Minimally Invasive Compared With Sternotomy in Redo Mitral Valve Surgery

Certainty assessment Summary of findings
Participants (studies) follow‐up Risk of bias Inconsistency Indirectness Imprecision Publication bias Overall certainty of evidence Study event rates (%) Relative effect (95% CI) Anticipated absolute effects
With sternotomy With Minimally invasive Risk with sternotomy Risk difference with minimally invasive
Perioperative mortality
3307 (13 nonrandomized studies) Serious* Not serious Not serious Not serious Not serious

⨁⨁◯◯

Low* , †

162/1855 (8.7%) 38/1452 (2.6%) OR 0.33 (0.23–0.48) 162/1855 (8.7%) 57 fewer per 1000 (from 66 fewer to 43 fewer)
Reoperation
2712 (11 nonrandomized studies) Serious† Not serious Not serious Serious‡ Not serious

⨁◯◯◯

Very low † , ‡

79/1544 (5.1%) 41/1168 (3.5%) OR 0.65 (0.44–0.98) 79/1544 (5.1%) 17 fewer per 1000 (from 28 fewer to 1 fewer)
Acute kidney injury
1836 (7 nonrandomized studies) Not serious§ Not serious Serious|| Serious Not serious

⨁◯◯◯

Very low § , || ,

91/864 (10.5%) 64/972 (6.6%) OR 0.58 (0.39–0.87) 91/864 (10.5%) 41 fewer per 1000 (from 61 fewer to 12 fewer)
Stroke
2906 (12 nonrandomized studies) Serious# Not serious Not serious** Very serious†† Not serious

⨁◯◯◯

Very low # , ** , ††

71/1650 (4.3%) 44/1256 (3.5%) OR 0.91 (0.57–1.48) 71/1650 (4.3%) 4 fewer per 1000 (from 18 fewer to 19 more)
Arrhythmia
1665 (8 non‐randomized studies) Serious‡‡ Serious§§ Not serious Serious|||| Not serious

⨁◯◯◯

Very low ‡‡ , §§ , ||||

173/759 (22.8%) 143/906 (15.8%) OR 0.75 (0.49–1.15) 173/759 (22.8%) 47 fewer per 1000 (from 102 fewer to 26 more)

OR indicates odds ratio.

*

6/13 studies have an overall moderate risk of bias; 7/13 studies have an overall serious risk of bias.

†

7/11 studies have an overall serious risk and 4/11 have moderate risk.

‡

The CI is too wide to be confident that a clinically important benefit truly exists.

§

4/7 studies have an overall moderate risk of bias and 3/7 have serious risk of bias.

||

Acute kidney injury can be defined by various criteria (eg, Kidney Disease Improving Global Outcomes; Risk, Injury, Failure, Loss of kidney function, and End‐stage kidney disease; serum creatinine rise). Variability in definition across studies introduces some indirectness in measurement.

The 95% CI (0.39–0.87) does not cross the line of no effect (OR=1), but the upper bound (0.87) is very close to 1, indicating the benefit could be very small. The lower bound (0.39) suggests a potentially important benefit. The total number of events is moderate (155 events). The CI is moderately wide.

#

7/12 studies have an overall serious risk of bias and 5/12 studies have an overall moderate risk of bias.

**

The outcome (stroke) is a clear, patient‐important event.

††

Very serious imprecision. The 95% CI (0.57–1.48) is very wide. It crosses the line of no effect (OR=1).

‡‡

4/8 moderate and 4/8 serious risk of bias.

§§

Moderate. I2=37% indicates low to moderate heterogeneity.

||||

The 95% CI (0.49–1.15) is wide.

Visual inspection of the funnel plot for perioperative mortality demonstrated no meaningful asymmetry (Figure S14). This was supported by Egger's regression test, which showed no evidence of small‐study effects (intercept −0.36; P=0.47). Consistent with these findings, Duval and Tweedie's trim‐and‐fill method did not impute any missing studies, and the adjusted effect estimate was identical to the observed estimate, indicating a low likelihood of publication bias.

DISCUSSION

Summary of Evidence

Our main findings were as follows:

  1. MI‐MVS for redo mitral valve procedures is associated with significantly lower perioperative mortality compared with redo STERN‐MVS.

  2. MI‐MVS is associated with reduced perioperative morbidity, including lower rates of reoperation and AKI, without an increased risk of stroke or postoperative arrhythmia.

  3. Time‐to‐event analysis demonstrates an apparent long‐term survival advantage for MI‐MVS in unadjusted cohorts, with early separation of KM survival curves.

  4. The apparent survival advantage was attenuated, with overlapping CIs and greater uncertainty in the propensity‐matched analyses, suggesting that baseline patient selection rather than procedural effect largely accounts for observed differences in long‐term survival.

  5. Meta‐regression analyses indicate that sex distribution, mean age, and baseline LVEF do not significantly moderate mortality outcomes.

Comments

Redo MVS is technically demanding and associated with increased perioperative risk due to adhesions, reentry injury, repeat cardiopulmonary bypass, and bleeding complications. 1 Our findings demonstrate that MI‐MVS mitigates several of these risks, translating into a substantial reduction in perioperative mortality with consistently low heterogeneity across studies. The observed reductions in reoperation and AKI further support the hypothesis that avoidance of repeat median sternotomy and reduced surgical trauma may improve early postoperative recovery. These benefits persisted across subgroups defined by surgical technique, including both endo‐MT and DV‐MT, as well as in propensity‐matched cohorts, reinforcing the robustness of the perioperative advantage associated with MI‐MVS. Importantly, sensitivity analysis using study‐adjusted Firth penalized logistic regression yielded nearly identical estimates to the primary random‐effects meta‐analysis, suggesting that the observed mortality benefit was not materially influenced by sparse‐event data or zero‐cell counts. The lack of difference in stroke and arrhythmia rates between approaches is noteworthy. Concerns regarding peripheral cannulation, retrograde perfusion, and limited exposure during MI reoperative surgery have historically raised questions about neurological risk. 25 Our pooled analysis suggests that, in contemporary practice, MI‐MVS can be performed without increasing cerebrovascular or rhythm‐related complications when appropriate patient selection and surgical expertise are applied.

Another key strength of this study is the integration of long‐term survival assessment using reconstructed IPD derived from published KM curves. In the overall cohort, time‐to‐event analysis demonstrated improved survival with MI‐MVS, with early separation of survival curves that persisted throughout follow‐up. However, when analyses were restricted to propensity‐matched cohorts, the apparent survival advantage was attenuated, with overlapping survival curves and CIs crossing the line of no effect, alongside substantial study‐level heterogeneity. These findings suggest that the long‐term survival benefit observed in unadjusted analyses likely reflects differences in baseline patient risk rather than a direct causal effect of surgical approach. Patients selected for MI‐MVS may be younger, have fewer comorbidities, or possess more favorable anatomy, factors that are difficult to fully account for in observational studies. The absence of effect modification by sex, age, or baseline LVEF in meta‐regression analyses supports the notion that unmeasured confounders and selection bias may play a dominant role in observed survival differences.

Several previous systematic reviews comparing MI and sternotomy approaches in redo‐MVS have largely focused on perioperative outcomes failed to evaluate long‐term survival. 26 , 27 Our findings extend this literature by demonstrating that, although MI‐MVS confers clear perioperative advantages, evidence for a durable long‐term survival benefit is attenuated once baseline patient risk is accounted for. By integrating reconstructed individual patient time‐to‐event analyses with conventional study‐level meta‐analysis, the present study provides a more comprehensive and methodologically robust assessment of long‐term outcomes than previous reports. In addition, we account for important heterogeneity in MI techniques, including endo‐MT and DV‐MT approaches, which has not been systematically addressed in prior analyses.

Several biologically plausible mechanisms may explain the perioperative advantages observed with MI‐MVS in re‐MVS. First, avoiding redo median sternotomy may reduce the need for extensive mediastinal dissection and lower the risk of catastrophic reentry injury to cardiac and vascular structures which are events that are a recognized driver of morbidity and mortality in reoperative cardiac surgery. 5 , 28 Secondly, right minithoracotomy approaches may reduce surgical trauma and bleeding compared with redo sternotomy because they avoid splitting the sternum and require less dissection of bony and soft‐tissue structures. 29 This minimization of surgical insult has been associated with reduced intraoperative blood loss and lower blood product transfusion requirements in MI mitral surgery series, particularly in reoperative settings. 25 Reduced perioperative bleeding and transfusion is clinically relevant because red blood cell transfusion has been consistently associated with postoperative AKI after cardiac surgery, 30 providing a mechanistic link between lower surgical trauma and the reduced AKI observed in our pooled analysis.

From a clinical perspective, these results support MI‐MVS as a safe and effective alternative to redo sternotomy in appropriately selected patients. The consistent perioperative benefits, coupled with comparable long‐term survival in propensity‐matched cohorts, suggest that MI‐MVS can reduce early morbidity and mortality without sacrificing durability. Surgical approach selection should therefore be guided by patient‐specific factors, anatomical considerations, and institutional expertise rather than concerns regarding long‐term survival disadvantage.

Limitations

Our meta‐analysis has several limitations. All included studies were observational, and residual confounding cannot be excluded despite the use of propensity‐matched analyses and meta‐regression. Heterogeneity in MI techniques, perioperative management, and outcome definitions may have influenced pooled estimates. Several perioperative outcomes also involved sparse event data and small cell counts, which may limit the robustness of conventional random‐effects meta‐analytic approaches that rely on large‐sample approximations. Alternative exact or Bayesian hierarchical methods were considered; however, these approaches would require additional assumptions regarding exchangeability and prior distributions that may be difficult to justify given the heterogeneity in study design, surgical techniques, and patient selection across observational cohorts. Furthermore, individual‐level binary outcome data were not consistently available across studies, precluding patient‐level modeling approaches. Although Firth‐penalized Cox regression may reduce small‐sample bias in sparse time‐to‐event data, it was not applied because the analysis required study‐level frailty terms and the fitted Cox models showed no evidence of nonconvergence or monotone likelihood.

In addition, complications related to peripheral cannulation, including vascular injury and limb ischemia, were inconsistently reported and could not be systematically analyzed, representing an important limitation in assessing the full safety profile of MI approaches. Reconstruction of IPD from KM curves, although validated and widely used, remains an indirect method and is subject to digitization error. Finally, long‐term outcomes beyond survival, such as valve durability and quality of life, were inconsistently reported and could not be systematically assessed.

Future research should focus on prospective, multicenter registries with standardized reporting of operative technique, patient risk profiles, and long‐term outcomes. Although randomized trials in this setting may be challenging, high‐quality observational studies with robust adjustment methods and longer follow‐up will be essential to further clarify the role of MI‐MVS in redo MVS.

CONCLUSIONS

In redo MVS, MI approaches are associated with improved perioperative outcomes without evidence of compromised long‐term survival compared with conventional STERN‐MVS. Although an apparent long‐term survival advantage was observed in unadjusted analyses, this effect was attenuated in propensity‐matched cohorts, suggesting that differences in baseline patient risk and selection may contribute substantially to observed survival differences. Given the observational nature of the available evidence, heterogeneity across studies, and limitations related to sparse event data, these findings should be considered hypothesis‐generating. Nevertheless, the current evidence supports MI techniques as a reasonable alternative to sternotomy in appropriately selected patients and experienced centers.

Sources of Funding

The authors report no funding for this work.

Disclosures

The authors report no conflicts of interest.

Supporting information

Figures S1–S14

Tables S1–S2

JAH3-15-e050284-s002.pdf (525.5KB, pdf)

PRISMA Checklist

JAH3-15-e050284-s001.pdf (194.6KB, pdf)

Acknowledgments

None.

This article was sent to John S. Ikonomidis, MD, PhD, Guest Editor, for review by expert referees, editorial decision, and final disposition.

For Sources of Funding and Disclosures, see page 12.

REFERENCES

  • 1. Schumacher K, Marin Cuartas M, de la Cuesta M, Noack T, Kiefer P, Leontyev S, Borger MA, Vollroth M, Misfeld M. Early and long‐term outcomes following redo mitral valve surgery in patients with prior minimally invasive mitral valve surgery. Interdiscip Cardiovasc Thorac Surg. 2024;38:ivae042. doi: 10.1093/icvts/ivae042 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Onnasch JF, Schneider F, Falk V, Walther T, Gummert J, Mohr FW. Minimally invasive approach for redo mitral valve surgery: a true benefit for the patient. J Card Surg. 2002;17:14–19. doi: 10.1111/j.1540-8191.2001.tb01214.x [DOI] [PubMed] [Google Scholar]
  • 3. Burfeind WR, Glower DD, Davis RD, Landolfo KP, Lowe JE, Wolfe WG. Mitral Surgery After Prior Cardiac Operation: Port‐Access Versus Sternotomy or Thoracotomy [Internet]. 2002. http://www.sts.org. [DOI] [PubMed]
  • 4. Sharony R, Grossi EA, Saunders PC, Schwartz CF, Ursomanno P, Ribakove GH, Galloway AC, Colvin SB. Minimally invasive reoperative isolated valve surgery: early and mid‐term results. J Card Surg. 2006;21:240–244. doi: 10.1111/j.1540-8191.2006.00271.x [DOI] [PubMed] [Google Scholar]
  • 5. Losenno KL, Jones PM, Valdis M, Fox SA, Kiaii B, Chu MWA. Higher‐risk mitral valve operations after previous sternotomy: endoscopic, minimally invasive approach improves patient outcomes. Can J Surg. 2016;59:399–406. doi: 10.1503/cjs.004516 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Kwon Y, Park SJ, Kim HJ, Kim JB, Jung SH, Choo SJ, Lee JW. Mini‐thoracotomy and full‐sternotomy approach for reoperative mitral valve surgery after a previous sternotomy. Interact Cardiovasc Thorac Surg. 2022;34:354–360. doi: 10.1093/icvts/ivab309 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Higgins J, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch V. Cochrane Handbook for Systematic Reviews of Interventions. 2nd ed. Chichester: John Wiley & Sons; 2019. [Google Scholar]
  • 8. Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta‐analysis. Stat Med. 2002;21:1539–1558. doi: 10.1002/sim.1186 [DOI] [PubMed] [Google Scholar]
  • 9. Rohatgi A. WebPlotDigitizer Version 4.5 .
  • 10. Liu N, Zhou Y, Lee JJ. IPDfromKM: reconstruct individual patient data from published Kaplan–Meier survival curves. BMC Med Res Methodol. 2021;21:111. doi: 10.1186/s12874-021-01308-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Firth D. Bias reduction of maximum likelihood estimates. Biometrika. 1993;80:27–38. doi: 10.1093/biomet/80.1.27 [DOI] [Google Scholar]
  • 12. Sterne JA, Hernán MA, Reeves BC, Savović J, Berkman ND, Viswanathan M, Henry D, Altman DG, Ansari MT, Boutron I, et al. ROBINS‐I: a tool for assessing risk of bias in non‐randomised studies of interventions. BMJ. 2016;355:i4919. doi: 10.1136/bmj.i4919 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Sterne JAC, Harbord RM. Funnel plots in meta‐analysis. Stata J. 2004;4:127–141. doi: 10.1177/1536867X0400400204 [DOI] [Google Scholar]
  • 14. Guyatt GH, Oxman AD, Vist GE, Kunz R, Falck‐Ytter Y, Alonso‐Coello P, Schünemann HJ; GRADE Working Group . GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336:924–926. doi: 10.1136/bmj.39489.470347.AD [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Olsthoorn JR, Heuts S, Houterman S, Maessen JG, Sardari Nia P; Cardiothoracic Surgery Registration Committee of the Netherlands Heart Registration . Minimally invasive approach compared to resternotomy for mitral valve surgery in patients with prior cardiac surgery: retrospective multicentre study based on The Netherlands Heart Registration. Eur J Cardiothorac Surg. 2022;62:ezac420. doi: 10.1093/ejcts/ezac420 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Hamandi M, Squiers JJ, Lanfear AT, Banwait JK, Meidan TG, Smith RL, Hutcheson K, DiMaio JM, Mack MJ, George TJ, et al. Minimally invasive mitral valve surgery after previous sternotomy: a propensity‐matched analysis. J Card Surg. 2021;36:3177–3183. doi: 10.1111/jocs.15711 [DOI] [PubMed] [Google Scholar]
  • 17. Zwischenberger BA, Gaca JG, Milano C, Carr K, Glower DD. Late survival after redo mitral operation with minithoracotomya compared with sternotomy. Ann Thorac Surg. 2024;117:353–359. doi: 10.1016/j.athoracsur.2023.08.017 [DOI] [PubMed] [Google Scholar]
  • 18. Patel NC, Hemli JM, Seetharam K, Graver LM, Brinster DR, Pirelli L, Scheinerman SJ, Hartman AR. Reoperative mitral valve surgery via sternotomy or right thoracotomy: a propensity‐matched analysis. J Card Surg. 2019;34:976–982. doi: 10.1111/jocs.14170 [DOI] [PubMed] [Google Scholar]
  • 19. Zhang H, Xu HS, Wen B, Zhao WZ, Liu C. Minimally invasive beating heart technique for mitral valve surgery in patients with previous sternotomy and giant left ventricle. J Cardiothorac Surg. 2020;15:122. doi: 10.1186/s13019-020-01171-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Monsefi N, Makkawi B, Öztürk M, Alirezai H, Alaj E, Bakhtiary F. Right minithoracotomy and resternotomy approach in patients undergoing a redo mitral valve procedure. Interact Cardiovasc Thorac Surg. 2022;34:33–39. doi: 10.1093/icvts/ivab228 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Ghoneim A, Bouhout I, Mazine A, Fortin W, El‐Hamamsy I, Jeanmart H, Pellerin M, Bouchard D. Beating heart minimally invasive mitral valve surgery in patients with patent coronary bypass grafts. Can J Cardiol. 2016;32:987.e1–987.e6. doi: 10.1016/j.cjca.2015.09.016 [DOI] [PubMed] [Google Scholar]
  • 22. Vallabhajosyula P, Wallen T, Pulsipher A, Pitkin E, Solometo LP, Musthaq S, Fox J, Acker M, Hargrove WC III. Minimally invasive port access approach for reoperations on the mitral valve. Ann Thorac Surg. 2015;100:68–73. doi: 10.1016/j.athoracsur.2015.02.039 [DOI] [PubMed] [Google Scholar]
  • 23. Bolotin G, Kypson AP, Reade CC, Chu VF, Freund WL, Nifong LW, Chitwood WR Jr. Should a video‐assisted mini‐thoracotomy be the approach of choice for reoperative mitral valve surgery? J Heart Valve Dis. 2004;13:155–158. [PubMed] [Google Scholar]
  • 24. Hiraoka A, Kuinose M, Totsugawa T, Chikazawa G, Yoshitaka H. Mitral valve reoperation under ventricular fibrillation through right mini‐thoracotomy using three‐dimensional videoscope. J Cardiothorac Surg. 2013;8:81. doi: 10.1186/1749-8090-8-81 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Modi P, Chitwood WR. Retrograde femoral arterial perfusion and stroke risk during minimally invasive mitral valve surgery: is there cause for concern? Ann Cardiothorac Surg. 2013;2:E1. doi: 10.3978/j.issn.2225-319X.2013.11.13 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Hanafy DA, Melisa S, Andrianto GA, Suwatri WT. Outcomes of minimally invasive versus conventional sternotomy for redo mitral valve surgery according to Mitral Valve Academic Research Consortium: a systematic review and meta‐analysis. Asian J Surg. 2024;47:35–42. doi: 10.1016/j.asjsur.2023.09.001 [DOI] [PubMed] [Google Scholar]
  • 27. Daemen JHT, Heuts S, Olsthoorn JR, Maessen JG, Sardari Nia P. Right minithoracotomy versus median sternotomy for reoperative mitral valve surgery: a systematic review and meta‐analysis of observational studies. Eur J Cardiothorac Surg. 2018;54:817–825. doi: 10.1093/ejcts/ezy173 [DOI] [PubMed] [Google Scholar]
  • 28. Valente T, Bocchini G, Rossi G, Sica G, Davison H, Scaglione M. MDCT prior to median re‐sternotomy in cardiovascular surgery: our experiences, infrequent findings and the crucial role of radiological report. Br J Radiol. 2019;92:20170980. doi: 10.1259/bjr.20170980 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Murzi M, Solinas M, Glauber M. Is a minimally invasive approach for re‐operative mitral valve surgery superior to standard resternotomy? Interact Cardiovasc Thorac Surg. 2009;9:327–332. doi: 10.1510/icvts.2009.203448 [DOI] [PubMed] [Google Scholar]
  • 30. Kindzelski BA, Corcoran P, Siegenthaler MP, Horvath KA. Postoperative acute kidney injury following intraoperative blood product transfusions during cardiac surgery. Perfusion. 2018;33:62–70. doi: 10.1177/0267659117712405 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figures S1–S14

Tables S1–S2

JAH3-15-e050284-s002.pdf (525.5KB, pdf)

PRISMA Checklist

JAH3-15-e050284-s001.pdf (194.6KB, pdf)

Articles from Journal of the American Heart Association: Cardiovascular and Cerebrovascular Disease are provided here courtesy of Wiley

RESOURCES