Minimally invasive versus conventional surgery of the ascending aorta and root: a systematic review and meta-analysis

Summary

Limited uptake of minimally invasive surgery (MIS) of the aorta hinders assessment of its efficacy compared to median sternotomy (MS). The objective of this systematic review is to compare operative and perioperative outcomes for MIS versus MS. Online databases Medline, EMBASE, Cochrane Library and Web of Science were searched from inception until July 2018. Both randomized and observational studies of patients undergoing aortic root, ascending aorta or aortic arch surgery by MIS versus MS were eligible for inclusion. Primary outcomes were 30-day mortality, reoperation for bleeding, perioperative renal impairment and neurological events. Intraoperative and postoperative timing measures were also evaluated. Thirteen observational studies were included comparing 1101 MIS and 1405 MS patients. The overall quality of evidence was very low for all outcomes. Mortality and the incidence of stroke were similar between the 2 cohorts. Meta-analysis demonstrated increased length of cardiopulmonary bypass (CPB) time for patients undergoing MS [standardized mean difference 0.36, 95% confidence interval (CI) 0.15–0.58; P = 0.001]. Patients receiving MS spent more time in hospital (standardized mean difference 0.30, 95% CI 0.17–0.43; P < 0.001) and intensive care (standardized mean difference 0.17, 95% CI 0.06–0.27; P < 0.001). Reoperation for bleeding (risk ratio 1.51, 95% CI 1.06–2.17; P = 0.024) and renal impairment (risk ratio 1.97, 95% CI 1.12–3.46; P = 0.019) were also greater for MS patients. There was substantial heterogeneity in meta-analyses for CPB and aortic cross-clamp timing outcomes. MIS may be associated with improved early clinical outcomes compared to MS, but the quality of the evidence is very low. Randomized evidence is needed to confirm these findings.

INTRODUCTION

Median sternotomy (MS) is the gold-standard surgical approach for dealing with thoracic aortic pathology, offering excellent exposure for access to the aorta and central cannulation [1]. The technical complexity and steep learning curves associated with minimally invasive surgery (MIS) of the aorta act as barriers, hindering the widespread adoption of these methods. Nevertheless, the proposed reduction in postoperative pain and hospital stay, alongside improved cosmesis in minimally invasive aortic valve surgery [2, 3] makes MIS techniques attractive.

Well-established operations of the aortic root, such as the Bentall–De Bono [4] and valve-sparing root replacement (David) [5] procedures, can now be performed via much smaller incisions. Additionally, minimal access techniques have proven to be diverse approaches, allowing the surgeon to carry out isolated or concomitant procedures of the aortic arch [6, 7]. Numerous case series assessing MIS have found it to be safe in selected patients [8–10]. However, the paucity of comparative studies investigating MIS versus MS makes it difficult for surgeons to assess the true benefit of minimally invasive techniques in thoracic aorta surgery.

The aim of this study is to comprehensively review the current body of evidence comparing MIS of the aorta with analogous procedures performed via MS. We performed a systematic review and meta-analyses to evaluate if MIS for pathologies of the aorta is a safe and feasible alternative to the current approach in terms of its perioperative outcomes.

MATERIALS AND METHODS

The protocol for this review can be found on the PROSPERO website, registration number: CRD42018102726.

Selection criteria

Both randomized and observational studies of patients undergoing aortic root, ascending aorta or aortic arch surgery comparing minimal access versus a MS were eligible for inclusion. Minimal access was defined as any incision type other than MS [11]. Studies were excluded if they did not have a comparison group, if they included patients receiving isolated aortic valve or abdominal aortic procedures only or if more than 10% of study participants were emergency cases or had previous cardiac surgery. Studies performing concomitant procedures were included if the data for patients undergoing procedures of interest could be identified, or if 80% or more of the study patients underwent procedures of interest. No restriction was made on language or study size. Where multiple publications were available for the same cohort study, we used the data from the publication reporting the largest cohort and/or the most up-to-date results. To reduce the risk of publication bias, studies presenting only an abstract without a full text were included.

Primary outcomes were 30-day mortality, reoperation for bleeding, perioperative renal impairment and neurological events. Intraoperative and postoperative timing measures were also evaluated.

Literature search strategy

Electronic searches were performed using Ovid Medline, EMBASE, the Cochrane Library and the Web of Science from inception until July 2018. We combined the terms: (aorta or aortic or aortic root or aortic arch or ascending aorta) AND (surgical or surgeries or replacement or operation or ministernotomy or hemisternotomy or hemi-sternotomy or mini-sternotomy). All terms were searched as both text words and subject headings. The full search strategy is supplied in Supplementary Material, Appendix S1. To look for further relevant literature, we used the phrases ‘minimally invasive aortic surgery’, ‘minimally invasive aortic root/arch surgery’ and ‘minimally invasive ascending aorta surgery’ to search websites and journals of relevance such as CTSnet and Annals of Cardiothoracic Surgery. The reference lists of included studies were reviewed to identify further potentially relevant studies. An expert cardiothoracic surgeon (H.A.V.) was consulted regarding the existence of any unpublished material.

Data extraction and critical appraisal of evidence

Two reviewers (T.A.R. and P.R.) independently reviewed retrieved citations using Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). For all relevant records, full papers were retrieved and read in full by 2 reviewers independently (T.A.R. and P.R.). Discrepancies were resolved by consensus, and where necessary inclusion of a third reviewer (J.S.). Data extraction was completed by T.A.R. and checked by P.R.

Statistical analysis

We calculated the weighted arithmetic mean of patient baseline characteristics to look for differences between groups. For binary outcomes, we estimated the summary risk ratio (RR) and 95% confidence intervals (CIs) from the reported number of events and participants from eligible studies. For continuous outcomes, we anticipated substantial variation between studies in terms of methods, technique and operations performed, making the raw mean difference less valid in a meta-analysis [12]. We therefore estimated the standardized mean difference (SMD) and its standard error from the reported means, standard deviations (SD) and numbers of participants [13], which accounts for some of these differences. If medians and interquartile ranges (IQR) were presented, the median was substituted for the mean and the SD was estimated from the IQR [14] if we considered the distribution looked normal (i.e. the IQR was reasonably symmetrical about the median). Both fixed-effect and random-effects models were estimated and presented. Because of the technical differences in surgery of the aortic root and ascending aorta when compared to the aortic arch, we performed subgroup analysis and meta-regression for each outcome to assess if there was evidence of a difference between studies including and excluding arch procedures. The I² statistic was used to estimate the percentage variation in the average treatment effect due to differences between studies [15]. We considered a value greater than 50% to represent substantial heterogeneity, and we considered potential reasons for such variation. Small-study effect and publication bias were assessed using visual inspection of funnel plots [16]. P-values were 2-tailed. Stata Version 15.1 (StataCorp LLC) was used for all statistical analysis.

Assessment and evaluation of the quality of evidence

The risk of bias was assessed using the Risk of Bias in Non-Randomized Trials- of Interventions (ROBINS-I) tool [17]. ROBINS-I examines 7 domains of bias: confounding, selection bias, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in the measurement of outcomes and bias in the selection of the reported result. Studies are judged to be at ‘low’, ‘moderate’, ‘serious’ or ‘critical’ for risk of bias. Studies judged ‘critical’ were excluded from synthesis. The quality of evidence for each of the main outcomes was assessed using the GRADE scoring system [18], using GRADEpro software (available from www.gradepro.org).

RESULTS

Study selection and characteristics of included studies

Literature searches of online databases yielded 4430 citations and an additional 33 records were found from other sources. Of these, 143 relevant articles were read in full and assessed against the inclusion criteria, and 15 studies were included in the review [19–33]. After assessment of the risk of bias, 2 studies were rated as having critical risk of bias and were not included in further analysis [27, 29], thus leaving 13 studies for descriptive analysis. One further study was excluded from quantitative synthesis, as no useable data existed for either binary or continuous outcomes [21]. Twelve studies were included in the quantitative synthesis, comprising 1101 patients in the MIS and 1405 in the MS group. This information is shown in Fig. 1 [34].

Figure 1:

PRISMA flow chart of the search and study selection process.

Table 1 illustrates the characteristics of the included studies. Three studies were only reported in abstracts from posters and conferences [20, 21, 30]. Less than 100 patients were investigated in 3 included articles [21, 23, 31]. Only one study involved more than 500 participants [25]. Mean follow-up time was provided for only 4 studies [20, 28, 31, 33].

Table 1:

Characteristics of studies included in this systematic review and meta-analysis comparing minimally invasive surgery of the aorta with median sternotomy

First author and year [Ref no.]	Study period	Country, treatment centre	Study design	n (MIS)	n (MS)	MIS incision	Mean follow-up time (months)		Comment
							MIS	MS
Abjigitova et al. (2018) [19]	1998–2016	The Netherlands, Rotterdam	OC, RSP	26	91	‘J’ ministernotomy or ‘inverted T’ ministernotomy
Aharon et al. (2017) [20]	1998–2016	USA, Wynnewood, PA	OC, RSPa	26	199	Ministernotomy	22.3	158.3	Type of ministernotomy not defined
Burdett et al. (2014) [21]	2012–2013	UK, Middlesbrough	OC, RSPa	7	9	Ministernotomy			Type of ministernotomy not defined
Hastaoglu et al. (2018) [22]	2010–2015	Turkey, Istanbul	MC	54	75	‘J’ ministernotomy
Hillebrand et al. (2018) [23]	2012–2016	Germany, Münster	OC, RSP	33	25	‘J’ ministernotomy
Lamelas et al. (2018) [24]	2009–2014	USA, Houston, TX	PSM	74	103	MI right thoracotomy OR right lateral thoracotomy
Levack et al. (2017) [25]	1995–2014	USA, Cleveland, OH	PSM	568	1259	‘J’ ministernotomy
Mikus et al. (2017) [26]	2010–2015	Italy, Ravenna	OC, RSP	53	185	‘J’ ministernotomy
Monsefi et al. (2018) [27]	1991–2015	Germany, Frankfurt	OC, RSP	90	206	‘J’ ministernotomy	36 –24	9 6–48	Critical Risk of Bias
Monsefi et al. (2018) [28]	1991–2016	Germany, Frankfurt	PSM	120	207	‘J’ ministernotomy	36–24	96–48
Shrestha et al. (2015) [29]	2011–2014	Germany, Hannover	OC, RSP	26	14	‘J’ ministernotomy	40–27	41–26	Critical Risk of Bias
Shrestha et al. (2018) [30]	2011–2016	Germany, Hannover	OC, RSPa	210	192	‘J’ ministernotomy
Sun et al. (2000)b [31]	1999–1999	China, Beijing	OC, RSP	8	21	‘J’ ministernotomy	3	3
Tabata et al. (2007) [32]	1996–2005	USA, Boston, MA	MC	128	93	‘J’ ministernotomy
Wachter et al. (2017) [33]	2007–2012	Germany, Stuttgart	MC	117	75	‘J’ ministernotomy	31–18	31–18

^aAbstract.

^bThe authors stated that patients were followed-up for at least 3 months for both cohorts.

MC: matched cohort; MIS: minimally invasive surgery; MS: median sternotomy; OC: observational cohort, PSM: propensity score matched; RSP: retrospective.

Patient characteristics

The weighted means of patient baseline characteristics were similar between MIS and MS groups (Supplementary Material, Table S1): for example, age (57.6 vs 58.0 years), sex (72.6% vs 74.6% male), left ventricular ejection fraction (58.8% vs 58.1%), New York Heart Association functional class ≥3 (9.5% vs 11.2%), bicuspid aortic valve (58.1% vs 59.1%), hypertension (61.4% vs 63.9%), diabetes mellitus (7.2% vs 7.7%) and chronic obstructive pulmonary disease (7.1% vs 7.7%). The percentage of patients with aortic insufficiency grade ≥3 was higher in the MIS group (57.3% vs 48.2%), although this was reported by only 2 studies [28, 33]. One study included 3 (1.5%) patients requiring emergency procedures [28], all remaining studies only included elective procedures.

Interventions

The indication, procedure and concomitant procedures performed in the studies are summarized in Supplementary Material, Table S2. The indication for operation varied between studies for the MIS and MS cohorts, though 10 articles reported aortic dilatation or aneurysm as an indication [19, 20, 23–26, 28, 31–33]. Aortic root replacement was performed in 12 institutions [19–23, 25, 26, 28, 30–32] and ascending aorta replacement was performed in 6 centres [22, 24, 25, 28, 30, 32]. Four studies reported operations of the aortic arch [24, 28, 32, 33], with only one explicitly stating that they performed complete arch replacement [28]. There were differences in the proportion of patients in the MIS and MS cohorts receiving each primary aortic intervention in 7 studies [20, 22, 23, 25, 28, 30, 32]. The Bentall procedure was performed by 6 institutions [19, 20, 22, 26, 30, 32], and 8 institutions operated on the aortic valve concomitantly [22–25, 28, 30, 32, 33]. Other additional procedures were performed by 3 institutions [23, 28, 33] and included mitral valve surgery and coronary artery bypass grafting. The proportion of patients receiving each of these concomitant procedures was in general greater for the MS cohort in 2 studies [23, 33], whilst in 1 study, MIS patients were more likely to undergo additional surgery [28].

The ‘J’ ministernotomy to the third or fourth intercostal space was used in all but 1 study, instead opting for a right or right lateral thoracotomy [24]. One study also performed MIS through an ‘inverted-T’ ministernotomy [19]. The cannulation technique and strategies for myocardial protection varied widely between studies. They are presented in Supplementary Material, Table S3. Only one study fully described their cannulation technique for both MIS and MS cohorts [22].

Five studies commented that they gained experience with aortic surgery via MS prior to progressing to MIS [18, 22, 24, 26, 28]. Four studies stated that a single surgeon performed the procedures at their institution for both MS and MIS groups [21, 22, 24, 26]. In 1 study, 5 surgeons performed aortic surgery via MS, whilst only 2 of this 5 operated on the MIS group [28]. The remaining studies did not report issues related to the surgical learning curve.

Risk of bias in included studies

All included studies were non-randomized and their risk of bias is shown in Supplementary Material, Table S4. We judged 2 studies to be at critical risk of bias due to the presence of strong unadjusted confounding [27, 29]. Ten included studies were at ‘serious’ risk of bias [20–26, 28, 29, 32], mainly due to confounding, 1 was at ‘moderate’ risk of bias [33] and 1 study provided insufficient information to make a risk of bias judgement [29]. Three studies undertook propensity score-matched analyses [24, 25, 28] and 3 studies used matched-pair analysis to control for specific patient baseline characteristics [22, 32, 33].

Synthesis of evidence by outcome

The timing outcomes and the main clinical findings for the included studies are presented in Supplementary Material, Tables S5 and S6, respectively. Results of meta-analyses for perioperative mortality, reoperation for bleeding, renal impairment, stroke, aortic cross-clamp (AoX) time, cardiopulmonary bypass (CPB) time and length of intensive care unit (ICU) and hospital stay are presented in Table 2. The quality of the overall body of evidence was very low for all outcomes as defined by GRADE criteria [18].

Table 2:

Summary of perioperative characteristics and outcomes with quality of evidence assessment for analysed outcomes by the Grades of Recommendation, Assessment, Development and Evaluation Working Group Approach (GRADE)

Outcome	Quality of evidence for outcome (GRADE) With justification(s)	Number of studies	Number of patients in MIS	Events in MIS group (%)	Number of patients in MS	Events in MS group (%)	RR (95% CI)		P-value for overall effect		Heterogeneity
							Fixed	Random	Fixed	Random	I2 (%)	P-value
Major outcomes
Mortality	⊕○○○ 1, 3, 4	9	1039	0.67	1328	1.73	1.96 (0.81–4.76)	1.74 (0.70–4.37)	0.14	0.24	0.0	0.99
Reoperation for bleeding	⊕○○○ 1, 3, 4, 5	12	1168	4.07	1470	7.10	1.61 (1.13–2.29)	1.51 (1.06–2.17)	0.008	0.024	0.0	0.83
Renal impairment	⊕○○○ 1, 3, 4	7	899	1.56	1194	3.52	1.99 (1.13–3.51)	1.97 (1.12–3.46)	0.017	0.019	0.0	0.99
Stroke	⊕○○○ 1, 3, 4	4	875	1.49	857	1.52	1.06 (0.50–2.25)	1.06 (0.50–2.26)	0.89	0.89	0.0	1.0
Operative outcomes							SMD (95% CI)
							Fixed	Random
AoX time	⊕○○○ 1, 2, 3	11	955		1275		0.26 (0.17–0.34)	0.16 (−0.03 to 0.36)	<0.001	0.091	70.7	<0.001
CPB time	⊕○○○ 1, 2, 3	11	955		1275		0.36 (0.15–0.44)	0.36 (0.15–0.58)	<0.001	0.001	76.5	<0.001
Length of ICU stay	⊕○○○ 1, 3	8	805		952		0.15 (0.06–0.25)	0.17 (0.06–0.27)	<0.001	<0.001	7.2	0.37
Length of hospital stay	⊕○○○ 1, 3	7	684		831		0.31 (0.21–0.41)	0.30 (0.17–0.43)	<0.001	<0.001	16.5	0.30

Minimally invasive surgery of the aorta versus median sternotomy

Population or patient: patients undergoing minimally invasive surgery of the aorta

Setting: inpatient hospital setting

Interventions: all minimally invasive procedures of the aortic root/arch and ascending aorta

Comparator: median sternotomy

Quality of evidence.

⊕○○○ = Very Low; ⊕⊕○○ = Low; ⊕⊕⊕○ = Moderate; ⊕⊕⊕⊕ = High.

Limitation in design: 1. Potential risk of bias; 2. Heterogeneity—possibly not explained; 3. Small number of events and/or small sample size and/or small number of studies reporting outcome; 4. Wide confidence intervals for effect estimate suggestive of imprecision; 5. Suspicion of publication bias confirmed by funnel plot.

AoX: aortic cross-clamp; CI: confidence interval; CPB: cardiopulmonary bypass; ICU: intensive care unit; MIS: minimally invasive surgery; MS: median sternotomy; RR: risk ratio; SMD: standardized mean difference.

The reported use of packed red blood cells (pRBC) suggested a skewed distribution, invalidating the method of converting medians to means, making meta-analysis unfeasible.

Perioperative mortality

There were more observed postoperative deaths in the MS cohort; however, the number of events occurring across all 12 studies was low and thus, there was little evidence that rates of postoperative mortality differed between MIS and MS (RR 1.74, 95% CI 0.70–4.37; P = 0.24; Fig. 2). There was no evidence that mortality was influenced by the inclusion of arch procedures (P for difference = 0.772). There was no evidence of heterogeneity (I² = 0.0%, P = 0.99). The funnel plot demonstrated no visual asymmetry (Supplementary Material, Fig. S1).

Figure 2:

Early postoperative mortality in patients undergoing MIS of the aorta versus MS. Solid squares for each study represent the RR with the size proportional to the weights in meta-analysis. The horizontal lines denote the 95% CI. A RR of 1 (vertical black line) indicates no difference between MIS and MS. The uppermost diamond represents the fixed-effect model weighted RR. The bottommost diamond illustrates the random-effects weighted RR. The horizontal tips of the diamond are the CI for the overall effect estimate. CI: confidence interval; D+L: DerSimonian–Laird test; M-H: Mantel–Haenszel test; MIS: minimally invasive surgery; MS: median sternotomy; RR: risk ratio.

Reoperation for bleeding and use of blood products

Reoperation for bleeding occurred more commonly in MS patients (RR 1.51, 95% CI 1.06–2.17; P = 0.024; I² = 0.0, P = 0.83; Fig. 3). There was some evidence that reoperation was influenced by the inclusion of arch surgery (RR 2.00, 95% CI 1.01–3.93 for studies including arch surgery, RR 1.36, 95% CI 0.89–2.07 for studies excluding arch surgery, P for difference = 0.0368). The funnel plot for the reoperation outcome demonstrated asymmetry which is suggestive of small-study effect or publication bias [35, 36] (Fig. 4).

Figure 3:

The requirement to reoperate for bleeding in patients undergoing MIS of the aorta versus MS. Solid squares for each study represent the RR with the size proportional to the weights in meta-analysis. The horizontal lines denote the 95% CIs. A RR of 1 (vertical black line) indicates no difference between MIS and MS. The uppermost diamond represents the fixed-effect model weighted RR. The bottommost diamond illustrates the random-effects weighted RR. The horizontal tips of the diamond are the CI for the overall effect estimate. CI: confidence interval; D+L: DerSimonian–Laird test; M-H: Mantel–Haenszel test; MIS: minimally invasive surgery; MS: median sternotomy; RR: risk ratio.

Figure 4:

Funnel plot for the reoperation for bleeding outcome. Individual blue circles indicate studies included in the present study. The position of these circles along the horizontal axis represents the effect-estimate/RR. This is plotted against the SE of the log-RR which is an estimate of study precision. Asymmetry is suggestive of small-study or publication bias causing overestimation of the effect size in a meta-analysis. RR: risk ratio; SE: standard error.

A greater number of pRBC units were transfused in the MS compared with MIS cohort, in 8 of the 9 studies reporting this outcome [19, 22, 24, 26, 28, 31–33]. Mean number of units transfused across studies ranged from 1.3 to 6.7 units to 0.89 to 4.9 units for MS and MIS patients, respectively.

Renal impairment and neurological events

There was some evidence to suggest that postoperative renal impairment was greater in the MS cohort (RR 1.97, 95% CI 1.12–3.46; P = 0.019; I² = 0.0, P = 0.99; Supplementary Material, Fig. S2A). There was no evidence that renal impairment was influenced by the inclusion of arch procedures (P for difference = 0.836). The funnel plot for the renal impairment outcome appeared symmetrical (Supplementary Material, Fig. S2B).

Four studies reported perioperative stroke [25, 28, 30, 32], but there were few events and so, there was no evidence of a difference in the incidence of stroke for MIS versus MS patients (RR 1.06, 95% CI 0.50–2.26; P = 0.887; I² = 0.0, P = 1.0; Supplementary Material, Fig. S3A). There was no evidence that the incidence of stroke was influenced by the inclusion of arch procedures (P for difference = 0.951). The funnel plot appeared symmetrical for the stroke outcome (Supplementary Material, Fig. S3B). One study found postoperative delirium to be increased for MS patients [33].

Aortic cross-clamp and cardiopulmonary bypass time

Patients undergoing MS for their aortic pathology had longer AoX times (SMD 0.16, 95% CI −0.03 to 0.36; P = 0.091; I² = 70.7, P < 0.001; Supplementary Material, Fig. S4A). However, there was substantial heterogeneity between the studies and there was little evidence of difference between groups in the random-effects model. The funnel plot appeared symmetrical (Supplementary Material, Fig. S4B).

There was some evidence to suggest that patients in the MS cohort were subject to increased CPB time, but the heterogeneity between studies was substantial (SMD 0.36, 95% CI 0.15–0.58; P = 0.001; I² = 76.5, P = 0.001; Supplementary Material, Fig. S5A). No asymmetry was observed in the funnel plot for this outcome (Supplementary Material, Fig. S5B).

There was no evidence the inclusion of arch procedures influenced the AoX (P for difference = 0.614) or CPB time (P for difference = 0.849).

Length of intensive care unit and hospital stay

Patients undergoing MS spent more time in ICU (SMD 0.17, 95% CI 0.06–0.27; P < 0.001; I² = 7.2%, P = 0.37; Supplementary Material, Fig. S6A). There was no strong evidence of a difference in ICU length of stay with the inclusion of arch procedures (P for difference = 0.085). There was no evidence of asymmetry in the funnel plot (Supplementary Material, Fig. S6B).

The length of hospital stay was longer for the MS group (SMD 0.30, 95% CI 0.17–0.43; P < 0.001; I² = 16.5, P = 0.30; Supplementary Material, Fig. S7A). There was no evidence the inclusion of arch procedures influenced the hospital length of stay (P for difference = 0.753). The funnel plot was symmetrical (Supplementary Material, Fig. S7B).

DISCUSSION

To the best of our knowledge, the present study represents the first systematic review and meta-analysis comparing outcomes of all aortic surgery by MIS versus MS. The overall quality of the body of evidence was very low [18] for all outcomes; thus, all findings should be interpreted with caution. We found no significant difference in mortality between MIS and MS, although MIS was associated with reduced rates of reoperation for bleeding, renal impairment, ICU stay, hospital length of stay and CPB time. There was no significant difference in AoX time between patient groups. The incidence of stroke was low and meta-analysis did not demonstrate a difference between MIS and MS patients. Although meta-analysis was not possible, fewer pRBC units were transfused for MIS patients in all but 1 study that reported the outcome [23]. We found no strong evidence that the inclusion of arch procedures influenced all outcomes except reoperation for bleeding. Our review highlights that MIS of the aorta is a highly versatile approach that facilitates surgery of the aortic root, ascending aorta and aortic arch for a diversity of indications. Despite the limitations of the available evidence, our findings suggest that MIS of the aorta may be a feasible alternative to MS. Robust randomized studies are needed to support this conclusion.

The strengths of this systematic review include the comprehensive search to identify all available evidence and the rigorous methods of study selection, with 2 independent reviewers. Our systematic review was conducted according to the highest standards of review conduct [37]. We designed a comprehensive and sensitive search strategy, with input from 2 professional information scientists, to identify as many relevant studies as possible and reduce the risk of publication bias. We searched multiple electronic databases, additional relevant sources, and references of relevant studies were inspected for further studies. We did not impose date or language restrictions. Study selection was performed independently by 2 reviewers and data extraction was carried out by 1 reviewer and checked by another. We used the ROBINS-I [17] tool to assess the risk of bias in included observational studies, the most comprehensive tool for assessing risk of bias in non-randomized studies of interventions. We assessed the overall quality of the body of evidence according to GRADE recommendations and followed Cochrane recommendations for conducting meta-analyses [13].

The reduction in the CPB time for MIS patients in our review contradicts current trends in minimally invasive cardiac surgery [2, 38]. It is well established that prolonged time on CPB increases the risk of neurological [39] and perioperative renal impairment [40]. There was substantial heterogeneity in this meta-analysis, with the Levack study [25] contributing the most weight to the estimate. We could not identify specific study characteristics that could explain the observed heterogeneity in CPB times across studies. One possible explanation for this finding is that patients receiving MIS may have undergone procedures that demanded less time on CPB when compared to the MS group. Moreover, many of the institutions in the included studies gained sufficient experience of aortic surgery via MS before graduating to MIS. This would have the effect on minimizing the surgeon learning curve for performing MIS of the aorta. Therefore, surgeons with less experience of MIS may require longer CPB time than in the included studies of this review. However, it is noteworthy that most institutions opted for a ministernotomy. This incision enables the surgeon to visualize a similar operating field when compared to MS. Therefore, the difference in CPB time should not vary considerably for MIS of the aorta versus MS, and the clinical significance of any difference is probably minimal.

Our study also reports a reduction in the number of patients undergoing reoperation for bleeding in the MIS group. Reoperation keeps patients in hospital, and brings with it the risks of reopening the chest [41]. Minimally invasive cardiac surgery has been theorized to reduce bleeding, possibly due to reduced sternal trauma and instability. However, the visually asymmetrical funnel plot indicates the presence of small-study effect or publication bias, the latter of which would result in a favourable interpretation of the benefits of MIS on the rate of reoperation. Selective reporting and publication bias precludes accurate interpretation of the potential benefits of MIS and so, it is key that surgeons report all data regardless of the outcome in future studies. Meta-regression analysis suggested that reoperation rates might be lower in studies which included aortic arch surgery. Though interesting, the proportion of arch procedures was relatively low in the included studies, so this finding is likely to be related to other differences between studies.

Although we were unable to quantitatively analyse the transfused pRBC outcome, fewer pRBC units were transfused in the MIS cohort in 8 of the 9 studies reporting the outcome. This may reflect a tendency of surgeons to pay closer attention to haemostasis in MIS compared to MS, and the possibility that the threshold for giving blood products may have differed for MIS and MS patients. Nevertheless, these results provide some reassurance that MIS of the aorta does not lead to a greater quantity of blood transfusion, which has the potential for minimizing morbidity [42] and cost to health services.

There was some evidence that MIS was associated with a reduction in both ICU and hospital length of stay. This finding is consistent with the current literature for minimal access cardiac surgery [2, 38, 43]. Prolonged periods in ICU are associated with perioperative morbidity and mortality [44], and so minimizing this would be an important advantage of MIS of the aorta. Whether the result in our review occurred because of the effect of MIS rather than differences in postoperative care for MIS and MS patients requires consideration. All included studies reporting the length of hospital stay found the time in hospital to be shorter for MIS patients. This could be a consequence of attenuated postoperative pain, although the lack of data on this outcome does not allow us to make firm conclusions. Future studies should endeavour to report this very important outcome.

It is challenging to recommend a means of approaching MIS of the aorta given the marked variation in the way surgeons undertake these procedures (e.g. cannulation and myocardial protection). This is more frequently dictated by surgeon preference, given their experiences with similar procedures performed through MS. Surgeons contemplating utilizing MIS may wish to first gain sufficient experience with aortic surgery via MS before undertaking MIS. Shrestha et al. [29] performed more than 500 David procedures via a MS at their institution and more than 200 minimal access aortic valve replacements prior to undertaking MIS of the aorta. This enabled them to adequately develop a routine approach to these procedures which minimizes the challenge of converting to MIS of the aorta. Moreover, the authors initially selected low-risk patients with isolated aortic disease to undergo MIS. We therefore emphasize the need for prolonged experience with MIS of the aorta and careful patient selection in the early stages of a MIS programme.

Limitations

A limitation of the evidence included in our review is that it is based on single-centre, non-randomized studies which are vulnerable to confounding and other biases. There was heterogeneity in the CPB and AoX time that was not explained by the inclusion of arch procedures. Therefore, it is likely that this variation occurred due to other confounding variables such as differences in indication, type of surgery and the performance of concomitant procedures between studies. To mitigate the impact of concomitant procedures such as aortic valve surgery on the outcomes of MIS, further studies should aim to compare isolated aortic surgery for MIS versus MS. The overall quality of the body of evidence was very low for all outcomes, as defined by the GRADE criteria [18]. As only a few of the studies had long-term follow-up, we were unable to evaluate the differences in long-term aortic complications between the 2 approaches. Moreover, we were not able to assess important measures of patient satisfaction such as quality of life and time to return to work. These outcomes should be addressed in future studies to establish whether MIS of the aorta is of benefit to patients.

CONCLUSION

Very low-quality non-randomized evidence suggests that MIS of the aorta may be associated with improved early clinical outcomes when compared to MS. Randomized controlled trials are essential to confirm these findings.

Presented at the Annual Meeting of the Society for Cardiothoracic Surgery (SCTS), QEII centre, London, UK, 10–12 March 2019.