Abstract
Background and Aims
The current study proposes a novel volume–outcome (V–O) meta-analytical approach to determine the optimal annual hospital case volume threshold for cardiovascular interventions in need of centralization. This novel method is applied to surgery for acute type A aortic dissection (ATAAD) as an illustrative example.
Methods
A systematic search was applied to three electronic databases (1 January 2012 to 29 March 2023). The primary outcome was early mortality in relation to annual hospital case volume. Data were presented by volume quartiles (Qs). Restricted cubic splines were used to demonstrate the V–O relation, and the elbow method was applied to determine the optimal case volume. For clinical interpretation, numbers needed to treat (NNTs) were calculated.
Results
One hundred and forty studies were included, comprising 38 276 patients. A significant non-linear V–O effect was observed (P < .001), with a notable between-quartile difference in early mortality rate [10.3% (Q4) vs. 16.2% (Q1)]. The optimal annual case volume was determined at 38 cases/year [95% confidence interval (CI) 37–40 cases/year, NNT to save a life in a centre with the optimal volume vs. 10 cases/year = 21]. More pronounced between-quartile survival differences were observed for long-term survival [10-year survival (Q4) 69% vs. (Q1) 51%, P < .01, adjusted hazard ratio 0.83, 95% CI 0.75–0.91 per quartile, NNT to save a life in a high-volume (Q4) vs. low-volume centre (Q1) = 6].
Conclusions
Using this novel approach, the optimal hospital case volume threshold was statistically determined. Centralization of ATAAD care to high-volume centres may lead to improved outcomes. This method can be applied to various other cardiovascular procedures requiring centralization.
Keywords: Volume–outcome relationship, Annual case volume, Optimal case volume, Hospital volume, Cardiovascular interventions, Acute type A aortic dissection
Structured Graphical Abstract
Structured Graphical Abstract.
Overview of the process to determine the volume-outcome relation and the optimal annual case volume in surgery for acute type A aortic dissection. Grey line: first quartile, yellow line: second quartile, blue line: third quartile, and red line: fourth quartile. ATAAD, acute type A aortic dissection.
See the editorial comment for this article ‘Acute type A aortic dissection: stay and play or load and run?’, by T. Berger et al., https://doi.org/10.1093/eurheartj/ehad517.
Introduction
Centralization of health care services is one of the most important contemporary themes in health care policy1 and may be of particular importance to high-risk and less frequently performed cardiovascular procedures. In recent years, centralization efforts have been made for a wide range of procedures within the medical spectrum, such as various oncological,2 vascular,3 and coronary interventions.4 In this paradigm, high-risk procedures should be performed in centres of expertise, for which the main criterion is annual case volume.5 Seminal historical and contemporary studies have established the relation between volume and outcome.6–8 Consequently, specific hospital volume thresholds have been formulated for common procedures, which in order define a high-volume centre.9
Still, this concept principally applies to elective procedures, as the re-routing of patients to expert centres might result in delay of treatment in patients with acute and life-threatening conditions. Moreover, when procedures are relatively uncommon, it might be even more difficult to establish this volume–outcome (V–O) relationship. In the current study, a novel method is proposed to evaluate this association in such infrequent high-risk cardiovascular interventions. Furthermore, we will define an optimal annual case load, with surgery for acute type A aortic dissection (ATAAD) as an illustrative example.
Acute type A aortic dissection is a devastating medical condition with an emergent indication for surgery,10–12 for which some studies have suggested a beneficial survival effect of transfer to a high-volume centre.13 Indeed, outcomes after ATAAD surgery seem superior when the intervention is performed in these high-volume centres.14 However, given the relative infrequency of ATAAD procedures, it is complex to determine an optimal annual case volume threshold for this intervention.
The aim of the current study was to establish this V–O association and the optimal case volume threshold for ATAAD surgery using this V–O meta-analytical approach.
Methods
Design
The current study was designed as a systematic review and meta-analysis to evaluate the real-world V–Os of high-risk cardiovascular interventions, with ATAAD surgery as an exemplary procedure. The study was registered in PROSPERO15 [registration date 18 July 2022 (CRD42022345024)] and adhered to the 2020 Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA 2020) guidelines.16
Eligibility criteria
Studies published between 1 January 2012 and 29 March 2023 were eligible and at least needed to report on: (i) the number of consecutive ATAAD patients (both DeBakey Types I and II), (ii) years of inclusion, and (iii) the primary outcome (in-hospital and/or 30-day mortality). Particularly single-centre studies were eligible for inclusion, as this allowed for a reliable determination of the annual case volume for that institution. Multi-centre studies were only eligible when reporting the number of patients per centre, years of inclusion, and outcome separately per institution. Both comparative and single-arm studies were eligible for inclusion, as long as they reported the aforementioned features. To reduce potential publication bias, abstracts of meeting presentations were eligible as well.
Studies reporting only on subsets of patients (i.e. only root replacement or only arch replacement) were excluded as they did not accurately reflect the actual number of annual ATAAD case load. Furthermore, studies reporting exclusively on endovascular strategies, non-ATAAD acute aortic syndromes, and non-acute dissections (>14 days from symptom onset) were excluded as well. Finally, duplicate study inclusion of the same centre was avoided by only including the study describing the largest sample size of that specific centre.
Information sources
PubMed, EMBASE, and the Cochrane Library were systematically searched for eligible articles. Furthermore, the reference lists of included articles were screened for potentially missed articles (cross-referencing). The search was initiated from 1 January 2012, and the last search was performed 29 March 2023 for all databases (M.J.K. and S.H.).
Search strategy
The databases were queried using an elaborate reproducible search incorporating a combination of the terms ‘type A aortic dissection’, ‘surgery’, and ‘mortality’ and all possible alternative spellings. Please see Supplementary data online, Material S1 for the comprehensive search strategy.
Selection process
The selection process was performed by two authors independently (M.J.K. and S.H.). Studies were screened based on title and abstract for eligibility [the mere mention of (i) ATAAD as a disease, (ii) number of patients and years of inclusion, and (iii) mortality as an outcome, sufficed]. We used the web-based application Rayyan (http://rayyan.qcri.org) for this phase,17 after which eligible full-texts were evaluated for final inclusion. Any potential disagreement was resolved through discussion.
Data collection process and items
Data collection was performed manually by three authors (M.J.K., J.R.O., and S.H.), using a pre-defined worksheet (see Supplementary data online, Material S2).
Outcomes and effect measures
The primary outcome was early mortality (defined as 30-day or in-hospital mortality), in relation to annual hospital case volume, as assessed by restricted cubic splines (see Data synthesis). Secondary outcomes were as follows: (i) the incidence of stroke and (ii) long-term survival and their relation to annual hospital case load. Studies were divided into quartiles based on annual case volumes, similar to previous V–O studies.18,19 Annual case volume in relation to mortality was presented graphically in a figure [x-axis, cases/year; y-axis, percentage (%) with variance] and in tabular form (as compared with a procedural volume of 10 cases/year). For the latter presentation, absolute risk reduction (ARR), relative risk reduction (RRR), and numbers needed to treat (NNTs) to save a life in a higher- vs. lower-volume centre were calculated to facilitate clinical interpretation.
Adjusted relative long-term survival differences were presented in hazard ratios (HRs, lowest volume quartile as reference), with corresponding 95% confidence intervals (CIs).
Risk of bias assessment
To objectively assess study quality, the Newcastle–Ottawa scale was utilized. The assessed items comprise Selection (representativeness of the exposed cohort, selection of the non-exposed cohort, ascertainment of exposure, outcome of interest not present at the start of the study), Comparability (comparability of cohorts on the basis of design or analysis), and Outcome (assessment of outcome, long enough follow-up, adequacy of follow-up). Furthermore, this scale allows to be modified on a specific subject to accommodate the assessment of single-arm studies. This adaptation concerns removing one of four questions in the selection process (i.e. Question 2 in Selection—selection of the non-exposed cohort), as described previously.20,21 This quality assessment was performed by two authors (M.J.K. and J.R.O.).
Data synthesis
Data of studies presenting medians and inter-quartile ranges were converted to mean and standard deviation, using Wan’s method.22 Studies were divided into quartiles (Qs, Q1–Q4), based on the annual case volume of these studies for descriptive purposes. Study characteristics were presented per quartile in the main tables and per study individually in the Supplementary data online. Between-quartile comparisons of baseline and procedural characteristics (weighted aggregate data) were compared using the χ2 test (categorical) or the one-way analysis of variance test (continuous), to obtain P-values. Between-quartile comparisons of outcomes were performed in analyses of pooled data (inverse variance method, random-effects). Statistical heterogeneity among studies was objectively assessed using the I2-metric and τ2.
For the primary analyses, the potential non-linear relation between continuous annual case volume and outcome was investigated using a restricted cubic spline model for meta-analysis, using three knots, and presented graphically (non-linear mixed effects model). Weights were assigned per institution based on the variance of data (larger spheres indicate more weight due to less variance). The V–O relation was expressed in terms of mortality (%). This approach was previously applied on a patient-level18,19 and on a study-level in dose–response meta-analyses.23,24
The modelled median absolute mortality risk was derived from the restricted cubic spline analysis, to calculate the absolute mortality risk, after which we derived the ARR, RRR, and NNT, facilitating clinical interpretation. For this clinical risk assessment, a volume of 10 cases/year was used as a reference, and ARR, RRR, and NNT were calculated for multiplications of this annual case volume.
To correct for potential between-quartile baseline differences, crude and adjusted early mortality rates were presented. The continuous mortality risk outcome was adjusted for age, sex, median inclusion year of the study, and geographical continent, on a study-level in a linear regression model, as proposed previously.25 The rationale for these covariates is provided in Supplementary data online, Material S3. These Supplementary data online also provide detailed information on other variables, which could not be adjusted for.
The optimal annual case volume cut-off for early mortality was specified by determining the ‘elbow of the curve’. This method mathematically determines the point in a curve where it starts to plateau.26,27 In short, the mathematical principle lies within the connection of a straight line between the extremes of the curve, after which the maximal distance from the straight line to the curve is calculated. At this point, an increase in case volume does not result in a further decrease of the mortality rate (see Supplementary data online, Material S4 for a visual representation of the ‘elbow method’).
For the secondary outcome of long-term survival, a meta-analysis of reconstructed Kaplan–Meier (KM)-derived individual patient data (IPD) was performed, as proposed by Liu et al.,28 using KM curves from articles reporting on this outcome. The reconstructed KM curves were presented according to the pre-specified quartiles. Study-level adjustment for age, sex, median inclusion year, and geographical continent was again performed, now in a non-parametric frailty Cox model for hierarchical time-to-event data,29 for ad hoc sensitivity analysis (with additional adjustment for study/centre). Of note, the proportional hazard assumption was assessed visually and tested using Schoenfeld residuals. For all survival analyses, T0 was defined as the time of surgery. The per-quartile 10-year survival rate was derived from the KM curves. Again, the lowest volume quartile (Q1) was used as a reference, and ARR, RRR, and NNT were calculated for the other quartiles to facilitate clinical interpretation.
For all outcomes, we performed a complete case analysis. In general, imputation methods can be used when data are missing at random, and cases are >60% complete for a specific parameter.30 However, unfortunately, multiple parameters with missing values which could be adjusted for had <60% case completion (see Supplementary data online, Material S3). Furthermore, these cases were not missing at random.
A more detailed description of the statistical analyses can be found in Supplementary data online, Material S5. All statistical analyses were performed with R Statistics Version 3.6.0 (R foundation, Vienna, Austria), using ‘metafor’, ‘meta’, ‘rms’, ‘mgcv’, ‘survival’, ‘survminer’, ‘maps’, ‘ggplot2’, ‘pathviewr’ (for the elbow analysis), and ‘discfrail’ (frailty Cox model) packages. The R codes used for analysis were included in Supplementary data online, Material S5 as well and shared on GitHub, allowing for future reproduction of the proposed method (https://github.com/samuelheuts/R-codes-and-manual-for-analyses).
Publication bias assessment and adjustment
Publication bias assessment was performed for the primary outcome (early mortality). Bias assessment was presented graphically using funnel plots and assessed statistically by Egger’s test. Presence of significant publication bias was corrected with Duval and Tweedie’s trim-and-fill procedure and presented in funnel plots for the overall population to obtain the expected mortality (vs. the observed mortality). However, the current study’s data and results may be subjected to further publication bias, as centres with sub-optimal results are less inclined to publish their outcomes. This would inherently also affect the V–O relation. Therefore, publication bias was also assessed per quartile (Q1–Q4) and again corrected by the trim-and-fill procedure per quartile in a post hoc sensitivity analysis (mirroring non-reported studies in the funnel plot). The primary study’s volume was used for the mirrored study’s volume, and the non-reported primary outcome rate (mortality) was incorporated in the publication bias-adjusted V–O model to test the robustness of the results.
Results
Study selection
The systematic search yielded 14 882 studies, of which 10 154 remained after removing duplicate studies. Based on title and abstract, we additionally excluded 9629 studies, leaving 525 studies for full-text screening, of which 31 could not be retrieved (n = 410). Of these 494 remaining articles, 140 were included for final analysis (exclusion n = 354, please see PRISMA flowchart in Figure 1 for detailed information on the reasons for exclusion).
Figure 1.
Preferred Reporting Items for Systematic reviews and Meta-Analyses 2020 flowchart for study inclusion. ATAAD, acute type A aortic dissection
Study characteristics and quartiles
The 140 included studies, originating from 140 individual centres (5 continents, 26 countries), comprised 38 276 patients undergoing surgery for ATAAD. None of the studies had a randomized design. Sample sizes ranged from 28 to 1522 patients per study (Figure 2A), and average annual case volume ranged from 4.26 to 152.12/year (Figure 2B). Geographical distribution is presented in Figure 2C and Table 1. Individual (per study) institutional, study, and procedural characteristics are presented in Supplementary data online, Materials S6–S9.
Figure 2.
Study sample sizes and geographical distribution. (A) Sample size per study, (B) annual case volume per study, and (C) geographical distribution
Table 1.
Study, baseline, and dissection characteristics, divided into annual case volume quartiles for short-term mortality
| Annual case volume quartiles | ||||||
|---|---|---|---|---|---|---|
| Variables | Overall | Quartile 1 (1–12) | Quartile 2 (12–17) | Quartile 3 (17–29) | Quartile 4 (29–152) | P-value |
| (n = 38 276) | (n = 4 449) | (n = 6833) | (n = 10 195) | (n = 16 799) | ||
| Study characteristics | ||||||
| Number of studies | 140 | 35 | 35 | 35 | 35 | NA |
| Number of patients, per continent (%) | 38 276 (100) | 4449 (100) | 6833 (100) | 10 195 (100) | 16 799 (100) | NA |
| Europe | 12 500 (32.7) | 1185 (26.6) | 3500 (51.2) | 6483 (63.6) | 1332 (8.0) | |
| North America | 8083 (21.1) | 1231 (27.7) | 1928 (28.2) | 483 (4.7) | 4441 (26.4) | |
| Asia | 17 036 (44.5) | 1745 (39.2) | 1405 (20.6) | 2860 (28.1) | 11 026 (65.6) | |
| South America | 145 (0.4) | 145 (3.3) | NA | NA | NA | |
| Australia and Oceania | 512 (1.3) | 143 (3.2) | NA | 369 (3.6) | NA | |
| Patient and surgical characteristics | ||||||
| Age, in years | 59.7 ± 5.5 | 60.4 ± 4.8 | 61.3 ± 4.2 | 61.1 ± 4.9 | 55.7 ± 6.2 | <.01 |
| Sex | 36 501 (100) | 4271 (100) | 6558 (100) | 9589 (100) | 16 083 (100) | <.01 |
| Male | 24 113 (66.1) | 2727 (63.2) | 4205 (64.1) | 6176 (64.4) | 11 005 (68.4) | |
| Female | 12 388 (33.9) | 1544 (36.2) | 2353 (35.9) | 3413 (35.6) | 5078 (31.6) | |
| DeBakey classification | 12 771 (100) | 1248 (100) | 1686 (100) | 4827 (100) | 5010 (100) | <.01 |
| Type I | 10 528 (82.4) | 981 (78.6) | 1418 (84.1) | 3856 (79.9) | 4273 (85.3) | |
| Type II | 2243 (17.6) | 267 (21.4) | 268 (15.9) | 971 (20.1) | 737 (14.7) | |
| Connective tissue disorder | 21 219 (100) | 2360 (100) | 4025 (100) | 5925 (100) | 8909 (100) | <.01 |
| Yes | 858 (4.0) | 133 (5.6) | 236 (5.9) | 167 (2.8) | 322 (3.6) | |
| No | 20 361 (96.0) | 2227 (94.4) | 3789 (94.1) | 5758 (97.2) | 8587 (96.4) | |
| Aortic root surgery | 26 689 (100) | 3505 (100) | 4972 (100) | 7790 (100) | 10 422 (100) | <.01 |
| Yes | 6446 (24.2) | 748 (21.3) | 1165 (23.4) | 2094 (26.9) | 2439 (23.4) | |
| No | 20 243 (75.8) | 2757 (78.7) | 3807 (76.6) | 5696 (73.1) | 7983 (76.6) | |
| Aortic arch replacement | 27 435 (100) | 2706 (100) | 5395 (100) | 8921 (100) | 10 413 (100) | <.01 |
| Yes | 9342 (34.1) | 496 (18.3) | 1141 (21.1) | 2336 (26.2) | 5369 (51.6) | |
| No | 18 093 (65.9) | 2210 (81.7) | 4254 (78.9) | 6585 (73.8) | 5044 (48.4) | |
| CPB, in minutes | 201.4 ± 39.0 | 207.1 ± 44.2 | 192.5 ± 33.8 | 207.9 ± 40.7 | 199.5 ± 37.8 | .46 |
| Cross-clamp time, in minutes | 117.1 ± 29.1 | 128.4 ± 27.1 | 108.2 ± 21.3 | 117.0 ± 35.3 | 118.9 ± 29.9 | .15 |
| CA, in minutes | 36.9 ± 14.7 | 41.4 ± 17.4 | 35.6 ± 13.3 | 38.1 ± 15.6 | 33.4 ± 12.2 | .35 |
| Short-term outcomes | ||||||
| Early mortality (%, 95% CI) | 13.6% (12.5%–14.7%) | 16.2% (14.0%–18.7%) | 14.2% (12.3%–16.4%) | 14.7% (12.5%–17.2%) | 10.3% (8.9%–11.8%) | NA |
| Early stroke (%, 95% CI) | 10.1% (9.1%–11.2%) | 10.6% (9.1%–12.3%) | 9.8% (8.1%–11.7%) | 12.3% (10.3%–14.7%) | 8.1% (6.2%–10.5%) | NA |
Bold values indicate statistical significance.
CA, circulatory arrest; CI, confidence interval; CPB, cardiopulmonary bypass; NA, not applicable.
Studies were divided into quartiles, based on annual case volume (1–12 cases/year: 35 studies, 12–17 cases/year: 35 studies, 17–29 cases/year: 35 studies, and 29–152 cases/year: 35 studies, respectively), as presented in Table 1.
Of note, elective aortic case volume and number of surgeons performing the ATAAD procedures were incompletely reported, and these parameters could therefore not be included in the analysis (merely reported by 12 and 27 studies, respectively, see Supplementary data online, Material S3).
Study quality assessment
Overall study quality ranged from low- to a high-risk of bias (low risk n = 86, intermediate risk n = 9, and high-risk n = 45), which is presented per individual study in Supplementary data online, Material S10. Although study quality differed, all studies were considered relevant for the primary outcome analysis, which was uniformly reported. Therefore, all studies were included for quantitative analysis.
Patient and procedural characteristics
Of the included patients, 66.1% (95% CI 65.6%–66.6%) were males, with a mean age of 59.7 years (±5.5 years). DeBakey Type I dissection was present in 82.4% of patients (95% CI 81.8%–83.1%). Patients in Q4 were significantly younger (60.4 ± 4.8 vs. 61.3 ± 4.2 vs. 61.1 ± 4.9 vs. 55.7 ± 6.2 years for Q1, Q2, Q3, and Q4, respectively, P < .01). Significantly more patients underwent extensive surgery (i.e. root replacement or arch surgery) in the higher volume quartiles (23% and 52% in Q4 vs. 21% and 18% in Q1, respectively, P < .01). Other dissection and operative variables, specified per quartile, can be found in Table 1. Unfortunately, time from symptom onset to surgery was incompletely reported and could therefore not be included in the analysis (reported by only 22 studies).
Early mortality
Early mortality differed significantly between the four quartiles [reported by all studies (140), n = 38 276 patients], with mortality being lowest in the quartile with the highest annual case volume [10.3% (95% CI 8.9%–11.8%, Q4) vs. 16.2% (95% CI 14.0%–18.7%, Q1), I2 = 87%, τ2 = 0.25; Table 1]. The early mortality rate was then risk-adjusted for available covariates (rationale can be found in Supplementary data online, Material S3). These covariates comprised age, sex, median inclusion year, and geographical continent. Modelled median institutional mortality was 13.3% (95% CI 11.8%–14.6%), corresponding to a modelled annual hospital volume of 24 cases/year [Table 2; Figure 3A (crude) and B (adjusted)]. Table 2 presents clinically intuitive numbers, including ARR, RRR, and NNT.
Table 2.
Volume–outcome relationship for fixed annual case volume (in steps of 10 cases/year, 10 cases/year as reference)
| Mean mortality retrieved from RCS for volume–outcome relation (Figure 3A) | Clinical estimates, as compared with an annual case volume of 10 cases/year | |||
|---|---|---|---|---|
| Annual case volume (cases/year) | Mean mortality (%, 95% CI) | ARR (%-points, 95% CI) | RRR (%, 95% CI) | NNT (95% CI) |
| 10 | 16.4 (14.7–18.2) | Ref. | Ref. | Ref. |
| 20 | 14.2 (13.0–15.3) | 2.2% (1.1%–3.4%) | 13.4% (6.7%–20.7%) | 45 (29–91) |
| 24 | 13.3 (11.8–14.6) | 3.1% (1.8%–4.6%) | 18.9% (11.0%–28.0%) | 32 (22–56) |
| 30 | 12.4 (10.7–14.1) | 4.0% (2.3%–5.7%) | 24.4% (14.0%–34.8%) | 25 (18–43) |
| 38 | 11.7 (9.8–13.5) | 4.7% (2.9%–6.6%) | 28.7% (17.7%–40.2%) | 21 (15–34) |
| 40 | 11.6 (9.7–13.4) | 4.8% (3.0%–6.7%) | 29.3% (18.3%–40.9%) | 21 (15–33) |
| 50 | 11.2 (9.4–13.0) | 5.2% (3.4%–7.0%) | 31.7% (20.7%–42.7%) | 19 (14–29) |
The annual volume of 24 cases/year contained the median modelled mortality (13.3%), 38 cases/year was the optimal case load at which mortality started plateauing (both in bold wording).
ARR, absolute risk reduction; CI, confidence interval; NNT, number needed to treat (to save a life) as compared with 10 cases/year; RCS, restricted cubic splines; Ref., reference; RRR, relative risk reduction.
Figure 3.
Volume–outcome relationship for crude early mortality (A), adjusted early mortality (B), covariates adjusted for: age, sex, median inclusion year, continent, and a visual representation of the elbow method to determine the optimal annual case volume for crude mortality (C), and for adjusted mortality (D). For a more in-depth elaboration on the elbow method, we refer to the references26,27 and Supplementary data online, Material S4. Of note, the percentual adjusted mortality rates (B and D) cannot be used for analyses of absolute and relative risk differences, or the calculation for numbers needed to treat, as these rates are not absolute
A significant non-linear association was found between annual hospital case volume and mortality (140 studies, n = 38 276 patients, P < .001, I2 = 74%, implying moderate to substantial heterogeneity, τ2 = 27.8), which is presented in Figure 3A and B. By application of the elbow method, the volume after which early mortality started plateauing (i.e. the optimal annual case volume) was determined at 38 cases/year (95% CI 37–40 cases/year, mean modelled mortality 11.7%, 95% CI 9.8%–13.5%, Figure 3C and D; Table 2). Figure 3C and D graphically depict the mathematical equation to retrieve the optimal annual case volume threshold, based on the elbow method. Of note, both the model for crude and risk-adjusted mortality provided a similar optimal annual case volume, confirming the robustness of the results. In the scenario of 38 cases/year, ARR is 4.7%-points, RRR is 28.7%, and the NNT to save a life is 21 patients, implying a life is saved per every 21 patients transferred to a centre with the optimal case volume (as compared with an annual case volume of 10 cases/year, Table 2).
Post hoc sensitivity analysis with exclusion of centres performing >50 cases annually was performed, providing similar results (122 studies, n = 29 412 patients, I2= 74%, τ2 = 27.8; Supplementary data online, Material S11). In addition to the geographical adjustment (continent) for early mortality, we also performed a subgroup analysis per continent for crude early mortality in the restricted cubic spline model. However, these analyses may lack statistical power because of the reduction in centres per analysis (see Supplementary data online, Material S12), and their results should be interpreted with caution.
Stroke
The overall stroke rate was 10.6% (101 studies, n = 3 054 patients, I2 = 87%, implying substantial heterogeneity, τ2 = 0.28). Peri-operative stroke differed significantly between the volume quartiles as well [Q4 8.1% (95% CI 6.2%–10.5%) vs. Q1 10.6% (95% CI 9.1%–12.3%), Table 1]. Again, a significant non-linear relationship between stroke and annual case volume was observed (P < .001, I2 = 78%, τ2 = 21.1; Supplementary data online, Material S13).
Long-term survival
Long-term survival was evaluated by KM-derived IPD. Eventually, 52 studies reported long-term survival with surgery as the starting point, by KM curves (n = 14 878 patients). Patient characteristics per quartile of these 52 studies are presented in Table 3 [individual study characteristics are presented in Supplementary data online, Material S7 (the 52 studies reporting on long-term survival were highlighted in the last column)]. Crude 10-year survival of the overall cohort was 60% (59%–62%). In Q4, 10-year survival was 69% (95% CI 66%–71%), compared with 51% (95% CI 48%–55%), in Q1 (P < .001 for Q4 vs. all quartiles, Figure 4). The crude ARR was therefore 18%-points for Q4 compared with Q1, implying an NNT of six transferred patients to a high-volume centre (Q4), to preserve a life at 10-year follow-up (Table 4). The proportional hazards assumption was met (Schoenfeld residuals P = .304).
Table 3.
Study, baseline, and dissection characteristics, divided into annual case volume quartiles for long-term survival
| Annual case volume quartiles | ||||||
|---|---|---|---|---|---|---|
| Variables | Overall | Quartile 1 (0–12) | Quartile 2 (12–17) | Quartile 3 (17–29) | Quartile 4 (29–152) | P-value |
| (n = 14 878) | (n = 1603) | (n = 3113) | (n = 3951) | (n = 6211) | ||
| Study characteristics | ||||||
| Number of studies | 52 | 13 | 14 | 14 | 11 | NA |
| Number of patients, per continent (%) | 14 878 (100) | 1603 (100) | 3113 (100) | 3951 (100) | 6211 (100) | NA |
| Europe | 4754 (32.0) | 514 (32.1) | 1720 (55.2) | 2043 (51.7) | 477 (7.7) | |
| North America | 5494 (36.9) | 602 (37.6) | 893 (28.7) | 678 (17.2) | 3321 (53.4) | |
| Asia | 4400 (29.6) | 257 (16.0) | 500 (16.1) | 1230 (31.1) | 2413 (38.9) | |
| South America | 87 (0.5) | 87 (5.4) | NA | NA | NA | |
| Australia and Oceania | 143 (1.0) | 143 (8.9) | NA | NA | NA | |
| Patient and surgical characteristics | ||||||
| Age, in years | 60.1 ± 4.8 | 58.9 ± 4.7 | 61.9 ± 4.6 | 60.9 ± 4.6 | 58.6 ± 5.1 | .26 |
| Sex | 14 559 (100) | 1603 (100) | 3113 (100) | 3951 (100) | 5892 (100) | <.01 |
| Male | 9441 (64.8) | 1077 (67.2) | 2016 (64.8) | 2466 (62.4) | 3882 (65.9) | |
| Female | 5118 (35.2) | 526 (32.8) | 1097 (35.2) | 1485 (37.6) | 2010 (34.1) | |
| DeBakey classification | 5311 (100) | 328 (100) | 743 (100) | 2481 (100) | 1759 (100) | <.01 |
| Type I | 4265 (80.3) | 254 (77.4) | 558 (75.1) | 2147 (86.5) | 1306 (74.2) | |
| Type II | 1046 (19.7) | 74 (22.6) | 185 (24.9) | 334 (13.5) | 453 (25.8) | |
| Connective tissue disorder | 10 176 (100) | 1115 (100) | 1827 (100) | 2838 (100) | 4396 (100) | <.01 |
| Yes | 388 (3.8) | 76 (6.8) | 97 (5.3) | 100 (3.5) | 115 (2.6) | |
| No | 9788 (96.2) | 1039 (93.2) | 1730 (94.7) | 2738 (96.5) | 4281 (97.4) | |
| Aortic root surgery | 12 055 (100) | 1358 (100) | 2444 (100) | 3461 (100) | 4792 (100) | <.01 |
| Yes | 2939 (24.4) | 336 (24.7) | 658 (26.9) | 837 (24.2) | 1108 (23.1) | |
| No | 9116 (75.6) | 1022 (75.3) | 1786 (73.1) | 2624 (75.8) | 3684 (76.9) | |
| Aortic arch replacement | 12 154 (100) | 682 (100) | 2973 (100) | 3901 (100) | 4598 (100) | <.01 |
| Yes | 3236 (26.6) | 71 (10.4) | 575 (19.3) | 971 (24.9) | 1619 (35.2) | |
| No | 8918 (73.4) | 611 (89.6) | 2398 (80.7) | 2930 (75.1) | 2979 (64.8) | |
| CPB time, in minutes | 199.1 ± 30.7 | 194.8 ± 33.7 | 188.6 ± 28.8 | 207.1 ± 26.1 | 206.6 ± 35.5 | .39 |
| Cross clamp time, in minutes | 115.6 ± 24.6 | 121.1 ± 25.9 | 109.8 ± 16.3 | 106.6 ± 21.4 | 134.3 ± 32.2 | .10 |
| CA time, in minutes | 35.4 ± 12.7 | 40.0 ± 18.5 | 34.7 ± 11.1 | 31.7 ± 11.4 | 39.0 ± 13.2 | .51 |
| Long-term outcomes | ||||||
| 1-year survival, (95% CI) | 82 (82–83) | 80 (78–82) | 80 (78–81) | 82 (81–83) | 85 (84–86) | <.01 |
| 3-year survival, (95% CI) | 79 (78–80) | 75 (73–77) | 75 (74–77) | 78 (77–80) | 82 (81–83) | <.01 |
| 5-year survival, (95% CI) | 74 (74–75) | 69 (66–71) | 70 (69–72) | 74 (72–76) | 78 (77–79) | <.01 |
| 10-year survival, (95% CI) | 60 (59–62) | 51 (48–55) | 57 (55–59) | 58 (55–60) | 69 (66–71) | <.01 |
Bold values indicate statistical significance.
CA, circulatory arrest; CPB, cardiopulmonary bypass; IQR, inter-quartile range; NA, not applicable.
Figure 4.
Long-term survival based on individualized patient data Kaplan–Meier meta-analysis. Shading indicates the 95% confidence interval
Table 4.
Volume–outcome relationship for long-term survival, per quartile (cases/year, Q1 as reference for absolute risk reduction, relative risk reduction, and number needed to treat)
| Annual case volume quartile (cases/year) | Median 10-year survival (%, 95% CI) | ARR (%-points, 95% CI) | RRR (%, 95% CI) | NNT (95% CI) |
|---|---|---|---|---|
| Q1 (0–12) | 51% (48%–55%) | Ref. | Ref. | Ref. |
| Q2 (12–17) | 57% (55%–59%) | 6% (4%–8%) | 12% (8%–16%) | 17 (13–25) |
| Q3 (17–29) | 58% (55%–60%) | 7% (4%–9%) | 14% (8%–18%) | 14 (11–25) |
| Q4 (29–152) | 69% (66%–71%) | 18% (15%–20%) | 35% (29%–39%) | 6 (5–7) |
ARR, absolute risk reduction; CI, confidence interval; NNT, number needed to treat (to save a life); Q, quartile; Ref., reference; RRR, relative risk reduction.
These long-term survival differences were also confirmed in a frailty Cox-model analysis with adjustment for center, age, sex, median inclusion year, and geographical continent (HR annual case volume per quartile 0.83, 95% CI 0.75–0.91, P < .001; Supplementary data online, Material S14).
Publication bias and adjustment
Publication bias assessment revealed the presence of significant publication bias for early mortality (Egger’s test: P = .002). The funnel plot showed an asymmetry suggesting an under-reporting of studies with greater mortality rates (see Supplementary data online, Material S15A). Trim-and-fill analysis revealed a higher actual mortality rate (16.2%) as compared with the mortality data obtained from the literature, for the overall cohort (see Supplementary data online, Material S15B).
As publication bias may be the most important limitation of the proposed approach, we performed additional per-quartile publication bias adjustment, for the primary outcome in the restricted cubic spline analysis. First, the trim-and-fill procedure was applied to the four quartiles separately and revealed non-reporting of studies with higher mortality rates (see Supplementary data online, Material S16). These non-reported studies were then incorporated in the restricted cubic spline analysis (see Supplementary data online, Material S17—publication bias-adjusted V–O relation) and consistently found a significant non-linear relation (P = .013) between annual case volume and early mortality. Furthermore, the robustness of the optimal annual case volume using the elbow method was confirmed in this post hoc sensitivity analysis as well (38 cases/year, 95% CI 37–50 cases/year).
Discussion
In the current study, we propose a novel methodology to evaluate the V–O relationship for infrequent high-risk cardiovascular interventions, with surgery for ATAAD as an illustrative example. By this method, we established a significant relation between annual hospital case volume and both early mortality and long-term survival, in conjunction with an optimal annual case load (Structured Graphical Abstract).
Implications
Centralization of care is a challenging process, particularly for highly specialized cardiovascular interventions. Currently, annual case load serves as the primary criterion for centralization,4,6 but it remains difficult to determine actual annual case volume thresholds for centres of expertise. Because of the incentive to concentrate care for these procedures, reproducible and objective methods are warranted to formulate such thresholds.
In emergency surgery for ATAAD, it is known that high-volume centres perform better in terms of mortality as compared with low-volume centres.14,31–33 Previous studies aiming to determine this V–O effect were confined to the categorical assessment of case volume (i.e. in tertiles or quartiles).14,32,33 However, in a categorical interpretation of case volume, a potential (non-linear) V–O relation might be missed, especially beyond the highest tertile/quartile.
The approach proposed in the current study circumvents some of the previous limitations. We performed a comprehensive literature search without geographical exclusion criteria, resulting in a real-world representation of ATAAD outcome assessment, with an in-depth analysis of the non-linear relation between volume and outcome. Our findings suggest that mortality and survival are improved in higher volume centres. Indeed, as can be appreciated in Figure 3, this non-linear relation confirms an important effect of annual case volume on outcome, and this is—intuitively—not infinite. In fact, after a certain number of procedures, the V–O effect is likely to plateau, implying accrual of more cases does not necessarily result in even better results. This ‘optimal’ volume was determined at 38 cases/year, but the steepest decrease in mortality was observed between the volume of 1 and 20 procedures/year. Moreover, for long-term survival, each consecutive quartile outperformed the previous quartile, with the fourth quartile of >29 cases/year performing the best (Figure 4).
Long-term survival is arguably the most important outcome in health-care but requires extensive and adequate follow-up in registries, which is often missing. By incorporating KM-derived IPD in the proposed concept, the V–O effect was assessed for long-term survival as well. Survival after surgery for ATAAD is affected by distal aortic events and false lumen patency.34 Consequently, more aggressive surgery during the ATAAD index procedure is associated with improved 10-year survival, as demonstrated in recent meta-analyses.35,36 In our current study, a continuous diversion of the KM volume curves was revealed, well beyond the first years. These findings imply a persistent effect of the V–O effect in the long term, which may be the result of a reduction in aortic events and re-interventions.35 The robustness of these findings was confirmed in multiple sensitivity analyses using risk-adjusted mortality and risk-adjusted survival, including a frailty Cox model.
Considerations
The relative and absolute survival differences derived from the annual case volumes, together with the NNTs to save a life at 30 days and 10 years (21 and 6, respectively), seem clinically meaningful. However, our findings should not be interpreted as a dichotomous conclusion that surgery for ATAAD should only be performed in centres performing more than 38 annual cases, and centralization should not solely depend on case volume. Several factors inherent to the treatment of ATAAD should be taken into account, in addition to a holistic consideration of the healthcare system. These considerations are outlined below.
First, in ATAAD, the patient’s presenting status is a major determinant of outcome.37 This status includes the presence of malperfusion or cardiac tamponade, amongst others.38 Although some studies have suggested that the primary resolution of malperfusion,39 or tamponade,40 could precede a surgical central aortic repair (after which a transfer to a high-volume centre can be initiated), this may not always be feasible.
Second, it may be overly simplistic to state that the mere accrual of more than 38 procedures will guarantee optimal results. Indeed, a centre of expertise does not only dispose of a high case load, but it is also dedicated to overall ATAAD care.41 Several studies have underlined the importance of a streamlined protocol with a focus on the rapid recognition and diagnosis of ATAAD.41,42 This also extends to the post-operative course and follow-up, incorporating imaging monitoring to screen for late ATAAD-related complications.43
Third, the transfer of a patient to a high-volume centre comes at the cost of treatment delay. Still, Goldstone et al.13 did not observe a negative effect of the inter-facility transport of ATAAD patients to high-volume centres, with a median travel distance of 80 km. Of note, similar outcomes were recently also observed in a Swiss cohort of re-routed patients with ruptured abdominal aneurysms.44 The transfer of ATAAD patients to a high-volume centre was also incorporated in the 2022 American guidelines for aortic disease (IIa recommendation, when patients are relatively stable).45
Fourth, cost-effectiveness of centralization should be taken into account, which includes hospital staffing and availability of dedicated aortic surgeons. Although data on the cost-effectiveness of centralization of elective surgery and care for patients with acute coronary syndromes exist,46,47 this is yet to be proven for an emergency surgical procedure like ATAAD.
Fifth, individual surgeon volume will play an increasingly important role, with the advent of specialized aortic teams. Indeed, while Bashir et al.48 found an improvement in risk-adjusted mortality when surgeons performed >4.5 cases/year, operative ATAAD mortality fell to 2.8% after the implementation of a thoracic aortic surgical programme in a single institution, as demonstrated by Andersen et al.49 Of note, the formation of such teams may also facilitate centralization to multi-centre collaborations, to guarantee the accrual of a sufficient case volume for these dedicated surgeons.
Sixth, based on our findings, only a few European and American centres currently operate >38 ATAAD patients annually. Ultimately, the decision on where to place a threshold should incorporate all key stakeholders. As centralization may lead to increased spatial inequality and socio-economic disparity,50 these centres should be appointed with care, taking geographic and socio-economic factors into consideration.
Nevertheless, the presented V–O meta-analytical approach provides an evaluation of the V–O effect, as outlined here, and facilitates the objective estimation of such volume thresholds. In turn, this facilitates health care policy-makers in the important decision-making process to concentrate care, ultimately improving the quality of care for patients with ATAAD.51
Future directions
Based on the contemporary dilemma of health care service centralization, the question remains how our results should be interpreted, and—potentially—implemented. Taking the above-presented considerations into account, we envision that our findings will encourage centres performing ATAAD surgery to co-operate more closely and potentially regionalize their care.
Furthermore, we believe that the proposed methodology has specific value for the evaluation of high-risk and less common cardiovascular interventions, as volume thresholds are currently arbitrarily formulated. Procedures that may be evaluated by our approach comprise complex coronary interventions, transcatheter valve interventions, minimally invasive cardiac surgical procedures, or use of temporary mechanical circulatory support, for example. To allow for reproduction, the R codes and statistical methodology were made transparent and available (through https://github.com/samuelheuts/R-codes-and-manual-for-analyses).
Limitations
The most important potential limitation of this methodology is the risk of publication bias, as higher volume centres are more inclined to publish their results, in contrast to centres with sub-optimal ATAAD outcomes. However, we made every effort to reduce the presence of publication bias, by the inclusion of published studies and abstracts of meeting presentations as well. Furthermore, we applied the trim-and-fill procedure to the quartiles separately and retrieved the annual case volume (by mirroring) and early mortality rate of ‘non-reported’ studies. The incorporation of these non-reported findings into the restricted cubic spline model, together with the elbow analysis, confirmed the robustness of our results. However, the trim-and-fill procedure is an imperfect method, and residual publication bias may still be present. Of note, due to the extensiveness of the analyses, publication bias was only assessed for the primary outcome (and not for stroke or long-term survival).
Although we found a statistically significant and clinically relevant association, hospital volume might not be the strongest volume predictor in ATAAD surgery, as individual surgeon case volume is also closely related to outcome.48,52 Regrettably, given the scarcity of studies on individual surgeon ATAAD volume, this parameter could not be included in the analysis. Furthermore, annual elective aortic surgical volume might also be reflective of the V–O relation in ATAAD surgery13 but was reported in the minority of studies.
Unfortunately, disease severity, which may be expressed by the extent of malperfusion,12,53 was only provided in a minority of studies as well and could therefore not be incorporated in the risk analysis (Supplementary data online, Material S3 describes the missingness of data). Inherently, these factors determine the urgency of surgery, and a transfer to a high-volume may therefore not always be feasible. In extent of this limitation, time of symptom onset to eventual surgery could not be included in the analysis either, due to sub-optimal reporting.
From a statistical point of view, heterogeneity seemed to be moderately to substantially present (I2 = 74%), but this was expected as I2 is an expression of the proportion of observed variance. In fact, the presence of statistical heterogeneity in outcomes between studies may be a reflection of the V–O effect. Therefore, we believe the presence of heterogeneity should not be regarded as a limitation of this methodology.
Based on the geographical distribution, as can be appreciated in Figure 2, high-volume centres are more frequently located in the East Asian region. As some reports have suggested a more favourable anatomy and risk profile of the Asian population,54 the superior results of high-volume centres might therefore not be exclusively related to case volume. Furthermore, the geographical distribution of centres, with a minority of centres in Europe and North-America performing more surgeries than the optimal case volume, can be considered a weakness of the study. Still, we have attempted to mitigate for both limitations by adjusting our analyses for geographical continent and by performing a sensitivity analysis with exclusion of centres performing >50 cases/year, providing similar results.
Conclusions
The current study proposes a novel methodological approach to evaluate the V–O relation for high-risk cardiovascular interventions, with ATAAD surgery as an illustrative example. In ATAAD, the highest volume centres exhibit superior results, in terms of early mortality and long-term survival. Despite the acute character of this disease, re-routing of patients and centralization of ATAAD care to high-volume centres may lead to improved results, but this should be interpreted in the light of important aspects such as presenting status, hospital staffing, infrastructure, and cost-effectiveness. The proposed method is not confined to ATAAD surgery and may be applied to various other cardiovascular procedures requiring centralization.
Supplementary Material
Contributor Information
Michal J Kawczynski, Department of Cardiothoracic Surgery, Maastricht University Medical Center (MUMC+), P. Debyelaan 25, 6629HX Maastricht, Limburg, The Netherlands; Cardiovascular Research Institute Maastricht (CARIM), Maastricht University, Universiteitssingel 50, Maastricht, PO Box 616, 6200 MD, Maastricht, The Netherlands.
Sander M J van Kuijk, Department of Clinical Epidemiology and Medical Technology Assessment, Maastricht University Medical Center (MUMC+), Maastricht, The Netherlands.
Jules R Olsthoorn, Department of Cardiothoracic Surgery, Maastricht University Medical Center (MUMC+), P. Debyelaan 25, 6629HX Maastricht, Limburg, The Netherlands; Department of Cardiothoracic Surgery, Catharina Hospital Eindhoven, Eindhoven, The Netherlands.
Jos G Maessen, Department of Cardiothoracic Surgery, Maastricht University Medical Center (MUMC+), P. Debyelaan 25, 6629HX Maastricht, Limburg, The Netherlands; Cardiovascular Research Institute Maastricht (CARIM), Maastricht University, Universiteitssingel 50, Maastricht, PO Box 616, 6200 MD, Maastricht, The Netherlands.
Suzanne Kats, Department of Cardiothoracic Surgery, Maastricht University Medical Center (MUMC+), P. Debyelaan 25, 6629HX Maastricht, Limburg, The Netherlands.
Elham Bidar, Department of Cardiothoracic Surgery, Maastricht University Medical Center (MUMC+), P. Debyelaan 25, 6629HX Maastricht, Limburg, The Netherlands; Cardiovascular Research Institute Maastricht (CARIM), Maastricht University, Universiteitssingel 50, Maastricht, PO Box 616, 6200 MD, Maastricht, The Netherlands.
Samuel Heuts, Department of Cardiothoracic Surgery, Maastricht University Medical Center (MUMC+), P. Debyelaan 25, 6629HX Maastricht, Limburg, The Netherlands; Cardiovascular Research Institute Maastricht (CARIM), Maastricht University, Universiteitssingel 50, Maastricht, PO Box 616, 6200 MD, Maastricht, The Netherlands.
Supplementary data
Supplementary data are available at European Heart Journal online.
Declarations
Disclosure of Interest
All authors declare no conflict of interest for this contribution.
Data Availability
All data, including coding for reproduction (also available through https://github.com/samuelheuts/R-codes-andmanual-for-analyses), are available upon reasonable request to the corresponding author.
Funding
All authors declare no funding for this contribution.
Ethical Approval
Ethical approval was not required as the current study comprises a systematic review and meta-analysis of existing literature.
Pre-registered Clinical Trial Number
This systematic review and meta-analysis was registered before initiation in PROSPERO: registration date 18 July 2022 (CRD42022345024).
References
- 1.Chhabra KR, Dimick JB. Hospital networks and value-based payment: fertile ground for regionalizing high-risk surgery. JAMA 2015;314:1335–6. 10.1001/jama.2015.9803 [DOI] [PubMed] [Google Scholar]
- 2.Sheetz KH, Dimick JB, Nathan H. Centralization of high-risk cancer surgery within existing hospital systems. J Clin Oncol 2019;37:3234–42. 10.1200/JCO.18.02035 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Budtz-Lilly J, Bjorck M, Venermo M, Debus S, Behrendt CA, Altreuther M, et al. Editor's choice—the impact of centralisation and endovascular aneurysm repair on treatment of ruptured abdominal aortic aneurysms based on international registries. Eur J Vasc Endovasc Surg 2018;56:181–8. 10.1016/j.ejvs.2018.01.014 [DOI] [PubMed] [Google Scholar]
- 4.Aikawa T, Yamaji K, Nagai T, Kohsaka S, Kamiya K, Omote K, et al. Procedural volume and outcomes after percutaneous coronary intervention for unprotected left main coronary artery disease-report from the national clinical data (J-PCI registry). J Am Heart Assoc 2020;9:e015404. 10.1161/JAHA.119.015404 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sheetz KH, Chhabra KR, Smith ME, Dimick JB, Nathan H. Association of discretionary hospital volume standards for high-risk cancer surgery with patient outcomes and access, 2005–2016. JAMA Surg 2019;154:1005–12. 10.1001/jamasurg.2019.3017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Birkmeyer JD, Siewers AE, Finlayson EV, Stukel TA, Lucas FL, Batista I, et al. Hospital volume and surgical mortality in the United States. N Engl J Med 2002;346:1128–37. 10.1056/NEJMsa012337 [DOI] [PubMed] [Google Scholar]
- 7.Luft HS, Bunker JP, Enthoven AC. Should operations be regionalized? The empirical relation between surgical volume and mortality. N Engl J Med 1979;301:1364–9. 10.1056/NEJM197912203012503 [DOI] [PubMed] [Google Scholar]
- 8.Begg CB, Cramer LD, Hoskins WJ, Brennan MF. Impact of hospital volume on operative mortality for major cancer surgery. JAMA 1998;280:1747–51. 10.1001/jama.280.20.1747 [DOI] [PubMed] [Google Scholar]
- 9.de Wilde RF, Besselink MG, van der Tweel I, de Hingh IH, van Eijck CH, Dejong CH, et al. Impact of nationwide centralization of pancreaticoduodenectomy on hospital mortality. Br J Surg 2012;99:404–10. 10.1002/bjs.8664 [DOI] [PubMed] [Google Scholar]
- 10.Erbel R, Aboyans V, Boileau C, Bossone E, Bartolomeo RD, Eggebrecht H, et al. 2014 ESC guidelines on the diagnosis and treatment of aortic diseases: document covering acute and chronic aortic diseases of the thoracic and abdominal aorta of the adult. The task force for the diagnosis and treatment of aortic diseases of the European Society of Cardiology (ESC). Eur Heart J 2014;35:2873–926. 10.1093/eurheartj/ehu281 [DOI] [PubMed] [Google Scholar]
- 11.Harris KM, Nienaber CA, Peterson MD, Woznicki EM, Braverman AC, Trimarchi S, et al. Early mortality in type A acute aortic dissection: insights from the international registry of acute aortic dissection. JAMA Cardiol 2022;7:1009–15. 10.1001/jamacardio.2022.2718 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Rylski B, Schilling O, Czerny M. Acute aortic dissection: evidence, uncertainties, and future therapies. Eur Heart J 2023;44:813–21. 10.1093/eurheartj/ehac757 [DOI] [PubMed] [Google Scholar]
- 13.Goldstone AB, Chiu P, Baiocchi M, Lingala B, Lee J, Rigdon J, et al. Interfacility transfer of Medicare beneficiaries with acute type A aortic dissection and regionalization of care in the United States. Circulation 2019;140:1239–50. 10.1161/CIRCULATIONAHA.118.038867 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Chikwe J, Cavallaro P, Itagaki S, Seigerman M, Diluozzo G, Adams DH. National outcomes in acute aortic dissection: influence of surgeon and institutional volume on operative mortality. Ann Thorac Surg 2013;95:1563–9. 10.1016/j.athoracsur.2013.02.039 [DOI] [PubMed] [Google Scholar]
- 15.Booth A, Clarke M, Dooley G, Ghersi D, Moher D, Petticrew M, et al. The nuts and bolts of PROSPERO: an international prospective register of systematic reviews. Syst Rev 2012;1:2. 10.1186/2046-4053-1-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. 10.1136/bmj.n71 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ouzzani M, Hammady H, Fedorowicz Z, Elmagarmid A. Rayyan-a web and mobile app for systematic reviews. Syst Rev 2016;5:210. 10.1186/s13643-016-0384-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Badhwar V, Vemulapalli S, Mack MA, Gillinov AM, Chikwe J, Dearani JA, et al. Volume-outcome association of mitral valve surgery in the United States. JAMA Cardiol 2020;5:1092–101. 10.1001/jamacardio.2020.2221 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Vemulapalli S, Carroll JD, Mack MJ, Li Z, Dai D, Kosinski AS, et al. Procedural volume and outcomes for transcatheter aortic-valve replacement. N Engl J Med 2019;380:2541–50. 10.1056/NEJMsa1901109 [DOI] [PubMed] [Google Scholar]
- 20.Ma LL, Wang YY, Yang ZH, Huang D, Weng H, Zeng XT. Methodological quality (risk of bias) assessment tools for primary and secondary medical studies: what are they and which is better? Mil Med Res 2020;7:7. 10.1186/s40779-020-00238-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Heuts S, Denessen EJS, Daemen JHT, Vroemen WHM, Sels JW, Segers P, et al. Meta-analysis evaluating high-sensitivity cardiac troponin T kinetics after coronary artery bypass grafting in relation to the current definitions of myocardial infarction. Am J Cardiol 2022;163:25–31. 10.1016/j.amjcard.2021.09.049 [DOI] [PubMed] [Google Scholar]
- 22.Wan X, Wang W, Liu J, Tong T. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 2014;14:135. 10.1186/1471-2288-14-135 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Shim SR, Lee J. Dose-response meta-analysis: application and practice using the R software. Epidemiol Health 2019;41:e2019006. 10.4178/epih.e2019006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Di Giuseppe D, Crippa A, Orsini N, Wolk A. Fish consumption and risk of rheumatoid arthritis: a dose-response meta-analysis. Arthritis Res Ther 2014;16:446. 10.1186/s13075-014-0446-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Buis ML. Predict and adjust with logistic regression. Stata J 2007;7:221–6. 10.1177/1536867X0700700206 [DOI] [Google Scholar]
- 26.Antunes M, Gomes D, Aguiar RL. Knee/Elbow estimation based on first derivative threshold. In: IEEE International Conference on Big Data Computing Service and Applications (BigDataService). Bamberg, Germany: Conference Publishing Services, 2018. [Google Scholar]
- 27.Satopaa V, Albrecht J, Irwin D, Raghavan B. Finding a “Kneedle” in a haystack: detecting knee points in system behavior. In: 2011 31st International Conference on Distributed Computing Systems Workshops. Minneapolis, MN: Conference Publishing Services, 2011. [Google Scholar]
- 28.Liu N, Zhou Y, Lee JJ. IPDfromKM: reconstruct individual patient data from published Kaplan-Meier survival curves. BMC Med Res Methodol 2021;21:111. 10.1186/s12874-021-01308-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Gasperoni F, Ieva F, Paganoni AM, Jackson CH, Sharples L. Non-parametric frailty cox models for hierarchical time-to-event data. Biostatistics 2020;21:531–44. 10.1093/biostatistics/kxy071 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Jakobsen JC, Gluud C, Wetterslev J, Winkel P. When and how should multiple imputation be used for handling missing data in randomised clinical trials—a practical guide with flowcharts. BMC Med Res Methodol 2017;17:162. 10.1186/s12874-017-0442-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Mariscalco G, Maselli D, Zanobini M, Ahmed A, Bruno VD, Benedetto U, et al. Aortic centres should represent the standard of care for acute aortic syndrome. Eur J Prev Cardiol 2018;25:3–14. 10.1177/2047487318764963 [DOI] [PubMed] [Google Scholar]
- 32.Knipp BS, Deeb GM, Prager RL, Williams CY, Upchurch GR Jr., Patel HJ. A contemporary analysis of outcomes for operative repair of type A aortic dissection in the United States. Surgery 2007;142:524–8; discussion 528 e1. 10.1016/j.surg.2007.07.012 [DOI] [PubMed] [Google Scholar]
- 33.Brescia AA, Patel HJ, Likosky DS, Watt TMF, Wu X, Strobel RJ, et al. Volume-outcome relationships in surgical and endovascular repair of aortic dissection. Ann Thorac Surg 2019;108:1299–306. 10.1016/j.athoracsur.2019.06.047 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Evangelista A, Salas A, Ribera A, Ferreira-Gonzalez I, Cuellar H, Pineda V, et al. Long-term outcome of aortic dissection with patent false lumen: predictive role of entry tear size and location. Circulation 2012;125:3133–41. 10.1161/CIRCULATIONAHA.111.090266 [DOI] [PubMed] [Google Scholar]
- 35.Sa MP, Jacquemyn X, Tasoudis PT, Van den Eynde J, Erten O, Sicouri S, et al. Long-term outcomes of total arch replacement versus proximal aortic replacement in acute type A aortic dissection: meta-analysis of Kaplan-Meier-derived individual patient data. J Card Surg 2022;37:4256–66. 10.1111/jocs.16852 [DOI] [PubMed] [Google Scholar]
- 36.Heuts S, Adriaans BP, Kawczynski MJ, Daemen JHT, Natour E, Lorusso R, et al. Editor's choice—extending aortic replacement beyond the proximal arch in acute type A aortic dissection: a meta-analysis of short term outcomes and long term actuarial survival. Eur J Vasc Endovasc Surg 2022;63:674–87. 10.1016/j.ejvs.2021.12.045 [DOI] [PubMed] [Google Scholar]
- 37.Augoustides JG, Geirsson A, Szeto WY, Walsh EK, Cornelius B, Pochettino A, et al. Observational study of mortality risk stratification by ischemic presentation in patients with acute type A aortic dissection: the Penn classification. Nat Clin Pract Cardiovasc Med 2009;6:140–6. 10.1038/ncpcardio1417 [DOI] [PubMed] [Google Scholar]
- 38.Trimarchi S, Nienaber CA, Rampoldi V, Myrmel T, Suzuki T, Mehta RH, et al. Contemporary results of surgery in acute type A aortic dissection: the international registry of acute aortic dissection experience. J Thorac Cardiovasc Surg 2005;129:112–22. 10.1016/j.jtcvs.2004.09.005 [DOI] [PubMed] [Google Scholar]
- 39.Yang B, Rosati CM, Norton EL, Kim KM, Khaja MS, Dasika N, et al. Endovascular fenestration/stenting first followed by delayed open aortic repair for acute type A aortic dissection with malperfusion syndrome. Circulation 2018;138:2091–103. 10.1161/CIRCULATIONAHA.118.036328 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Cruz I, Stuart B, Caldeira D, Morgado G, Gomes AC, Almeida AR, et al. Controlled pericardiocentesis in patients with cardiac tamponade complicating aortic dissection: experience of a centre without cardiothoracic surgery. Eur Heart J Acute Cardiovasc Care 2015;4:124–8. 10.1177/2048872614549737 [DOI] [PubMed] [Google Scholar]
- 41.Harris KM, Strauss CE, Duval S, Unger BT, Kroshus TJ, Inampudi S, et al. Multidisciplinary standardized care for acute aortic dissection: design and initial outcomes of a regional care model. Circ Cardiovasc Qual Outcomes 2010;3:424–30. 10.1161/CIRCOUTCOMES.109.920140 [DOI] [PubMed] [Google Scholar]
- 42.Harris KM, Strauss CE, Eagle KA, Hirsch AT, Isselbacher EM, Tsai TT, et al. Correlates of delayed recognition and treatment of acute type A aortic dissection: the international registry of acute aortic dissection (IRAD). Circulation 2011;124:1911–8. 10.1161/CIRCULATIONAHA.110.006320 [DOI] [PubMed] [Google Scholar]
- 43.Heuts S, Schalla S, Ramaekers M, Bidar E, Mihl C, Wildberger JE, et al. Imaging surveillance for complications after primary surgery for type A aortic dissection. Heart 2022;109:96–101. 10.1136/heartjnl-2022-320881 [DOI] [PubMed] [Google Scholar]
- 44.Meuli L, Menges AL, Stoklasa K, Steigmiller K, Reutersberg B, Zimmermann A. Inter-hospital transfer of patients with ruptured abdominal aortic aneurysm in Switzerland. Eur J Vasc Endovasc Surg 2023;65:484–92. 10.1016/j.ejvs.2022.12.009 [DOI] [PubMed] [Google Scholar]
- 45.Isselbacher EM, Preventza O, Hamilton Black JIII, Augoustides JG, Beck AW, Bolen MA, et al. 2022 ACC/AHA guideline for the diagnosis and management of aortic disease: a report of the American heart association/American college of cardiology joint committee on clinical practice guidelines. Circulation 2022;146:e334–e482. 10.1161/CIR.0000000000001106 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Concannon TW, Kent DM, Normand SL, Newhouse JP, Griffith JL, Cohen J, et al. Comparative effectiveness of ST-segment-elevation myocardial infarction regionalization strategies. Circ Cardiovasc Qual Outcomes 2010;3:506–13. 10.1161/CIRCOUTCOMES.109.908541 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Ahola R, Sand J, Laukkarinen J. Pancreatic resections are not only safest but also most cost-effective when performed in a high-volume centre: a Finnish register study. Pancreatology 2019;19:769–74. 10.1016/j.pan.2019.06.007 [DOI] [PubMed] [Google Scholar]
- 48.Bashir M, Harky A, Fok M, Shaw M, Hickey GL, Grant SW, et al. Acute type A aortic dissection in the United Kingdom: surgeon volume-outcome relation. J Thorac Cardiovasc Surg 2017;154:398–406 e1. 10.1016/j.jtcvs.2017.02.015 [DOI] [PubMed] [Google Scholar]
- 49.Andersen ND, Ganapathi AM, Hanna JM, Williams JB, Gaca JG, Hughes GC. Outcomes of acute type a dissection repair before and after implementation of a multidisciplinary thoracic aortic surgery program. J Am Coll Cardiol 2014;63:1796–803. 10.1016/j.jacc.2013.10.085 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Huguet M. Centralization of care in high volume hospitals and inequalities in access to care. Soc Sci Med 2020;260:113177. 10.1016/j.socscimed.2020.113177 [DOI] [PubMed] [Google Scholar]
- 51.Vogel J, Barkhause M, Pross C, Geissler A. Defining minimum volume thresholds to increase quality of care: a new patient-oriented approach using mixed integer programming. Eur J Health Econ 2022;23:1085–104. 10.1007/s10198-021-01406-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Benedetto U, Dimagli A, Kaura A, Sinha S, Mariscalco G, Krasopoulos G, et al. Determinants of outcomes following surgery for type A acute aortic dissection: the UK national adult cardiac surgical audit. Eur Heart J 2021;43:44–52. 10.1093/eurheartj/ehab586 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Sievers HH, Rylski B, Czerny M, Baier ALM, Kreibich M, Siepe M, et al. Aortic dissection reconsidered: type, entry site, malperfusion classification adding clarity and enabling outcome prediction. Interact Cardiovasc Thorac Surg 2020;30:451–7. 10.1093/icvts/ivz281 [DOI] [PubMed] [Google Scholar]
- 54.Schoenhoff FS, Tian DH, Misfeld M, Perreas KG, Spielvogel D, Mohr FW, et al. Impact of reimplantation technique of supra-aortic branches in total arch replacement on stroke rate and survival: results from the ARCH registry. Eur J Cardiothorac Surg 2018;54:1045–51. 10.1093/ejcts/ezy194 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data, including coding for reproduction (also available through https://github.com/samuelheuts/R-codes-andmanual-for-analyses), are available upon reasonable request to the corresponding author.