A multicentre review comparing long term outcomes of endoscopic vein harvesting versus open vein harvesting for coronary artery bypass surgery

Patients

Between January 2007 and December 2019, 27,024 patients (OVH, n=13,230; EVH, n=13,794) who had CABG surgery with at least one long saphenous vein from Royal Brompton and Harefield (n=14,717), Wythenshawe (n=6,360) and Blackpool (n=5,947) hospitals were included in this study (Figure 1). All patients’ demographics, intraoperative and postoperative details were prospectively collected and entered into the database by healthcare practitioners. Data was obtained from a prospectively maintained institutional registry (Dendrite Clinical Systems, Henley-on Thames, UK) and NICOR UK database. In addition to this database, some of the intervention details were obtained from local cardiology centres and General Practitioner letters. All mortality and survival data were obtained from the national censor database, UK.

Figure 1. Patient recruitment and flow chart.

Ethical approval

The Edge Hill University Health Research ethical committee (University REC ref: ETH2021-0066) approved this audit study protocol, which is in accordance with the principles of the Declaration of Helsinki. The ethical committee waived the need to obtain informed consent from the patients. In addition, the use of these data for this study was approved by the Manchester Foundation NHS Trust hospital institutional review board (registration no: 9477) and it was registered in the clinical governance audit departments. Only symptomatic patients who came back to hospital for reintervention were analysed as a secondary analysis for this study. All three centres cover large geographical areas and patients who get treated outside of the territory were unable to be included in this analysis.

Surgical techniques

Data was included from patients whose vein was retrieved by experienced practitioners (previous experience of a minimum of 100 OVH cases and 50 EVH cases) at three hospitals. However, the Blackpool EVH dataset included patients’ outcomes following surgery by harvesters transitioning from OVH to EVH during the period of this study. The surgical techniques for open vein and endoscopic vein harvesting are standard surgical practice, as explained in previous publications^3,11. All EVH patients received at least 2500 to 5000iu heparin before the start of EVH to avoid any intra-luminal clot formation¹¹. Harefield and Wythenshawe hospitals used the Maquet Vasoview® Hemopro 1 and 2 (MAQUET, Inc, Wayne, NJ) vein harvesting systems. Blackpool hospitals have also used Maquet® Vasoview Hemopro 1 and 2 for 60% of their retrievals, and the remaining 40% were using the Terumo® VirtuoSaph (Terumo Cardiovascular, Ann Arbor, Mich).

Study end point

The primary end points of this study were mortality, and the secondary end points were postoperative complications, length of hospital stay and repeat revascularisation for symptomatic patients.

Statistical analyses

All patient characteristics were described by using count (percentage) for categorial variables and mean (standard deviation) for continuous variables. The primary outcome measure was death from any cause after surgery. Patients were followed from the date of surgery until the date of death. All statistical analyses and data management were performed by an independent statistician using R- software (Rx64 version 4.0.3)^12,13. The population were divided into EVH vs OVH and data was coded for analysis to blind the statistician. To investigate the effect of the two surgical techniques (OVH vs EVH) on mortality, we used Inverse Probability of Treatment Weights (IPTW) to balance demographics and preoperative comorbidities between the two groups.

Missing variables were imputed with multiple imputation using all baseline variables in Table 1, plus intervention (open/endoscopic), duration in the study (days), year of operation and survival status in February 2020 (dead/alive). Ten multiply imputed data were created. To assess the intervention effect, each imputed data was weighted using IPTW, intervention effect was calculated in each weighted dataset using cox proportional hazard regression in which time to event was regressed against intervention. Results were pooled using Rubin’s rule.

IPTW model. The propensity scores¹⁴ were estimated using a logistic regression model in which intervention assignment (OVH or EVH) was regressed on the 25 covariates and their pairwise interactions, as listed in Table 1. Also restricted cubic splines with 5 knots were entered for age, BMI and logistic EuroSCORE. Pairwise interactions between multi-level (more than 2 levels) categorical variables were removed as they had very low balance at some levels and created large weights. Stabilised weights to assess Average Treatment Effect (ATE) were calculated for each imputed set.

Balance diagnostics in weighted datasets. It has been suggested that standardised differences in excess of 10% may be indicative of meaningful imbalance in a covariate between the two groups and less than 10% is negligible imbalance¹⁵. Balancing criteria were considered as standardised mean difference less than 10%, variance ratio between treatment groups for continuous variables less than 2 and Kolmogorov–Smirnov threshold of less than 4% for equality of the continuous variables between treatment groups. The balance was assessed for each variable in Table 1, squared and cubic order of continuous variables and 2x2 interaction between continuous variables¹⁶.

Pre-operative data

During the study period, a total of 13,794 patients underwent EVH and 13,230 underwent OVH for CABG surgery. Patient characteristics and pre-operative data are included in Table 1. The majority of patients in the EVH group underwent surgery at Harefield hospital, although the ratio of OVH cases was more evenly split between hospitals. Table 1 is the summary of baseline and pre-operative comorbidities. Add EuroSCORE had a skewed distribution and therefore logistic (Add EuroSCORE) was used. Continuous variables are reported by mean(SD) and categorical variables with count (percentage). A total of 7 out of 25 preoperative variables had standardised difference above 10%, indicating imbalance between the two groups (Table 1). IPTW was applied to 10 imputed datasets. Mean stabilized weights for the 10 imputed data were 1.006-1.007 with standard deviations 0.66-0.67. Stabilized weights were all smaller than 10. Maximum absolute standardized differences between the two intervention groups in weighted datasets was 0.036 (Table 1), maximum KS difference for continuous variables was 0.017 and maximum variance ratio for continuous variables was 1.04.

Peri-operative data

Peri-operative details are displayed in Table 2. The length of hospital stay is similar for both groups at median 6 days (p=0.86). Significant differences are observed in the type of surgery undertaken, with a greater number of EVH patients also receiving valve surgery or additional other cardiac surgery compared to the OVH group (p<0.001). A greater number of EVH patients also had the mammary artery used compared to the OVH group (p=0.009).

Mortality

The primary outcome measure for the study was post-operative mortality. Across the follow-up period of the study (median 1656 days for EVH and 2191 days for OVH), deaths from any cause occurred for 13.8% of EVH patients and 20.8% of OVH patients. In-hospital death occurred at similar rates (2.3% EVH vs. 2.1% OVH, p=0.12; Table 3). The crude hazard ratio of death (EVH to OVH, HR (95% CI)) was computed as 0.902 (0.850, 0.957) for raw sample data, 0.892 (0.839, 0.948) for complete case data and 0.851 (0.789, 0.919) for weighted complete case data. Kaplan-Meier plots for the raw data, complete case data and weight complete case data are depicted in Figure 2, with greater survival in the EVH group compared to the OVH group (p<0.001 for all).

Figure 2. Kaplan-Meier survival curves for the study cohort.

Since the missing data was not completely at random, multiple imputation was used to impute missing data for Euroscore and previous MI. Each imputed set was weighted and hazard ratio for mortality was obtained for each weighted dataset. Results were pooled using Rubin’s rule.

Overall hazard ratio (95% CI) of death EVH:OVH was 0.823 (0.767, 0.884). As a sensitivity analyses for unmeasured confounders, once those with weights outside overlapped region and once those with weights above 99 percentile and below 1 percentile of all weights were excluded, remaining data were re-weighted and hazard ratios were calculated and pooled again. Hazard ratio (95% CI) from excluding non-overlapped weights was 0.829 (0.772, 0.889). Hazard ratio (95% CI) from excluding lower and upper 1% weights was 0.837 (0.778, 0.900). There seem to be a slight bias for unmeasured confounder, but nevertheless the difference between the two interventions stays significant.

Secondary outcome measures

Post-operative complications and requirement for repeat operation is displayed in Table 3. Repeat operation was required more frequently in the EVH group (3.9% vs. 2.7%, p<0.001), however this was not as a result of graft occlusion (EVH vs. OVH: 0.1% vs. 0.2%). There was a significant difference in post-operative arrhythmias between the groups (p<0.001). Whilst a similar proportion of patients were free of arrhythmias in each group (EVH vs. OVH: 79.6% vs. 79.5%, respectively), there was a difference in the nature of arrhythmias recorded between the groups, with a greater requirement for permanent pacing in the EVH group (1.4% vs. 0.9%). A similar effect was observed in post-operative gastrointestinal complications with 98.2% free from problems in both groups. However, the nature of the complications observed differed between the groups (p=0.003). Finally, pulmonary complications were observed at a higher rate in the EVH group compared to OVH (14.7% vs. 12.8%, p<0.001), driven predominantly by a more common occurrence of collapse and consolidation (5.8% vs. 2.9%). Importantly, there were also a significantly greater number of surgical site infections in the OVH group, driven primarily by leg wound infection (p<0.001).

Study limitations

This is an observational retrospective study rather than a prospective randomised controlled trial. Propensity score matching requires that the sample of untreated subjects be larger than the sample of treated subjects. Ideally, there should be substantially more untreated subjects than treated subjects. Therefore, for this study we have used IPTW, which considered weights based on the propensity score to create a synthetic sample in which the distribution of measured baseline covariates is independent of treatment assignment. A subject’s weight is equal to the inverse of the probability of receiving the treatment that the subject received. Propensity score weighting assumes that there is no unmeasured confounding factor (e.g. all confounding factors have been considered in the model).

Another important limitation is that there are not angiographic details for the entire cohort of patients. Limited angiographic data was available only for symptomatic patients who have come back to the base hospital for reintervention, and these are listed on the Table 4. Patients who have attended peripheral hospitals were not captured due to large geographical catchment areas. It is difficult to accurately capture major adverse cardiac events in both groups due to lack of routine planned follow up in specialist cardiac centres. Importantly, we need to take into consideration that it is a multicentre observational study where there is likely to be a degree of variations in vein harvesting and surgical techniques by different operators.

Underlying data

The raw patient data were obtained from the Royal Brompton and Harefield, Wythenshawe and Blackpool hospitals. The hospitals and Edgehill University gave permission to conduct analysis on the data for this study only and not future studies. For further research using these datasets, additional permission must be sought from these hospitals. Researchers who require access to the raw data will need to contact the audit/research department of each hospital; contact details are available as follows: Royal Brompton and Harefield Research office, https://www.rbht.nhs.uk/research/research-office, Wythenshawe, mft.rd@manchester.ac.uk; Blackpool, https://www.bfwh.nhs.uk/onehr/research-development/. Intermediary data can be found in the article.