Arrhythmias and ischemic heart disease comprise a significant percentage of sudden cardiac death especially in the older population   . Patients with ischemic heart disease presenting with ventricular arrhythmias (VA) represent a high risk population compared to the rest of the patients undergoing surgical revascularisation  . Randomised control trial on this subgroup of patients is not feasible due to ethical and clinical restraints and although a number of studies assessed short and long term outcomes of these ischemic patients presenting with VAs none seems to have a proper control group    . The absence of a control group raises concerns over any conclusions. In this study we aim to assess the short and long term outcomes of patients undergoing surgical revascularisation following presentation with VA against an equivalent control group by propensity matching. Propensity matching empowers the retrospective study. It allows comparable results and conclusions to those of a randomised controlled study  .
Given the increased incidence of annual VA (8%) even after revascularisation  , we also reviewed the impact of completeness of revascularisation and ICD insertion on these outcomes in our data.
This is a retrospective study of all patients that underwent coronary artery bypass surgery (CABG) between January 1999 and September 2015 and included in our database at Hull and East Yorkshire Hospitals NHS Trust was examined. We identified a total of 9609 patients who underwent CABG with 54 (5.62%) of them having survived an ischemic ventricular arrhythmia (IVA). Operative mortality was defined as occurring during the operative hospitalization or within 30 days of operation for discharged patients. The preoperative characteristics were as defined by the EuroSCORE I criteria  .
Other secondary outcomes were: Perioperative intra-aortic balloon pump, sternal resuturing, postoperative myocardial infarction, postoperative arrhythmias and permanent pacing, pulmonary complications, wound infections, gastrointestinal complications and multiorgan failure.
2.2. Propensity Score Matching
Using “R” program  ; patient’s preoperative characteristics (Age, gender, BMI), comorbidities (neurologic, renal, pulmonary), Ejection fraction, urgency of operation and logistic EuroSCORE were used to develop a score (propensity score) for each case. The 54 post IVA patients were propensity matched to patients that underwent CABG without prior IVA in a 1:2 ratio (n = 108).
Individual informed consent was waived due to the fact that the data was already available for clinical and research purposes. Furthermore, the data was in a format that would not allow any individual patient to be identified (anonymised).
2.3. Statistical Analysis
1) Univariate analysis:
Categorical data are summarized using percentages. Continuous data are presented as the mean value (±standard deviation). Categorical data were compared by Chi-squared test. Continuous data were tested against “Normal” Plot and “Q-Q” plots; those of normal distribution were compared using student t-test while non-normal distributed data were compared by “Mann-Whitney test”.
2) Multivariate analysis:
Cox multivariate analysis was used to identify the impact of the significant factors from the univariate analysis on the long term survival.
2.4. Long Term Survival
Long term survival for each group was analysed using Kaplan-Meier curves and then compared by the log rank test to calculate the p-value.
A p-value of <0.05 was considered statistically significant. The 95% confidence interval is calculated according to Brookmeyer & Crowley.
Statistical analyses were performed using MedCalc for Windows, version 13.1.2 (MedCalc Software, Ostend, Belgium).
The baseline criteria for all the cases that underwent cardiac surgery in our unit are shown in Table 1.
Study flow chart:
The flow chart (Figure 1) shows how the study and the control groups were obtained and their relevant sizes.
Propensity scoring and matching:
Using R computer programming each patient from the patient population was given a score depending on the factors mentioned in the methods section. Based on the score, each patient from the study group was matched with the closest two patients from the control group (Figure 2).
Analysis of the pre-, intra- and postoperative factors identified post operative VF and permanent pacemaker insertion were significantly higher in the study group (Table 2).
Table 1. Baseline criteria for the database (n = 9609). Absolute numbers (percentage of total) or mean ± standard deviation (SD).
Figure 1. Flow chart showing the control and study groups.
Figure 2. Bar charts and dot graph showing the propensity matched control and study groups.
Cox multivariate analysis (predictors of long term survival):
Multivariate analysis revealed that the only factor impacting on in hospital mortality was postoperative VF (HR 1.5) as seen in Table 2.
Significantly higher in the study group. Multivariate analysis showing Postoperative VF the only factor impacting on in-hospital mortality.
Table 2. Univariate analysis showing postoperative VF and permanent pacing were significantly higher in the study group. Multivariate analysis showing Postoperative VF the only factor impacting on in-hospital mortality.
Long term survival:
Kaplan Meier analysis of the long term survival at 15 years postoperatively showed 7/108 (control group) versus 6/54 (study group) alive. The median survival was 13 (control group) and 10 years (study group). This was not significant by log rank test (p = 0.33) as demonstrated in Figure 3.
Subgroup analysis of Patients with persistent VA in the study group:
12 (22.2%) patients in the IVA group had postoperative VF. 2 died in hospital (No EPS performed). 5/10 discharged patients had ICDs inserted (Mean 5.86 years survival). 3/10 discharged did not require ICD (Mean 5.66 years survival). 2/10 needed PPM (Mean 2.85 years survival). The relation between ICD/PPM insertion and LV function (in patients that had postoperative VF and were discharged home alive) is shown in Table 3.
The ratio between the extent of coronary artery disease to number of bypass grafts was as follows:
Preoperative sinus rhythm and postoperative non-VA: 0.95, preoperative sinus rhythm and postoperative VA: 1.03, preoperative IVA and postoperative non-VA: 0.99 and preoperative IVA with postoperative IVA: 0.97.
VAs are a fatal complication of ischemic heart disease. Many earlier studies claimed better short   and long term survival following revascularisation  . However, these studies failed to compare this outcome with an “equivalent” control group with suboptimal deduction of conclusions.
The difference in the preoperative characteristics between the database and the propensity matched control group highlights the relevance of performing propensity matching, allowing comparison between homogenous groups, not only performing a similar operative procedure, but with similar characteristics.
To the best of our knowledge, this is the first study to compare the short and long term outcomes following surgical revascularisation with a similar population, risk profile-wise; by propensity scoring and matching. We shall base our discussion on the following questions:
1) Ischemic patients with preoperative VA have up to 8% annual recurrence of their VA (MADT-CRT)  even following there revascularisation. Should this fact impact on conduit choice and completion of grafting (e.g., arterial revascularisation)?
2) With such a high risk profile, rapid surgical revascularisation of the culprit lesions; as performed in primary PCI, should suffice and allow these high risk patients to be discharged home from a quick operation rather than end up with prolonged ITU stay and in-hospital mortality while the remaining CAD may be dealt with by hybrid PCI?
The two groups had similar short term outcomes (except for a higher incidence of postoperative VA and pacemaker insertion). There hospital stay was similar (9 days) which was either due to prolonged ITU stay (from Inotropic, IABP, respiratory and renal causes) and ward stay (due to wound break down
Figure 3. Kaplan Meier curves for the control and the study groups showing no significant difference in the long term survival.
Table 3. Relation between postoperative VF (in the IVA group) and LV function.
and infection). Both groups had complete revascularisation (Table 2). In hospital mortality for both groups was statistically similar (p = 0.15). Hence; prolonged bypass and cross clamp time for the sake of adequate revascularisation has not impacted negatively on the short term outcomes.
Long term survival was similar between the control and study group (p = 0.33). From the significant factors on univariate analysis, only persistent postoperative VA significantly affected long term survival (HR 1.5) in the multivariate analysis. Causes of postoperative VA are multiple and interrelated: electrolyte disturbance, poor perfusion secondary to low cardiac output, inadequate revascularisation, arrhythmogenic inotropes and even genetic  . It is difficult to diagnose the exact cause in most cases and some of these causes are unavoidable, such as genetic predisposition. The most important avoidable factor in theatres is inadequate revascularisation which we investigated in more detail by comparing the number of ischemic vessels with the number of grafts in the various subgroups (Table 2 and the last paragraph in the results section) and have not found any difference between the patients that had postoperative VA from the study and control groups as compared to those that had no postoperative VA in both groups. Hence we again conclude that complete revascularisation and choice of conduits should not be jeopardised as patients with preoperative VA have a similar long term survival equivalent to high risk non VA patients. Other studies have suggested the relevance of surgical revascularisation over ICD insertion  .
All patients that needed postoperative ICD had impaired LV but one. The patient with a good LV underwent ICD insertion 1 month postoperatively based on electrophysiologic studies performed preoperatively. This was prior to the recommendations of MADIT-CRT and SOLVD in 2006   -  , which stated that the benefits of ICD insertion are after 6 months and a significant proportion of patients, especially those with preserved LV, will remodel and avoid the need for an ICD during this period. Also the mean survival of patients with postoperative VF in the study group that needed an ICD (5.86 years) was similar to those that did not need an ICD (5.66 years). This highlights that even with persistent aetiologies of post operative VA that cannot be reversed, long term survival may be preserved with ICD insertion  . The 2 patients that needed postoperative pacing had a shorter mean survival (2.85 years). We could not identify the reason for this and the causes of death could not be identified weather cardiac or non cardiac.
Strengths & Limitations
Propensity scoring and matching allowed us to compare the effect of treatment (surgical revascularisation) on observational data. The advantage of using a linear combination of covariates for a single score allows balancing a test group with a control group without losing a large number of observations (cases). In our case the “pool” of data allowed a “greedy” match for the chosen confounding variables.
One disadvantage is that the matching is only for the chosen variables and the rest of the risk factors and comorbidities remain unmatched. Nevertheless, PSM has been assessed by other studies to validate the effect of treatment on observational data    .
The intrinsic limitations of selection bias are inherent to all observational data. Also, failure to attain the cause related mortalities in long term studies is a common drawback for long term studies.
1) Coronary Artery Bypass surgery on patients presenting with IVA can be performed safely yielding short and long term results equivalent to non-IVA cases.
2) Persistence of Postoperative Ventricular Fibrillation is the single predictor for long term mortality in this population.
3) Cardiology referral to assess the need for ICD insertion is essential postoperatively.
These patients should not be denied surgery with consideration of good long term outcome.
There is no funding to declare.
Conflicts of Interest Statement
There is no conflict of interest to declare.