Disturbances in native cardiac electric conduction after cardiac surgery manifest as transient or persistent bradyarrhythmia, the latter necessitating permanent pacemaker (PPM) implantation in a minority of patients after observation for 5–7 days.1 Accurate pre-operative risk stratification for PPM requirement may reduce bradyarrhythmic complications and length of hospital stay, but this is done suboptimally in the clinical setting, and studies have reported mixed findings of predictors.2–6 We aim to review and identify predictors of PPM implantation after cardiac surgery at our centre.

Methods

All patients who had cardiac surgery at Auckland City Hospital during June 2014–October 2016 who had PPM implantation during the same admission and after cardiac surgery were reviewed. A random sample of the identical number of patients of other cardiac surgery patients were obtained as controls. Clinical characteristics including demographics, past history, presentation, pre-operative ECG, surgery type and times were extracted from computerised records and EuroSCORE II7 was calculated.

Pre-operative ECGs were divided into 4 categories defined as 1: normal, 2: arrhythmia (in the absence of any heart block), 3: low-grade block (not a standalone pacemaker indication such as first degree, left or right bundle branch, bifasiscular block, Wenckebach), and 4: high grade block (that is pacemaker indication, eg, 2nd degree type 2 or complete heart block). Surgery types were divided into category 1: isolated coronary, 2: 1 valve+/-coronary, 3: 2 valves+/-coronary and 4: other (such as aortic surgery, congenital heart surgery, triple valve surgery).

Mean+/-standard deviation and percentage (frequency) were used to present quantitative and qualitative variables, and Hosmer-Lemeshow and Fisher’s Exact Tests used for univariable analysis of these respectively. Logistic regression was used to identify independent predictors of PPM implantation, and construct an additive model for predicting this. Its accuracy is assessed using the area under the receiver-operator characteristics curve c-statistic. P-values <0.05 were deemed statistically significant and all tests were two-tailed. SPSS (Version 17.0, SPSS Inc., Chicago, IL, USA) was used for statistical analyses.

Results

Of 2,446 cardiac surgeries undertaken, 4.0% (98) received a post-operative PPM during the study period, so 98 controls were selected for comparison, with characteristics presented in Table 1. PPM patients were on average older, had higher prevalence of previous surgery and lower prevalence of previous myocardial infarction, higher EuroSCORE II, higher proportion with higher grade block and valvular surgery in univariate analysis. In particular, PPM incidence was 0.6% (7/1165) for coronary surgery, 7.1% (45/636) for single valve+/-coronary surgery, 9.0% (12/133) for double valve+/-coronary surgery and 7.0% (34/489) for other cardiac surgeries.

Table 1: Clinical characteristics and outcomes of cardiac surgery receiving PPM or not (controls).

Multivariable analysis performed based on the above univariable analysis parameters, and combining category 4 and 2 of operation type as one group, identified pre-operative ECG and operation type as independent predictors of PPM implantation after cardiac surgery. Odds ratios (95% confidence intervals) per category were 1.65 (1.08–2.53) and 5.2 (2.47–10.9) respectively. As such, a simple additive risk model out of seven of these two variables was created: a) pre-operative ECG score 0 for normal, 1 for arrhythmia, 2 for low-grade block and 3 for high-grade block; and b) type of surgery score 0 for isolated coronary, 2 for 1 valve+/-coronary or other, 4 for 2 valve+/-coronary. C-statistics of this model and individual parameters for PPM implantation predictor are shown in Table 2. The risk model had the highest c-statistic of 0.78, and if the score cutpoint was set at 3, gave sensitivity of 70.4% and specificity of 77.6%.

Table 2: Area under the curve (c-statistic) for predicting PPM implantation after cardiac surgery.

‍

*Change to risk model is based on two parameters and total score of 7: a) pre-operative ECG score 0 for normal, 1 for arrhythmia, 2 for low-grade block and 3 for high-grade block; and b) type of surgery score 0 for isolated coronary, 2 for 1 valve+/-coronary or other, 4 for 2 valve+/-coronary.

Discussion

PPM was implanted in 4.0% of our cardiac surgery cohort, comparable to the 1.4–8.5% incidence reported in contemporary studies.2–6 Pre-operative ECG findings of bundle branch block or other heart block,2–5 and valvular heart surgery4,5 were also found to be predictors in other studies like ours, although we found that multi-valve surgery had independently higher risk than single-valve surgery. Other predictors previously identified include redo operation and pulmonary hypertension,5 age, emergency admission, diabetes, renal impairment and heart failure,6 which are common risk factors to operative mortality.7 To our knowledge, we have developed the first risk score for predicting PPM requirement after cardiac surgery, based on the two important predictors of pre-operative ECG and surgery type, which performed moderately well with c=0.78. The score out of 7 can be clinically utilised as 0–1 as low risk, 2–3 as moderate risk and 4–7 as high risk. The clinical implication is that in low-risk patients, temporary epicardial wires could be turned off and removed earlier, and those moderate and particularly high-risk patients would need close monitoring early on for, and ongoing monitoring with low threshold of PPM implantation of, any conduction disturbance post-operatively. The simplicity of the score should aid rather than hinder utility even if performance maybe somewhat hindered.

This study has several limitations. It is a single-centre retrospective observational study. The study sample size was small so power was limited. It was a case-control rather than cohort study design, and the controls, although randomly selected, weren't matched to the pacemaker group, however this allowed us to assess and highlight the differences in characteristics between the two groups. We did not specifically examine long-term outcomes, including survival, pacing burden and late pacemaker dependency, but focused only on what factors predicted patients getting PPM in the first place. The score has not been externally validated, which would be a next step. Nevertheless, our analysis identified two important parameters of pre-operative ECG and type of cardiac surgery and constructed a user-friendly additive risk model to predict PPM after cardiac surgery with moderately good accuracy.

‍