Mediastinitis after coronary artery bypass grafting increases the incidence of left internal mammary artery obstruction

Abstract

Mediastinitis after coronary artery bypass grafting (CABG) gives a longstanding chronic inflammation and has a detrimental negative effect on long‐term survival. For this reason, we aimed to study the effect of mediastinitis on graft patency after CABG. The epidemiologic design was of an exposed (mediastinitis, n = 41) versus non‐exposed (non‐mediastinitis, controls, n = 41) cohort with two endpoints: (i) obstruction of saphenous vein grafts (SVG) and (ii) obstruction of the internal mammary artery (IMA) grafts. The graft patency was evaluated with coronary CT‐angiography examination at a median follow‐up of 2·7 years. The number of occluded SVG in the mediastinitis group was 18·9% versus 15·5% in the control group. Using generalized estimating equations model with exchangeable matrix, and confounding effect of ischaemic time and patients age, we found no significant association between presence of mediastinitis and SVG obstruction [rate ratio (RR) = 0·96, 95% CI (0·52–2·67), P = 0·697]. The number of occluded IMA grafts was 10·5% in the mediastinitis group and 2·4% in the control group. Using the Poisson regression model, we estimated RR = 5·48, 95% CI (1·43–21·0) and P = 0·013. There was a significant association between mediastinitis and IMA graft obstruction, when controlling for the confounding effect of ischaemic time, body mass index, presence of diabetes mellitus and the number of diseased vessels. Presence of mediastinitis increases the risk of IMA graft obstruction. This may confirm the importance of inflammation as a major contributor to the pathogenesis of atherosclerosis and explain the negative effect of mediastinitis on a long‐term survival.

Introduction

The incidence of mediastinitis after coronary artery bypass grafting (CABG) is around 1%, and increases the risk of morbidity and mortality 1, 2, 3. During follow‐ up time of 10 years, survival of patients who suffered from mediastinitis was 49·5 ± 5·0% versus 71·0 ± 2·2% for patients without this condition, P < 0·01 3.

However, it is well established that atherosclerosis is an inflammatory disease 4. The inflammatory response itself may have a profound effect on lipoprotein movement within the artery, which is evident in the accumulation of macrophages at sites of plaque, plaque rupture and thrombosis. Considering the long‐term effect of mediastinitis on survival and the fact that atherosclerosis is mainly an inflammatory disease, we hypothesized that mediastinitis has a detrimental effect on patency of saphenous vein graft (SVG) and internal mammary artery (IMA) grafting after CABG. This effect must be early in the follow‐up time after operation.

Materials and methods

Study population

This study represents a multi‐centre collaboration among Rikshospitalet University Hospital, Feiring Heart Clinic, Oslo Heart Center, Akershus University Hospital and the University of Oslo. The clinical institutions maintain the same diagnostic criteria and treatment with active and prospective epidemiologic surveillance of hospital infections. Between September 2005 and April 2010, a total of 6·620 adult patients undergoing CABG surgery were considered as the source population for this study. In this period, patients suffering from deep sternal wound infection were treated with vacuum‐assisted closure (VAC). The diagnosis was based on the criteria established by the Centres for Disease Control and Prevention (www.cdc.gov) 5, 6.

Epidemiological design and statistical methods

This is a cohort study of 82 patients undergoing CABG surgery. The epidemiologic design is of an exposed (mediastinitis, n = 41) versus non‐exposed (non‐mediastinitis, controls, n = 41) cohort with two endpoints: (i) occlusion of SVG and (ii) occlusion of IMA graft (Figure 1).

Figure 1.

Design of the study, an exposed/non‐exposed cohort.

All CABG patients who had developed postoperative mediastinitis during the period of 4·7 years, from September 2005 to April 2010, were considered as exposed patients. The non‐exposed group was a random control sample of 41 patients without mediastinitis, collected from the same source population and the same period 7.

Evaluation of the endpoints was performed via coronary CT angiography (CTA) during a period of 9 months, from April to December 2010. Blinded to the status of the patients as far exposition to mediastinitis, the evaluation of the images was performed by two independent radiologists.

Median observation time from primary CABG to coronary CTA was 2·6 (range 0·5–5·2) years in the mediastinitis group and 2·8 (range 0·7–4·6) years in the control group, P = 0·87.

Endpoint SVG obstruction

SVG obstruction was defined as total occlusion of the vein graft between the last distal anastomosis and the proximal anastomosis. Patients with CABG usually have more than one SVG graft. A major issue in this study is that there is a correlation in graft obstruction within each patient, because of similar quality of saphenous veins and distal vessels grafted into similar flow dynamics and characteristics of patient blood. The hypothesis of this study is that mediastinitis leads to graft obstruction. Henderson et al. 8 pinpointed a method that control for this independence, which is based on the cluster sampling and the ratio estimates. This method could not control for the risk factors of graft patency, which have a confounding effect on this association. Other investigators such as Kleinbaum et al. 9 and Abdelnoor et al. 10 pinpointed clustering in occlusion of SVG. Generalized estimating equations (GEE) model was used to control for SVG occlusion clustering considering as a nuisance and the risk factors of SVG obstruction with confounding effect on the association between mediastinitis on the incidence of SVG obstruction. Exchangeable correlation matrix was considered to indicate that the correlation between different saphenous graft obstructions is assumed to be the same in a specific patient.

Endpoint IMA graft obstruction

IMA graft obstruction was defined as total occlusion (100%) of the graft from the subclavian artery to the anastomosis of left descending artery (LAD). For this endpoint, we do not have the problem of clustering nuisance as every patient had only one IMA graft. The quantification of the effect of mediastinitis on IMA graft obstruction was performed by the rate ratio (RR) and the patient time model. This will control for variability in follow‐up time for patients. The adjusted effect controlling for confounders and variability in follow‐up time was performed using Poisson regression model 11.

A priori power analysis

A research of the international and national literature was performed as far as estimation of graft patency at a mean follow‐up of 3 years after CABG. We considered two publications from the study by Goldman et al. 12, 13. In their publication 12, they showed that the 10‐year patency was 61% for SVG and 85% for IMA (P < 0·001).

As they defined occlusion as 100% stenosis (article page 2150), the estimated occlusion at 3 years of SVG was 23% and for IMA graft, 10%. Predictors of graft occlusion were non‐users of aspirin after bypass, younger age, higher serum cholesterol and higher Canadian functional class with highest occlusion rate after the first postoperative year.

The other study by Goldman et al. 13 has showed an IMA graft occlusion rate of 10·3% and an SVG occlusion rate of 19·7% for patient using aspirin at 3 years after CABG. Other studies with randomized design including 610 CABG patients at our hospital from the study by Eritsland et al. 14, 15 pinpointed 13% occlusion rate of IMA graft and 33% for SVG at 1 year follow‐up.

Extensive research of the literature was performed for the estimation of the strength of the association between mediastinitis and graft patency. Only one study, Grmoljez et al. 16, pinpointed a strong association between presence of mediastinitis and IMA graft occlusion at 1 year of follow‐up. It was estimated as [RR = 8·6, 95% CL (3·5–21·6), P = 0·001], no association with SVG occlusion was found [RR = 1·01, 95% CL (0·42–2·37), P = 0·995].

Taking all these results into consideration, we hypothesized that the incidence of IMA occlusion at 3 years of follow‐up will be around the mean of the three studies (10% + 10·3% + 13%)/3 = 11·1%. However, we hypothesized also that the magnitude of the effect of mediatinitis on IMA graft occlusion to be the lowest value of the 95% confidence interval as estimated by Grmoljez et al. namely 3·5. For a type I error of 5%, a power of 80% and an RR = 3·5, we need 37 in every group, a total of 74 patients. Because of possible dropout of patients for attending the CT imaging control, we increase the sample to 82 patients.

With regard to the incidence of SVG occlusion, we considered the mean values of the three studies, which was around (20% + 19·7 + 33)/3 = 24·2%, and we hypothesized a weaker strength of association of RR = 2·0, a type I error of 5% and a power of 80%; thus, we need 37 patients in every group, a total of 74 patients.

To perform this comparative study, we need around 41 patients with mediatinitis and 41 random control patients from the same source population. As the prevalence of mediastinitis is around 0·6% 3, we need a source population of 6620 patients with CABG over a period of 5 years.

CTA data acquisition and analysis

Coronary CTA examinations were performed with a 256‐multislice CT scanner (Brilliance ICT; Philips Medical Systems, Eindhoven, The Netherlands), with a volumetric covering of 8 cm per single gantry rotation and a gantry rotation time of 0·27s. Patients were examined in the cranio‐caudal direction, covering the region from clavicula to the diaphragm and performed as helical scan with retrospective ECG gating. Oral β‐blocking medication was administered 1 h before examination, if the heart rate exceeded 65 beats/min. Two blinded experienced observers analysed the CTA images by consensus. All the main coronary arteries, their secondary branches and the arterial and venous bypass grafts were evaluated for degree of stenoses. In the evaluation of the graft, the distal anastomosis was regarded as part of the graft, and the recipient vessels distal to the graft anastomosis as separate vessels. In case of an occluded graft, patency of coronary vessel sections proximal to the site of anastomosis of the graft was also evaluated.

The newest generation ultrafast multidetector CT scanners have an improved diagnostic accuracy around 93% compared with invasive coronary angiography as gold standard 17, 18, 19.

Surgical technique in the CABG operation

In all patients, CABG was performed through a median sternotomy, using cardiopulmonary bypass (CPB), systemic moderate hypothermia and antegrade crystalloid cardioplegia. Standard technique included routine use of left IMA, with supplemental vein grafts to obtain complete revascularization. Mediastinal shed blood was re‐transfused.

The patients received standard antibiotic prophylactics with four doses of intravenous (i.v.) cephalotin 2 g for at least 24 hours or until all drains or monitor lines had been removed.

Surgical revision of mediastinitis

All patients underwent VAC as a single‐line therapy followed by sternal rewiring. Initial revision was performed with removal of all sternal wires, and debridement of necrotic and infective bone tissue. To prevent right ventricle rupture, sternum was freed from the right ventricle, and paraffin gauze was placed in between.

Results

In this cohort study, the clinical profile of the two groups is summarized (Table 1). The table shows a difference between groups in frequency of unstable angina (UAP), body mass index (BMI), previous myocardial infarction (MI), diabetes, functional NYHA‐classification, left ventricular ejection fraction (EF), ventilation time and blood transfusion. This indicates the presence of potential confounders on the association between exposition to mediastinitis and graft obstruction. The mean number of coronary arteries involved was 2·7 (± 05) in the mediastinitis group, and 2·8 (±0·4) in the control group, p = 071.

Table 1.

Clinical profile of patients with mediastinitis and random control without mediastinitis

	Mediastinitis (n = 41)	Controls (n = 41)	P‐value*
Preoperative
Age (years)	67·0 (±9·7)	64·0 (±9·9)	<0·30
Male gender (%)	87·8	78·0	<0·13
NYHA(III–IV) (%)	68·2	49·1	<0·09
Hypertension (%)	64	61	<0·60
Three vessels disease (%)	75·0	80·4	<0·59
Unstable angina (%)	43·9	24·4	<0·05
Body mass index (kg/m2)	30·1 (±4·0)	27·0 (±3·9)	<0·01
Ejection fraction (%)	60 (14·5)	66·2 (9·8)	<0·29
Previous myocardial infarction (%)	65·9	34·1	<0·01
Preoperative atrial fibrillation (%)	17·1	9·8	<0·52
Chronic obstructive pulmonary disease (%)	22·0	14·6	<0·28
Diabetes (%)	41·5	7·3	<0·01
Operative
Aortic cross‐clamp time (minutes)	36·1 (±15·6)	32·5 (±13·9)	<0·32
Cardio‐pulmonary bypass time (minutes)	60·8 (±28·5)	50·6 (±18·5)	<0·12

Endpoint SVG obstructions

Our results showed that the number of occluded SVG grafts in the mediastinitis group was 10/53 (18·9%) versus 9/58 (15·5%) in the control group. Other investigators have found that SVG occluded in cluster within patients, and there is need to control for this interdependence (nuisance) in estimating the effect of mediastinitis on SVG graft occlusion.

Table 2 shows the results using GEE model to quantify the effect of mediastinitis on the incidence of SVG occlusion. We considered the exchangeable correlation matrix, indicating that the correlation between saphenous grafts was assumed to be the same in a specific patient.

Table 2.

Effect of mediastinitis on SVG obstruction controlling for nuisance and confounders using the GEE model at a median follow‐up of 2·7 years

Variables*	RR	SE	95% CI	P value
A. Crude effect
Presence of mediastinitis	1·17	0·49	0·52–2·67	0·697
B. Controlling for nuisance and confounders
Presence of mediastinitis	0·96	0·402	0·42–2·18	0923
Age (years)	096	0·0184		009
Ischaemic time (minutes)	1·03	0·0099		0·002

GEE, generalized estimating equation with counting data; SVG, saphenous vein grafts; RR, rate ratio.

^*The unit of age is 1 year and ischaemic time is 1 minute.

The effect of mediastinitis on SVG occlusion when controlling for nuisance was RR = 1·17 95% CI (0·52—2·68), P = 0·697. When controlling both for nuisance and confounding effect of age and ischaemic time, the GEE model gave the results as RR = 0·96 95% CI (0·42—2·18), P = 0·923.

Endpoint IMA graft obstructions

The number of occluded IMA grafts was 4/38 (10·5%) in the mediastinitis group and 1/41 (2·4%) in the control group. Using the Patient Years model to control for variability in follow‐up time between patients and groups, we estimated the RR, quantifying the association between presence of mediastinitis and incidence of IMA graft occlusion. Our results in Table 3 showed an RR of 4·5 with 95% CI (0·50—40·30) and mid‐exact P = 0·087. Using the Poisson regression model, controlling for variability in follow‐up time, confounding effect of ischaemic time, BMI, presence of diabetes, and number of diseased vessels, we estimated a final adjusted rate ratio using robust methods an RR of 5·48, 95% CI (1·43–210), P = 0·013. This indicates that patients with mediastinitis had five times higher risk of IMA graft occlusion than patients without this condition at a median follow‐up time of 2·7 years after CABG.

Table 3.

Effect of mediastinitis on IMA graft obstruction at a median follow‐up of 2·7 years

A. Crude rate ratio (RR) from the patient time model
	Mediastinitis
	Yes	No
IMA occluded	4	1
Patients (days)	36128	40694
Rate ratio = 4·5, 95% CI (0·50–40·3); mid‐exact P = 0·087}

B. Adjusted RR controlling for confounders using the Poisson model
Variables	Coefficient	SE*	P value	RR	95% CI
Mediastinitis, yes/no	−1·7029	0·6845	0·013	5·48	1·43–21·00
Ischaemic time (minutes)	−0·1455	0·0490	0·003
Body mass index (kg/m2)	−0·2584	0·0455	0·0001
Diabetes, yes/no	−0·6839	0·4274	0·0550
Number of diseased vessels	−1·2345	0·3957	0·002

^*Robust estimates.

Discussion

Internal validity considerations

Mediastinitis significantly increases the risk of IMA graft occlusion in patients undergoing CABG. First, we have to ask if an apparent association between presence of mediastinitis and IMA graft occlusion is real. It could be merely an artefact of the study, because of bias and random variation.

Primarily, we needed to evaluate the internal validity of this study. Our first attempt was to check that the association was not due to selection bias, residual confounding bias or information bias.

As the probability of selection bias is low, the control patients without mediastinitis were a random sample from the same source population as the case patients with mediastinitis, undergoing CABG surgery in the same period. In confounding bias, we controlled for BMI and ischaemic time, simultaneously using multivariate models, although unmeasured confounders are still a problem.

Considering information bias a misclassification of end point is always a possibility as ultrafast CT has a specific sensitivity and specificity as compared with the golden standard of invasive coronary angiography. Meanwhile, this misclassification is non‐differential between the two groups and would have weakened the rate ratio and introduced a bias towards the null effect 7.

We have, however, a concern as far the precision of the estimate of IMA graft occlusion. Our results showed an association with a rate ratio of 5·4 and a large 95% confidence interval, indicating a lack of precision of the estimate. An increase in the follow‐up time would have increased the incidence of IMA graft obstruction in both groups of the study, and would have generated shorter confidence interval for the rate ratio. It has been established that graft obstruction in the follow‐up time is related to different independent risk factors 12. They are the amount of aspirin use, the level of cholesterol, the age of the patients and Canadian functional class, and the diameter of the grafted vessels. All these factors could interact with mediastinitis during the follow‐up time in increasing the risk of graft occlusion.

Surgical considerations

However, IMA grafting adversely affects the sternal blood flow, predisposing the patients to infection 2. The left IMA is most often used in revascularization of the LAD artery, in order to supply the greatest territory of the heart.

Long‐term patients' outcome is directly related to coronary artery bypass durability 12. Inflammatory changes involve the sternum, the aortic wall, cardiac chamber and bypass grafts. Anatomically, the IMA graft is involved in this area of inflammation, hence the risks of thrombosis and increased tissue fragility. Vein grafts, running more laterally or to the posterior of the heart, are more distant from the inflammation area and are less exposed to damage during surgical revision. Coronary vasospasm is also one of the major causes of ischaemic heart conditions, and affects the patency rates of CABG, mostly the arterial grafts 20.

The VAC‐method is based on topical negative pressure through a sponge across the sternal wound to promote bacterial removal and improve wound healing 21. Although not confirmed by others one may speculate that this could lead to mechanical stress on IMA and to a lesser degree on SVG due to the distance.

Consistency of our results with other investigators

Temporality of this association is evident as mediastinitis is occurring before graft obstruction, but the consistency of our results with other investigations on different populations, places and period is not evident. Three investigators (Table 4) have tried to investigate this hypothesis.

Table 4.

Comparative results of studies of mediastinitis and graft patency after CABG

Authors	Graft nature	Patency mediastinitis (%)	Patency control	Follow‐up years
Grmoljez et al. 16	SVG (16 grafts)	75	75%*	1
	IMA (9 pats)	55	95%*
Macmanus et al. 22	SVG (8 grafts)	75	70%†	1·3
Mansuroglu et al. 23	SVG (17 grafts)	70·6	X‡	2·5
	Lima (16 pats)	100	X‡
This study	SVG (53 grafts)	81·1	84·5%§	2·7
	IMA (41 pats)	88·8	97·5%§

SVG, saphenous veins graft; pats; IMA, internal mammary artery; Time years, time from the operation to the angiography or CT; PATS, patients.

^*Controls are from the same source population as the cases.

^†The estimation of the patency in the control group is from the medical literature and not reported by the authors.

^‡There is no information on the control, their number and their patency rate.

^§Controls are a random sample from the same source population as cases.

Macmanus et al. 22 found a patency rate per graft of 6/8 = 75%, 95% CI (34·9–96·8%), for a mean follow‐up of 6·7 months. They compared this incidence with the anticipated patency of the total population of patients in their centre, which was 70% (as 70% was included in the 95% confidence limits interval). Their conclusion was that mediastinal infection does not necessarily adversely affect aortocoronary saphenous vein bypass graft patency.

Grmoljez et al. 16 found difference in obstruction incidence for IMA graft between two groups, but no difference in SVG graft obstruction. Their estimate of RR was close to our result, and their data showed that patients with mediastinitis had reduced patency of 55% (45% obstruction frequency), compared with the control patency of 95% (5% frequency of obstruction), which was statistically significant. Grmoljez et al. quantified this relative risk as RR = 8·6, 95% CI (3·5–21·6), it was stronger than the estimated of 5·4 adjusted for confounders and follow‐up variability.

Mansuroglu et al. 23 compared graft patency at a mean follow‐up of 2·5 years for a group of 16 CABG patients with mediastinitis to the graft patency of a selected control from the same source population. For the mediastinitis group, he found 100% patency for the IMA graft and 70·5% patency for SVG. Comparing his results with graft patency of patients without infection, they concluded that mediastinitis does not affect the graft patency rates adversely. It is important to underline that in this study, there was not enough indication on the nature of the control group considered, its source population and the period of time, and no adjustment for variables with confounding effect.

Biological plausibility of this association

Mediastinitis may lead to an accelerated coronary ischaemia and artery spasm with fibrosis and thrombosis, ultimately resulting in ischaemic cardiomyopathy and severe haemodynamic consequences. Biological mechanisms explain our findings, as it is well known that atherosclerosis is an inflammatory disease 4. The nature of the inflammatory response is on different level, the first one interaction among endothelial cells, monocytes and T cells, instability and rupture of plaque. Macrophage accumulation may be associated with increased plasma concentrations of both fibrinogen and C‐reactive protein (CRP), markers of inflammation, thought to be early signs of atherosclerosis. Plaque rupture and thrombosis may be responsible for obstruction of the IMA graft. Some investigators have pinpointed that SVG and IMA grafts behave differently, anatomically as well as physiologically. It has been highlighted a huge difference in the anatomic properties (endothelial fenestrations and internal elastic lamina) and physiological properties (flow reserve, nitricoxide/protacyclin production, vasomotor response to thrombin and number of fibroblast receptors) of IMA graft and vena saphena magna 24, 25. Considering these anatomical and physiological properties, it is not strange that a profound mediastinal inflammatory process has different effects on IMA and SVG grafts in the follow‐up time of patients with CABG with regard to graft thrombosis.

Conclusion

In early time of follow‐up after CABG, presence of mediastinitis increases the risk of IMA graft obstruction, but not the SVG. This may confirm the importance of inflammation as a major contributor of the pathogenesis of atherosclerosis, and explain why postoperative mediastinitis is a major contributor of reduced long‐term survival after CABG.