Vacuum‐assisted closure therapy for deep sternal wound infections: the impact of learning curve on survival and predictors for late mortality

Abstract

The aim of this study was to evaluate the possible learning curve effects on survival during the introduction of vacuum‐assisted closure (VAC) therapy in patients with deep sternal wound infection (DSWI). Furthermore, predictors of late mortality were analysed and causes of late death were examined. Fifty‐three patients (early Group, n = 26, January 1999 to July 2001 versus late group, n = 27, August 2001 to March 2003) were all treated with VAC for DSWI. A follow‐up was carried out in September 2006. Multivariate analyses were used to evaluate the predictors of late mortality. The 90‐day mortality was 0% in both groups. The survival rates at 5 years were 69·2 ± 9·1% (early group) versus 58·5 ± 11·7% (late group), P = ns (non significant). The time interval from cardiac surgery to diagnosis of DSWI and prolonged VAC therapy were identified as independent predictors of late mortality. Our concept for VAC therapy in DSWI seems to be readily introduced in clinical practice. There was no difference in survival between our initial cases and later cases. Late diagnosis and prolonged wound therapy were identified as predictors for late mortality.

Introduction

Deep sternal wound infection (DSWI) following cardiac surgery is a serious complication with a reported early rate of mortality between 5% and 25% 1, 2. During recent decades, several forms of treating DSWI, such as closed irrigation, pectoral muscle flaps and omental flap techniques, have been introduced 3, 4, 5. The use of these conventional wound healing techniques is now well established but still associated with unsatisfactorily high early mortality rates, recurrent infections or flap‐related complications 6, 7.

Vacuum‐assisted closure (VAC) therapy is a relatively new modality in wound healing management, and recent studies have demonstrated promising results when VAC is applied to DSWI 8, 9, 10, 11, 12. Furthermore, the results of a recent study by our research group suggested an improved long‐term outcome after VAC‐treated DSWI (13). However, there are always obstacles when establishing new methods in the surgical community, even if the treatment gives superior results. Surgeons are usually somewhat reluctant to abandon established techniques with a well‐known outcome in favour of new surgical techniques (14). This may, at least partly, be explained by the learning curve, which may be reflected in poor outcome during initial procedures. Previous studies regarding learning curve effects in surgical procedures have suggested a plateau in learning after initial cases (15).

Therefore, the aim of the present study was to evaluate the results of our initial cases during the introduction of VAC therapy and to compare them with later cases, once the clinical concept of VAC in DSWI had been established at our clinic. Our intention was to identify possible learning curve effects with negative impact on survival. We also evaluated several predictors of mortality after VAC‐treated DSWI and identified causes of late death.

Patients and methods

Patient population and data collection

Between January 1999 and April 2003 53 patients were treated with VAC therapy for culture‐verified DSWI following cardiac surgery at the Department of Cardiothoracic Surgery in Lund University Hospital. The incidence of culture‐verified DSWI during this time period was 0·9% at our department (8). DSWI was diagnosed prior to discharge or after readmission from the referring hospital. Our department was the only cardiothoracic surgery centre in the region during the period studied, and patients suffering from sternal infections were readmitted to our hospital. DSWI was defined according to the guidelines of the Center for Disease Control and Prevention (16); however, only patients with both positive tissue cultures and clinical signs of DSWI were included in the present study. Patients with negative substernal tissue cultures at the first revision or superficial sternal wound infections not requiring reoperation with removal of the sternal wires were not included.

Preoperative variables, including the EuroSCORE, for our initial 26 patients (early group, January 1999 to July 2001) and the following 27 patients (late group, August 2001 to March 2003) were collected in the department database in a prospective manner and are presented in Table 1. The majority (36/53) of the patients underwent coronary artery bypass grafting (CABG). Other surgical procedures in the early group included aortic valve replacement (one patient), aortic valve replacement with concomitant CABG (three patients), mitral valvuloplasty (one patient), mitral valvuloplasty with concomitant CABG (one patient), Ross procedure (one patient) and a reconstruction of the right outflow tract with a pulmonary artery homograft (one patient). In the late group, following procedures, except CABG, were included: aortic valve replacement (one patient), aortic valve replacement with concomitant CABG (five patients), pseudarthrosis procedure following CABG (one patient), heart transplant (one patient) and mitral valvuloplasty with concomitant CABG (one patient). The medical records of all patients showing DSWI were also reviewed retrospectively, and information on risk factors considered to be relevant to poor wound healing, such as diabetes mellitus, obesity and systemic immunosuppressive therapy, was collected from the patients’ medical records. The presence of bilateral mammary artery grafts was not included since this technique was practised very rarely at our department. The DSWI was also classified according to the criteria proposed by El Oakley and Wright (17). Our standard prophylactic antibiotic regimen during the relevant period was three doses of isoxazolyl penicillin, 2 g intravenously on the day of operation and on the first postoperative day. When DSWI was diagnosed, the antibiotic therapy usually commenced with vancomycin administered intravenously, and this continued until the results of the tissue cultures became available. The antibiotic therapy was then adjusted according to bacterial sensitivity and bacterial strain.

Table 1.

Preoperative characteristics in the early and late VAC‐treated groups

Variable	Early group		Late group		P value
	n	(%)	n	(%)
Number of patients	26	49	27	51
Age* (years)	67·8 ± 9·5		68·2 ± 11·4		ns
Open heart surgery
CABG only	18	69	18	67	ns
Other procedure(s)	8	31	9	33	ns
Female gender	8	31	6	22	ns
Diabetes mellitus	14	54	11	41	ns
Obesity (BMI >30)	12	46	12	44	ns
COPD	7	27	4	15	ns
LVEF <30%	6	23	6	22	ns
Renal failure†	5	19	1	4	ns
Preoperative stroke	5	19	1	4	ns
MI <90 days preoperation	11	42	10	37	ns
Emergency surgery	2	8	5	19	ns
El Oakley class ≥III	22	85	15	56	<0·05
EuroSCORE*	7·7 ± 4·4		7·3 ± 3·5		ns

BMI, body mass index; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; LVEF, left ventricular ejection fraction; MI, myocardial infarction; ns, non significant; VAC, vacuum‐assisted closure.

^* Mean ± standard deviation.

^† Serum creatinine >200 μmol/l (2·27 mg/dl).

VAC therapy

In January 1999, the standard therapy at our department in cases of DSWI was changed from conventional treatment, for example closed irrigation, pectoral muscle flaps and omental flap techniques, to VAC therapy. The VAC therapy was not subject to any significant changes during the study period. During surgical revision, sternal wires were removed and five substernal tissue cultures were sent for microbiological analysis prior to debridement. Sharp sternal edges were debrided with a sharp spoon and necrotic bone was removed if necessary, but extensive sternectomy was avoided. Following haemostasis, three or four layers of paraffin gauze dressing (Jelonet^®; Smith and Nephew Medical Ltd, Hull, UK) were placed at the bottom of the wound covering the right ventricle, bypass grafts and visible lung tissue. Both sternal edges were dissected free from underlying tissue, along their full length, with a 1‐ to 2‐cm margin laterally, and the paraffin gauze was carefully placed under the sternal edges. This was performed to further protect the heart and to facilitate sternal rewiring later on. A polyurethane foam dressing (KCI, San Antonio, TX) was placed between the sternal edges. A second layer of polyurethane foam was placed subcutaneously, covering the wound completely. This layer was secured subcutaneously to the surrounding skin edge with a running suture (Dermalon 2‐0; Davis and Geck, St Louis, MO). The open wound was sealed with a transparent adhesive drape (KCI), and the wound was connected to a vacuum source (KCI) by a tube. A continuous negative pressure of −125 mmHg was applied. The majority of patients were extubated immediately after VAC therapy initiation and left the intensive care unit after 2–3 hours. The polyurethane foam and protective layers were changed three times a week until tissue cultures were negative, and the C‐reactive protein levels were between 50 and 70 mg/l, and the wound was covered with granulation tissue. The sternotomy was rewired with interrupted steel wires (Stahldraht, Johnson and Johnson, Brussels, Belgium) and partial Robicsek wires were used to secure sternal fractures. All DSWI‐related surgical procedures (revision of the wound, including initiation of VAC therapy, procedures including changing the polyurethane foam and the final rewiring of the sternum) were performed in the operating theatre under general anaesthesia.

Follow‐up

The protocol for the present study was approved by the Ethics Committee for Clinical Research at Lund University, Sweden. The length of hospital stay was calculated according to Domkowski and coworkers (10), that is after the onset of DSWI if it occurred during the same period of hospitalisation as the initial cardiac surgery or the period of hospitalisation because of separate admission for DSWI. Follow‐up was performed in September 2006 and included a total of 238 patient‐years (mean 4·5 ± 1·8; range 0·3–7·6 years). No patient was lost to follow‐up. The mortality data during follow‐up were provided by the National Board of Health and Welfare in Sweden.

Statistical analysis

Continuous variables are expressed as means ± standard deviation. Categorical variables are presented as absolute values in addition to percentages. Univariate analysis for continuous variables was conducted with Student’s t‐test. Categorical variables were analysed using the chi‐squared test, except when expected frequencies were lower than 5, when Fisher’s exact test was used. Multivariate analysis was performed using stepwise Cox regression analysis to determine the independent predictors of late mortality after VAC‐treated DSWI. The inclusion criterion for the full model was P < 0·20, and the limit for stepwise backward elimination was P < 0·10. Survival was plotted using the Kaplan–Meier method, and the log‐rank test was used to compare the difference in survival between the two groups. Statistical analysis was performed, and graphs were plotted with the Intercooled Stata version 9.2 statistical package (Stata Corporation, College Station, TX). A P value <0·05 was considered statistically significant.

Results

The 90‐day mortality was 0% in both the early group (n = 26) and the late group (n = 27). There was no significant difference in patient characteristics between the early and late groups, except that there were significantly more patients with severe DSWI (El Oakley class ≥III) in the early group (Table 1). The majority of tissue cultures (39/53) from patients with VAC‐treated DSWI demonstrated coagulase‐negative Staphylococci strains, either as a single pathogen or in combination with other bacteria.

All 53 patients (100%) treated with VAC for DSWI (early group and late group) underwent rewiring of the sternum without the need for pectoral muscle or omental flaps. The early group underwent a mean of 3·3 ± 3·2 DSWI‐related surgical procedures (median 2; range 2–17), and the late group underwent a mean of 3·2 ± 1·2 procedures (median 3; range 2–7), P = ns (non significant). The DSWI‐related procedures included initial revision of the wound with simultaneous application of VAC therapy, changing the polyurethane foam and final rewiring of the sternum. No major VAC‐related complications occurred during therapy.

The mean period between cardiac surgery and the diagnosis of DSWI was 20·1 ± 13·9 days (median 15; range 3–71) in the early group and 12·9 ± 6·0 days (median 11; range 4–27) in the late group (P = 0·02). There was no significant difference (P = ns) in the number of days of active VAC therapy between the early group (12·6 ± 12·7 days, median 9·5; range 2–66 days) and the late group (11·2 ± 4·9 days, median 10·5; range 4–21 days). The number of days during which VAC therapy was applied in patients with DSWI (early group and late group), according to year of surgery (1999–2003), is shown in Figure 1A. The total length of stay in hospital in patients with DSWI (early group and late group) was 28·6 ± 23·3 days (median 20·5; range 8–103 days) and 21·7 ± 8·4 days (median 21; range 9–44 days), respectively (P = ns) (Figure 1B).

Figure 1.

(A) Duration of vacuum‐assisted closure (VAC) therapy for deep sternal wound infection presented according to year of surgery (1999–2003) (* indicates outlier). (B) Total length of stay in hospital of patients treated with VAC according to year of surgery (1999–2003) (* indicates outlier).

There were a total of 18 deaths during late follow‐up in the study population (early group and late group). The causes of death are listed in Table 2. There was no significant difference (P = ns) in overall survival between the two groups (Figure 2). The rates of survival at 1 and 5 years were 92·3 ± 5·2% and 69·2 ± 9·1% in the early group, and 92·6 ± 5·0% and 58·5 ± 11·7% in the late group, respectively. The univariate analysis (Table 3) did not identify any significant difference in pre‐, peri‐ and postoperative variables between patients who died during follow‐up (late death) and patients who still alive (control). However, the multivariate analysis identified two independent predictors of late mortality after VAC‐treated DSWI: number of days from cardiac surgery to DSWI diagnosis (odds ratio 1·07; 95% confidence interval 1·03–1.11; P = 0·0001) and prolonged VAC therapy (odds ratio 1·13; 95% confidence interval 1·05–1·21; P = 0·001).

Table 2.

Cause of late death after VAC‐treated deep sternal wound infection

Cause of late death	n
Total	18
Cardiac causes
Heart failure	2
Myocardial infarction	4
Non cardiac causes
Stroke	2
COPD	1
Pneumonia	2
Prostate carcinoma	1
Pancreatic carcinoma	1
Subarachnoidal haemorrhage	1
Ruptured abdominal aneurysm	1
Gastrointestinal embolism	1
Urosepsis	1
Pancreatitis	1

COPD, chronic obstructive pulmonary disease; VAC, vacuum‐assisted closure.

Figure 2.

Overall survival in the early group versus the late group (P = non significant).

Table 3.

Results of univariate analysis of predictors of death after VAC‐treated deep sternal wound infection

Variable	Late death		Late survivors		P value	OR	95% CI
	n	(%)	n	(%)
Number of patients	18		35
Age* (years)	70·5 ± 10·1		66·7 ± 10·5		0·204	NA	NA
EuroSCORE*	8·3 ± 4·3		7·1 ± 3·7		0·297	NA	NA
Time to diagnosis (days)*	19·4 ± 16·9		13·7 ± 5·0		0·065	NA	NA
Duration of VAC therapy* (days)	15·2 ± 14·7		10·2 ± 4·4		0·068	NA	NA
Total length of stay* (days)	31·2 ± 24·7		22·0 ± 11·7		0·070	NA	NA
Open heart surgery
CABG only	11	61	25	71	0·446	0·629	0·16–2·51
Other procedure(s)	7	39	10	29	0·446	0·629	0·16–2·51
Age >80 years	3	17	4	11	0·678	1·55	0·20–10·4
Female gender	3	17	11	31	0·333	0·44	0·68–2·07
Diabetes mellitus	8	44	17	49	0·776	0·85	0·23–3·07
Obesity (BMI >30)	8	44	16	46	0·930	0·95	0·26–3·44
COPD	6	33	5	14	0·105	3·00	0·62–14·7
LVEF <30%	3	17	9	26	0·730	0·58	0·88–2·84
Renal failure†	3	17	3	9	0·397	2·13	0·25–17·6
Preoperative stroke	3	17	3	9	0·397	2·13	0·25–17·6
Peripheral artery disease	5	28	6	17	0·478	1·86	0·37–8·78
Emergency surgery	3	17	4	11	0·678	1·55	0·20–10·4
MI <90 days preoperation	7	39	14	40	0·938	0·95	0·25–3·51
El Oakley class ≥III	13	72	24	69	0·784	1·19	0·30–5·34
VAC changes ≥4	7	39	9	26	0·323	1·84	0·45–7·23

BMI, body mass index; CABG, coronary artery bypass grafting; CI, confidence interval; COPD, chronic obstructive pulmonary disease; LVEF, left ventricular ejection fraction; MI, myocardial infarction; NA, not applicable; OR, odds ratio; VAC, vacuum‐assisted closure.

^* Mean ± standard deviation.

^† Serum creatinine >200 μmol/l (2·27 mg/dl).

Discussion

The present study suggests that VAC therapy can be safely implemented in a full‐service cardiac surgery centre for the treatment of DSWI. We found no negative learning curve effects on early or late survival. Furthermore, we identified two independent predictors of late mortality (number of days from cardiac surgery to DSWI diagnosis and prolonged VAC therapy).

A learning curve effect is usually defined as an improvement in performance over time. The improvement tends to be most rapid at first and then tails off. The effect of a learning curve may cause poor outcome initially (18). It has previously been shown that this phenomenon may cause a problem when designing randomised studies because surgeons are reluctant to randomise until they are proficient in a technique. However, once they are convinced of the value of a technique, surgeons argue that it is too late to randomise (14). In addition, it can be difficult to measure the learning curve effect in a surgical procedure, as variables that measure patient outcome (e.g. survival after VAC‐treated DSWI) are usually studied together with variables that reflect the progress of the surgical process, such as number of foam changes during VAC therapy or the number of days of active VAC therapy.

In our series, the 90‐day overall mortality was 0%, and all 53 patients had their sternum rewired without soft tissue flaps. The late group showed significantly fewer days from cardiac surgery to DSWI diagnosis, which may reflect a more proactive approach to signs of DSWI at our department resulting from the promising early results of VAC. There were no significant differences in the number of VAC‐related procedures between the two groups, but during the initial use of VAC, two patients in the early group were treated with VAC for a long time and underwent several foam changes (Figure 1A). In one case, VAC therapy was used as a salvage procedure after previous failure of conventional treatment, and this demanded long intensive care and a prolonged hospital stay. In the other early case, an adult patient with a complex congenital heart disease suffered from DSWI after redo surgery. In the initial patients, several foam changes were made mainly because of the lack of data at that time on outcome after VAC therapy monitored with C‐reactive protein (CRP) levels. This particular case reflects a common uncertainty among surgeons when introducing VAC therapy in clinical practice, namely, when is it safe to stop the VAC treatment and rewire the sternum without recurrent mediastinitis? Recurrent mediastinitis is a very difficult and dangerous clinical condition with poor clinical outcome (19). The outcome data in the present study support previous studies by our research group, suggesting that the C‐reactive protein level should be between 50 and 70 mg/l to safely rewire with a very low risk of recurrent infection (20). We found no significant difference in the number of days of active VAC therapy or the total length of stay between the early group and the late group, but there was a trend towards a shorter length of stay in hospital during the period studied (Figure 1B). The initial prolonged length of stay was probably because of an initial desire to monitor the patient cautiously, in order to identify early signs of recurrent infection.

The multivariate analysis of patients in Table 3 identified the number of days from cardiac surgery to DSWI diagnosis and prolonged VAC therapy as independent predictors of late mortality following VAC‐treated DSWI. Both these factors probably reflect aggressive infections in patients predisposed to poor wound healing because of comorbidity. Such infections have a high El Oakley class (17) and usually a poor prognosis.

VAC therapy per se is not a complicated procedure surgically compared with conventional treatment involving the transfer of pectoral muscle or omental flaps. However, the procedure should still be performed with care, as should any surgery performed on the heart and large vessels and should preferably not be performed by surgeons without sufficient experience. In order to avoid major complications, we have consequently used the same treatment algorithm since January 1999, including an organ‐protecting barrier (i.e. JeloNet^©) placed under the sternal edges. Right ventricle rupture, a potentially fatal complication, is well described in previous literature during treatment for DSWI with conventional methods 21, 22. This devastating complication has also been described during VAC‐treated DSWI 23, 24. However, this may be explained by an inappropriate surgical technique and the absence of an organ‐protecting barrier, such as paraffin gauze or a silicon dressing. The use of an organ‐protecting barrier prevents the heart from adhering to the VAC foam and also protects the free right ventricular wall from sharp sternal edges.

A limitation on the present study may be extrinsic factors not related to the procedure, which may cause possible bias because this is not a randomised study. However, randomised studies are not always feasible in the surgical setting because of practical and ethical reasons. The generalisability of our results can be discussed; however, the patients studied represent a case mix in a full‐service department on tertiary level, including acute cardiac surgery, heart transplant procedures and grown‐up congenital heart surgery.

Conclusions

In our opinion, VAC therapy is an excellent adjunct for wound healing in DSWI without significant learning curve effects on early or late mortality, and no differences in the cause of late deaths between the two groups could be identified. However, we observed a trend towards a shorter length of stay towards the end of the study period that may indicate an increased confidence in the method.