Mediastinitis After Coronary Artery Bypass Graft Surgery
Risk Factors and Long-term Survival
Background Mediastinitis is a severe complication of coronary artery bypass graft surgery (CABG). The purpose of the present study was to determine preoperative and intraoperative variables that predict mediastinitis and to determine the impact of this complication on long-term survival.
Methods and Results Data on 20 preoperative and intraoperative variables were collected prospectively on 6459 consecutive patients who underwent CABG between January 1987 and January 1994. Eighty-three patients (1.3%) developed mediastinitis postoperatively, and a total of 24 patients (29%) died. Multivariate analysis identified 4 of the 20 variables as highly significant independent predictors for the development of mediastinitis: obesity (P=.0002), New York Heart Association congestive heart failure class (P=.002), previous heart surgery (P=.008), and duration of cardiopulmonary bypass (P=.05). A comprehensive review of the literature identified 13 other studies that evaluated 48 factors as predictors of mediastinitis; these data were critically analyzed and compared with the results from this series. In this series, postoperative interval mortality during the first 90 days after surgery for the patients with mediastinitis was 11.8% compared with 5.5% for the patients without mediastinitis. Interval mortality between 1 and 2 years after surgery remained high for the mediastinitis group (8.1%) relative to the nonmediastinitis group (2.3%). These differences were not eliminated by adjusting for important variables that influenced late survival in this population.
Conclusions The present study and a review of the literature suggest that obesity and duration of surgery are the most important predictors of mediastinitis. Furthermore, although the early increase in mortality has been well described, the present study documents for the first time that mediastinitis has a significant negative influence on long-term survival independent of the patient’s preoperative condition.
Mediastinitis after cardiac surgery continues to represent an important complication associated with tremendous morbidity and cost.1 2 The exact mechanism by which mediastinitis develops is unknown and multifactorial. In the present report, a wide variety of preoperative and operative variables are examined with multivariate analysis to determine independent predictors of mediastinitis. A total of 6459 consecutive patients were analyzed; all data were collected prospectively. In an effort to define consensus risk factors for the development of mediastinitis, a comprehensive review of the literature was performed, and results from this analysis were compared with results from previous reports.1 2 3 4 5 6 7 8 9 10 11 12 13
Modern management of mediastinitis with early, aggressive debridement followed by delayed closure has been reported to reduce early mortality to less than 20%.2 8 The effect of mediastinitis and its current treatments on long-term survival has not been investigated; whether long-term sequelae exist in patients who survive acute therapy is unknown. Therefore, we examined whether the rates of mortality up to 2 years after surgery were different for patients who developed mediastinitis than for patients who did not have this complication. Mortality rates in the two groups were adjusted with the use of a Cox proportional hazards model for a number of variables that influenced late survival; this adjustment helped to identify the influence of mediastinitis on long-term survival.
The study group consisted of 6459 consecutive patients who underwent both cardiac catheterization and CABG at Duke University Medical Center between January 1987 and January 1994. Patients undergoing CABG with other procedures (valve repair, valve replacement, and so on) were included in the analysis. Twenty preoperative and intraoperative variables were collected prospectively by either the cardiologist, cardiology fellow, cardiac surgeon, or cardiac surgery fellow. Data were recorded at the time of catheterization or at the time of the operative procedure. For most variables, data were missing for fewer than 0.5% of patients. For three variables (renal failure, COPD, and other comorbid conditions), data were collected for only 3 years (1991 through 1993).
Age, ejection fraction, duration of preoperative stay, and duration of CPB were analyzed as continuous variables. Internal mammary artery graft (none, unilateral, or bilateral) and the NYHA CHF class (0-IV) were analyzed as graded variables. The Coronary Artery Disease Index, a previously described grading system for coronary angiograms ranging from 0 (least severe disease) to 100 (most severe disease), was also included in the analysis.14 Hemostasis at the time of closure was graded by the surgeon as either adequate or poor. Operation type was categorized as CABG alone versus CABG with other cardiac procedures. Obesity was defined as a weight-to-height ratio of more than 50% above the ideal ratio (Metropolitan Life tables). DM included both insulin-dependent and non–insulin-dependent types. History of prior cardiac surgical procedure was included as a variable. Preoperative insertion of an intra-aortic balloon pump and cardiogenic shock (defined as a preoperative cardiac index of <2.0 L · min−1 · m−2) were included as parameters of acuity. Peripheral vascular disease was defined as the presence of symptomatic disease (claudication, rest pain, and so on). COPD included patients whose disease limited daily activity. Renal failure was defined as a serum creatinine level of more than 2.0. The variable comorbid condition included the presence of one or more other pulmonary, gastrointestinal, or hepatic diseases or any malignant disease process that interfered with the patient’s daily life or was likely to limit life expectancy. In addition, rates of mediastinitis for the five attending surgeons who performed the most cases were compared with the use of Student’s t test with Bonferroni’s correction.
Patients who developed mediastinitis were retrospectively identified through medical records coding, which required that the attending physician confirm the diagnosis. These records were then checked against the hospital infection control records that had been maintained independently by the Division of Infectious Disease. Charts of both the mediastinitis group and the uncomplicated group were then sampled for review. All patients in the mediastinitis group had purulent deep sternal wound infections requiring extensive debridement and drainage. Superficial wound infections that did not involve the sternum and did not require extensive debridement and drainage were not classified as mediastinitis.
All variables were initially evaluated with univariate logistic regression analysis as predictors of mediastinitis. Variables with univariate P<.1 value were then used in a multivariate analysis. Variables were considered independent predictors of mediastinitis if final multivariate P≤.05. A model was generated with the three variables whose final multivariate P values were most significant; this model was validated with a standard boot-strapping method.
Survival for both the mediastinitis and nonmediastinitis groups was expressed and compared with the use of a Kaplan-Meier plot. Complete follow-up was obtained on all patients with mediastinitis except for two patients who refused further follow-up at 6 and 18 months, respectively. With a Cox proportional hazards model, Kaplan-Meier plots were then adjusted for seven variables that had been previously demonstrated to be determinants of late survival for patients with coronary artery disease14 : age, ejection fraction, Coronary Artery Disease Index, presence of cerebrovascular or peripheral vascular disease, angina, mitral regurgitation, and acute myocardial infarction. The adjusted Kaplan-Meier plot included only the patients for whom all seven of these variables were recorded. Interval mortality rates were also calculated for both groups by dividing the number of deaths in a given postoperative interval by the number of patients living at the beginning of the interval for whom follow-up through the given period was available.
To further study the contribution of preoperative and operative risk factors in the development of mediastinitis, a comprehensive review of the literature was performed by searching the National Library of Medicine’s MEDLINE database with the use of the cd plus search engine with the Ovid interface. The years 1966 to October 1994 were searched by exploding and combining the terms “heart surgery,” “heart,” and “heart disease”; this set was then combined with a second set generated with the terms “sternotomy,” “sternum,” and “mediastinitis.” From this search, 63 articles were identified that were investigations of risk factors for mediastinitis after adult cardiac surgery. These studies were reviewed and excluded if 10 or fewer cases of mediastinitis were reported or if patient enrollment, data analysis, or variables studied were incompletely defined. A total of nine observational and four case-control studies were included in the final review. A total of 33 preoperative and 15 intraoperative variables were analyzed in these 13 studies.
Eighty-three of 6459 patients studied (1.3%) developed postoperative mediastinitis; 73 of the 83 mediastinitis patients (88%) were treated with soft tissue flaps. The average postoperative length of stay was 25 days for the mediastinitis group compared with 10 days for the nonmediastinitis patients. Table 1⇓ describes the prevalence of the preoperative and intraoperative variables studied. The majority of patients were male and underwent CABG with a single internal mammary graft without other procedures.
Rates of mediastinitis by attending surgeon ranged from 0.49% to 2.31%. These differences were compared with the use of Student’s t test with Bonferroni’s correction and were not statistically significant (P>.05). Attending surgeon was, therefore, not included in the multivariate analysis.
Seven variables were associated with increased risk of mediastinitis by univariate analysis (P<.1). In order of significance, these were obesity, NYHA CHF class, DM, prior heart surgery, duration of CPB, presence of comorbid conditions, and poor hemostasis at the time of closure (Table 2⇓). These seven variables were then included in a multivariate logistic regression analysis. Four independent predictors of mediastinitis were identified: NYHA CHF class (P=.002), obesity (P=.0002), prior heart surgery (P=.008), and duration of CPB (P=.05). As prior heart surgery and duration of CPB were closely linked variables, two final models were generated with NYHA CHF class, obesity, and either prior heart surgery (Table 3⇓, top) or duration of CPB (Table 3⇓, bottom). Higher CHF class related to the incidence of mediastinitis in a near-linear manner from class 0 through IV. For asymptomatic patients, for example, the incidence of mediastinitis was less than 1%, whereas patients with class IV CHF had a more than 3% incidence.
Median duration of follow-up was 1.5 years for the mediastinitis patients and 3.4 years for the nonmediastinitis patients. During the period of follow-up, a total of 24 of the mediastinitis patients died (28.6%). The Figure⇓ (A) illustrates the Kaplan-Meier survival curve for the mediastinitis and nonmediastinitis patients. The Figure⇓ (B) also illustrates adjusted mortality for both groups. Interval mortality rate was also determined and remained higher for the mediastinitis group throughout the 2-year postoperative follow-up period (Table 4⇓).
Review of the literature identified nine observational studies and four case-control studies that analyzed risk factors for the development of mediastinitis after open heart surgery.1 2 3 4 5 6 7 8 9 10 11 12 13 A total of 33 preoperative and 15 intraoperative variables were analyzed in these studies; both univariate and multivariate analysis results are listed (Table 5⇓). Studies used either univariate or both univariate and multivariate analyses to determine risk factors. Included in Table 5⇓ is the present study, which represents the largest group of mediastinitis patients in which both univariate and multivariate analyses were used to determine risk factors.
The exact mechanism by which mediastinitis develops is unknown and multifactorial. Intraoperative wound contamination has been conclusively demonstrated in a small number of cases15 16 and probably represents an important source of many infections. In addition, a variety of patient characteristics have been associated with an increased incidence of mediastinitis, suggesting that certain factors may predispose patients to the development of this complication.1 2 3 4 5 6 7 8 9 10 11 12 13 These risk factors, however, have varied widely among different studies, resulting in confusion. Many of the studies evaluating different variables were conducted on small numbers of patients with mediastinitis and therefore have limited applicability. Frequently, only univariate analysis was used, and the independent ability of statistically significant risk factors to predict the development of mediastinitis was not evaluated. In addition, some studies with relatively small patient populations used inappropriately large numbers of variables in their multivariate models. Furthermore, postoperative conditions, which are themselves complications, were analyzed as potential predictors of mediastinitis. To help identify true independent predictors of mediastinitis and avoid further confusion, we included a large group of patients and used a multivariate analysis to study preoperative and operative variables as potential predictors of mediastinitis. In addition, a comprehensive review of the literature is presented, and the results of our series are evaluated relative to these previous studies.
The most important independent predictor of mediastinitis in this study was the presence of obesity, and other studies strongly support this finding (Table 5⇑, top). Four other studies using multivariate analysis evaluated obesity as a potential risk factor; three concluded that it was an independent predictor of mediastinitis. Nagachinta et al4 identified obesity as the strongest predictor, with 14% of the obese patients developing mediastinitis. Loop et al2 identified obesity as the second most important independent predictor of mediastinitis, defining obesity as only a 20% increase in height-adjusted weight. The exact mechanisms by which obese patients are predisposed to wound infection are not known. Perioperative antibiotics are generally not adjusted for the increased volume of distribution in these patients at this institution, and tissue drug levels may be inadequate. Furthermore, deep skin folds in obese patients make skin preparation difficult and probably at times inadequate. The inframammary creases and the lower extremity vein harvest site can be particularly difficult areas for preparation and require special attention. Large amounts of adipose tissue probably also serve as a better substrate for wound infection. Attention to sterile technique and efforts to reduce the length of the procedure are probably critical in obese patients in whom the risk of mediastinitis is increased.
Prior heart surgery and prolonged CPB time were closely linked variables in our study; both were significant independent predictors of mediastinitis. Four other studies have evaluated prior heart surgery: by univariate analysis, three of the four found it to be predictive. However, only one of these studies found it to be predictive by multivariate analysis (Table 5⇑). Repeat median sternotomy generally requires more extensive dissection because of mediastinal adhesions, and this may increase the length of the procedure and the risk for intraoperative wound contamination. Repeat median sternotomy and CPB time are both markers for the overall duration of the procedure, and two other large studies2 10 found parameters that assess duration of the procedure to be important predictors by multivariate analysis. Based on the results of the present study and from these two other large studies, duration of the procedure is probably an important independent predictor of mediastinitis.
Another independent predictor of mediastinitis in this study was NYHA CHF class. In the review of the literature, only two other studies evaluated CHF (Table 5⇑, top), and one small study concluded that it was not predictive of mediastinitis. However, in a large cooperative Veterans Administration study, CHF was one of only two variables that predicted increased risk of mediastinitis.3 Furthermore, the American Society of Anesthesiology score, which includes CHF as a component, has also been shown to be an important predictor of postoperative wound infection.17 Together, these data suggest that CHF may be an important risk factor for mediastinitis.
Bleeding at the time of closure was only a significant univariate predictor for the development of mediastinitis in this study. However, in the two previous studies, parameters of bleeding were significant predictors by multivariate analysis (Table 5⇑, bottom). Although the degree of bleeding at the conclusion of surgery is difficult to quantify, these data suggest that bleeding may predispose a patient to the development of mediastinitis. This effect, however, is probably less important than the obesity or the duration of the procedure.
DM in this study was found to be a predictor by univariate but not multivariate analysis. Loop et al2 similarly concluded that DM was not a predictor of mediastinitis by multivariate analysis.2 However, three of five other studies that evaluated DM found it to be a predictor of mediastinitis by multivariate analysis (Table 5⇑, top). In two of these three studies, obesity was simultaneously evaluated and found to be a considerably stronger predictor than DM.4 9 These data suggest that DM predisposes to the development of mediastinitis but probably is less of a risk factor than duration of the procedure or obesity.
In the present study, bilateral IMA grafting was not a predictor of mediastinitis by either univariate or multivariate analysis. Furthermore, the combination of bilateral IMA grafting and DM was not predictive of mediastinitis (Table 2⇑). Only two other studies2 6 used multivariate analysis to determine whether bilateral IMA grafting was a risk factor for mediastinitis. In a very large series, Loop et al2 demonstrated that in the absence of other risk factors, bilateral IMA grafting was not a risk factor for mediastinitis. In a much smaller series, Grossi et al6 concluded that bilateral IMA grafting was a risk factor. Of note is that techniques for IMA takedown may vary between institutions and may affect the preservation of collateral blood supply to the sternum.6 Bilateral IMA grafting probably has a minimal effect on the development of mediastinitis, and its influence on the development of mediastinitis may be important when other risk factors are also present,2 but in general, denying a patient this procedure because of this association is not warranted.
Other reports associated the incidence of postoperative mediastinitis with different intraoperative techniques. For example, more discriminate use of electrocautery was reported to markedly decrease the incidence of mediastinitis.18 In this study comparing five different surgeons, each with a slightly different operative technique, statistically significant differences in the rates of mediastinitis were not present. In addition, in the review of the literature, surgeon and surgical experience were not reported as significant risk factors for mediastinitis (Table 5⇑, bottom). This suggests that although differences in technique may affect the incidence of mediastinitis, this effect is probably overshadowed by other, more powerful risk factors.
The model generated in the present report allowed the risk of mediastinitis to be calculated for individual patients. Patients with NYHA CHF class III or IV, obesity, or prior heart surgery each account for ≈5% of the total CABG population (Table 1⇑). The model-predicted incidence of mediastinitis for patients with one of these risk factors is approximately three times greater than that of patients with no risk factors. Patients with all three of the risk factors identified by the model had a predicted incidence of mediastinitis almost 20 times greater than patients with no risk factors.
Although many studies have reported the in-hospital or early mortality associated with mediastinitis, the independent effect of mediastinitis and its treatment on long-term mortality and morbidity has not been described. This is important because a variety of treatment strategies are available, and appropriate follow-up care for survivors has not been determined. For each of the four postoperative intervals studied in this report, patients with mediastinitis experienced much higher raw mortality. For example, interval mortality between 1 and 2 years after surgery was almost four times greater for the mediastinitis group (8.1% versus 2.3%) (Table 4⇑). After adjustment for common risk factors for late mortality among CABG patients, mediastinitis remained a significant independent predictor of late mortality (Figure⇑, B). This study is the first to demonstrate that patients with mediastinitis have significantly increased late mortality, independent of other factors that affect mortality in this population. These data confirm the suggestion by Loop et al2 that mediastinitis has a long-term negative influence on survival. Despite control of sepsis and acute wound management, the effects of mediastinitis on cardiac, pulmonary, and renal functions are probably significant, and this condition is best viewed as a chronic medical condition. Finally, these data suggest the need to evaluate treatments for mediastinitis not only acutely but also with regard to late morbidity and mortality.
Selected Abbreviations and Acronyms
|CABG||=||coronary artery bypass graft surgery|
|COPD||=||chronic obstructive pulmonary disease|
|IMA||=||internal mammary artery|
|NYHA CHF||=||New York Heart Association congestive heart failure|
- Received March 14, 1995.
- Accepted May 3, 1995.
- Copyright © 1995 by American Heart Association
Hammermeister KE, Burchfiel C, Johnson R, Grover FL. Identification of patients at greatest risk for developing major complications at cardiac surgery. Circulation. 1990;82(suppl IV): IV-380-IV-389.
Nagachinta T, Stephens M, Reitz B, Polk BF. Risk factors for surgical wound infection following cardiac surgery. J Infect Dis. 1987;156:967-973.
Smith JM, Glaser RS, Osborne BJ, Buckley DC, Rath R, Schreiber JT. Sternal wound complications after open heart surgery: results from 3524 consecutive operative procedures. Contemp Surg. 1993;43:197-202.
Smith LR, Harrell FE, Rankin JS, Califf RM, Pryor DB, Muhlbaier LH, Lee KL, Mark DB, Jones RH, Oldham HN, Glower DD, Reves JG, Sabiston DC Jr. Determinants of early versus late cardiac death in patients undergoing coronary artery bypass graft surgery. Circulation. 1991;84(suppl III):III-245-III-253.
Gaynes R, Marosok R, Mowry-Hanley J, Laughlin C, Foley K, Friedman C, Kirsh M. Mediastinitis following coronary artery bypass surgery: a 3 year review. J Infect Dis. 1991;163:117-121.
Culver DH, Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG, Banerjee SN, Edwards JR, Tolson JS, Henderson TS, Hughes JM, and the National Nosocomial Infections Surveillance System. Surgical wound infection rates by wound class, operative procedure and patient risk index. Am J Med. 1991;91(suppl 3B):3B-152-3B-157.