Surgical treatment of infective endocarditis in active intravenous drug users: a justified procedure?

From January 1993 to July 2013 a total of 451 patients underwent surgery for IE at University Hospital Heidelberg. Out of this cohort, 20 active IVDU patients were identified and included in the study. The study was approved by the Ethics Committee, Medical Faculty University of Heidelberg, application number S-198/2013. Patients were only included in the study if the endocarditis was considered to be “definite” according to the modified criteria proposed by Duke[7, 8] and there was documented acknowledgement by the patient of active intravenous drug abuse (heroin) until the day of admission. Additionally, other drug use history was obtained including type of drug (marijuana, cocaine, amphetamines, ecstasy and methadone). Patient demographics, preoperative medical history, intra-operative and post-operative course were collected retrospectively from hospital medical records. Valve-related complications were documented according to the guidelines for reporting morbidity and mortality after cardiac valvular surgery[9]. Transesophageal echocardiography was performed in all survivors. Mortality was determined using the bureau of vital statistics database if such information was unobtainable from the medical records.

The primary endpoint in the study cohort was survival at 90 days after surgical treatment in our institution. The secondary endpoints were perioperative clinical characteristics and adverse events, which could have an impact on early postoperative mortality. All the patients completed a follow-up period of at least 90 days and were divided into two groups depending on 90-day survival. The demographic and perioperative variables of the 90-day survivors and non-survivors were compared to identify the predictors of 90-day overall mortality.

The diagnosis of IE was made using a combination of clinical and laboratory findings and was confirmed by transesophageal echocardiography in all patients according to the modified Duke criteria. Mandatory preoperative studies included abdominal ultrasound or computer assisted tomography to rule out septic foci. Absolute indications for surgical treatment during the active phase of IE, according to our overall surgical principles at the University Hospital of Heidelberg, included the following: progressive cardiac failure, unmanageable infection despite antibiotic therapy, recurrent embolic events, acute renal failure and prosthetic valve endocarditis. Each patient in our cohort met at least one of these criteria.

All patients underwent surgery with cardiopulmonary bypass with mild hypothermia. Myocardial protection was achieved with antegrade cold crystalloid cardioplegia. Tricuspid valve surgical procedures were performed with occlusion of both caval veins while the heart was beating or fibrillating.

Our treatment policy for IE is radical debridement of infected tissue followed by generous irrigation with antibiotic solution. All infected or necrotic structures surrounding the valves were resected irrespective of whether the conduction system was damaged or large defects in the myocardial or fibrous tissue were created. The valvular annulus and adjacent structures were carefully examined with the aim of detecting extension of the infective process. Abscesses were resected as completely as possible so as not to leave any macroscopically infected tissue in situ. Large evacuated abscess cavities were filled with gentamycin-fibrin glue. Ventricular septal defects caused by extension of abscess cavities were closed with Dacron patches. Allografts were preferred for aortic root reconstruction in cases of annular destruction and periannular invasion. If there was a requirement to buttress the suture line of a prosthetic valve, autologous pericardium was the material of choice. Defects were subsequently closed with pericardial patches or if possible, by direct suturing. Generally, the use of foreign material was kept to a minimum.

If the valves had to be replaced, the type of valve implanted was generally determined by the preference of the surgeon. Prostheses, patch material, and sutures were soaked in gentamicin solution before implantation.

In addition to hemodynamic stabilization, an important goal of postoperative intensive therapy was to control the local and systemic infective process. In general, the antibiotic therapy was specifically directed against the infecting microorganisms. In the presence of methicillin-resistant staphylococci, vancomycin was used in combination with tobramycin or rifampicin. Methicillin-susceptible staphylococci were treated with a semisynthetic penicillinase-resistant penicillin and rifampicin. If no rapid disappearance of infective signs could be achieved, the therapy was changed to vancomycin alone or in combination. If no microorganisms could be isolated, broad spectrum antibiotic therapy with imipenem and vancomycin or teicoplanin was used. In each case, intravenous antibiotic therapy was maintained for 6 weeks postoperatively.

Post-operative transesophageal echocardiography was performed regularly to exclude recurrent vegetations or paravalvular leaks. These findings in combination with evidence of persisting infection would give a strong suspicion of ongoing or recurrent endocarditis and per se would justify re-operation.

Data are presented as continuous or categorical variables. Continuous data were evaluated for normality using the one sample Kolmogorov-Smirnov test. As all continuous variables were normally distributed, they were analyzed with the Student t-test and expressed as the mean ± standard deviation. Pearson’s χ ² or Fisher exact tests were used for categorical data dependent on the minimum expected count in each cross tab. The categorical data are expressed as total numbers and percentages. The ANOVA with repeated measurements was applied for the comparison of survivors and non-survivors regarding perioperative inflammatory markers. Kaplan-Meier actuarial survival estimate was generated to analyse survival of the entire cohort. Multivariate logistic regression analysis was performed on univariate predictors for 90-days mortality with an entry criterion of p < 0.05. All data were analyzed using the Statistical Package for Social Sciences, version 21.0 (SPSS Inc., Chicago, Illinois).

The mean age of the entire population was 35 ± 7.7 years and 35% (n = 7) of the population were female. All patients in the study cohort were active IVDUs and had advanced heart failure with the mean NYHA stage of 3.4 ± 0.9. One patient from the cohort had previous aortic valve endocarditis and had undergone a previous aortic valve replacement. Further past medical history included active Hepatitis B (70.6%, n = 12), renal insufficiency (35%, n = 7), insulin dependent diabetes mellitus (5%, n = 1) and history of cerebrovascular event (5%, n = 1). The mean duration of i.v. drug abuse was 15.2 ± 10.6 years. The affected valves were aortic (20%, n = 4), mitral (10%, n = 2), tricuspid (40%, n = 8), combined aortic and mitral (10%, n = 2), combined aortic and tricuspid (5%, n = 1), combined mitral and tricuspid (10%, n = 2) and combined aortic, mitral and pulmonary (5%, n = 1) valves. Eighty per cent of the cohort (n = 16) had positive blood cultures and 85% (n = 17) had valve vegetations confirmed by transesophageal echocardiography. The mean size of the valve vegetations was 25 ± 9.4 mm. The surgery was performed through a median sternotomy using cardiopulmonary bypass (CPB) with a mean CPB time of 139 ± 78 min and cross clamp time of 92 ± 49 min. In cases of valve replacement, mechanical or biological prostheses were used in 55% (n = 11) and 40% (n = 8), respectively. In five percent of cases (n = 1) the valve was reconstructed. Moreover, one patient with severe aortic valve/root destruction had several abscess cavities including periannular invasion that were filled with gentamycin-fibrin glue and subsequently closed with pericardial patches. After that an allograft was anchored in the reconstructed aortic root. The other patient with extensive, infiltrative aortic and tricuspid valve endocarditis needed closure of a ventricular septal defect with a Dacron patch caused by extension of abscess cavities. Large parts of the walls of the atria and the right ventricle were destroyed by the infection, so that additional reconstruction with Dacron patches had to be performed. In 12 patients splenectomy was performed simultaneously for severe abscess formation, which was diagnosed preoperatively.

Thirty five per cent of cases (n = 7) were urgent, 60% (n = 12) were emergent and one case was performed as a salvage procedure. Perioperatively, patients received 1440 ± 991 mL of red blood cells, 410 ± 428 mL of fresh frozen plasma and 287 ± 330 mL of platelets. The mean operation duration was 236 ± 96 min. Postoperatively, all patients were transferred to the intensive care unit (mean stay 3.9 ± 3.1 days). Having achieved hemodynamic stability and after extubation, the patients were transferred to the high dependency unit (mean stay 12.6 ± 14 days) before they were discharged to normal surgical ward.

The mean follow-up was 2504 ± 1842 days. The overall survival of the patient cohort was 90% at one year, and remained at 85% from 3 years to the end of follow-up (Figure 1). Seventeen patients survived until the end of follow-up, whereas 3 patients died at post-operative day 6, 8 and 591 days. The cause of death was severe sepsis in the first two cases and sudden cardiac death in the third. Eleven patients experienced recurrent IE during follow-up and were managed with i.v. antibiotics without the need for surgery. The most common postoperative adverse events were atrial fibrillation in 65% (n = 13), atrioventricular block in 25% (n = 5) with the need for permanent pacemaker implantation in 10% (n = 2), renal failure in 35% (n = 7), coagulation disorder (disseminated intravascular coagulation) in 20% (n = 4) and a cerebrovascular event (ischemic stroke) in 5% (n = 1). No myocardial infarction or deep sternal infections occurred postoperatively in the study cohort.

Staphylococcus aureus was the pathogen most commonly identified from blood cultures accounting for 40% (n = 8) of infections, followed by Enterococcus spec. 10% (n = 2), coagulase negative staphylococci 10% (n = 2) and streptococci 20% (n = 4). 20% (n = 4) of patients were culture-negative. The mean follow-up for the entire cohort was 2504 ± 1842 days. No patients were lost to follow-up.

A subgroup analysis of 90-day survivors and non-survivors is presented in Tables 1 and2. There were no statistically significant differences with respect to preoperative demographics and most clinical baseline characteristics including patients’ preoperative blood culture profile. However, non-survivors had significantly higher EuroSCORE II (p = 0.036), higher NYHA class (p = 0.005) and a significantly higher rate of HIV infection (p = 0.047). Also, non-survivors had a significantly longer operating time (p = 0.016) and higher RBC transfusion requirement (p = 0.020). Figures 2,3, and4 show the postoperative profile of white cell count (WCC), CRP and body temperature reflecting patients’ perioperative inflammatory status. In this respect, 90-day non-survivors had significantly higher body temperature over the first week postoperatively compared to 90-day survivors, while there were no significant differences regarding WCC and CRP levels. However, none of the univariate predictors reached statistical significance in multivariate logistic regression analysis.

The main limitation of this study is its retrospective single-arm, non-randomized design involving a limited number of patients from a single institution leading to a low statistical power. Given the rare nature of the condition, it was not possible to get a large number of patients and so, a prospective multi-centre long-term study should be conducted to further characterize IE in IVDUs in a larger patient sample. In spite of these limitations, we were able to include consecutive patients and were able to retrieve complete data records from the majority. Theoretically, it would have been desirable to compare outcomes of study patients with a medically treated control group although this was deemed to be ethically infeasible; all patients were deemed to be too sick for further antibiotic therapy and thus were treated surgically. The onset of IE did not necessarily occur at the time of the initiation antibiotic therapy. As most patients were admitted by external hospitals, and many of them were treated by their primary physicians, the exact onset of the endocarditis could not be defined in each patient.