Prognostic factors in left-sided endocarditis: results from the andalusian multicenter cohort

All patients diagnosed of IE in seven hospitals in Andalusia (South of Spain) were consecutively registered in a uniform database from January 1984 throughout December 2006. Five are tertiary referral hospitals for cardiac surgery and 2 are community hospitals; the latter transferred patients evaluated as being at higher risk for mortality to the referral centres. All cases of left-sided IE defined according to Duke criteria [16] with later modifications [17] for definite and possible IE were included. Patients registered before 1994 were retrospectively evaluated for diagnostic criteria. In the case of relapses, only the first episode was included. Cases with insufficient follow-up (not longer than one month) were excluded.

Data were recorded prospectively by members of the Infectious Diseases services or units in the participating centres using a structured questionnaire and introduced into a common database. The cases were detected from routine review of microbiology, echocardiography and surgery reports. The primary outcome measure was overall mortality during hospitalisation (in-hospital death). All cases with no signs of infection and a negative post-antimicrobial treatment culture were considered cured. Relapses, defined as an IE episode occurring within 6 months of a previous one and caused by the same microorganism, were also recorded. Recurrence was defined as a new IE episode in the same patient and related to a different microorganism. The following independent variables were recorded: age, sex, health care-associated acquisition, type and severity of underlying comorbidity evaluated according to the age-adjusted Charlson index [18], duration of symptoms before diagnosis, septic shock, heart failure, embolism, neurologic manifestations, acute renal failure, echocardiographic findings (vegetations and size, abscess, prosthetic dehiscence and valvular dysfunction), microbiologic data (microorganisms isolated from blood cultures and/or valve tissues, and serologic tests), and treatment (medical and surgical). For analysis purposes, the study period was arbitrary divided in two: 1984 - 1995 and 1996 - 2006.

The additive euroSCORE index [19] were calculated for operative risk evaluation in every patient. Predisposing valvulopathy was defined as any previous valvular heart lesion that predispose to IE. Prosthetic endocarditis was considered as early if it occurred within one year of valve implantation and late otherwise. Heart failure was defined according to Framingham's criteria [20] and was evaluated at admission and during evolution. Septic shock was defined according to standard criteria [21]. Perivalvular extension was defined as the presence of an abscess and/or fistula and/or pseudoaneurism diagnosed by echocardiogram, in surgery or at autopsy. A major symptomatic embolism was considered if the central nervous system, spleen or extremities were affected. Neurologic manifestations were considered only in patients with clinical manifestations related to encephalopathy, meningitis, cerebral infarct or a haemorrhage, and demonstrated by cerebrospinal fluid analysis, radiological test or at autopsy. Severe renal insufficiency was defined as a serum creatinin level >2.5 mg/dl. Left-sided IE were considered nosocomially acquired using previously published criteria [22]. After 1994 transesophagic echocardiography (TEE) was performed in cases with a high clinical probability of IE when transthoracic echochardiography (TTE) was not diagnostic, and routinely in prosthetic IE or when cardiac complications were suspected.

This study was approved by ethical and research committees of all participating hospitals.

The association of the different variables with the outcome was firstly analysed by univariate analysis. Continuous variables were expressed like mean ± standard deviation or like median and interquartile rank in case of abnormal distribution and were analysed using the Student's T test or the Mann-Whitney U test as appropriate. Categorical variables were expressed like absolute numbers and proportion of cases and were compared using the chi-squared or Fisher's test when appropriate. Relative risks with 95% confidence interval for mortality were calculated. Then, multivariate analysis was performed using logistic regression. The multivariate model was built including all variables with a significant association in univariate analysis (p < 0.05) and those considered potentially clinically relevant; modification effects between variables were also studied. The selection of variables was performed using a stepwise backward procedure. All statistical studies were performed using SPSS software (v12.0, Chicago, IL, USA).

Seven hundred and five case-patients of left-sided IE were included in the study. Only other 8 patients were excluded because insufficient evolution data. Among included cases, 624 (88%) were considered as definite and 81 (12%) as possible IE. Mean (± SD) post-hospitalisation follow-up was 15.75 ± 25.05 months (range: 1-192). The median age of the patients was 56 years (interquartile rank from 41 to 68 years), only 5 patient were ≤ 15 years old. Four hundred and eighty six (69%) patients were male. Native valves were affected in 534 (76%) cases. Of 171 patients (24%) with prosthetic IE, 69 were considered early and 102 late. The aortic valve was affected in 320 (45%) cases, the mitral valve in 299 (42%), and both valves were affected in 86 (11%) patients. Globally, 488 (69%) had a predisposing valvulopathy for IE and 57 (8%) had suffered at least one previous episode of IE. The mean (± SD) age-adjusted Charlson index was 1.84 ± 2.27 (range: 0-10), and the most frequent underlying diseases were: diabetes mellitus (35 patients, 5%), non-valvular heart diseases (32 patients, 5%), chronic liver diseases (32 patients 5%); neoplasia (28 patients, 4%); chronic renal insufficiency (24 patients, 3%); and chronic lung disease (16 patients, 2%). Forty-six patients (7%) were intravenous drug abusers.

Aetiology was established by isolating the causative microorganism from blood culture in 598 (85%) patients, a surgically-explanted valve culture in 70 (10%) and by serologic test in 35 (5%). Streptococcus viridans group was the most frequent microorganism isolated, followed by Staphylococcus aureus, coagulase-negativestaphylococci and Enterococcus spp. (table 1). In 71 cases (10%), aetiology was unknown, because all microbiological studies were negatives.

Echocardiography was performed on 698 patients: TTE alone on 409 (58%), TEE alone on 16 (2%) and both TTE and TEE on 273 patients (39%). In 7 patients, no echocardiographic study was performed because endocarditis was not suspected until surgery or necropsy. Overall, TTE was diagnostic in 448 of 682 patients (66%) in which it was performed, and TEE in 251 of 289 (87%). In 156 patients where a TTE study was not diagnostic, a TEE study obtained a defined diagnosis of IE in 122 (78.2%).

Many patients developed complications: 282 (40%) had cardiac failure, 194 (27%) presented neurologic manifestations, 280 (40%) had systemic embolisms (in 147 cases, the central nervous system was affected), 190 (27%) developed acute renal insufficiency, 86 (12.2%) septic shock, and 87 (13%) perivalvular extension (perivalvular abscess, 70; fistula 15; and pseudoaneurism 2).

Valvular surgery was performed in 269 patients (38%). The most frequent indications for surgery were severe heart failure (in 171 patients, 63.5%), followed by valve dysfunction (37 patients, 13.7%), persistent sepsis (21 patients, 7.8%), and perivalvular extension (8 patients, 2.9%). There were several indications for surgery in 27 patients (10.0%). Fifty and three (7.5%) was not operated because of critical situation and high risk intervention.

Two hundred and eight patients died during their hospital stay (mortality rate, 29.5%). Relapses occurred in 17 patients (2.4%), and the 480 remaining cases (68%) were considered cured. Principal causes of death was: heart failure in 44 patients (21.5%) sepsis in 28 (13.4%) neurologic complications in 22 (10.5%), and by several of these complications in 71(34.1%).

The variables significantly associated with increased mortality in univariate analysis were age, nosocomial acquisition, severity of underlying disease, chronic liver disease, prosthetic valve endocarditis, early prosthetic valve endocarditis, S. aureus, Streptococcus agalactiae and fungal aetiology, lesser duration of symptoms before diagnosis, development of septic shock, progressive or refractory heart failure, acute renal insufficiency, the presence of cardiac abscess or fistula, and non performed surgery although indicated because of high associated risk (tables 2 and 3). Conversely, Streptococcus viridans group aetiology was associated with lower mortality rates. The final multivariate model is shown in table 4; the p value of the Hosmer-Lemeshow goodness-of-fit test for the final model was 0.514.