Clinical Presentation, Etiology, and Outcome of Infective Endocarditis in the 21st Century: The International Collaboration on Endocarditis–Prospective Cohort Study

Background  We sought to provide a contemporary picture of the presentation, etiology, and outcome of infective endocarditis (IE) in a large patient cohort from multiple locations worldwide.

Methods  Prospective cohort study of 2781 adults with definite IE who were admitted to 58 hospitals in 25 countries from June 1, 2000, through September 1, 2005.

Results  The median age of the cohort was 57.9 (interquartile range, 43.2-71.8) years, and 72.1% had native valve IE. Most patients (77.0%) presented early in the disease (<30 days) with few of the classic clinical hallmarks of IE. Recent health care exposure was found in one-quarter of patients. Staphylococcus aureus was the most common pathogen (31.2%). The mitral (41.1%) and aortic (37.6%) valves were infected most commonly. The following complications were common: stroke (16.9%), embolization other than stroke (22.6%), heart failure (32.3%), and intracardiac abscess (14.4%). Surgical therapy was common (48.2%), and in-hospital mortality remained high (17.7%). Prosthetic valve involvement (odds ratio, 1.47; 95% confidence interval, 1.13-1.90), increasing age (1.30; 1.17-1.46 per 10-year interval), pulmonary edema (1.79; 1.39-2.30), S aureus infection (1.54; 1.14-2.08), coagulase-negative staphylococcal infection (1.50; 1.07-2.10), mitral valve vegetation (1.34; 1.06-1.68), and paravalvular complications (2.25; 1.64-3.09) were associated with an increased risk of in-hospital death, whereas viridans streptococcal infection (0.52; 0.33-0.81) and surgery (0.61; 0.44-0.83) were associated with a decreased risk.

Conclusions  In the early 21st century, IE is more often an acute disease, characterized by a high rate of S aureus infection. Mortality remains relatively high.

Infective endocarditis (IE) is a disease characterized by high morbidity and mortality. Although first described in the mid-16th century, the Gulstonian lectures by Osler^1-3 to the Royal College of Physicians in 1885 created the impetus for the systematic study of IE. Beginning in the early 1900s, investigators have frequently reported on the manifestations of this disease.^4-11 However, despite advances during the past century in diagnosis,¹² medical therapy,¹³ and surgical treatment,^14,15 mortality rates have not changed substantially in the past 25 years.^5,9,16-18 The current in-hospital mortality rate for patients with IE is 15% to 20%,^5,16 with 1-year mortality approaching 40%.^16,18,19 This is in stark contrast to sustained and ongoing improvements observed in other cardiovascular diseases such as myocardial infarction.²⁰

Unfortunately, definitive studies of IE have been limited by its relative infrequency, a problem compounded by the wide range of causative organisms, at-risk populations, and underlying risk factors for infection. Most studies have consisted of case reports or single-center studies that limit the scope and statistical power necessary for definitive conclusions. Moreover, the lack of multinational studies has prevented an understanding of how geographic differences in patient characteristics and disease management affect outcome in patients with IE.

A prospective multicenter approach is essential for addressing the limitations associated with prior investigations of IE and, importantly, for examining therapeutic choices in a definitive way. Therefore, the International Collaboration on Endocarditis (ICE) was established to facilitate a multinational, multicenter approach to the study of IE. From this collaboration, the ICE–Prospective Cohort Study (ICE-PCS) was designed to assess the current characteristics of patients with IE. In this study, we describe this large cohort of patients, with particular emphasis on the current clinical presentation, microbial etiology, and outcomes of patients with IE.

The ICE began in June 1999. The ICE investigators later developed the ICE-PCS.²¹ Enrollment in ICE-PCS began on June 1, 2000, and for the purposes of this study was closed on September 1, 2005; the present study includes data from 58 sites in 25 countries.

All patients 18 years or older with IE from sites that met criteria for participation were included in the study. Sites had to meet the following criteria: (1) minimum enrollment of 12 cases per year in a center with access to cardiac surgery; (2) patient identification procedures in place to ensure consecutive enrollment and to minimize ascertainment bias²¹; (3) high-quality data, including query resolution; and (4) institutional review board and/or ethics committee approval or waiver based on local standards.

The ICE-PCS database is maintained at the Duke Clinical Research Institute, which serves as the coordinating center for the ICE studies, with institutional review board approval from Duke University School of Medicine.

Patients were prospectively identified at each site to ensure consecutive enrollment.²¹ A total of 3284 patients were enrolled into ICE-PCS, of whom 2781 had definite IE by the modified Duke criteria (Table 1).²² The 2781 patients with definite IE were included in this analysis.

A case report form of 275 variables was developed by the ICE group according to standard definitions.^21,23,24 Data were collected prospectively by site investigators during the index hospitalization and were then sent to the coordinating center for data entry or were entered directly by the site investigators through a secure Internet data entry system. Queries were developed on critical variables and were distributed to the sites for reconciliation. Once complete, the reconciled queries were returned to the coordinating center for final data entry.

Definitions of the variables included in the ICE-PCS case report form have been reported in detail elsewhere.²³ Community-acquired IE was defined as IE diagnosed at the time of admission (or within 48 hours of admission) in a patient who did not fulfill the criteria for health care–associated infection. Health care–associated IE was defined as nosocomial IE or nonnosocomial health care–associated IE. Nosocomial IE was defined as IE that developed in a patient who was hospitalized for more than 48 hours before the onset of signs or symptoms consistent with IE. Nonnosocomial health care–associated IE was defined as IE diagnosed within 48 hours of admission in an outpatient with extensive health care contact as reflected by any of the following criteria: (1) receipt of intravenous therapy, wound care, or specialized nursing care at home within the 30 days before the onset of IE; (2) attendance at a hospital or hemodialysis clinic or receipt of intravenous chemotherapy within the 30 days before the onset of IE; (3) hospitalization in an acute care hospital for 2 or more days in the 90 days before the onset of IE; or (4) residence in a nursing home or long-term care facility. In an effort to group centers according to geographic similarities, regions were defined as follows: North America (10 sites from the United States), South America (8 sites from Brazil, Argentina, and Chile), Europe (22 sites from Croatia, France, Germany, Italy, the Netherlands, Spain, Sweden, Ireland, Romania, Russia, Slovenia, and the United Kingdom), and other (18 sites from Australia, Israel, India, Lebanon, Malaysia, New Zealand, Singapore, Thailand, and South Africa).

Continuous variables are presented as medians with 25th and 75th percentiles. Categorical variables are presented as frequencies and percentages of the specified group. Univariable comparisons were made with the χ² test or Kruskal-Wallis test as appropriate. To account for the possibility that patients referred to study hospitals from other health care facilities may represent a different population than those who were admitted directly, data from the latter group only were analyzed separately where indicated.

We used a generalized estimating equation method to determine factors associated with in-hospital mortality. Age, sex, transfer from another health care facility, and variables found to have a univariable association with in-hospital mortality (P < .10) were entered into the final exploratory model. The generalized estimating equation method produces consistent parameter estimates that measure the association between in-hospital death and the baseline covariates while accounting for the correlation in outcomes of patients from the same hospital. Likelihood ratio tests were used to compare models with and without interaction terms. Final parameter estimates were converted to odds ratios with corresponding 95% Wald confidence intervals. The model was validated using the bootstrap procedure. Some 200 estimates were obtained by fitting the generalized estimating equation model to 200 data sets obtained by randomly selecting 2781 observations with replacement from the actual data. Bootstrap estimates were computed by averaging the 200 parameter estimates, and bootstrap confidence intervals were computed sorting the parameter estimates in ascending order and selecting the 5th estimate for the lower confidence limit and the 195th estimate for the upper confidence limit.

Statistical analyses were performed using commercially available software (STATA, version 8.2; StataCorp, College Station, Texas).

Patients were enrolled in ICE-PCS from the following regions: North America (n = 597 [21.5%]), South America (n = 254 [9.1%]), Europe (n = 1213 [43.6%]), and other (n = 717 [25.8%]). Baseline characteristics and predisposing factors are shown in Table 2. The median age of the cohort was 57.9 (mean, 56.5; interquartile range, 43.2-71.8) years. Most of the patients in the cohort (72.1%) had native valve IE, and most patients (77.0%) were admitted within 1 month of the initial signs of illness. The most common underlying condition was diabetes mellitus (16.2%), but 9.9% of the South American patients had diabetes, compared with 26.7% of North American patients. Similarly, less than 5% of patients from outside North America were receiving hemodialysis, compared with 20.8% of North American patients.

Predisposing conditions were common in patients with definite IE (Table 2). Although intravenous drug use remains important (9.8%), the most common predisposing conditions were related to valvular heart disease. Degenerative valve disease (eg, significant mitral [43.3%] and/or aortic [26.3%] valve regurgitation) was the most frequent native valve predisposing factor. In contrast, rheumatic heart disease was uncommon; only 92 patients (3.3%) had rheumatic mitral valve disease. Valvular predisposing conditions also included the presence of a prosthetic valve in 618 patients (22.2%).

Chronic intravenous access was as common as intravenous drug use in the overall cohort; 148 of 244 patients (60.7%) in this study with chronic intravenous access were from North America (Table 2).

Clinical and laboratory findings on admission are presented in Table 3. The classic signs that are often considered diagnostic for IE were infrequent.

In 2756 of the 2781 patients (99.1%), blood samples were cultured to determine the causative microorganism. Of the 310 patients (11.1%) with negative blood culture yields, 192 (61.9%) had received antibiotics within 7 days of the blood culture. In addition to blood culture information, serologic tests and valve cultures were performed in a minority of cases. Of the 2781 patients, 277 (10.0%) had cultures and serologic tests that were negative for IE.

The causative microorganisms isolated from blood cultures are shown in Table 4. Gram-positive organisms were predominant (81.5%), with Staphylococcus aureus accounting for 31.2% of all infections.Staphylococcus aureus was also the most common organism in each major risk group, including intravenous drug users and those with intracardiac devices (Table 5). Positive serologic tests for Coxiella burnetii were reported for 27 patients (17 from Europe, 2 from North America, 1 from South America, and 7 from other regions), but only 9 were reported to have reciprocal antibody titers of more than 800. Similarly, 22 patients had positive serologic tests for Bartonella species (18 from Europe, 1 from South America, and 3 from other regions), but only 3 were reported to have reciprocal antibody titers of more than 800. One case of infection was due to Tropheryma whippelii.

Staphylococcus aureus was the most common organism in 3 of 4 regions, whereas viridans group streptococci were the most common organisms isolated from patients in South America. The frequency of Streptococcus bovis–associated IE was much higher in Europe and South America compared with the other regions, and IE due to the group of bacteria consisting of Haemophilus species, Aggregatibacter (formerly Actinobacillus) actinomycetemcomitans, Cardiobacterium hominis, Eikenella corrodens, and Kingella species (HACEK bacteria) was relatively uncommon in North America. Most of the C burnetii and Bartonella infections were from Europe.

The location of acquisition was determined in 94.5% of patients; community acquisition (71.5%) was more common than nosocomial (13.7%) or nonnosocomial (9.3%) health care–associated IE in the total cohort (Figure). North America had a much higher proportion of health care–associated infections (38.1%) compared with other regions, mainly owing to a larger proportion with nonnosocomial health care–associated IE. The microbial causes of IE varied with location of acquisition, with a higher proportion who had staphylococcal IE and a lower proportion who had viridans streptococcal IE among those with health care–associated IE. Among patients with community-acquired infection, 34.3% had staphylococcal IE and 22.7% had viridans streptococcal IE, whereas the corresponding figures for nosocomial infection were 69.8% and 0.8%, respectively, and for nonnosocomial health care–associated infection were 67.4% and 4.3%, respectively.

Echocardiography was used in most patients (99.2%). More than one-half (59.0%) of the patients had undergone transthoracic and transesophageal echocardiography. Of the 2781 patients, 87.1% had echocardiographic evidence of vegetation, whereas new, significant valvular regurgitation was discovered in 63.8% of patients. Abscess was the most common paravalvular complication (14.4% of patients), whereas 34.1% of patients with prosthetic valve IE had evidence of a prosthetic valve complication such as dehiscence or new paravalvular regurgitation.

Congestive heart failure was the most common complication in all regions (Table 6). In general, the highest complication rates occurred in North America and Europe.

There were also geographic differences in treatment and outcome, although the magnitude of this variation was not large (Table 6). Surgical treatment was common for the entire cohort (48.2%), and in-hospital mortality was 17.7%. Table 7 shows the results of the regression modeling for in-hospital mortality with the estimates from bootstrap validation. The following variables were independently associated with an increased risk of in-hospital death: involvement of a prosthetic valve, increasing age, radiographic pulmonary edema, S aureus infection, coagulase-negative staphylococcus infection, presence of mitral valve vegetation, and paravalvular complications. Variables independently associated with a decreased risk of in-hospital death were elevated erythrocyte sedimentation rate (ESR), infection with a viridans group streptococcus, and surgery during the current IE episode. The estimates obtained by bootstrap validation were similar to those of the original model and support the validity of this model. Differences between models were noted for the following 4 variables: diabetes mellitus, health care–associated acquisition, coagulase-negative staphylococcus infection, and presence of a mitral valve vegetation.

Of the total cohort of patients with definite IE, 1174 (42.2%) had been transferred to a study hospital from another health care facility. Analysis of the data after excluding these patients revealed few differences from analysis of the whole cohort (Tables 2, 4, and 6). Notable differences were that transferred patients were more likely to undergo surgery (63.4% of transferred patients vs 37.1% of nontransferred patients [P < .001]) and were more likely to have congestive heart failure as a complication (39.3% vs 27.1% [P < .001]). In-hospital mortality (18%) and microbial etiology were similar for both groups of patients.

Despite more than a century of study and recent advances in diagnosis and treatment, IE remains an incompletely understood disease with high morbidity and mortality. Textbook descriptions of the clinical features and epidemiology of IE are still largely based on data obtained several decades ago. Lack of progress is partly related to the fundamental difficulty in studying this type of disease. By necessity, most studies are derived from case reports or small case series from single sites, with few large cohort studies or randomized trials. A shift in approach is necessary to further the understanding of endocarditis and to definitively study therapeutic choices. The ICE-PCS represents a new effort in broadening our understanding of endocarditis. To our knowledge, this study is by far the largest prospective cohort study of IE to date. The size of the cohort coupled with the multinational perspective has enabled several important observations to be made.

Our findings reveal that, in much of the world, IE is no longer a subacute or chronic disease occurring primarily in younger patients with rheumatic valvular abnormalities. In contrast, most patients in this investigation presented early and demonstrated few of the classic clinical findings traditionally associated with IE. For example, in the 1960s and 1970s, Osler nodes were recorded in 11% to 23% and splenomegaly in 20% to 44% of patients with IE.^9,10,25,26 In our study, predisposing valvular conditions were common but were primarily owing to the presence of degenerative valve disease or a prosthetic valve rather than rheumatic heart disease. Forty years ago, approximately 50% of cases of IE in the United States were superimposed on preexisting rheumatic lesions,²⁷ compared with less than 5% in the present study. Prosthetic valve endocarditis was present in one-fifth of our patients, as discussed in detail elsewhere.²⁴

An emerging population at risk for IE consists of patients with health care–associated infections. Overall, IE was attributed to a health care–related exposure in nearly 25% of the patients. These findings confirm those of recent reports from small single-center studies^16,28 and provide evidence that these population changes are occurring in many regions of the world. The health care setting will continue to gain importance in relation to complications such as IE, mainly owing to aging societies that rely on increasingly invasive medical care.^29,30

Our analysis has provided evidence of geographic differences for several important characteristics in patients with IE. For example, although the overall IE population characteristics were influenced by contact with health care services and medical interventions, this specific finding was not observed in the centers from South America. In addition, the association between health care–associated IE was most striking in North America.

Another observation arising from this investigation is the shift in the microbiologic characteristics of IE. Staphylococcus aureus is now the most common cause of IE in much of the world, confirming several recent investigations^5,16,31 and the earlier findings of the ICE-PCS.²³ This shift is due in part to the global presence of risk factors for S aureus–associated IE (eg, intravenous drug use, health care contact, and invasive procedures). Given the growing antimicrobial resistance of S aureus,³² including to vancomycin,^33-35 the importance of this pathogen as a potentially lethal infection is cause for concern.

We also noted a substantially higher prevalence of S bovis–associated IE in Europe, that HACEK-associated IE was relatively uncommon in North America, and that most cases of Q fever and Bartonella-associated IE came from Europe. Whether these findings reflect differences in patient characteristics, regional health care access, diagnostic bias, or other factors remains to be determined. For IE due to microorganisms that are difficult to culture, geographic differences may, at least partially, reflect variation in the threshold for performing additional diagnostic tests. This may be the case for Q fever and Bartonella-associated IE, which often rely on serologic and/or nucleic acid amplification tests for diagnosis.³⁶ However, it is also clear that there are geographic differences in the incidence of these 2 infections.³⁷

These changes in the patients and pathogens have important implications for the diagnosis and management of IE. For example, new risk groups have been identified that necessitate careful diagnostic attention in the presence of fever and bacteremia. In addition, the acute nature of IE in the modern era may require an accelerated evaluation strategy that provides the opportunity for early diagnosis and treatment decisions in patients at high risk for complications and death.

We have found several factors that were independently associated with in-hospital mortality. Some of these factors, such as increasing age, presence of pulmonary edema, and paravalvular complications, were not surprising. In addition, prosthetic valve IE and staphylococcal IE were also associated with an increased risk of in-hospital death, whereas there was a decreased risk associated with viridans streptococcal IE. An elevated ESR was associated with a decreased risk of death, although the reason for this is unclear. Elevated ESR may be associated with more chronic infection, thereby signifying a more chronic clinical course. We have found that early surgery may be critical in improving survival in patients with definite IE. This finding adds detail to recent reports supporting early surgical intervention^38,39 and adds credence to the practice of a combined medical and surgical approach from admission for patients with IE, specifically in those with congestive heart failure and prosthetic valve infections. Our finding that nearly 50% of patients had surgery indicates that the threshold for early surgical treatment has lowered.

This is an observational study of patients from centers with a particular interest in IE. These hospitals are typically referral centers with cardiac surgical programs. Consequently, the study population is unlikely to be a true population-based sample, thereby limiting epidemiologic inferences. This potential selection bias may be less evident in some sites (eg, New Zealand), where most cases of IE within the catchment area would be eligible for enrollment in the study. It might be expected that patients transferred from other health care facilities would represent a different population than those who presented directly to study hospitals. In particular, the former group may have more complicated disease and greater indications for surgery. However, when the 2 groups were compared, patients transferred from other facilities had characteristics similar to those presenting directly to study hospitals, with notable exceptions being that a larger proportion of the former group underwent surgery during their initial hospitalization and had congestive heart failure as a complication. Consequently, we believe it is important to present data from both groups of patients and that exclusion of referred patients may create a greater selection bias.

Although study sites spanned all non-Antarctic continents, there was a heavy weighting toward wealthy countries in Europe, North America, and Australasia, with few sites in Asia and Africa. There would undoubtedly be greater geographic differences in patient and microbiologic characteristics of IE if sampling was able to more closely resemble the global population distribution. The study lacked long-term follow-up of patients, thereby limiting the ability to analyze outcome beyond initial hospitalization. The precise timing of all complications was not recorded and may affect the ability to determine the clinical significance of some findings.

Infective endocarditis remains a serious and deadly disease despite recent advances in diagnosis and treatment. Of particular note, IE has shifted to a disease in which the presentation is more acute than previously described and, throughout much of the world, is characterized by a high rate of S aureus infection in patients with previous health care exposure. More care must be taken to effectively treat all patients with S aureus bacteremia and to identify patients with high potential for complications.⁴⁰ We have documented geographic differences in the presentation, microbial etiology, treatment, and outcome of patients with IE. In addition, we have found initial evidence that early surgery may be important in improving patient outcomes. Because nearly 50% of patients with IE undergo surgery, early identification of surgical indications may improve mortality. More research also needs to focus on stroke prevention (eg, surgical therapy for vegetations), the identification of the most effective therapy (eg, the role of new antibiotics and combination treatment), and understanding reasons for the high prevalence of S bovis–associated IE in Europe and the near absence of HACEK-associated IE in North America.

Correspondence: David R. Murdoch, MD, MSc, Department of Pathology, University of Otago, Christchurch, PO Box 4345, Christchurch, New Zealand (david.murdoch@cdhb.govt.nz).

Accepted for Publication: September 14, 2008.

Author Contributions: Dr Murdoch had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All of the authors have seen and approved the final version of the article. Study concept and design: Murdoch, Corey, Hoen, Miró, Fowler, Bayer, and Cabell. Acquisition of data: Murdoch, Hoen, Miró, Fowler, Olaison, Chambers, Holland, Jones, Klein, Raymond, Read, Tripodi, Utili, and Wang. Analysis and interpretation of data: Murdoch, Corey, Hoen, Miró, Fowler, Bayer, Karchmer, Olaison, Pappas, Moreillon, Chu, Falcó, Holland, Klein, Tripodi, Utili, Woods, and Cabell. Drafting of the manuscript: Murdoch, Olaison, and Holland. Critical revision of the manuscript for important intellectual content: Murdoch, Corey, Hoen, Miró, Fowler, Bayer, Karchmer, Pappas, Moreillon, Chambers, Chu, Falcó, Holland, Jones, Klein, Raymond, Read, Tripodi, Utili, Wang, Woods, and Cabell. Statistical analysis: Murdoch and Pappas. Obtained funding: Wang. Administrative, technical, and material support: Murdoch, Hoen, Fowler, Chambers, Chu, Holland, and Jones. Study supervision: Murdoch, Hoen, Olaison, Holland, Tripodi, and Utili.

Financial Disclosure: Dr Fowler has received grant or research support from Cubist, Cerexa, Merck, Theravance, Inhibitex, Nabi, and the National Institutes of Health; has been a paid consultant for Cubist, Inhibitex, Leo Pharm, Merck, and Johnson & Johnson; has been on the speaker's bureau for Cubist; has received honoraria from Arpida, Astellas, Cubist, Inhibitex, Merck, Theravance, and Ortho-McNeil; and has been a member of an advisory committee or review panel for Cubist.

Funding/Support: This study was supported in part by grants AI-068804 (Dr Fowler) and K23 HL70861-01 (Dr Cabell) from the National Institutes of Health; grant AHA BGIA 0265405U from the American Heart Association (Dr Cabell); grants REIPI RD06/0008 and FIS 05/0170 from the Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III, Madrid (Spain)–Red Española de Investigación en Patología Infecciosa (Dr Miró); the Fundación Privada Máximo Soriano Jiménez (Dr Miró); and the Institut d’Investigacions Biomèdiques August Pi i Sunyer and the Conselleria de Salut de la Generalitat de Catalunya (IDIBAPS) (Dr Miró).

Role of the Sponsors: The sponsors played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.

Additional Contributions: In addition to all of the named ICE investigators at each site, the personnel at each site and at the coordinating center provided support for this study that has allowed the project to move forward.