Infective Endocarditis Epidemiology Over Five Decades: A Systematic Review

Aims To Assess changes in infective endocarditis (IE) epidemiology over the last 5 decades. Methods and Results We searched the published literature using PubMed, MEDLINE, and EMBASE from inception until December 2011. Data From Einstein Medical Center, Philadelphia, PA were also included. Criteria for inclusion in this systematic review included studies with reported IE microbiology, IE definition, description of population studied, and time frame. Two authors independently extracted data and assessed manuscript quality. One hundred sixty studies (27,083 patients) met inclusion criteria. Among hospital-based studies (n=142; 23,606 patients) staphylococcal IE percentage increased over time, with coagulase-negative staphylococcus (CNS) increasing over each of the last 5 decades (p<0.001) and Staphylococcus aureus (SA) in the last decade (21% to 30%; p<0.05). Streptococcus viridans (SV) and culture negative (CN) IE frequency decreased over time (p<0.001), while enterococcal IE increased in the last decade (p<0.01). Patient age and male predominance increased over time as well. In subgroup analysis, SA frequency increased in North America, but not the rest of the world. This was due, in part, to an increase in intravenous drug abuse IE in North America (p<0.001). Among population-based studies (n=18; 3,477 patients) no significant changes were found. Conclusion Important changes occurred in IE epidemiology over the last half-century, especially in the last decade. Staphylococcal and enterococcal IE percentage increased while SV and CN IE decreased. Moreover, mean age at diagnosis increased together with male:female ratio. These changes should be considered at the time of decision-making in treatment of and prophylaxis for IE.


Introduction
Infective endocarditis (IE) extols a high cost for society worldwide, with a US incidence of 10,000 to 15,000 cases each year [1]. IE is associated with prolonged hospitalization, can require surgery [2], and impairs quality of life [3]. IE was initially described in 1885 by Osler [4] as a disease of patients with preexisting valvular abnormalities. Since then, notable improvements in IE diagnosis and treatment have been made. However, in-hospital mortality is still close to 20 percent [5,6].
Risk factors for IE have changed over time. There have been widespread changes in health-care delivery in the last five decades, which have impacted the clinical spectrum of IE.
These include the use of intracardiac devices [7], prosthetic valves [8], hemodialysis [9], and an increase in the elderly population [10]. Furthermore, changes in antibiotics have led to alterations in patterns of infection and bacterial resistance in both the US [11,12] and Europe [13].
In recent decades, several studies have noted an increase in the proportion of IE caused by staphylococcal species [14,15]. However, others have not [16]. A systematic review of population-based studies including 15 studies and 2,371 cases found no significant changes in the causative organism over time [17]. However, significant limitations were present in this study, including a low power to detect changes; nor did it cover the last decade. Moreover, to the best of our knowledge, there are no systematic reviews of hospital based studies.
Proper understanding of IE epidemiology is paramount, as different organisms produce varied complications and may require different treatment and prophylaxis [18]. The objective of this research was to assess whether there have been changes in IE epidemiology globally over the last half century. Towards this end, we performed a systematic review of both population and hospital-based studies.

Data Sources and Searches
We searched, with no language restrictions, PubMed, OVID/ MEDLINE and EMBASE electronic databases from their inception to December 2011, for studies reporting infective endocarditis microbiology. We used the term 'infective endocarditis' for the Mesh keyword. Date last search performed was December 1 st , 2011. We supplemented the search with references from articles reviewed and correspondence with other researchers, including experts in the field. When a reference was deemed potentially suitable for inclusion, a fulltext copy was obtained and reviewed according to predefined criteria (listed below). We followed the PRISMA guidelines for systematic reviews.

Study Selection and Data Collection
Prospective and nonprospective studies reporting the frequency distribution of infective endocarditis in the last five decades were included in this systematic review. Two investigators had the protocol for study selection (LS and CD) and independently assessed the studies for eligibility. Inclusion criteria were: (1) a clear definition of the population studied; (2) a clear definition of the time period; (3) a clear definition of infective endocarditis; and (4) a clear description of the frequency distribution of the microbiology encountered. In order to avoid bias, we excluded studies limited to specific populations (e.g. HIV or intravenous drug users). When there was difference of opinion, a third investigator (NC) resolved the disagreement. The authors are fluent in English, Spanish, Portuguese, and Italian; papers in other languages were translated by collaborating physicians who were native speakers. When two studies reported data from the same cohort and time frame, only the most complete one was included. If a study reported data for various time frames, data were analyzed separately for each decade. Kappa (k) for inclusion was calculated from a sample of 10 randomly selected papers. Results were compared and inconsistencies were resolved by consensus. Dr. Andrew Wang from Duke University was consulted for reviewing the list of included studies for completeness.
For each study included, the following information was extracted: first author's last name, journal, year of publication, IE definition used, certainty (possible IE vs. definite IE), countries, time-frame, multi-center vs. single center, sample size, age, gender, mortality, intravenous drug abuse (IVDA), intracardiac device or prosthetic valve, Staphylococcus aureus (SA), coagulase-negative staphylococcus (CNS), enterococci, Streptococcus viridans (SV) and culture negative (CN) IE.
We also included patients from Einstein Medical Center (EMC), a tertiary 440-bed city hospital in Philadelphia, PA, US; for the years 2000 -2010. Data were retrospectively extracted per ICD code and then included only if patients met 'definite' or 'possible' modified Duke criteria [19]. Clinical characteristics were extracted for each patient as mentioned above. This data was added as one more study to the last decade. Sensitivity analyses showed that subtracting this data from the rest did not modify results significance.

Quality assessment
Two reviewers (LS and CD) independently assessed the quality of the manuscripts using the approaches recommended by Khan and colleagues [20] and Stroup and colleagues [21] for cohort studies. The main criteria were: (1) prospective study, (2) Duke or Von Reyn definition of IE, (3) definite IE, and (4) number of patients above 40. Quality was assigned as A, or excellent, with 4 points, B or good, with 2-3 points, and C, or suboptimal, with 0-1 points. Data was weighted for quality for SA and SV without affecting results statistical significance.

Statistical analysis
Data were extracted independently by two researchers (LS and CD) and collected on an Excel spreadsheet. Data were allocated to a decade according to the midpoint date for the time frame studied. For example if the time frame was 1979 to 1983, data was assigned to the 80s. Results are shown as mean +/-SD, percentages for each decade, and the 95% CI. Main variables studied were the frequency distribution of pathogens, patient age and gender, and in-hospital mortality. ANOVA and Chi 2 test were used to compare studies by decades. Each group was compared to the rest using paired Students t-tests. Samples were weighted for size. Results from our own hospital were included in the last decade and weighted for size. Sensitivity analysis showed that subtracting this data did not affect results statistic significance. Moreover subtracting the biggest study by Murdoch et al. from the International Collaboration on Endocarditis did not affect statistical significance of results. Sub-analyses were performed for continent and IVDA. Correlation between IVDA and SA was performed using Spearman correlation test. Hospital-based studies and population-based studies were included. As both types of studies may be prone to bias (hospital-based studies to referral bias [22] and population-based studies to selection bias) hospital-based and population-based studies were analyzed separately. Sensitivity analyses were performed for size, single vs multi-center and quality without affecting significance of results. A p<0.05 was considered statistically significant. Analyses were performed using JMP version 10.0 (SAS Institute, Cary, NC, US).
Data were presented in a graph as mean (in green, centerline of diamond) and variance (size of diamond) for each variable studied in each decade, with standard deviations (blue). Each dot in a column represents a particular study percentage. N below decades represents total number of patients in each decade. Every patient included in each study was diagnosed with IE as described above.

Results
Candidate studies included 24,415 articles identified in PubMed, 10,421 in Medline, and 4,528 in EMBASE; one hundred sixty studies met all inclusion criteria (see flow diagram in Figure 1). Of these, 142 were hospital-based, including a total of 23,606 patients (Table 1), and 18 were population-based, including a total of 3,477 patients ( Table 2). Investigators (LS and CD) were in agreement on which articles were to be included (k=1).

Hospital-based Studies
Among hospital-based studies, IE epidemiology changed over the last 5 decades (Figure 2). Patients were significantly older ( Figure 2A     .01) while in Europe there was a decrease that did not reach statistical significance (p=0.06). No significant changes were seen in Latin America and Oceania (p=0.9, p=0.32, respectively). Insufficient data were available from Africa for separate analysis.
Subgroup analyses for changes in IVDA IE percentage are shown in Figure 5. No significant changes were seen on a global basis. However, a significant increase in IVDA related IE frequency distribution was observed in North America in the last decade ( Figure 5C; 1980s: 17.3%, CI 10.7-23.9 vs 2000s: 50.7%, CI 28.5-73.0, p<0.05). Conversely, we observed a significant decrease in IVDA related IE percentage in Europe in the last decade ( Figure 5D; 1990s: 21.1%, CI 12.3-29.8 vs 2000s: 6.8%, CI 3.5-10.2, p<0.01). We found a positive correlation between SA IE and IVDA. Interestingly, this correlation lost strength in the last decade (1990s rs=0.82, p=0.001 vs 2000s rs=0.40 p=0.05; 1990s vs 2000s, Fisher r-toz transformation, p<0.001). We further analyzed the studies that reported microbiology for the IVDA IE group. Twenty-five studies and 1288 patients were included in this sub-analysis. No significant temporal trends in IVDA IE microbiology were found. In this subset of patients, SA represented the main pathogen (1970s: 69 In-hospital mortality rate due to IE decreased following the 1960s and remained stable thereafter ( Figure 6; 1960s: 30.6%, CI 24.4-36.8 vs 2000s: 19.7%, CI 17.8-21.6, p=0.01). On subgroup analysis by continent, no regional differences were observed.

Population-based Studies
Among population-based studies, no significant trends were observed regarding IE microbiology as shown in Table 3 (SA p=0.82; SV p=0.14; enterococci p=0.33). These studies included populations in the US and Europe primarily. Only one study was from Asia. Of note, the majority of data from the US is from Olmstead County, Minnesota.

2000s data from Einstein Medical Center, Philadelphia, PA, US
We identified a total of 261 cases from 2000 to 2010 (

Discussion
The main finding of this study is that the epidemiology of IE has changed worldwide over the last half century. Furthermore, the observed changes in IE microbiology varied by continent. These findings stemmed from analyses of hospital-based reports. In a separately analyzed smaller group of populationbased studies (most of them from the US), no consistent changes in IE microbiology frequency distribution over time were observed.
Most notably, the global percentage of SA IE has nearly doubled in the last five decades (18% in the 1960s to nearly 30% in the 2000s). When analyzed by continent this increase was largely due to an increased frequency of SA in North America (from 25% in the 1960s to 52% in the 2000s) with no significant change among reports from other continents. This finding has important implications as SA infections are associated with longer length of stay, higher death rates [23], increase hospitalizations [24], and elevated costs [24]. An increase in IVDA IE percentage in North America as compared to Europe may partially explain these changes in SA frequency distribution. However, the number of studies in the last decade in this analysis is small and therefore this finding should be studied further. Other potential contributors include increases in the elderly population [10], increased numbers of chronically-ill patients [25], increased contacts with the health-care system [26,27], and increasing use of intracardiac and vascular devices. Benito and colleagues [27] found a high percentage of health-care associated infections (related to catheters, dialysis, or immunosuppressive therapy) among US patients with native valve IE; SA was the most common organism isolated. Though the present study was not able to specifically track cardiac device implantation, another recent study found an increased prevalence of staphylococcal IE in these patients [28]. In absolute numbers Bor et al, reported an incidence of infective endocarditis in the US close to 40,000 cases/year [29]. Furthermore, at least when measured by ICD codes the total number of SA cases seems to be increasing in the US [15]. In addition, certain subgroups may behave differently; a welldesigned population based study found that SA frequency has increased in Europe in patients without previously known valve disease [5]. It is important to clarify that changes in individual countries may not necessarily follow the trends at a continent level. The present study also documents a substantial decline in the frequency of SV IE over the last five decades (27.4% in the 1960s to 17.6% in the 2000s). This finding was statistically significant for North America and Asia with a strong trend in Europe. Therefore, it appears that changes in the epidemiology of this organism are more widespread than for SA.
Paralleling the increase in SA IE frequency, there was also an increase in CNS IE percentage over time. It is known that CNS infections are often related to the use of intravascular catheters and prosthetic vascular grafts [30]. Thus, the rise in CNS IE may well be health-care related.
Enterococcal IE frequency increased in the last decade of the study. Enterococcal infections typically affect elderly patients and those with prior valvular damage, diabetes mellitus, indwelling catheters, or who are on hemodyalisis [31]. This finding is extremely important given the high prevalence of multidrug resistant enterococci and therefore the implications on treatment options. Lastly, culture negative endocarditis percentage decreased in the last decade. This is likely because of improved laboratory techniques and culture methods.
Worldwide, the present study found increases in age among IE patients. This has important implications for treatment and use of health care resources as elderly patients have more comorbidities and may be more prone to infection with certain organisms, such as enterococci. Consistent with the general perception [18], the present study found that in-hospital mortality rate of IE remains high with no significant decrease observed since the 1960s.
A limitation of the present study is the lack of individual patient level data. This data was not available from older studies and including it for only the last decades would have changed one bias for another one, without adding accuracy. Another limitation is that most of the findings come from hospital-based studies, whereas no significant changes were seen in the population based-studies over time. One possible explanation for this is a lack of power (18 population-based studies covering 3,477 patients vs 142 hospital-based studies covering 23,606 patients). Population-studies are also subject to sample bias: the population studied may not truly represent the general population. They can be subject to underreporting, as many times they rely on surveys. Moreover, populationstudies of IE in the US are mainly from the Olmstead County, a population that is unlikely to represent the total US population. Hospital-based studies can suffer from referral bias as well, with sicker patients being referred to specialized centers. Thus, these results might not apply to community hospitals. However, Kanafani and colleagues [32] found only a slight difference between referred and non-referred patients, with higher SA IE in the non-referred patients. Thus, had this played a role in the present study, it would have likely decreased SA frequency. Therefore, it is unlikely to explain our findings. Finally, the   definition of IE has changed over time, as well as culture quality, which could have caused heterogeneity in the cases included.

Conclusion
The present study represents the largest systematic review of IE epidemiology to date. Important findings include an increase in staphylococcal IE frequency over the last halfcentury, particularly in North America, and a worldwide decrease in SV IE percentage. In the last decade SA IE and enterococci IE frequencies have increased while culture negative IE has decreased. Patients with IE are getting older and the male to female ratio is increasing. Mortality has changed little in the last four decades.