Article Text

Download PDFPDF
Original research
Contemporary epidemiology and outcomes in recurrent infective endocarditis
  1. Afonso B Freitas-Ferraz1,
  2. Gabriela Tirado-Conte1,
  3. Isidre Vilacosta1,
  4. Carmen Olmos1,
  5. Carmen Sáez2,
  6. Javier López3,
  7. Cristina Sarriá2,
  8. Carlos Nicolás Pérez-García1,
  9. Daniel García-Arribas1,
  10. Marianela Ciudad2,
  11. Pablo Elpidio García-Granja3,
  12. Raquel Ladrón3,
  13. Carlos Ferrera1,
  14. Salvatore Di Stefano3,
  15. Luis Maroto1,
  16. Manuel Carnero1,
  17. J Alberto San Román3
  1. 1 Servicio de Cardiología, Instituto Cardiovascular, Hospital Clínico San Carlos, Instituto de Investigación Sanitaria del Hospital Clínico San Carlos (IdSSC), Madrid, Spain
  2. 2 Servicio de Medicina Interna-Infecciosas, Instituto de Investigación Sanitaria del Hospital Universitario de la Princesa, Madrid, Spain
  3. 3 Servicio de Cardiología, Instituto de Ciencias del Corazón (ICICOR), CIBERCV, Hospital Clínico Universitario de Valladolid, Valladolid, Spain
  1. Correspondence to Dr Afonso B Freitas-Ferraz; afonsobff{at}


Objective Recurrent infective endocarditis (IE) is a major complication of patients surviving a first episode of IE. This study sought to analyse the current state of recurrent IE in a large contemporary cohort.

Methods 1335 consecutive episodes of IE were recruited prospectively in three tertiary care centres in Spain between 1996 and 2015. Episodes were categorised into group I (n=1227), first-IE episode and group II (n=108), recurrent IE (8.1%). After excluding six patients, due to lack of relevant data, group II was subdivided into IIa (n=87), reinfection (different microorganism), and IIb (n=15), relapse (same microorganism within 6 months of the initial episode).

Results The cumulative burden and incidence of recurrence was slightly lower in the second decade of the study (2006–2015) (7.17 vs 4.10 events/100 survivors and 7.51% vs 3.82, respectively). Patients with reinfections, compared with group I, were significantly younger, had a higher frequency of HIV infection, were more commonly intravenous drug users (IVDU) and prosthetic valve carriers, had less embolic complications and cardiac surgery, with similar in-hospital mortality. IVDU was found to be an independent predictor of reinfection (HR 3.92, 95% CI 1.86 to 8.28).

In the relapse IE group, prosthetic valve endocarditis (PVE) and periannular complications were more common. Among patients treated medically, those with PVE had a higher relapse incidence (4.82% vs 0.43% in native valve IE, p=0.018). Staphylococcus aureus and PVE were independent predictors of relapse (HR 3.14, 95% CI 1.11 to 8.86 and 3.19, 95% CI 1.13 to 9.00, respectively) and in-hospital-mortality was similar to group I. Three-year all-cause mortality was similar in recurrent episodes compared with single episodes.

Conclusion Recurrent IE remains a frequent late complication. IVDU was associated with a fourfold increase in the risk of reinfection. PVE treated medically and infections caused by S. aureus increased the risk of relapse. In-hospital and long-term mortality was comparable among groups.

  • endocarditis
  • valvular heart disease

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Infective endocarditis (IE) carries a high risk of morbidity and mortality, especially during the active phase of infection.1 Patients surviving a first episode of IE have a significantly worse survival than the general population, mainly due to late complications such as heart failure, increased need for valve surgery and a higher risk of recurrence.2 However, only a small number of studies have primarily focused on patients with recurrent IE and most information regarding long-term and short-term prognosis, risk factors and the predominant causative agents in this setting comes from outdated studies that may not reflect the current situation.3–5

The aim of this study was to evaluate the current state, temporal trends and prognostic impact of recurrent IE and to identify predictive factors for subsequent episodes.


Study population

The current study is the result of a collaborative ongoing research on IE between three tertiary care centres in Spain with on-site surgical facilities. These centres have been working together since 1995, with standardised protocols, uniform data collection and an identical diagnostic and therapeutic approach (please refer to online supplementary figure 1 for additional information regarding the geographical situation of the participating Hospitals).

Supplemental material

From January 1996 to December 2015, 1335 consecutive episodes of IE were prospectively recruited using standardised protocols.6 The diagnosis was established according to Duke criteria until 2002,7 and the modified Duke criteria thereafter.8 This study complies with the Declaration of Helsinki and was approved by the local ethics committee.

For comparison purposes, episodes were categorised into group I, first episodes of IE, and group II, episodes in patients with a history of IE. Group II was further subdivided into IIa, reinfection, and IIb, relapse. Group I was then compared with the other two groups. In order to calculate the incidence rate and predictors of reinfection in our cohort, the analysis was performed for individual patients instead of by number of episodes.

Long-term follow-up was performed retrospectively, through review of medical records or by phone contacts. The flow diagram is depicted in figure 1A.

Figure 1

(A) Patients flow diagram. Out of 1271 patients enrolled in our database, 1335 consecutive episodes of infective endocarditis (IE) were prospectively identified (grey square outline). In these 1271 patients, the first registered episode was a recurrence in 44 and a first episode of IE in 1227 patients. Additionally, a first recurrence was identified in 57 patients and a second recurrence in seven patients (the yellow square outline features patients with repeat episodes registered in our database). During follow-up, 26 additional recurrent episodes that were not registered in the database were identified: 19 patients with a first recurrence, two with a second recurrence and five experiencing three recurrences. The red rectangle highlights patients in whom the first IE episode was registered in our database and were subsequently taken into account to analyse the IE incidence, long-term survival and predictors of recurrences. The green rectangles point out how patients registered in our dedicated database were categorised into two groups: group I, first episodes of IE (n=1227), and group II, episodes in patients with a history of IE (n=108). After excluding six cases in which it was not possible to distinguish between relapse and reinfection, group II was subdivided into group IIa, reinfections (n=87), and group IIb, relapses (n=15). (B) Factors independently associated with reinfection and relapse. First table: intravenous drug users was found to be an independent predictor of reinfection. Second table: Staphylococcus aureus and prosthetic valve endocarditis were independent predictors of relapse.


Patients with more than one episode of IE were classified as having a recurrence.

Relapse was defined as an additional IE episode caused by the same microorganism or by a microorganism of the same species within 6 months after the initial episode. Reinfection was defined as a new episode of IE caused by a different microorganism or by a microorganism of the same species after a period of time greater than 6 months from the initial episode.9 Nosocomial IE, non-nosocomial healthcare-associated IE and community-acquired IE have been previously defined in the literature.6 9

A diagnosis of chronic kidney disease was established based on a cut-off value of glomerular filtration rate estimate of <60 mL/min/1.73 m2. Acute-onset IE was considered to exist when the time between the appearance of symptoms and diagnosis was <15 days.10 Septic shock was defined according to the Third International Consensus Definitions for Sepsis and Septic Shock.11 Cardiac death was defined as death from any immediate cardiac cause, and sudden death from unknown causes.

Statistical analysis

Categorical variables were expressed as a number (percentage) and continuous variables as mean (SD) or median (IQR: 25–75th percentile). Assessment of normality was performed using the Shapiro-Wilk test. Qualitative variables were compared with the X2 or Fisher’s exact test. Quantitative variables were analysed with a two-sided Student’s t-test and median test. Mortality curves were calculated using the Kaplan-Meier method, and the log-rank test was applied to compare groups.

We estimated the cumulative burden of IE recurrence (overall and according to the two study periods) using the method of mean cumulative count, which estimates the mean number of recurrent events in a cohort over time in the presence of competing risk events. Bootstrap percentile method was used to calculate 95% CI.12 Cause-specific incidence and predictive factors of reinfections and relapses were estimated with a competing risk model using time-dependent weights.13 14 All-cause mortality served as the competing event.

Univariable analysis was performed to identify variables from the first IE episode associated with reinfection/relapse. Those variables with p<0.10 and those considered clinically significant were entered into the competing risk model and the results were expressed as sub-HR. The proportional subhazards assumption was tested by including time-varying covariates in the model, which demonstrated an absence of interaction between the predictors and the survival time and between the predictors and the logarithmic function of survival time.

To calculate the likelihood and predictors of recurrences, we only took into account the second episode of IE that occurred in patients for which we had the first episode registered in our database and survived the event. Subsequent episodes included those recorded in our database and those found during follow-up (figure 1A).

All data were analysed with Stata V.14.2 and R V.3.6.0.

Patient and public involvement statement

This research was done without patient and public involvement.


Out of 1335 episodes of IE, 1227 (91.9%) were first episodes and the remaining 108 (8.1%) were recurrences, 15 of which were relapses and 87 were reinfections. There were six cases in which it was not possible to distinguish between relapse and reinfection due to lack of relevant data. One episode of recurrence was recorded in 94 patients and two episodes in seven patients.

Follow-up was completed in 1026 out of 1271 patients (80.7%). The median follow-up of all patients was 20.8 months (IQR 1.3–72.6), 19.9 months (IQR 1.2–71.1) in patients with a single episode and 37.8 months (IQR 1.0–89.0) in those with recurrences.

At 3 years follow-up, overall cumulative burden of recurrence and cause-specific cumulative incidence was 5.70 (95% CI 4.37 to 7.21) events per 100 survivors and 5.43% (95% CI 4.16% to 7.07%), respectively. A slightly higher cumulative burden and cause-specific incidence of recurrences was found in the first (1996–2005) compared with the second (2006–2015) decade of the study: 7.17 (95% CI 4.79 to 9.61) versus 4.10 (95% CI 2.63 to 5.81) events per 100 survivors, p=0.12, and 7.51% versus 3.82% at 3-year follow-up, p=0.28, respectively (figure 2).

Figure 2

Cumulative incidence of recurrence and cumulative burden of recurrence with 95% CI according to the first (1996–2005) and second (2006–2015) decades of the study.

Reinfections compared with a single episode of IE

Baseline characteristics and microbiological profile

Cause-specific cumulative incidence of reinfections was 4.29% at 3 years (95% CI 3.12 to 5.89). Median time until reinfection was 3.0 years (IQR 1.0–6.6 years). Baseline characteristics of patients with a single episode of IE and those with reinfections are depicted in table 1.

Table 1

Demographic and clinical characteristics at admission of patients with a first episode of IE and patients with reinfection

The reinfection group had a final diagnosis of possible IE more frequently (6.9 vs 16.1%, p=0.002). Intravenous drug users (IVDU) were more frequent in the reinfection group at the expense of the first decade of the study (1996–2005) (group I: 9.8 vs group IIa: 22.2%, p=0.021), in which the prevalence of IVDU was much higher compared with the second period (2006–2015) (group I: 1.8% vs group IIa: 4.0%, p=0.255). HIV infection was also more prevalent in patients with reinfections, but only during the first decade of the study (first decade: 6.7% vs 20.0%, p=0.004; second decade: 1.8% vs 0%, p=1.0). Degenerative valvular heart disease was more frequent in group I, whereas prosthetic valve endocarditis (PVE) was more common in patients with reinfections. The same location (valve/device) was involved in the first and recurrent episode in 72% of patients.

Clinical presentation was similar between groups, with the exception of acute onset IE, which was more common in the reinfection group (47.0% vs 59.8%, p=0.021).

Pathogenic microorganisms had a similar distribution between patients with and without reinfections (online supplementary table 1). There were no significant differences in the microbiological profile between the first and the second study periods (online supplementary table 2).

Echocardiographic findings

Echocardiographic detection of vegetations was more frequent in group I. Periannular complications were documented with a similar occurrence in both groups; however, prosthetic dehiscence was much more commonly found in patients with reinfection and there was also a clear tendency to a higher prevalence of fistulas in this group (online supplementary table 3).

In-hospital outcomes

In-hospital outcomes were similar in both groups (table 2). Among patients with a single IE episode, embolisms were more commonly documented. These differences persisted even when only definite IE episodes were analysed (central nervous system embolisms: 21.7 vs 11.0%, p=0.029 and peripheral embolisms: 26.3 vs 16.4%, p=0.062).

Table 2

Clinical events during in-hospital evolution

Cardiac surgery was less frequently performed in the reinfection group (59.1 vs 45.4%, p=0.013). However, surgical treatment differed according to the side of the heart involved. Patients with right-sided IE from group I underwent surgery more frequently than those with reinfections (64.7 vs 30.8%, p=0.019). Conversely, patients with left-sided IE underwent surgical treatment in a similar percentage in both groups (58.9 vs 50.0%, p=0.139).

Surgical management was more common in the second decade of the study, but without reaching statistical significance (first decade: 35.1% vs 53.1%, p=0.098).

No differences were found in in-hospital mortality between groups (26.6 vs 20.9%, p=0.245).

Independent predictors of reinfection

Factors associated with reinfection were age, IVDU and HIV infection. Age was not included in the multivariable analysis due to a strong collinearity with IVDU. After adjustment, only IVDU remained as an independent predictor of reinfection (HR 3.92; 95% CI 1.86 to 8.28, p<0.001) (table 3 and figure 1B).

Table 3

Clinical characteristics from the first episode found to be independent predictors of reinfection

Relapses compared with single episodes of IE

Fifteen patients experienced a relapse during the study period, corresponding to a cause-specific cumulative incidence of 1.53% (95% CI 0.92% to 2.53%). Median time until relapse was 2.9 months (IQR 1.3–4.0). We found no differences regarding age, sex, origin, comorbidities or portal of entry. The same location was involved in the first and recurrent episode in 82% of patients. A detailed description of the valves/devices involved in the first and subsequent relapsed IE episodes are provided in online supplementary table 4.

PVE accounted for 60% of the IE episodes observed in the relapse group and for 31.3% of those that occurred in group I (p=0.017). A higher incidence of relapse was found in patients whose first IE episode was a PVE compared with those with native valve infection (2.44 vs 0.78%, p=0.045), without significant differences between mechanical and biological prostheses (2.29 vs 2.90%, p=0.677). When patients were analysed according to the treatment received, those with PVE treated medically had more episodes of relapses compared with those with native valve IE medically treated (4.82 in PVE vs 0.43% in native valve IE, p=0.018).

Staphylococcus aureus was the predominant aetiological agent in patients with relapse and was also more prevalent in this group compared with patients with a single IE episode (40.0 vs 20.6%, p=0.066). Compared with patients from group I, there was a higher incidence of periannular complications (46.7 vs 24.1%, p=0.043) and a tendency to a higher frequency of septic shock (20.0 vs 6.2%, p=0.064) in the relapse group. Nonetheless, we found no statistically significant differences in in-hospital mortality between relapse and single episodes of IE (40.0 vs 26.6%, p=0.245).

Independent factors associated with relapses were S. aureus infection (HR 3.14; 95% CI 1.11 to 8.86) and PVE (HR 3.19; 95% CI 1.13 to 9.00) (figure 1B).

Follow-up survival analysis: recurrences (reinfections and relapses) compared with single episode

Recurrent episodes had a 3-year freedom from all-cause mortality similar to single episodes (59.4 vs 56.8%, log rank p=0.819). At 1 and 3 years, no statistically significant differences were found in cardiac death between the aforementioned groups (figure 3).

Figure 3

Freedom from all-cause mortality (A) and cardiac death (B) at 3 years follow-up among patients with recurrences compared with those with a single infective endocarditis (IE) episode.


This study is the largest contemporary cohort of patients with IE analysing the profile of repeat endocarditis. Major strengths of our work include: (1) it is a multicentre study with prospective recruitment; (2) a standardised diagnostic and therapeutic approach have been used since the beginning of the study; (3) long-term follow-up has been achieved in the majority of patients and (4) reinfections and relapses were analysed separately due to their different physiopathology.

Recurrent IE constituted 8.1% of the total IE episodes in our database. Previous studies have reported a variable recurrence rate, ranging between 4.7% and 22.5%.3–5 15–17 Herein, a trend towards a lower cumulative burden and incidence of recurrences was found in the second decade of the study, especially in regard to relapses, which may reflect a better management of the disease in the current era.

Conceptually, relapses represent a failure of treatment, whereas reinfections are more related to patients’ clinical and immunological profile, as they correspond to a new infection by another microorganism.

Reinfections compared with single episodes of IE

Patients with reinfections were younger compared with those with a single episode of IE. Earlier series showed that an older age and male gender were risk factors for reinfection.4 However, a more recent report found that patients with recurrences were younger and without gender differences.16 A plausible explanation for this observation is that IVDU and HIV, which are known risk factors for reinfections, are more prevalent among younger patients.16–18

IVDU is also a recognised predisposing factor for the development of IE and, in accordance to previous reports, was found to be an independent predictor of reinfection associated with a threefold increase in the risk of presenting a new IE episode.16 18 Repeated exposition to particulate matter causing endothelial valve damage in the setting of injected ‘bacterial loads’ seems part of the explanation for the higher incidence of reinfections in this group.19

A final diagnosis of possible IE was more common in patients with reinfections. In this group, the percentage of patients with negative blood cultures was slightly higher, vegetation detection rate was lower and prosthetic valve dehiscence and fistula were more frequent. These findings are possibly related to a higher prevalence of reinfection in prosthetic valve carriers. It is well known that detecting vegetations in the setting of PVE can be more challenging,6 20 and the application of Duke criteria less useful.21 22 Additionally, PVE is characterised by a lower incidence of vegetations and a higher incidence of abscesses and perivalvular complications.23

Interestingly, in our cohort, patients with reinfections had a lower incidence of embolic events, especially central nervous system embolisms, alongside with a lower vegetation detection rate. This group of patients had a significantly lower need for cardiac surgery, with similar mortality rates. The apparent better prognosis of patients with reinfections could be attributed to excess mortality in high-risk patients during the first episode, hereby selecting a ‘low-risk’ population vulnerable for future reinfections (survivor bias).

Relapses compared with single episodes of IE

Relapsing IE suggests failed patient management and mandates a search for a persistent focus of infection (local or metastatic), to reconsider whether the antibiotic treatment has been appropriate, and the need for surgery. In our series, the incidence of relapse was 1.53%, similar to that of a recent Spanish cohort of patients with IE.24 This low relapse rate, especially in the last decade, may reflect a better management of these patients. There are three features of relapse IE that deserve a special mention. First, patients with PVE had more relapses than patients with native valve endocarditis. In fact, one-third of the relapses occurred in patients in which the first episode was a PVE. This strengthens the results of previous studies that pointed out PVE as a risk factor for the development of relapses.3

Second, relapses were associated with a particular clinical and microbiological scenario, such as having S. aureus as the main causative agent and a higher incidence of periannular complications and septic shock. The higher incidence of S. aureus infection in patients with relapses was also described by the International Collaboration on Endocarditis-Prospective Cohort Study.16

S. aureus PVE has been associated with an increased risk of developing periannular extension of the infection, septic shock and higher in-hospital mortality.25–27 As such, the higher frequency of the above-mentioned complications found in relapses might be explained by the fact that S. aureus was the predominant etiological agent in this subgroup of patients.

Lastly, patients with PVE treated medically had a higher incidence of relapse than patients with native valve IE medically treated. Nevertheless, a selection bias might possibly occur, as it is well known that patients with PVE who do not undergo surgery have higher mortality. A recent meta-analysis composed of 32 studies found a similar rate of recurrence in patients with PVE regardless of treatment strategy.28 However, those studies did not discern relapses from reinfections.

Follow-up survival analysis: recurrences compared with single episode

Recurrent IE is considered to be a risk factor for death.2 4 16 In our cohort, freedom from all-cause mortality at 3 years follow-up was similar among recurrent and single IE episodes. Other authors have found similar results (figure 3).29 Understandably, this does not mean that having a recurrent episode does not carry out an increased mortality in patients with IE, it simply denotes that the risk of each episode is similar.

However, in our study, a trend towards a lower cardiac mortality was observed in the recurrence group. In this scenario, the presence of survival bias cannot be excluded, and it is important to consider that patients with recurrences were younger, more frequently male, with a higher percentage of possible IE and intravenous drug use.


Our study has several limitations. First, we were not able to achieve long-term follow-up for the totality of patients, which could have influenced our results, as well as the estimation of recurrences. Second, in patients with relapse, we did not perform molecular analysis to confirm that the same microorganism was present. Third, it is well-known that some patients with prosthetic material, such as pacemakers, may have relapses with late presentation (>6 months). However, in this study, patients with a new IE episode after 6 months, by a microorganism of the same species or with negative blood cultures, were considered to have a reinfection.


The cumulative incidence and burden of recurrent IE in our series showed a trend towards a decrease over time. However, it still remains a relatively frequent complication of patients surviving a first episode of IE. IVDU was found to be an independent predictor of reinfection. Patients with relapse had the worst in-hospital outcomes owing to a higher incidence of septic shock and periannular complications. A higher risk of suffering a relapse was seen in patients with S. aureus IE and in those cases where PVE was treated conservatively. However, neither reinfection nor relapse were associated with increased in-hospital mortality nor with a worse long-term outcome when compared with patients without recurrence.

Key messages

What is already known on this subject?

  • Patients surviving a first episode of infective endocarditis (IE) have a significantly worse prognosis than the general population, mainly due to late complications. Despite its clinical relevance, the majority of data comes from outdated studies, many of which do not focus on patients with recurrent IE and do not differentiate relapses from reinfections.

What might this study add?

  • This contemporary updated study shows that, even though there was a trend towards a decrease in the incidence of recurrent IE during the past decade, it still remains a frequent late complication.

  • Reinfection was almost four times more frequent in intravenous drug users (IVDU).

  • Staphylococcus aureus and prosthetic valve endocarditis (PVE) increased the risk of relapse threefold.

  • Patients with prosthetic valve endocarditis treated conservatively had a higher risk of relapse and worse in-hospital outcomes.

How might this impact on clinical practice?

  • IVDU with endocarditis, as well as medically treated PVE, may take advantage from a close monitoring in order to promptly detect recurrent episodes.



  • ABF-F and GT-C contributed equally.

  • Contributors ABFF, GTC, IV and CO made substantial contributions to the conception and design of the study. ABFF and GTC have drafted the manuscript and analysed the data. IV, CO, CS, JL, CS, CNPG, DGA, MC, PEGG, RL, CF, SDS, LM, MCA and JASR have revised the manuscript for important intellectual content.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval The study was approved by the local ethics committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available upon reasonable request.