Kaplan-Meier curve relating survival (percentage [SE]) according to the period of infective endocarditis management (hazard ratio, 0.41; 95% confidence interval, 0.21-0.79 [P = .008]).
Botelho-Nevers E, Thuny F, Casalta JP, Richet H, Gouriet F, Collart F, Riberi A, Habib G, Raoult D. Dramatic Reduction in Infective Endocarditis–Related Mortality With a Management-Based Approach. Arch Intern Med. 2009;169(14):1290-1298. doi:10.1001/archinternmed.2009.192
Copyright 2009 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2009
Despite improvements in medical and surgical therapy, infective endocarditis (IE) is still associated with a severe prognosis and remains a therapeutic challenge. We aimed to evaluate the impact of a standardized diagnostic and therapeutic protocol on mortality and to correlate the outcome with compliance with our management-based protocol.
We conducted an observational before-after study that included 333 consecutive patients treated for IE at a referral center from 1991 to 2006, which was divided into 2 periods: period 1 (1991-2001), before implementation of our therapeutic protocol (n = 173), and period 2 (2002-2006), after implementation of our protocol (n = 160). Our protocol was created by a multidisciplinary task force including a sampling of biological specimens, the use of only 4 antimicrobial agents, a standardized duration of treatment, standardized surgical indications, and 1 year of close follow-up. Because our protocol was based on a local consensus by physicians and surgeons, it was not possible to randomize the study.
The 1-year mortality significantly decreased from 18.5% during period 1 to 8.2% during period 2 (hazard ratio, 0.41; 95% confidence interval, 0.21-0.79 [P = .008]). After multivariable analysis, the management during period 2 remained a strong protective factor (adjusted hazard ratio, 0.26; 95% confidence interval, 0.09-0.76 [P = .01]). During period 2, we observed a statistically significantly better compliance in antimicrobial therapy and fewer cases of renal failure. Deaths by embolic events and multiple organ failure syndrome also significantly decreased during period 2.
A dramatic reduction in mortality was observed during this study, suggesting that a management-based approach has a significant impact on IE outcome.
Infective endocarditis (IE) remains among the deadliest infectious disease. Despite changes in the profile of IE over the years1,2 and better medical3- 5 and surgical management,6 in-hospital mortality has not declined, still ranging from 11% to 36%.1,6- 8 Many factors have an impact on the outcome of IE, including the virulence of the microorganism, the characteristics of the patients, the presence of underlying disease, the delay in diagnosis and/or treatment, the surgical indications, and the timing of surgery. In addition, many professionals are involved in the management of IE including cardiologists, microbiologists, infectious diseases specialists, and surgeons, complicating the already difficult treatment, and in fine no 2 patients receive the same treatment. Moreover, the deviations from the published guidelines9- 11 may have a negative impact on the patient's prognosis. Preliminary results from our center showed that there were unfounded modifications in antimicrobial therapy and that surgical indications differed from one surgeon to another.
Therefore, to improve the prognosis of IE, we decided to implement a new strategy based on both a local consensus on the management of IE formed between microbiologists, cardiologists, infectious diseases specialists, and surgeons and on the development of medical-surgical local guidelines modeled according to the protocols developed to treat cancers, which have been shown to be extremely successful in improving the outcome of oncology patients. In such instances, all patients benefit from the same adequate treatment, whoever the physician in charge. Our consensual protocol was characterized by the administration of few antimicrobial agents (amoxicillin sodium or oxacillin sodium or vancomycin hydrochloride and gentamicin sulfate) and clearly defined surgical indications. This approach has been our standardized diagnostic strategy since 1994.12 Our approach to the management of IE is also similar to the type of intervention developed by Pronovost et al13 to decrease catheter-related bloodstream infections in the intensive care unit by using an evidence-based intervention. As in the study by Pronovost et al,13 our protocol was evidence based and without randomization because it would have been unethical to deliberately expose patients with IE to potentially risky therapeutic methods. Instead, we chose to evaluate the impact of our protocol by assessing IE-associated mortality before and after the implementation of our evidence-based management strategy. We also compared our results with those found in the literature.
This study took place in the Cardiology Department of La Timone Hospital, Marseille, France, a teaching tertiary care hospital for adults. All the consecutive patients with suspected IE who were admitted to the hospital between January 1991 and December 2006 were eligible for study entry. From 1991 to 1994, clinical data were retrospectively abstracted from the medical records. Since 1994, clinical data were prospectively collected using a standardized questionnaire.14 A distinction was made between cases involving presumably healthy valves and patients carrying intracardiac material. Comorbidity was assessed by using the Charlson comorbidity index.15 For each case, a diagnostic kit12 was used, and after a sampling period of 4 hours, a treatment was initiated when suspicion of IE was high. Echocardiography was performed by experienced cardiologists, by the transthoracic approach at the beginning of the study and later by the systematic transthoracic and transesophageal approach. The modified Duke criteria16,17 were applied to all suspected cases of IE, which were characterized as definite, probable, or excluded. Only patients with definite IE were included.
Patients with IE related to gram-negative rods, fungi, or rare causes (Legionella species, Chlamydia species, Brucella species, Coxiella burnetii, and Tropheryma whipplei) were excluded from the study because our therapeutic protocol was not adapted to those microorganisms.
Since 1992, the management of IE at our hospital has been multidisciplinary, involving cardiologists, cardiac surgeons, infectious diseases specialists, and microbiologists. In 1994, a new strategy was implemented that aimed to standardize the diagnosis of IE by using a diagnosis kit including systematic serologic testing, blood cultures, and microbiological and histological analysis of the removed valves, as fully described elsewhere.12,14,18,19
The study was divided into 2 periods: period 1, before implementation of therapeutic protocol (1991-2001), and period 2, after implementation of our protocol (2002-2006). The ethics committee of Institute Fédératif de Recherche 48 approved the study.
In 2001, to establish a local consensus in the treatment of IE used and applied by all physicians in charge, a simplified and standardized protocol was elaborated in a multidisciplinary approach, following a review of evidence-based studies. A consensus was obtained for all characteristics of the protocol (duration, antibiotic compounds, surgical indications, and follow-up). Microbiologists, cardiologists, infectious diseases specialists, and surgeons were involved in the development of this protocol.
The protocol, which included only 4 intravenous antimicrobial agents (amoxicillin, vancomycin, gentamicin, and oxacillin) (Table 1 and Table 2), was applied during a testing period of 6 months before being reviewed, modified (notably surgical indications), and definitively applied since 2002. The protocol is a medical-surgical guide for the management of IE, taking the following into account: (1) the localization of IE; (2) the principal causative microorganisms (streptococci, enterococci, staphylococci, and HACCEK [Haemophilus species, Actinobacillus actinomycetemcomitans, Cardiobacterium hominis, Capnocytophaga species, Eikenella corrodens, and Kingella species]) microorganisms; and (3) IE of unknown etiology.1,6,20 The indications for surgery (Table 3) are those reported in the literature21 as well as 3 other indications regarding Staphylococcus epidermidis IE involving prosthetic cardiac valve, early postsurgical IE (<8 weeks), and enterococcal endocarditis. For each case of IE, the therapeutic decision was taken in a multidisciplinary way.
Patients with definite IE were hospitalized while receiving antimicrobial therapy. Before 2002, patients were followed up at regular, nonstandardized consultations with cardiologists and cardiac surgeons. After 2002, the patients were systematically seen for consultation on the same day by the cardiologist and the infectious diseases consultant after 1, 3, 6, and 12 months following the end of treatment. Blood testing was performed (complete blood cell count, erythrocyte sedimentation rate, C-reactive protein, transaminases, creatinine, and rheumatoid factor, with or without blood culture and systematic serologic testing if relapse was suspected) as well as a thoracic radiography and an echocardiography. Patients who did not attend all the consultations were contacted by telephone and summoned to a consultation. The patients considered lost to follow-up were those who could not be reached after the hospital discharge.
The primary end point was all-cause mortality at 1 year after baseline (day of definite diagnosis). The causes of deaths were investigated and classified in different categories (heart failure, multiple organ failure syndrome, embolic complications, and others) with the intent to analyze if there was a difference in the causes of death between the 2 periods studied. The secondary end point was the evaluation of the compliance with the protocol. Compliance with the protocol was evaluated as follows: (1) administration of antimicrobial agents according to the literature guidelines during period 1 or to the protocol during period 2 (yes/no); (2) duration of treatment according to the same guidelines (yes/no); and (3) surgical indications according to the same guidelines (yes/no).
Proportions were compared by using the χ2 test or the Fisher exact test, while continuous variables were compared by using the unpaired t test or the Mann-Whitney test. One-year survival was estimated by the Kaplan-Meier method. The impact of the management protocol on 1-year mortality was tested using the Cox proportional hazards model. Multivariable analysis was systematically adjusted for the main prognostic factors on the basis of epidemiological knowledge (ie, age, sex, the Charlson comorbidity index, mechanical prosthetic valve IE, renal failure, Staphylococcus aureus, acute heart failure, stroke, abscess, and valvular surgery) and for the baseline clinical and microbiological characteristics of the patients, which had changed statistically significantly between the 2 periods. In addition, the calendar year was added as a term of the Cox model to limit confounding by other modifications in endocarditis care over time. The proportional hazards and the linearity assumptions were assessed using residual methods as proposed by Therneau and Grambsch.22 No violations were observed with these methods. Statistical analyses were performed with SPSS software version 10 (SPSS Inc, Chicago, Illinois).
To compare our mortality rate with those reported in the literature, we conducted a literature search using MEDLINE, without time limits, cross-referencing the following terms: mortality rate, in-hospital mortality, and outcome infective endocarditis. Only articles written in English and French and dealing with human cases were reviewed.
During the study period, 466 cases of IE were managed at our hospital, and 133 of them were excluded from the study because they did not meet the inclusion criteria. Among the patients excluded, 47 had a definite IE related to gram-negative rods, fungi, or other rare causes. Therefore, 333 episodes of IE occurring in 321 patients were included in the study. A total of 173 cases of IE were included during period 1 (1991-2001), and 160 cases of IE were included during period 2 (2002-2006), following the implementation of the therapeutic strategy.
The characteristics of the patients included in the study are given in Table 4. Their mean (SD) age was 59.4 (15.9) years, 74.5% were male, and IE occurred in the native valve in 59.7%. Several differences were observed in patients' characteristics between the 2 periods of the study. Patients were significantly older during period 2 (P = .02) than in period 1. A significantly lower number of patients with presumably healthy valves were included during period 2 (P < .001) vs period 1 (Table 4). Moreover, significantly more patients had a Charlson comorbidity index higher than 2 during period 2 compared with period 1 (26.9% in period 2 vs 8.7% in period 1; P < .001). There were significantly more cases of IE due to enterococci during period 2 (P = .008) and significantly fewer cases of IE due to nondigestive streptococci (P = .01). After 1994, the date of implementation of the standardized diagnosis strategy, a significant decrease in cases of IE of unknown etiology was observed (Table 4). The incidence of the cardiac and cerebral complications was similar between the 2 periods.
The rate of administration of an appropriate antimicrobial agent increased significantly from 31.6% during period 1 to 95% during period 2 (P < .001). Therefore, the chance for a patient to receive a complete appropriate antimicrobial therapy was 1.7 times (95% confidence interval [CI], 1.4-2.0 times) higher during period 2 than during period 1. The number of patients treated with both appropriate antimicrobial agent and appropriate duration of treatment also significantly increased from 22.7% during period 1 to 61.8% during period 2 (P < .001) (Table 4). The reasons for noncompliance to the therapeutic protocol during period 2 are detailed in the eTable (http://www.archinternmed.com). The compliance rose from 43.5% in 2002 to 82.3% in 2006 (P < .001). Renal failure significantly decreased in period 2 compared with period 1 (P = .01) (Table 4).
The overall surgery rate (valvular surgery + ablation of pacemaker) was not significantly different between period 1 and period 2 (Table 4) as well as the rate of valvular surgery. Surgery was performed at a median time (range) of 14 (0-150) days and 11 (0-76) days after the beginning of antibiotic therapy in the 2 periods, respectively (P = .35). The surgical protocol was applied in all but 7 patients during period 2 (compliance with the protocol, 92%). No differences were noticed in preoperative and overall use of critical care units between period 1 and period 2, and critical care procedures have not been modified between both periods.
The overall 1-year mortality significantly decreased from 18.5% during period 1 to 8.2% during period 2 (hazard ratio [HR], 0.41; 95% CI, 0.21-0.79 [P = .008]) (Figure). After adjustment for relevant prognostic factors (age, sex, Charlson comorbidity index, mechanical prosthetic valve IE, renal failure, S aureus, acute heart failure, stroke, abscess, and valvular surgery) and the baseline clinical and microbiological characteristics of patients, which had changed significantly between the 2 periods (presumably healthy valve, enterococci, nondigestive streptococci, and IE of unknown etiology), the treatment during period 2 remained independently and strongly predictive of 1-year survival (adjusted HR, 0.25; 95% CI, 0.12-0.52 [P < .001]). This beneficial effect remained when the calendar year was added to the Cox model (adjusted HR, 0.26; 95% CI, 0.09-0.76 [P = .01]) (Table 5).
In-hospital mortality significantly decreased from period 1 to period 2 (12.7% vs 4.4%, respectively; P = .007). The mortality occurring after hospital discharge did not significantly differ between the 2 periods. There was no difference in the time of death from the onset of the disease between periods 1 and 2 (56 days vs 100 days, respectively; P = .11). This decrease in mortality remained when the cases of IE with pacemaker were removed from the analysis (in-hospital mortality rate of 12.9% in period 1 and 4% in period 2; P = .007). Compliance with the surgical protocol was also significantly higher in period 2 than in period 1 among the patients who died (P = .04). Among the patients who died, the compliance with antimicrobial therapy was significantly higher in period 2 compared with period 1, as well as in the whole population. The causes of death during the study are given in Table 6. The deaths from multiple organ failure syndrome and embolic events significantly decreased in period 2 compared with period 1 (P = .003 and .01, respectively [Table 6]).
No relapse was observed during period 2. However, there was no significant difference in relapse and reinfection rates of IE between the 2 periods (Table 4). In addition, there were significantly less patients lost to follow-up in period 2 compared with period 1 (P = .002) (Table 4).
When the patients with rare causes (not eligible to the protocol) were added into the study population, 1-year mortality was 17.7% during period 1 and 9.3% during period 2 (P = .02).
The present study shows a dramatic reduction in mortality between the 2 periods analyzed, suggesting a beneficial impact of our multidisciplinary management strategy of IE. This result might be related to the creation of a local task force, leading to the implementation in 2002 of a therapeutic protocol including both medical and surgical management of IE. Before that, we had implemented a standardized sampling strategy using a diagnostic kit12,14,18,19 that led to a significant decrease in the number of cases of IE of unknown etiology. Thanks to these strategies, it was possible to start empirical treatment of IE as soon as 4 hours after a diagnosis of IE was suspected.
This spectacular improvement in the outcome of IE occurred while the population was significantly older and had more comorbidities23,24 in period 2. In our study, older age was independently associated with mortality, confirming the efficacy of our strategy. This improvement also occurred in a population with statistically significantly fewer presumably healthy valves, which usually had the best outcome, and with a higher rate of coagulase-negative staphylococcal IE, which is usually associated with a bad outcome.25 Compared with the data already published,6- 8,20,23,26- 38 our mortality rate is the lowest reported (Table 7), even when patients not eligible to the protocol were added.
This drop in mortality was due to an important reduction in in-hospital mortality. Posthospital mortality remained stable and could not be changed, contrary to in-hospital mortality. During period 2, patients died significantly less often from multiple organ failure syndrome and embolic events. We also noticed a decrease in renal failure during period 2. The better outcome of IE obtained at our center might be attributed to the extensive experience acquired over the years by cardiologists, surgeons, and microbiologists and to their close collaboration, a fact that has already been reported by other authors.39,40 Anguita Sánchez et al40 have also published an improvement of in-hospital mortality due to a better collaboration with heart surgeons. Thus, these modifications seem to be also related to better global management during the acute stage of the disease. For example, our experience led us to keep patients at rest and to avoid moving them during the acute stage of the disease because we had noticed cases of sudden death during intrahospital transfers for radiological examinations. Moreover, during the acute phase of IE, we decided recently to reduce the indications for computed tomographic scanning because the use of contrast product could potentialize the nephrotoxicity of the antimicrobial agents. Besides, this better management was also due to the simplification of the therapeutic schedule in our protocol (use of 4 antibiotics, one daily dose of aminoglycosides) as well as its standardization. Moreover, in this study, the rate of surgery was higher than those previously reported (from 20% to 52%).6,20,41 Recently, early surgery was reported to be associated with a decreased mortality,42 primarily in S aureus IE43; however, this association was not found in our study.
We succeeded to improve the management of IE because our medical protocol was simplified compared with the published guidelines. When compared with the published recommendations, our medical protocol is easier to use with only 4 antibiotics, whereas up to 8 molecules may be found in some guidelines with several different dosages that make the treatment of IE confusing.44 This complexity and confusion has led to a discordance between guidelines and clinical practice, as reported in the literature10,11 and also found in period 1 of our study. However, despite the fact that our protocol was simple and used by a limited number of physicians in a center specializing in cardiology, the level of compliance with the guidelines was lower than expected, even if the compliance increased from 2002 to reach 82.3% in 2006. The lack of compliance was mainly due to mistakes in the duration of antibiotic therapy and, in only 8 cases, in the nature of the antimicrobial. This reemphasizes the need for a clear and simple protocol to avoid wrong antimicrobial prescriptions. Compliance with surgical indications was higher, probably because they were applied by a restricted number of surgeons.
Finally, our approach to the management of IE is similar to the type of intervention developed by Pronovost et al13 to decrease catheter-related bloodstream infections in the intensive care unit by using an evidence-based intervention. By doing that they were able to obtain a large and sustained reduction in rate of catheter-related bloodstream infection that was maintained throughout the 18-month study.13 The improvement in the outcome of IE obtained by our standardized therapeutic protocol suggests that there is no need to harm patients by subjecting them to randomized studies. Our project was, as Baily45 wrote in the New England Journal of Medicine about the study by Pronovost et al,13 a combination of quality improvement and research on organization. Our project also was not designed to use patients with IE as human subjects to test a new, possibly risky therapeutic method but to promote clinicians' use of procedures already shown to be safe and effective for the purpose.
There were some limitations to this study. This study was nonrandomized, and time trends in patient mix, organisms, and general clinical care could have influenced the results. However, adjusting for calendar year has limited the influence of these factors even if they cannot be entirely excluded. Although it remains difficult to definitely prove a direct relationship between the reduction of mortality and the use of our protocol, we observed an indisputable dramatic decrease in mortality from a disease for which the death rate has remained constant for years.6,46 Bacteriological exclusion criteria could be a potential bias in this mortality rate. However, these microorganisms are rare and cause only 5% of endocarditis cases.1,6,18 Data about the duration of hospitalization at “nonexperienced” hospitals prior to transfer to our center are too lacking to compare whether there is a difference between the 2 periods.
The management of IE by a multidisciplinary medical-surgical team using a standardized protocol to treat IE was associated with a dramatic decrease in mortality, with this rate being among the lowest described in the literature. This standardized management is similar to oncological protocols, with drastic monitoring of compliance and codified and standardized therapeutic indications. The risk of dying from IE is important, and we believe that it is necessary to have a global standardized management to treat this disease; our management-based protocol led to a 2.43-fold decrease in mortality in a 15-year period. Our study also shows the benefit of the creation of a multidisciplinary task force to design protocols and to manage IE in patients on a daily basis. This experience was possible in our hospital because of a good agreement between cardiologists, cardiac surgeons, and microbiologists. Thus, the creation of an “IE committee,”11 like the one at our hospital, should be tested in other centers.
Correspondence: Didier Raoult, MD, PhD, Unité des Rickettsies, CNRS UMR 6236, IFR 48, Faculté de Médecine, Université de la Méditerranée, 27 Blvd Jean Moulin, 13385 Marseille CEDEX 5, France (firstname.lastname@example.org).
Accepted for Publication: April 30, 2009.
Author Contributions: Drs Botelho-Nevers and Thuny contributed equally to this work and had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Thuny and Raoult. Acquisition of data: Botelho-Nevers, Thuny, Casalta, Gouriet, Collart, Riberi, and Raoult. Analysis and interpretation of data: Botelho-Nevers, Thuny, Richet, Habib, and Raoult. Drafting of the manuscript: Botelho-Nevers, Thuny, Casalta, Richet, Gouriet, Collart, Riberi, Habib, and Raoult. Critical revision of the manuscript for important intellectual content: Botelho-Nevers, Thuny, Richet, and Raoult. Statistical analysis: Thuny and Richet. Study supervision: Raoult.
Financial Disclosure: None reported.
Additional Information: An eTable is available at http://www.archinternmed.com.
Additional Contributions: Larry Baddour, MD, reviewed the manuscript. We thank all the infectious diseases residents for collecting the data (Clarisse Rovery, MD, Stéphanie Branger, MD, Séverine Genot, MD, Bema Coulibaly, MD, and Sophie Bayle, MD).