Skip Navigation



European Heart Journal Advance Access published online on February 9, 2009
European Heart Journal, doi:10.1093/eurheartj/ehp008
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow E-letters: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when E-letters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Bannay, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bannay, A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Published on behalf of the European Society of Cardiology. All rights reserved. © The Author 2009. For permissions please email: journals.permissions@oxfordjournals.org

The impact of valve surgery on short- and long-term mortality in left-sided infective endocarditis: do differences in methodological approaches explain previous conflicting results?

Aurélie Bannay1,2,3, Bruno Hoen4, Xavier Duval5, Jean-François Obadia6,7, Christine Selton-Suty8, Vincent Le Moing9, Pierre Tattevin10, Bernard Iung11, François Delahaye12, François Alla1,2,3,* and for the AEPEI Study Group{dagger}

1 Nancy-Université, Faculté de médecine, EA4003, Nancy 54000, France
2 Inserm, CIC-EC, Nancy 54000, France
3 CHU Nancy, Epidemiologie, CO No. 34, Nancy Cedex 54035, France
4 CHU Besançon, Maladies Infectieuses et Tropicales, Besançon 25000, France
5 APHP, Hôpital Bichat Claude Bernard, Centre d'Investigation Clinique, Maladies Infectieuses et Tropicales, Paris 75018, France
6 HCL, Hôpital Cardiovasculaire et Pneumologique Louis Pradel, Chirurgie Cardiothoracique et Transplantation, Lyon-Bron 69500, France
7 HCL, Hôpital Cardiovasculaire et Pneumologique Louis Pradel, Laboratoire de Physiologie Lyon Nord, UCBL1 Inserm U886 ‘cardioprotection’, Lyon, France
8 CHU Nancy, Cardiologie, Nancy 54000, France
9 CHU Montpellier, Maladies Infectieuses et Tropicales, Montpellier 34000, France
10 CHU Pontchaillou, Maladies Infectieuses, Rennes 35000, France
11 APHP, Hôpital Bichat Claude Bernard, Cardiologie, Paris 75018, France
12 HCL, Hôpital Cardiovasculaire et Pneumologique Louis Pradel, Cardiologie, Lyon-Bron 69500, France

Received 16 June 2008; revised 27 November 2008; accepted 5 January 2008.

* Corresponding author. Tel: +33 383852163, Fax: +33 383851205, Email: f.alla{at}chu-nancy.fr


   Abstract
Top
 Abstract
Introduction
 Methods
 Results
 Discussion
 Funding
 Appendix
 Acknowledgements
 References
 
Aims: The aim of this study was to evaluate the effect of valve surgery (VS) in infective endocarditis (IE) on 5-year mortality and to evaluate whether conflicting results reported by previous studies could be due to differences in their methodological approaches.

Methods and results: Four hundred and forty-nine patients with a definite left-sided IE were selected from a prospective, population-based study. Association between VS and 5-year mortality was examined with a Cox model. To determine the impact of different methodological approaches, we also analysed the relationship between VS and mortality in our database, according to each method used in the five previous studies. Valve surgery was performed in 240 patients (53%). It was associated with an increase in short-term mortality [within the first 14 post-operative days; adjusted hazard ratio (HR), 3.69; 95% confidence interval (CI), 2.17–6.25; P < 0.0001] and a decrease in long-term mortality (adjusted HR, 0.55; 95% CI, 0.35–0.87; P = 0.01). At least 188 days of follow-up were required for VS to provide an overall survival advantage. When applying each study's method to our database, we obtained results similar to those reported.

Conclusion: Previous conflicting results appear to be related to differences in statistical methods. When using appropriate models, we found that VS was significantly associated with reduced long-term mortality.

Key Words: Infective endocarditis • Valve surgery • Mortality • Propensity analysis


   Introduction
Top
 Abstract
 Introduction
Methods
 Results
 Discussion
 Funding
 Appendix
 Acknowledgements
 References
 
Treatment of left-sided infective endocarditis (IE) is based on aggressive antibiotic therapy, with or without valve surgery (VS). Over the last years, the rate of VS performed during the course of antibiotic therapy has increased in most industrialized countries and is believed to help reduce mortality in patients with IE.1 However, no randomized controlled trial has ever been conducted to confirm that VS may improve the outcome of IE. Therefore, guidelines for the surgical management of IE are based only on the results of observational studies and expert opinions.2,3 In addition, although the rate of VS has increased over time, the overall mortality in patients with IE has not changed substantially over the last 15–20 years.4

Five recently published observational studies analysed the impact of VS on survival, using propensity analysis, in order to better adjust for potential confounding factors.5–9 Results of these studies are conflicting. Two studies reported that VS was independently associated with reduced mortality.5,8 Two others found no significant relationship between VS and in-hospital survival.6,9 Finally, in the most recent study, VS was associated with increased mortality.7

We conducted a long-term prospective follow-up study of a representative population-based cohort of patients with definite IE, in order (i) to evaluate the potential benefit of VS on 5-year mortality in patients with left-sided IE and (ii) to evaluate whether prior studies' conflicting results could be due to differences in methodological approaches (i.e. patient selection, follow-up duration, and modelling methods).


   Methods
Top
 Abstract
 Introduction
 Methods
Results
 Discussion
 Funding
 Appendix
 Acknowledgements
 References
 
Patient population
Cases of IE were collected during a cross-sectional prospective population-based survey conducted between 1 December 1998 and 31 March 2000 in seven French regions: Franche-Comté, Gironde, Ile-de-France, Lorraine, Marne, Nouvelle-Calédonie, and Rhône-Alpes (total population: 16 million inhabitants). The design of this study has been described in detail previously.10 All consecutive patients with IE suspicion were assessed for inclusion by their physicians. Among the 819 patients assessed for eligibility, 260 were not included (195 patients with no definite IE, 51 patients non-resident in the study area, three non-adult patients, and 11 patients with uncompleted case report form). Of the 559 adult patients with definite IE (according to the Duke criteria11) and living in the study area who were enrolled, 449 patients had left-sided IE and were included in the current study.

Ethics and legal considerations
This study was approved by the Comité consultatif sur le traitement de l'information en matière de recherche dans le domaine de la santé (CCTIRS), the Commission nationale de l'informatique et des libertés (CNIL), and the Institutional Review Board of the Centre Hospitalier Universitaire de Besançon. Patients were informed and gave consent for the study.

Collected data
The following parameters were collected during initial hospital stay: demographic data, comorbidities, predisposing heart conditions (native valve disease, prosthetic valve, previous IE), manifestations of IE (vascular events, immunological manifestations, secondary septic location, splinter haemorrhage, septic shock, fever, C-reactive protein, aseptic meningitis, heart failure), echocardiographic findings (presence and size of vegetation, presence of intracardiac abscess, degree of or presence of a new valvular regurgitation, prosthesis dehiscence, left ventricular ejection fraction), IE location as determined by echocardiographic and/or surgical findings, causative microorganisms, medical and surgical treatment, and outcome.

Pre-existing severe comorbidity was reported by the patient's attending physician and involved at least one of the following diseases: ischaemic cardiomyopathy (previous coronary thrombosis and/or angina pectoris with electrocardiographic signs and/or confirmed coronary lesion), congestive heart failure (presence of heart failure was defined by a class 3 or 4 New York Heart Association status or, when this information was not available, by the need for initiation or increase of diuretic treatment), peripheral vascular disease, previous stroke, chronic pulmonary disease (PaO2 < 60 mmHg during baseline conditions), renal failure (serum creatinine >200 µmol/L and/or regular dialysis), connective tissue disease, immunodeficiency, liver disease, and malignant disease.

Valve surgery was defined by VS performed during initial hospitalization and during the initial antibiotic therapy and within 60 days after admission.

Outcomes
The primary outcome endpoint was the 5-year all-cause mortality rate (follow-up time origin was hospital admission for IE suspicion, follow-up cut-off date was 1 January 2005). Ascertainment and confirmation of deaths were collected prospectively for patients (i) born in France, from administrative registries, and (ii) born abroad, by contact with their general practitioners.

Statistical analysis
Descriptive analysis
Descriptive statistics for quantitative variables are expressed in terms of median and interquartile range except for age (mean and standard deviation) and for qualitative variables as percentages (%). Probabilities of survival were estimated using the Kaplan–Meier method.

Predictors of surgery and prognostic factors
Characteristics of patients who underwent VS and those who did not were compared using Fisher's exact test for categorical variables and unpaired t-test or median test for continuous variables. Bivariable and multivariable ascending stepwise Cox proportional hazard models were used to determine: (i) independent predictors of VS and (ii) independent 5-year survival predictors. All variables listed in Table 1 were candidates to these analyses.


View this table:
[in this window]
[in a new window]

  		Table 1 Patient characteristics (449 patients with a definite left-sided IE)

 
Modelling of association between surgery and mortality
The association between VS and 5-year death rate was examined through an ascending stepwise multivariable Cox proportional hazard regression model, adjusted for survival prognostic factors and predictors of VS, with VS as a time-dependent covariate. A time-dependent variable is a variable whose value may change over time.12 For instance, patients who underwent VS remained in the non-surgical group until the date of surgery, i.e. the ‘surgery’ variable was set to 0 until the date of surgery and to 1 afterwards.

Proportional hazard assumption was assessed by plotting Schoenfeld residuals over time. Log-linearity assumption was assessed by linearity {chi}2 test.

We observed a violation of the proportional hazard assumption for the ‘surgery’ variable, i.e within 14 days following VS, the mortality rate was higher in the surgery group than in the non-surgery group and vice versa thereafter. Consequently, we partitioned the ‘surgery’ variable into two components, according to hazard ratio (HR) distribution over time. Two time-dependent covariates were created. The first covariate indicated whether the patient had undergone VS within the preceeding 14 days, which reflects the short-term effect of VS. The second covariate indicated whether the patient had had surgery more than 14 days before, which reflects the long-term effect of VS. Thus, adjustment for short-term surgical effect can reveal long-term surgical effect.7

Equity point estimation
To estimate a global effect of VS, we estimated the point in time at which early high post-operative risk (adjusted HR of death within 14 days following VS) was offset by later low risk (adjusted HR of death within the 14th day after VS and the end of follow-up) and the point in time at which VS provides an overall survival advantage (Figure 1). This point in time was defined as the ‘equity point’ (EP).13,14 In a graphic of death hazard function over time, the EP is the time at which the area between the surgical group curve and the non-surgical group curve during the short-term period is equal to the area between the surgical group curve and the non-surgical group curve during the long-term period. The 95% confidence interval (CI) of EP time was estimated by Monte Carlo simulation technique.15 This method consists in randomly generating a high number (i.e. 10 000) of indicators (HR and death hazard function), from a normal distribution with observed mean and standard deviation, then calculating corresponding EPs and describing their distribution [median (interquartile range)].


Figure 1
View larger version (14K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 1 Death hazard functions over time. Equity point is the time at which the area between the surgical group curve and the non-surgical group curve during the short-term period (area A) is equal to the area between the surgical group curve and the non-surgical group during the long-term period (area B).

 
Subgroup analyses
We performed a priori planned subgroup analyses using adjusted Cox regression analysis. The association between VS and 5-year survival was examined per following subgroups: gender (men and women), age (< and ≥60 years), history of valvular disease (history of native valve disease and prosthetic valve), IE location (mitral IE, aortic IE, mitral and aortic IE), causative microorganisms (streptococci, Staphylococcus aureus, coagulase-negative staphylococci, others). Subgroups values were compared using {chi}2 tests of interaction.

Comparison with previous studies
To compare our results with those from the five recently published studies,5–9,11 we carried out five different analyses of our database using their methodology. We studied the relationship between VS and death rate, according to the inclusion criteria (native valve IE, prosthetic valve IE, or both), follow-up duration (in-hospital, 6-month, or 5-year survival), modelling method (Cox or logistic regression), and surgery coding (binary or time-dependent), used by each of these five reports. For the comparison with studies that were not restricted to left-sided IE, we used our whole database of 559 patients, which was described previously.16 Specifically, for comparisons with study by:

  1. Tleyjeh et al.,7 we carried out a multivariable Cox modelling of 6-month death rate, with a partitioned time-dependent coding for VS, in our sample of 449 patients with left-sided IE;
  2. Aksoy et al.,5 we carried out a multivariable Cox modelling, for 5-year death rate, with a binary coding for VS, in the whole sample of 559 patients with IE;
  3. Cabell et al.,6 we carried out a multivariable logistic regression modelling of in-hospital mortality, with a binary coding for VS, in a subsample of 447 patients with native valve IE;
  4. Wang et al.,9 we carried out a multivariable logistic regression modelling of in-hospital mortality, with a binary coding for VS, in a subsample of 82 patients with prosthetic valve IE;
  5. Vikram et al.,8 we carried out a multivariable Cox modelling of 6-month death rate, with a binary coding for VS, in a subsample of 372 patients with complicated left-sided native valve IE.
For all these modelling methods, we used the best multivariable adjustment model including prognostic factors and VS predictors.

Power calculation
Two-sided P-value <0.05 was considered to be statistically significant for all analyses. Given the mortality rate, the proportion of surgery, an alpha error of 5%, and an expected protective HR of 0.6 for VS, the power for main analysis was 90%.

All analyses were performed using SAS software version 9.1 (SAS Institute).


   Results
Top
 Abstract
 Introduction
 Methods
 Results
Discussion
 Funding
 Appendix
 Acknowledgements
 References
 
Patient characteristics
Of the 559 adult patients with definite IE who had been enrolled, 449 patients had left-sided IE and were included for primary analysis. Among these patients, 74% were male and mean age was 61 (15) years. About half (48%) of them had no previously known heart disease, 37% had a native valve disease, and 16% had a prosthetic valve. Thirty-seven per cent of the patients had mitral IE, 45% aortic IE, and 17% both mitral and aortic IE. The most frequently identified microorganisms were group D streptococci (27%), followed by oral streptococci (19%) and S. aureus (16%). In 6% of the cases, no microorganism was identified. Other patients' characteristics are displayed in Table 1.

In-hospital and long-term follow-up
Median duration of initial hospital stay was 42 days, and in-hospital lethality was 19%. Two hundred and forty patients (53%) underwent VS, after a median of 20 days after admission. The median follow-up was 5.0 years (the loss to follow-up rate was 12.5% at 5 years). During this period, 160 patients died (including in-hospital lethality), resulting in a 41% 5-year mortality rate.

Predictors of valve surgery
Table 2 shows the differences between patient characteristics according to their treatment group.


View this table:
[in this window]
[in a new window]

  		Table 2 Patient characteristics according to treatment (449 patients with a definite left-sided infective endocarditis)

 
Independent predictors of VS were younger age, tobacco use, the presence of peripheral vascular emboli, heart failure, presence of vegetation >15 mm or absence of vegetation, the presence of intracardiac abscess, and first symptoms of IE not appeared during a hospital stay for other reasons (Table 3).


View this table:
[in this window]
[in a new window]

  		Table 3 Independent predictors of valve surgery (449 patients with a definite left-sided infective endocarditis, adjusted Cox model, n = 436)

 
Prognostic factors
Variables independently associated with 5-year death rate were older age, a high number of serious comorbid diseases, a history of valvular disease, the occurrence of vascular events, the presence of septic shock, and C-reactive protein >120 mg/L (Table 4). All covariates, except septic shock, fulfilled proportional hazard assumption. For patients with septic shock, examination of HR distribution revealed a higher risk of death only during initial hospitalization. In order to take into account this non-proportional hazard, we separated the septic shock variable into two variables: the first one indicated the effect of septic shock on in-hospital mortality and the second one indicated this effect on post-hospital mortality.


View this table:
[in this window]
[in a new window]

  		Table 4 Independent prognostic factors of 5-year death rate (449 patients with a definite left-sided infective endocarditis, adjusted Cox model, n = 449)

 
Relationship between valve surgery and 5-year survival
During follow-up, 61 patients died (25.4%) in the surgical group vs. 99 (47.4%) in the non-surgical group. Five-year survival rates were, respectively, 69.6 and 48.0% (crude P < 0.0001, Figure 2). Valve surgery, assessed as a time-dependent variable, was associated with non-significantly reduced mortality (HR 0.78, 95% CI 0.56–1.1, P = 0.170). This association disappeared after adjustment for prognostic factors and VS predictors (HR 1.01, 95% CI 0.67–1.51, P = 0.958) (Table 5). However, there was an interaction between HR of death and time (P = 0.019); within 14 days following intervention, mortality was higher in the surgery group than in the non-surgery group; but thereafter, it was lower in the surgery group than in the non-surgery group. The partitioned coding revealed that VS was associated with a significant increase in short-term mortality (within 14 days following VS) with an adjusted HR of 3.69 (95% CI 2.17–6.25, P < 0.0001) and with a significant decrease in long-term mortality (95% CI 0.35–0.87, P = 0.01) with an adjusted HR of 0.55 (Table 5). For the whole sample, the EP was obtained after 188 (129–292) days following VS. At least 188 days of follow-up were required for the long-term beneficial effect of VS to compensate for the high post-surgery mortality rate (Figure 1).


Figure 2
View larger version (12K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 2 Overall 5-year survival according to the treatment group (Kaplan–Meier curves).

 

View this table:
[in this window]
[in a new window]

  		Table 5 Five-year death rate hazard ratio (95% confidence interval) of valve surgery under different modelling conditions (449 patients with a definite left-sided infective endocarditis, adjusted Cox model)

 
Subgroup analyses
Despite lack of significant P-values in all subgroups, VS was associated with increased short-term mortality and decreased long-term mortality for all predefined subgroups, except for patients with prosthetic valve IE for whom VS was associated with a non-significant increased long-term mortality. A statistically significant interaction existed between the short-term effect of VS and gender and IE location and between the long-term effect of VS and embolic events (Table 6). Valve surgery was associated with a significantly higher risk of death during the early post-surgery phase in women and in patients having bivalvular left-sided endocarditis. It seemed to have a higher long-term beneficial effect in patients with a history of embolic event.


View this table:
[in this window]
[in a new window]

  		Table 6 Five-year death rate hazard ratio (95% confidence interval) of valve surgery (449 patients with a definite left-sided infective endocarditis, adjusted Cox model): subgroup analysis

 
Comparison with previous studies
Through recreating identical statistical and methodological conditions—in particular, in terms of patient selection, follow-up duration, and surgical variable coding—we obtained results consistent with those of the five previously published reports. Thus, when using a binary coding for surgery variable, we observed a protective effect of VS on mortality, significant at 6 months and 5 years and non-significant during hospitalization. However, when using a partitioned time-dependent coding for surgery variable, we showed, as Tleyjeh et al., a non-significant relationship between VS and 6-month mortality after adjustment for short-term mortality (HR 0.65, 95% CI 0.35–1.21) (Table 7 and Figure 3).


View this table:
[in this window]
[in a new window]

  		Table 7 Relationship between valve surgery and death rate, according to statistical methods

 

Figure 3
View larger version (15K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 3 Relationship between valve surgery and death. Comparison between results from previous reports (black squares) and results from re-analysis from our database using the same methods (white squares). The squares and horizontal lines correspond to the study-specific adjusted hazard ratio (odds ratio for Wang and Cabell), and their 95% confidence interval. aShort-term mortality, 0–7 days after surgery; bmid-term mortality, 8 days to 6 months after surgery.

 

   Discussion
Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
Funding
 Appendix
 Acknowledgements
 References
 
The new major findings of clinical importance in our survey concern the relationship between VS and mortality. Using appropriate models, we found that VS was significantly associated with reduced long-term mortality in patients with left-sided IE. This protective effect remained significant after adjustment for early post-operative mortality and confounding factors. The high early post-operative risk of death was offset by lower late post-operative risk of death as of 6 months after VS for IE. Moreover, our analysis strongly suggests that results between the five major recently published reports5–9 could be only apparently conflicting and in fact be due to differences in the methods used, especially the coding of the surgery variable (binary or time-dependent) and duration of follow-up.

Comparison with previous reports
The five previous studies analysed the impact of VS on survival using propensity score adjustment, matching, and/or stratification to take into account confounding factors. Two of them reported that VS was independently associated with reduced mortality. Both studies used binary surgery coding: one included left-sided native valve IE followed over 6 months8 and the other included patients with left-sided IE followed over 5 years.5 Two other studies, from the International Collaboration on IE Merged Database, used binary coding for surgery and did not include a post-hospital follow-up. They did not find any statistically significant relationship between VS and in-hospital survival for patients with, respectively, native6 and prosthetic9 valve IE. Finally, one study found that VS for patients with left-sided IE provided no survival benefit and could even be associated with increased mortality.7 This last study was the only one that accounted for the impact of the timing of surgery after diagnosis and thus for the possibility of a ‘survivor treatment selection bias’ in the analysis.17 When taking into account the methodological differences between these studies, we showed that the results of these previous studies fit within a logical framework. In particular, our study suggests that the negative impact of VS on mortality observed by Tleyjeh et al. could be explained by an insufficient follow-up period. Indeed, the duration of follow-up in their study was only 6 months, whereas we showed herein that the benefit of VS only appears later, because of early post-operative mortality.

Methodological considerations
The analysis of the reasons for the apparently divergent results of these five studies underlines the need for the use of rigorous and standardized methods. We would like to suggest the following recommendations for further studies on this topic:

(i) To take into account the survival bias by appropriate coding of the ‘surgery’ variable. An analysis in which VS is not regarded as a time-dependent factor ignores the fact that many patients who underwent VS were different from those who did not, simply by virtue of having survived until the date of surgery.18 In other words, longer initial survival may increase the patient's probability of undergoing VS. This may lead to a ‘survivor selection bias’.17 It is worth noting that a recent systematic review showed that survivor selection bias is common in studies published in top medical journals. This bias frequently affects key factors as well as the conclusion of the study: in more than half of the identified investigations, correction of this bias would have changed the study conclusions.19 This bias can be controlled by entering the time of surgery into a Cox model using a time-dependent covariate analysis.7 Time-dependent bias can be addressed with the proper use of time-dependent covariate analysis. Such an analysis, when performed correctly, removes bias that is due to the time-dependant nature of the data.19

(ii) To control potential confounding factors using multivariable models. In an observational study, treatment is not randomly assigned. Therefore, the effect of confounders must be minimized. Confounding factors are factors related to both exposition (surgery) and event (death). Analysis should also adjust for factors that are associated with the treatment assignment (predictors of surgery or propensity factors) and/or factors that are associated with survival rate (prognostic factors). In our study, we confirmed that the methods used to control confounding factors did not have a major influence on the results. In fact, our results with prognostic-based adjustments were similar to those of other studies that used propensity matching. Two recent reviews already showed that in observational studies, using a conventional prognostic model or a propensity model to control confounding factors may lead to similar results.20,21

However, since the decision for VS was not made at inclusion but could occur throughout the follow-up period according to patient evolution, a propensity score calculation, generally obtained from a logistic regression model where surgery is a binary outcome, is not appropriate.22 Apparent predictors of VS could in fact be predictors of early mortality. The following example, obtained from our data, illustrates this point: patients with septic shock were operated on less often than others (42 vs. 55%). However, these patients also died more often (in-hospital mortality was 63% in patients with septic shock vs. 15% in patients without septic shock). When taking into account this early mortality in a Cox regression model, we showed that septic shock was no longer associated with VS (P = 0.698).

However, the use of propensity-based modelling in the case of time-varying treatment would be possible when applying a specific method, as suggested recently.22

(ii) To extend the duration of follow-up. In our study, the protective effect of VS on mortality was not significant until after the first 6 months of follow-up.

Clinical implications
For more than 40 years, VS has been the mainstay in the treatment of IE.23–26 However, decisions about VS are often problematic due to the lack of evidence from prospective randomized controlled trials. Guidelines for the surgical management of IE are only based on results of observational studies and expert opinion.2,3 On the one hand, predictors of VS we found largely reflect practices in accordance with the guidelines. Actually, most of these predictors of VS are recognized as well-identified indications for surgery, particularly presence of heart failure, intracardiac abscess, and risk of systemic embolization (large vegetations and vascular emboli). On the other hand, poor patients' conditions (i.e. old age and IE onset during hospitalization) were related to a lower rate of surgery. More surprising was the relationship between the absence of vegetation and VS. In our database, the absence of vegetation was associated with some characteristics of patients (e.g. prosthetic valve and presence of intracardiac abscess). It can be assumed that the absence of vegetation is a variable that combines several patient characteristics, not taken into account by the adjustment. Likewise, the relationship between tobacco consumption and VS might be explained by such a confounding bias. Therefore, except for patients with a clear indication or contra-indication for valvular replacement, choosing if and when to perform surgery is often a difficult and complex clinical decision. Illustration of this difficulty can be found in the wide variations among studies in the use of VS during IE (from 27 to 51% of patients).10,27 Results from our observational study suggest that current VS practices are beneficial in terms of long-term survival. However, we must be aware that VS in the present study was associated with favourable outcome in those patients for whom IE presentation and patient characteristics led the practitioner to choose this course of therapy. In other words, it does not mean that VS is of benefit and must be performed in all patients with IE. This survey has assessed the effect of VS based on current recommendations and was not designed to propose changes, particularly extensions, of surgery indications.

Furthermore, when considering VS, one should also take into account the balance between negative short-term effects and protective long-term effects of surgery. Further studies are required to develop and validate clinical decision-making tools that integrate short-term risk and long-term benefit, according to patients' characteristics.

Our subgroup analysis showed that patients who developed an embolic event, heart failure, and/or intracardiac abscess during hospitalization were those who benefited the most from VS (the interaction test was significant only for embolic event). These patients with complicated IE also had higher surgery rates. These results are in accordance with major current indications for VS during IE. Thus, patients who seem to benefit most from surgery are those for whom management guidelines indicate surgery.2,3

Interestingly, subgroup analysis suggested that the protective effect of VS on mortality may be restricted to patients with native valve IE (interaction test non-significant). Likewise, Tleyjeh et al.7 also observed, in a subgroup analysis, a negative effect of surgery on mortality for patients with prosthetic valve IE (adjusted HR 2.10, 95% CI 1.07–4.14 vs. adjusted HR 1.17, 95% CI 0.60–2.29 for patients with native valve IE). Whether the subgroup of prosthetic valve patients benefits or not from VS is an important question that could not be answered by our study and that requires further investigation. In fact, results of our prosthesis subgroup analysis were limited by a small number of events in this subgroup, as our study included only 71 patients with prosthesis valve IE.

Strengths and limitations
Compared with other studies in this field, the major strength of our study is that it is the first prospective study to use a contemporary, large database with a comprehensive population-based recruitment and a short accrual period. Patients were recruited from all types of hospital settings (private and public facilities, primary-, secondary-, and tertiary-care hospitals) in a 16-million inhabitant area, over a 16-month period. That makes our database the most extensive representative sample used to study this issue. A population-based survey can limit referral bias,28 which can otherwise lead to misinterpretation of changes in the epidemiological profile of IE.29 Moreover, the prospective 5-year follow-up brings an unprecedented opportunity to accurately assess the long-term impact of valvular surgery on patients' prognosis. However, as in any observational non-randomized study, our findings may still be hampered by biases related to unmeasured or hidden factors and incomplete and/or inexact adjustment,30–32 despite the fact we used rigorous methods to adjust for confounding factors. Besides, all our prognostic factors and predictors of surgery, except two (tobacco use and time between first symptoms and admission), were previously known from the literature.

Conclusion
We performed the first prospective study to assess the relationship between VS and mortality in patients hospitalized for left-sided IE, based on a large unselected patient population, using a long-term follow-up. This study provides the strongest observational evidence to date that VS is significantly associated with reduced long-term mortality in these patients.

Further investigation is required to understand the pathophysiology of the protective effect of VS on long-term survival. Moreover, because of the potential ethical and logistical constraints of conducting a randomized controlled trial of surgery vs. medical therapy alone in patients with IE, further well-designed observational prospective studies are needed to confirm and complete our results; particularly (i) to better estimate the influence of VS on prognosis; (ii) to define which patients would benefit the most from VS; (iii) to define the best timing for VS during the acute phase; and (iv) to compare different surgical procedures.


   Funding
Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 Funding
Appendix
 Acknowledgements
 References
 
This work was funded by the Programme Hospitalier de Recherche Clinique (grant PHRC 1997-RC30) and supported by the following professional organizations: Société de Pathologie Infectieuse de Langue Française, Société Française de Microbiologie, Société Nationale Française de Médecine Interne, Société de Réanimation de Langue Française, Société Française de Gérontologie, Société Française de Cardiologie, Société Française de Chirurgie Thoracique et Cardiovasculaire, Société Française d'Anesthésie-Réanimation, and the Fédération Française de Cardiologie.

Conflict of interest: none declared.


   Appendix
Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 Funding
 Appendix
Acknowledgements
 References
 
The AEPEI (Association pour l'étude et la prévention de l'endocardite infectieuse) study group:Principal investigators: B.H., C. Leport; other members: F.A., I. Béguinot, A. Bouvet, S. Briançon, P. Bruneval, N. Danchin, F.D., X.D., J. Etienne, V. Goulet, V.L.M., J.L. Mainardi, R. Roudaut, R. Ruimy, R. Salamon, C.S.-S., J. Texier-Maugein, and F. Vandenesch. Region study coordinating investigators: Y. Bernard, F. Duchêne, and P. Plésiat (Franche-Comté); T. Doco-Lecompte, C. Selton-Suty, and M. Weber (Lorraine); I. Béguinot, P. Nazeyrollas, and V. Vernet (Marne); B. Garin, F. Lacassin, and J. Robert (New Caledonia); A. Andremont, E. Garbaz, V. Le Moing, C. Leport, J.L. Mainardi, and R. Ruimy (Paris and Ile-de-France); and C. Chidiac, F. Delahaye, J. Etienne, and F. Vandenesch (Rhône-Alpes). Clinical research assistants: S. Boucherit, Y. Bourezane, W. Nouioua, and D. Renaud. Centre National de Référence des Streptocoques: A. Bouvet, G. Collobert, B. Merad, and L. Schlegel. Centre National de Référence des toxémies à Staphylocoques: M. Bes, J. Etienne, and F. Vandenesch.


   Acknowledgements
Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 Funding
 Appendix
 Acknowledgements
References
 
F.A. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.


   Footnotes
 
{dagger} See Appendix for the AEPEI study group. Back


   References
Top
 Abstract
 Introduction
 Methods
 Results
 Discussion
 Funding
 Appendix
 Acknowledgements
 References
 

  1. Slater MS, Komanapalli CB, Tripathy U, Ravichandran PS, Ungerleider RM. Treatment of endocarditis: a decade of experience. Ann Thorac Surg (2007) 83:2074–2079.[Abstract/Free Full Text]
  2. Baddour LM, Wilson WR, Bayer AS, Fowler VG Jr, Bolger AF, Levison ME, Ferrieri P, Gerber MA, Tani LY, Gewitz MH, Tong DC, Steckelberg JM, Baltimore RS, Shulman ST, Burns JC, Falace DA, Newburger JW, Pallasch TJ, Takahashi M, Taubert KA, Committee on Rheumatic Fever, Endocarditis, Kawasaki Disease; Council on Cardiovascular Disease in the Young; Councils on Clinical Cardiology, Stroke, Cardiovascular Surgery and Anesthesia; American Heart Association; Infectious Diseases Society of America. Infective endocarditis: diagnosis, antimicrobial therapy, and management of complications: a statement for healthcare professionals from the Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease, Council on Cardiovascular Disease in the Young, and the Councils on Clinical Cardiology, Stroke, and Cardiovascular Surgery and Anesthesia, American Heart Association: endorsed by the Infectious Diseases Society of America. Circulation (2005) 111:e394–e434.[Abstract/Free Full Text]
  3. Horstkotte D, Follath F, Gutschik E, Lengyel M, Oto A, Pavie A, Soler-Soler J, Thiene G, von Graevenitz A, Priori SG, Garcia MA, Blanc JJ, Budaj A, Cowie M, Dean V, Deckers J, Fernández Burgos E, Lekakis J, Lindahl B, Mazzotta G, Morais J, Oto A, Smiseth OA, Lekakis J, Vahanian A, Delahaye F, Parkhomenko A, Filipatos G, Aldershvile J, Vardas P, Task Force Members on Infective Endocarditis of the European Society of Cardiology; ESC Committee for Practice Guidelines (CPG); Document Reviewers. Guidelines on prevention, diagnosis and treatment of infective endocarditis executive summary: the task force on infective endocarditis of the European Society of Cardiology. Eur Heart J (2004) 25:267–276.[Free Full Text]
  4. Cabell CH, Abrutyn E. Progress toward a global understanding of infective endocarditis. Early lessons from the International Collaboration on Endocarditis investigation. Infect Dis Clin North Am (2002) 16:255–272.[CrossRef][Web of Science][Medline]
  5. Aksoy O, Sexton DJ, Wang A, Pappas PA, Kourany W, Chu V, Fowler VG Jr, Woods CW, Engemann JJ, Corey GR, Harding T, Cabell CH. Early surgery in patients with infective endocarditis: a propensity score analysis. Clin Infect Dis (2007) 44:364–372.[CrossRef][Web of Science][Medline]
  6. Cabell CH, Abrutyn E, Fowler VG Jr, Hoen B, Miro JM, Corey GR, Olaison L, Pappas P, Anstrom KJ, Stafford JA, Eykyn S, Habib G, Mestres CA, Wang A, International Collaboration on Endocarditis Merged Database (ICE-MD) Study Group Investigators. Use of surgery in patients with native valve infective endocarditis: results from the International Collaboration on Endocarditis Merged Database. Am Heart J (2005) 150:1092–1098.[CrossRef][Web of Science][Medline]
  7. Tleyjeh IM, Ghomrawi HM, Steckelberg JM, Hoskin TL, Mirzoyev Z, Anavekar NS, Enders F, Moustafa S, Mookadam F, Huskins WC, Wilson WR, Baddour LM. The impact of valve surgery on 6-month mortality in left-sided infective endocarditis. Circulation (2007) 115:1721–1728.[Abstract/Free Full Text]
  8. Vikram HR, Buenconsejo J, Hasbun R, Quagliarello VJ. Impact of valve surgery on 6-month mortality in adults with complicated, left-sided native valve endocarditis: a propensity analysis. JAMA (2003) 290:3207–3214.[Abstract/Free Full Text]
  9. Wang A, Pappas P, Anstrom KJ, Abrutyn E, Fowler VG Jr, Hoen B, Miro JM, Corey GR, Olaison L, Stafford JA, Mestres CA, Cabell CH, International Collaboration on Endocarditis Investigators. The use and effect of surgical therapy for prosthetic valve infective endocarditis: a propensity analysis of a multicenter, international cohort. Am Heart J (2005) 150:1086–1091.[CrossRef][Web of Science][Medline]
  10. Hoen B, Alla F, Selton-Suty C, Béguinot I, Bouvet A, Briançon S, Casalta JP, Danchin N, Delahaye F, Etienne J, Le Moing V, Leport C, Mainardi JL, Ruimy R, Vandenesch F, Association pour l'Etude et la Prévention de l'Endocardite Infectieuse (AEPEI) Study Group. Changing profile of infective endocarditis: results of a 1-year survey in France. JAMA (2002) 288:75–81.[Abstract/Free Full Text]
  11. Durack DT, Lukes AS, Bright DK. New criteria for diagnosis of infective endocarditis: utilization of specific echocardiographic findings. Duke Endocarditis Service. Am J Med (1994) 96:200–209.[CrossRef][Web of Science][Medline]
  12. Fisher LD, Lin DY. Time-dependent covariates in the Cox proportional-hazards regression model. Annu Rev Public Health (1999) 20:145–157.[CrossRef][Web of Science][Medline]
  13. Charman SC, Sharples LD, McNeil KD, Wallwork J. Assessment of survival benefit after lung transplantation by patient diagnosis. J Heart Lung Transplant (2002) 21:226–232.[CrossRef][Web of Science][Medline]
  14. Luckraz H, Sharples LD, Charman SC, Tsui SS, Wallwork J, Parameshwar J, Large SR. Does heart transplantation confer survival benefit in all risk groups? J Heart Lung Transplant (2005) 24:1231–1234.[CrossRef][Web of Science][Medline]
  15. Fishman GS. Monte-Carlo Concepts, Algorithms, and Applications (1996) New York: Springer.
  16. Duval X, Alla F, Doco-Lecompte T, Le Moing V, Delahaye F, Mainardi JL, Plesiat P, Célard M, Hoen B, Leport C, Association pour l'Etude et la Prévention de l'Endocardite Infectieuse (AEPEI). Diabetes mellitus and infective endocarditis: the insulin factor in patient morbidity and mortality. Eur Heart J (2007) 28:59–64.[Abstract/Free Full Text]
  17. Tleyjeh IM, Baddour LM. The potential impact of survivor treatment selection bias on the perceived efficacy of valve surgery in the treatment of infective endocarditis. Clin Infect Dis (2007) 44:1392–1393.[CrossRef][Web of Science][Medline]
  18. Austin PC, Mamdani MM, van Walraven C, Tu JV. Quantifying the impact of survivor treatment bias in observational studies. J Eval Clin Pract (2006) 12:601–612.[CrossRef][Web of Science][Medline]
  19. Van Walraven C, Davis D, Forster AJ, Wells GA. Time-dependent bias was common in survival analyses published in leading clinical journals. J Clin Epidemiol (2004) 57:672–682.[CrossRef][Web of Science][Medline]
  20. Shah BR, Laupacis A, Hux JE, Austin PC. Propensity score methods gave similar results to traditional regression modeling in observational studies: a systematic review. J Clin Epidemiol (2005) 58:550–559.[CrossRef][Web of Science][Medline]
  21. Sturmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. J Clin Epidemiol (2006) 59:437–447.[Web of Science][Medline]
  22. Lu B. Propensity score matching with time-dependent covariates. Biometrics (2005) 61:721–728.[CrossRef][Web of Science][Medline]
  23. Bonow RO, Carabello BA, Chatterjee K, de Leon AC Jr, Faxon DP, Freed MD, Gaasch WH, Lytle BW, Nishimura RA, O'Gara PT, O'Rourke RA, Otto CM, Shah PM, Shanewise JS, Smith SC Jr, Jacobs AK, Adams CD, Anderson JL, Antman EM, Fuster V, Halperin JL, Hiratzka LF, Hunt SA, Lytle BW, Nishimura R, Page RL, Riegel B. ACC/AHA 2006 guidelines for the management of patients with valvular heart disease: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (writing Committee to Revise the 1998 guidelines for the management of patients with valvular heart disease) developed in collaboration with the Society of Cardiovascular Anesthesiologists endorsed by the Society for Cardiovascular Angiography and Interventions and the Society of Thoracic Surgeons. J Am Coll Cardiol (2006) 48:e1–e148.[Free Full Text]
  24. Braniff BA, Shumway NE, Harrison DC. Valve replacement in active bacterial endocarditis. N Engl J Med (1967) 276:1464–1467.[Web of Science][Medline]
  25. Mullany CJ, McIsaacs AI, Rowe MH, Hale GS. The surgical treatment of infective endocarditis. World J Surg (1989) 13:132–136.[CrossRef][Web of Science][Medline]
  26. Wallace AG, Young WG Jr, Osterhout S. Treatment of acute bacterial endocarditis by valve excision and replacement. Circulation (1965) 31:450–453.[Abstract/Free Full Text]
  27. Chu VH, Cabell CH, Benjamin DK Jr, Kuniholm EF, Fowler VG Jr, Engemann J, Sexton DJ, Corey GR, Wang A. Early predictors of in-hospital death in infective endocarditis. Circulation (2004) 109:1745–1749.[Abstract/Free Full Text]
  28. Steckelberg JM, Melton LJ III, Ilstrup DM, Rouse MS, Wilson WR. Influence of referral bias on the apparent clinical spectrum of infective endocarditis. Am J Med (1990) 88:582–588.[CrossRef][Web of Science][Medline]
  29. Tleyjeh IM, Abdel-Latif A, Rahbi H, Scott CG, Bailey KR, Steckelberg JM, Wilson WR, Baddour LM. A systematic review of population-based studies of infective endocarditis. Chest (2007) 132:1025–1035.[Abstract/Free Full Text]
  30. Grimes DA, Schulz KF. Bias and causal associations in observational research. Lancet (2002) 359:248–252.[CrossRef][Web of Science][Medline]
  31. Laupacis A, Mamdani M. Observational studies of treatment effectiveness: some cautions. Ann Intern Med (2004) 140:923–924.[Free Full Text]
  32. Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med (1997) 127:757–763.[Abstract/Free Full Text]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow E-letters: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when E-letters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Bannay, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bannay, A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?