Summary
We do not know whether surgery during the active phase of infective endocarditis improves prognosis, as no randomized trial data exist. Several observational studies published recently have examined the influence of surgery on prognosis by performing a propensity score analysis. The aim of the present paper is to review these studies, in order to determine whether or not early surgery decreases mortality in adult patients with infective endocarditis. Among nine published studies, 4199 patients were included overall. The rate of surgery during the active phase of infective endocarditis ranged from 23 to 53%. Surgery was significantly beneficial in six studies (adjusted hazard ratios or odds ratios ranging from 0.27 to 0.47), neutral in two studies and without benefit in one study (hazard ratio 1.9; 95% confidence interval 1.1–3.2). Conflicting results appear to be related to differences in statistical methods. When using appropriate models, surgery is significantly associated with reduced long-term mortality. Results from these observational studies suggest that current surgical practices in infective endocarditis are beneficial in terms of long-term survival. However, we cannot conclude that surgery is beneficial and must be performed in all patients with infective endocarditis. Surgery was associated with a favourable outcome in those patients in whom infective endocarditis presentation and patient characteristics led the physicians to perform surgery. Patients who seem to benefit most from surgery are those who fulfil management guidelines (embolic event, heart failure and/or intracardiac abscess).
Résumé
Nous ne savons pas si l’intervention chirurgicale pendant la phase active d’une endocardite infectieuse (EI) en améliore le pronostic car il n’existe pas d’essai randomisé. Plusieurs études d’observation ont été publiées récemment, qui ont examiné l’influence de la chirurgie sur le pronostic de l’EI grâce à des analyses utilisant le score de propension. Le but de cet article est de passer ces études en revue, afin de conclure quant à la diminution, ou non, de la mortalité grâce à l’intervention chirurgicale chez des sujets adultes victimes d’EI. Neuf études ont été publiées, au total 4199 patients ont été inclus. Le taux d’intervention chirurgicale pendant la phase active d’EI allait de 23 à 53 %. La chirurgie était significativement bénéfique dans six études ( odds ratios ou hazard ratios entre 0,27 et 0,47), sans effet dans deux études, et significativement non bénéfique dans une étude ( hazard ratio 1,9 ; intervalle de confiance à 95 % 1,1–3,2). Ces résultats discordants semblent être liés à des différences de méthodes statistiques. Lorsque des modèles appropriés sont utilisés, la chirurgie est significativement associée à une réduction de la mortalité à long terme. Les résultats de ces études d’observation suggèrent que les pratiques chirurgicales actuelles dans l’EI sont bénéfiques en termes de survie à long terme. Cependant, nous ne pouvons pas conclure que la chirurgie est bénéfique, et doit être réalisée, chez tous les patients avec EI. La chirurgie est associée à une issue favorable chez les patients chez lesquels leurs caractéristiques et la présentation de l’EI incitaient les médecins à réaliser l’intervention chirurgicale. Les patients qui semblent bénéficier le plus de la chirurgie sont ceux qui correspondent aux recommandations d’intervention chirurgicale (embolie, insuffisance cardiaque, abcès intracardiaque).
Background
Although relatively rare, IE remains a severe disease; the in-hospital mortality rate is around 20% overall, but is much higher in complicated cases. Prognosis can be improved in several ways, one of which is avoiding delay to cardiac surgery in complicated cases; this remains the main issue for physicians involved in the care of patients with IE.
Historically, the dogma was to avoid surgery during the active phase because tissues are inflamed and infected, which makes surgery very difficult and leads to high postoperative mortality and a high risk of valve dysfunction. This belief has changed dramatically over the past two decades owing to improvements in surgical technique. Cardiac surgeons no longer refuse to operate at a very early stage in the course of the disease. In France, the rate of surgery increased from 30% in 1991 to 50% in 1999 . In a systematic review of 15 studies, the rate of surgery increased by 7% per decade between 1969 and 2000 .
Indications for surgery are well defined in guidelines, but they are based on expert opinion . We still do not know whether surgery during the active phase of IE improves a patient’s prognosis, as no randomized trial data exist. Several observational studies published recently have examined the influence of surgery on prognosis by using propensity score analysis . In 1983, Rosenbaum and Rubin proposed ‘propensity scores’ as a method of controlling for selection (here, treatment-selection) bias in observational studies . A patient’s propensity score is his/her conditional probability of a particular exposure (here, surgery) versus another, given the observed confounders; it can be estimated using logistic regression, modelling the exposure as the dependent variable and the potential confounders as the independent variables . The most recent studies controlled for another bias, survivor bias, which occurs because patients who live longer are more likely to undergo surgery than those who die early.
The aim of the present paper is to review these studies, in order to determine whether or not early surgery is beneficial in adults with IE. Papers were identified using Medline ( http://www-ncbi-nlm-nih-gov.easyaccess1.lib.cuhk.edu.hk/sites/pubmed ). The search strategy selected articles that contained ‘propensity’ and ‘endocarditis’ as text words or Medical Subject Headings, and papers cited in the reference list of the selected articles. Papers are presented in chronological order of publication. Data from the studies are summarized in Table 1 .
| Reference | Setting | Inclusion period | Population | Number of patients | Percentage operated (%) | Outcome measure | Bias adjusted for | Effect of surgery on mortality | |
|---|---|---|---|---|---|---|---|---|---|
| Univariate analysis | Propensity + multivariable analysis | ||||||||
| Vikram et al. | Seven hospitals in Connecticut, USA | 1990–99 | Complicated, left-sided NVE | 513; 218 propensity-matched | 45 | 6-month all-cause mortality | Treatment-selection bias | Operated patients 16%; non-operated patients 33%; HR 0.43; p < 0.001 | HR 0.40; 95% CI 0.18–0.91; p = 0.03 |
| Mourvillier et al. | ICUs in French teaching hospital | 1993–99 | NVE or PVE hospitalized in an ICU | 228 NVE; 54 propensity-matched | 46 | In-hospital mortality | Treatment-selection bias | NVE: operated patients 29%; non-operated patients 47%; p = 0.01 PVE: operated patients 44%; non-operated patients 64%; p = 0.04 | OR 0.47; 95% CI 0.22–1.00; p = 0.05 |
| Cabell et al. | International registry | 1985–99 | NVE | 1516; 1497 in the propensity groups | 40 | In-hospital mortality | Treatment-selection bias | Operated patients 13.6%; non-operated patients 16.4%; p = 0.14 | Group 1 (lowest likelihood for surgery): medical therapy, 9.5%; surgical therapy, 20.0%; p = 0.16 Group 5 (strongest likelihood for surgery): medical therapy, 38.0%; surgical therapy, 11.2%; p < 0.001 |
| Wang et al. | International registry | 1985–99 | PVE | 355; 136 propensity-matched | 42 | In-hospital mortality | Treatment-selection bias | Operated patients 25.0%; non-operated patients 23.4%; p = 0.73 | OR 0.56; 95% CI 0.23–1.36; p = 0.20 |
| Aksoy et al. | Duke University, Durham, NC, USA | 1996–2002 | Left-sided NVE or PVE without intracardiac device | 333; 102 propensity-matched | 23 | 5-year all-cause mortality | Treatment-selection bias | Estimated; operated patients 30%; non-operated patients 47%; p = ? | HR 0.27; 95% CI 0.13–0.55; p = ? |
| Tleyjeh et al. | Mayo Clinic, Rochester, MN, USA | 1980–98 | Left-sided NVE or PVE | 546; 186 propensity-matched | 24 | 6-month all-cause mortality | Treatment-selection bias; survivor bias | Operated patients 27.1%; non-operated patients 23.7%; p = ? | Matched cohort: HR 1.3; 95% CI 0.5–3.1; p = 0.56 Whole cohort, surgery as a time-dependent covariate: HR 1.9; 95% CI 1.1–3.2; p = 0.02 After adjustment for early (operative) mortality: HR 0.9; 95% CI 0.5–1.8; p = ? |
| Bannay et al. | Prospective, population-based study in several French regions | 1999 | Left-sided NVE or PVE | 449 | 53 | 5-year all-cause mortality | Treatment-selection bias; survivor bias | Operated patients 30%; non-operated patients 52%; p < 0.0001 | Within 14 days after intervention, mortality was higher in the surgery group: adjusted HR 3.69; 95% CI 2.17–6.25; p < 0.0001 Thereafter, it was lower in the surgery group: adjusted HR 0.55; 95% CI 0.35–0.87; p = 0.01 |
| Sy et al. | Two university teaching hospitals in Sydney, Australia | 1996–2006 | Left-sided NVE or PVE | 223 | 28 | 5.2-year all-cause mortality | Treatment-selection bias; survivor bias | Operated patients 32%; non-operated patients 51%; p = 0.02 | After adjustment for baseline differences in propensity for surgery and risk of mortality: HR 0.50; 95% CI 0.28–0.88; p = 0.02 After time-dependent analysis: HR 0.77; 95% CI 0.42–1.40; p = 0.39 |
| Lalani et al. | International registry | 2000–2005 | NVE | 1552; 1238 in the propensity groups | 46 | In-hospital mortality | Treatment-selection bias; survivor bias; hidden bias | Operated patients 12.1%; non-operated patients 20.7%; p < 0.001 | ARR, −11.2%; OR 0.44; 95% CI 0.33–0.59; p < 0.001 |
Study review
Vikram et al. conducted a retrospective, observational cohort study of 513 patients with complicated, left-sided NVE . Complication was defined as ‘clinical complication for which surgery is considered in current clinical practice: CHF; new valve regurgitation; refractory infection; systemic embolization to vital organs; or presence of a vegetation on echocardiography’. Forty-five percent of patients underwent surgery. In the 6-month period after baseline (defined as the date of surgery or the date of decision not to operate), 26% of the patients died. In unadjusted analyses, surgery was associated with reduced mortality (16% vs 33%; HR 0.43; CI 0.29–0.63). After adjustment for heterogeneity (baseline features that were associated with mortality: hospital site; comorbidity assessed using the Charlson index ; CHF; microorganism; immunocompromised state; abnormal mental status; and refractory infection), surgery remained associated with reduced mortality (HR 0.35 [0.23–0.54]; p < 0.001), as in further analyses of 218 patients matched by propensity scores (15% vs 28%; HR 0.45 [0.23–0.86]; p = 0.02). After additional adjustment for confounding (patient selection bias due to non-randomized assignment of treatment) within the propensity-matched group, surgery remained significantly associated with reduced mortality (HR 0.40 [0.18–0.91]). In this propensity-matched group, patients with moderate-to-severe CHF showed the greatest reduction in mortality with surgery (14% vs 51%; HR 0.22 [0.08–0.53]; p = 0.01), whereas there was no benefit of surgery among patients with none-to-mild CHF. Vikram et al. did not study the impact of surgery on the prognosis of all IE: they included only patients with complicated IE (those for whom surgery is usually advised) and they excluded prosthetic valve IE (PVE) and right-sided IE. We do not know which subgroup analyses other than CHF the investigators performed. All hospitals participating in the study had surgical facilities. This first paper using propensity analysis confirmed what we, as clinicians, intuitively knew about the beneficial effect of surgery in complicated IE.
A retrospective study was performed in the two medical intensive care units of a French teaching hospital . To evaluate the potential role of surgery in the prognosis of NVE, the investigators performed a nested case-control study comparing 27 operated with 27 non-operated patients using propensity scores. Among 228 patients with definite IE (146 NVE and 82 PVE), 46% underwent cardiac surgery. The overall in-hospital mortality rate was 45%. In patients with NVE, mortality was significantly lower in operated patients than in non-operated patients ( p = 0.014). Cardiac surgery was independently associated with in-hospital mortality (OR 0.47). In contrast, in the nested case-control study, operated patients did not have lower in-hospital mortality than non-operated patients (OR 0.96; not significant). In patients with PVE, mortality was lower in the surgery group ( p = 0.04). However, surgery was not an independent predictor of lower mortality. The authors concluded that surgery appears to improve in-hospital mortality. The in-hospital mortality rate was very high because patients were recruited from intensive care units. There were very few propensity-matched patients, which may explain why the authors did not observe a beneficial effect of surgery in these patients, whereas there was a significant beneficial effect of surgery in the cohort as a whole.
The ICE merged database is a large, multicentre, international registry of patients with definite IE, as defined by the Duke criteria. Surgery was performed in 40% of 1516 patients with NVE . In-hospital mortality was 13.6% in the surgery group and 16.4% in the no surgery group (not significant). The investigators constructed a multiple logistic regression model to create equally sized propensity groups for the likelihood of surgery. Within group 1 (lowest likelihood of surgery), medical therapy had a non-significant advantage over surgery in terms of in-hospital mortality ( Fig. 1 ). This observation was reversed in groups with a moderate likelihood of surgery (groups 2 to 4), with a suggestion of survival benefit with surgery. In the group of patients with the strongest indicators for surgery (group 5), surgery had a significant survival benefit compared with medical therapy alone ( p < 0.001) even after adjustment for multiple comparisons. Patients with the fewest predictors for surgery (group 1) were more likely to be female and to have an oral streptococcal infection compared with the other groups. Patients with the most predictors for surgery (group 5) were more likely to be male, and to have aortic valve involvement, CHF or an intracardiac abscess. The authors concluded that the benefits of surgery are not seen uniformly in all patients with NVE, but are mostly demonstrated within a target population. Surgery was performed in 42% of the 355 patients with PVE . In-hospital mortality was similar for patients treated with surgery compared with those treated with medical therapy alone (25.0% vs 23.4%; not significant). After adjustment for factors independently related to surgery (age, microorganism, intracardiac abscess, CHF), in-hospital mortality was predicted by brain embolization (OR 11.1 [4.2–29.7]) and Staphylococcus aureus infection (3.7 [1.4–9.7]), with a trend toward a beneficial effect of surgery (0.56 [0.23–1.36]). The ICE database is by far the largest database of those presented here. The ICE investigators separated NVE and PVE. There was no HR or OR given for surgery in NVE, but Fig. 1 clearly shows that surgery is not beneficial in uncomplicated cases but is beneficial in complicated cases.
