Analysis of the Impact of Early Surgery on In-Hospital Mortality of Native Valve Endocarditis
Use of Propensity Score and Instrumental Variable Methods to Adjust for Treatment-Selection Bias
Background— The impact of early surgery on mortality in patients with native valve endocarditis (NVE) is unresolved. This study sought to evaluate valve surgery compared with medical therapy for NVE and to identify characteristics of patients who are most likely to benefit from early surgery.
Methods and Results— Using a prospective, multinational cohort of patients with definite NVE, the effect of early surgery on in-hospital mortality was assessed by propensity-based matching adjustment for survivor bias and by instrumental variable analysis. Patients were stratified by propensity quintile, paravalvular complications, valve perforation, systemic embolization, stroke, Staphylococcus aureus infection, and congestive heart failure. Of the 1552 patients with NVE, 720 (46%) underwent early surgery and 832 (54%) were treated with medical therapy. Compared with medical therapy, early surgery was associated with a significant reduction in mortality in the overall cohort (12.1% [87/720] versus 20.7% [172/832]) and after propensity-based matching and adjustment for survivor bias (absolute risk reduction [ARR] −5.9%, P<0.001). With a combined instrument, the instrumental-variable–adjusted ARR in mortality associated with early surgery was −11.2% (P<0.001). In subgroup analysis, surgery was found to confer a survival benefit compared with medical therapy among patients with a higher propensity for surgery (ARR −10.9% for quintiles 4 and 5, P=0.002) and those with paravalvular complications (ARR −17.3%, P<0.001), systemic embolization (ARR −12.9%, P=0.002), S aureus NVE (ARR −20.1%, P<0.001), and stroke (ARR −13%, P=0.02) but not those with valve perforation or congestive heart failure.
Conclusions— Early surgery for NVE is associated with an in-hospital mortality benefit compared with medical therapy alone.
Received March 31, 2009; accepted December 21, 2009.
Native valve endocarditis (NVE) is associated with mortality rates of 15% to 30%, and despite advances in diagnosis and treatment, mortality rates remain largely unchanged.1–5 Consensus guidelines for the treatment of NVE advocate the use of early valve surgery for complications such as congestive heart failure, systemic embolization, or intracardiac damage, but there are insufficient data to support such recommendations.6 Ethical, logistical, and financial issues create major challenges to the performance of randomized controlled treatment trials for this relatively infrequent disease. Hence, investigators have focused on observational studies to compare outcomes in patients treated with surgery versus medical therapy for NVE using techniques such as propensity analyses to control for bias related to measured patient characteristics. However, these techniques do not adjust for survivor bias (patients who live longer are more likely to undergo surgery than patients who die early) or hidden bias (unmeasured patient characteristics that affect both the decision to treat and the outcome).7,8
Editorial see p 960
Clinical Perspective on p 1013
To date, 6 propensity analyses evaluating treatment strategies for NVE have been performed.2–5,9,10 Two studies reported a significant reduction in 6-month and 5-year mortality associated with valve surgery.2,3 A third study reported that the in-hospital mortality benefit of surgery was limited to patients with the highest propensity scores for surgery.5 In contrast, other studies have demonstrated either no benefit or increased mortality associated with surgery.4,9,10 This disparity in results is compounded by methodological limitations, including retrospective data collection, small sample sizes, and single-center studies. Finally, although the reported propensity-score analyses have controlled for overt or measured bias, none of these studies have adjusted for hidden bias.7
The objectives of the present study were to assess whether early surgery is associated with lower in-hospital mortality compared with medical therapy and to determine whether this association varied by propensity or specific indications for early surgery. We used a prospective, multinational cohort of patients with NVE and used statistical methods to control for treatment-selection bias, survivor bias, and hidden bias.
Study Population and Clinical Data
The cohort for the present study was obtained from the International Collaboration on Endocarditis–Prospective Cohort Study (ICE-PCS) database, which contains 2760 patients with definite infective endocarditis (IE) as defined by the modified Duke criteria.11 The background and inclusion criteria of this prospective, multicenter, international registry of IE have been reported previously.1,12 Briefly, data on patients with IE from 61 centers in 28 countries were collected prospectively between June 2000 and August 2005. The study was approved by the institutional review board or ethics committee at all of the participating sites.
Only patients who had definite left- or right-sided NVE, based on modified Duke criteria,11 were included in the present study. Patients with the following characteristics were excluded: injection drug use, prosthetic valves, non–native-valve IE (eg, pacemaker IE); receipt of surgery before admission; and missing values for gender, receipt of surgery, and in-hospital death. To preserve the assumption of independence of observations, only the first episode of IE recorded for an individual patient was used. For missing data in ICE-PCS, sites and their investigators were queried to complete data collection. All of the variables related to complications or outcomes of NVE had data collected for ≥97% of patients. Missing values for clinical outcomes were imputed with the negative category for categorical variables.
Definitions of the variables have been reported elsewhere.13 Early surgery was defined as replacement or repair of the affected valve during the initial hospitalization for IE. Chronic illness was defined as the presence of chronic comorbidities such as diabetes mellitus, cancer, immunosuppression, hemodialysis dependence, chronic obstructive pulmonary disease, and cirrhosis. Paravalvular complication was defined as the presence of an intracardiac abscess or fistula by transthoracic or transesophageal echocardiography. Systemic embolization was defined as embolism to any major arterial vessel, excluding stroke. Healthcare-associated IE consisted of either nosocomial or nonnosocomial healthcare–associated infection.14
We used an observational cohort to estimate the impact of early surgery on mortality with statistical methods to control for overt and hidden biases. Overt treatment bias, related to covariates measured in the study, was addressed by use of propensity-score matching and multivariate regression analysis. To eliminate survivor bias, each patient in the medical therapy group was required to have survived at least as long as the time to surgery in the matched surgically treated patient. Instrumental variable analysis was used to control for all types of potential bias, including hidden or unmeasured bias. We also performed a subgroup analysis stratifying patients by propensity-score quintiles, paravalvular complications such as an abscess or fistula, valve perforation, systemic embolization, stroke, congestive heart failure, or Staphylococcus aureus infection.
Standardized differences between the 2 treatment groups were calculated as the differences in the means divided by the pooled standard deviation and expressed as a percentage. The primary outcome was all-cause mortality during initial hospitalization. Differences in mortality between treatment groups are reported in terms of absolute risk reduction (ARR) and odds ratios, in accordance with recent recommendations.15,16 A 2-sided probability value <0.05 was considered statistically significant. All analyses were performed with STATA software, version 10 (StataCorp, College Station, Tex).17
A propensity score, which was the estimated probability that a patient would undergo early surgery, was calculated for each patient. The propensity score was computed with nonparsimonious multivariable logistic regression, with early surgery as the dependent variable; it incorporated 25 clinically relevant covariates and 3 interaction terms as the independent variables (Table I in the online-only Data Supplement).
Patients undergoing early surgery were matched on a 1-to-1 basis with patients treated medically on the basis of the following: (1) propensity score, by use of nearest-neighbor matching with replacement (each medical therapy patient could be used more than once for matching, whereas surgically treated patients were matched once only) and a matching tolerance (caliper) of 0.05; and (2) follow-up times, such that each patient in the medical group survived at least as long as the time to surgery in the surgical patient.
Instrumental Variable Analysis
The present study used an observational cohort, and therefore, assignment to early surgery or medical therapy depended on patient characteristics rather than randomization. Given the limitations of observational data collection, omission of covariates that influence treatment assignment and the associated outcome may have occurred. The conventional probit (or logit) approach assumes that after controlling for measured characteristics of the patient, there are no unmeasured characteristics that influence both the decision to treat and the outcome of treatment. Instrumental variable analysis is an econometric method used to control for the possible existence of hidden bias (ie, caused by the omission of relevant covariates).7 We evaluated several candidate instrumental variables (separately and as a combined instrument) with the following key characteristics: (1) high correlation with early surgery and (2) no effect on mortality independent of its effect through early surgery. A combined instrument consisting of the following variables fulfilled these criteria: evidence of NVE on transthoracic echocardiography, echocardiography performed at the referral hospital, duration of symptoms greater than 1 month before presentation, site-specific rate of early surgery for NVE, transfer from another hospital, and performance of transesophageal echocardiography. The specific instrumental variable estimation method we used is a STATA routine called BIPROBIT. This is a simultaneous maximum-likelihood procedure that estimates equations for mortality and treatment simultaneously, to enhance the efficiency (precision) of the estimates.18,19 This bivariate probit (biprobit) model also explicitly takes into account the fact that both the outcome (mortality) and the treatment (surgery) are “0, 1” indicator variables. To evaluate whether early surgery was an exogenous variable in the mortality model (ie, the effect of early surgery on mortality was independent of all measured and unmeasured covariates), we performed a χ2 test of the null hypothesis that early surgery is exogenous.
Clinically plausible variables known to affect the decision to perform valve surgery were used to perform subgroup analysis to determine characteristics associated with maximum mortality benefit. The propensity-matched, survivor-bias–adjusted cohort (n=1238 patients) was divided into quintiles based on the propensity scores for surgical patients, and the differential in-hospital mortality between matched surgical and medically treated patients in each quintile was then computed. Patients were also stratified by presence or absence of paravalvular complications, valve perforation, systemic embolization, stroke, S aureus infection, or congestive heart failure. The subgroup analysis was performed by adding interaction terms to the logistic regression model. In the propensity analysis of the effects of surgery, clustered standard errors were estimated to account for matching with replacement of controls, and the reported probability values reflect this.
The ICE-PCS cohort consisted of 2760 patients, including 1859 (67%) with NVE. Of these, 1552 (83%) qualified for the present study (Figure I in the online-only Data Supplement). Seven hundred twenty (46%) patients underwent early surgery, and 832 (54%) were treated with medical therapy (Table 1). Patients who were treated with early surgery were younger; more likely to be male and to have transferred from another medical facility; and more likely to have complications such as congestive heart failure, intracardiac abscess, and paravalvular complications. The median time from admission to surgery was 7 days (quintile 1 to quintile 3: 2 to 17 days). S aureus was the most common pathogen in patients receiving medical therapy. Early surgery was associated with a significant reduction in mortality in an unadjusted univariate analysis (12.1% versus 20.7%; ARR −8.6%, P<0.001) and after controlling for all of the other measured covariates by standard logistic regression analysis (ARR −5.9%, P<0.001; Table II in the online-only Data Supplement). The latter model had a concordance index of 0.83, which indicates a strong ability to discriminate between possible outcomes (ie, in-hospital survival versus death).
To control for observed differences in patient characteristics, we estimated the probability of surgery (propensity score) for each patient. After nearest-neighbor propensity-score matching with replacement and adjustment for survivor bias, 619 patients who underwent early surgery were each matched with a medical therapy patient to yield a quasi-randomized sample. Of the 619 medical therapy control subjects used for matching with replacement, 120 (19.4%) were selected more than once. Standardized differences between covariates in the 2 treatment groups decreased substantially after propensity matching and survivor bias adjustment (Table 1). The mortality benefit associated with early surgery persisted after propensity matching, adjustment for survivor bias, and controlling for confounders (ARR −5.9, P<0.001; Table 2).
Next, instrumental variable analysis (with the biprobit technique) was performed to adjust for hidden bias. Candidate variables for this composite instrumental variable were selected on the basis of clinical judgment that each would influence or increase the probability of surgical treatment without otherwise being associated with in-hospital mortality. The c-statistic for the surgery equation in the biprobit model was 0.86. The Hosmer-Lemeshow test statistic of 10.91 and associated probability value of 0.207 implied that we could not reject the null hypothesis of no difference between the observed values of surgery and the values predicted by the surgery equation. The instrumental-variable–adjusted mortality reduction with early surgery versus medical therapy was −11.2% (P<0.001). We were unable to reject the hypothesis that early surgery was an exogenous variable (ie, its effects on mortality were independent of all of the measured and unmeasured covariates) in the mortality model (χ2=0.51, P=0.48). A comparison of the adjusted odds ratios and ARRs associated with early surgery and medical therapy is shown in Table 2.
To determine the impact of early surgery on mortality across different strata, subgroup analyses were performed on the matched cohort (Table 3). For the quintile-stratification analysis, patients were divided into 5 subgroups based on the propensity scores of surgical patients. The distribution of key characteristics across the propensity score quintiles is depicted in the Figure. With 1238 patients, this yielded approximately 248 patients (about half surgical and half medical therapy patients) per quintile. A differential benefit of surgery that favored patients with a higher propensity for surgery compared with those with a lower propensity score was observed (ARR −10.9% [P=0.002] for patients in quintiles 4 and 5 versus −2.4% in patients in quintiles 1 to 3 [P=0.2]; P for difference=0.029). A mortality benefit associated with early surgery was also found in patients with paravalvular complications, systemic embolization, S aureus NVE, and stroke but not in those with valve perforation or congestive heart failure. Finally, a differential, greater benefit of surgery was observed in the presence of paravalvular complications, systemic embolization, and S aureus NVE than in NVE without these characteristics.
Early surgery is performed in a high percentage of patients with NVE, generally in patients with a complicated clinical course for whom medical therapy is deemed inadequate. During the last 3 decades, observational studies have yielded conflicting conclusions regarding the use of early surgery for complicated NVE, and optimal patient selection has not been determined.2–5,9,10 The results of the present study demonstrate that early surgery is associated with significant in-hospital mortality benefit compared with medical therapy, even after adjustment for important biases such as treatment selection, survivorship, and hidden biases.
Table 4⇓ summarizes the studies that have used propensity analysis to evaluate the effect of surgery for NVE. Although most studies have shown a mortality benefit associated with early surgery, others have demonstrated no benefit or possible harm. Tleyjeh et al4 reported that surgery offered no significant reduction in 6-month mortality (hazard ratio 0.92, 95% CI 0.48 to 1.76) after adjusting for treatment and survivor bias. There are several methodological differences in the design of the present study compared with these prior studies that may explain the disparate results. Prior investigations used retrospective, single-center data; included both NVE and prosthetic valve endocarditis in their analyses; and had different end points (eg, inpatient versus 6-month mortality).3–5,9,10 Propensity-based matching reduces sample size (most studies have been limited to matched cohorts of ≤100 patients), thus reducing the power to detect small differences in mortality and to evaluate the efficacy of treatment strategies in the different propensity quintiles or other subgroups. In contrast, the present study used the largest contemporary, multinational cohort of prospectively enrolled patients and did not include patients with prosthetic valve endocarditis, for whom indications for surgery may differ from NVE. Even after matching on the basis of propensity scores and survival times, each treatment group had 619 patients. The present results indicate that early surgery was associated with an absolute reduction in mortality in the overall cohort of 5.9% to 11.2% compared with medical therapy; the number of patients needed to treat with early valve surgery to prevent 1 in-hospital death ranged from 9 to 17. This beneficial effect of early surgery is also discrepant from results reported recently by Bannay et al20 of a study in which valve surgery was associated with an early increase in mortality within 14 days of surgery and the survival benefit of surgery was not evident until >6 months after surgery.
On a broader level, the present study used the technique of propensity modeling in a rigorous manner to evaluate nonrandomized treatment with surgery, which reflects the concerns raised by Austin15 in his assessment of 44 cardiovascular studies. In addition to the large sample size and prospective nature of the present investigation, other advantages include (1) a thorough description of the matching method; (2) complete reporting of the balance in baseline variables between surgically treated and untreated patients; (3) appropriate statistical methods for estimating the effect of early surgery on in-hospital mortality, particularly adjustment for survival bias; and (4) appropriate statistical methods for subgroup analysis, including the use of interaction terms to account for any treatment differences in outcome that may be related to the patient subgroup and the use of clustered standard errors to account for matching with replacement of controls.4
Although surgery was shown to confer a mortality benefit over medical therapy alone for NVE in the overall cohort, important caveats are suggested from the subgroup analyses. Similar to results previously published by our group,5 the benefit of early surgery was not distributed uniformly among all of the propensity quintiles. A differential benefit of surgery was observed in patients with strong indications for surgical intervention (ie, those in the combined fourth and fifth quintiles). Surgery was also found to confer a survival benefit among patients with paravalvular complications, systemic embolization, and stroke but not among those with valve perforation or congestive heart failure. In addition, a significant differential benefit of early surgery was observed in patients with S aureus NVE, yet our unadjusted analysis and prior reports show that patients with S aureus NVE are less likely to undergo early surgery because of healthcare-associated infection and multiple comorbid conditions.2,3,5 Even after adjustment for treatment and survivor bias, patients with S aureus NVE who underwent early surgery had a 20.1% ARR in mortality compared with patients who received medical therapy (number needed to treat=5).
Propensity adjustment may be influenced by unmeasured variables, which necessitates the use of instrumental variable analysis to adjust for hidden biases. With the use of a single-instrument approach, it is less likely that the variable is inadvertently associated with mortality except via its effect on the intervention.7,21 Even in a relatively large cohort of NVE patients, we were unable to predict the instrumented values of early surgery with adequate precision using a single variable. With the use of a combined instrument, the adjusted ARR in mortality associated with early surgery was 11.2% (P<0.001). Using the χ2 test for exogeneity, we were unable to reject the null hypothesis that surgery was exogenous. This finding implies that in the present sample, the effect of early surgery on mortality was independent of all measured and unmeasured covariates. The estimated effects of surgery before correction for propensity score and survivor bias thus should be similar to the estimated effects after these corrections. Indeed, the estimated effects of surgery were consistent across techniques, which supports our conclusion that early surgery is an independent predictor of in-hospital mortality and is associated with a mortality benefit compared with medical therapy.
The present study has several limitations. The ICE cohort may be influenced by referral bias, because most centers are tertiary care centers with voluntary participation. Thus, the results of the present study may not be generalizable to the global epidemiology, treatment, and outcomes of NVE. Limitations associated with data collection were also present. Although the ICE-PCS case report form captures the occurrence of events such as congestive heart failure and stroke, the timing of such events is not recorded, which potentially affects the reliability of the surgical propensity model. Such complications are more likely to occur soon after hospitalization and determine whether surgical intervention is indicated. Surgery was not found to confer a survival benefit for patients with heart failure. Evaluation of the effect of surgery as a function of heart failure severity, previously reported to be limited to patients with moderate or severe heart failure,3 was not feasible in the present cohort owing to incomplete collection of this variable (based on New York Heart Association classification). Although early surgery was associated with mortality reduction in NVE complicated by stroke, the effect of timing of surgery on outcome could not be evaluated. The end point of in-hospital mortality does not reflect long-term outcome, yet this early benefit may extrapolate to a significant survival benefit in longer follow-up, on the basis of the results of previous studies.3
Randomized controlled trials of surgery in NVE are lacking but may reduce differences in patient characteristics and treatment biases between groups. Two randomized trials evaluating the use of surgery in patients with NVE are reportedly under way.22,23 It may be challenging, however, to define an intermediate-risk group for whom surgery is not required for complications of NVE yet the benefit of surgery is uncertain, and studies such as the present investigation may help to define these criteria. Furthermore, the results of the observational studies are important to evaluate the effectiveness of early surgery for NVE in clinical practice.
In conclusion, early surgery for NVE is associated with a significantly lower in-hospital mortality rate than medical therapy. The mortality benefit associated with surgery was observed in patients with a high propensity for surgery and specifically those with paravalvular complications, systemic embolization, stroke, or S aureus infection. Careful assessment for these complications and prompt surgical intervention may improve the outcome of this serious disease. In addition, given the high and increasing prevalence of S aureus NVE in the contemporary era, additional studies are needed to evaluate the use and outcome of surgery in these patients.
Sources of Funding
Grants were received from the Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III, Madrid, Spain, the Spanish Network for the Research in Infectious Diseases (REIPI RD06/0008), and Fundación Privada Máximo Soriano Jiménez, Barcelona, Spain (grant supporting the Hospital Clínic Endocarditis database; Dr Moreno); and the National Institutes of Health (Dr Fowler; grant no. R01-AI068804). This study was supported by a grant from the American Heart Association (3830047) to Dr Wang. The funding source had no role in the design or conduct of the study. The ICE project has also received funding from investigator donations (Bruno Barsic, Dr Corey, Dr Fowler, David Gordon, Dr Wang) and educational grants (Cubist Pharmaceuticals, International Society of Cardiovascular Infectious Diseases).
Dr Fowler is a member of the Cubist Pharmaceuticals advisory committee; has served as a consultant for Astellas, Biosynexus, Cubist, Inhibitex, Merck, Johnson & Johnson, and Leo Pharmaceuticals; received research support from Cerexa, Cubist, Inhibitex, Theravance, and Merck; participated as a speaker for Cubist; and received honoraria from Arpida, Astellas, Biosynexus, Cubist, Inhibitex, Merck, Nabi, Pfizer, Theravance, and Ortho-McNeil. Dr Corey is a consultant for Theravance, Cubist, Implicit, Inimex, Arpida, Pfizer, Targanta, AstraZeneca, Merck, and Cerexa. He is on the advisory board for Pfizer, Cubist, Inhibitex, Merck, Vicuron, and Johnson & Johnson and has received grants or research support from Theravance, Innocoll, Cerexa, and Cempra. The remaining authors report no conflicts.
Aksoy O, Sexton DJ, Wang A, Pappas PA, Kourany W, Chu V, Fowler VG, 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.
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.
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. 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.
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, 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.
Stukel TA, Fisher ES, Wennberg DE, Alter DA, Gottlieb DJ, Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA. 2007; 297: 278–285.
Sy RW, Bannon PG, Bayfield MS, Brown C, Kritharides L. Survivor treatment selection bias and outcomes research: a case study of surgery in infective endocarditis. Circ Cardiovasc Qual Outcomes. 2009; 2: 469–474.
Li JS, Sexton DJ, Mick N, Nettles R, Fowler VG Jr, Ryan T, Bashore T, Corey GR. Proposed modifications to the Duke criteria for the diagnosis of infective endocarditis. Clin Infect Dis. 2000; 30: 633–638.
Fowler VG Jr, Miro JM, Hoen B, Cabell CH, Abrutyn E, Rubinstein E, Corey GR, Spelman D, Bradley SF, Barsic B, Pappas PA, Anstrom KJ, Wray D, Fortes CQ, Anguera I, Athan E, Jones P, van der Meer JT, Elliott TS, Levine DP, Bayer AS. Staphylococcus aureus endocarditis: a consequence of medical progress. JAMA. 2005; 293: 3012–3021.
Chu VH, Woods CW, Miro JM, Hoen B, Cabell CH, Pappas PA, Federspiel J, Athan E, Stryjewski ME, Nacinovich F, Marco F, Levine DP, Elliott TS, Fortes CQ, Tornos P, Gordon DL, Utili R, Delahaye F, Corey GR, Fowler VG. Emergence of coagulase-negative staphylococci as a cause of native valve endocarditis. Clin Infect Dis. 2008; 46: 232–242.
Austin PC. Primer on statistical interpretation or methods report card on propensity-score matching in the cardiology literature from 2004 to 2006: a systematic review. Circ Cardiovasc Qual Outcomes. 2008; 1: 62–67.
Stata Statistical Software, Release 10 [computer program]. College Station, Tex: StataCorp LP; 2007.
Stata Base Reference Manual. College Station, Tex: Stata Press; 2009; 1: 186–192.
Greene WH. Econometric Analysis. 5th ed. New York, NY: Prentice Hall; 2002: 715–719.
Bannay A, Hoen B, Duval X, Obadia JF, Selton-Suty C, Le Moing V, Tattevin P, Iung B, Delahaye F, Alla F; for the AEPEI Study Group. 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? Eur Heart J. Published before print February 9, 2009. doi:10.1093./eurheartj/ehp008.
Suying L, Jiannong L, Gilbertson D, McBean M, Dowd B, Collins A. An instrumental variable analysis of the impact of practice guidelines on improving quality of care and diabetes-related outcomes in the elderly Medicare population. Am J Med Qual. 2008; 23: 222–230.
San Roman JA, Lopez J, Revilla A, Vilacosta I, Tornos P, Almirante B, Mota P, Villacorta E, Sevilla T, Gomez I, Del Carmen Manzano M, Fulquet E, Rodriguez E, Igual A. Rationale, design, and methods for the Early Surgery in Infective Endocarditis Study (ENDOVAL 1): a multicenter, prospective, randomized trial comparing the state-of-the-art therapeutic strategy versus early surgery strategy in infective endocarditis. Am Heart J. 2008; 156: 431–436.
Kang DH, Park DW. Early Surgery Versus Conventional Treatment in Infective Endocarditis (EASE), 2006. http://www.clinicaltrials.gov/ct2/show/NCT00750373?term=endocarditis&rank=4. Accessed January 2, 2009.
A challenging clinical dilemma in the treatment of patients with native valve endocarditis (NVE) is selection for valve surgery during the acute period of hospitalization. Although consensus guidelines provide recommendations for surgery in endocarditis, such recommendations are based on observational retrospective studies. Some previous studies have shown a mortality benefit associated with early surgery, whereas others have found that surgery offers no survival benefit or even potential harm. This discrepancy is due in part to methodological limitations, including lack of adjustment for survivor bias (ie, patients who live longer are more likely to undergo surgery) or hidden bias (factors associated with mortality and probability of receiving surgical treatment that are not measured in the data set). The present study used the largest prospective cohort of patients with NVE and evaluated the effect of early surgery (during initial hospitalization) versus medical therapy alone on in-hospital mortality. In a cohort of 1552 patients with NVE, by use of the methods to reduce biases mentioned previously, valve surgery was associated with a significant in-hospital survival benefit compared to medical therapy alone. In subgroup analysis, surgery was found to confer a survival benefit among patients with a high propensity (or indications) for surgery and those with paravalvular complications, systemic embolization, and stroke. The survival benefit of surgery was evident in patients with NVE caused by Staphylococcus aureus, despite the lower use of surgery among patients with S aureus endocarditis. These findings support current guidelines recommending early surgery for NVE in patients with paravalvular complications or systemic embolization and suggest a beneficial effect of surgery in controversial settings such as stroke or S aureus NVE.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
The online-only Data Supplement is available with this article at http://circ.ahajournals.org/cgi/content/full/CIRCULATIONAHA.109.864488/DC1.
All of the members of the International Collaboration on Endocarditis–Prospective Cohort Study (ICE-PCS) are listed in the Appendix in the online-only Data Supplement.