Dental and Cardiac Risk Factors for Infective Endocarditis: A Population-Based, Case-Control Study

Infective endocarditis is uncommon but potentially fatal. Administration of antibiotic prophylaxis is conventional [1], but data supporting its effectiveness derive solely from anecdotal reports, studies of bacteremia after dental and other procedures, and animal models. The low incidence of disease [2] has made randomized human trials of antibiotic effectiveness impractical. Even if effective, antibiotic prophylaxis should be reserved for patients at increased risk, such as those with cardiac abnormalities who are undergoing dental procedures. However, controlled human studies of risk factors are lacking. Previous case series indicate that approximately 15% of patients with infective endocarditis caused by mouth organisms had undergone a recent dental procedure [3], but the comparable percentage from a general population is unknown. The single hospital-based casecontrol study did not find an elevated risk associated with dental therapy, except for a borderline increase with dental scaling [4]. We are unaware of controlled human studies that quantify the risk for infective endocarditis associated with cardiac valve abnormalities other than mitral valve prolapse. We therefore conducted a population-based casecontrol study to evaluate and quantify risk factors for infective endocarditis, especially those considered by the American Heart Association (AHA) to be indications for antibiotic prophylaxis [1]. Methods Participants From August 1988 to November 1990, we maintained surveillance for infective endocarditis in 54 hospitals of the Delaware Valley Case-Control Network, a population-based network of hospitals in the eight counties that constitute the Philadelphia Metropolitan Statistical Area and the county of New Castle, Delaware. Patients with a putative diagnosis of infective endocarditis were identified by hospital personnel and were reported to study nurses, who also actively sought cases. To assess the completeness of ascertainment, five high-yield hospitals and three low-yield hospitals were asked to list all patients discharged with a diagnosis of endocarditis over 3 months. These lists were compared with those obtained from our surveillance; charts were reviewed when differences were identified. We obtained informed consent from physicians and case-patients, then used structured forms to abstract medical records, including echocardiographic reports and hospital laboratory information on the infecting organism. We deleted information on purported risk factors for infective endocarditis and submitted these records for review by three of the authors, who are consultants in infectious diseases recognized for their expertise in infective endocarditis [5, 6]. These experts used their own global clinical judgment to classify potential cases as definite, probable, or possible cases or probable noncases. Agreement of two of the three reviewers was required to make the determination of a case or a noncase [6]. One control from the community was selected for each case-patient by using a modification of the Waksberg random-digit dialing method [7]. Controls and case-patients were matched for age (in 5-year strata), sex, and neighborhood of residence (by using area code, telephone exchange, and the first subsequent digit of the case-patient's telephone number). We excluded from these analyses patients with infective endocarditis who were younger than 18 years of age, intravenous drug users, and patients who developed endocarditis in the hospital. This study received separate institutional review board approval at the University of Pennsylvania and all 54 participating hospitals. Data Collection Information was obtained from case-patients by conducting a structured telephone interview after hospital discharge. The date of hospital admission served as the study date for case-patients; for controls, the date of the telephone interview was used. Telephone interviewers collected information on demographic characteristics; diagnostic and therapeutic medical and dental procedures in the year before the study date; potential host risk factors, including preexisting cardiac lesions, preexisting local infection, risk factors for oral or dental disease, diabetes mellitus, immune deficiencies, family history of endocarditis, alcoholism, malignant conditions, and autoimmune disease; previous antibiotic use; and other recent illnesses. For each host risk factor, we requested the date of diagnosis, diagnostic method (for example, echocardiography for mitral valve prolapse), and the name of the practitioner who made the diagnosis. For each medical and dental procedure, we sought information about the procedure, the month and year in which it was performed, and the practitioner. We requested medical and dental records describing procedures and validating individual diagnoses. Study Variables Case-patients were considered infected with dental flora if the organism found was viridans streptococci; nutritionally variant streptococci; Actinobacillus species; Cardiobacterium hominis; anaerobes; -hemolytic streptococci (not group D); unspecified streptococci; or Haemophilus, Eikenella, Kingella, or Neisseria species. Because this study focused on indications for antibiotic prophylaxis, we examined host characteristics reported by patients as the primary risk factor variables, reflecting the information that would be available to a practitioner about to perform a procedure for which prophylaxis might be indicated. A variable called any valvular heart abnormality was defined as the presence of any of the following self-reported, preexisting conditions: mitral valve prolapse, congenital heart disease, history of rheumatic fever with heart involvement, prosthetic heart valve, previous episode of endocarditis, or other valvular heart disease. Dental visit information was obtained solely from dental records. Dental hygiene care was defined as preventive oral health services and therapeutic services, including coronal scaling and polishing. Consistent with AHA guidelines, invasive dental procedures were defined to include dental hygiene care, extractions, periodontal treatment (including scaling and root planing), endodontic treatment, mouth or gingival surgery, and treatment of tooth abscess. Noninvasive dental procedures were simple restorations, prosthetic and restorative dentistry, fluoride treatment, and other procedures (prosthetic services, including adjustments and suture removal). Unless otherwise specified, dental treatment refers to all dental treatment and is not limited to invasive procedures. Initial analyses focused on dental procedures performed at any time in the 3 months before the study date. Analyses were then narrowed to 2 months and 1 month before the study date. Time frames are approximate because the onset of infective endocarditis is often difficult to determine with certainty. We therefore chose the date of hospital admission as the study date, collected procedural data based on month rather than on a specific date, and calculated time frames under the assumption that procedures were performed on the 15th of the month. Statistical Analysis Frequencies and cross-tabulations between casecontrol status and potential risk factors were obtained. Conditional logistic regression was used to determine the independent effects of the various potential risk factors and the possibility of any interactions among factors [8]. Variables were included in multiple regression models if they were significant (P < 0.2) in unadjusted (matched) analyses (such as kidney disease and diabetes), if their inclusion had a substantial effect (>15% change) on coefficients for variables already in the model (such as insurance status) [9], or if they were strongly suspected a priori of being confounders (such as ethnicity). For analyses specific to participants with known cardiac valvular abnormalities, odds ratios and CIs were calculated from a model that included main effects for cardiac valvular abnormalities and dental treatment and the interaction between those variables. The odds ratio for various dental therapy variables among participants with cardiac valvular abnormalities was estimated by combining coefficients for the dental therapy variable and the interaction term. The CI for this combination of coefficients was estimated by using the appropriate variance and covariance terms [8]. With the interaction terms, participants with and those without valvular abnormalities were included in these analyses. Exact odds ratios and CIs, stratified on the matching variables, were calculated when data were too sparse for conditional logistic regression [10]. We used SAS statistical software (SAS Institute Inc, Cary, North Carolina) for data management and to obtain frequencies and cross-tabulations. We used EGRET (Epidemiological Graphics, Estimation and Testing software, version 0.25.1, Cytel Software Corp., Cambridge, Massachusetts) for conditional logistic regressions and exact analyses. All CIs are 95%, and all P values are two-sided. The sample size for the study was chosen so that by assuming an level of 0.05 (two-sided) and a power of 80%, we would be able to detect associations with an odds ratio of 2.0 or more for risk factors with a prevalence between 0.1 and 0.8. Results Participants We identified and recruited 416 potential case-patients (Figure 1). Our assessment process confirmed that more than 90% of true cases of infective endocarditis had been identified. The expert panel judged 379 patients to have definite, probable, or possible infective endocarditis; 37 (9%) were judged to be probable noncases and were excluded. Agreement among judgments was high, ranging from 92% to 96% [6]. Figure 1. Enrollment of case-patients. Of these 379 patients, 287 had community-acquired infective endocarditis not associated with intravenous drug use (248 on native valves and 39 on prosthetic valves), 27 had nosocomial infective endocarditis (18 on nativ