Clinical Features of Staphylococcus aureus Endocarditis:  A 10-Year Experience in Denmark FREE

INFECTIVE ENDOCARDITIS is a constantly evolving disease. During the last decades, several important aspects of the disease have changed. The patients are older, and the spectrum of predisposing heart conditions has shifted; ie, fewer patients have underlying rheumatic valve disease, but more patients now have prosthetic valves.^1- 6 It also appears that the incidence of Staphylococcus aureus endocarditis is increasing.^3,5- 7 Both morbidity and mortality from S aureus endocarditis continue to be high, which to a certain extent reflects the difficulty in diagnosing the disease clinically. Espersen and Frimodt-Møller⁸ previously reported the clinical characteristics of patients with S aureus endocarditis over a 6-year period (1976-1981), and since then, only a few studies involving large series of patients with this disease have been published.

The purpose of the present study was to analyze the epidemiological characteristics, clinical features, antibiotic treatment, complications, and outcome of S aureus endocarditis occurring in a nation-wide population of non–drug addicts during a 10-year period (1982-1991).

Almost all S aureus strains isolated from blood cultures in Denmark are referred to the Staphylococcus Laboratory, Statens Serum Institut, Copenhagen, for phage typing. The hospital departments involved are asked to provide further clinical information. For our study, the clinical data were reviewed by a panel of 2 clinical microbiologists (N.F.-M. and F.E.) and 1 infectious disease specialist (P.S.), and the medical records, including autopsy reports, were requested in all cases in which the diagnosis of endocarditis was reported or suspected either by the clinicians or by the panel.

During the period from 1982 to 1991, a total of 8514 cases of bacteremia involving S aureus were reported, and in 485 cases, the presence of endocarditis was possible. One hundred three episodes of S aureus endocarditis in drug addicts were excluded because the clinical aspects of endocarditis in this subpopulation were reported in detail in 1995.⁹ Sixty-seven other cases were excluded either because no medical records were obtained (41 cases) or because the information was insufficient (26 cases). Thus, 315 cases were categorized as endocarditis according to the criteria outlined below. Fifty-five cases failed to fulfill the criteria; therefore, the present study comprises 260 cases of endocarditis.

The new criteria for the diagnosis of infective endocarditis proposed by Durack et al¹⁰ were applied. Thus, endocarditis was defined as definite by pathologic criteria if valvular vegetation or intracardiac abscess was present at autopsy or surgery and if active endocarditis or staphylococci were demonstrated by histologic examination or culture. It was defined as definite by clinical criteria if 2 major criteria, 1 major and 3 minor criteria, or 5 minor criteria were fulfilled. The definitions of major and minor criteria are listed in Table 1. Endocarditis was defined as possible if there were findings consistent with endocarditis that fell short of definite by clinical criteria. This category includes a wide range of indeterminate cases with very different probabilities of endocarditis. Therefore, and also because the study was conducted retrospectively, we decided that 2 more criteria had to be fulfilled for inclusion in this category: the diagnosis of endocarditis had to be suspected by the clinicians, and the patient had to be treated accordingly. Cases that did not meet these criteria were rejected, as were those in which the symptoms of endocarditis resolved with antibiotic therapy for 4 days or less and those in which no pathologic evidence of endocarditis was present at surgery or autopsy after therapy for 4 days or less.

The medical records, including the autopsy report, for each patient were reviewed, and the following data were listed: sex; age; date of admission; underlying diseases, including predisposing heart conditions, such as prosthetic cardiac valves; previous bacterial endocarditis; congenital cardiac malformations; rheumatic and other acquired valvular dysfunction¹¹; and all symptoms and physical findings. For community-acquired cases, symptoms were those experienced from onset until admission, and physical findings were those present on admission. For nosocomial cases, symptoms as well as physical findings were recorded as those present at the time of the first positive blood culture. Also, all cases were examined for localized foci of staphylococcal infection. Lesions from which S aureus organisms were cultured and present on initial evaluation were considered to be primary. Finally, we recorded the type and number of complications, valvular involvement, results of echocardiographic investigations, details of antibiotic treatment, cardiac surgery, and the outcome of the infection.

The following data were listed regarding antibiotic treatment: date of initiation of appropiate antistaphylococcal therapy, drug regimen, duration of treatment, and dosage. We considered appropriate treatment to be the parenteral administration of benzyl penicillin sodium (if the infecting strain was susceptible to penicillin), penicillinase-stable penicillins, cephalothin sodium and cefuroxime, and vancomycin. Thus, the administration of fusidic acid, rifampin, and erythromycin alone or in combination was considered inappropiate treatment.

The Fisher exact test was used to compare proportions, and the Mann-Whitney U test was used for comparison of nonpaired observations. A P value of less than .05 was considered significant.

Of the 260 patients, 145 (56%) were male and 115 (44%) were female. The age ranged from 2 to 97 years (median age, 67.5 years). The female patients were significantly older than the male patients (median age, 73 vs 64 years; P<.001). The number of endocarditis cases per year was between 19 and 30 (median number of cases, 27), or between 2.3% and 3.7% of the total number of S aureus bacteremia cases reported. The distribution of endocarditis cases according to the diagnostic categorization is shown in Table 2. One hundred fifty-two cases (58%) of endocarditis were pathologically confirmed; 70 (27%) were classified as definite by clinical criteria; and 38 (15%) were classified as possible. In more than half (78 [51%]) of the cases in which the diagnosis was confirmed pathologically, no clinical diagnosis of infective endocarditis was made while the patient was alive. In retrospect, 15 (19%) of these 78 cases met the definition of clinical definite endocarditis ante mortem; echocardiographic data were not available in any of these cases. Among the whole group of pathologically confirmed cases, a total of 66 cases (43%) fulfilled the Duke criteria of clinically definite endocarditis. Overall, the diagnosis of endocarditis was not suspected clinically in 83 cases (32%). One hundred seventy-two cases (67%) were community acquired, and in these cases, the median duration of illness prior to admission was 3 days (range, 0-50 days). In nosocomial cases, the median duration from admission to onset of disease was 12 days (range, 0-95 days). Overall, 56 patients (22%) had a predisposing heart condition that qualified as a minor criterion in the diagnostic categorization. The most frequent predisposing heart conditions were prosthetic cardiac valves (24 patients), congenital heart disorder (9 patients), aortic stenosis (9 patients), and previous bacterial endocarditis (7 patients). Only 4 patients had mitral valve prolapse, and 3 patients had known rheumatic valve disease. A number of patients had predisposing factors other than heart diseases. Thirty-five patients (13%) had diabetes mellitus; 28 patients (11%) were treated with steroids; 12 patients (5%) had cirrhosis; 11 patients (4%) had a malignant disease; and 10 patients (4%) had a chronic renal disease. In 103 (60%) of the patients with community-acquired infection, but in only 25 (29%) of those with nosocomial infection, no portal of entry was identified (P<.001). The most common primary focus among patients with community-acquired infection was the skin (23 patients), followed by pneumonia (13 patients) and urinary tract infection (11 patients). Infected intravascular devices (23 patients) were the most frequent portals of entry in nosocomial cases.

The presentation was often quite nonspecific and no symptom was particularly frequent; ie, all complaints were shared by fewer than half of the patients. Gastrointestinal symptoms were relatively common, and almost one tenth of the patients complained of abdominal pain. Many of these patients were primarily admitted to surgical departments. Ninety-five percent of the patients presented with a temperature higher than 38°C, while a new heart murmur could be detected in only 81 patients (31%) at presentation. Sixty-one patients (23%) presented initially with signs of involvement of the central nervous system. These signs included major tromboembolic events, meningitis, and toxic confusion.¹² Only fifty-six patients (22%) had signs of congestive heart failure at the onset of disease. Vascular phenomena suggestive of infective endocarditis, such as Janeway lesions (16 patients), conjunctival hemorrhage (9 patients), and ecchymoses (2 patients), were relatively uncommon, while petechiae or splinter hemorrhages, which we considered to be nonspecific manifestations, were observed in 24 patients.

The sites of endocarditis are shown in Table 3. Native left-sided endocarditis composed 78% of all cases, and mitral valve involvement was more common than aortic valve involvement. Native right-sided endocarditis included 11 cases of tricuspid valve endocarditis and 2 cases of pulmonary valve endocarditis. Twenty-four patients had prosthetic valve endocarditis. Six patients with congenital heart disorder and 16 others had no heart murmur and either had normal echocardiographic findings or did not undergo echocardiography, which explains why the valvular involvement in these patients was classified as unknown. Two-dimensional, transthoracic echocardiograms were obtained in 128 (49%) of the 260 cases of infective endocarditis. In 63 (49%) of these 128 cases, vegetations were seen on the echocardiograms. Three more patients had echocardiographic evidence of endocardial involvement that qualified as a major criterion: 1 had a partial dehiscence of a prosthetic valve and 2 had an intracardiac abscess. When comparing the 63 patients who had vegetations detected on echocardiograms with the 114 patients who also had an antemortem clinical diagnosis of endocarditis but no vegetations detected, we found no differences in outcome parameters, such as congestive heart failure (56% vs 53%; P=.83), central nervous system involvement (35% vs 32%; P=.77), and cardiac surgery (21% vs 12%; P=.21). However, the vegetation-positive group had a significantly lower mortality rate than those without vegetations (33% vs 53%; P=.02). The difference in mortality rates remained significant when the patients who received only medical therapy were considered (38% vs 57%; P=.04).

Table 4 shows some complications of S aureus endocarditis. A new heart murmur was eventually detected in precisely 50% of the patients. Congestive heart failure, which initially was present in only one fifth of the patients, developed during the disease in a further 83 patients. It is of interest to consider a new cardiac murmur and congestive heart failure simultaneously, 2 features that are often regarded as hallmarks of staphylococcal endocarditis. Only 8% (n=20) of the patients had both a new cardiac murmur and signs of congestive heart failure at the time of presentation, and this percentage rose to only 29% (n=75) when the whole course of hospitalization was considered. On the other hand, 25% (n=66) of the patients did not have a murmur or any signs of heart failure during their illness. The 91 patients with complications involving the central nervous system have previously been described in detail.¹² Altogether, 57 patients experienced cerebral embolism: 29 at the onset and 28 during the course of the disease. Only 4 patients had noncerebral emboli: 1 patient had a renal embolism; 1 patient had a mesenteric artery occlusion, which prompted an acute surgical embolectomy; and 2 patients each had an embolism in a lower limb, one of which eventually resulted in amputation. Vascular phenomena were relatively uncommon both at onset and during the disease. Roth spots were recorded in only 1 patient.

The overall mortality rate among the patients who were diagnosed clinically was 46% (81/177). Many of the patients in the present study were primarily admitted at smaller provincial hospitals, and we questioned whether this could be a factor attributing to the high mortality rate. We found no difference, however, in mortality rates between patients admitted to university hospitals in the 3 largest cities in Denmark and patients admitted to all other hospitals (51% [37/72] vs 42% [44/105]; P=.28). The time course of mortality was analyzed with regard to duration of therapy for the group with an antemortem clinical diagnosis of endocarditis (177 patients). A total of 3 patients in this group did not receive relevant antistaphylococcal treatment. Two patients with nosocomial infection died 1 and 2 days after onset, respectively, and the third patient, with community-acquired infection, died 2 days after admission. Of the remaining 174 patients, 16 (9%) died within 7 days of initiation of therapy. The mortality rate had increased to 20% (35/174) at 14 days after initiation of antibiotic treatment, with further increases to 30% (52/174) and 34% (59/174) after 21 and 30 days, respectively. Nineteen patients died more than 30 days after initiation of treatment.

Of the 174 patients with an antemortem clinical diagnosis of endocarditis who received appropiate antistaphylococcal treatment, 147 were treated with medical therapy only. We compared the survivors with the nonsurvivors in this group to identify risk factors for mortality that might predict the need for early valve replacement. The results are shown in Table 5. Congestive heart failure developing during the disease was significantly associated with mortality, as opposed to heart failure at disease onset. Involvement of the central nervous system and age were the 2 other factors associated with a higher mortality rate. Sixty percent (58/96) of the patients aged 60 years or older died, as opposed to 29% (15/51) of the patients younger than 60 years. The mortality rate and frequency of S aureus endocarditis in each age group for all 260 patients are shown in detail in Figure 1.

The infecting strain of S aureus was not resistant to methicillin sodium in any of the 260 patients. Table 6 gives a summary of the appropriate antibiotic treatment regimens in the group of patients who received medical therapy only and who had an antemortem clinical diagnosis of endocarditis. The treatment regimens showed great variability, and many patients were treated with several antibiotics simultaneously or in succession. Only 18 patients were treated with vancomycin, and 17 of these were also treated with other drugs for varying periods: 16 with a β-lactam antibiotic and 8 with an aminoglycoside. Only 9 patients were treated with vancomycin for 3 weeks or longer. Thus, treatment with vancomycin played only a minor role, and the vast majority of patients were treated with a β-lactam antibiotic, often in combination with another drug. We found no difference in mortality rates between patients treated with β-lactam antibiotics and those treated with a combination of β-lactam antibiotics and aminoglycosides (54% [27/50] vs 54% [32/59]). In addition to aminoglycosides, fusidic acid was administered to 10 patients in the combination group and erythromycin was administered to 1 patient in the combination group. The duration of treatment was undoubtedly another factor of importance for the outcome of infection, but evaluation of this parameter was hampered by the fact that many patients died early during appropriate treatment. Thus, 51 (70%) of 73 patients died within the first 3 weeks of treatment. When the patients who survived and the duration of parenteral treatment were considered, the minimum duration was 17 days and only 5 patients were treated for less than 3 weeks. Twenty-three patients were treated from 3 to 4 weeks, and 23 patients were treated from 29 days to 6 weeks. Many patients continued with oral treatment for extended periods after parenteral treatment was discontinued. Finally, we tried to examine the significance of dosages of penicillinase-stable penicillins. Among 85 adult patients who received a known dosage of penicillinase-stable penicillins, we found no difference in mortality rates between those treated with 6 g/d or more and those receiving less than 6 g/d (62% [28/45] vs 55% [22/40]; P=.65).

In 83 patients, the diagnosis of endocarditis was not suspected clinically. Table 7 shows some characteristics of diagnosed vs undiagnosed cases. The undiagnosed cases occurred in older patients, and it seemed that the probability of the disease being diagnosed clinically decreased as age increased. Thus, no clinical diagnosis was made in 39% of the patients (70 of 178) aged 60 years or older, in 44% of the patients (53 of 120) aged 70 years or older, and in 50% of the patients aged 80 years or older. Not surprisingly, a new heart murmur was present less frequently in the undiagnosed cases, but a cardiac murmur was present with the same frequency in old and young persons, ie, 87 (51%) of 171 patients older than 60 years compared with 43 (48%) of 89 patients aged 60 years or younger. Vascular phenomena suggestive of infective endocarditis were detected only rarely: Janeway lesions in 6 patients, ecchymoses in 2 patients, and conjunctival hemorrhages in 0 patients. A predisposing heart condition was present less frequently in the patients with no clinical diagnosis (8% [7/83] vs 28% [49/177]; P<.001). Also, one characteristic separated the undiagnosed from the diagnosed cases: a primary focus of infection was identified more frequently. Thus, in 29 (85%) of 34 undiagnosed nosocomial cases, a primary focus of infection—most frequently intravascular devices, lungs, and infections in the skin—was identified, compared with only 33 (62%) of 53 diagnosed nosocomial cases (P=.03). In community-acquired cases, a primary focus of infection was identified in 53% (26/49) of the undiagnosed cases, compared with 35% (44/124) of the diagnosed cases (P=.05).

Twenty-one (25%) of the 83 patients did not receive appropiate antistaphylococcal treatment. Many of these patients were initially treated with a combination of ampicillin and an aminoglycoside and died shortly after admission. Thus, 25 (51%) of 49 patients with community-acquired infection died within 1 week of admission and 35 (71%) within 2 weeks. Thirteen (38%) of 34 patients with nosocomial infection died within the first week after the onset of the disease and 22 (65%) within 2 weeks. Undiagnosed cases tended to be more common among female patients than among male patients (38% [44/115] vs 27% [39/145]; P=.07).

When comparing the results in studies dealing with infective endocarditis, a number of factors must be taken into consideration, the most important being the definition of endocarditis and the study population. In the past, one approach has been to use strict case definitions; in particular, those formulated by von Reyn et al¹³ have been applied often in the last 15 years. In a previous Danish study on S aureus endocarditis, Espersen and Frimodt-Møller⁸ used these case definitions, but since then, echocardiography has been introduced as a routine noninvasive cardiac examination. Therefore, we thought it more appropriate to use the new diagnostic criteria proposed by Durack et al,¹⁰ although the echocardiographic criteria formulated by Durack and colleagues are actually specific for transesophageal echocardiography and the patients in our study only underwent transthoracic echocardiography. We are well aware that the application of these criteria by no means is without problems and especially that the inclusion of the vaguely defined "possible" category inevitably increases the sensitivity at the expense of specificity.¹⁴ However, the justification for including this category lies in the fact that the inclusion was restricted to those patients in whom the clinicians suspected the diagnosis of endocarditis. Although the reasons for this decision in a few instances may have been questionable, it always had therapeutic implications. As regards the study population, it is well recognized that the clinical spectrum of endocarditis in patients from referral centers differs from that seen in the general population.¹⁵ Our study includes 260 cases of S aureus endocarditis in non–drug addicts, representing a total of 63 hospitals from all over the country, thereby avoiding the specific referral bias that may hamper studies from a single institution. The present study is also unique in that it was confined to cases of infective endocarditis caused by S aureus and to cases in which the microorganism was found in blood cultures.

The number of endocarditis cases was between 2.3% and 3.7% of the total number of S aureus bacteremia cases reported. Thus, the relative frequency of endocarditis has remained unchanged since the last survey, which was carried out between 1976 and 1981,¹⁶ but the absolute number has increased from a mean of 20 cases to a mean of 26 cases annually. While the incidence of S aureus endocarditis seems to be a constant fraction of the incidence of S aureus bacteremia, the absolute number of S aureus bacteremia cases has increased markedly over the years, an increase that is closely related to the increasing admission rate to Danish hospitals.¹⁷

The clinical picture of S aureus endocarditis remains relatively uncharacteristic, and a number of nonspecific symptoms of infection (eg, chills, nausea, vomiting, myalgia, and arthralgia) were each reported only infrequently. It has obvious implications to distinguish clinically uncomplicated S aureus bacteremia from endocarditis. The absence of either a cardiac murmur or signs of congestive heart failure by no means excludes the presence of endocarditis; therefore, a major question that needs to be answered is, "What clinical factors should arouse suspicion that endocarditis is present in a patient with S aureus bacteremia?" This question was addressed more than 20 years ago by Nolan and Beaty.¹⁸ In a retrospective study of 105 cases of S aureus bacteremia, they found that 24 of 26 cases of endocarditis were characterized by the following triad: (1) community-acquired S aureus bacteremia that (2) arose from an inapparent primary focus and (3) was associated with metastatic sequelae. In contrast, only 2 cases of endocarditis were acquired nosocomially from an obvious primary focus with no metastatic sequelae. From the few studies conducted since then with the purpose of verifying the utility of these criteria, it appears that the Nolan and Beaty parameters have insufficient predictive value in a population of non–drug addicts with S aureus bacteremia to be used as the only tools for diagnosing endocarditis.¹⁹ The insufficiency of these criteria agrees with the results of our study. First, approximately one third of the 8500 cases of S aureus bacteremia reported during the study period were community acquired. This means that even if the frequency of endocarditis is higher when community-acquired infection is compared with nosocomial infection, the difference is only a factor of 4. Second, although found less frequently in community-acquired cases, a primary focus was still identified in 40% of these patients. Third, an obvious primary focus was identified in 71% of the nosocomially acquired cases of endocarditis, and metastatic involvement of the central nervous system was observed with the same frequency as in the community-acquired cases.¹² In this context, however, it must be mentioned that there was no follow-up beyond the hospital stay for the patients. This lack of follow-up represents a limitation to our study, as serious complications due to endocarditis, including metastatic sequelae, can occur several months after treatment has ended. Of particular interest, however, is the fact that an intravascular catheter was identified as a primary focus in 25% of the patients with nosocomial infection. Thus, even though it is true that most cases of S aureus bacteremia acquired in the hospital and relating to indwelling intravascular devices rarely represent or eventuate in endocarditis,^19- 20 these cases nevertheless represent one tenth of the patients in the present study. A more frequent identification of a primary focus did indeed characterize the patients who were not diagnosed clinically. Thus, it is extremely important to emphasize that a primary focus of the infection by no means excludes the coexistence of endocarditis. In conclusion, in many cases, it is impossible to differentiate uncomplicated S aureus bacteremia from endocarditis clinically, and it is imperative to increase diagnostic efficiency by using echocardiography. The question is, "Which patients should undergo screening by transthoracic echocardiography?" Based on our results, it is difficult to make firm recommendations, but it does not seem justified to restrict such screening to patients with community-acquired S aureus bacteremia.

ONE OF THE KEY ISSUES in S aureus endocarditis is the treatment strategy: what is the optimum antibiotic treatment regimen and which patients should be selected for early valve replacement? Concerning the antimicrobial management of S aureus endocarditis, some recent articles have suggested that vancomycin may result in suboptimal clinical outcomes.^21- 22 Contrary to the present study, which deals exclusively with a population of non–drug addicts predominantly with left-sided endocarditis involving native valves, these previous studies have mainly been conducted among intravenous drug users with right-sided endocarditis^21- 22 or among patients infected with methicillin-resistant strains of S aureus, none of which were encountered in the present study.²³ We found no indication of decreased efficiency of vancomycin-based treatment regimens, but it must be emphasized that our study carries little weight in this matter, as only a few patients with a proper diagnosis were treated with vancomycin, and the majority of these patients were also treated with other drugs. Another point of particular interest concerning the antibiotic treatment of S aureus endocarditis is the potential benefit of combining β-lactam antibiotics and aminoglycosides. In a prospective multicenter study of S aureus endocarditis, the addition of aminoglycosides for the first 2 weeks of a 6-week regimen of nafcillin therapy failed to alter morbidity and mortality, but the clearance of bacteremia in non–drug addicts occurred significantly faster in the combination group.²⁴ In the hope that a more rapid eradication of bacteremia will result in a reduced incidence of metastatic infection and reduced damage to the heart valve, it has been advocated that patients with left-sided S aureus endocarditis be treated with combined β-lactam–aminoglycoside regimens for the first 3 to 5 days²⁴ or the first treatment week.¹⁹ An ad hoc writing group appointed by the American Heart Association to provide guidelines for the treatment of endocarditis in adults made the addition of gentamicin sulfate to the regimen for the first 3 to 5 days optional.²⁵ In the present study, we were also unable to demonstrate a beneficial effect of a combination regimen: the mortality rates were exactly the same in the combination-treated group as in the group treated with β-lactam antibiotics alone. This finding is in agreement with the results of a survey that covered the period of 1976 to 1981.²⁶ With regard to duration of therapy, a 4- to 6-week course seems reasonable, as proposed by others.^19,24- 25

Concerning surgical treatment, a total of 27 patients underwent cardiac valve replacement, and only 5 patients (18%) died. Thus, the clinical outcome was substantially improved in those patients who were treated surgically, but such a direct comparison is strongly biased, considering the nonrandom way patients were selected for this treatment and the many confounding factors that cannot be controlled in a retrospective study like ours. The question then remains as to which patients should be treated surgically, as well as how the decision can be made early enough to make a difference. To address this problem, we focused particularly on that subset of patients who both received appropiate antistaphylococcal therapy and were diagnosed clinically. Three factors differed significantly between the survivors and nonsurvivors in this group of patients: (1) Late congestive heart failure occurred more frequently among the nonsurvivors, which is not surprising, as this condition is one of the leading causes of death among patients with infective endocarditis. (2) Older patients were encountered more frequently among the nonsurvivors, but this is probably a mere reflection of age being a general risk factor in serious diseases. (3) However, we found that involvement of the central nervous system occurred significantly less frequently among the survivors. A metastatic involvement occurs approximately in one third of the patients with S aureus endocarditis, and as we have reported previously,¹² these neurological events often occur at the onset of the disease or shortly thereafter, and recurrent embolism is rare. Thus, based on our data, involvement of the central nervous system may constitute a relative indication of early valve replacement, as these patients with central nervous system involvement have a poor prognosis when they are treated only with antimicrobial chemotherapy.

The overall mortality rate of the patients who were diagnosed clinically was 46% (81/177). This figure is somewhat higher than the values of 30% to 45% that have been reported by others^{4,7,13,27- 30} and almost twice as high as the 26% reported by Watanakunakorn.³¹ The high mortality rate may be partly explained by several factors. First, drug addicts, in whom S aureus endocarditis constitutes a more benign clinical syndrome, were excluded in the present study. Second, the inclusion of endocarditis cases from all over the country may have played a role. Many small provencial hospitals cared for only a few patients during the study period, indicating a lack of experience. However, we were unable to demonstrate a difference in mortality rates when comparing patients primarily admitted to university hospitals in the 3 largest cities in Denmark with the remaining group of patients. Finally, it is also conceivable that the lack of standardization of treatment exemplified by the great variability in antibiotic treatment regimens may have influenced the mortality rate.

The fact that almost one third of the cases were not suspected clinically makes the present study unique. The inclusion of these clinically undiagnosed cases is explained by the high autopsy rates, which previously were common in Denmark: in 1982, autopsies were performed on approximately 65% of the patients who died in hospital, whereas the figure in 1991 had declined to approximately 30%.

In conclusion, the clinical diagnosis of S aureus endocarditis is still extremely difficult; the mortality rate remains high; and the disease will therefore continue to be a major clinical challenge. We believe that an improvement in the prognosis partly depends on a higher degree of suspicion of the disease combined with a raised awareness of the paucity of clinical findings so often connected with staphylococcal endocarditis. Also of great importance, however, is a more frequent use of echocardiography as a screening method in patients with S aureus bacteremia, and not only among those patients lacking an obvious primary focus of infection. While the exact role of cardiac valve replacement in the treatment ofS aureus native valve endocarditis remains to be defined, the option must be available. Therefore, transfer to a hospital with a cardiovascular surgery unit should be considered early in the course of S aureus endocarditis.

Accepted for publication July 15, 1998.

This work was supported by a grant (12-1618) from the Danish Medical Research Council, Copenhagen.

Reprints: Bent L. Røder, Department of Clinical Microbiology, Hillerød Sygehus, Helsevej 2, DK-3400 Hillerød, Denmark.