Pérez-Vázquez A, Fariñas MC, García-Palomo JD, Bernal JM, Revuelta JM, González-Macías J. Evaluation of the Duke Criteria in 93 Episodes of Prosthetic Valve EndocarditisCould Sensitivity Be Improved?. Arch Intern Med. 2000;160(8):1185-1191. doi:10.1001/archinte.160.8.1185
Since publication of the Duke criteria for diagnosing endocarditis, several articles have confirmed their sensitivity when native and prosthetic valves are considered together.
To compare the differences between the older von Reyn criteria and the Duke criteria in prosthetic valve endocarditis only, and to determine if the latter's sensitivity could be improved by adding 2 minor criteria: new-onset heart failure and presence of conduction disturbances.
We retrospectively evaluated 93 episodes of prosthetic valve endocarditis from January 1986 to January 1998 in a teaching hospital, and then analyzed the 76 surgically confirmed episodes to compare the differences between the von Reyn and Duke diagnostic criteria.
The von Reyn criteria rejected the diagnosis in 16 of the confirmed episodes, compared with 1 diagnosis missed by the Duke criteria and 1 missed using our suggested modifications. Definite diagnosis (Duke) was established in 60 episodes, compared with a diagnosis of probable (von Reyn) in 36 episodes (P<.001). Our modifications improved the sensitivity of the Duke criteria, diagnosing 70 episodes as definite (P=.02).
As was the case with native valve endocarditis, the Duke criteria proved to be more sensitive than the von Reyn criteria in prosthetic valve endocarditis. The addition of 2 minor criteria (new-onset heart failure and presence of conduction disturbances) could improve the diagnostic sensitivity of the Duke criteria.
OVER THE past few years, surgical repair of cardiac valves has become the therapy of choice for an increasing number of patients, offering them a higher survival rate and better quality of life.1,2
Prosthetic material infection is, among the complications of surgical repair of cardiac valves, the one with the highest morbidity and mortality, and it often challenges physicians' clinical diagnostic and therapeutic skills.2- 5
The current approach to patients with prosthetic valves and suspected endocarditis has been determined by (1) echocardiography, mainly transesophageal (TEE); (2) the higher sensitivity of diagnostic criteria3,6; and (3) new surgical techniques. We reviewed our experience during the past 12 years, focusing on sensitivity assessment of the Duke criteria in episodes of prosthetic valve endocarditis (PVE), and tried to improve these criteria.
The present study was conducted at a 1200-bed community and teaching hospital. All patients with PVE admitted from January 1986 to January 1998 were included. Data were obtained by a retrospective review of the information registries from the Cardiovascular Department, Infectious Diseases Unit, and Admissions Department of Marqués de Valdecilla University Hospital, Santander, Spain. During the chart review phase, we considered the Duke criteria6 when establishing a diagnosis of PVE. Early infections were defined as those in which clinical symptoms or signs were recorded during the first 60 days after surgery.
Surgically confirmed PVE was defined as episodes with positive culture or surgical specimens histologically compatible with endocarditis and those with a valve appearance at surgery compatible with endocarditis (abscess, dehiscence, or vegetations).3
Medical records of patients were reviewed for data on age, sex, place of origin (our community or reference from other institutions); preoperative functional class according to the New York Heart Association classification; underlying valve diseases and previous endocarditis with native or prosthetic valves, previous valve surgery, date of latest surgical repair, and prostheses characteristics (size, position, and type); preoperative and postoperative antibiotic drug use; clinical, analytic, echocardiographic, microbiologic, and therapeutic data; pathological examination results; and clinical outcome.
Specimens taken during surgery were transported to the bacteriology laboratory in a closed syringe. Valvular and perivalvular specimens were incubated under aerobic conditions in thioglycolate broth and on MacConkey, chocolate, and blood agar media. For cultures of anaerobes, the material was plated onto prereduced vitamin K1–enriched Brucella blood agar, onto an anaerobic blood agar containing colistin sulfate and nalidixic acid, and was inoculated into enriched thioglycolate broth (containing hemin and vitamin K1). For each blood culture, 10 mL of blood was inoculated directly into bottles of Bactec NR 730 blood culture system (Becton-Dickinson, Cockeysville, Md). Isolated microorganisms were identified by standard microbiologic techniques, and antimicrobial sensitivity assays were performed with Microscan system panels (Dade International Inc, West Sacramento, Calif).
Two-dimensional echocardiography was performed with an ultrasound transducer operating at 2.5 MHz (Sonos 2500 or 5500; Hewlett Packard, Palo Alto, Calif).
Data were analyzed using statistical software (SPSS for Windows; SPSS Inc, Chicago, Ill). Qualitative data were studied by means of a χ2 test or Fisher exact test (when expected frequencies were <5), and t tests were used for quantitative variables.
Quantitative data are expressed as mean ± SD. Two-sided tests of significance were performed with an α of .05.
During the study period, 93 episodes of PVE were treated at Marqués de Valdecilla University Hospital (76 patients had 1 episode each, 7 patients had 2 episodes each, and only 1 patient had 3 episodes). Patients were aged 58 ± 14 years (range, 26-51 years). The male-female ratio was 7:3.
Twenty-two patients (26%) had a history of endocarditis before the latest surgical repair of their valves (12 with native valves and 10 with prostheses). Twenty-seven patients had previously undergone valve replacement surgery at least once.
Rheumatic valvulopathy was the most frequent underlying disease (48%), followed by a prosthetic dysfunction (25%), aortic degenerative disease (11%), congenital valvulopathies (12%), and mixomatous degeneration (3.3%).
Fifty-three episodes of PVE involved porcine bioprostheses (23 Hancock, 18 Carpentier-Edwards, 7 Ionnescu, 2 Xennotech, 2 MMI, and 1 Wessex) and 40 involved mechanical valves (18 Björk, 17 Carbomedics, 4 St Jude, and 1 Starr-Edwards). The mechanical prostheses were implanted mainly after 1986, when the relative frequency of biological valves decreased. The time course of PVE development was different for recipients of mechanical valves than for those with porcine valves. Of the 19 early infection episodes, 3 involved porcine bioprostheses and 16 involved mechanical valves, whereas of 74 episodes of late PVE, 50 involved porcine bioprostheses and 24 involved mechanical valves.
The replaced valve was aortic in 49 episodes and mitral in 27. Episodes involving multiple prosthetic valves included 16 with mitral and aortic and 1 with mitral and tricuspid valves.
The New York Heart Association functional class before surgery showed the following distribution: class I, 4 episodes (5%); class II, 17 (20%); class III, 49 (58%); and class IV, 15 (18%). In 8 episodes, these data were unknown. The risk of early PVE was significantly higher when patients were in class III or IV (P=.002).
In 19 episodes (20%) involving 18 patients, the symptoms of infection began during the first 60 days after surgery. Fever was the most frequent finding (17 episodes). In 12 (63%) of these 19 episodes patients developed congestive heart failure, and in 3 episodes (16%) patients had new arrhythmias (recent atrioventricular conduction block not noticed just after surgery). The frequency of other symptoms is shown in Table 1. At least 1 echocardiogram was performed on all but 1 patient (data not shown). Eleven transthoracic explorations, 1 TEE, and 6 using both techniques were performed. The ejection fraction was 54.6% ± 11.2%; other echocardiographic findings are detailed in Table 2.
Time after surgery that blood cultures were drawn was 42.8 ± 29.9 days. Microorganisms were identified in the blood of 17 patients (Table 3). Sixteen patients underwent valve replacement and medical treatment. At surgery, vegetations were found in 7 valves, dehiscence of the prostheses in 11, and valve-ring abscess in 9. Seven (39%) of 18 patients died, with heart failure being the leading cause (4 patients). The microorganisms isolated in the blood cultures of patients who died were Staphylococcus aureus and Staphylococcus epidermidis in 2 patients each and Enterococcus faecalis in 1 patient; the cultures were negative for the other 2 patients.
In 74 episodes (involving 66 patients), the symptoms began later than 60 days after surgery. Fever was also a frequent finding (70 episodes). Heart failure was present at the time of diagnosis or developed later in 48 episodes (65%), and 19 (26%) had some degree of recent-onset conduction disturbances. Other symptoms and their frequencies are shown in Table 1. Echocardiography was performed on all patients: the transthoracic technique was used in 56 episodes, and both transthoracic echocardigraphy and TEE were used in 20. Ejection fraction was 55.1% ± 10.8%. Table 2 lists other echocardiographic findings. Time between surgery and blood cultures was 5.7 ± 4.0 years (range, 82 days to 16 years). The results of blood cultures taken in all the episodes are shown in Table 3. Sixty-five patients underwent valve replacement and antibiotic drug treatment. Only 9 patients received medical treatment alone. The microorganisms isolated in these episodes were S aureus and Streptococcus species in 2 episodes each and E faecalis and Streptococcus bovis in 1 each; in the other 3, no pathogens were isolated. Of patients who underwent surgery, vegetations were found in 49, dehiscence of the prostheses in 37, and ring-valve abscesses in 25.
Twenty-four (36%) of 66 patients died; in 15 cases, death was attributed to cardiogenic shock. The distribution of infecting organisms in patients who died was as follows: S aureus, 7 patients; Streptoccocus viridans, 6 patients; S epidermidis, Streptococcus species, and negative culture, 2 patients each; and coagulase-negative staphylococci, E faecalis, Candida albicans, Escherichia coli, and group G β-hemolytic streptococci, 1 patient each.
We analyzed differences between early and late episodes of PVE. Because the classic limit accepted by most authors is 60 days, we performed the comparison using this criterion for all variables except the analysis of infecting microorganisms, where we also considered the limit of 120 and 365 days because the results were somewhat unexpected. When the limits between early and late PVE were defined as 60, 120, and 365 days, early episodes represented 20%, 25%, and 31%, respectively.
Patients with early PVE were significantly younger than those with late PVE (51 ± 14 vs 59 ± 13 years) (P=.046).
A history of endocarditis, with native or prosthetic valves, was not statistically associated with early or late PVE. Although systemic embolisms occurred more frequently in early than in late PVE, there were no statistically significant differences (P=.06). The presence of dehiscence of the prostheses seen in the echocardiogram was more frequent in late than in early PVE; however, the 2 groups did not differ significantly (P=.07). Analysis of surgical findings indicated that the presence of vegetations was significantly more frequent in late than in early PVE (P=.007). Ring-valve abscesses showed a tendency to affect aortic valves more than mitral valves (P=.054). Some of these findings had been previously found to be significant; however, in the present study they did not reach statistical significance, probably because of the small sample size.
Our analysis of microbiologic etiology found no statistically significant difference between the isolation of staphylococci in early PVE or streptococci in late PVE (P=.37), even when the limit of early infection was modified to 4 or 12 months.
No statistically significant differences were found in mortality between early and late PVE (P=.83).
Patients underwent surgical treatment in 81 of 93 episodes. In 76 of these 81 episodes, the diagnosis of infection was confirmed at surgery or autopsy by macroscopic, microscopic, or microbiologic findings. We compared these 76 episodes using the von Reyn7 and Duke6 criteria.
Twelve of these 76 episodes were confirmed as PVE histologically or with their valve cultures and the remainder with abscesses, vegetations, or dehiscence found at surgery.
In this comparison of Duke and von Reyn clinical criteria, we established a correspondence between definite and probable diagnosis, respectively (Table 4, Table 5, Table 6, and Table 7). Accordingly, we found a higher definite clinical diagnostic rate using the Duke criteria (79%) compared with probable diagnoses (47%) (P<.001) (Table 8). In addition, as many as 21% of the surgically confirmed diagnoses would be rejected by the von Reyn criteria, whereas only 1 diagnosis would be rejected using the Duke criteria.
Because we found a high frequency of patients with heart failure and conduction disturbances, we analyzed a possible modification of the Duke criteria to obtain more sensitive criteria for the diagnosis of PVE. We decided to add heart failure and atrioventricular conduction disturbances to its minor criteria, maintaining the previous criteria and the requisites for establishing the diagnosis of definite and possible endocarditis. Results are shown in Table 8.
The modifications we suggest significantly improve the sensitivity of the Duke diagnostic criteria (P=.02). This was applicable mainly to late PVE (with the Duke criteria, 82% of late PVE would be definite, whereas 97% would be definite adding the suggested minor criteria; P=.02). However, in early PVE, although a higher percentage was definite when the suggested criteria were added (73% compared with 67% using the Duke criteria), we found no statistical differences (P=.71).
Prosthetic valve endocarditis is a serious complication that can occur at any time after the surgical implantation of a device. According to the rates of the different causative microorganisms, and their possible relationship with infection mechanisms, we differentiate between early and late infections. Although the classic limit accepted by most authors is 60 days, others consider late episodes to be those occurring 4, 6, or even 12 months after surgery. The rates of early and late infections in our hospital (20% and 80%, respectively) are similar to those reported by other authors8,9 when the criterion is 2 months; this proportion is similar if 4 and even 12 months are considered.
The proportion of early infections in our patients was higher in those with mechanical prostheses; however, the proportion of late infections was higher with use of porcine valves. These findings are similar to those reported in the literature10,11; nevertheless, other authors12 have found no differences, and actuarial survival curves are similar for biological and mechanical prostheses.10,12 This difference in the time course for the development of PVE may be because of differences in the pathogenesis of infection between mechanical and porcine valves.10
Our microbiologic results were unexpected because we found no significant differences when comparing late and early infections according to streptococci and staphylococci, even when the early infection was considered in the first 4 or 12 months. As reported by other authors, S epidermidis was the causal agent in many episodes, with early and late presentation. Staphylococcus aureus was responsible for a high proportion of late infections in our patients; this is not the experience of Wilson et al,13 who found only 12 late infections due to S aureus (11% of late PVE) in their review of 5 previous series with a total of 184 episodes. Our results are closer to those of more recent reviews,14,15 in which S aureus caused more than 20% of late PVE. These differences might be caused, as proposed by Wolf et al,14 by a reference bias toward S aureus infections at tertiary care institutions, since severe complications are more frequently seen with this microorganism (in our series, only 4 of 11 late infections by S aureus were from the community).
We found a high number of culture-negative endocarditis episodes, mainly in early presentation (29%), diagnosed echocardiographically. Fourteen of 17 episodes of culture-negative endocarditis were confirmed at surgery as definite episodes. In 33% of early and 24% of late PVE, a TEE had to be performed after a negative or inconclusive transthoracic exploration finding; only 1 TEE had inconclusive findings. These findings confirm the value of echocardiography in diagnosing endocarditis, especially TEE in prosthetic infection. In fact, some authors16 suggest that every patient with a prosthetic valve and suspected endocarditis (whether with low or high clinical probability) should undergo a TEE because it contributes significantly to a more accurate diagnosis.
In our series, 84% with early and 88% with late PVE underwent both medical and surgical treatment. These figures are significantly higher than those reported previously,8,11,12,14,17 in which 28% to 62% of patients underwent surgical repair of infected prostheses. Although assessment of the impact of surgery on mortality is beyond the scope of the present study, we believe that these high reoperation rates correlate with the low mortality found in early and late infections. This is in agreement with the opinion of other authors,4,14 who recommend higher interventionism and aggressive surgical techniques to eradicate as much infected prosthetic material and tissue as possible.
Since Durack et al6 (the Duke criteria) reported their new criteria for the diagnosis of endocarditis, several studies18- 23 have evaluated their sensitivity and specificity. Almost all previous studies, including that by Durack et al,6 analyzed native and prosthetic endocarditis together. Lamas and Eykyn3 considered native and prosthetic infections separately, but they reported a small number of PVE. Nettles et al,23 in a series of 25 episodes of pathological confirmation of PVE, proved the superiority of Duke criteria over von Reyn criteria with similar findings to our actual study. Our series evaluates the Duke criteria in PVE only in more patients with surgically confirmed PVE, and our results suggest the possibility of improving the sensitivity of the Duke criteria.
Clinical information provides a major clue for the diagnosis of patients with suspected endocarditis, as reflected in 4 of the minor Duke criteria. Lamas and Eykyn3 suggested 9 additional minor criteria to improve the sensitivity of the Duke criteria, but, although they consider native and prosthetic infections separately, with good results in both cases, specific physiopathologic characteristics of PVE are not taken into account.
We found conduction disturbances in 24% of our patients (17 in aortic and 4 in mitral infections). Although this percentage is lower than that reported by other authors,24 its high frequency and its anatomical substratum lead us to hypothesize that including atrioventricular conduction disturbances among the minor diagnostic criteria would improve the sensitivity of the Duke criteria in PVE. Moreover, conduction disturbances are more frequent in prosthetic than in native endocarditis,24 probably because of extension of the infection along the suture lines into the paravalvular area.
In our patients, heart failure developed in 63% with early PVE and in 66% with late PVE episodes, rates similar to those in previous studies. Lindner et al16 found a significant association between heart failure and the probability of a clinical diagnosis of endocarditis (although they included native and prosthetic endocarditis). Heart failure was present in 25% to 50% of patients with endocarditis when native and prosthetic infections were considered together,7,25,26 but frequency increased significantly in studies in which only prosthetic infections were taken into account, ranging from 43% to 70%.14,26,27 In addition, prosthesis dysfunction (whether due to dehiscence of the valve or to stenosis) correlates with the development of heart failure.27
Therefore, we hypothesized that new conduction disturbances and heart failure would be 2 clinical data with enough relevance to contribute to the diagnosis of prosthetic endocarditis. With this in mind, we analyzed comparatively the von Reyn criteria7 and the Duke criteria,6 later adding 2 new minor criteria (new conduction disturbances and heart failure), keeping the requirements of establishing definite and possible endocarditis to improve the sensitivity of the Duke criteria in PVE. For this analysis, we took into account only the 76 episodes in which surgery confirmed the accuracy of the diagnosis.
In our series, as in previous studies,3,18- 23 the Duke criteria proved to be significantly more sensitive than the von Reyn criteria because as many as 21% of our patients would have been rejected according to the latter. Despite the improvement in sensitivity in the Duke criteria, 20% of the episodes would still have escaped definite diagnosis—a high percentage when we consider the high mortality rate of PVE and the role of clinical diagnosis in surgical decisions that may modify the prognosis.
When we compared the Duke criteria with our suggested modifications, we found that more than 90% of our episodes are definite PVE when new-onset heart failure and conduction disturbances are included as minor criteria. This means a statistically significant improvement in the sensitivity of the Duke criteria (P=.002). The fact that in episodes of early PVE the addition of these 2 new minor criteria does not reach statistical significance may be because of the smaller number of patients and the lower percentage of atrioventricular conduction disturbances in this subgroup. In addition, because episodes rejected by the Duke criteria were the same as those rejected when the suggested minor criteria were added, we think specificity would be maintained, and the specificity of the Duke criteria has been shown to be high.21
The conclusions of our study are limited by several features. The first limitation is the retrospective analysis of charts. Second, our definition of surgically confirmed PVE is not as rigorous as the von Reyn or Duke6 pathological criteria. However, previous clinicopathologic studies27,28 found macroscopic evidence of abscesses, vegetations, or dehiscence to be specific. In fact, medically treated PVE, at the time of surgery, might yield no histological findings of active endocarditis but only chronic inflammatory reactions.
In summary, we found the Duke criteria to be significantly more sensitive than the von Reyn criteria in PVE diagnosis. Nevertheless, the inclusion in the former criteria of 2 easily clinically recognizable variables, ie, new-onset heart failure and conduction disturbances, would significantly improve the sensitivity of the Duke criteria, establishing a definite clinical diagnosis in a high number of episodes. These preliminary findings require further confirmation to be applied to clinical trials and decision-making algorithms.
Accepted for publication August 3, 1999.
Reprints: M. Carmen Fariñas, MD, PhD, Infectious Diseases Unit, Department of Internal Medicine, Hospital Universitario Marqués de Valdecilla, Santander 39008, Spain.