Author + information
- Received November 7, 2012
- Revision received January 3, 2013
- Accepted January 21, 2013
- Published online June 11, 2013.
- Ludivine Saby, MD*,
- Olivia Laas, MD†,
- Gilbert Habib, MD*,* (, )
- Serge Cammilleri, MD, PhD†,
- Julien Mancini, MD, PhD‡,
- Laetitia Tessonnier, MD†,
- Jean-Paul Casalta, MD§,
- Frederique Gouriet, MD, PhD§,
- Alberto Riberi, MD⋮,
- Jean-Francois Avierinos, MD*,
- Frederic Collart, MD⋮,
- Olivier Mundler, MD, PhD†,
- Didier Raoult, MD, PhD§ and
- Franck Thuny, MD, PhD*,§,¶,* ()
- *Département de Cardiologie, Hôpital de la Timone, AP-HM, Aix-Marseille Université, Marseille, France
- †Service Central de Biophysique et Médecine Nucléaire, Pôle Imagerie, Hôpital de la Timone, AP-HM, Aix-Marseille Université, Marseille, France
- ‡LERTIM and UMR 912 INSERM-IRD, Faculté de Médecine, Aix-Marseille Université, Marseille, France
- §Unité de Recherche sur les Maladies Infectieuses et Tropicales Emergentes, UM63, CNRS 7278, IRD 198, Inserm 1095, Faculté de Médecine, Aix-Marseille Université, Marseille, France
- ⋮Service de Chirurgie Cardiaque, Hôpital de la Timone, AP-HM, Aix-Marseille Université, Marseille, France
- ¶Département de cardiologie, Hôpital Nord, AP-HM, Aix-Marseille Université, Marseille, France
- ↵*Reprint requests and correspondence:
Prof. Franck Thuny OR Prof. Gilbert Habib, Département de Cardiologie, Hôpital de la Timone, 264, rue Saint Pierre 13385 Marseille Cedex 5, France.
Objectives This study sought to determine the value of 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) for diagnosing prosthetic valve endocarditis (PVE).
Background The diagnosis of PVE remains challenging. In PVE cases, initial echocardiography is normal or inconclusive in almost 30%, leading to a decreased diagnostic accuracy for the modified Duke criteria.
Methods We prospectively studied 72 consecutive patients suspected of having PVE. All of the patients were subjected to clinical, microbiological, and echocardiographic evaluation. Cardiac PET/CT was performed at admission. The final diagnosis was defined according to the clinical and/or pathological modified Duke criteria determined during a 3-month follow-up.
Results Thirty-six patients (50%) exhibited abnormal FDG uptake around the site of the prosthetic valve. The sensitivity, specificity, positive predictive value, negative predictive value, and global accuracy were as follows (95% confidence interval): 73% (54% to 87%), 80% (56% to 93%), 85% (64% to 95%), 67% (45% to 84%), and 76% (63% to 86%), respectively. Adding abnormal FDG uptake around the prosthetic valve as a new major criterion significantly increased the sensitivity of the modified Duke criteria at admission (70% [52% to 83%] vs. 97% [83% to 99%], p = 0.008). This result was due to a significant reduction (p < 0.0001) in the number of possible PVE cases from 40 (56%) to 23 (32%).
Conclusions The use of 18F-FDG PET/CT was helpful for diagnosing PVE. The results of this study support the addition of abnormal FDG uptake as a novel major criterion for PVE.
Prosthetic valve infective endocarditis (PVE) remains a diagnostic challenge (1). The poor prognosis of this disease is associated, in particular, with deficiencies in identifying prosthetic and periprosthetic damage via echocardiography, which can lead to a delay in the application of antibiotic and surgical therapies (2,3).
Currently, the modified Duke criteria are considered to be the gold standard for diagnosing infective endocarditis. Based on clinical, echocardiographic, microbiological, and pathological findings, these criteria provide a diagnostic probability, classified as definite, possible, or rejected endocarditis (4). This classification has a sensitivity of approximately 80% when the criteria are evaluated at the end of patient follow-up in epidemiological studies (2,5). However, in clinical practice, the modified Duke criteria show a lower diagnostic accuracy for early diagnosis, especially in the case of PVE, for which echocardiography is normal or inconclusive in almost 30% of cases (6,7).
Recently, 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) has been employed to identify inflammatory and infectious processes (8). Thus, several studies have demonstrated the role of PET/CT in detecting vascular prosthesis (9–11) and implantable pacemaker/defibrillator infections (12–14). Moreover, some case reports have suggested a possible role for this technique in diagnosing PVE (15–18), but no study has addressed this issue to date.
Therefore, we aimed to assess the diagnostic value of PET/CT in PVE and to determine whether the addition of abnormal FDG uptake as a major diagnostic criterion would improve the early sensitivity of the Duke classification upon the admission of patients with suspected PVE.
From January 2011 to June 2012, all consecutive adults admitted to the Department of Cardiology at Timone Hospital (Marseille, France) with a suspected diagnosis of PVE were prospectively eligible for entry to the study. In patients with a prosthetic valve, endocarditis was suspected in the presence of at least 1 of the following criteria: 1) unexplained persistent or recurrent fever >38°C; 2) unexplained increase in serum C-reactive protein (CRP) level >10 mg/l; 3) positive blood cultures, independent of the echocardiographic analysis results; 4) positive serological testing for Coxiella burnetii, Bartonella spp., Mycoplasma pneumoniae, Legionella pneumophila, Aspergillus spp., or Tropheryma whipplei; and 5) a mass or a new partial prosthetic valve dehiscence observed using echocardiography. The patient exclusion criteria were pregnancy, inability to lie flat, need for urgent cardiac surgery, hemodynamic instability, cardiac surgery <1 month ago, and blood glucose level >1.8 g/l. All of the patients had clinical, microbiological, and echocardiographic evaluations and PET/CT. Patients with poor PET/CT image quality were also excluded.
Written informed consent was obtained from all of the participating patients using an approved protocol, which was required by the institutional review board.
Clinical, microbiological, and echocardiographic data
The following data were collected from the patients at admission and during hospitalization: age, sex, and Charlson comorbidity index (19); history of diabetes, cancer, intravenous drug use, underlying heart disease, chronic renal insufficiency, and heart failure; leukocyte count, serum CRP level, and causative pathogens (determined by blood culture, serology testing, valve culture, or polymerase chain reaction from a valve specimen according to international guidelines ); and indications for valve surgery. Transthoracic echocardiographic (TTE) and transesophageal echocardiographic (TEE) analyses were performed in all patients at admission, and the data were interpreted by experienced senior cardiologists as previously described (21). TTE and TEE were repeated weekly during the hospitalization period. Briefly, the echocardiographic data included the presence and maximal length of vegetations. Periannular complications were defined as an abscess, pseudoaneurysm, or fistula according to accepted definitions (2). The data were stored electronically and noted the time of the original examination without alteration. Valvular regurgitation was semiquantitatively assessed (22). The left ventricular ejection fraction was quantified using the Simpson biplane method.
After eating a meal rich in fat and very low in carbohydrates to reduce the physiological uptake of FDG in the myocardium, the patients fasted for at least 12 h prior to PET/CT. Intravenous administration of 5 MBq/kg 18F-FDG was performed after the blood glucose level (<1.8 g/l) was checked. PET and whole-body CT scans were carried out consecutively using a Discovery ST PET/CT scanner (General Electric, Milwaukee, Wisconsin) 1 h after 18F-FDG injection. The PET data were acquired in 3-dimensional mode, and the images were corrected for attenuation and reconstructed using the OSEM (ordered-subset-expectation-maximization) algorithm. The PET data were linked with the CT data. The data analysis (Xeleris, General Electric) was based on visual interpretation and quantitative measurement of FDG uptake using the following criteria:
Visual analysis: hypermetabolic intensities in the prosthetic and periprosthetic areas were considered to be abnormal. This uptake had to be confirmed in the uncorrected images. This visual analysis defined whether the PET/CT was positive or negative.
Semiquantitative analysis: the intensity of FDG uptake was determined by measuring the maximal standardized uptake value (SUVmax), which was obtained by normalizing the tissue concentration of FDG activity and the patient’s weight. SUVmax was calculated using an average of 3 measurements from 3 volumes of interest (5 mm3) introduced in the prosthetic region at equal distances from each other. To obtain a background value for FDG uptake, blood SUV was determined by introducing 35-mm3 volumes of interest within the right atrium in locations devoid of significant spillover activity. The prosthetic valve-to-background ratio was calculated by dividing the prosthetic valve SUVmax by the atrial blood SUVmax. This ratio allowed possible biases related to differences in the FDG metabolism to be overcome.
In addition, whole-body acquisition was performed to detect silent embolic events and septic metastases defined as extracardiac abnormal uptakes. Some abnormal uptakes could also correspond to the incidental discovery of tumors. The calculated mean radiation dose for a whole-body 18F-FDG PET/CT was 15 mSv.
The images were interpreted by 2 nuclear medicine physicians who were blinded to the clinical data and the applied imaging convention. In the cases of disagreement, a consensus was reached via a third physician. Patient management was performed independently of the PET/CT results.
The final diagnosis was defined by an expert team according to the clinical and/or pathological modified Duke criteria determined from the data collected during a 3-month follow-up after admission (4). Follow-up after hospital discharge was actively conducted based on consultations at 1 and 3 months, TTE and/or TEE, blood cultures, and biological samples in our department or by contacting the patients or their physicians.
Continuous variables were expressed as the median (interquartile range [IQR]) and compared using the Mann-Whitney U test. Categorical data were expressed as numbers (percentages) and compared using the chi-square or Fisher exact test.
In the first analysis, we determined the diagnostic value of PET/CT (presence or absence of abnormal FDG uptake at the prosthetic valve site). The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy, as well as the 95% confidence interval (CI), were calculated using the final diagnoses of definite PVE and rejected PVE as the outcomes. The prosthetic SUVmax and SUVmax/right atrium ratio were compared among the definite, possible, and rejected final diagnoses using the Kruskal-Wallis test and post hoc Dunn test.
In the second analysis, we added abnormal FDG uptake at the prosthetic valve as an additional major criterion in the modified Duke classification. Then we compared the early diagnostic value at admission of this novel Duke classification (Duke-PET/CT) with the classic Duke classification (Duke) (Table 1). The sensitivities and specificities were compared using the McNemar test. The net reclassification index was calculated to assess the improvement in the diagnosis categorization (23).
The analysis was performed according to the Standards for Reporting of Diagnostic Accuracy guidelines (24). Three patients of the present population were previously described in a recent case report publication (15).
The interobserver variability of PET-CT was assessed using the kappa test.
The analyses were performed using SPSS version 16.0 (SPSS Inc., Chicago, Illinois). All of the tests were 2 sided. A p value <0.05 was considered to be significant.
Seventy-two patients with suspected PVE were included during the study period (Fig. 1). The median age of the patients was 67 years (IQR: 60 to 76 years), and 61% were male. The baseline characteristics are summarized in Table 2. At the end of follow-up, the final diagnosis was classified as definite PVE in 30 patients, rejected PVE in 20, and possible PVE in 22. Fifty-nine patients received antibiotic treatment during hospitalization. Of the 29 patients who underwent cardiac surgery, PVE was pathologically proven in 13 individuals. The indications for surgery were heart failure (n = 16), periannular abscess (n = 12), and high embolic risk (n = 8). Surgery consisted of implantation of bioprosthesis (n = 22), homograft (n = 5), and mechanical prosthesis (n = 2).
Diagnostic value of cardiac PET/CT
PET/CT was performed at a median time of 6 days (IQR: 4 to 11 days) after admission, 30 days (IQR: 6 to 46 days) after onset of symptoms, and 1,484 days (IQR: 526 to 3,396 days) after implantation of the prosthetic valve. In the 55 patients who received antibiotic therapy, PET/CT was performed at a median time of 9 days (IQR: 5 to 19 days) after the beginning of this treatment. A total of 36 of the patients (50%) showed abnormal FDG uptake at the prosthetic valve (positive PET/CT). There was no difference in the rate of positive PET/CT between patients with biological and mechanical prosthetic valves (52% vs. 46%; p = 0.63). The interobserver reproducibility was 86% (kappa = 0.71).
The results of PET/CT according to the final diagnosis are summarized in Table 3. The sensitivity, specificity, positive predictive value, negative predictive value, and global accuracy were 73% (95% CI: 54% to 87%), 80% (95% CI: 56% to 93%), 85% (95% CI: 64% to 95%), 67% (95% CI: 45% to 84%), and 76% (95% CI: 63% to 86%), respectively. Among the 13 patients with pathologically proven PVE, the PET/CT results were positive in 12 patients (92%), whereas echocardiography on admission was positive in only 7 patients (54%). Among the 7 patients who underwent surgery with a negative PET/CT, none had a pathologically proven PVE.
PET/CT was positive in 4 patients with a rejected PVE diagnosis (false positive 15%). Two of these individuals showed abnormal FDG uptake in an ascending aortic graft implanted with a prosthetic valve. In these 2 cases, it was difficult to determine whether the abnormal FDG uptake involved the prosthetic valve or only the aortic root. Additionally, 1 patient exhibited mechanical prosthesis thrombosis with negative tissue culture results after surgery and 1 presented an acute mitral chordae rupture associated with low FDG uptake in the aortic bioprosthetic valve, which appeared normal based on echocardiography. Eight patients with definite PVE showed normal cardiac PET/CT results (false negative 33%). PET/CT was conducted in these patients at a median time of 8.5 days (IQR: 4 to 13 days) after starting antibiotics; in the patients with definite PVE and positive PET/CT, this analysis was performed after a median time of 7 days (IQR: 5 to 16 days) (p = 0.85). The serum CRP level was higher in patients with positive PET/CT, but the difference was not statistically significant (56 mg/l [IQR: 10 to 116 mg/l] vs. 64 mg/l [IQR: 12 to 158 mg/l]; p = 0.43]). Figure 2 shows examples of PET/CT results.
Among the 30 patients with definite PVE, 15 patients (50%) exhibited both a positive initial echocardiogram at admission and positive PET/CT, 7 patients (23%) presented a positive PET/CT despite a negative initial echocardiogram, and 8 patients (27%) exhibited a negative PET/CT despite a positive initial echocardiogram. Thus, no patient with definite PVE displayed both a negative initial echocardiogram and negative PET/CT. Of the 12 patients with a positive PET/CT and negative initial echocardiography, 7 had a definite PVE, and a periprosthetic abscess was identified in 6 individuals during repeated echocardiographic study (Fig. 3) performed at a median time of 8 days (IQR: 7 to 23 days).
The PET/CT was positive in 11 of the 20 patients (55%) with only a vegetation shown in the initial echocardiogram, 5 of the 8 patients (63%) presented periannular complications, and 7 of the 13 patients (54%) exhibited new partial dehiscence of the prosthetic valve (Table 4).
The semiquantitative analysis of FDG uptake showed a significant higher prosthetic valve SUVmax in patients with definite PVE in comparison with the other groups and trend toward a higher prosthetic valve-to-background SUVmax ratio (Fig. 4).
Cardiac PET/CT as a major Duke criterion
The addition of abnormal FDG uptake at the site of the prosthetic valve as a novel major criterion significantly increased the sensitivity of the modified Duke criteria established at admission (70% [95% CI: 52% to 83%] vs. 97% [95% CI: 83% to 99%]; p = 0.008). This result was the consequence of a significant reduction (p < 0.0001) in the number of possible PVE cases from 40 (56%) to 23 (32%). With this strategy, the specificity was not significantly decreased (50% [95% CI: 30% to 70%] vs. 40% [95% CI: 22% to 61%]; p = 0.5) (Table 5). The net reclassification index was 10.3%. The results of the Duke and Duke-PET/CT criteria are summarized for all of the patients in the Online Appendix, Online Table 1.
Whole-body PET/CT to detect emboli, distant infection, and portal of entry
Whole-body PET/CT identified emboli and distant infection in 8 patients (3 splenic emboli and 5 spondylodiscitis). Only one of these lesions was silent. Thus, the results of applying this technique did not significantly improve the value of the modified Duke criteria. In addition, 8 patients had a colonic abnormal FDG uptake. Among them, colonoscopy with biopsies identified cancer in 2 patients and low-grade dysplasia in 3 patients.
PVE represents a diagnostic challenge. The present study tested the use of 18F-FDG PET/CT for diagnosing PVE. We demonstrated that this imaging technique has good diagnostic value, especially when abnormal FDG uptake around the prosthetic valve was added as a major criterion to the modified Duke classification established within a few days after admission. The inclusion of this new criterion significantly increased the sensitivity of the modified Duke classification and allowed for an earlier diagnosis, especially when echocardiography was normal or doubtful.
The early diagnosis of endocarditis is of crucial importance because a delay in antibiotic therapy and cardiac surgery has negative effects on clinical outcomes (3,25). However, diagnosis of this condition is often difficult, especially in the presence of foreign materials such as pacemaker/defibrillator leads and prosthetic valves. In such a situation, echocardiography shows a decreased sensitivity and specificity (2,26). Therefore, efforts have been made to develop new imaging strategies to accelerate the diagnostic process. Although the interest in cardiac CT for use in some situations has increased (27), similar to echocardiography, cardiac CT is still a morphological imaging modality that lacks functional data, which is a major limitation when microbiological data are negative and in the case of absent or doubtful structural lesions. Labeled leukocyte scintigraphy is another option for imaging, but this technique is time consuming (28). By measuring metabolic tissue activity, 18F-FDG PET/CT has emerged as a promising method for diagnosing cardiovascular infections (29). In a large contemporary cohort of patients with presumed PVE, we found that the sensitivity and specificity of this technique were 73% and 80%, respectively. These results were similar to data from Bensimhon et al. (12) on pacemaker lead endocarditis. Although these results are good, we cannot conclude that PET/CT alone is the “magic” modality capable of diagnosing any PVE episodes. Indeed, such an imaging test is difficult to imagine for a complex disease such as endocarditis. However, when the results of PET/CT were taken into account together with other clinical, microbiological, and echocardiographic parameters, our findings showed that the sensitivity of the modified Duke criteria dramatically increased to 97%, a rate never reported previously for this disease. This result was obtained without compromising specificity. Moreover, the inclusion of cardiac PET/CT in the diagnostic strategy demonstrated the potential of the modified Duke criteria to provide a diagnosis earlier in the course of the disease because the criteria were initially developed to define cases of endocarditis for epidemiological studies conducted at the end of patient follow-up (4).
As previously suggested, the major contribution of PET/CT to the early diagnosis of PVE appears to be related to cases for which the initial echocardiographic analysis is negative (15). In addition, abnormal FDG uptake was observed before the appearance of infectious damage in echocardiography in several patients, which highlights the power of PET/CT to reveal infection before significant damage has occurred. Our medical team has recently proposed the inclusion of PET/CT in the strategy of the difficult diagnosis of Q fever endocarditis, for which echocardiography usually shows atypical lesions (30).
The improvement in the sensitivity of the modified Duke criteria obtained by including the PET/CT data resulted from a significant decrease in the rate of possible PVE diagnoses. The possible PVE group has represented a problem in therapeutic decision making because of the potential delay in initiating antibiotic treatment, which can lead to severe complications, such as embolism, cerebral hemorrhage, and acute heart failure. By reclassifying the majority of possible PVE cases to definite PVE, PET/CT might have an impact on therapeutic strategies and clinical outcomes in the future.
Although the results of the PET/CT analysis are encouraging, we did observe false-positive and false-negative tests. The few definite PVE cases with negative PET/CT might be explained by the patients presenting disease with a lower inflammatory activity or PET/CT that was performed too long after initiating antibiotic therapy. The optimal timing for performing PET/CT, according to the onset of antibiotic treatment, should be determined in future studies. In addition, false-positive results of PET/CT may occur when this technique is performed too early after the implantation of the prosthetic valve (31). In the present study, the patients with cardiac surgery performed <1 month earlier were not included to avoid these false-positive results related to the early post-operative inflammation around the sewing ring. We identified other possible causes of false positives, such as severe prosthetic thrombosis or the presence of an aortic root graft. In the latter case, abnormal FDG could be related to the BioGlue surgical adhesive (a topically applied mixture of bovine serum albumin and glutaraldehyde) used to seal the aortic root graft, as reported by Schouten et al. (32). The limited spatial resolution of the PET/CT used in this study did not allow us to determine whether the FDG uptake was present only on the aortic root graft or also involved the prosthetic valve.
The final diagnosis determined according to the results of modified Duke criteria at the end of the follow-up was used as the gold standard in this study. Although this choice could be criticized, this decision allowed the inclusion of patients in whom surgical specimens could not be systematically obtained because of a nonsurgical therapeutic strategy. Furthermore, the favorable results regarding PET/CT sensitivity were confirmed when a surgical specimen positive for infection was used as the gold standard.
The 18F-FDG PET/CT imaging technique was useful for diagnosing PVE. This imaging modality is not a substitute for clinical, microbiological, and echocardiographic evaluation but could be implemented in the global assessment of patients with suspected PVE. By demonstrating the potential role of this methodology in early diagnosis, especially in the case of initial negative echocardiography results, we showed the significant diagnostic impact when positive PET/CT was used as a novel major Duke criterion. Moreover, whole-body imaging was also useful for detecting emboli, metastatic infection, and occult primary tumors. Based on these results, we have proposed an algorithm for incorporating PET/CT in the evaluation of patients with suspected PVE (Fig. 5). Future large-scale studies will be necessary to confirm whether this strategy is cost effective.
For a supplemental table, please see the online version of this article.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- C-reactive protein
- computed tomography
- interquartile range
- positron emission tomography
- prosthetic valve endocarditis
- maximal standardized uptake value
- transesophageal echocardiography
- transthoracic echocardiography
- Received November 7, 2012.
- Revision received January 3, 2013.
- Accepted January 21, 2013.
- American College of Cardiology Foundation
- Habib G.,
- Badano L.,
- Tribouilloy C.,
- et al.
- Li J.S.,
- Sexton D.J.,
- Mick N.,
- et al.
- Habib G.,
- Derumeaux G.,
- Avierinos J.F.,
- et al.
- Vieira M.L.,
- Grinberg M.,
- Pomerantzeff P.M.,
- et al.
- Meller J.,
- Sahlmann C.O.,
- Scheel A.K.
- Sarrazin J.F.,
- Philippon F.,
- Tessier M.,
- et al.
- Saby L.,
- Le Dolley Y.,
- Laas O.,
- et al.
- Habib G.,
- Hoen B.,
- Tornos P.,
- et al.
- Thuny F.,
- Di Salvo G.,
- Belliard O.,
- et al.
- Lodise T.P.,
- McKinnon P.S.,
- Swiderski L.,
- et al.
- Evangelista A.,
- Gonzalez-Alujas M.T.
- Feuchtner G.M.,
- Stolzmann P.,
- Dichtl W.,
- et al.
- Erba P.A.,
- Conti U.,
- Lazzeri E.,
- et al.
- Alavi A.,
- Saboury B.,
- Basu S.