Surgery for blood culture-negative infective endocarditis: outcomes and the role of molecular biological imaging as diagnostic approach^†

Abstract

OBJECTIVES

The study aimed to analyse outcomes of surgery for blood culture-negative infective endocarditis (BCNIE) and to evaluate the role of molecular biological imaging.

METHODS

Patients undergoing surgery for native or prosthetic valve endocarditis from 2013 to 2022 were analysed regarding blood culture-positive infective endocarditis (BCPIE) and BCNIE. For laboratory diagnostics in BCNIE, excised valves or prostheses underwent conventional microbiological culture and fluorescence in situ hybridization combined with 16S rRNA-gene polymerase chain reaction and sequencing (FISHseq).

RESULTS

Of 521 patients overall, we identified 473 patients (342 males, mean age 63 ± 14.4 years) with preoperative blood cultures: 396 with BCPIE (83.7%) and 77 with BCNIE (16.3%). Preoperative characteristics and operative procedures were comparable between groups and the calculated perioperative risk (EuroSCORE II) was identical (BCNIE: 12.2 ± 8.8%, BCPIE: 12.9 ± 11.9%, P = 0.788). At surgery, signs of infective endocarditis were present in 71%, missing in 18% and inconclusive in 10% of patients with BCNIE. While valve cultures alone identified pathogens in 32% of BCNIE patients, the combination with FISHseq confirmed the infective aetiology in 98% and identified causative pathogens in 52%. Overall, early mortality was similar in BCNIE (16.9%) and BCPIE (18.2%, P = 0.620), but increased in 37 BCNIE patients without pathogen identification compared to 40 patients with pathogen identification (27.0% vs 7.5%, P = 0.032).

CONCLUSIONS

Integrating FISHseq as molecular biological imaging technique into valve analysis algorithms in patients undergoing surgery for BCNIE increases diagnostic gain and potentially improves outcome.

Open in new tabDownload slide

INTRODUCTION

Approximately half of patients with infective endocarditis (IE) require surgical treatment and in up to 42% of them preoperative blood cultures (BC) do not detect a causative microorganism [1–5]. Compared to blood culture-positive IE (BCPIE), the overall outcomes of blood culture-negative IE (BCNIE) have been shown to be inferior, whereas the results after surgery for BCNIE are still not clear enough [3, 5–8].

Although pathogen identification is essential for targeted antibiotic therapy, diagnostics in BCNIE remain challenging [1]. Recently, the molecular imaging technique FISHseq (fluorescence in situ hybridization combined with 16S rRNA-gene PCR and sequencing) for examination of infected heart valve tissue and prostheses has been shown to improve pathogen detection by conventional valve culture (VC) in approximately 30% overall [9–11].

The objectives of this study were to analyse outcomes in patients undergoing surgery for BCNIE and to clarify the role of FISHseq as diagnostic technique in these patients. In particular, we addressed the following research questions: (1) Is early mortality different between patients with BCPIE and BCNIE? (2) What impact has FISHseq on pathogen identification in patients with BCNIE? and (3) Is early mortality different if pathogens were identified in BCNIE or not?

PATIENTS AND METHODS

Ethical statement

All patients undergoing surgery for native/prosthetic valve IE at the Department of Cardiovascular Surgery at Charité between January 2013 and December 2022 were retrospectively reviewed. The Institutional Ethics Committee approved the study (EA4/092/22, 23.06.2022) and waived additional informed consent. Age <18 years and ongoing intravenous drug abuse were exclusion criteria. Of 521 patients, we identified 473 consecutive patients (342 males, mean age 63 ± 14.4 years) with preoperative BC: 77 patients with BCNIE and 396 patients with BCPIE.

Definitions

According to the ESC 2015 modified criteria, diagnosis of IE was based on clinical findings, laboratory testing and imaging results [2]. If no pathogen was detected in three BC sets, it was defined as BCNIE. Infections exceeding the cusps/leaflets of native valves or the prosthesis causing dehiscence or perivalvular destruction were considered locally uncontrolled. The perioperative risk was defined by EuroSCORE II [12].

Surgical treatment

All operations were performed through a median sternotomy using cardiopulmonary bypass and cardioplegic arrest. Infected native and prosthetic valves were removed and abscesses and fistulas were thoroughly debrided and disinfected. In the presence of large tissue defects, a pericardial patch repair was performed. Indicated concomitant procedures were performed according to standard techniques.

Medical treatment

Preoperatively, antibiotic treatment followed BC findings and guidelines [2]. In BCNIE, either the combination of ampicillin, flucloxacillin and gentamicin (for community-acquired IE of native valves and prosthetic valve endocarditis [PVE] ≥12 months post-surgery) or the combination of vancomycin, rifampin and gentamicin (for healthcare-associated IE and PVE <12 months post-surgery) was used [2]. Post-operatively, antibiotic regimens were adapted based on VC and FISHseq findings and continued for at least 2 weeks. In BCNIE and failed pathogen identification, treatment was individualized after consultation with an infectious disease specialist from the endocarditis team.

Laboratory diagnostics in BCNIE

VC and FISHseq of excised valves or prostheses were sequentially as well as simultaneously (predominantly) performed as previously described [10, 11]. Briefly, heart-valve samples were processed, Gram-stained and cultured for VC. If positive, microorganisms were identified by routine techniques. For FISHseq, fixated and processed samples underwent hybridization and epifluorescence microscopy. Sections were first screened with a pan-bacterial, 16S rRNA-directed probe. The nucleic acid stain 4’,6-diamidino-2-phenylindole (DAPI) was used to visualize host cell nuclei and ribosome-depleted microorganisms. Upon detection of microorganisms, FISH-probes were applied for identification at the genus- or species-specific level. Detection of FISH-positive microorganisms was regarded as active IE and FISH-negative microorganisms as treated IE. Adjacent sections were submitted to gene sequencing. If images were DAPI-negative and neither FISH nor PCR detected microorganisms, IE was excluded.

Data analysis

Data reporting and analysis followed published definitions and guidelines [13]. Categorical variables are presented as absolute and relative frequencies. For continuous data, means and standard deviations or if there was evidence of skewed data distribution, gauged using a histogram, medians with lower and upper quartiles were calculated. Continuous variables were compared between groups using the Mann–Whitney U-test and categorical data by Pearson chi-squared test. A P-value <0.05 was considered to be statistically significant. All the statistical analyses were performed using IBM SPSS Statistics for Windows, version 29.0 (IBM Corp., Armonk, NY, USA).

RESULTS

Preoperative status

Baseline characteristics are listed in Table 1. There was no difference in demographic variables except for a higher proportion of females in BCNIE. Fever was more common in BCPIE and congestive heart failure more common in BCNIE. In both groups, the aortic and mitral valve were predominantly infected and approximately 30% were PVE. Apart from a trend towards more locally uncontrolled infections in BCPIE, valvular manifestations (vegetations and lesions) did not differ. The pattern of predisposing and comorbid conditions was also comparable. Considering embolic events and shock, systemic manifestations of IE were also similar affecting a significant number of patients. Thus, the estimated perioperative risk (EuroSCORE II) was remarkably high, but not different.

Operative data

Procedural details are given in Table 2. There was no significant difference between groups. In BCPIE and BCNIE, 187 (47.2%) and 36 (46.8%) patients (P = 0.940) underwent concomitant procedures. Four patients died intraoperatively due to intractable bleeding (n = 2), ventricular rupture (n = 1) and vasoplegia (n = 1). By direct inspection, most cases with BCNIE (n = 55, 71%) were classified as positive (clear signs of IE), whereas 8 (10%) were inconclusive and 14 (18%) were negative (Fig. 1, upper part).

Postoperative laboratory findings

Figure 1 visualizes the results of further laboratory diagnostics in BCNIE according to intraoperative findings. Valve specimens from 73 patients (95%) underwent VC and pathogens were identified in 23 (32%), a rate that compares with 34% in BCPIE. Table 3 shows detailed results (pathogens in BCPIE are listed in Supplementary Material, Table S1). Although VC confirmed IE in 43% of macroscopically negative findings, it mostly failed to detect microorganisms (Fig. 1, middle part). Of 50 negative VC, 41 (82%) underwent FISHseq. As an example, Fig. 2 shows pathogen identification using this technique. FISHseq excluded IE in only one patient (2%), meaning that IE was present in 63/64 specimens (98%) from BCNIE analysed by VC or FISHseq. Positive FISHseq findings classified IE as active (n = 9, 23%), treated (n = 11, 28%) and not excluded (n = 20, 50%). Pathogen identification succeeded in nine cases of active IE (100%) and in eight of treated IE (73%). Combining the results of VC and FISHseq, 42 microorganisms were identified in 40 patients with BCNIE (Table 3). Compared to BCPIE, the spectrum of pathogens was dominated by Gram-positive bacteria different from classical IE causing pathogens (staphylococci, streptococci and enterococci) and by Gram-negative bacteria (Fig. 3). When FISHseq identified pathogens, antimicrobial therapy was adapted essentially following guideline recommendations [2].

Operative outcomes

Table 4 lists variables describing post-operative results. Comparing BCPIE and BCNIE overall, mortality was high, but not different. The duration of intensive care unit and hospital stay was also comparable. The re-exploration and reoperation rates as well as the incidence of wound infections were higher in BCNIE. Except for a statistical trend towards increased postoperative renal failure in BCNIE, there was no group difference regarding other complications. Analysing 37 patients with BCNIE and no proof of pathogens versus 40 patients with pathogen identification—either by VC or FISHseq—revealed inferior results regarding mortality when pathogen identification failed (27.0% vs 7.5%, P = 0.032). There was no significant difference in relevant preoperative characteristics between these subgroups (Supplementary Material, Table S2).

DISCUSSION

The main results from this analysis of 473 patients undergoing surgery for IE are the following: (i) 16% of patients had BCNIE; (ii) although differences were observed regarding gender ratio and the presence of fever, heart failure and uncontrolled infection, the perioperative risk (EuroSCORE II) was identical in BCNIE and BCPIE; (III) at surgery, most patients with BCNIE showed signs undoubtedly positive for IE; (IV) whereas VC alone identified pathogens in 32% of BCNIE, the combination with FISHseq confirmed infective aetiology in 98% and identified causative microorganisms in 52%; (V) the spectrum of pathogens eventually identified in BCNIE was dominated by Gram-positive bacteria different from classical IE causing pathogens and by Gram-negative bacteria; (VI) although early mortality was not different in BCNIE and BCPIE overall, inferior survival was observed in 37 patients with BCNIE in whom pathogen identification by BC, VC or FISHseq failed.

Operative outcome in BCNIE

The potential impact of BCNIE on outcomes after surgery remains a matter of debate. Looking only at BC findings, we found no difference in early mortality for BCNIE compared to BCPIE. This aligns with an analysis of 2715 Italian patients undergoing surgery for IE between 2000 and 2015 demonstrating that BCNIE was not associated with early mortality [14]. Renzulli et al. also could not find an association of negative BC and VC with early and long-term mortality when analysing 232 patients (114 with negative BC and VC) undergoing surgery for IE between 1978 and 1998 [6]. In contrast, an unfavourable impact of BCNIE on survival was reported in smaller surgical series [5, 8]. These results could be related to inhomogeneities of study cohorts as significantly more patients (31% and 42%, respectively) had BCNIE therein. Recently, an ancillary analysis of the ESC-EORP EURO-ENDO registry (3113 patients overall, 523 patients with negative BC and VC) provided new insights [3]. Although different from the main study analysis, the authors report that culture negativity had a negative impact on 30-day and 1-year mortality both in medically and surgically treated patients when sensitivity analyses on mortality were performed using Cox regression models with LASSO penalty for variable selection [3]. Of note, the present study also suggests inferior results regarding early mortality in BCNIE when pathogen identification fails. Similarly, the Spanish GAMES study (2000 patients with IE) demonstrated that the absence of pathogen detection (by blood, valve and other cultures, polymerase chain reaction, and serology) was an independent predictor of in-hospital mortality [15]. As targeted antibiotic therapy—besides surgical debridement whenever indicated an essential element of effective IE treatment [1]—requires pathogen identification and unspecific or inappropriate antibiotic treatment may lead to unfavourable results, comprehensive diagnostics are imperative to clarify the aetiology in BCNIE.

Identification of causative pathogens in BCNIE

Here, most patients with BCNIE presented with clear signs of IE intraoperatively. Ideally, pathogens can be identified from BC and/or VC and subsequently tested for antibiotic susceptibility. In fact, conventional microbiological techniques frequently fail in endocarditis: BC are negative in up to 42% and the positive detection rates of VC (herein 32%), has been reported to be as low as 6–26% overall [3–5, 16]. In recent years, molecular biological techniques, mostly based on nucleic acid amplification and sequencing, have been increasingly used to support diagnostics [1, 9–11, 15, 17]. Using broad range and specific real-time PCR for example, a study of 119 surgically explanted valves from 283 patients with BCNIE identified causative microorganisms in 44% and 38%, respectively [17]. FISHseq, the combination of PCR with the microscopic method of fluorescence in situ hybridization, furthermore allows to (i) distinguish between active and resting bacteria, (ii) exclude contamination and (iii) detect biofilms [9–11]. In this study, FISHseq confirmed IE in 98% of tested specimens and identified pathogens in 43%. Together with VC, the infective aetiology of endocarditis was proven in 86% of tested patients with BCNIE and causative microorganisms were identified in 52%. Moreover, information about IE activity was valuable to determine the duration of antibiotic therapy in 20 patients.

The spectrum of microorganisms eventually identified in BCNIE differed from BCPIE (Fig. 3). The role of Gram-positive cocci as BCNIE agents has been previously observed in other series and is—inter alia—attributable to the early administration of antibiotics [17]. For the same reason, FISHseq detected streptococci more frequently than culture (Table 3). Fastidious pathogens (Bartonella spp. and Tropheryma whipplei), being not detectable in BC and difficult to cultivate, were rather detected by FISHseq (Fig. 2). The higher rate of Gram-negative bacteria in BCNIE patients herein can be explained by detection of Bartonella spp (4/7 Gram-negative bacteria) and the generally increasing prevalence of Gram-negative pathogens, being frequently cleared faster from the bloodstream by antibiotics [17]. Also, because the pattern of microorganisms causing BCNIE is highly variable (local epidemiology, study period, patient characteristics etc), empirical antibiotic treatment may be inappropriate in the individual patient. Therefore, obtaining infected tissue or prosthetic material that allows pathogen identification, can be an additional argument for surgical treatment of BCNIE.

Molecular biological imaging in clinical routine

As the efficacy of conventional microbiological methods regarding confirmation or exclusion of infective aetiology in BCNIE is limited, molecular biological techniques are increasingly used in clinical practice, today [1, 2]. For IE requiring surgical treatment, the benefit of molecular biological imaging using FISHseq was shown with this and other studies [9–11]. Certainly, it is associated with additional costs (depending on the health care system, about twice the price of PCR/sequencing alone), but facilitates a gain in diagnostics in particular in BCNIE. Based on the results herein, integrating FISHseq as molecular biological imaging technique into algorithms for valve analysis and clinical decision making (endocarditis team) as recently shown by Hopf et al. [18] may be recommended.

Limitations

This is a retrospective, non-randomized study with inherent limitations. It reports results from a tertiary surgical centre as part of a large medical university, but patients have been referred from different hospitals. Thus, variations in preoperative management are inevitable. The generalizability of findings is limited by the single centre nature of the study. Furthermore, the evolution of laboratory techniques and variations in perioperative treatment during the study period should be considered. Presumably, a higher rate of pathogen identification and a more precise specification of FISHseq would have been achieved, if all resected valves or prostheses had undergone microbiological and FISHseq analysis. By studying patients undergoing surgery for BCNIE in a 10-year period however, we were able to cumulate a sufficiently sized cohort potentially reducing interpretation bias within the confines of a retrospective study. Nevertheless, the observation of increased mortality in patients with BCNIE without pathogen identification could be limited by small patient numbers.

CONCLUSION

Overall, early mortality after surgery for BCNIE did not differ from BCPIE, but excess mortality was observed in patients with BCNIE when pathogen identification failed. Although most patients with BCNIE showed clear signs of IE at surgery, VC alone identified pathogens in only 32%, whereas the combination with FISHseq confirmed the infective aetiology of endocarditis in almost all cases and identified the causative pathogen in 52%. Further research is required to determine how integrating FISHseq as molecular biological imaging technique into valve analysis algorithms in patients undergoing surgery for BCNIE can affect outcome.

Footnotes

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

FUNDING

This research received no grant from funding agencies in the public, commercial or not-for-profit sectors.

Conflict of interest: Annette Moter is founder and head of Moter Diagnostics and discloses lecture honoraria (BioMerieux and Chiesi), support for invited talks (BÄMI, ECCMID, SGM, ERASMUS+ and REMMDI), and patents pending for “Control Preparation for FISH Methods in Microbiology” (PCT/EP2016/051630). Annette Moter and Judith Kikhney founded and hold shares in MoKi Analytics GmbH and report German Federal Ministry of Research and Education grants 13N15824 (FIELD), 13GW0414A (PROCEED), 13N15815 (TEAM) and European Union ITN grant 814168 (GROWTH). Judith Kikhney discloses meeting/travel support (German Society of Hygiene and Microbiology). Dinah on Schöning received meeting participation fees (Insmed Germany GmbH). Volkmar Falk declares institutional financial activities (educational grants including travel support, fees for lectures/speeches, professional consultation and research funds) with the following commercial entities: Medtronic GmbH, Biotronik SE&Co, Abiomed GmbH, Abbott GmbH&CoKG, Boston Scientific, Edwards Lifesciences, Berlin Heart, Novartis Pharma GmbH, JOTEC/CryoLife GmbH, LivaNova and Zurich Heart. All other authors have nothing to disclose.

DATA AVAILABILITY

The data underlying this article are available in the article and in its online supplementary material.

ACKNOWLEDGMENT

We are thankful to Ms. Kathrin Krüger, who provided excellent administrative support.

Author contributions

Herko Grubitzsch: Conceptualization; Formal analysis; Investigation; Methodology; Project administration; Visualization; Writing—original draft; Writing—review & editing. Johannes Erik Puritz: Data curation; Formal analysis; Investigation; Writing—review & editing. Dustin Greve: Data curation; Validation; Writing—review & editing. Hector Rodriguez Cetina Biefer: Conceptualization; Methodology; Software; Writing—review & editing. Dinah von Schöning: Methodology; Supervision; Writing—review & editing. Judith Kikhney: Methodology; Supervision; Writing—review & editing. Annette Moter: Methodology; Supervision; Writing—review & editing. Stefanie-Teodora Sima: Data curation; Software; Writing—review & editing. Matthias Schneider-Reigbert: Supervision; Validation; Writing—review & editing. Miriam Songa Stegemann: Supervision; Validation; Writing—review & editing. Frieder Pfäfflin: Supervision; Validation; Writing—review & editing. Volkmar Falk: Project administration; Resources; Supervision; Writing—review & editing

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks Mateo Marin-Cuartas and the other anonymous reviewers for their contribution to the peer review process of this article.

REFERENCES

Abbreviations

Abbreviations
 
• BC

  Blood culture

 
• BCNIE

  Blood culture-negative infective endocarditis

 
• BCPIE

  Blood culture-positive infective endocarditis

 
• FISHseq

  Fluorescence in situ hybridization combined with 16S rRNA-gene polymerase chain reaction and sequencing

 
• IE

  Infective endocarditis

 
• PVE

  Prosthetic valve endocarditis

 
• VC

  Valve culture