Discussion
In this 11-year study in high-volume cardiac surgery centers, the incidence of fungal mediastinitis after cardiac surgery was low, accounting for around 0.05% of initial surgical procedures. Candida spp. were the main causative agents. Prognosis was poor, with almost two-thirds of patients dying within a month of diagnosis.
Our data bring new insights to the current literature, as this is to date the largest multicenter study on post-cardiac surgery fungal mediastinitis. Indeed, mycological data in critically ill patients are poorly reported and are mainly available for invasive Candida and Aspergillosis infections, or in hematological population.
While some risk factors are shared between fungal and bacterial mediastinitis, such as malnutrition, obesity or diabetes mellitus [
11], our study attempted to identify specific factors associated with higher mortality in the fungal variants.
As previously described, a higher Euroscore II and a higher SAPS II score seem to be associated with higher complication rates in patients with fungal mediastinitis. Also, more than half of cases were previously hospitalized before surgery, suggesting a possible association between fungal infection and comorbidities requiring multiple hospital attendance.
With only half of the patients showing local signs at clinical presentation, fungal mediastinitis appears more indolent than bacterial mediastinitis (Table
2). The latter is almost systematically associated with local signs [
12]. Finally, time to diagnosis of fungal mediastinitis after surgery looks prolonged compared to bacterial mediastinitis, with a median delay of 38 days [
13]. The non-bacterial nature and delayed onset of mediastinitis may explain a lower rate of septic shock in our population, given that some data suggest that early onset of mediastinitis is associated with septic shock [
13].
Thirty percent of cases were cardiac transplant recipients with immunosuppressive agents, making them at higher risk of fungal colonization and opportunistic infections. Previous data suggest that the detection of post-cardiac transplant bacterial mediastinitis is very challenging, with lower white blood cell count and fever [
14]. Keeping a high level of suspicion for bacterial and non-bacterial healthcare-associated infections is of paramount importance to improve early diagnosis and prognosis. Our data suggest that the diagnosis of fungal mediastinitis may be significantly delayed and that septic shock in this population occurs latter in about 50% of cases. In order to reduce the diagnosis delay, the fungal wall biomarker β-D-glucan could regularly be assayed in high-risk patients, notably V-A ECMO or heart transplants recipients. If this biomarker could be useful for
Aspergillosis spp. and
Candida spp., it has never been evaluated in this setting and cannot be used in case of
Trichosporon spp. infections.
After initial surgery, one in five patients was supported by V-A ECMO for cardiogenic shock when mediastinitis was diagnosed. Although the higher prevalence of fungal infections on V-A ECMO remains controversial [
15,
16], circulatory support reflects a more severe patient condition, leading to a potential increased susceptibility to infections.
Regarding ventricular mechanical supports, it is important to highlight that five patients received LVAD before initial surgery, and that all benefited from transplantation. Three of them developed postoperative mediastinitis due to fungal pathogens, mostly non-
Candida species. LVAD as destination therapies are at risk of infectious complications, through infection of the percutaneous site or pocket infection [
17], and prior LVAD before heart transplantation has been identified as a supplementary risk factor of bacterial mediastinitis after cardiac transplantation [
1]. These facts may suggest a specific vulnerability to fungal mediastinitis among transplanted patients previously on prolonged heart assistance [
18]. Notably, the role of per and postoperative antifungal prophylaxis to prevent infection in patients receiving long-term assist device or heart transplantation is still a matter of debate.
We presume that airborne contamination from
Aspergillus spp. spores may occur during the surgical procedure [
19]. Spreading in the mediastinal area from a contiguous source or an hematogenous invasion is also conceivable, especially among immunosuppressed patients [
20]. For
Candida spp. and
Trichosporon spp., direct inoculation from skin barrier rupture during surgery or cross-contamination is theoretically possible [
21].
The most common strains were
Candida spp. The subspecies were consistent with the current
Candida distribution described in critically ill patients [
22]. To note,
Aspergillus mediastinitis is a very rare condition after cardiac surgery, with only few case reports [
23‐
25]. In our study, only four patients had postoperative
Aspergillus mediastinitis and none of them died in ICU.
We reported five cases of postoperative mediastinitis caused by Trichosporon spp.; four out of five patients died from these infections, which makes this opportunistic pathogen the deadliest strain with non-albicans Candida, with a reserve of anecdotal evidence. However, the number of cases prevents us from drawing any conclusion about the respective virulence of these pathogens.
There are no specific recommendations to guide the management of fungal mediastinitis. In our study, all but one patient had surgical treatment, which is a cornerstone of the management of postoperative bacterial mediastinitis [
26]. Moreover, half of the patients required at least a second surgical debridement, underlining the difficulties in obtaining satisfying source control.
Most of patients were treated with echinocandin or azoles as a first-line antifungal therapy. However, we found that 15% of fungal strains were resistant to azoles, which mainly involved
C. glabrata and
C. parapsilosis. Whereas fluconazole resistance is already described for these two species [
27,
28], our data strengthen the need to carefully choose the empirical antifungal therapy.
There is increasing concern in the literature, suggesting that non-
albicans Candida can generate a biofilm, yielding issues to remove the fungal burden [
29]. However, it seems that these strains remain sensitive to echinocandin [
30]. Additionally, our results suggest the need to consider non-
albicans Candida species when choosing first-line antifungal treatment. However, most of therapeutic suggestions are adapted from other deep fungal surgical site infections, and proposed from local experiences and based on limited series. No randomized trial could be built to answer to this too rare condition, due to a likely ultra-low recruitment rate. The limited evidence for medical treatment can be extrapolated from recommendations for the management of osteomyelitis and endocarditis candidiasis.
Postoperative
Candida mediastinitis should be treated by surgical debridement, followed by echinocandin or fluconazole in the absence of invasive circulatory support and in the absence of underlying aorta prosthesis. As recommended for endocarditis, the preferred treatment among patients with invasive devices or vascular prosthesis is either lipid formulation amphotericin B (sometimes associated with flucytosine), or high dose echinocandin. Step-down therapy to fluconazole can be considered in patients who have fluconazole-susceptible
Candida isolates and quite stable clinical condition, that is after 2 weeks of initial amphotericin or echinocandin treatment. Optimal treatment duration is unknown, but usually requires several months. For ventricular assist devices that cannot be removed, chronic suppressive therapy with fluconazole is recommended, if the isolate is susceptible, as long as the device remains in place [
31].
In cases of
Aspergillus spp. mediastinitis, voriconazole and liposomal amphotericin B should be preferred as first-line antifungal therapy, analogous to the treatment of extrapulmonary aspergillosis. Long-term treatment with oral voriconazole is recommended after initial therapy [
32]. Eventually,
Trichosporon spp
. mediastinitis medical treatment relies on voriconazole or posaconazole, given the fact that these yeasts are intrinsically resistant to echinocandins [
33]. This point is of utmost importance in the critically ill patients, because guidelines recommend favoring empiric echinocandins treatment in patients with invasive fungal infection [
31].
In addition, high levels of suspicion for fungal mediastinitis should be kept in patients with perioperative clinical evidence of mediastinitis and negative bacteriological cultures, especially in those who received heart transplantation. In this population, intraoperative surgical samples should be sent for mycological analysis, in addition to the usual bacteriological analysis.
Whereas the usual mortality rate of postoperative mediastinitis ranges from 30 to 50% [
3], our study highlights a higher mortality rate of nearly 60%. Fungal infections in the critically ill patients are associated with a high mortality rate, even if this reflects a large spectrum ranging from putative pulmonary aspergillosis to invasive candidiasis in hematological patients. Indeed, in the overall critically ill patients, invasive
Candida infections are associated with roughly 50% mortality rate [
34]. A recent single-center retrospective study focusing on
Candida spp. postoperative mediastinitis underlined a significantly lower survival than bacterial mediastinitis (43 ± 8% vs. 80 ± 6.3%, respectively;
p < 0.0001) [
35]. One explanation might be the high prevalence of ECMO-supported patients (62%) and the over-representation of cardiac transplantation recipients in this cohort. However, this hypothesis must be interpreted with caution, since specific data on fungal mediastinitis is limited.
In our study, a short delay between surgery and infection was related to mortality. This finding was previously described in patients with post-sternotomy bacterial mediastinitis [
13]. The early postoperative decrease and loss of function of lymphocytes [
34] leads to an increased vulnerability to infection. Except for
Aspergillus spp. infections which have a longer time to onset, we did not find any difference in terms of infection delay between the causative agents.
Notably, we found an increased trend of post-cardiac surgery fungal mediastinitis between 2009 and 2019. This observation corroborates a similar tendency in pulmonary fungal infections and in general fungal disease [
36]. This could be related to an increased population of immunocompromised patients, including those who receive immunomodulatory agents. The indications of V-A ECMO have also largely increased worldwide over the last decade, exposing this high-risk population to nosocomial infections. It is noteworthy that the lack of record of actual number of VA-ECMO, LVAD or heart transplantation during the study period prevents us from providing trends of surgical procedures and specific complications in our centers. Moreover, the improvement of the diagnosis techniques of fungal infections [
37] may have contributed to a greater identification rate of fungal mediastinitis, and clinicians awareness’s may have been raised by previous experiences. To note, no change in national or local antibioprophylaxis policy occurred during the study period. So, the changes in case mix over time are likely the predominant factor explaining the increasing incidence.
Overall, evidence-based recommendations about treatments and survival following this cardiac surgical complication are likely to be uncertain, due to methodology issues and extremely low incidence. Our initiative opens the door to a larger sample experience with recruitment of international centers, in order to better appreciate “real-life” epidemiology, outcome and treatment algorithms.
Our study presents several limitations. First, the inclusion period of these 40 cases lasted 11 years in 10 centers, involving potential changes related to surgical and medical management over the years. However, the extremely low incidence in our cohort would be unsuitable for prospective studies. Identification of mediastinitis may have differed between centers, and cases may have been underdiagnosed or diagnosed in other centers. So, this incidence should be viewed as an estimation. Whatever, it is very likely that the incidence would stay very low. Similarly, due to the sample size, no multivariable regression was possible to identify risk factors of onset. Second, our population was heterogeneous, with cardiac transplantation representing a specific immunocompromised population which should probably be considered aside. Our study would have benefited from comparing patients after each specific procedures, notably, coronary/valvular surgery, ECMO/LVAD/heart transplantation recipients, so as to better identify specific patient or procedure risk factors. Larger database will have to indicate the volume of each surgical procedure. Third, we focused on post-cardiac surgery fungal mediastinitis, which does not allow the generalization of our results to other postoperative mediastinitis, including mediastinitis after esophageal or cervico-facial surgery. Finally, in the absence of a control population of bacterial mediastinitis, we could not draw firm comparison between fungal and non-fungal mediastinitis: assertions regarding clinical presentations, risk factors and outcomes should be cautiously considered.
Acknowledgements
We thank the following colleagues for screening patients with potential fungal mediastinitis in their respective centers: Dr. Michael Piagnerelli, Département de Médecine Intensive, CHU-Charleroi, Charleroi Belgium. Pr. Pierre Wauthy, Département de chirurgie cardiaque, Centre Hospitalier Universitaire Brugmann, Bruxelles, Belgium. Pr. David De Bels, Réanimation polyvalente, CHU Brugmann, Bruxelles, Belgium. Dr. Mélanie Dechamps, Soins intensifs cardio-vasculaires, Clinique Universitaire St-Luc, Bruxelles, Belgium. Dr. Vincent Fraipont, Unité de Soins Intensifs, Centre Hospitalier Régional de la Citadelle, Liège, Belgium. Pr. Pierre Damas, Département de Médecine Intensive, CHU Sart Tilman Liège, Belgium. Dr. Yves Bouckaert, Service des Soins Intensifs, CHU Tivoli, La Louvière, Belgium. Pr. Isabelle Michaux, Unité de soins intensifs cardiothoraciques, Mont-Godinne, CHU UCL Namur, Belgium. Dr. Matthieu Jabaudon and Pr. Emmanuel Futier, Anesthésie-réanimation, Nouvel Hôpital d’Estaing, Clermond-Ferrand, France. Dr. Emmanuel Rineau, Département d’Anesthésie-Réanimation, CHU Angers, France. Dr. Thomas Kerforne, Réanimation Chirurgicale Cardio-Thoracique et Vasculaire, CHU Poitiers, France. Dr. Philippe Gaudard and Dr. Norddine Zeroual, Département d’Anesthésie-Réanimation Arnaud de Villeneuve, CHU Montpellier, France. Pr Olivier Collange and Dr. Anne-Claude Roche, Nouvel Hôpital Civil, Hôpitaux Universitaires de Strasbourg, France. Pr. Guillaume Besch, CHU de Besançon, France. Dr Diane Lena, Institut Arnault Tzanck, France. Pr Marc-Olivier Fischer, CHU de Caen, France. Pr. Jérôme Morel, CHU de Saint-Etienne, France. Pr. Belaid Bouhemad, CHU de Dijon-Bourgogne, France. Dr. Wulfran Bougouin, réanimation polyvalente, hôpital privé Jacques Cartier, Massy, France. Pr. Bertrand Rozec, hôpital nord Laennec, CHU de Nantes, France. Dr. Osama Abou-Arab, CHU d’Amiens, France. Dr. Fatou Dramé, CHU Beaujon, Clichy, France. Pr. Françoise Botterel and Dr. Françoise Foulet, laboratoire de mycologie, CHU Henri Mondor, Créteil, France. Pr. Alexandre Ouattara, Service d’Anesthésie-réanimation GH Sud, CMC Magellan, Hopital Haut-Lévêque, CHU de Bordeaux, France.