Active Surveillance Program to Increase Awareness on Invasive Fungal Diseases: the French RESSIF Network (2012 to 2018)

ABSTRACT The French National Reference Center for Invasive Mycoses and Antifungals leads an active and sustained nationwide surveillance program on probable and proven invasive fungal diseases (IFDs) to determine their epidemiology in France. Between 2012 and 2018, a total of 10,886 IFDs were recorded. The incidence increased slightly over time (2.16 to 2.36/10,000 hospitalization days, P = 0.0562) in relation with an increase of fungemia incidence (1.03 to 1.19/10,000, P = 0.0023), while that of other IFDs remained stable. The proportion of ≥65-year-old patients increased from 38.4% to 45.3% (P < 0.0001). Yeast fungemia (n = 5,444) was due mainly to Candida albicans (55.6%) with stable proportions of species over time. Echinocandins became the main drug prescribed (46.7% to 61.8%), but global mortality rate remained unchanged (36.3% at 1 month). Pneumocystis jirovecii pneumonia (n = 2,106) was diagnosed mostly in HIV-negative patients (80.7%) with a significantly higher mortality than in HIV-positive patients (21.9% versus 5.4% at 1 month, P < 0.0001). Invasive aspergillosis (n = 1,661) and mucormycosis (n = 314) were diagnosed mostly in hematology (>60% of the cases) with a global mortality rate of 42.5% and 59.3%, respectively, at 3 months and significant changes in diagnosis procedure over time. More concurrent infections were also diagnosed over time (from 5.4% to 9.4% for mold IFDs, P = 0.0115). In conclusion, we observed an aging of patients with IFD with a significant increase in incidence only for yeast fungemia, a trend toward more concurrent infections, which raises diagnostic and therapeutic issues. Overall, global survival associated with IFDs has not improved despite updated guidelines and new diagnostic tools.

pneumonia [PJP]) of 27.2 cases/100,000 patients per year with an all-cause mortality of 17.0% at 42 days (16). Differences may be related to the populations studied and the sample size (35,876 versus 3,374 cases). However, both studies acknowledged that electronic data extraction is likely to result in some inaccuracies, even after extensive manual validation (15,16). Indeed, misdiagnosis and incorrect coding can result in moderate performances at least in specific populations (17). Furthermore, some diagnoses require the identification of the causal agent (16). The advent of new diagnostic tools and the grouping of laboratories in large platforms may also substantially impact temporal trends.
Active and sustained surveillance systems based on local multidisciplinary teams should overcome some limitations, but such networks are rare. In France, the assignments of NRCs supervised by Santé Publique France (SPF) include expertise and surveillance on specific microorganisms and contribution to outbreak investigations. The NRC for Invasive Mycoses and Antifungals (NRCMA) deals with all pathogenic fungi and is also in charge of surveying all IFDs. Up to 2012, it relied on passive surveillance except for yeast fungemia for which the active surveillance YEASTS program was implemented (18)(19)(20)(21). In 2012, a network called RESeau de Surveillance des Infections Fongiques (RESSIF) was launched to actively survey all proven and probable IFDs and underlying conditions and to characterize the isolates of uncommon species and/or unusual phenotypes at the NRCMA. RESSIF is still ongoing. Our purpose here is to describe its organization and the IFD burden in France between 2012 and 2018. It is also to officially launch the open access database Institut Pasteur FungiBank (https://fungibank.pasteur.fr/) recording data (antifungal susceptibility profiles, sequences, and major clinical settings) associated with the uncommon species responsible for IFDs in France.

RESULTS
Global picture of IFDs in France. Over the 7 years of the study, a total of 10,886 IFDs were recorded in 10,154 patients. Overall, the global incidence of IFD was 2.21/ 10,000 hospitalization days with a slight increase over time (2.16 to 2.36, P = 0.0562) in relation with an increased incidence of fungemia (1.03 to 1.19/10,000, P = 0.0023), while that of other IFDs remained stable (Fig. 1A). There was a 6% decrease in the number of hospitalization days from 2014 until 2018 in the participating centers. Overall, there was a significant increase in concurrent infections diagnosed when considering all IFDs (3.4% to 5.6%, P = 0.0092) or only mold IFDs (from 5.4% to 9.4%, P = 0.0115).
The means of diagnosis differed according to IFDs and underlying conditions (Table 1 and 2) and evolved over time ( Fig. 1D and E and below). PCR-positive samples were mostly respiratory (2,015/2,359, 85.4%) and blood (n = 273, 11.6%) samples.   Results are n/n total (%) unless otherwise stated.
Active Surveillance of Invasive Fungal Diseases in France mBio Global mortality at 3 months was 38.8% (3,364/8,669 patients for whom the information was available) with significant variations according to the IFD diagnosed (see below) and the main underlying condition. The global mortality rate did not change over time nor did that of each IFD (Fig. 1F).
Concurrent IFDs were diagnosed in 74 cases (including 54 IAs, 13 cryptococcoses, and 9 fungemia). They occurred more often in the same anatomical site in patients with malignancy than in the other groups (P = 0.037). Global mortality at 1 month was higher in patients with (21/66, 31.8%) than without (330/1,807, 18.3%; P = 0.006) concurrent infections. The proportion of proven cases was 74.3% (304/409) and 37.9% (643/1,697) among HIV-positive and -negative patients, respectively (P , 0.0001). The means of diagnosis were used unevenly in the four groups (Table S5). The trends in the use of diagnostic tools were oriented inversely with microscopy decreasing, while PCR-based diagnosis increased (P , 0.0001) (Fig. 1E and F). Only 5.7% (n = 119) of the patients received PCP prophylaxis. Cotrimoxazole was used for treatment in 86.8% of the cases.

DISCUSSION
We showed here the potential of a university hospital's network based on active and perennial collaboration with well-trained mycologists and infectious disease clinicians for surveying IFDs. The IFDs recorded through RESSIF include fungemia but also more challenging diagnoses, such as invasive mold infections or PJP. Of note, each  Even if medical specialties are not distributed evenly across centers altering the distribution of populations at risk of IFDs, we can reasonably expect that the sustained surveillance and the aggregation of data collected in 29 academic centers totaling almost 11,000 cases smoothened the center and even geographical effects. The design of the RESSIF network allowed us to describe the landscape of IFDs in France between 2012 and 2018 and their own dynamics. IFDs consisted mainly of fungemia (49%), PJP (20%), and IA (19%) confirming the previous ranking in France (15), but mucormycosis is now more often diagnosed than cryptococcosis. Despite the availability of new diagnostic procedures and publications of guidelines (23,24), the global 3-month mortality remained unchanged over the 7 years, as recently shown for yeast fungemia (25). Overall, the global incidence was 2.21 IFDs/10,000 hospitalization days, with a significant increase over time (Fig. 1A). This increase paralleled the French policy to reduce hospital beds. This reduction may have contributed to an increase in the proportion of more fragile/susceptible patients in hospitals, as shown before (15). This increase is probably linked to the global aging of the French population, as observed in RESSIF, with an increased proportion of patients with diabetes or autoimmune disorders and of elderly patients.
The increase in IFD incidence was supported mainly by an increase in fungemia following a similar observation at the national (2001 to 2010) (15) or regional (Paris area, 2010 to 2017) (18) levels. There is no microbiological explanation for it and no dramatic change in the species diagnosed or their proportions even if preexposure to antifungals influenced the species recovered as previously shown (18,19,21). The population diagnosed with yeast fungemia changed over time with more elderly, more patients with type 2 diabetes, and fewer patients with recent surgery or ICU stay. However, the proportion of patients with malignancies remained stable. As shown in the Paris area (18), antifungal prescriptions evolved with more echinocandins and less fluconazole, but the proportion of resistant isolates did not change (26). Global mortality rate associated with yeast fungemia was 47.8% at 3 months. Whether this result is related to more fragile patients dying with their IFD rather than from the IFD remains a major question.
In contrast to fungemia, there was a significant decrease in the proportion of IA (from 17.7% to 13.6%), without a significant change in its incidence (from 0.38 to 0.32). Underlying conditions remained stable over time, including HM, age, and ICU stay, as well as the species involved that differed according to the risk factors. Azole resistance remained anecdotal (2.1%) for A. fumigatus isolates recovered in patients with IA, being different from chronically colonized patients (27). The use of PCR-based diagnosis was still limited and depended on local practice and population at risk. Global mortality (34.5% and 42.5% at 6 weeks and 3 months, respectively) was in line with previous reports (15) and unfortunately did not change over time. Concurrent fungal infections were recorded in 11.1% of the cases. Similar to findings by Danion et al., who showed the deleterious effect of coinfections during IA (28), global mortality associated with IA at 6 weeks (34.5%%) was much higher for those diagnosed with concurrent fungemia or mucormycosis, whereas PJP or other IFDs had no impact. The increased proportion of coinfections over time could explain the lack of decrease in global mortality despite advances in the management of IA. The relative decrease in the incidence of IA may result from more efficient antifungal prophylaxis in high-risk patients (replacement of the posaconazole solution by tablets providing better absorption [29]). An in-depth analysis is requested to better analyze the factors contributing to these trends. We also acknowledge that IA associated with chronic respiratory diseases was certainly underreported for a lack of clear-cut criteria. It is unlikely to have decreased over time given the high proportion of smokers in France (32% of the population [30]).
Compared with the previous report (15), the data recorded through RESSIF showed a nonsignificant increased proportion of mucormycosis without significant trends in terms of incidence or global mortality over time. The 3-month mortality recorded here (59%) is higher than that in the 2005 to 2007 RetroZygo study (44%) (31), possibly resulting from a higher proportion of patients with malignancies who have the worse prognosis (50% in RetroZygo, 62.3% here). The cases diagnosed in 2000 to 2010 and now are probably different by the underlying diseases themselves with more severely ill patients. This difference may also be a paradoxical consequence of diagnosis by PCR rather than culture in these patients. Indeed, the lack of prognosis improvement contrasts with major changes in the tools used for the diagnosis of mucormycosis with the advent of PCR-based methods in routine laboratory practice (32) and the decreased use of culture. The reason why an earlier diagnosis, as provided by the Mucorales PCR, would not impact global mortality is not trivial and will deserve further analysis, keeping in mind that up to 2017, the PCR test was only confirmatory and that some tests are outsourced for a lack of local resources, thus limiting the benefit of an earlier diagnosis. The incidence of PJP remained stable over the study period. Although PJP has been associated mainly with HIV infection, it is not the case anymore (33). The HIV-infected patients accounted for only 19.3% of the patients diagnosed with PJP. The diagnosis relies nowadays more on PCR than on microscopy even though the significance of microscopy-negative and PCR-positive samples is sometimes questioned by clinicians, especially in HIV-negative patients. Indeed, the fungal load is lower in these patients (34), as shown here by a higher proportion of probable PJP in HIV-negative patients. However, the quantification as inferred by PCR assays was not considered here for a lack of consensus on threshold and method (35). Importantly, global mortality associated with PJP was significantly higher in HIV-negative than in HIV-positive patients, suggesting that even low P. jirovecii DNA loads should not be neglected (36).
Using the RESSIF data to extrapolate incidences at a national scale is challenging. Since RESSIF encompassed 45% of national capacity of university hospitals in terms of hospitalization days, it could be done for IA, mucormycoses, and PJP; these IFDs occur mainly in immunocompromised patients mostly treated in university hospitals. Fungemia however can occur in nonimmunocompromised patients in hospitals not included in the RESSIF network for a lack of a referent mycologist. According to the administrative data for France, a mean of 1,556 candidemia/year was recorded between 2001 and 2010, leading to an annual incidence of candidemia of 2.5 to 3.5/100,000 population/year (15). Here, we collected in average 737 cases/year suggesting that we missed at least 53% of the candidemia knowing that indeed the figures are probably higher with the increase of the French population and its aging. Overall, it is likely that the IFDs burden in France is more than twice what we recorded in RESSIF since we also excluded IFDs following surgery, such as fungal peritonitis (37), and those for which clear criteria are still missing, such as in patients with chronic respiratory diseases.
Even though RESSIF was designed to overcome limitations of surveillance programs based on administrative data, it still has some, as follows: (i) it does not cover the entire country (see above with the estimations at a national level) as opposed, for example, to the Danish survey on candidemia (8); (ii) despite providing clinical data which are usually missing in other surveillance systems, global and attributable mortality analysis would have required additional information, such as preemptive and prophylactic therapies, therapeutic follow-up, and severity grading for ICU patients, which cannot be collected reliably unless specific studies are set up; and (iii) the electronic case report form (e-CRF) includes predefined variables potentially preventing the recording of new events. Interactions with our colleagues and SPF permit uncovering such cases.
Altogether the e-CRF flexibility and interactions with SPF allowed to rapidly evaluate the Saprochaete clavata outbreak in 2012 (38), to assess the small number of Candida auris in France (12), or more recently to investigate the burden of IFDs in the context of COVID-19 (39). The e-CRF is indeed easily amendable through the VOOZANOO platform to record new parameters (e.g., COVID-19 context). (iv) The last issue concerns the completeness of the recordings for a lack of a double check with the hospital databases. We regularly assessed the reporting evolution for each CC-NCRMA and challenged suspicious annual variation. However, to consider the possible weariness of some participants that could lead to incomplete reporting in the future, we are planning to reduce the e-CRF.
We are convinced that implementing this program and involving clinicians and mycologists in the recording helped increasing the awareness on IFDs both at the government agency and the hospital levels. Increasing the knowledge on the epidemiology of IFDs may in turn improve patient prognosis by triggering new prophylactic measures or targeted diagnostic tests for specific populations. RESSIF is already an invaluable tool to use to monitor IFDs and their evolution in France and to allow for an in-depth analysis of unusual settings, including rare pathogens (40) or specific conditions (e.g., diabetes, cirrhosis, and elderly). These analyses could help define new diagnostic or therapeutic strategies and contribute to a better knowledge of IFDs. They may also pave the way to new hypotheses on IFD pathogenesis and new avenues for investigating host/pathogen interactions.
Another benefit of data centralization is the update of taxonomic names based on the NRCMA expertise. While such a change rarely affects therapeutic management, it is important for studies on the relationship between genera/species/clinical presentation and/or outcome for specific groups of pathogenic fungi. In that regard, a request through Institut Pasteur FungiBank could trigger collaborations.

MATERIALS AND METHODS
Network. RESSIF is based on active, sustained, and voluntary participation of collaborative centers (CC-NRCMA; n = 13 in 2012, 29 later on, and 21 active between 2013 and 2018). These CC-NRCMAs correspond to 21/32 university hospitals totaling approximately 45% of hospitalization days of all French university hospitals and covering 15/18 French regions, including overseas territories (see Fig. S1 in the supplemental material). Each CC-NRCMA aggregates several hospitals/wards dealing with adult and pediatric patients with various underlying conditions. A referent well-trained medical mycologist is responsible for the accuracy and completeness of the records based on local diagnosis in collaboration with clinicians. A signed agreement between the CC-NRCMA and the NRCMA determined the duration of the collaboration, the duties, and the benefits for each party. Participating centers and individuals had no financial reward.
Information system VOOZANOO and database organization. The real-time Web-based electronic case report form (e-CRF) was developed using VOOZANOO, an open-source platform to rapidly create questionnaires and surveillance applications (http://www2.voozanoo.net/). The database is saved daily on a server approved for medical data storage. The architecture is hierarchical allowing the administrator (NRCMA) to define levels (NRCMA/city/hospital/ward), and roles with specific rights (referent mycologist and collaborator/clinician) for each account. The rights concern the files (creation, update, deletion, and download) and the tools (creation or simple use of queries and variables available for comparison). Each CC-NRCMA has full access to its own data, is free to perform any in-house study, and can request access to part of the whole data set (32). The coordinating committee, including representatives of CC-NRCMA and NRCMA, oversees validation of the projects. An annual meeting held at Institut Pasteur allows representatives to share updates on the network and results of substudies.
Questionnaire. Data on the patient (gender, date of birth, ward, underlying conditions, immunosuppressive drugs, foreign devices, and travel history) and on the episode of IFD (signs, symptoms, imaging, means of diagnosis, direct examination, culture, histology, PCR, antigen detection, specimens studied, prior antifungal prescription, and coinfections), initial therapy within the first 48 hours (antifungals and surgery), and outcome at 3 months (dead, alive, or lost) with date of last news/death were recorded. Most variables were categorical with unique or multiple choices or drop-down menus to facilitate secure recording and analysis. Some free text entries allowed one to add relevant information. Missing or ambiguous information for outstanding data and failure to send the isolate triggered queries (through e-CRF, mail, or phone call). All edits were traced. A unique "case identifier" allowed one to link episodes for a given patient.
Strain identification. The CC-NRCMAs are responsible for fungal identification. Nevertheless, they agreed to send their isolates for central characterization at the NRCMA, except for the common species (Candida albicans, Candida glabrata, Candida tropicalis, Candida parapsilosis, Candida krusei, Candida kefyr, and Aspergillus fumigatus), unless the isolate exhibited an unusual antifungal susceptibility profile.
The NRCMA provided polyphasic identification based on morphology, matrix-assisted laser desorption ionization-time of flight (MALDI-TOF; Bruker biotyper, Germany) mass spectrometry, DNA barcoding (41)(42)(43), and antifungal susceptibility testing based on the EUCAST broth microdilution method (42,44). Results of the final identification and MICs were mailed individually and recorded in the e-CRF. In case of discrepancy for fungal identification, that achieved by the NRCMA was used for the analysis. DNA sequences and MIC values were also made available through Institut Pasteur FungiBank. Aggregated yearly updated data are available online (https://www.pasteur.fr/fr/file/39845/download). Data analysis. Each record was assigned to one of the six main exclusive categories of risk for IFD (using the following hierarchy: malignancy, solid organ transplantation [SOT], recent surgery [,1 month] excluding curative management of the IFD, HIV infection, and other conditions known as risk factors for specific IFDs [e.g., diabetes, corticosteroid therapy, and systemic inflammatory disease]) or none known. Complex cases were reviewed by two NRCMA staff members.
IFDs were recorded as invasive aspergillosis (IA), PJP, mucormycosis, cryptococcosis, fusariosis, endemic mycosis, fungemia (if not included in the specific IFDs listed), or other IFDs (rare localizations due to common species and infections due uncommon species not listed above). Abscesses following surgery, peritonitis, esophagitis, and pyelonephritis were not recorded. Some patients experienced multiple IFDs that were considered concurrent if within #15 days for pulmonary infections and #5 days for other IFDs or subsequent otherwise. Concurrent coinfections involved different body sites (e.g., pulmonary IA and yeast fungemia) and/or multiple species or genera at the same anatomical site. Recurrence was considered 6 months after mycological cure (10 days for yeast fungemia).
IFDs were classified as proven and probable according to the 2008 European Organization for Research and Treatment of Cancer and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) criteria (45). Possible cases were not analyzed given the high degree of diagnostic uncertainty (45). IA cases diagnosed only by positive PCR tests were included, anticipating the new EORTC/MSGERC criteria (2). Even if they were not yet approved, we used the same approach for mucormycosis. For PJP, classification relied on microscopy and PCR results, as follows: proven with positive microscopy and probable with negative microscopy and a positive qPCR. Global mortality was assessed 1 and 3 months after diagnosis.
Data were analyzed anonymously using Stata/SE 15.1 for Mac (StataCorp., College Station, TX). To avoid autocorrelation, characteristics of the patients, including death, were analyzed for the first episode recorded for the global analysis and for the first episode of the corresponding IFD for the analysis of specific IFDs. Trends and incidence were calculated for the 21 CC-NRCMA perennials between 2013 and 2018. Incidence rates were calculated per 10,000 hospitalization days using annual hospital activity data (SAE administrative data, Ministry of Health, https://www.sae-diffusion.sante.gouv.fr/sae-diffusion/recherche.htm). Official data on the French population were available online (https://www.insee.fr/fr/statistiques/2381476). Their evolution was analyzed by linear regression. The trends of specific parameters were analyzed using the chi-square test for trends. Comparisons were based on chi-square or Fisher's exact test when needed for discrete variables.
Ethical considerations. The research was carried out in compliance with the French law and the Declaration of Helsinki (as adopted in 2000). The surveillance of the NRCMA was approved by the Institut Pasteur Institutional Review Board 1 (2009-34/IRB) and the "Commission National de l'Informatique et des Libertés" according to the French regulation.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, TIF file, 0.7 MB.

ACKNOWLEDGMENTS
This work was supported by recurrent financial support from Santé Publique France and Institut Pasteur. The funders had no role in study design, data collection, analysis, interpretation of data, or the decision to submit the work for publication.
We thank Didier Che (Direction des Maladies Infectieuses, Santé Publique France) for his continuous support and interest in invasive fungal infections and Etienne Sevin (EpiConcept, Paris, France) for his contribution in the development of the RESSIF website using the VOOZANOO platform.
We are thankful for the technical help of Catherine Blanc, Anne Boullié, Cécile Gautier, Virginie Geolier, and Damien Hoinard (Institut Pasteur) for the characterization of the isolates received at the NRCMA. We declare no conflict of interest. The French Mycoses Study Group includes collaborators who contributed to these data by their involvement in the management of patients, expertise in diagnostic tools,