Clinical and economic burden of invasive fungal diseases in Europe: focus on pre-emptive and empirical treatment of Aspergillus and Candida species

Twenty-one studies were identified across ten countries, including one European-wide study, which explicitly reported IFD-related or attributable mortality. However, a clear definition of IFD-related death was lacking in many studies, with criteria ranging from all deaths during the study to patients with hyphal invasion on autopsy. Many patients had underlying conditions; differentiating mortality related to these conditions from deaths directly caused by IFD was problematic. Furthermore, the proportion of IFD-attributable deaths was calculated using a variety of methods (e.g. in relation to the total number of deaths or in relation to the total population). Supplementary Table 3 reports the rates of IFD-related mortality and risk factors by European country.

The overall 28- or 30-day mortality burden in patients with IFDs ranged from 1.5 to 82.4 %, with most estimates falling between 35 and 45 %. Three studies fell outside this range, with mortality estimates of <30 and >45 % [20, 33, 39]. These mortality rates may be explained by specific patient populations, with the highest 28-day mortality rates (66.7–82.4 %) in patients in an intensive care unit (ICU) with probable or proven IA [20].

The 12-week overall mortality ranged from 22 to 47 % [15, 19, 33, 51, 61]. Other time frames for mortality measures were 7, 14, 42, 60 and 100 days, 4 months and 4 years, and mortality estimates ranged from 5 % in haematopoietic stem cell transplant (HSCT) patients (at 100 days) [82] to 70.6 % (at 14 days) in an ICU population with IA [20]. Twenty-one studies reported mortality over an unspecified time frame. The highest overall mortality rate (90 %) was reported in paediatric patients with haematological malignancies and IA [76], and in ICU patients with Candida non-albicans infection [45].

The overall mortality due to Candida infection ranged from 20 to 90 %; however, after the exclusion of ICU populations, this narrowed to 20–32 %. In a study by Almirante and colleagues, the rates of overall mortality of 23 and 43 % were reported for patients with C. parapsilosis and C. albicans, respectively (p = 0.003) [64].

These results indicate that mortality rates in patients receiving antifungal treatment depend on the time period over which mortality is measured, the responsible organism and the underlying disease characteristics of the patient.

Galactomannan testing for the diagnosis of Aspergillus infections was reported in 20 studies across ten countries. All of the studies used serum specimens, and three also used bronchoalveolar lavage (BAL) fluid [15, 22, 61]. BAL testing was preferred in the study by Slobbe and colleagues [61], and serum was only used when a bronchoscopy could not be performed. The testing frequency ranged from twice weekly [28, 78, 79] to daily monitoring [17] as part of a screening strategy. Four studies measured galactomannan following a triggering event as part of an intensive diagnostic workup (IDWU) [18, 51, 57, 61]. Optical density cut-offs ranged from >0.5 to ≥1.5; the most frequently used criteria for positivity was an optical density index ≥0.5 for two consecutive tests.

Reports varied regarding the utility of galactomannan testing in the diagnosis of IFD. Barnes and colleagues investigated the use of a febrile neutropaenia care pathway that used routine biomarker testing in patients with acute leukaemia, refractory disease undergoing aggressive chemotherapy or undergoing HSCT [79]. In this study, 61 patients were positive for Aspergillus by PCR; of these, 25 (41 %) were positive using the galactomannan assay. Thirty-two patients were positive by galactomannan testing and all but seven were also positive by PCR. Thirty-six patients were positive by Aspergillus PCR alone. Antigen testing (galactomannan and mannan) was less sensitive than PCR but demonstrated good specificity. Combined antigen and PCR testing gave a sensitivity of 100 and 87.5 % for single and multiple results, respectively, and a specificity of 100 %.

Girmenia and colleagues used galactomannan testing as part of an IDWU and reported detection in 89 % of cases [51]. Galactomannan assays contributed to diagnosis in 26 of 100 IDWU as a result of at least two positive samples, while 42 of 100 IDWU computed tomography (CT) findings led to IFD diagnosis. Both galactomannan tests and high-resolution CT (HRCT) scans were positive in the same IDWU in 74 % of patients who were diagnosed with pulmonary IFD. The detection of galactomannan occurred after a mean of 4.7 days (range 3–7 days) following the CT results in four cases, and galactomannan test results were negative despite positive culture findings in three cases [51]. No further data regarding the specificity or sensitivity were provided.

Mannan testing for the diagnosis of Candida infections was reported in two studies, each using a minimum cut-off of 0.5 ng/ml [56, 79]. In the aforementioned study by Barnes and colleagues [79], testing was performed twice weekly in conjunction with other biomarker tests. Eleven patients had positive Candida PCR results, five of whom also had a positive mannan antigen assay.

Posteraro and colleagues reported success with the use of Candida mannan screening versus Candida culture surveillance in BAL in the neonatal ICU [56]. Sixteen infants with positive mannan assays were compared with 16 historical controls with positive surveillance cultures. The incidence of IFD in the surveillance culture group was 23 %; no cases of IFD were reported in the group undergoing mannan screening.

β-D-glucan is not specific for Aspergillus species and was only referred to in a nationwide Aspergillus registry in Austria, in which (1,3)β-D-glucan testing was used in 3 % of cases as a diagnostic tool [15].

Studies conducted in the UK and Austria reported that PCR testing was routinely performed in 4 and 16 % of cases, respectively [15, 41]. PCR assays were used in a variety of ways in addition to screening/pre-emptive strategies. For example, in a study conducted in Germany, a positive PCR assay of BAL specimens was used in conjunction with suspicious CT findings for infection diagnosis [38]. By contrast, Rubio and colleagues performed PCR testing as a confirmation assay in patients who were positive for invasive filamentous fungal infection by pathology [76].

A prospective study conducted in Germany compared outcomes in patients following allogeneic HSCT who were treated empirically with antifungals versus those treated pre-emptively after PCR testing [39]. Patients treated with antifungals after a positive PCR assay had significantly decreased mortality at 30 days (p = 0.015), but this survival benefit did not persist at 100 days. Therefore, while pre-emptive treatment following a positive PCR assay appears to confer survival benefits, routine use in the clinical setting may present challenges [39].

Barnes and colleagues reported on the reproducibility and rapidity of PCR test results versus antigen testing by performing assays for both Aspergillus and Candida. All cases of proven disease were positive by both PCR and antigen testing [79]. A comparison of the PCR and galactomannan results revealed simultaneous positive results in six patients, a positive PCR prior to galactomannan testing in 15 patients and a positive galactomannan test prior to PCR in four patients [79].

Most studies that reported use of HRCT were conducted in oncology patients; few details were provided regarding reasons for performing an HRCT scan (e.g. clinical suspicion or as part of routine monitoring). Weisser and colleagues reported on the use of HRCT scans in 161 episodes of infection in 107 patients with haematological malignancies [78]. Scans were performed once weekly or when clinically indicated. The halo sign, air crescent sign or cavitatory lesions were classified as major signs from HRCT scans, whereas all other infiltrates were classified as minor signs. Minor signs were reported in 43 % of cases and major signs in 7 % of cases; no infiltrate was seen with 50 % of the infection episodes.

Two studies reported the use of diagnostic HRCT scans in populations other than oncology patients. In one study, a suggestive HRCT was noted in 41 % of patients in a general hospital who had a positive pulmonary isolate for Aspergillus [63]. In the second study, which focused on critically ill patients with neutropaenia, an HRCT scan was performed in 17/67 patients with probable or proven IA with abnormalities on chest X-ray. Of these, three patients had a halo sign and nine had cavitation. The remaining five patients had non-specific changes [17].

The use of antifungal prophylaxis ranged from 28 to 100 % in 21 identified studies across 12 countries. All azole- and amphotericin-type products were used, and fluconazole was the most frequently reported prophylactic agent. Most studies did not address breakthrough infections, although a rate of 30 % was reported in an Austrian registry that defined breakthrough infection as proven fungal infection after 7 days of prophylaxis [15]. Therapeutic drug monitoring was not performed in this study; therefore, low drug concentrations could not be excluded as the reason for breakthrough infection.

Lafaurie and colleagues conducted a prospective chart review of empirical therapy in a heterogeneous population of patients admitted to haematology, oncology, ICU or infectious-disease wards in a French hospital [31]. Empirical treatment was initiated for refractory fever (39 %), recurrent fever (48 %), clinical signs of sepsis (5 %) and tachycardia with elevated C-reactive protein (CRP) (4 %) [31]. Caspofungin was the first-line treatment in 71 % of episodes, followed by liposomal amphotericin B (18 %) and amphotericin B deoxycholate (11 %) [31]. Eleven percent of patients had breakthrough infection (any IFD identified after 3 days of empirical treatment), although this could be attributed to the high-risk population.

In a second study focusing on empirical therapy, a retrospective chart review of patients undergoing induction of salvage chemotherapy for leukaemia given antifungal therapy on day 4 of fever with negative cultures was conducted [57]. The fungal-attributable mortality was 2.5 %, leading the authors to conclude that, while empirical therapy has become a controversial treatment modality, it may be of value in selected high-risk populations [57].

Eight identified studies reported some form of a pre-emptive strategy, although definitions were diverse. Criteria for treatment initiation and the specific approaches used were different for all pre-emptive studies (Table 1). The presence of clinical signs and symptoms were triggers for treatment initiation in two studies [28, 79]; HRCT scans alone or with clinical microbiological or positive galactomannan tests triggered treatment in five studies [17, 51, 74, 79, 82]; galactomannan testing was used in four studies [17, 25, 51, 79], with one study using the Candida mannan assay [79]; PCR testing was used in two studies [39, 79]. A variety of treatment regimens were given to patients. Seven studies pre-defined drug treatment, while one study did not [51].

Overall, pre-emptive treatment was initiated in 7.7–51.7 % of patients [17, 25, 39, 51, 74, 79, 82]. Most studies reported pre-emptive antifungal use of between 39.2 and 51.7 %. Three studies reported low rates of pre-emptive treatment initiation [17, 51, 82]. Maertens and colleagues had the lowest number of patients receiving therapy due to strict requirements for treatment initiation (two positive galactomannan assays or positive microbiological results with suggestive CT findings) [17]. Dignan and colleagues had strict definitions for antifungal treatment initiation, which resulted in 17 % of patients receiving pre-emptive treatment [82]. Caspofungin was started: if there was a positive CT and neutropaenic fever; if a CT could not be performed within 24 h; or at the discretion of the physician in patients who developed respiratory failure with high dependency or were in the ICU [82]. In another study, only one patient in the pre-emptive therapy arm received treatment secondary to a negative IDWU and worsening clinical condition [51].

These findings indicate a lack of consensus and understanding of the definition of pre-emptive therapy. Therefore, the applicability of currently available data to the wider clinical setting is unclear.

Ten studies across six countries reported on some form of empirical or pre-emptive therapy. A summary of non-randomised, observational studies of empirical and pre-emptive treatments in adult patients is presented in Table 2. Two prospective randomised studies conducted in oncology patients directly compared empirical and pre-emptive strategies (Table 3) [25, 39].

In the comparative studies, the incidence of IFDs ranged between 2.7 and 8.2 % in empirically treated patients and between 8.2 and 9.1 % in the pre-emptively treated patients [25, 39]. In the study conducted by Cordonnier and colleagues, the difference in the IFD rate between the treatment groups was statistically significant (pre-emptive 9.1 % vs. empirical 2.7 %, p < 0.02) [25]. The rate of IFDs was significantly higher in a subgroup of patients who received induction therapy (16.4 %) versus patients who received consolidation therapy/autologous stem cell transplantation (3.9 %, p < 0.01), with 15/17 IFD cases occurring in the induction group and most infections occurring in the pre-emptive arm (12 cases of IFD). Despite the randomised study design, there were differences between the two treatment groups, which must be taken into account when interpreting these data. For example, greater morbidity was recorded for patients in the pre-emptive group versus the empirical group. Outcomes were similar between pre-emptive and empirical strategies in the study by Hebart and colleagues, where patient demographics were more consistent, all patients were on prophylaxis at baseline and only liposomal amphotericin B was used [39].

The overall mortality in the two comparative trials ranged from 1.5 to 16.5 % and IFD mortality ranged from 0 to 13.2 % [25, 39]. The study by Cordonnier and colleagues demonstrated non-inferiority in the overall population between the empirical and pre-emptive groups with regards to survival at 2 weeks after recovery from neutropaenia [25]. In a subgroup analysis of patients who received induction therapy, the inferiority of pre-emptive versus empirical treatment in terms of 2-week survival could not be ruled out, but the study was not powered to prove this. This may be related to a longer median duration of neutropaenia in the pre-emptive versus the empirical arm in patients receiving induction therapy (26 days vs. 12 days, respectively). By contrast, non-inferiority was demonstrated for empirical versus pre-emptive treatment in a subgroup of patients who received consolidation therapy/autologous stem cell transplantation. While empirical treatment may be comparatively beneficial versus pre-emptive treatment in patients receiving induction therapy (but not in patients receiving consolidation therapy), the differences between the groups in this study must be considered.

In the Hebart study, mortality at Day 100 did not differ between the pre-emptive and empirical groups, although the pre-emptive group had better survival at Day 30 [39]. Pre-emptive treatment was initiated after a positive PCR test; after Day 30, regular PCR testing became more difficult, as most patients had been discharged from hospital. A study with consistent PCR testing beyond 30 days would be instructive in delineating whether pre-emptive treatment confers a tangible survival benefit over empirical therapy.

The studies reviewed herein indicated that IFD cases may be missed less frequently when patients are treated pre-emptively versus empirically. In total, seven IFD cases were missed in three studies. Four proven cases of IFD were missed in the empirical group and one proven case of IFD was missed in the PCR-triggered pre-emptive group reported by Hebart et al. [39]. Maertens and colleagues identified one case of invasive zygomycosis where a patient in the empirical group did not receive antifungal treatment [17]. This patient did not present with fever or other signs of IFDs. By contrast, ten patients were treated pre-emptively due to positive galactomannan testing, despite being non-febrile or having another source of fever identified. These patients had no evidence of IFDs and would not have received treatment via an empirical fever-driven strategy.

Cordonnier and colleagues defined breakthrough infections as infections documented ≥24 h after the first dose of antifungal treatment. The occurrence of breakthrough infections was similar for the empirical and pre-emptive treatment arms (empirical arm Aspergillus infections, 1.3 % of patients; pre-emptive arm Aspergillus species or Candida species, 1.4 % of patients for both organisms) [25]. Two cases of breakthrough candidaemia with C. glabrata were identified with blood cultures in a study of pre-emptive treatment by Maertens and colleagues [17].

Overall, excluding the study by Hebart et al. [39], empirical or potentially empirical antifungal use ranged from 35 to 61.3 %, while pre-emptive antifungal use ranged from 7.7 to 39.4 %. Thus, five out of six studies demonstrated an overall decrease in antifungal use by 43–78 % with the use of a pre-emptive treatment versus empirical strategy [17, 25, 51, 74, 79, 82].

Limited data are available comparing the cost-effectiveness of pre-emptive versus empirical strategies for IFDs. However, a benefit in terms of cost and LOS in high-risk patients has been suggested for pre-emptive treatment [79].

These results emphasise the diverse diagnostic methods and therapeutic modalities that can be used within a pre-emptive strategy. Various IFDs were represented in the studies, including Candida species. However, when patient populations were consistent, all patients received prophylaxis at baseline and amphotericin B was used by all patients, and outcomes were similar for pre-emptive and empirical treatment strategies [39]. The rates of breakthrough infection were comparable. The rates of antifungal use were lower with pre-emptive treatment and infection may be detected more frequently using this strategy. Additional studies of pre-emptive antifungal treatment are warranted.

The hospital perspective costs varied depending on the resources included in the analysis. Costs generally ranged from €8,351 to €11,821 when evaluating incremental hospitalisation and antifungal drug expenditure only, €3,930–€7,314 for antifungal drugs, €8,252–€51,760 for hospital bed day costs and €26,596–€83,300 when all direct costs for management were included [35‐37, 58, 59, 61, 67, 68, 77, 86].

Bruynesteyn and colleagues used a decision tree to compare the cost-effectiveness of caspofungin versus liposomal amphotericin B from the UK National Health Service perspective [80]. The average direct treatment costs were £9,763 for caspofungin and £11,795 for liposomal amphotericin B, where the drug costs were £4,601 and £6,395, respectively [80]. The difference in other direct costs, including hospitalisation and drug costs related to the management of adverse events, was £239 in favour of caspofungin up to a weight of 77 kg [80]. Van Campenhout and colleagues compared cost data from an observational study with a model utilising voriconazole for IA therapy [19]. The average total costs of treatment when considering only hospital days associated with fungal infection was €12,376, and most of this was associated with antifungal use [19].

Two studies reported costs from the societal perspective, although one study reported a narrow societal perspective and only included direct costs [40, 60]. Both studies used a Markov model and compared voriconazole with conventional amphotericin B for IA, with one study also including itraconazole [40, 60]. The mean treatment costs for voriconazole and conventional amphotericin B ranged from €25,353 to €26,974 in a 12-week model and €30,026–€33,616 in a life-long model [40, 60].