Introduction

Invasive fungal infections (IFI) of the central nervous system (CNS) are notoriously difficult to treat and mostly affect immunocompromised patients [1, 2]. Major factors impacting the outcome include the severity of immunosuppression, the susceptibility pattern of the infecting fungus and, importantly, the ability of antifungal agents to cross the blood–brain barrier [3]. Amphotericin B, the echinocandins, itraconazole and posaconazole are large molecules (>700 Da) with limited CNS penetration. Fluconazole and 5-fluorocytosine penetrate well into the CNS, but exhibit a narrow spectrum of activity. In contrast, voriconazole displays broad antifungal activity and enhanced CNS penetration. Peak voriconazole cerebrospinal fluid levels >1 μg/mL and concentrations in human brain tissue or brain abscess material exceeding 1 μg/g have been repeatedly reported [4].

In a retrospective analysis of 81 patients with cerebral aspergillosis, an improved rate (35%) of complete and partial responses with voriconazole treatment was recorded and 31% of patients survived the infection for a median time of 390 days [5]. Interestingly, there was a significant trend towards improved survival in patients who had neurosurgical interventions.

A growing number of publications indicate that voriconazole has substantial clinical activity in other, non-Aspergillus CNS fungal diseases, including those caused by Scedosporium spp. [6], various yeasts and rare or emerging fungi, but most published data are limited to case reports and small case series. Thus, we systematically reviewed the cases from the voriconazole clinical trials database and literature data to evaluate the clinical efficacy of voriconazole against various CNS fungal infections.

Patients and methods

The Pfizer voriconazole clinical trials database (plus supplementary data for Scedosporium, see Troke et al. [6]) was queried for CNS IFI. Database patients came from those enrolled in a Phase III, comparative invasive aspergillosis study [7], a Phase III non-comparative salvage therapy study of IFI [8], a Phase II study of acute invasive aspergillosis [9], a Phase II study of chronic IFI [10] and unpublished compassionate studies. In addition, published data (all were individual case reports or small case series), in English, from 1st January 2002 (the year voriconazole was marketed) to 31st December 2008, inclusive, were identified via Medline and using the search terms as follows: voriconazole, brain, CNS, cerebral, cerebrospinal fluid and Aspergillus, Blastomyces, Candida, Coccidioides, Cryptococcus, Fusarium, Histoplasma, hyalohyphomycetes, phaeohyphomycetes, Pseudallescheria, Scedosporium and zygomycetes.

All cases were reviewed individually (by S.S. and P.F.T.) and only those that we classed as satisfying the recent EORTC/MSG criteria for proven or probable CNS infections were included [11]. Efficacy was based on the investigator’s assessment at the end of therapy and, in line with current practice [12], a complete or partial response was classed as success, while all other responses were classed as failure.

For the purposes of this analysis, salvage therapy patients were defined as those receiving at least one day of appropriate antifungal therapy (i.e. not fluconazole in cases of aspergillosis) prior to commencing voriconazole.

Survival is defined throughout as from the first day of voriconazole therapy until death or the last day that the patient was known to be alive.

For statistical analyses, the distributions of categorical variables were compared using the Chi-square or Fisher’s exact tests, as appropriate. All-cause mortality and survival was assessed using the Kaplan–Meier curves and the log-rank test was used to detect statistically significant differences. There was no adjustment for multiplicity of statistical testing. Analyses were conducted in SAS, version 8.2.

Results

Baseline characteristics and microbiology

Patients with confirmed proven or probable infection were included in this analysis after a review of over 200 cases. The 192 selected patients (database n = 114 and published cases n = 78) included 127 males and 65 females, and 41 (21%) were less than 18 years of age (Table 1). There were 137 proven and 55 probable infections. One probable CNS aspergillosis infection from the previously published dataset [5] was downgraded to possible infection and excluded.

Table 1 Baseline and clinical characteristics of 192 patients with central nervous system (CNS) fungal infections and voriconazole therapy
Full size table

The majority of patients were infected with Aspergillus species (n = 120) and the 53 infections speciated were predominantly A. fumigatus (89%), including one triple infection with A. fumigatus, A. fischeri and A. nidulans. In addition, there were 34 infections caused by Scedosporium species and 38 caused by various yeasts and rare or emerging fungi (Table 1).

Most patients were salvage therapy cases (157/192, 82%) and had, thus, received prior antifungal therapy. A single patient with aspergillosis received only prior fluconazole and was considered to be receiving voriconazole as the primary therapy. Known prior therapies were predominantly amphotericin B (either deoxycholate; n = 86 or a lipid formulation; n = 73). However, azoles (itraconazole or fluconazole; n = 68), 5-fluorocytosine (n = 19), echinocandins (caspofungin or anidulafungin; n = 9) and a few other treatments (rifampicin or interferon or granulocyte colony-stimulating factor [G-CSF]; n = 6) were also used. A total of 91/157 (58%) salvage therapy patients were known to have received multiple therapies either sequentially or as combinations.

For all Phase III database cases, and the published cases where this was clearly recorded, the initial voriconazole dose regimen was 6 mg/kg i.v. × 2 on Day 1, followed by 4 mg/kg i.v. q12 h. Patients might then switch to 200 mg p.o. b.i.d. The two Phase II aspergillosis studies used different initial regimens (either 6 mg/kg × 2 on day 1, then 3 mg/kg i.v. q12 h., then 200 mg p.o. b.i.d. [9] or 200 mg b.i.d. p.o. [10]).

Response to therapy

A successful response was recorded in 93/192 (48%) patients (Table 2), with a median duration of voriconazole therapy of 93 days (range 1–1,128). Patients receiving at least 5 days of voriconazole therapy exhibited a 51% (88/174) response rate. There was no significant difference in the rates of complete response for patients with proven versus probable infections (13/34, 38% vs. 6/23, 26%; p = 0.4). However, 36/93 (39%) of patients with a successful response had insufficient information (generally, imaging data) for a clear assignment to complete or partial responders, thereby, precluding this analysis.

Table 2 Outcome by various efficacy groups
Full size table

Paediatric patients showed a somewhat better response rate than adults (63 vs. 44%, p = 0.06, NS) (Table 2). The median duration of voriconazole therapy was 231 days (range 3–1,128 days) and 86 days (range 1–900 days) for paediatric and adult patients (p = 0.004), respectively. Slightly fewer paediatric patients than adults received primary voriconazole therapy (7/41, 17% vs. 28/151, 19%; p = 1 NS) or suffered from aspergillosis (22/41, 54% vs. 98/151, 65%; p = 0.2 NS), while more paediatric patients than adults underwent HSCT or suffered from haematologic malignancies (17/41, 41% vs. 53/151, 35%; p = 0.47 NS), received combination therapy (11/41, 27% vs. 26/151, 17%; p = 0.18 NS) or underwent neurosurgical interventions (17/41, 41% vs. 55/151, 36%; p = 0.59 NS). The response rates of paediatric patients from the clinical trials database or the literature were better than the responses for their corresponding adult populations (12/23, 52% vs. 25/91, 27%; p = 0.04 and 14/18, 78% vs. 42/60, 70%; p = 0.77, respectively). However, the 78 patients from the literature had larger median voriconazole therapy durations (Table 1) and also exhibited significantly better response rates compared with the 114 patients from the database (56/78, 72% vs. 37/114, 32%, p < 0.001) (Table 2).

Response by pathogen and underlying condition

Responses were seen in 56% (19/34) of patients with scedosporiosis, 47% (56/120) of patients with aspergillosis and 47% (18/38) of patients infected with other fungi (Table 2). Amongst the 38 patients infected with other fungi were 12 with cryptococcosis, nine of whom had acquired immune deficiency syndrome (AIDS) as their underlying condition. The three non-AIDS cases were all treated successfully. The nine AIDS patients had durations of voriconazole therapy ranging from 7 to 148 days, but 8/9 eventually failed therapy. These therapy failures were all salvage cases and were known to have previously received fluconazole (5), amphotericin formulations (7), 5-fluorocytosine (4) or itraconazole (1). However, most were not receiving triple therapy for their AIDS infection.

The response rates also varied across risk groups, from 14% (5/35) in patients with HSCT to 72% (41/57) in patients with other underlying conditions and absent or a less severe immunosuppression (Table 3). The clinical response rate of patients with HSCT was significantly less than all of the other groups (p < 0.001), while patients with other underlying conditions had significantly better response rates compared to all of the others combined (p < 0.001) (Table 3). There was no difference in response rates between patients with probable or proven infections (69/137, 50% vs. 24/55, 44%; p = 0.39 NS) (Table 2).

Table 3 Outcome by underlying conditions
Full size table

Adjunctive treatments

In 37 patients, caspofungin (n = 18), amphotericin B lipid formulations (L-AMB, n = 17), amphotericin B deoxycholate (n = 6), 5-fluorocytosine (n = 5), terbinafine (n = 4), itraconazole (n = 2), γ-interferon (n = 2), G-CSF (n = 1) or posaconazole (n = 1) were given in addition to voriconazole or as sequential therapy (termination of voriconazole and continuation of alternate therapy). A slight, but not significant, trend towards an improved response in favour of patients with these additional treatments versus voriconazole monotherapy was detected (23/37, 62% vs. 70/155, 45%, p = 0.09 NS) (Table 2). However, only a single patient with adjunctive treatment had previously received HSCT.

Treatments which included L-AMB appeared to show a somewhat better response rate than those which did not (12/17, 71% vs. 11/20, 55%; p = 0.5 NS), but none of the 17 L-AMB patients underwent HSCT and only three suffered from a haematologic malignancy. Treatments with or without caspofungin showed similar response rates (11/18, 61% vs. 12/19, 63%; p = 0.15 NS).

In addition, 72 patients underwent various neurosurgical interventions, including ventricular drainage/shunt (7), abscess biopsy/drainage (28) and other (mostly craniotomy or abscess resection, 37). Patients receiving these neurosurgical interventions showed significantly higher response rates compared with those without/unknown neurosurgery (43/72, 60% vs. 50/120, 42%; p = 0.017). In all, 74% (53/72) of the patients who underwent neurosurgery were receiving voriconazole as salvage therapy (salvage therapy success 30/53, 57%; primary therapy success 13/19, 68%).

Survival

A total of 112/188 (60%) patients were known to have survived for 90 days or more from the start of voriconazole therapy and 54/188 (29%) survived for at least 365 days (four patients had missing survival times). Only 10/54 (19%) of these long-term survivors had HSCT or a haematological malignancy as their predisposing underlying condition. The median survival time for all 188 patients was 297 days (range 3 to >2,000, Table 4). There was a trend towards shorter survival in patients with aspergillosis, although this was not significant (Table 4).

Table 4 Known survival
Full size table

When survival by underlying condition was examined, there was a significant difference between the groups (p < 0.001), and patients with HSCT also had a significantly poorer survival than all of the other groups (p < 0.001) (Fig. 1). There was also a significant difference in survival between those patients with chronic immunosuppression or other underlying conditions and the remaining patients (p < 0.001).

Fig. 1
figure 1

Kaplan–Meier curves of survival by underlying condition: grey dotted line haematopoietic stem cell transplant; black dotted line haematologic malignancy; grey dashed line solid organ transplant; black line chronic immune suppression; grey line other; circles censored patients

Full size image

Paediatric patients (p = 0.014), patients with neurosurgical interventions (p = 0.0399) or those from the publications (p < 0.0001) exhibited significantly better survival compared to adults, those without/unknown neurosurgery or database patients, respectively (Table 4; Fig. 2). There was also evidence for improved survival in those patients who received voriconazole with other antifungal agents over those that did not (p = 0.0149) (Table 4).

Fig. 2
figure 2

Kaplan–Meier curves of survival by neurosurgical intervention: grey dashed line neurosurgery; black line no/unknown; circles censored patients

Full size image

Discussion

In a previous retrospective study of 81 voriconazole-treated patients with CNS aspergillosis from the voriconazole clinical trials database, an encouraging efficacy rate of 35% was observed, despite most patients in this series suffering from profound immunosuppression [5]. Voriconazole has since become the drug of choice for treating this devastating infection [85, 86]. However, apart from case studies, there is no substantive examination of the role of voriconazole for the therapy of other fungal infections of the CNS. In this second, larger retrospective analysis, we update the efficacy of voriconazole against CNS aspergillosis and summarise the available data for other fungal CNS infections. The analysis utilises the voriconazole database until the drug was launched in 2002 and includes additional cases from publications spanning January 2002 to December 2008. Over 200 patients were identified and reviewed using the current definitions of proven and probable fungal infections [11]. The result was 192 cases for final analysis. The 47% response rate for the 120 patients with CNS aspergillosis is an improvement over that previously recorded [5], and this encouraging response rate is maintained for all 192 patients (48%). However, comparison of the database cases with those from the literature revealed highly significant response and survival differences in favour of the published cases, which is likely due to a publication bias (only one patient with HSCT among all literature cases), possibly skewed, at least in part, by an underrepresentation of unsuccessfully treated patients in the literature and improvements in patient management over the 6 years of the literature survey.

Despite this concern, the analysis also revealed some interesting findings. Both response and survival of the 41 paediatric cases were better than those for the 151 adults, with statistically significant differences in survival. These differences were maintained when the database and literature subsets were analysed separately, suggesting a real effect. The fact that fewer paediatric cases than adults suffered from aspergillosis while more received combination therapy and neurosurgical interventions may account, at least in part, for the improvements. It has been suggested that early diagnosis and aggressive intervention in paediatric patients with CNS aspergillosis results in improved outcome [87, 88].

Neurosurgical interventions did seem to favour an improved response rate and survival in these patients, which is consistent with the published data [5, 46]. However, there was no predefined algorithm for selecting an adjunct neurosurgical approach and patients with lesions localised to CNS regions more suitable for surgery might have carried a lower risk of deterioration and, therefore, have received a neurosurgical intervention more frequently. Interestingly, neurosurgical cases with primary voriconazole therapy responded slightly better than those with salvage therapy (68 vs. 57%) and, possibly, were more likely having received an effective “pre-operative antifungal therapy”. However, we were unable to analyse the sequence of medical and surgical therapies due to limitations of the available data. Recently published data, mostly in immunocompetent patients with cerebral aspergillosis, indicate that antifungal treatment prior to neurosurgery is associated with an improved outcome [89].

There was also limited, but not statistically significant, evidence for improved response and survival amongst the 37 patients who received both voriconazole and other antifungal agents, particularly for treatments that included L-AMB (but not caspofungin) with voriconazole. However, these 37 patients formed less than 20% of the dataset and appeared to be somewhat less severely immunocompromised (only 27% with HSCT or haematologic malignancy) than the remaining patients (36% with HSCT or haematologic malignancy). Not surprisingly, patients with the least severe immune compromisation (other underlying conditions) responded the best to therapy (72%) and HSCT the worst (14%). Thus, the benefits of combination therapies in this analysis are unclear and require further study in formal trials.

Except in cryptococcosis, where an amphotericin B formulation combined with 5-fluorocytosine or fluconazole is the therapy of choice [90, 91], meaningful clinical data supporting a combination therapy in CNS fungal infections are lacking. However, two small comparative studies suggest that caspofungin combined with L-AMB may be better than L-AMB alone in invasive aspergillosis or mucormycosis, including a few patients with rhinocerebral mucormycosis [92, 93].

Indeed, there are few substantive studies of the efficacy of any antifungal agent in CNS infections other than cryptococcosis. Conventional amphotericin B as monotherapy for CNS aspergillosis or mono- and combination therapy with 5-fluorocytosine for phaeohyphomycosis gives poor outcomes [94, 95]. However, conventional or lipid formulations of amphotericin B followed by an oral azole are effective in CNS blastomycosis [96], while fluconazole monotherapy of CNS coccidioidomycosis is often effective [97]. Consequently, the data presented in this retrospective study are the most comprehensive by far for any antifungal agent and suggest that voriconazole not only shows promise for CNS fungal infections due to Aspergillus spp., but also those caused by a variety of other, non- Aspergillus fungi. A neurosurgical management was associated with improved outcomes and should always be considered in individual patients with CNS fungal infections, whereas the impact of combination therapies is less clear and these should be assessed in future trials.

We also detected a publication bias and evaluated patients from a study database, which is a clear limitation of this analysis. Other limitations include the wide heterogeneity of the reported cases due to the retrospective analysis, differences in design of the different data sources (prospective clinical trials with well-standardised therapy, case reports), missing information on neutropenic status, the relative severity and initial stage of disease and, in some instances, the reason for switching to voriconazole (non-response vs. toxicity).

However, this approach appears to be currently the best way to gain a deeper insight into the effects of more refined treatment options in patients with CNS fungal infections. The analysis also supports the use of voriconazole for these life-threatening fungal infections.