Cryptococcal Neuroradiological Lesions Correlate with Severity during Cryptococcal Meningoencephalitis in HIV-Positive Patients in the HAART Era

Cryptococcal meningoencephalitis has an overall global mortality rate of 20% in AIDS patients despite antifungals. There is a need for additional means of precise assessment of disease severity. We thus studied the radiological brain images available from 62 HIV-positive patients with cryptococcocal meningoencephalitis to analyse the brain lesions associated with cryptococcosis in relationship with disease severity, and the respective diagnostic contribution of magnetic resonance (MR) versus computed tomography (CT). In this retrospective multicenter analysis, two neuroradiologists blindly reviewed the brain imaging. Prospectively acquired clinical and mycological data were available at baseline and during follow-up. Baseline images were abnormal on 92% of the MR scans contrasting with 53% of the CT scans. MR/CT cryptococcosis-related lesions included mass(es) (21%/9%), dilated perivascular spaces (46%/5%) and pseudocysts (8%/4%). The presence compared to absence of cryptococcosis-related lesions was significantly associated with high serum (78% vs. 42%, p = 0.008) and CSF (81% vs. 50%, p = 0.024) antigen titers, independently of neurological abnormalities. MR detected significantly more cryptococcosis-related lesions than CT for 17 patients who had had both investigations (76% vs. 24%, p = 0.005). In conclusion, MR appears more effective than CT for the evaluation of AIDS-associated cerebral cryptococcosis. Furthermore, brain imaging is an effective tool to assess the initial disease severity in this setting. Given this, we suggest that investigation for cryptococcosis-related lesions is merited, even in the absence of neurological abnormality, if a high fungal burden is suspected on the basis of high serum and/or CSF antigen titers.


Introduction
Cryptococcus neoformans is an encapsulated yeast responsible for severe opportunistic meningoencephalitis mostly in patients with acquired immunodeficiency syndrome (AIDS) [1][2][3]. C. neoformans var grubii is by far the predominant serotype in HIV-infected patients worldwide. The main presentation is a disseminated meningoencephalitis [1][2][3]. Retrospective radiological studies involving a limited number of HIV-infected patients with cerebral cryptococcosis have been performed in the pre-highly active antiretroviral therapy (HAART) era [4][5][6][7]. They describe the abnormal cerebral images during cryptococcal meningoencephalitis. The introduction of HAART has significantly modified the radiological presentation of other opportunistic infections [8]. Given this, it is possible that HAART may also have had an impact on the radiological appearances of cerebral cryptococcosis. This is a particularly interesting theory when considering the demonstrated effect of protease inhibitors on some opportunistic pathogens and, specifically, the impact of indinavir or tipranavir on cryptococcal virulence [9][10]. Radiological data obtained during the post-HAART era is therefore important, as all the data published so far consists of case reports or small series (n#4) of HIV-infected patients [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26].
Cryptococcal meningoencephalitis is still associated with an overall 20% mortality rate despite appropriate antifungal therapy, underlining the urgent need for improved management strategies. Most HIV-infected patients with acute neurological symptoms will undergo radiological brain evaluation. However no study has, to date, specifically evaluated the potential utility of neuroimaging to assess the initial severity of AIDS-associated cryptococcosis. Furthermore, analysis of the respective contribution of cranial computed tomography (CT) versus magnetic resonance (MR) in detecting cryptococcal lesions is scarce, in contrast to other opportunistic infections where the beneficial contribution of MR has been clearly established [4,27].
The aim of the present study was thus to answer these questions using data from a large prospective cohort of HIV-infected patients with culture-proven cryptococcal meningoencephalitis and for whom brain images were available at baseline and during follow-up [3]. In addition, our results were compared with published data on brain imaging during cryptococcosis after a systematic review of the literature.
HAART intake at baseline was not associated with any specific radiological pattern (data not shown). Based on 59 CT images, there was no difference in the percentage of abnormal lesions as a whole, nor specifically of cryptococcosis-related lesions according to the infecting serotype A (n = 49) or D (n = 10) (data not shown).
Relationship between baseline cerebral images and initial severity or outcome. We then assessed whether any of the radiological features were associated with initial severity or specific outcome. High serum and CSF antigens titers at baseline were recorded in 21/27 (78%) and 21/26 (81%) patients with cryptococcosis-related lesions respectively, but only in 13/31 (42%) and 14/28 (50%) patients without (p = 0.008 and p = 0.024, respectively). In the multivariate analysis, parameters independently associated with cryptococcosis-related cerebral lesions were high serum antigen titers ( No statistically significant association was found between cryptococcosis-related lesions as a whole or any specific lesions seen at baseline and the occurrence of mycological failure at the week 2 and month 3 workups or further development of IRIS. Indeed, the 4 patients with IRIS had various baseline CT lesions (intracerebral masses, aspecific caudate nucleus hypodensities, cortico-subcortical atrophy and no lesion, n = 1 each).

Radiological findings during the course of cryptococcal meningoencephalitis
Clinical and mycological data were available for 57 patients during follow-up. Of these, 12 (21.5%) died before month 3, with death related to cryptococcosis for 7/12 (58%). Four patients were subsequently diagnosed with IRIS within a median of 11 months (range, 3-38 months) and neurological sequelae were reported in 6 of the 45 survivors (14%).
Of the 57 patients, 24 patients had repeated radiological procedures, allowing comparison of consecutive images over 3 months. At baseline, these 24 patients were more severely infected with more frequently higher serum antigen titers than the other patients (18/23 (78%) vs. 46%, p,0.001). They also had more subsequent neurological sequelae (5/16 (31%) vs. 1/28 (4%), p = 0.018).  No correlation was found between radiological evolution (exacerbation, stability or improvement of cryptococcosis-related or unrelated lesions) and outcome at month 3 including occurrence of IRIS.
Radio-pathological comparisons underlined the overall low performance of both MR and CT to detect fungal meningitis which was a constant at autopsy [5,26,36]. Several studies pointed out the lack of radiological hydrocephalus in cases of high opening CSF pressure [21,29,[37][38][39][40]. Radiological cerebral atrophy was frequently noted ($30% [4,31]). CT images were reported as normal in more than 40% of cases in the 10/12 series and MRI reported as normal in around 10% of cases in the 4 MR pre-HAART series.

Discussion
We analysed the radiological features of cerebral cryptococcosis in a large prospective cohort of HIV-infected patients in the HAART era and compared brain images with parameters assessing disease severity [3].
The pattern of cryptococcosis-related lesions recorded by CT consisted, here, of predominantly dilated VR spaces, masses and meningitis and on MR images, of mass(es) and pseudocysts with, rarely, radiological meningitis and hydrocephalus, as reported in the pre-HAART era in the available series (Table 3) [4][5][6]26,[28][29][30][31][32][33][34]. Radiological cerebral atrophy however appeared less frequently here than previously reported (#13% here vs. $30% [4,31]), probably because of the diminished incidence of HIVassociated dementia and because cryptococcosis is more often revelatory of AIDS than in the pre-HAART era [41]. Thus, it would appear that the advent of HAART has had no influence on the development of cryptococcal-related lesions (including inflammatory lesions) despite its demonstrated effect on fungal virulence and on the local production of IL-8 [9][10]42]. Strikingly, and in agreement with pre-HAART studies, normal brain imaging (47% by CT and 8% by MR) did not rule out cryptococcal meningoencephalitis. Finally, as reported elsewhere [43], other opportunistic infections can be concomitantly diagnosed (as was the case in 2 of our patients).
The value of brain imaging for assessing AIDS-associated cryptococcosis' initial severity was analyzed. We have previously shown that abnormal brain images were associated with a higher risk of death within three months after the diagnosis [3]. Here, a significant association was found between the existence of cryptococcosis-related radiological brain lesions at baseline and high serum or CSF antigen titers, two major prognostic markers [3]. This association was found even in patients with normal neurological examinations. The Infectious Diseases Society of America guidelines for the management of cryptococcosis, published in 2000, recommend performing neuroimaging prior to lumbar puncture in cases of neurological abnormalities [44]. Our study suggests that: (i) neuroimaging, especially MR, should be considered as a valuable tool to assess cryptococcosis' initial severity in HIV-infected patients, along with other prognostic markers and that (ii) cryptococcosis-related lesions may deserve appraisal even in the absence of neurological abnormalities if a high fungal burden is suspected on the basis of high serum and/or CSF antigen titers.
The question of which investigation should be performed is often answered by which is available faster. MR has clearly been demonstrated to be more accurate than CT for the investigation of other cerebral lesions [29,[45][46]. Two preliminary comparisons of CT and MR in,respectively, 10 and 8 immunocompetent/ compromised patients with cryptococcal meningoencephalitis suggested a higher efficiency of MR over CT for the visualization of VR spaces [4,47] possibly because of the limited inflammation [5,26]. Here, cryptococcosis-related lesions were significantly more frequently observed on MR than on CT images for 62 HIVinfected patients including 17 for whom both investigations were performed. Of note, the VR spaces which appear as the main anatomical site involved, radiologically, during cerebral cryptococcosis are also the site of brain invasion associated with fungemia in a relevant murine model of disseminated cryptococcosis [48][49][50]. Considering its high performance, cerebral MR imaging of infected mice should be a powerful tool for further dissection of cerebral cryptococcosis pathophysiology.
The current guidelines do not comment on the need to repeat radiological investigations [44]. Twenty four patients in our cohort had multiple neuroimaging. They were more likely to have more severe disease and a poorer outcome, an observation consistent with daily practice. Our data from this large subgroup of patients does not support repeated neuromaging in clinically and mycologically stable patients. However, neurologically unstable patients would benefit from further radiological evaluations, keeping in mind the possibility of new opportunistic infections and IRIS [51][52][53].
In conclusion, our study suggests that brain imaging, especially by MR, is an additional effective tool in the assessment of initial disease severity in AIDS-associated cryptococcosis. The absence of neurological abnormality should not preclude neuroimaging especially in cases of suspected high fungal burden on the basis of high antigen titers.

Study population
Sixty two patients with culture-proven cryptococcal meningoencephalitis and available brain imaging were analysed from 21 hospitals in Paris area. These patients were enrolled onto the nationwide multicentric prospective CryptoA/D study [3]. Written informed consent had been obtained in accordance with the French Ethical Committee Recommendations.
Clinical, biological and mycological data were available at baseline for all patients and at 2 weeks and 3 months of antifungal therapy for 57 patients. Initial severity and prognosis was evaluated, as already described [3]. Briefly, initial severity was assessed according to the existence of neurological abnormalities and high ($512) serum or CSF cryptococcal antigen titers. Low CSF white-cell count (,20/ml), elevated CSF protein concentration or hyponatremia were also recorded. Mycological failure after 2 weeks of antifungal therapy corresponded to the persistence of viable yeasts in at least one body system. Treatment failure was defined for the patients followed up to 3 months after the diagnosis of cryptococcosis and consisted of death, persistence of viable yeasts in culture or neurological sequelae. Immune reconstitution inflammatory syndrome (IRIS) was defined according to the literature [52].

Radiological investigations
Neuroimaging was performed according to local practices. All brain CT and MR images available from baseline to 3 months after the diagnosis were collected and blindly analyzed by neuroradiologists experienced in the field of AIDS-associated opportunistic central nervous system infections. Brain lesions were recorded following a pre-established checklist.
Baseline MR and CT images were obtained within 7 days before and up to 14 days after the mycological diagnosis of cryptococcal meningoencephalitis. A minimum of 2 weeks was required between two consecutive procedures to evaluate potential changes. Dual exploration (CT and MR for the same patient) was defined if both investigations were performed within a period of two consecutive weeks. Radiological evolution was evaluated on images obtained by the same procedure (CT or MR) and classified as an exacerbation (i.e., increase in size and/or number of lesions or apparition of any new lesions), a complete or partial response (i.e., disappearance or decrease of more than 50% in the number and/or size of lesions), or otherwise as stable.

Review of reported series in the HIV-infected population
We reviewed and analyzed the literature through a Medline search (up to October 2007). All series including more than 5 HIV-infected cases published in English or French were selected using combinations of brain/cerebral, magnetic resonance imaging/computed tomography, X-Ray, and cryptococcosis/cryptococcal, meningitis/meningoencephalitis as key words. Cases due to C. gattii were excluded because of the reported distinct pattern in terms of geographical distribution, underlying diseases and types of tissue lesions [59].

Statistical analysis
The prevalence of every radiological feature at baseline was determined for all the CT and MR images collected. Then, the association between the presence of cryptococcosis-related lesions at baseline and initial severity parameters with the subsequent occurrence of mycological failure, death, neurological sequelae or IRIS was studied using results obtained with either CT or MR (and considering only MR results in cases of dual exploration). Statistical analysis was performed by using the STATA 8.0 statistical package (Stata Corporation, College Station, Texas, USA). Continuous variables were compared using the nonparametric Mann-Whitney rank sum test. The chi-square or Fisher's exact test was used to assess significant relationships between discrete variables. The Mac Nemar exact test was used to compare the proportion of positive findings with CT and MR in cases of dual exploration. Two-sided p values less than 0.05 were regarded as significant. Multivariate analysis with logistic regression was performed to determine factors associated with cryptococcosisrelated lesions. Odds ratios (OR) and their 95% confidence intervals [95%CI] were determined by means of logistic regression analysis. All variables that were clinically relevant with p value ,0.25 in univariate analysis were entered simultaneously into the initial model. A backward stepwise procedure was used to remove variables until all variables retained in the final model had p values ,0.05. Interactions were explored by means of interaction terms added to the logistic regression model.