Outcomes of Cryptococcus meningoencephalitis and associated magnetic resonance imaging ndings

Woo-Jin Lee Seoul National University Hospital Young Jin Ryu Seoul National University Bundang Hospital Jangsup Moon Seoul National University Hospital Soon-Tae Lee Seoul National University Hospital Keun-Hwa Jung Seoul National University Hospital Kyung-Il park Seoul National University Hospital Manho Kim Seoul National University Hospital Sang Kun Lee Seoul National University Hospital Kon Chu (  stemcell.snu@gmail.com ) Seoul National University Hospital


Introduction
Cryptococcus neoformans meningoencephalitis is a serious central nervous system (CNS) complication in immunocompromised patients and is associated with a high mortality rate. [1][2][3][4] Although protocols including the administration of intravenous amphotericin B combined with ucytosine for acute induction treatment and uconazole for consolidation and long-term maintenance have been established as standard regimen, 2-5 the clinical outcomes are considerably heterogeneous and a signi cant portion of patients with mild baseline severity neurologically deteriorate and end up with death or permanent sequelae. 1,3,6,7 Prognostic factors for poor outcomes include old age, higher antigen titer in the cerebrospinal uid (CSF), larger ex vivo capsule size of the fungus, increased or decreased intracranial pressure (ICP), high peripheral white blood cell (WBC) count, low body weight, anemia, and features that constitute encephalitis such as reduced Glasgow coma scale (GCS) scores or presence of a seizure; however, these markers do not account for the neurological outcomes. 1,8,9 Additionally, a marker that re ects the disease pathomechanism and estimates the risk of disease progression and poor neurological outcome is still lacking. 6,7 The major route of entry of Cryptococcus into the CNS might be the key to explain the mechanism of disease progression and subsequent poor outcomes. Leukocyte-bound or free Cryptococci can exit the small-sized vessels in the brain and are accumulated in the perivascular space of the CNS, especially the peri-venular space. 10 Considering that the peri-venular space lacks pial membrane, 11 it can be postulated that the degree of peri-venular ow stagnation caused by the accumulated Cryptococcus might determine the risk of Cryptococcus invasion into the brain parenchyma, manifesting as the progression of disease. 4,10 Enlarged perivascular space (ePVS) is a common brain magnetic resonance imaging (MRI) feature associated with Cryptococcus meningoencephalitis. 12,13 Given that ePVS might re ect the perivascular CSF ow stagnation caused by Cryptococcus accumulation, its degree might predict the risk of disease progression and poor outcomes. Similarly, other MRI ndings such as parenchymal cryptococcoma or hydrocephalus might be utilized to monitor the neurological deterioration due to disease progression. 12,13 In this study, we hypothesized that brain MRI ndings might re ect the pathomechanism underlying disease progression and predict the outcomes of Cryptococcus meningoencephalitis, and analyzed the brain MRI ndings, their serial changes, and its association with the disease progression and outcomes.

Study subjects
This retrospective cohort study initially included all consecutive individuals admitted to the neurology department of the Seoul National University Hospital between January 2000 and December 2019 who were diagnosed with Cryptococcus meningoencephalitis. Among the initially included 117 individuals, the nal study population was de ned according to the following criteria: (1) underwent baseline brain MRI evaluation; (2) availability of clinical, treatment, laboratory, and long-term (> 6 months) neurological outcome data. According to the criteria, 33 patients without brain MRI evaluations and eight with inadequate data were sequentially excluded and the remaining 76 individuals were included in the study analysis. Diagnosis of Cryptococcus meningoencephalitis was based on the detection of the Cryptococcus antigen in CSF by latex agglutination or by lateral ow assay along with or without detecting Cryptococcus in CSF by culture or India ink assay. 2,14,15 The design of this study was approved by the institutional review board of the Seoul National University Hospital (SNUH) and the study was performed in compliance with the SNUH IRB regulations and the International Conference on Harmonisation guideline for Good Clinical Practice. Written informed consent was obtained from each patient or the patient's legal surrogate.

Clinical and laboratory evaluation
Along with the demographic information, patients' underlying immune status was reviewed and the causes of immunode ciency were categorized as follows: Human Immunode ciency Virus (HIV) infection, hematologic malignancy, solid organ cancer, post-transplant status, and long-term use of highdose immune suppressants (for indications other than cancer treatment or post-transplantation immunosuppression). 4,16 At baseline, encephalitis feature was de ned according to the 2013 Consensus Statement of the International Encephalitis Consortium diagnostic criteria as: (1) altered mental status lasting more than 24 hours without an alternative cause and (2) 3 or more of the followings: documented fever (> 38.0 °C); seizures not fully attributable to a preexisting seizure disorder; new onset of focal neurologic ndings; CSF WBC count ≥ 5/mm 3 ; and abnormal brain MRI ndings suggestive of encephalitis. 17 Baseline GCS score and modi ed Rankin Scale (mRS) score data were also obtained from the patients' medical records. 1 CSF analysis included the evaluation of protein levels, WBC counts, and the elevation in the opening pressure (≥ 20 cmH2O). 8 CSF Cryptococcus antigen titer was evaluated semi-quantitatively, and high antigen titer was de ned as antigen detection in > 1:1 000 dilution. 1 Treatment pro le analysis Intravenous amphotericin (0.7-1.0 mg/kg/day) with or without ucytosine (100 mg/kg/day) or uconazole (400-800 mg/kg/day) was used for the induction treatment period (within 2 weeks from the treatment initiation). Oral uconazole was used during the consolidation (8 weeks after the induction treatment) and long-term maintenance treatment period in most patients (74/76, 97.4%). [2][3][4][5] Treatment pro les with the durations of each treatment regimen were reviewed.

Outcome analysis
The scores on mRS was obtained at the time of treatment initiation, at 2 weeks, 10 weeks, and at 6 months. As a primary outcome, a mRS score of > 2 was designated as 6-month poor neurological outcome. Serial follow-up CSF data at 2 weeks (window time of ± 3 days), at 10 weeks (window time of ± or multiple discrete T2/FLAIR hyperintensity lesions with T1 hypointensity in brain parenchyma. 18 Hydrocephalus was de ned as the Evans' index (the ratio of the maximal frontal horn width of lateral ventricle to the transverse inner skull diameter) of ≥ 0.3 ( Fig. 1). To evaluate the serial changes in the MRI parameters, follow-up MRIs at 2 weeks (window time of ± 3 days), at 10 weeks (window time of ± 2 weeks), and at 6 months (window time of 5 months to 10 months) were analyzed, if available.
Statistical analysis SPSS 25.0 (SPSS Inc., Chicago, IL, USA) was used for the statistical analyses. Data were reported as numbers (percentages), means ± standard deviations, or medians [interquartile ranges, IQR]. In univariate analyses, Pearson's chi-square test and Student's t-test were used. To evaluate factors associated with a poor neurological outcome, logistic regression analyses were performed including the parameters with a P < .10 in univariate analyses using a backward elimination method. Age was included in the nal model of every regression analysis. Regression analyses were also performed separately for the subgroup without an encephalitis feature at baseline. Regression analysis for mortality was not performed due to its low frequency. Receiver operating characteristic (ROC) curve were drawn to evaluate the prognostic value of the factors derived from the regression analysis and to designate a cut-off value for predicting a poor outcome. For every analysis, a P value < .05 was considered statistically signi cant. Inter-reader reliability for the MRI parameters were evaluated using Cohen's κ values

Data Availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on request. The cumulative number of patients who achieved CSF antigen clearance at 10 weeks was 50 (65.8%). At 6 months, median mRS score was 2 [0-4] and mortality rate was 15 (19.7%). Detailed clinical, CSF, baseline MRI, treatment, and outcome pro les are described in Table 1. Numbers of patients with magnetic resonance image (MRI) evaluations included at each time points were 76, 40, 49, and 59 at baseline, 2-week, 10-week, and 6-months, respectively..

Results
In the univariate analysis, poor 6-month outcome was associated with baseline encephalitis feature (P<.001), elevated CSF opening pressure (P<.001), the presence of periventricular extension (P=.001), cryptococcoma (P=.022), hydrocephalus (P<.001), and higher scores of GB-, CS-, and total ePVS (all P<.001) in the baseline MRI. Demographics, underlying immune status, treatment pro les were comparable between the groups with poor or good outcomes (Table 1). Subsequently, logistic regression analysis indicated that total ePVS score (Odds ratio [OR]: 5.068, 95% CI: 1.627−15.785 for 1 score increment, P=.005) was independently associated with a poor 6-month outcome. The association of the periventricular lesion extension was marginal (P=.056). When total ePVS score was dichotomized, an ePVS score of ≥5 In the univariate analysis for the factors associated with 6-month mortality, the mortality of 6-month was associated with baseline encephalitis feature (P=.013), elevated CSF opening pressure (P=.032), periventricular lesion extension (P<.001), cryptococcoma (P=.013), and higher scores of GB-, CS-, and total ePVS (all P<.001) in the baseline MRI (Table 3). Due to low frequency (n=15), multivariate analysis for mortality was not performed.
We compared the clinical pro les and the outcomes between the groups with or without baseline encephalitis feature. The group with baseline encephalitis feature was associated with a higher frequency of HIV infection (P=.015), GCS score <15 (P<.001), elevated CSF opening pressure (P<.001), the presence of periventricular extension (P<.001), cryptococcoma (P=.015), and hydrocephalus (P=.021) in the baseline MRI, and lower baseline mRS scores and higher scores of GB-, CS-, and total ePVS, compared to the group without encephalitis feature (all P<.001). The 6-month mRS score was also lower in the subgroup with baseline encephalitis feature (P<.001, Supplemental Table 2).
Risk score for a poor 6-month outcome was calculated by summing up the number of the factors associated with poor outcomes (encephalitis feature, ePVS score ≥5, and periventricular lesion extension), with a score range of 0−3. In ROC curve analysis for the total study population, the risk score predicted a poor 6-month outcome with area under the curve (AUC) of 0.978 (95% CI: 0.950−1.000, P<.001) and 6-month mortality with AUC of 0.836 (95% CI: 0.745−0.927, P<.001 , Fig 2A and 2B). The risk score of 2 predicted a poor 6-month outcome with a sensitivity of 94.1% and a speci city of 95.2%, and 6month mortality with a sensitivity of 93.3% and a speci city of 67.2%. For the subgroup without baseline encephalitis feature, the risk score predicted a poor 6-month outcome with AUC of 0.952 (95% CI: 0.896−1.000, P<.001) and 6-month mortality with AUC of 0.870 (95% CI: 0.764−0.978, P=.003, Fig 2C and  2D). In this subgroup, the risk score of 2 predicted a poor 6-month outcome with a sensitivity of 85.7% and a speci city of 95.0%, and 6-month mortality with a sensitivity of 83.3% and a speci city of 81.2%.
Two-week follow-up MRI evaluation data were available for 40 (59.7%), 10-week follow-up MRI for 49 (64.5%), and 6-month follow-up MRI for 59 (77.6%) patients. The median number of MRI evaluations analyzed per patient was 3 . When the serial changes in the MRI parameters were analyzed in association with the changes in mRS scores, the subgroup with baseline ePVS score ≥5 showed gradual deterioration in the mRS score along with progressive increment of the frequency of cryptococcoma and hydrocephalus (Fig 3A), whereas the subgroup with baseline ePVS score <5 showed gradual improvement in the mRS score and maintained a low frequency of periventricular lesion extension, cryptococcoma, and hydrocephalus in the follow-up MRIs (Fig 3B). A similar trend was observed in the subgroup without baseline encephalitis feature (Fig 3C and 3D, see Fig 4 for representative cases). The pro les of the MRI parameters between the groups evaluated using 1.5-T or 3.0-T MRI machines were comparable (Supplemental Table 4).

Discussion
This study demonstrated clinical and MRI parameters associated with the progression and the poor outcomes of Cryptococcus meningoencephalitis. Along with the baseline encephalitis feature, a high ePVS score and periventricular lesion extension were independently associated with poor outcomes. Especially, presence of two or more risk factors at baseline showed high association with poor neurological outcomes and mortality, indicating that these might serve as prognostic markers. Given that the association was still valid for the subgroups without baseline encephalitis feature, these prognostic factors not only re ect the disease severity but also predict the risk of progression. Additionally, neurological deterioration manifested in brain MRI data as development of cryptococcoma and progression of hydrocephalus, which suggests that these MRI markers can also be used for monitoring the progression of disease.
According to a large-sized prospective study of 501 patients with HIV infection, high fungal burden in the CSF, altered mental status, old age, high peripheral WBC count, low body weight, anemia, and low CSF opening pressure were associated with 10-week mortality. 1 Further, combination treatment of amphotericin and ucytosine at induction period reduced 10-week mortality while uconazole-based induction treatment was associated with higher mortality. 3 However, these studies mainly focused on mortality or fungal clearance in the CSF, while the neurological outcome of Cryptococcus meningoencephalitis has not been investigated in depth. In this regard, this is the rst study to describe the dynamic neurologic course of the disease, and demonstrate the early accessible MRI factors that are useful to predict or monitor the neurological outcomes.
Notably, the outcome predictors in the current study re ect the distinct pathomechanism underlying the progression of Cryptococcus meningoencephalitis and can therefore be related with the previously reported prognostic factors for mortality. High fungal antigen titer and larger fungus capsule size might also contribute to a mechanical stagnation of CSF ow, especially in the peri-venular space which has small diameter. 1,9,10 Therefore, these factors can be correlated with the enlarged PVS in the baseline MRI which re ects the degree of CSF stasis. Altered mental status is a factor constituting a baseline encephalitis feature and is also related to MRI indicators of the parenchymal invasion of the Cryptococcus, such as periventricular lesion extension and cryptococcoma. 1 Increased ICP might also be the consequence of CSF recirculation failure resulting from the wide-spread cryptococcus accumulation over the perivascular space and manifest as the progression of hydrocephalus in MRI. 8,9 In addition to outcome prediction and disease monitoring, the ndings of the current study can also be useful for risk estimation and deciding the treatment strategy. For the patients with ≥ 2 baseline risk factors, higher combination or higher dose of anti-fungal treatment could be used to prevent the patient deterioration. [19][20][21] Additionally, frequent follow-up brain imaging to monitor the progression might be bene cial for the timely detection and early intervention to lower the ICP or other neurological complications.
The current study has several limitations. First, as a retrospective study, the number and the interval of CSF and MRI evaluations, and treatment regimen were heterogeneous and not standardized. Second, the study population with baseline MRI evaluation might bear a potential source of selection bias, as this criteria might exclude patients with severe or unstable baseline clinical status. Different subpopulations included in each time point of the serial MRI data analysis also warrant a careful interpretation of the result. Third, this study included both 1.5-T and 3.0-T machines, although baseline and follow-up images were obtained from the same scanner and the MRI parameter pro les between the groups with different MRI powers were comparable. Fourth, the rate of fungal clearance in CSF, one of the major parameters for determining treatment outcome, was not analyzed in this study. 22 Additionally, careful discrimination ePVS and periventricular lesion extension from aging-related cerebral small-vessel disease is warranted, although their associations with outcomes were signi cant after adjusting for age.  Tables   Table 1. Comparison of the clinical, laboratory, and treatment profiles between the groups with or without poor outcomes.