A Multicenter, Longitudinal Cohort Study of Cryptococcosis in Human Immunodeficiency Virus–negative People in the United States

Abstract

Background

Cryptococcosis is increasingly recognized in people without human immunodeficiency virus (HIV).

Methods

A multicenter, prospective cohort study was performed in 25 US centers. Consenting patients were prospectively followed for ≤2 years. Neurological morbidities were assessed with longitudinal event depiction and functional scores (Montreal Cognitive Assessment [MoCA]). Risks of death were analyzed using Cox regression.

Results

One hundred forty-five subjects were enrolled. Most were male (95; 65.5%) and had immunosuppression (120; 82.8%), including solid organ transplant (SOT; 33.8%), autoimmunity (15.9%), and hematologic malignancies (11.7%). Disease involved the central nervous system (CNS) in 71 subjects (49%). Fever was uncommon, documented in 40 (27.8%) subjects, and absence was associated with diagnostic delay (mean: 48.2 vs 16.5 days; P = .007). Abnormal MoCA scores (<26) were predictive of CNS disease; low scores (<22) were associated with poor long-term cognition. Longitudinal event depiction demonstrated frequent complications in people with CNS disease; 25 subjects (35.2%) required >1 lumbar puncture and 8 (11.3%) required ventriculostomies. In multivariable models, older age (>60 years) was associated with higher risks of death (hazard ratio [HR], 2.14; 95% confidence interval [CI], 1.05–4.38; P = .036), and lower risks were noted with underlying hematologic malignancy (HR, 0.29; 95% CI, 0.09–0.98; P = .05) and prior SOT (HR, 0.153; 95% CI, 0.05–0.44; P = .001).

Conclusions

Despite aggressive antifungal therapies, outcomes of CNS cryptococcosis in people without HIV are characterized by substantial long-term neurological sequelae. Studies are needed to understand mechanism(s) of cognitive decline and to enable better treatment algorithms.

cryptococcus, cryptococcosis, transplant, outcomes

In the United States, treatment of human immunodeficiency virus (HIV) has resulted in fewer cases of cryptococcosis [1]. However, infections in people receiving biological immunosuppressive therapies are increasingly recognized [2–11]. We have little data to understand clinical presentations, therapeutic approaches, and outcomes of cryptococcosis in people without HIV [12]. Informing treatment algorithms by host context is important, because experimental models show that cellular inflammatory responses mediate a parabolic balance dictating microbial clearance and destructive inflammatory responses [13]. With pathology driven by immune response, or dynamics of reconstitution, therapeutic strategies may be optimized by understanding the host variables that drive outcomes. For instance, case series show that some HIV-negative patients with cryptococcosis develop a unique postinfectious inflammatory response syndrome associated with alternative macrophage activation [14]. In this context, damaging inflammatory responses are dulled by timely administration of corticosteroids; however, adjuvant steroids are not supported by results of trials in HIV-infected people [7, 14–17].

We performed a multicenter prospective cohort study to further our understanding of presentations and outcomes of cryptococcosis diagnosed in people without HIV infection. The prospective design enabled illustration of longitudinal outcomes, treatments, and functional disability scores. Results demonstrate delayed diagnoses and frequent neurological sequelae, especially in people with impaired cognition at presentation.

METHODS

Twenty-five centers with varied geographic distribution activated the study (Figure 1; Acknowledgments). People who had cryptococcosis without HIV were consented for longitudinal follow-up, under a protocol approved by the John Hopkins University (JHU) Institutional Review Board (protocol NA_00051263). Subjects who had no apparent immunosuppression could also provide consent for referral to the National Institutes of Health (NIH) Clinical Center for a separate study of immunological responses, using an “opt-in” clause in informed consent. A steering committee associated with JHU and NIH met quarterly to monitor progress and operations.

Figure 1.

Location of CINCH sites. Geographical location of sites active in the CINCH study. Each dot represents a site that activated the CINCH protocol during the study period, with the exception of 2 opened in Portland, Oregon. No sites were opened in Hawaii, Alaska, or Puerto Rico. Abbreviation: CINCH, Cryptococcus Infection Network in non-HIV Cohort. Modified from https://www.census.gov/geo/reference/webatlas/regions.html.

Clinical Cohort

Subjects were older than 2 years of age with proven, probable, or possible cryptococcosis and a negative HIV test within 7 months. Cryptococcal disease was defined as “proven” with compatible radiographic abnormalities and culture positivity from sterile fluid or biopsy confirmation; cryptococcosis was defined as “probable” with antigen positivity in blood or cerebral spinal fluid (CSF). Microbial recovery from sputum or bronchoalveolar lavage in the absence of compatible clinical and/or radiographic findings was considered “possible” pulmonary disease [18]. Disease was considered to involve the central nervous system (CNS) with Cryptococcus in histology, antigen, or culture of brain or CSF.

Subjects were followed with voice or in-person contact every 3 months for the first year and every 6 months during the second year. Demographic characteristics, clinical findings, comorbidities, predisposing conditions, CD4 T-cell counts, diagnostics, radiographic descriptors, antifungal management, and outcomes were collected in a Research Electronic Data Capture (REDCap) database [19]. Therapies and neurological outcomes were assessed longitudinally.

The RAND 36-item health survey, measuring self-reported quality of life in 8 domains pertaining to physical, emotional, and social functioning, was performed at baseline and 1 and 2 years after diagnosis [20]. The modified Rankin scale was used to assess neurological function at baseline [21, 22]. The Montreal Cognitive Assessment (MoCA) score was calculated at each visit, measuring short-term memory, visuospatial abilities, executive functioning, concentration, and language and orientation [23].

Blood, urine, and saliva were collected at enrollment and every 3 months during the first 1 year of the study and shipped to a central laboratory for storage. Microbial samples were collected when available. Cryptococcus species identification was confirmed using the MALDI Biotype CA system (Bruker Corporation, Billerica, MA) at JHU.

Statistics

Descriptive statistics were performed to summarize demographic characteristics and presentation. Outcomes were assessed according to whether disease was documented as disseminated to the CNS or not. P values were calculated by Fisher’s exact and Student t or Mann–Whitney U tests for continuous variables. Kaplan–Meier survival curves were drawn for people with or without CNS disease. Cox proportional hazards analysis was performed to determine risks of death. All statistical tests were 2-tailed and significance was set at α = .05. Analyses were performed using R software version 3.4.4 (http://cran.r-project.org/).

RESULTS

Investigators at 20 sites consented 152 subjects between July 2013 and May 2016. Seven subjects died before baseline data collection; 138 subjects confirmed to have Cryptococcus infection were followed longitudinally. Eight subjects were lost to follow-up and 9 withdrew consent; they were censored from outcomes assessments.

Cohort Characteristics at Presentation

Subject demographic characteristics, exposures, and immunosuppressive conditions are summarized in Table 1. The majority of the cohort was male, nonsmoking, resided in the southern United States, and reported significant outdoor exposure associated with vocation and/or hobbies. Most common occupations included construction and landscaping (n = 12; 8.3%) and farming (n = 8; 5.5%). Hobbies included harvesting, hiking, camping, and horseback riding; 14 (9.7%) subjects reported hobbies directly involving birds. The majority had immunosuppression associated with solid organ transplants (SOTs), autoimmune diseases, and hematologic malignancies. A large proportion of subjects were receiving corticosteroids as part of antirejection regimens (27; 18.6%) or autoimmune conditions (16; 11%), the most common including rheumatoid arthritis (n = 4), systemic lupus erythematosus (n = 3), and sarcoidosis (n = 3). Seven of 17 (41.2%) patients with hematologic malignancies were receiving antibody-based therapy targeting B cells or lymphocyte signaling (tyrosine kinase/Janus-associated kinases inhibitors).

Table 1.

Demographic Characteristics of 145 Human Immunodeficiency Virus–negative Patients With Cryptococcosis

	Values
Mean age, y (range)	56.8 (17–89)
Male sex, no. (%)	95 (65.5)
Race, no. (%)
Caucasian	108 (74.5)
Black	24 (16.6)
Asian	6 (4.1)
Other/unknown	7 (4.8)
Smoker, no. (%)	45 (32.1)
Region of residence in the United States,a no. (%)
South	84 (57.9)
Midwest	36 (24.8)
West	16 (11.0)
Northeast	7 (4.8)
Other	2 (2.4)
Significant outdoor exposure, no. (%)
Hobby	120 (78.9)
Vocational	22 (14.5)
Underlying disease,b no. (%)
SOT	49 (33.8)
Autoimmune syndromes	23 (15.9)
Hematologic malignancy	17 (11.7)
Decompensated liver disease	14 (9.7)
Solid tumor	8 (5.6)
Primary immunodeficiency	3 (2.1)
HSCT	2 (1.3)
Miscellaneous	4 (2.8)
None	25 (17.2)
Immunosuppressive medications,c no. (%)
Glucocorticoid therapy	69 (47.6)
Cytotoxic chemotherapy	60 (41.4)
Calcineurin/mTOR inhibitors	42 (29.0)
Antimetabolites	36 (24.8)
Targeted antibodies	10 (6.9)
Other	6 (1.3)

Cohort, N = 145.

Abbreviations: HSCT, hematopoietic stem cell transplant; mTOR, mammalian target of rapamycin; SOT, solid organ transplant.

^a“Other” includes Puerto Rico (n = 1) and Paraguay (n = 1).

^bSOT includes kidney (n = 24), liver (n = 10), heart (n = 8), kidney/pancreas (n = 3), lung (n = 3), and kidney/heart (n = 1). Autoimmune syndromes include systemic lupus erythematosus (n = 3), rheumatoid arthritis (n = 2), eosinophilic syndromes (n = 2), sarcoidosis (n = 2), myasthenia gravis (n = 2), inflammatory colitis (n = 2), autoimmune hepatitis (n = 1), primary biliary cirrhosis (n = 1), multiple sclerosis (n = 1), idiopathic thrombocytopenia (n = 1), polyarteritis nodosa (n = 1), polymyositis (n = 1), Wegener’s disease (n = 1), polyarthropathy (n = 1), psoriasis (n = 1), and unknown (n = 1). Hematologic malignancies (without HSCT) included lymphoma (n = 7), chronic lymphocytic leukemia (n = 4), myelodysplastic syndrome or acute myelogenous leukemia (n = 3), myeloma (n = 2), and acute lymphocytic leukemia (n = 1). Solid tumors included lung (n = 3), breast (n = 2), prostate (n = 1), rectal (n = 1), and liver (n = 1). Primary immunodeficiencies included idiopathic lymphocytopenia (n = 2) and amylogenesis imperfecta (n = 1). HSCT included 1 autologous and 1 allogeneic graft recipient. Miscellaneous disease included diabetes mellitus (n = 2) and steroid receipt after pneumonia presentation (n = 2).

^cSome subjects reported receiving >1 primary regimen. Descriptions of cytotoxic regimens were not collected. Calcineurin/mTOR inhibitors include tacrolimus (n = 37), cyclosporine (n = 2), sirolimus (n = 1), and tacrolimus/sirolimus (n = 2). Antimetabolites include mycophenolate mofetil (n = 30), azathioprine (n = 3), and methotrexate (n = 3). Targeted antibodies include anti-CD20 (n = 3), anti–tumor necrosis factor α (n = 3), and kinase inhibitors (n = 4). “Other” includes cyclophosphamide (n = 4), bortezomib (n = 1), and lenolidomide (n = 1).

Of 145 subjects, 71 had disease involving the CNS and 74 did not. Pulmonary disease was most common; a minority of people with CNS disease had concurrent radiographic abnormalities consistent with underlying lung infection (Table 2). CSF cultures were positive in 50 individuals (50/145; 34.4%), and biopsy revealed proven disease in 3 individuals (2.1%). Forty-five of 51 (88.2%) culture-positive subjects evaluated for CSF cryptococcal antigens had positive results and 3 (6.3%) had negative results. An additional 15 CSF antigens were positive at baseline, with cultures either negative (n = 12) or not performed (n = 3). Serum antigens were evaluated in 37 subjects with culture-confirmed CNS disease; 33 had positive pretreatment serum antigens (89.2%) and 4 (10.8%) had negative antigen results. Sixteen people with positive serum antigens and CNS signs or symptoms lacked CSF culture or antigen confirmation, indicating “probable” disease (Table 2). Fifty people with non-CNS disease had positive serum antigens at diagnosis (67.6%). Antigen titers, when reported, were higher in people with CNS disease (Table 2).

Table 2.

Disease Presentation of 145 Human Immunodeficiency Virus–negative Subjects With Cryptococcosis

	Total Cohort (N = 145)	CNS (n = 71)	Non-CNS (n = 74)	P Valuea
Diagnosis, no. (%)				.003
Proven	93 (64.1)	54 (76)	39 (52.7)
Probable	50 (34.5)	16 (22.5)	34 (45.9)
Possible	2 (1.4)	0	2 (2.7)
Concurrent (non–CNS) site of infection, no. (%)				.01
Lung	93 (64.1)	24 (33.8)	69 (93.2)
Skin	7 (4.8)	4 (5.6)	3 (4.1)
Bloodstream	7 (4.8)	0	7 (4.8)
Bone / joint	2 (1.4)	2 (1.4)	0
Peritoneum	1 (0.7)	0	1 (0.7)
Baseline serum CrAg titer,b no. (%)				.005
1:2–1:8	10 (8.9)	1 (1.4)	9 (12.2)
1:16–1:512	54 (48.2)	29 (40.8)	25 (33.8)
≥1:1024	21 (18.8)	15 (21.1)	6 (8.1)
Clinical presentation, no. (%)
Fever	40 (27.6)	20 (28.2)	20 (27.0)	1
Hypotension	17 (11.7)	6 (1.4)	11 (14.9)	.30
Headache	74 (51.0)	53 (74.6)	20 (27.0)	<.0001
Neurological symptoms	47 (32.4)	30 (42.3)	17 (23.0)	.02
Vomiting	39 (26.9)	26 (36.6)	13 (17.6)	.01
Cough	56 (38.6)	13 (18.3)	43 (58.1)	<.0001
Sputum production	18 (12.4)	3 (4.2)	15 (21.6)	.004
Dyspnea	40 (27.6)	11 (15.5)	29 (39.2)	.002
Skin lesions	11 (7.6)	4 (5.6)	7 (9.5)	.53
Functional scoresc
MoCA score <22, no. (%)	26 (30.6)	17 (47.2)	9 (18.4)	.009
Rankin score, mean (SD)	1.2 (1.8)	1.6 (2.0)	0.9 (1.6)	.07
RAND score, mean (SD)	42.7 (19.2)	34.6 (15.9)	49.4 (19.3)	<.001
Time to diagnosisd
Mean number of days (range)	39.2 (0–730)	29.8 (1–365)	47.3 (0–730)	.26
Delayed diagnosis >1 month, no. (%)	30 (23.6)	13 (21.3)	17 (25.8)	.70

Abbreviations: CNS, central nervous system; CrAg, cryptococcal antigen; MoCA, Montreal Cognitive Assessment.

^aP values obtained from t test (continuous variables) or Fisher’s exact test (categorical variables) for testing differences between CNS and non-CNS groups.

^bTiters not available for 26 subjects with CNS disease and 34 with non-CNS disease.

^cReported results from 100 subjects who underwent Rankin and RAND 36-point health surveys and 86 subjects who underwent MoCA testing at baseline.

^dTwo people with lung disease had incidental findings at surgery, reported as 0 days.

Bronchoalveolar lavage or other respiratory samples were positive in 28 subjects (19.3%). Five isolates were reported (and confirmed) as Cryptococcus gattii, recovered from subjects in the Pacific Northwest (n = 3) and the Southeast (n = 2); 4 of 5 subjects infected with C. gattii had no prior or ongoing underyling disease.

Diagnoses were delayed in both groups, with 101 (65.4%) diagnosed more than 1 week after the onset of symptoms. Almost one-quarter had diagnoses delayed more than 1 month after symptom onset (Table 2). Diagnoses were delayed in subjects with lung disease compared with CNS disease (mean: 47.3 vs 29.8 days; P = .26). Delay was more common in the absence of fever (mean: 48.2 vs 16.5 days; P = .007).

Presenting symptoms are detailed in Table 2. Fever was unusual, occurring in only 40 (27.6%) subjects. Gait disorders (n = 53), headache (n = 32), and hearing deficits (n = 13) were common noncognitive complaints in people with CNS disease; 8 people with CNS disease (11.3%) presented with pulmonary complaints only. Signs and symptoms of CNS infection, including headache, were also present in a sizeable proportion of people who ultimately had no documented CNS disease.

RAND scores illustrated poor subjective quality of life, especially in people with confirmed CNS disease. Baseline neurological disability scores using the Modified Rankin scale did not reveal significant impairment in either group (Table 2). Baseline MoCA scores, which represent a more sensitive indicator of cognitive impairment, were lower in subjects with CNS disease than in others (Table 2 and Figure 2A). Impaired cognition at baseline was moderately predictive of CNS disease; area under the curve (AUC) of the MoCA-generated receiver operating characteristic (ROC) curve estimated 0.71 (Figure 2B). Abnormal scores (MoCA <26) demonstrated high sensitivity (85.9%) but low specificity (36.6%) for predicting CNS disease.

Figure 2.

Baseline distributions and diagnostic utility of MoCA scores in the cohort. A, Histogram depiction of MoCA scores among people with CNS infection and other (non-CNS). Scores >26 are considered “normal.” Shading represents areas of severe (<22) and mild impairment (22–26). B. Receiver operating characteristic curve of baseline MoCA performance as a measure to predict CNS disease. Cutoffs at scores of 22 and 26 are shown (AUC, 0.71). The results of the smoothed curve (shown) did not differ from the empirical curve (not shown). Abbreviations: AUC, area under the curve; CNS, central nervous system; MoCA, Montreal Cognitive Assessment.

Treatment and Morbidities

Therapies are summarized in Table 3. Most people with CNS disease received primary regimens that contained amphotericin B formulations (±5-flucytosine), later switched to azoles. The median duration of primary therapy was 14 days (range: 0–278 days); antifungal therapy was administered for a median of 261 days (range: 0–740 days). Eleven people (15.1%) received “re-induction” therapy with amphotericin after switching to azoles for maintenance. More people without CNS involvement received primary therapy with azoles. The median duration of primary therapy in people with non-CNS disease was 197 days (range: 0–774 days).

Table 3.

Antifungal Regimens and Adjunctive Treatment of 145 Human Immunodeficiency Virus–negative Patients With Cryptococcosis

	Total Cohort (N = 145)	CNS (n = 71)	Non-CNS (n = 74)
Antifungal regimen
Amphotericin B + 5FCa	63 (43.4)	51 (71.8)	12 (16.2)
Azole monotherapy	44 (30.3)	1 (1.4)	43 (58.1)
Amphotericin B + azole	11 (7.6)	8 (11.3)	3 (4.1)
Amphotericin B monotherapy	9 (6.2)	2 (2.8)	7 (9.4)
Azole + 5FC	6 (4.1	3 (4.2)	3 (4.1)
Triple therapyb	3 (2.1)	2 (2.8)	1 (1.3)
5FC monotherapy	1 (0.7)	1 (1.4)	0
Echinocandin monotherapy	1 (0.7)	0	1 (1.3)
None	7 (4.8)	3 (4.2)	4 (5.4)
Adjunctive management
High-dose corticosteroidsc	8 (5.5)	7 (9.9)	1 (1.4)
Serial LP	26 (18.2)	25 (35.2)	1 (1.4)
CSF shunt/ventriculostomy	8 (5.5)	8 (11.3)	0
Surgical lung resection	8 (5.5)	0	8 (10.8)

Abbreviations: CNS, central nervous system; CSF, cerebrospinal fluid; LP, lumbar puncture; 5FC, flucytosine.

^an = 48 received an azole antifungal after a primary regimen of amphotericin B + 5FC.

^bRegimens included amphotericin + 5FC + echinocandin and amphotericin + 5FC + azole (both in CNS) and echinocandin + 5FC + azole (non-CNS).

^cExcludes low-dose steroids (<1 mg/kg/d prednisone equivalent) and that received for treatment of an underlying condition. One subject with non-CNS disease received steroids immediately for hemodynamic instability.

Serum antigens were tested more than once in 68 of 113 (60.1%) subjects who had at least 1 positive result. Twenty-seven subjects had their last positive result recorded at diagnosis (day 0) or within 2 weeks of therapy (Figure 3). Many had positive serum antigens 3 months or longer after diagnosis.

Figure 3.

Timing of last positive serum antigen. Frequency distribution is depicted from 68 subjects who had more than 1 positive antigen result recorded, according to day of diagnosis (day 0), in months.

Figure 4A depicts neurological complications and treatments of CNS disease. Twenty-five (35.2%) required more than 1 lumbar puncture (LP) to manage high CNS pressures (Table 3); of these, 20 (28%) required 2 or more LPs within the first 2 weeks of therapy. Nine (12.8%) developed hydrocephalus and 4 (5.6%) developed cranial nerve palsies. Eight (11.3%) received ventriculostomies and/or shunts. Seven (9.9%) with CNS disease received high-dose corticosteroids to manage CSF inflammation. Thirty-three required LPs and shunting more than 2 weeks after diagnosis, and 24 required these procedures more than 4 weeks after diagnosis (Figure 4A).

Figure 4.

Longitudinal assessments of neurological complications in people with CNS disease. A, Event depiction for management of CNS pressure complications. Each subject is represented by a line, with day 0 as day of diagnosis. Black dots represent “alive” but censored for follow-up. B. Dynamics of MoCA assessments are shown for people with documented CNS disease who had tests obtained for at least 1 year, according to severity of baseline dysfunction. Trendlines are shown in bold. Abbreviations: CHS, central nervous system; CSF, cerebrospinal fluid; LP, lumbar puncture; MoCA, Montreal Cognitive Assessment.

Prospective evaluation of the cohort enabled capture of MoCA scores to characterize cognitive functioning. Figure 4B plots serial MoCA scores for the subjects with CNS disease who had data captured for at least 1 year after diagnosis. Among those who presented with very low scores (<22), only 1 of 5 normalized (>26), several improved, and 2 subsequently declined. All subjects who had low MoCA scores required multiple LPs to manage intracranial pressures and/or required another invasive procedure (eg, shunts). People with mild or no impairment typically maintained or improved cognition (Figure 4B).

Mortality and Risks

Kaplan-Meier curves demonstrated worse survival with CNS disease (log-rank P value = 0.16; Figure 5). Among the entire cohort, univariate analysis demonstrated higher risks of death in people with older age (>60 years; hazard ratio [HR], 1.6; 95% confidence interval [CI], 0.82–3.15; P = .17), high pretreatment serum antigen titers (≥1:128; HR, 2.25; 95% CI, 0.91–5.52; P = .078), and CNS disease (HR, 1.8; 95% CI, 0.91–3.62; P = .09). People who had received a prior SOT (HR, 0.22; 95% CI, 0.08–0.61; P = .004) had low risks of death and underlying hematologic malignancy was less significant (HR, 0.44; 95% CI, 0.1–1.8; P = .25) in univariate testing. In multivariable models, age remained a significant predictor of death (HR, 2.1; 95% CI, 1.05–4.37; P = .036), and hematologic malignancy (HR, 0.29; 95% CI, 0.09–0.98; P = .05) and SOT (HR, 0.153; 95% CI, 0.05–0.44; P = .001) were associated with lower risks of death. The impact of delayed diagnosis was explored in additional models, but the variable was collinear with site of disease. When modeled within the limited cohort of subjects with CNS disease, older age remained the only significant predictor of death (HR, 3.85; 95% CI, 1.2–12.2; P = .02).

Figure 5.

Kaplan–Meier 1-year survival curves. Survival is shown after diagnosis of CNS and non-CNS disease (log-rank P value = .0086). Abbreviation: CNS, central nervous system.

DISCUSSION

Cryptococcosis has decreased in incidence in HIV-infected patients, but disease and related mortality are increasing in other immunosuppressed populations [2–11, 24]. We undertook this longitudinal cohort study as an NIH intramural–extramural effort, positioning 25 centers to identify cases and report longitudinal outcomes. Longitudinal assessment enabled depiction of clinical presentations and outcomes over time, incorporating functional assessments that were not clinical practice. Results demonstrate atypical presentations and delayed diagnoses. People with poor neurological functioning at presentation had long-term cognitive impairment.

Underlying diseases in the cohort generally reflect other population-based analyses [9]. Results reflect an increasing trend in targeted biological therapies (anti–tumor necrosis factor α and interleukin 6) [7, 25–29]. Patients with hematologic malignancies had frequently received targeted monoclonal antibodies and small-molecule signaling inhibitors, anti-CD20 (rituximab), ruxolitinib [30, 31], and ibrutinib [32], the latter also associated with aspergillosis [33]. A sizeable proportion of people had decompensated liver disease as a sole risk, consistent with other reports [2], and potentially indicative of complex immunodeficiency [4, 34, 35].

This study used a novel mechanism to refer patients for enhanced immunologic investigations at the NIH, using an “opt-in” clause for assessment for transfer. These patients contributed to detailed reports on neuroinflammatory outcomes [14, 15] and immunologic predispositions [36].

Five cases of C. gattii were identified in this cohort, with geographic distribution in the Northwest and Southeast, and lack of apparent underlying disease [37]. Because both microbiology and underlying diseases are representative of the centers participating in the cohort, the design limits conclusions on wider US epidemiology.

Many subjects related vocational or recreational activities that may predict environmental exposure. It is difficult to assess significance, because they are subject to recall bias and the natural history of infection is not well understood. Prior serologic studies have shown that many in certain geographic areas are infected with Cryptococcus neoformans early in life, and a substantial proportion of cases that are recognized after SOT reflect reactivation of latent infection [38, 39].

Disease was diagnosed late in this cohort, consistent with atypical, afebrile presentation. Increased awareness of both pulmonary and CNS manifestations in the diverse group of people with immunocompromising conditions may hasten diagnosis.

Compared with HIV-associated cryptococcosis, CNS disease is proportionately less common, occurring in ~50% in this cohort, but associated with high, and recurrent, neurological morbidities, including shunt requirements, serial LPs, and pressure-related complications. Similarly, high neurological morbidities have been described in case series [16, 24], typically driven by pathological inflammatory responses [40]. Paradoxical worsening in HIV-negative patients is associated with defective alternative (M2) macrophage activation, proinflammatory cytokine release, and intrathecal T-cell activation, with resultant axonal damage [14, 15]. In some studies, imaging revealed choroid plexitis and ependymitis [14, 15]. Effective ventricular drainage and immunosuppressive therapy, including high-dose corticosteroids, can abate paradoxical worsening associated with dysregulated inflammatory responses [7, 14–17, 41–43]. Substudies are currently being performed to characterize neuroinflammation and imaging in this CNS cohort; future efforts will be necessary to define the optimal strategy to predict, and minimize, pathological CNS inflammation.

In this study, prospective enrollment with active longitudinal follow-up enabled the capture of data over time, providing an understanding of recurrent events that go beyond time-to-event analyses. Incorporation of cognitive assessments provided evidence of sustained morbidity in people with CNS disease. Low MoCA scores in this group predicted poor long-term cognitive functioning and need for recurrent procedures. The MoCA is a well-recognized instrument that has enabled detection of functionally relevant cognitive impairment associated with age-related dementia and neurocognitive diseases [23, 44–47]. Use of the MoCA score in CNS disease may thus assist further study of therapeutic strategies to minimize long-term cognitive morbidity. It is likely that scores reflect infection-related morbidity; however, underlying disease(s) and other host risks may have contributed to cognitive decline.

In this heterogeneous cohort of people without HIV infection, survival was, not surprisingly, lowest in people with CNS disease. Low risks of death were noted among SOT recipients and people with hematologic malignancy, perhaps due to a more rapid diagnosis or modulation of pharmacologic immunosuppression [3, 48]. While attributable mortality is difficult to adjudicate in settings of complex illness, older age as a predictor for death had been reported [3]. Diagnosis was delayed in a substantial number of people, but we were unable to determine the significance on outcome because of collinearity; delay was common in people with pulmonary disease. Analysis of immunologic markers and larger cohort studies may provide actionable data to minimize mortality within a specific host context.

Cryptococcal serum titers were higher in people with CNS disease, as reported in recent HIV-infected and noninfected cohorts [49, 50]. Sequential serum assays were positive for long durations in a substantial proportion of patients. Because these were reported only when clinically obtained, we did not calculate time to clearance; more studies are warranted to understand dynamics of clearance and prognostic utility in different disease states.

Management recommendations for cryptococcosis are generated largely from randomized trials performed in HIV-infected people [12]. Here, primary antifungal regimens typically followed recommendations for both CNS and non-CNS disease; however, people with CNS disease received multiple inductions and long-term maintenance therapy as well as adjunctive corticosteroid therapy for the control of cerebral edema or possible neuroinflammation. This approach is supported by small observational studies [41, 42] and will require further evaluation of safety and efficacy in people without HIV infection. Because neuroinflammation is driven by different dynamics of infection and immune responses, observations from people with HIV may not be generalizable. Caution is warranted in generalizing recommendations pertaining to immunomodulators, including those focused on enhancing and suppressing T-cell immunity [12]. While studying “rare” infections in host context is difficult, the use of cohort designs, functional assessments, and biomarkers may assist in the development of more-informed therapeutic approaches.

The cohort design also has limitations. While it enables assessment of long-term outcomes in a limited cohort of people, it cannot generate estimates of prevalence or geographic distribution, because this is also influenced by site selection. As with other noninterventional studies, conclusions about clinical strategies are limited by bias introduced by clinical practices.

The findings of this study substantiate cases and series suggesting increased recognition of disease in people with complex immunosuppression, including that associated with transplant and use of targeted biological therapies. We observed a high proportion of people with CNS-pressure complications and chronic cognitive impairment despite appropriate antifungal therapies, consistent with reports documenting atypical neuroinflammatory syndromes that influence poor outcomes. More studies will be necessary to develop prevention strategies and to optimize diagnostic and treatment approaches for cryptoccocal disease in people who have complex biological immunosuppression.

Notes

The Cryptococcus Infection Network Cohort study working group members . Jennifer Lyons, Brigham and Women’s Hospital; Adarsh Bhimraj, Cleveland Clinic; Robin Trotman, Cox Health; John Perfect, Duke University; G. Marshall Lyon, Emory University; Jose Vazquez, Georgia Regents; Julia Piwoz, Hackensack University; Kieren Marr, Johns Hopkins; Steven Spindel, Kaiser Northwest; Dannah Wray, Medical University of South Carolina; John Bennett, National Institutes of Health; Julia Garcia-Diaz, Ochsner Medical Institutions; Lynne Strasfeld, Adults, Oregon Health Sciences University; Dawn Nolt, Doernbecher Children’s Hospital; Aruna Subramanian, Stanford University; Peter Pappas, University of Alabama; Joanna Schaenman, University of California, Los Angeles; Randy Taplitz, University of California, San Diego; Marisa Miceli, University of Michigan; Samuel A. Lee, University of New Mexico Veterans Healthcare System; Hong Nguyen, University of Pittsburgh; Pia Pannaraj, Children’s Hospital of Los Angeles; Rodrigo Hasbun, University of Texas, Houston; Ajit Limaye, University of Washington; and William Powderly and Andrej Spec, Washington University, St. Louis.

Financial support. This work was supported by the National Institutes of Health (intramural–extramural grant number U01 AI109657 and extramural grant numbers AI001123-01 and AI001124-01; to A. P. ’s institution).

Potential conflicts of interest. K. A. M. reports personal fees from Cidara and Merck outside the submitted work; A. P. reports being employed as Director at Merck & Co., Inc., since April 2017. A. S. reports grants and personal fees from Astellas Pharma, Inc.; grants from Scynexis; grants from IMMY; grants and personal fees from Mayne Pharma; and grants and personal fees from Minnetronix outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.