Cryptococcal Antigen in Serum and Cerebrospinal Fluid for Detecting Cryptococcal Meningitis in Adults Living With Human Immunodeficiency Virus: Systematic Review and Meta-Analysis of Diagnostic Test Accuracy Studies

Abstract

Cryptococcal antigen (CrAg) detection could direct the timely initiation of antifungal therapy. We searched MEDLINE and Embase for studies where CrAg detection in serum/cerebrospinal fluid (CSF) and CSF fungal culture were done on adults living with human immunodeficiency virus (HIV) who had suspected cryptococcal meningitis (CM). With Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2), we evaluated the risk of bias in 11 included studies with 3600 participants, and used a random-effects meta-analysis to obtain summary sensitivity and specificity of serum and CSF CrAg, as well as agreement between CSF CrAg and CSF culture. Summary sensitivity and specificity of serum CrAg were 99.7% (97.4–100) and 94.1% (88.3–98.1), respectively, and summary sensitivity and specificity of CSF CrAg were 98.8% (96.2–99.6) and 99.3% (96.7–99.9), respectively. Agreement between CSF CrAg and CSF culture was 98% (97–99). In adults living with HIV who have CM symptoms, serum CrAg negativity may rule out CM, while positivity should prompt induction antifungal therapy if lumbar puncture is not feasible. In a first episode of CM, CSF CrAg positivity is diagnostic.

In patients with symptoms suspicious of cryptococcal meningitis (CM) associated with human immunodeficiency virus, a positive serum cryptococcal antigen (CrAg) is highly presumptive of culture-confirmed CM and a positive cerebrospinal fluid CrAg is diagnostic of a first episode of CM.

Cryptococcal meningitis (CM), a life-threatening systemic, opportunistic fungal infection, occurs mainly in patients with defective cellular immunity [1, 2]. Consequent to acquired profound immune depression associated with the human immunodeficiency virus (HIV) pandemic [3, 4], there has been a surge in the burden of CM, especially in low- and middle-income countries (LMIC), where more than 90% of CM is HIV-related [5]. In 2014, an estimated 223 100 cases of CM occurred, of which 181 100 were fatal, accounting for about 15% of all-cause HIV-associated mortality [6].

The reference standard for diagnosing CM is the direct identification of the encapsulated yeast Cryptococcus spp. by microscopy of Indian ink–stained preparations of cerebrospinal fluid (CSF) or of yeast colonies cultured from CSF on Sabouraud’s dextrose agar [7, 8]. Consequently, confirmation of the diagnosis of CM requires specialized equipment and clinical and technical expertise, which are not always available in most LMICs. More so, patients’ acceptance of lumbar puncture (LP) in such settings is not guaranteed [9–12]. Therefore, the poor outcomes associated with delayed diagnosis emphasize the need for alternative and reliable methods for timely diagnosis of CM [8].

Cryptococcus spp. is characterized by the presence of a polysaccharide capsule containing cryptococcal antigen (CrAg) and surrounding the cell wall. CrAg is shed into biological milieus during infection and constitutes a biomarker of cryptococcosis. Within the last half-century, growing interest in CrAg detection has resulted in the development of commercial CrAg tests, each based on antibody-antigen interactions, using latex agglutination (LA) assays, enzyme-linked immunosorbent assays (ELISA) [13, 14], or more recently, immunochromatographic lateral flow assays (LFA) [15, 16].

In 2011, the United States Food and Drug Administration approved a point-of-care (POC) immunochromatographic CrAg LFA test [15]. CrAg LFA is affordable (about 2.5 US dollars per test) [17], detects all cryptococcal serotypes, has no constraints on reactant storage or technical expertise, and provides results within 10 minutes [15, 16]. This POC CrAg test is currently recommended by the World Health Organization for routine systematic screening for cryptococcosis in the blood of asymptomatic patients living with HIV who present with less than 100 CD₄⁺ cells/μL, before initiation of antiretroviral therapy (ART) [8].

A recent systematic review and meta-analysis evaluating the clinical utility of routine CrAg screening in asymptomatic patients living with HIV who are without symptoms of central nervous system (CNS) disease revealed that up to a third of patients whose serum was CrAg positive had CM [11]. Such an evaluation in patients with symptoms suggestive of CNS disease could greatly improve the timeliness of clinical decision-making, and hence improve patient outcomes. This systematic review was designed to determine the diagnostic accuracy of CrAg detection in serum and CSF, as well as the prevalence of culture-confirmed CM in adults living with HIV who have symptoms suggestive of CM.

METHODS

This systematic review was registered at PROSPERO (www.crd.york.ac.uk/PROSPERO) as CRD42017069664, conducted according to the Cochrane guidelines [18], and reported following the Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies statement (Supplementary Appendix 1) [19].

Eligibility Criteria

We included randomized trials and both cross-sectional and cohort (prospective and retrospective) studies, irrespective of country, region, continent, or level of care (primary, secondary, or tertiary). In these studies, CrAg detection had to be performed in blood or CSF of adults (age > 18 years) with confirmed HIV serology, presenting with signs and symptoms suggestive of CM, using either LA, ELISA, or LFA. In these patients, the reference standard for establishing the diagnosis of CM was direct yeast identification by microscopy of CSF or of colonies cultured from CSF and stained with India ink, as defined by the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycosis Study Group Consensus Group [7]. Participants with a positive cryptococcal culture and/or Indian ink stain in CSF were considered as having proven CM; those with a negative cryptococcal culture and negative Indian ink stain were considered as not having CM. Studies published in English, French, and Spanish were assessed for inclusion, and those published in other languages were considered for translation into English. Case-control studies were excluded due to their high risk of bias (RoB) [20]. We included published and unpublished studies (eg, conference abstracts).

Search Strategy and Study Selection

A comprehensive search strategy was developed by a medical information specialist (R. S.) and adapted for MEDLINE (via PubMed) and EMBASE. Medical subject headings and other search words included: cryptococcal antigen, cryptococcal surface polysaccharide, cryptococcal meningitis, HIV, AIDS, LA, ELISA, and LFA (see search details in Supplementary Appendix 2). Searches were run from 1981 (year of first HIV case description) through 17 September 2019. We did not use methodological filters, to avoid omitting relevant studies [21]. We also searched for included studies on Google Scholar, for reports that cited these studies. Conference proceedings of the International Conference on Cryptococcus and Cryptococcosis, Conference on Retroviruses and Opportunistic Infections, and International AIDS Society were screened from 2010 onwards.

During the study selection process, 2 review authors (E. T. and J. J. B. R.) independently screened citations for eligibility, first by perusing the title and abstracts. Studies irrelevant to the review question were excluded, and the full texts of relevant articles were retrieved for data extraction. Discrepancies were discussed and arbitrated by a third author (J. F. C.) to achieve consensus.

Data Extraction

Authors E. T. and J. J. B. R. independently extracted data from included studies into a previously piloted data collection form. Studies where more than 1 type of index test or the same index test on both serum and CSF had been evaluated were subdivided by index test and sample type into diagnostic cohorts (see Supplementary Appendix 3 for a detailed list). In this review, results of index tests and reference standards were considered as binary outcomes (positive or negative). Data on semiquantitative CrAg titres or CSF fungal colony unit counts were not extracted, because they were not relevant to the review question.

Information extracted from each study included study characteristics (first author, year of publication, design, setting), participant characteristics (number of participants, mean or median age, proportion of males, proportion of ART-naive participants, mean or median CD₄ counts, survival history), CrAg test characteristics (commercial name, test principle [LA, ELISA, or LFA], types of biological samples used [serum, CSF, or both], total number of samples tested, technical specifications for testing [heat inactivation, pronase pretreatment, and dilutions prior to testing]), reference standard characteristics (commercial name, underlying principle, technical specifications, and component tests, if a composite reference standard was used), data from 2 x 2 contingency tables (numbers of true positives, false positives, true negatives, and false negatives; number of indeterminate results, when reported), and any other information of relevance (eg, funding source).

Quality Assessment

RoB was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool [22]. This 4-domain tool was adapted to suit the review question (Supplementary Appendix 4). For each of the first 3 domains (patient selection, index test, and reference standard), the RoB as well as the applicability to the review question were evaluated and classified as either “low risk,” “high risk,” or “unclear” (if insufficiently reported details). For the fourth domain (flow and timing), only RoB was evaluated.

Statistical Analysis and Data Synthesis

The prevalences of serum and CSF CrAg positivity, as well as of culture-confirmed CM among patients with symptoms suggestive of CNS disease, were estimated by a standard random-effects meta-analysis for proportions using the Freeman-Tukey double arcsine transformation [23]. Then, we fitted bivariate random-effects models to obtain summary estimates of sensitivity and specificity of CrAg in serum and CSF, and their 95% confidence intervals (CI). When the bivariate model could not be fitted because the number of studies was small (less than 4), univariate random-effects models were used to obtain separate summary estimates of sensitivity and specificity. Meta-analysis results were presented by CrAg test (LA, ELISA, or LFA) and sample type (serum, CSF, or both). A random-effects meta-analysis was also used to obtain summary estimates of agreement between CSF CrAg and CSF cultures in the study population: that is, the proportion of tests that gave similar results between CSF CrAg and culture. Heterogeneity was evaluated by inspecting the forest plots and Reciever Operating Characteristics (ROC) space, and by calculating I² statistics (when applicable). We performed a meta-regression to investigate sources of heterogeneity across CrAg test (LA, ELISA, LFA) and sample types (serum vs CSF), by incorporating covariates in the bivariate or univariate model, as appropriate. We also performed sensitivity analyses using only studies judged as having a “low” RoB. The statistical analysis involved the use of Stata 16.0 (Statacorp).

RESULTS

Search Results

The electronic search performed on 17 September 2019 identified 1972 citations (147 duplicates), of which 1794 were excluded based on title and abstract screening (Figure 1). Further assessment of 31 citations resulted in the inclusion of 11 studies [14, 24–33]. Nonelectronic searches did not identify any additional study.

Figure 1.

Flow diagram of the study selection process. Abbreviation: CM, cryptococcal meningitis.

Study Characteristics

The studies included for meta-analyses were published between 1990 and 2018 and conducted in 8 countries (including 6 LMICs) on 3600 adults living with HIV who were clinically suspected of having CM (Table 1). The median number of participants per study was 146 (interquartile range, 99–465), and they were predominantly male (71%). When reported, the median age and CD₄⁺ count were 35.5 years and 27 cells/µL, respectively.

Table 1.

Characteristics of Included Studies

Author, year	Country	Population	CrAg test(s) evaluated	Reference standard considered for the review	Number of participants	Comments
Nelson et al, 1990 [25]	United Kingdom	Consecutive sample of HIV patients presenting with fever and meningism in a hospital setting.	LA system IMMY Diagnostics	Nigrosin, Gram stain, and fungal culture in Sabouraud dextrose agar	828	CrAg on both serum and CSF.
Temstet et al, 1992 [14]	France	Consecutive sample of HIV patients from University hospitals with suspected CM.	LA Meridian Biosciences; Crypto-LA International Biological Labs; and Pastorex LA Sanofi Pasteur Diagnostics	Fungal culture	87	CrAg detection performances of 3 latex agglutination tests on both serum and CSF.
Asawavichienjinda et al, 1999 [24]	Thailand	Consecutive sample of patients living with HIV suspected of CNS infections in a hospital setting.	Pastorex LA Sanofi Pasteur Diagnostics	Indian ink stain and/or culture of CSF	100	Serum CrAg to identify LA cut-off point for the screening and diagnosis of CM.
Boulware et al, 2014 [29]	South Africa and Uganda	Stored samples from 2 cohorts of HIV patients suspected of CM.	LA Meridian Biosciences; LA system IMMY Diagnostics; and LFA IMMY Diagnostics	India ink and/or CSF fungal culture	832	3 index tests were evaluated in CSF. Use of a composite reference standard.
Kabanda et al, 2014 [31]	Uganda	Prospective cohort of patients living with HIV suspected of CM in a hospital setting.	LA Meridian Biosciences; and LFA IMMY Diagnostics	Indian ink and/or fungal culture	112	2 index tests evaluated on CSF. Use of composite reference standard.
Lourens et al, 2014 [33]	South Africa	Consecutive sample of patients living with HIV with signs and/or symptoms of meningitis.	LA Remel Inc. Lenexa USA; and LFA IMMY Diagnostics	CSF fungal culture	465	2 index tests were evaluated in CSF.
Williams et al, 2015 [28]	Uganda	Consecutive sample of HIV patients suspected of CM in a hospital setting.	LFA IMMY Diagnostics	CSF fungal culture	207	Index test evaluation on serum and CSF. Use of a composite reference standard.
Kammalac Ngouana et al, 2015 [32]	Cameroon	Consecutive sample of patients living with HIV suspected of CNS infections in a hospital setting.	LA Fumouze Diagnostics	Indian ink stain and culture	146	Index test evaluated on CSF.
Dharmshale et al, 2016 [30]	India	Sample of HIV patients with signs and symptoms suggestive of meningitis.	LA Meridian Biosciences	Indian ink stain, fungal culture, and polymerase chain reaction	99	Index test evaluation on CSF.
Mpoza et al, 2018 [26]	Uganda	Consecutive sample of patients from 4 cohorts clinically suspected of meningitis.	LFA StrongStep Liming Bio China	CSF fungal culture	282	Evaluation of a new test in both serum and CSF. Use of a composite reference standard.
Ssebambulidde, 2018 [27]	Uganda	Consecutive sample of HIV patients suspected of meningitis.	LFA IMMY diagnostics	CSF fungal culture in Sabouraud dextrose agar	1201	Evaluation of diagnostic performance in serum and CSF.

Data are from 11 studies.

Abbreviations: CM, cryptococcal meningitis; CNS, central nervous system; CrAg, cryptococcal antigen; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; LA, latex agglutination; LFA, lateral flow assay

Across the 11 studies, the following commercial CrAg tests were evaluated: Pastorex (Sanofi Diagnostic Pasteur, France), cryptococcal antigen latex agglutination system (CALAS, Meridian Biosciences), latex agglutination CrAg (IMMY Diagnostics), crypto-latex agglutination (Crypto-LA, International Biological Labs, Cranberry, NJ), cryptococcal latex agglutination (Fumouze, France), CrAg LFA (IMMY Diagnostics), and StrongStep (Liming Bio, Nanjing, Jiangsu, China). Those studies that evaluated more than 1 commercial CrAg tests were subdivided for data extraction for each test: there were 3 tests each in 2 studies [14, 29] and 2 tests in 1 study [31]. These studies were further subdivided by sample type (serum or CSF), yielding a total of 24 diagnostic cohorts (8 on serum and 16 on CSF; Supplementary Appendix 3).

In terms of CrAg detection technologies, 7 of 11 (63.6%) studies evaluated LA (613 participants) [14, 24, 25, 30–32], 4 (36.4%) evaluated LFA (2987 participants) [26–29, 33], and none evaluated ELISA (Table 1). CrAg was assessed in both serum and CSF of the same participants in 5 studies (1846 participants) [14, 25–28], only on serum in 1 study (100 participants) [24], and only in CSF in 5 studies (1654 participants) [29–33].

In all 11 studies, CSF fungal culture was the reference standard for confirming CM. Both culture and direct microscopy of CSF were used in 5 studies (45.5%; 1259 participants) [24, 25, 29, 31, 32]. However, in 4 studies (1433 participants) [26, 28, 29, 31], a composite reference standard comprising culture, India ink staining, CrAg tests, or polymerase chain reaction was considered. No study relied solely on India ink positivity as the reference standard.

Methodological Quality of Included Studies

We determined a high RoB in 1 study (9%) with respect to the patient selection process [14], in 2 (18.2%) studies due to how the index test was performed [14, 33], in 4 studies (36.4%) [26, 28, 29, 31] based on the reference standard (because of composite reference standards), and in 1 study (9%) [29] based on the flow and timing of tests (Supplementary Appendix 5). The 4 studies (36.4%) [26, 28, 29, 31] which used composite reference standards were also judged to be at high risk of applicability concerns (Supplementary Appendix 6). Overall, 5 studies (45.5%) were considered as having a low RoB [24, 25, 27, 30, 32].

Prevalence of CrAg Positivity and Culture-confirmed CM

The summary prevalence of serum CrAg in patients presenting with CNS symptoms was 63% (95% CI, 45–81; I² = 98.7%; Figure 2A). In CSF, the summary prevalence of CrAg was 37% (95% CI, 25–48; I² = 99.2%; Figure 2B). Across studies, the prevalence of culture-confirmed CM ranged between 6% [33] and 63% [29]. The summary prevalence of culture-confirmed CM was 43% (95% CI, 26–59; I² = 99.2%; Figure 3).

Figure 2.

Prevalence of CrAg positivity in (A) serum and (B) CSF in adults living with HIV who have central nervous system symptoms. Abbreviations: CI, confidence interval; CrAg, cryptococcal antigen; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus.

Figure 3.

Prevalence of confirmed CM in adults living with HIV who have central nervous system symptoms. Abbreviations: CI, confidence interval; CM, cryptococcal meningitis; HIV, human immunodeficiency virus.

Diagnostic Accuracy of CrAg

In serum, across 8 diagnostic cohorts of 1946 participants [14, 24–28], the sensitivity of CrAg detection ranged from 83 to 100%, and the specificity ranged from 72 to 100% (Figure 4A). The summary estimates of sensitivity and specificity of serum CrAg for detecting CM were 99.7% (97.4–100) and 94.1% (88.3–98.1), respectively.

Figure 4.

Forest plots of serum (8 cohorts) and CSF (16 cohorts) CrAg sensitivity and specificity for CM diagnosis in adults living with HIV who have central nervous system symptoms. Abbreviations: CI, confidence interval; CM, cryptococcal meningitis; CrAg, cryptococcal antigen; CSF, cerebrospinal fluid; HIV, human immunodeficiency virus; TP, True positive; FP, False positive; FN, False negative; TN, True negative..

In CSF, across 16 diagnostic cohorts of 3500 participants [25–33], the sensitivity of CrAg detection ranged from 80 to 100%, and the specificity ranged from 82 to 100% (Figure 4B). The summary estimates of sensitivity and specificity of CSF CrAg for detecting CM were 98.8% (96.2–99.6) and 99.3% (96.7–99.8), respectively. In these 16 diagnostic cohorts (3500 participants) where CSF CrAg was compared with CSF culture, the summary agreement between CSF CrAg and CSF culture results was 98% (97–99); Table 2.

Table 2.

Agreement Between Cerebrospinal Fluid Cryptococcal Antigen and Cerebrospinal Fluid Culture Results Across Diagnostic Cohorts

	CrAg test	n	n CrAg (+) and culture (+)	n CrAg (−) and culture (+)	n CrAg (+) and culture (−)	n CrAg (−) and culture (−)	Raw agreement between CSF CrAg and CSF culture results, % (95% CI)
Nelson et al, 1990 [25]	LA (IMMY)	69	16	0	0	53	100 (94.8–100)
Temstet et al, 1992 [14]	LA (Pastorex)	77	30	2	0	45	97.4 (90.1–99.7)
	LA (International biological)	41	30	0	2	9	95.1 (83.4–99.4)
	LA (Meridian)	41	30	0	2	9	95.1 (83.4–99.4)
Boulware et al, 2014 [29]	LFA (IMMY)	666	435	3	2	226	99.2 (98.3–99.8)
	LA (IMMY)	749	452	14	0	283	98.1 (96.9–99.0)
	LA (Meridian)	279	176	4	14	85	93.5 (90.0–96.1)
Kabanda et al, 2014 [31]	LFA (IMMY)	112	47	0	0	65	100 (96.8–100)
	LA (Meridian)	112	47	1	0	64	99.1 (95.1–100)
Lourens et al, 2014 [33]	LFA (IMMY)	465	23	3	8	431	97.6 (95.8–98.8)
	LA (Remel Inc.)	465	26	7	0	432	98.4 (96.9–99.4)
Kammalac et al, 2015 [32]	LA (Fumouze)	185	40	10	1	134	94.1 (89.6–97.0)
Williams et al, 2015 [28]	LFA (IMMY)	207	126	0	12	69	94.2 (90.1–97.0)
Dharmshale et al, 2016 [30]	LA (Meridian)	99	42	0	5	52	94.9 (88.6–98.3)
Mpoza et al, 2018 [26]	LFA (StrongStep)	142	101	0	0	41	100 (97.4–100)
Ssebambulidde, 2018 [27]	LFA (IMMY)	1201	671	3	0	527	99.8 (99.3–100)
Random-effects meta-analysis	…	3500	…	…	…	…	98.0% (97.0–99.0)

Data are from 16 cohorts.

Abbreviations: CI, confidence interval; CrAg, cryptococcal antigen; CSF, cerebrospinal fluid; LA, latex agglutination; LFA, lateral flow assay.

Investigations of Heterogeneity

In serum, the LA (5 diagnostic cohorts, 256 participants) summary sensitivity was 100% (99.5–100) and the summary specificity was 96.7% (93.8–98.9); while for LFA (3 diagnostic cohorts, 1690 participants) the summary sensitivity was 97.9% (87.9–100) and the specificity was 89.5% (74.3–98.5). LA showed similar sensitivity in serum as LFA (P = .08) and there was no statistically significant difference in specificity (P = .14; Table 3).

Table 3.

Summary of Diagnostic Accuracy Findings

		Quantity of evidence		Summary estimates
Sample	Test type	Cohorts, n	Participants, n	Sensitivity, % (95% CI)	Specificity, % (95% CI)
Serum	LA	5	256	100 (99.5–100)a	96.7 (93.8–98.9)a
	LFA	3	1690	97.9 (87.9–100)a	89.5 (74.3–98.5)a
	Overall serum CrAg	8	1946	99.7 (97.4–100)b	94.1 (88.3–98.1)b
	P valuec	8	-	.08	.16
CSF	LA	10	1810	97.1 (91.9–99.0)b	99.1 (93.8–99.9)b
	LFA	6	3099	99.5 (97.2–99.9)b	99.5 (94.2–100)b
	Overall CSF CrAg	16	3500	98.8 (96.2–99.6)b	99.3 (96.7–99.9)b
	P valuec	16	-	.07	.54

Abbreviations: CI, confidence interval; CrAg, cryptococcal antigen; CSF, cerebrospinal fluid; LA, latex agglutination; LFA, lateral flow assay.

^aUnivariate random-effects model.

^bBivariate random-effects model.

^cUnivariate logit-normal random-effects meta-regression model.

In CSF, LA (10 diagnostic cohorts, 1810 participants) had a summary sensitivity of 97.1% (91.9–99.0) and a specificity of 99.1% (93.8–99.9) and LFA (6 diagnostic cohorts, 3099 participants) had a summary sensitivity of 99.5% (97.2–99.9) and a specificity of 99.5% (94.2–99.9). Though there was some weak statistical evidence that LFA may have better sensitivity in CSF (P = .07) than LA, their specificities were comparable (P = .54; Table 3).

In 7 diagnostic cohorts comprising 1846 participants [14, 25–28], CrAg detection was performed in both serum and CSF in the same participants, which allowed a direct head-to-head comparison. There was no evidence that sensitivity and specificity differed between CrAg in serum and CrAg in CSF (sensitivity, 99.7% [86.8–100] and 99.9% [97.1–100], respectively [P = .33]; specificity, 95.2% [87.7–98.2] and 99.5% [86–100], respectively [P = .77]; Figure 5; Supplementary Appendix 7.

Figure 5.

Direct head-to-head comparisons of serum and CSF CrAg performed in the same patients (7 cohorts). Circles and diamonds represent serum and CSF CrAg, respectively. The curved lines represent the summary Receiver Operating Characteristics (ROC) curves of sensitivity and specificity. Abbreviations: CrAg, cryptococcal antigen; CSF, cerebrospinal fluid.

Sensitivity Analysis

A sensitivity analysis using only studies judged to be of low RoB confirmed the robustness of results: in serum, the CrAg sensitivity was 100% (100–100) and the specificity was 94% (87.8–98.3), and in CSF, the CrAg sensitivity was 99% (84–99.9) and the specificity was 99.7 (91.9–100).

DISCUSSION

Main Findings

In this systematic review encompassing 11 studies (24 diagnostic cohorts, 3600 participants), we investigated the diagnostic accuracy of CrAg for detecting CM in adults living with HIV and presenting with CNS symptoms. We found that: (1) the prevalence of serum CrAg is about 60%; (2) the prevalence of culture-confirmed CM is about 40%; (3) the sensitivity and specificity of serum CrAg are 99% and 95%, respectively; (4) the sensitivity and specificity of CSF CrAg are 99% and 99%, respectively; and (5) agreement between the results of CSF CrAg and CSF culture is 97%.

Implications for Practice

In routine practice, the utility of a medical test depends on its role in guiding clinical decisions that could impact patient outcomes. Tests used in CM, an extremely severe disease with a high fatality, must be highly sensitive to ensure the timely initiation of induction antifungal therapy [34, 35]. Concomitantly, the high cost of currently recommended induction treatment—as well as potential amphotericin B–related severe adverse events, which are not easy to monitor and manage in LMICs [36, 37]—requires these tests also to be highly specific.

Among patients living with HIV who have CNS symptoms, we found that serum CrAg was highly predictive of confirmed CM [38] and was able to rule out CM when negative. As such, in LMIC settings with a high burden of CM and no facilities for CSF analysis, systematically screening symptomatic patients for serum CrAg should become routine practice. If serum CrAg is positive, empirical inductive antifungal combination therapy should be started, unless the patient was previously known to have had cryptococcal infection. Thus, treatment is not delayed, although an LP is still required in order to measure and manage CSF pressure; this provides the opportunity to confirm the diagnosis, and to confirm whether or not there is an active infection in previously treated cases. Currently, systematic serum CrAg screening is recommended only for ART-naive patients, prior to ART initiation [8]. However, with long-term ART interruption and therapy failure accounting for the majority of CM cases among ART-experienced patients [39, 40], systematic serum CrAg is warranted in all CNS symptomatic patients living with HIV. As such, among those with serum CrAg positivity and a negative CSF CrAg, other causes of CNS infection could be considered.

Relying on India ink staining of CSF and/or culture for confirmation of the diagnosis of CM requires a laboratory setting, trained technicians, and sustainable reagents and equipment. Moreover, India ink staining of CSF, which showed relatively low sensitivity in some studies (as low as 86% [29, 41]), is only positive in the presence of a high fungal count and requires CSF centrifugation for the highest sensitivity. Fungal culture, though reliable, requires viable organisms in CSF and laboratory incubation at 30˚C for several days to ensure fungal growth. This is not always logistically feasible and may delay diagnosis and treatment. In this meta-analysis, a positive agreement between CSF CrAg and reference-standard results was 98%. With such high accuracy, the increasing availability of the LFA CrAg test, and its ease of performance, we suggest that in contexts where there is limited ability to analyze CSF, CSF CrAg is an alternative to conventional fungal culture, especially for first episodes of CM.

Implications for Research

In CrAg-positive patients living with HIV who have asymptomatic underlying CM, a serum CrAg titre of at least 1:160 is associated with culture-confirmed CM [9, 42]. Though we did not investigate serum CrAg titres in this review due to a scarcity of data, their potential role as a biomarker of culture-confirmed CM in CNS symptomatic patients is of high clinical importance, warranting further evaluation.

Limitations

Our review had some limitations. In some of the studies on CSF, diagnostic accuracy might have been overestimated because of composite reference standards. Due to a low number of studies, we had to use univariate random-effects models to separately estimate the sensitivity and specificity of CrAg in serum, as well as in the meta-regression analysis of sources of heterogeneity. A comparison of the performances of LA and LFA was indirect, as only 1 study evaluated both tests in CSF, limiting our ability to draw firm conclusions.

CONCLUSIONS

On average, the accuracy of CrAg detection in the serum and CSF of adults living with HIV who have signs and symptoms suggestive of CM is very high when compared with conventional fungal culture and microscopy following India ink staining. In settings without facilities for CSF analysis or with low LP uptake, CrAg detection in serum may be sufficiently sensitive to rule out CM, and sufficiently specific to start antifungal therapy in cases with a positive result. In settings where LP is feasible but where laboratory equipment is limited, CSF CrAg could replace culture and India ink staining for establishing the diagnosis of first episodes of HIV-associated CM.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. J. F. C. and O. L. contributed equally to this work. E. T. and O. L. designed the review question. E. T., O. L., T. S. H., and J. F. C designed the study and drafted the protocol. R. S. performed the literature search. E. T., J. J. B. R., and J. F. C performed the study selection and data extraction. E. T. and J. F. C. performed the statistical analyses. E. T., O. L., J. F. C., and T. S. H. drafted the manuscript. A. L., T. C. S., T. C., J. P., and P. G. P. proofread the manuscript and edited it for important intellectual content. All authors approved the final manuscript for submission.

Acknowledgments. The authors thank Francoise Dromer and Henry Namme Luma for their constant support and guidance, and Peter Donnelly for proofreading the manuscript.

Financial support. This work was supported by the French National Agency for Human Immunodeficiency Virus and Hepatitis Research (to E. T.’s PhD program; predoctoral bursary number 33/CSS6/AO 2013-1).

Potential conflicts of interest. J. P. has received grants from Merck, Astellas, Minnetronix, Amplyx, and Pfizer; is on advisory boards for Merck, F2G, Scynexis, Amplli Amplyx, Minnetronix, and Matinas; and provides consulting for Scynexis and Amplli, all outside the submitted work. T. S. H. reports nonfinancial support from Immunomycologics, during the conduct of the study. O. L. reports personal fees from Pfizer, Merck, Astellas, and Gilead, outside the submitted work, and has been involved with the development of the Biosynex CryptoPS test, with no associated patent or royalties. 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.

Presented in part: 30^th European Conference on Clinical Microbiology and Infectious Diseases (ECCMID), Paris, France, 18–21 April 2020. Abstract Number 5965.