All of the studies that assessed the diagnostic accuracy of Mn and/or A-Mn antibody detection using immunoenzymatic methods in any patient population, with either prospective or retrospective data collection, were eligible. The tests under evaluation were the commercially available sandwich ELISAs (Platelia™) for detecting Mn and A-Mn antibodies in serum. Studies addressing detection in other fluids are discussed briefly. Studies that used tests other than ELISA were not included in this review to minimise the problem of comparing results obtained with different assays.

The target condition of this review is candidemia and any other form of IC. The following reference standards can be used to define the target condition: autopsy or the criteria of the European Organization for Treatment and Research of Cancer and Mycoses Study group (EORTC/MSG) for defining invasive fungal infections 7 9 . According to these criteria, proven candidiasis is defined as histopathologic, cytopathologic or direct microscopic examination of a specimen obtained by needle aspiration or biopsy from a normally sterile site (other than mucous membranes). The specimen must have evidence of yeast cells or recovery of a yeast by culture of a sample obtained using a sterile procedure (including a freshly placed drain) from a normally sterile site showing a clinical or radiological abnormality consistent with an infectious process.

The definition of probable invasive, that is, hepatosplenic, candidiasis, has changed during the past 6 years. In the first version of EORTC/MSG diagnostic criteria, probable Candida infection was diagnosed in patients with risk factors who had small, peripheral target-like abscesses (that is, bull's-eye lesions) in liver and/or spleen demonstrated by CT, MRI or ultrasound, as well as an elevated serum alkaline phosphatase level; supporting microbiological criteria were not required for probable category 7 . On the contrary, the EORTC/MSG criteria published in 2008 defined disseminated hepatosplenic candidiasis as the presence, in high-risk patients, of characteristic lesions in the liver or spleen after an episode of candidemia within the previous 2 weeks. However, this definition is problematic because blood cultures are frequently negative in these patients, despite repeated attempts to culture a large volume of blood and each lumen of intravenous catheters.

Patients with proven or probable invasive candidiasis defined according to EORTC/MSG criteria were considered as true positive subjects with IC. Subjects without IC were considered as true negatives. Patients with possible IC, that is, the presence of highly suggestive symptoms without microbiological documentation, were not included in the assessment of the performance of the test because of the uncertainty whether they represent true or false positives.

The MEDLINE electronic database was searched with the following terms: Candida, candidiasis, candidemia, antigen, antibody, diagnosis, mannan antigen, anti-mannan antibodies, ELISA, and Platelia™ entered both as text word and MeSH terms if present. The literature search was performed by one of the authors (MM), and the studies published between 1 January 1998 and 9 January 2010 were considered. To identify additional studies, we entered relevant studies selected from the above sources into PubMed and then used the related articles feature and checked the reference lists of all relevant manuscripts. Additionally, review articles and abstracts from the main conferences from the past 5 years (American Society of Hematology Annual Meeting, Interscience Conference on Antimicrobial Agents and Chemotherapy, Infectious Diseases Society of America Annual Meeting, European Congress of Clinical Microbiology and Infectious Diseases and Congress on Trends in Medical Mycology) were screened for any other relevant studies. The following articles were excluded from the review: animal or in vitro studies, articles in languages other than English, case reports and studies that included less than 10 patients (including cases and controls).

Our reference standard was the set of EORTC/MSG criteria. To calculate tests' accuracy and to reflect the categories that are used in clinical practice, we considered the patients with proven and probable IC as having invasive Candida infection (true positives) and patients without candidiasis as the control group (true negatives). This resulted in two-by-two tables: positive or negative Mn antigen, A-Mn antibody or both Mn and A-Mn in each of two groups. The data in the two-by-two tables were used to calculate sensitivity and specificity for each study, while 95% confidence intervals (95% CI) were calculated using the Freeman-Tukey test. For the number of true positives, true negatives, false positives and false negatives that were reported, all of the following were calculated: the diagnostic odds ratios (DORs) with 95% CI. In case of two-by-two tables containing zeroes, 0.5 was added to all counts in the table, which is a commonly used method to calculate an approximation of DOR 22 23 . Median values of sensitivity, specificity and DOR were calculated for all of the available studies.

Individual study results, together with overall pooled results, were presented graphically by plotting the estimates of sensitivity, specificity and DOR (and their respective 95% CIs) in forest plots. The heterogeneity of the studies was investigated using a χ² test. P values of 0.5 or lower were considered statistically significant.

Overall, 556 literature search results were retrieved and screened for relevant information. There were 22 studies that described the use of Mn and A-Mn in various patient populations. Eight studies used tests other than ELISA and were not included in this review. Thus, 14 studies on Mn/A-Mn immunoenzymatic tests were reviewed. The description of these 14 studies is outlined in Table 1 in chronological order.

The number of patients included in the studies varied from 7 to 105, with a median of 25 patients per study. Four studies were performed exclusively in patients with haematological malignancies, three studies were conducted in patients mostly with cancer or haematological malignancy and the remaining seven studies included mostly or exclusively patients from intensive care unit (ICU) or surgery wards (among them was one study from a neonatal ICU). Overall, among 453 case patients described, 123 (27%) had haematological disorders. A control group was included in only 11 of 14 studies and most frequently consisted of patients with similar risk factors, but without IC and sometimes with other documented infections. Four studies included healthy blood donors as control samples.

All of the studies performed included Mn antigen testing, while only 11 of them also searched for A-Mn antibodies. Thus, the sensitivity of the test could be evaluated in all studies (14 for Mn, 10 for A-Mn and combined Mn/A-Mn), but the specificity could be evaluated in only 11 studies that included a control group. In all 11 studies, Mn specificity was evaluated, while the specificity of A-Mn or combined Mn/A-Mn testing was reported in only 7 and 6 papers, respectively.

For Platelia™ Mn antigen and A-Mn antibody testing, the values of 0.5 ng/mL for Mn and 10 arbitrary units (AU)/mL for A-Mn are defined as positive according to the manufacturer, while the values 0.25-0.5 ng/mL for Mn and 5-10 AU/mL for A-Mn were considered indeterminate. Most studies defined a positive result according to the cutoff value recommended by the manufacturer in a single serum sample. In two studies, a result was regarded as positive if in two samples Mn or A-Mn or Mn in one and A-Mn in one were above intermediate cutoff thresholds 24 25 . Additionally, one prospective study differed significantly as far as sampling and threshold values are concerned. In the study by Ellis et al. 26 , the cutoff used for A-Mn was two to four times lower than the others (2.6 vs. 5 or 10), but different criteria were used to define a positive Mn/A-Mn result, that is, two consecutive samples positive for both Mn and A-Mn. Therefore, for A-Mn and Mn/A-Mn testing, the results obtained in this way are reported. For each study, the cutoff values used are reported in Table 1.

The per-patient values of sensitivity and specificity (with 95% CI) with respect to the reference diagnostic method of Mn, A-Mn and combined Mn/A-Mn testing are reported in Table 2. The overall pooled results, together with the results of single studies and their respective weight in meta-analysis, are reported as forest plots in Figures 1, 2 and 3 for Mn, A-Mn and combined Mn/A-Mn testing, respectively.

The median sensitivity of all the studies was 62%, ranging from 31% in the study by Prella et al. 25 to 100% in the study by Sendid et al. that reported seven cases of candidemia due to C. tropicalis 27 . The overall pooled per-patient sensitivity of Mn was evaluated in 14 studies in 453 patients and resulted in 58% sensitivity (95% CI, 53-62). Antibody testing was performed in 10 studies with a total of 284 patients, and the sensitivity of the antibody assay was 59% (95% CI, 54-65), with the median value of 57%, ranging from 44% to 100% 17 26 . The median sensitivity of combined Mn/A-Mn testing (that is, when either Mn- or A-Mn-positive results were considered, except for the study by Ellis et al. 26 , in which a different definition of positivity was used as reported above) was 86%, ranging from 71% to 100%. The pooled overall sensitivity of Mn/A-Mn was 83% (95% CI, 79-87).

Eleven studies included a control group, allowing the assessment of specificity and the calculation of DOR. The specificity of Mn testing was performed in all 11 studies and resulted in a range from 65% in the study by Ellis et al. 26 to 100% in the study reported by Alam et al. 28 , with an overall pooled specificity of 93% (95% CI, 91-94). For A-Mn testing, the specificity was evaluated in seven studies and with a pooled overall result of 83% (95% CI, 79-97), ranging from 38% to 100% (the lowest value was reported in the study by Ellis et al. 26 ). The overall specificity of combined Mn/A-Mn assay was 86% (95% CI, 82-90).

The DORs were calculated for Mn, A-Mn and Mn/A-Mn testing and differed significantly between the studies (Figures 1, 2, 3). Overall, DOR was the highest in the case of combined Mn/A-Mn testing (58; 95% CI, 27-122), followed by Mn testing (18; 95% CI, 12-28) and A-Mn (12; 95% CI, 7-21).

Significant heterogeneity of the studies was detected for the sensitivity of Mn (P < 0.0001) and A-Mn (P = 0.0002); the specificity of Mn, A-Mn and Mn/A-Mn (P < 0.0001 for all); and the DORs of Mn (P = 0.004) and A-Mn (P = 0.01). To reduce the heterogeneity, the same pooled values were calculated with the exclusion of the one study that differed significantly from the others, that is, the study by Ellis et al. 26 . However, even with this study excluded, the heterogeneity remained significant (data not shown). When the studies were grouped by underlying disease (haematological or cancer and ICU or surgery), the heterogeneity disappeared in some of the subgroups, but this effect might be due to the low number of studies included in each subgroup (data not shown). Therefore, the final overall pooled results are reported for all the studies available (Figures 1, 2, 3).

Additionally, per-sample values were reported in five studies in which there were more samples than patients. The per-sample values were not considered significantly different from per-patient data for the study by Alam et al. 28 , where only five patients had two samples instead of one. The overall per-sample sensitivity was lower than the per-patient sensitivity, but the specificity remained high (Table 3).

The sensitivity of both Mn and A-Mn varied for different Candida species, and it was the highest for C. albicans, followed by C. glabrata and C. tropicalis 27 29 . In particular, according to the results reported by Sendid et al. 17 , the sensitivity for the detection of Mn was 58%-70% for infections caused by C. albicans, C. glabrata and C. tropicalis, while it was 25%-30% for infections caused by C. parapsilosis and C. krusei (Table 4). The difference was even more pronounced in the study by Fujita et al. 29 , where the sensitivity for C. albicans was 78% compared to 15% and 0 for C. parapsilosis and C. krusei, respectively. Even though the sensitivity varied among the studies, it was clearly lower in cases of C. parapsilosis and C. krusei, probably because of the lower amount of Mn produced and released by these species 19 30 .

Obviously, the most interesting patients are those with possible IC, where culture is probably not sensitive enough to detect candidemia and where a more sensitive method, such as antigen testing, might prove extremely helpful. The fact that Mn is more sensitive than culture is indirectly proved by the fact that Mn sensitivity in groups of patients with possible candidemia is higher than that in controls but lower than that in culture-positive IC. For example, in the study by White et al. 32 , 5 of 18 patients with possible Candida infection had positive results on Mn testing. Similarly, in 39 patients with clinically suspected IC, Mn and A-Mn were present in 16% and 29% of patients, respectively 28 .

The colonisation with Candida, particularly if multiple sites are colonised, has always been feared to be a potential reason for the lower specificity of Mn or A-Mn testing. Indeed, lower specificities are generally observed in colonised subjects, and Candida colonisation has been reported to result in detectable A-Mn antibody levels in approximately 30% of uninfected patients 16 . Therefore, we reviewed the data on test performance in patients with Candida colonisation.

Overall, four studies included patients with Candida colonisation in their control population. In the study by Verduyn Lunel et al. 31 , 19 of 21 patients and 20 of 30 controls were colonised with Candida (mostly C. albicans) as detected by two consecutive samples from mouthwashes and/or faeces. However, in the logistic regression analysis, neither prior colonization nor superficial Candida infections were associated with the detection of Mn or A-Mn. On the contrary, A-Mn was detected in patients with Candida colonisation in the study by Sendid et al. 16 , in which one of the control groups comprised 23 ICU patients, of whom 19 had Candida colonisation. In this group, only one patient (4%) had positive Mn, but 6 (26%) of 19 had positive A-Mn results 16 . Similarly, in a control group of 10 patients with vaginal candidiasis, 2 patients (20%) had a positive Mn result 28 , while among 15 ICU patients colonised with Candida at two sites or more, only 1 patient (7%) had a positive A-Mn result 33 . Higher rates of false positives were reported in the study by Persat et al. 34 , where 18 of 38 control patients had Candida colonisation, 4 (22%) had positive results for Mn and 8 (44%) had positive results for A-Mn. Finally, in the study by Ellis et al. 26 , where 60% of 74 control patients had Candida colonisation, the specificities of both Mn and A-Mn were significantly lower than reported in other studies. In particular, the specificity was only 21% if two consecutive positive results for either Mn or A-Mn were evaluated 26 . Such a low specificity differs from the results of the other studies and may be related to the fact that a particularly low cutoff value was used for A-Mn testing.

Even though ELISA is licensed to be used in serum only, Verduyn Lunel et al. 35 reported an interesting use of Mn testing in cerebrospinal fluid (CSF) in five patients with Candida meningitis. In fact, four of them tested positive for Mn in CSF. Additionally, a recent study performed in preterm infants found that Mn detection in bronchoalveolar lavage fluid might be useful for early identification and preemptive treatment of candidemia in these patients 36 .