Accuracy of rapid diagnosis of Talaromyces marneffei: A systematic review and meta-analysis

Background To examine the accuracy of Rapid Diagnosis of Talaromyces marneffei (RDTM) in order to improve diagnosis and treatment for clinical measures and reduce the mortality due to associated infections. Methods In this systematic review and meta-analysis, we screened PubMed, Ovid (Cochrane library) and Web of Science, Chinese database CNKI and Wanfang for articles published between 1956 and December, 2017. Data were taken from cross-sectional studies as well as from baseline measurements in longitudinal studies with clinical follow-up. Articles were excluded if they did not contain a cohort with T. marneffei and a control cohort or a cohort with standard fungus culture. Data were extracted by two authors and checked by three for accuracy. For quality assessment, modified QUADAS-2 criteria were used. Results The 26 included diagnostic studies enrolled 5,594 objectives in 632 patients with T. marneffei infections and 2,612 negative controls between 1996 and 2017 in Thailand, Vietnam and China. The total combined sensitivity and specificity of rapid diagnosis of T. marneffei was 0.82 (95% CI: 0.68–0.90) and 0.99 (95% CI: 0.98–1.00). According to the experimental method, the included studies can be divided into three subgroups, including PCR-based, ELISA-based and others. The results showed these three subgroups had a highly pooled specificity of 1.00 (95% CI: 0.99–1.00), 0.99 (0.98–1.00) and 0.97 (95% CI: 0.91–1.00), respectively, while combined sensitivity was 0.84 (95% CI: 0.37–0.98), 0.82 (95% CI: 0.64–0.92) and 0.77 (95% CI: 0.54–0.91), respectively. Conclusions Although serological methods with a high specificity is essential for potential rapid diagnostic, false-negative results can be obtained in the serum samples, there is no suitable rapid serological test to refer to as is the case with TM infection.


Introduction
Talaromyces marneffei (TM), formerly named Penicillium marneffei (also called PSM, or P.M) [1,2], was first discovered in 1956 from the hepatic lesions of a bamboo rat (Rhizomys sinensis) which dying of disseminated mycosis that had been maintained in captivity for experimental infections at the Pasteur Institute of South Vietnam [3]. T. marneffei is endemic to Southeast Asia and South China as an opportunistic infectious disease in immunocompromised individuals [4]. The incidence of systemic T. marneffei infection has grown up rapidly in recent years, in consistent with the increasing incidence of HIV infections [5]. In endemic areas, for instance northern Thailand, it has become the third most common opportunistic infection in patients with AIDS, after tuberculosis and cryptococcosis [6]. The proportion of HIV patients infected with T. marneffei currently is 30% in Thailand, and~10% in southern China [4,5]. In addition, in these areas, about 50,000 HIV positive patients are newly infected by T. marneffei each year, leading to an annual death rate of up to 10% [7]. T. marneffei infected AIDS present serious systemic disease with fever, anemia, weight loss, skin lesion, respiratory signs, generalized lymphadenopathy, and hepatosplenomegaly. Clinical experiences show that treatment using amphotericin B, itraconazole, ketoconazole, and fluconazole alone or combined appropriately during the early stage can effectively combat T. marneffei [8,9].
However, in light of a reported sensitivity ranging from 10 to 100%, whereas specificity usually exceeds 95%, it is unclear which rapid diagnosis technique is preferable. The aim of this review is to investigate whether RDTM is sufficiently specific or sensitive in order to enable early treatment. This study has the potential to provide a rapid diagnostic method for T. marneffei infection, thus enabling early therapeutic management. A study was included in the meta-analysis when it met all following criteria: (1) a study for diagnosis of Talaromyces marneffei; (2) use of the reference standards; (3) use of a control group; (4) provision of true positive, false positive, true negative and false negative directly or indirectly.

Data extraction and quality assessment
Two trained reviewers independently screened the articles and judged research eligibility; disagreements were eliminated through team discussion. Data retrieved from the articles included publication year, country, participants, specimen resource, study design, RDTM method (PCR, ELISA, LA or other rapid serodiagnosis methods), blinding, true positive, false positive, true negative, false negative, sensitivity and specificity data, positive and negative predictive values. Table 1 shows the details of the extracted data. We assessed the methodological quality of each article on case selection, study design, golden standard diagnosis, blinding and so on, according to the Quality Assessment of Diagnostic Accuracy Studies tool (QUADAS2) [46].

Data collection and analysis
Data were extracted either on article or study level when possible to reconstruct the 2 × 2 tables which were used to calculate sensitivity and specificity. The studies included were grouped by type of RDTM. For meta-analyzed the accuracy of RDTM, we selected to perform bivariate randomeffects regression models, as recommended by the Cochrane Diagnostic Test Accuracy Working Group for which I 2 >50%, and combined the sensitivity and specificity estimates. Publication bias was investigated with visual inspection of funnel plots. The bivariate random-effects model to estimate takes into consideration the potential trade-off between sensitivity and specificity by explicitly incorporating this negative correlation in the analysis. Positive and negative of predictive values were computed and diagnostic likelihood ratios (DLR) were directly generated from pooled pair sensitivity and specificity estimates. The data was also used to plot summary receiver-operating characteristic (SROC) curves by establish the true positivity and false positivity (1-specificity) of each study. The closer the curve is to the upper left-hand corner, with the exact area under the curve (AUC) of the SROC curve plot, the better the overall accuracy of the test. Summary sensitivity and specificity estimates for each variate and subgroup were generated, along with 95% CIs (confidence intervals). All analyses were conducted using Stata 12.0 and Revman 5.0.

Study selection and characteristic of included studies
We identified 368 records from PubMed/Ovid/Web of Science and 510 and 244 articles from CNKI and Wanfang, respectively. If these articles, 1046 were excluded because their research contents did not comply with our criteria. We excluded another 61 articles because of specificity/sensitivity was absence or dual publication. Finally, due to some articles evaluated more than one RDTM, 15 full-text articles, including 26 studies were available for the meta-analysis (Fig 1). The 26 studies included diagnostic studies enrolling 632 patients infected with T. marneffei and 2,612 negative controls between 1996 and 2016 in Thailand, Vietnam and China ( Table 1). The methodological quality of the included studies was showed in S1 Fig. There were only 3 studies described the reference test results were interpreted blind to the results of the index test or blinding is dictated by the test order, and the other trails without any blinding method ( Table 1). The funnel plots for publication bias (Fig 2) show symmetry and the Deek test was not significant (p = 0.65). These results indicated that there was no potential for publication bias.

Overall accuracy of rapid diagnosis
To determine the sensitivity and specificity of the studies, the authors totally test 5,594 cases or control samples to identify the T. marneffei infections. The pooled specificity and sensitivity of the rapid diagnosis assay could be a useful tool for prompt diagnosis. Combined sensitivity, specificity, DLR positive and DLR negative were 0.82 (95% CI: 0.68-0.90), 0.99 (95% CI: 0.98-1.00), 38.92 (95% CI: 19.17-79.00) and 0.24 (95% CI: 0.16-0.36), respectively, after analysing Rapid diagnosis of T. marneffei the different types of rapid diagnosis, regardless of the detection methods (Fig 3 and S2 Fig). According to the experimental method, the included trials can be divided into three subgroups, the PCR subgroup, the ELISA-based subgroup and the others.

Molecular diagnosis
Four articles, including seven trials, used the PCR-based experimental technique to detect T. marneffei. In these studies, 108 patients and 414 controls between 2003 and 2016 in Thailand, China and Vietnam were recruited in the studies. As shown in Fig 3, specificity seemed to be more consistent across the studies than sensitivity, with sensitivity estimates ranging from 0.10 to 1.00 and specificity estimates achieving a stable 1.00. Overall, for all RDTM of PCR, combined sensitivity was 0.84 (95% CI: 0.37-0.98), while combined specificity was 1.00 (95% CI: 0.99-1.00).

ELISA based serodiagnosis assays
The ELISA-related experimental methods were also employed to detect the fungus. The methods were analysed in eight articles (14 studies) comprising 320 patients with T. marneffei and 1,873 controls in Thailand and China between 1996 and 2016. A recombinant T. marneffei mannoprotein (Mp1p) for the serodiagnosis of T. marneffei infection was developed to be a common use ELISA-based antibody RDTM in clinical. Evaluation of this test revealed high specificity (100%) and approximately 80% (95% CI, 52%-96%) sensitivity in HIV seropositive patients infected with T. marneffei. Compiled with antibody and antigen tests for the diagnosis of T. marneffei infection had a higher sensitivity of 88%, with a positive predictive value of 100% and a negative predictive value of 96% (Table 1). Pooled sensitivity and specificity were 0.82 (95% CI: 0.64-0.92) and 0.99 (0.98-1.00) (Fig 3), respectively.

Other RDTMs
Along with PCR-and ELISA-related experimental methods, researchers also developed other detection technologies to diagnose T. marneffei. Five different experimental methods were considered in the meta-analysis, including latex agglutination (LA), G test, spectrophotometric detection and Giemsa staining. Pastorex Aspergillus is a latex agglutination test kit using a monoclonal antibody to detect Aspergillus fumigatus galactomannan in serum specimens from patients with aspergillosis. The reagent was used to detect galactomannan in an experimental infection with T. marneffei. However, the titer of antigen detected was lower than that been used effectively for specific detection of T. marneffei. In addition, our meta-analysis revealed that nested PCR was more accurate in detecting T. marneffei in clinical culture samples or fresh tissues. It can improve the detection accuracy when using certain measures, for instance, blood fungus culture for 3 or 4 days, to increase the DNA extraction yield from white blood cells. However, the ELISA-based RDTM showed high specificity and approximately 80% sensitivity, even in non-culture serum. Until now, rapid diagnosis of T. marneffei infections has been a significant challenge, although serodiagnostic assays have been reported as powerful tools for detecting pathogenic fungi.
Molecular tools for the detection of the fungus is based on taxon-specific primers designed from the internally transcribed spacer and ribosomal RNA gene for the fungal strains, which enables early and rapid diagnosis [21,22]. The specificity of T. marneffei primers was tested in a nested PCR from series case reports. This biology approach was nearly 100% successfully to amplify T. marneffei DNA and have been developed to identify T. marneffei from a skin biopsy. Use of new modified PCR-based techniques for the rapid detection of the infection needs to be studied and implemented further. Moreover, most of T. marneffei exist in white blood cells, and the authors used liquid nitrogen grinding crushing method to break cell wall, which cannot break the cell wall completely. Thus, the cell wall broken method can only extract a small amount of DNA and reduce the efficiency of PCR amplification seriously. Compared with nested PCR, the detection sensitivity of traditional PCR that the authors used was lower for the lower amplification sensitivity.
The use of monoclonal antibody based sandwich enzyme-linked immunosorbent assay offers advantages for the highly specific and sensitive detection of T. marneffei antigens in clinical specimens from patients with laboratory confirmed T. marneffei infection. Whereas the IgM from clone 8C3 responses was immobilized onto the wall of microtiter plates were higher against both yeast and mycelial antigens [47]. The antigen in serum or urine was significantly higher against with biotinylated polyclonal rabbit anti-T. marneffei antibody. Especially T. marneffei often occurs in immune dysfunction, it is easily combined with other opportunistic infection, which would greatly affect the ELISA accuracy in the practical use. Also ELISA procedures the antigen is detected clearly in the early phase whereas the antibody only can detect during the intermediate stage, it would be impact on the different phases. In our study, the positivity rate for antigen was significantly higher for antibody as compared to the early diagnosis. When used in combination with antigen and antibody tests, it improves the detection rate substantially.
Several additional methods for detecting circulating T. marneffei antigens have been developed with the same polyclonal antibody and compared with the ELISA for the detection of T. marneffei urinary antigen. A dot blot ELISA and a latex agglutination (LA) test were developed in which urine specimens from 37 patients with culture-proven penicilliosis and 300 controls (52 healthy subjects and 248 hospitalized patients without penicilliosis) were tested. The overall sensitivities of the tests were as follows: dot blot ELISA, 94%; ELISA, 97%; LA test, 100% [31]. However, a potential future application would be to assess responses to antifungal therapy in serial clinical samples testing during treatment and subsequent follow-up without the isolation.
Our studies also had several potential methodological limitations. In particular, only three of the included studies reported blinded assessment of the RDTM assays and the control subjects in most studies were comprised by healthy subjects and hospital patients without T. marneffei. Although RDTM gives a dichotomous yes/no answer, discordance lines may be due to one or the other factor of false-positive results. It is difficult assess the risk of bias, although we did not find any discordance exists in reported accuracy between whether blinded versus unblinded. Moreover, this review could have been affected by publication bias, although we searched several sources and updated our searches, but we may have missed some eligible studies in south Asia for searched articles only published in Chinese and English. The last but not the least, cost-effectiveness analysis would seem to be an absolute necessity to evaluate whether potential benefits offset the added costs of routine use of RDTM. Nevertheless, further development of these tests need to be concerned, a larger number of specimens from patients with T. marneffei and other virulence potential favoring disease need to be tested.