We conducted a systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [7]. We identified English language studies, without temporal restriction or study type restriction, in Scopus and PubMed databases. Studies published from inception up until 31st October 2022 were included in the search criteria. Our search duration of the two databases lasted from 5th September 2022 until 31st October 2022. We used the following free text search terms on PubMed: “tuberculosis” and its MeSH terms, AND “adenosine deaminase” and its MeSH terms, AND “Lactate Dehydrogenase” and its MeSH terms. We modified the search terms to suit the Scopus search algorithm. This systematic review was registered in the PROSPERO database under CRD42022359917.

We included studies of adult patients having both TPE and PPE, which reported on the diagnostic performance of pleural fluid LDH/ADA ratio. Observational studies (cross-sectional, cohort, and case-control studies) were included. The study titles and abstracts were screened by two review authors to determine if the study fulfilled eligibility criteria. The two review authors independently evaluated the risk of bias in the eligible studies based on the Quality Assessment of Diagnostic Accuracy Studies Tool 2 (QUADAS-2) [8]. Any disagreement between the two review authors over the eligibility of any studies was resolved through discussion with a third review author. We reviewed full-text publications to identify studies for inclusion in the analysis after the reviewers agreed that the cited publication met the eligibility criteria, and all the disagreements were resolved. Information regarding the study design, methodology, participant demographics, baseline characteristics, and measures of effects were extracted from the studies, and data was recorded in an Excel spreadsheet (Microsoft, Redmond, WA, USA).

The main outcomes were summary diagnostic accuracy (sensitivity, specificity, positive and negative likelihood ratios, and PPV and NPV) of pleural fluid LDH/ADA ratio in differentiating TPE from PPE.

We derived summary diagnostic accuracy estimates from all the studies. MetaDiSc version 1.4 (https://meta-disc.software.informer.com/1.4) was used to analyze and derive a summary estimate of the diagnostic accuracy of the LDH/ADA ratio [9]. Pooling estimates by random effects model were derived using the DerSimonian-Laird method. The chi-square test was used to determine significant heterogeneity with alpha set at 0.1, whereby p-values of <0.1 indicated a significant difference with the null. I² was used to estimate and quantify the degree of heterogeneity. The I² value ranges from 0% to 100%, with 75% or more expressing a high degree of heterogeneity [10].

No ethical approval was sought because data was retrieved from previous published studies in which informed consent was obtained by primary investigators.

Based on the electronic database search, 110 studies were found, eight of which were excluded as duplicates (Figure 1). Fifty-nine studies were excluded based on their titles and abstracts. The remaining 43 studies were subjected to full-text assessment. One study was not retrievable. We excluded 35 of these studies because they did not mention the LDH/ADA ratio (n=24), of the absence of TPE or (PPE) (n=12), and the absence of accuracy testing (n=1). Finally, a total of five studies were deemed eligible for the systematic review, encompassing a total of 2,407 patients (Figure 1). These studies were published between 2017 and 2022. All of the five studies were observational, single-center studies, and only one was a prospective study [6]. The studies were conducted in South Africa, New Zealand, Taiwan, and China [4-6,11,12]. There were no randomized control trials. Of note, one of the studies only recruited patients with ADA value >40 U/L [12]. Criteria for TPE were consistent in all but one study, which only included TB culture-positive samples [11]. Meanwhile, criteria for PPE differed between studies; one study did not explicitly state the PPE criteria [6]. Variation in PPE criteria included a documented clinical diagnosis in the absence of alternative causes [11], positive pleural or sputum culture with preceding clinical pulmonary infection [12], and clinical pulmonary infection in the absence of TPE [4,5]. The characteristics of the included studies and their patients are presented in Table 1.

Varying ADA assays, such as the Giusti-Galanti method and the Diazyme method were used [6,11]. ADA assays used in two studies were not detailed.

Cited LDH/ADA ratio cutoff values for TPE differed in each study, ranging from <7.5 (sensitivity 64%, specificity 96%) in a South African cohort [6], to <25 (sensitivity 97%, specificity 62% in a South African cohort [6]; sensitivity 100%, specificity 61.6% in a New Zealand cohort [11]) PPV was much lower (ranging from 8.5 to 57.3 at cutoff values of <15 to <25, with ADA >15 or >30) but NPV was high (99.5 to 100) in the New Zealand cohort, where TB prevalence is very low [11]. Specific cutoff values for PPE were >14.5 (sensitivity 79.9%, specificity 78.5%) in the Taiwanese cohort [12]. LDH/ADA ratio was found to be superior to ADA alone in differentiating TPE from alternative diagnoses (area under the receiver operating characteristic curve [ROC], 0.92 [p<0.001] vs. 0.88 [p<0.001]) [6,12], while other studies demonstrated noninferiority [11]. However, the LDH/ADA ratio had less PPV and NPV than ADA alone in a setting of lymphocyte-predominant pleural effusion [6]. In studies that employed multiple cutoff values, PPV and sensitivity increased as the LDH/ADA ratio cutoff values decreased, while NPV and specificity decreased accordingly. LDH/ADA ratio is significantly lower in TPE than in PPE [5]. Detailed information can be found in Table 2.

When compared to LDH/ADA ratio alone, no difference was found with the combination of LDH/ADA ratio and lymphocyte-predominance, with the authors proposing that a ratio <7.5 precludes the need for fluid lymphocyte testing [6]. Combined LDH/ADA ratio and elevated fluid ADA increased specificity and PPV [11], while a combination of LDH/ADA, ADA, serum albumin, serum LDH, and pleural fluid LDH/total protein provided a 100% sensitivity and 98.7% specificity [4]. More details on this can be found in Table 2. Median values for ADA, LDH, and LDH/ADA ratios are presented in Table 3.

Selection bias was high for all the studies identified, as convenience sampling was used in all five studies. None of the studies were randomized. A summary table of risk of bias assessments is described in detail in Supplementary Table S1.

Patient selection was conducted via convenience sampling from a database or registry [4,6,12], lab samples [11], or consecutive patients presenting to a healthcare center [5]. In all studies, none of the sampling methods were randomized. Inclusion and exclusion criteria varied greatly between the studies. One study included all exudative pleural effusions [4], one study included both exudative and transudative effusions [11], one study included only effusions with ADA levels above 40 [12], and three studies were conducted on patients with an established diagnosis. Most studies excluded subjects with incomplete clinical data. Two studies did not define their exclusion criteria [4,5]. One study excluded undiagnosed exudates [6]. The risk of bias for patient selection was high in all studies and may have led to underestimation of diagnostic accuracy.

In all studies, the LDH/ADA ratio was interpreted with knowledge of the reference standards, as they assessed patients with a known diagnosis based on reference standard to compare their respective LDH/ADA ratios [4-6,11,12]. The LDH/ADA ratio was assessed retrospectively after the diagnosis was made; hence, there was a high risk of bias. Of note, the studies had varying analysis groups. One study subclassified PPE into uncomplicated and complicated groups and included two categories that were not found in other studies— connective tissue disease-related effusion and chronic, nonspecific pleuritis [4]. A few studies had preset LDH/ADA ratio thresholds. The New Zealand cohort used a ratio <25 and <15; these values were determined based on historical data from an internal audit conducted at the study site [11], but the threshold was determined after data collection. ROC curves were used to identify the optimal cutoff points in other studies [4-6,12]. Two studies excluded patients with an ADA level of below 40 [12]. This may have led to overestimated test performance.

PPE was explicitly defined in all but one study [6]. Meanwhile, pleural TB was consistently defined in all studies. The authors have low concerns regarding both whether the reference standard is likely to correctly classify the target condition and whether the target condition will not match the review question. This is because, in all studies, LDH/ADA ratios were calculated and interpreted retrospectively in patients with the already known diagnoses; hence, there is no risk of bias.

There were no concerns regarding verification bias, as all studies retrieved ADA and LDH samples in the same pleural fluid sampling time frame. In one study only 68.7% of the pleural fluid samples were cultured for TB, as the practice in the center was only to culture those with ADA >15 [11]. A few studies had significant exclusion of patients. In one study, 39 out of 267 patients were excluded from analysis due to missing data (n=24) or undiagnosed effusion (n=15) [6]. In another study, 42 out of 353 patients were excluded from analysis due to missing data [12]. Some studies had no described exclusions while two studies mentioned excluding patients who did not fulfill inclusion criteria but did not mention how many patients were excluded [4]. A summary of all studies is presented in Table 1.