Predictive Ability of Sonographic Characteristics Alone to Diagnose Tuberculous Pleural Effusions in a Resource‐Limited Setting: A Cross‐Sectional Pilot Study in Zambia Authors: Lukonde Zimba, Salomé‐Joëlle Gass, Mpundu C Makasa, Matthew S Haldeman Published: 2026 ABSTRACT ABSTRACTBackgroundTuberculous pleural effusion (TPE), the second most common form of extra‐pulmonary tuberculosis (TB), remains difficult to diagnose due to low‐yield, invasive and often unavailable diagnostics in resource‐limited settings. This study aimed to evaluate the diagnostic accuracy of sonographic characteristics alone to predict TB aetiology of pleural effusions using point‐of‐care ultrasound (POCUS).MethodsThis cross‐sectional study included 45 participants in Zambia who underwent ultrasound scanning and diagnostic thoracentesis. Key sonographic characteristics of pleural effusions—presence of septations, presence of plankton sign and unilateral location—were evaluated for their association with TB, using pleural fluid adenosine deaminase (ADA) ≥ 40 IU/L, combined with Xpert MTB/RIF and lipoarabinomannan, as the gold standard. Sensitivity, specificity, positive predictive values and negative predictive values were calculated for each characteristic alone and in combination and multivariate logistic regression was used to determine statistical significance. Interrater reliability between clinician and expert image interpretation was calculated using a kappa statistic.ResultsSignificant associations were found between TB positivity and unilateral location (either side) (p = 0.037), right‐sided location (p = 0.043) and any two POCUS signs (p = 0.044). Diagnostic performance varied across the three POCUS signs, with the presence of any two POCUS signs demonstrating the highest sensitivity (90.9%) but moderate specificity (41.7%) and the presence of all 3 POCUS signs showing strong specificity (83.3%) but poor sensitivity (45.5%). Agreement also varied, with the lowest kappa score observed for the plankton sign and the highest for unilateral location.ConclusionsMultiple sonographic characteristics, including unilateral location (either side), unilateral right‐sided location and any two positive signs were significantly associated with TB aetiology. Limitations included small sample size, potential selection bias, imperfect gold standard options, and inability to fully assess potential confounders. With additional study, these findings may be used to inform development of a non‐invasive tool to predict TB aetiology of pleural effusions—and expedite initiation of anti‐TB therapy—in high TB settings globally.Keywords: pleural effusion, point‐of‐care ultrasound, tuberculosis, Zambia Abbreviations Adenosine deaminase Acid‐fast bacilli Extra‐pulmonary tuberculosis Focused Assessment with Sonography for HIV‐associated Tuberculosis Lipoarabinomannan Multi‐drug resistant/rifampin‐resistant tuberculosis Negative predictive value Point‐of‐care ultrasound Positive predictive value Resource‐limited settings Tuberculosis Tuberculous pleural effusions World Health Organization Xpert Mycobacterium tuberculosis /Rifampicin test Introduction Despite significant progress in reducing tuberculosis (TB) morbidity and mortality globally, TB remains an urgent global health concern, ranking as the second leading infectious cause of death [1]. According to the World Health Organization (WHO), there were 8,200,000 new TB cases and 1,250,000 related deaths in 2023 [2]. Zambia is among the countries most affected, with an incidence rate of 283 per 100,000 in 2023 and is designated by WHO as high‐burden for TB, HIV‐associated TB and multi‐drug resistant/rifampin‐resistant TB (MDR/RR‐TB) [3]. Despite this high burden, the most recent TB prevalence survey in Zambia found that over 90% of individuals found to be TB positive had gone undiagnosed, including some who had previously sought medical care [4]. This diagnostic gap reflects the challenges faced by clinicians in resource‐limited settings (RLS), where limited diagnostic options may even necessitate the initiation of empiric TB treatment without laboratory confirmation, despite its prolonged duration and potential adverse effects [5]. Therefore, improved access to effective, point‐of‐care diagnostic tools is essential to enhance TB case detection and reduce missed or delayed diagnoses. Pleural TB, which often manifests as tuberculous pleural effusions (TPE), is the second most common manifestation of extra‐pulmonary TB (EPTB) and is particularly challenging to diagnose given the limitations of current diagnostic options. Laboratory assessment of pleural fluid—typically obtained via invasive procedures such as thoracentesis or chest tube placement—has low diagnostic yield: acid‐fast bacilli (AFB) staining is positive in fewer than 20% of cases and the sensitivity of Xpert MTB/RIF (GeneXpert) is only 46.4% [6, 7]. While adenosine deaminase (ADA) testing offers high sensitivity and specificity for TPE diagnosis [4, 8], it remains inaccessible in many RLS, including Zambia [8]. Point‐of‐care ultrasound (POCUS) offers a promising solution to this diagnostic gap. It has already demonstrated significant diagnostic value in high‐resource settings for a variety of applications and its potential impact in RLS could be even more substantial [9, 10]. Given its portability, affordability, excellent safety profile, low maintenance requirements and ability to operate independent of consumable resources, POCUS is an ideal imaging modality in RLS [9, 10, 11]. Additionally, POCUS has shown great promise in applications unique to these settings, such as the FASH (Focused Assessment with Sonography for HIV‐associated TB) scan protocol, which has been effectively used to detect signs of EPTB in individuals with HIV [12, 13]. However, data describing sonographic findings of TPE are limited. Preliminary studies show that TPE are almost always exudative [14, 15], usually unilateral [14, 16, 17] and the majority have septations [18, 19]. Most are usually small to moderate in size [15, 16] and most are not associated with parenchymal TB disease [14, 15]. While the majority show some degree of pleural thickening, only a minority demonstrate pleural nodularity [18]. Previous authors have shown that sonographic septations, when combined with pleural fluid ADA levels, are predictive of TB aetiology [7] and septations can also help differentiate TPE from malignant effusions [19]. However, no study to date has investigated the ability of sonographic characteristics alone to predict TB aetiology of pleural effusions. This study sought to evaluate whether sonographic characteristics of pleural effusions—independent of laboratory analysis—can be used to predict TB aetiology of pleural effusions in a high TB setting. These findings may be used by RLS clinicians to reduce diagnostic delays and expedite appropriate initiation of anti‐TB medications. Materials and Methods Study Setting and Participants This was a cross‐sectional study performed among adult patients in Lusaka, Zambia. Ethical clearance was obtained from the University of Zambia Biomedical Research Ethics Committee (UNZABREC) and the Zambian National Health Research Authority (NHRA) prior to data collection. Study sites included two government hospitals in Lusaka: Chilenje Level 1 and Kanyama Level 2 hospitals. The sample size was calculated using Cochran's formula, based on an expected prevalence of 0.011% for TB effusions. Patients with suspected pleural effusion—based on physical examination, clinical presentation or chest X‐ray—were recruited from each hospital's emergency department, inpatient wards and HIV clinic. The study lead was then notified to perform bedside POCUS to confirm the presence of effusion. Patients with a POCUS‐confirmed effusion were eligible for the study if they were aged ≥ 18 years, had an effusion of suitable size for thoracentesis (per clinician's judgement, but generally interpleural distance of ≥ 2–3 cm) and provided written consent. Exclusion criteria include participants already diagnosed with TB infection and/or currently undergoing treatment with anti‐TB medications. All participants meeting inclusion criteria and enrolled in the study underwent thoracentesis, regardless of the effusion's sonographic characteristics or the clinician's level of suspicion for TB aetiology. Data Collection All scans were performed using a Butterfly iQ+ (Burlington, MA, USA) handheld ultrasound device. All participants were scanned bilaterally in the mid‐axillary line to screen for the presence of effusion at the costophrenic angles; further views were obtained transcostally over the flank and lower back for further characterisation, if necessary. Sonographic characteristics of the effusion were recorded, including the presence of septations, the presence of plankton sign (swirling, dot‐like heterogeneous echogenicities suggestive of exudate) and whether the effusion was unilateral or bilateral (see Figure 1) [20, 21]. Other sonographic characteristics of TPE, such as pleural nodularity, presence of adjacent parenchymal disease and size estimation, were not included in the protocol due to difficulty of technique or lack of an objective measure. Diagnostic thoracentesis was performed immediately post‐scan and pleural fluid was sent for Xpert MTB/RIF Ultra and ADA analysis. In addition, sputum and urine samples were obtained for Xpert MTB/RIF and lipoarabinomannan (LAM) analysis, respectively. Participants were considered ‘probable positive’ for TPE if pleural fluid ADA level ≥ 40 IU/L (considered the current diagnostic gold standard for TB in this study setting [22, 23]) but Xpert MTB/RIF and LAM testing were negative. Participants were considered ‘confirmed positive’ if any of Xpert MTB/RIF and LAM testing was positive. For the purpose of this study, ‘TB positive’ participants included both probable and confirmed positives. Limited participant demographic information was also recorded. Pleural effusion demonstrating plankton sign (left image) and septations (right image). Participants were assigned a study number and de‐identified data, including scan results, laboratory results and limited demographic information, were entered into an electronic database. All sonographic images were uploaded and stored securely on the encrypted Butterfly Network cloud‐based platform. Subsequently, all images were over‐read by a POCUS expert, who was blinded to ADA results. The initial scan readings were utilized for analysis, regardless of any discrepancies with the expert interpretation. Statistical Analysis Descriptive statistics were used to summarize participant characteristics by TB status and sonographic findings by TB status. Differences in proportions between groups (TB‐positive vs. TB‐negative) were assessed using Fisher's exact test due to small sample sizes. The association between sonographic findings and TB was assessed using Firth's penalized multivariable logistic regression to reduce small‐sample bias. In addition, we restricted our models to three key covariates (age, gender and HIV status) based on clinical and epidemiological relevance. Diagnostic performance of individual and combined POCUS findings for TB was evaluated by calculating sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV), along with 95% confidence intervals. In addition, kappa scores were calculated to assess the level of agreement between the initial and expert image interpretation for each sonographic characteristic. All analyses were conducted using Stata 18 (StataCorp, College Station, TX, USA). Results Participant Characteristics A total of 47 patients with effusions were screened for enrollment; two declined participation, none were < 18 years old and none were already diagnosed or undergoing treatment for TB. All participants were referred for screening based on evidence of effusion on a previous chest X‐ray, so all effusions were found on scan to be suitable in size for thoracentesis. A total of 45 participants were included in the analysis (Table 1). Of these, 73.3% (N = 33) were classified as TB positive based on ADA levels, GeneXpert and/or urine LAM results. The sample was predominantly male (70%) with a mean age of 39.3 years and there were no significant differences in gender or age distribution between the TB‐positive and TB‐negative groups. Employment status was nearly evenly distributed, with 51% (N = 23) of the sample employed, including those self‐employed or working as casual labourers and 44% (N = 20) unemployed; this distribution did not differ significantly by TB status. Marital status also did not differ significantly by TB status; however, a greater proportion of single individuals was observed among those who were TB positive (47% vs. 25%). A greater proportion of individuals in the TB‐positive group were HIV positive (39% vs. 25%), though this was not significant. TABLE 1Participant characteristics by TB status.FactorLevelTB positive (N = 33)TB negative (N = 12) p GenderFemale10 (30%)4 (33%)1.00Male23 (70%)8 (67%)Age18–34 years15 (47%)4 (33%)0.7035–49 years11 (34%)4 (33%)50–64 years3 (9%)2 (17%)65 years and older3 (9%)2 (17%)OccupationEmployed/self‐employed/casual labourer16 (48%)7 (58%)0.44Unemployed16 (48%)4 (33%)Retired1 (3%)1 (8%)Marital statusMarried17 (53%)8 (67%)0.14Single15 (47%)3 (25%)Widowed0 (0%)1 (8%)HIV statusNon‐reactive20 (61%)9 (75%)0.49Reactive13 (39%)3 (25%) Participant characteristics by TB status. Sonographic Findings As summarized in Table 2, most participants provided sputum and urine samples for microbiologic testing. However, the yield of microbiologic confirmation was low overall. Sputum GeneXpert positivity was rare, identified in only a few participants across both groups and urine LAM positivity was slightly more common among those classified as probable TB positive (16.7%) than probable TB negative (6.7%). TABLE 2Probable versus confirmed TB positives.Probable TB positive (ADA ≥ 40) (n = 30)Probable TB negative (ADA < 40) (n = 15)Sputum collected?23 (76.7%)11 (73.3%)Sputum GeneXpert result 3 (10.0%) positive 2 (13.3%) positive Urine collected?22 (73.3%)11 (73.3%)Urine LAM result5 (16.7%) positive1 (6.7%) positiveTotal confirmed positives8 (26.7%)3 (20.0%) Probable versus confirmed TB positives. 3 (10.0%) positive 2 (13.3%) positive When examining the association between sonographic findings and TB positivity (Table 3), unilateral location on either side (aOR = 4.9, p = 0.037), unilateral right‐sided location (aOR = 5.1, p = 0.043) and any two POCUS signs (aOR = 5.0, p = 0.044) were all significantly associated with TB in adjusted models. Septations and plankton signs were not significantly associated with TB in either unadjusted or adjusted models Figure 1. TABLE 3Association of POCUS findings with TB positivity.TB positive (n = 33)TB negative (n = 12)Unadjusted OR p Adjusted OR p Plankton sign27 (81.8%)8 (66.7%)2.2 (0.538, 9.331)0.2682.0 (0.489, 8.486)0.329Septation(s)24 (72.7%)6 (50.0%)2.6 (0.689, 9.657)0.1602.7 (0.583, 12.509)0.204Unilateral: Either side27 (81.8%)5 (41.7%)5.8 (1.432, 23.239)0.0144.9 (1.099, 21.499)0.037Unilateral: Right‐sided19 (57.6%)2 (16.7%)5.6 (1.214, 26.285)0.0275.1 (1.056, 24.201)0.043Any two POCUS signs present30 (90.9%)7 (58.3%)6.4 (1.343, 30.417)0.0205.0 (1.044, 23.953)0.044All three POCUS signs present15 (45.5%)2 (16.7%)3.5 (0.758, 16.341)0.1082.7 (0.579, 12.544)0.206Any POCUS sign present33 (100%)10 (83.3%)16.0 (0.708, 359.277)0.08116.5 (0.543, 503.461)0.108 Association of POCUS findings with TB positivity. Diagnostic performance varied substantially across individual and combined POCUS signs, with notable trade‐offs between sensitivity and specificity (Table 4). Overall, the presence of any POCUS sign demonstrated the highest sensitivity (100%), correctly identifying all TB‐positive cases, but this came at the cost of poor specificity (16.7%), indicating a high false‐positive rate. In contrast, unilateral right‐sided location and the presence of all three POCUS signs offered greater specificity (both 83.3%), suggesting these signs were more reliable in identifying true TB‐positive individuals, though their sensitivities were lower. Combinations of signs, particularly the presence of any two POCUS signs, yielded a strong balance, with high sensitivity (90.9%) and moderately good predictive values. Among individual findings, the plankton sign and unilateral location had relatively high sensitivity and positive predictive value, but modest specificity. TABLE 4Diagnostic performance of POCUS signs for TB diagnosis.POCUS signSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Plankton sign81.8 (64.5, 93.0)33.3 (9.9, 65.1)77.1 (59.9, 89.6)40.0 (12.2, 73.8)Septation72.7 (54.5, 86.7)50.0 (21.1, 78.9)80 (61.4, 92.3)40.0 (16.3, 67.7)Unilateral: Either81.8 (64.5, 93.0)58.3 (27.7, 84.8)84.4 (67.2, 94.7)53.8 (25.1, 80.8)Unilateral: Right57.6 (39.2, 74.5)83.3 (51.6, 97.9)90.5 (69.6, 98.8)41.7 (22.1, 63.4)Two POCUS signs90.9 (75.7, 98.1)41.7 (15.2, 72.3)81.1 (64.8, 92.0)62.5 (24.5, 91.5)Three POCUS signs45.5 (28.1, 63.6)83.3 (51.6, 97.9)88.2 (63.6, 98.5)35.7 (18.6, 55.9)Any POCUS sign100.0 (89.4, 100.0)16.7 (2.1, 48.4)76.7 (61.4, 88.2)100.0 (15.8, 100.0) Diagnostic performance of POCUS signs for TB diagnosis. While overall agreement was relatively high across all signs, the strength of agreement (as measured by kappa statistic) varied considerably (Table 5). The plankton sign showed moderate agreement (κ = 0.53), indicating some inconsistency in detecting this sign. Septations had a higher kappa (κ = 0.66), suggesting substantial agreement, while the strongest reliability was seen with unilateral location (κ = 0.94), reflecting near‐perfect agreement between readers. TABLE 5Agreement on POCUS signs between initial scan read and expert scan read.Initial scan readExpert scan readAgreement (%)Kappa (SE)Plankton signNo35 (78%)25 (56%)77.780.5263 (0.1313)Yes10 (22%)20 (44%)SeptationNo30 (67%)29 (64%)84.440.6557 (0.1489)Yes15 (33%)16 (36%)UnilateralNo32 (71%)33 (73%)97.780.9446 (0.1488)Yes13 (29%)12 (27%) Agreement on POCUS signs between initial scan read and expert scan read. Discussion This study assessed the diagnostic utility of POCUS findings for identifying tuberculous aetiology among patients presenting with pleural effusions in a high‐TB burden setting. The results indicate that no single sonographic feature demonstrates concurrently high sensitivity and specificity. Rather, diagnostic accuracy is improved through the combination of multiple sonographic findings. Notably, unilateral (either side) and right‐sided location, as well as the presence of any two concurrent POCUS findings, were independently associated with TB positivity and exhibited relatively high positive predictive values, suggesting their potential utility in clinical contexts where confirmatory laboratory diagnostics are limited or unavailable. In contrast to previous studies, our results did not show a significant association between plankton sign and sonographic septations with TB aetiology of pleural effusions, though both trended towards higher odds of TB positivity [18, 19]. We believe this is most likely a factor of small sample size and those larger studies would indeed show the presence of septations and plankton sign to be predictive of TB aetiology. Importantly, Qureshi and colleagues also found sonographic septations to be a useful discriminator between TB and malignant etiologies of exudative lymphocytic pleural effusions in a high TB setting [19]. This suggests that septations may be helpful to minimize the confounder of malignancy if this POCUS‐based diagnostic strategy were employed clinically in the future. Other data points, such as details from patient history, may also be useful to reduce confounding as well (e.g., shorter duration of symptoms for parapneumonic effusion). Combining sonographic findings with key patient history details may be an ideal method to optimize predictive ability of this non‐invasive strategy to diagnose TB effusions. Additional study will be required to further elucidate the utility of this strategy. The diagnostic performance was enhanced when sonographic findings were considered in combination. Specifically, the presence of any two POCUS signs provided the most favourable balance, achieving high sensitivity and a moderate positive predictive value. Although the concurrent presence of all three signs demonstrated high specificity, its low sensitivity limits its applicability as a screening tool. Conversely, the presence of any POCUS finding exhibited high sensitivity but poor specificity, indicating that while its absence may effectively exclude TB, its presence alone is insufficient for definitive diagnosis. These observations underscore the inherent tradeoff between sensitivity and specificity and suggest that a composite assessment of sonographic features may improve case detection without compromising specificity. All analyses were conducted using interpretations from the initial scan reader, thereby reflecting the real‐world use of POCUS in frontline clinical settings. Inter‐rater agreement was high for unilateral location but lower for other findings such as the plankton sign, emphasizing the necessity for standardized training and interpretation guidelines to enhance diagnostic consistency. For the RLS clinician in a high TB setting, such a POCUS‐based strategy for TPE diagnosis could complement the existing laboratory‐based diagnostic algorithm, which—even in many better‐equipped healthcare facilities—has poor TPE diagnostic yield and can be challenging to access. A POCUS‐based strategy may offer several advantages, including its non‐invasive nature, independence from consumables such as laboratory reagents and the ability to provide immediate, bedside results in real time. In light of this, a POCUS‐based strategy may serve as a viable alternative to laboratory confirmation (until the latter is available), improving clinicians' diagnostic acumen for TPE when no other options are available. Studies have shown that RLS clinicians can be readily trained in POCUS [24]. The POCUS scan techniques used in this study are easy to perform and the sonographic characteristics chosen are easily discernable even by POCUS novices. Such a strategy may reduce diagnostic delays and expedite initiation of TB medications, thereby leading to improved clinical outcomes for patients. Taken together, our findings suggest that POCUS could be an impactful strategy for diagnosing TPE in high‐burden, low‐resource settings. Further study with larger sample sizes will be necessary to fully determine the clinical utility of POCUS in TPE diagnosis. Limitations This study has several limitations. First, the small sample size limits the generalizability of findings and the statistical power to potentially detect associations. Some POCUS findings (e.g., plankton sign and septations) may be associated with TPE but this couldn't be detected in this study. Additionally, the sample size was too small to compare probable TB and confirmed TB cases. Second, this was a facility‐based study conducted in a capital urban city and may not reflect patterns of disease in rural populations or other geographic regions. Third, selection bias may have occurred. Patients were only included if they had pleural effusions large enough to allow thoracentesis and were willing to participate. This may have excluded individuals with small or atypical effusions. However this bias was deemed acceptable, as diagnostic thoracentesis was required to perform gold standard testing. Additionally, patients were only recruited if a patient was symptomatic and pleural effusion was suspected; this strategy may have excluded asymptomatic patients whose effusions may have differing characteristics. The complexity of gold standard testing was another limitation. By itself, elevated pleural fluid ADA level is an imperfect gold standard [4, 8], but direct methods of TB testing had major limitations, including poor sensitivity [6, 7], some patients' inability to produce sputum for Xpert MTB/RIF testing and inconsistent availability of LAM testing at the study's clinical site, all of which limited TB confirmation. Our ability to evaluate potential confounders was limited by the lack of comprehensive clinical data, including information on malignant and parapneumonic effusions, both of which may exhibit sonographic similarities with TPE, as well as comorbid conditions, history of TB and other patient characteristics. Further research with a larger sample size is warranted to investigate the role of these factors and better adjust for confounders. Finally, this study's results are applicable only to high‐TB settings; these findings should not be generalized to settings with lower TB prevalence. Funding The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest. Acknowledgements The authors would like to thank Seed Global Health and University of Zambia for their continued support of the University of Zambia Family Medicine program, as well as Chilenje and Kanyama hospital staff for their support with data collection. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. References Associated Data Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.