Comparison of stool and gastric content samples analyzed with genexpert ultra for the diagnosis of pulmonary tuberculosis in children under five in Niger: a multi-center diagnostic study

Abstract

Background

Pediatric tuberculosis is challenging to diagnose due to the limited presence of bacteria and difficulties in obtaining high-quality sputum samples. This study assessed the effectiveness of stool samples versus gastric contents analyzed with GeneXpert Ultra for diagnosing pulmonary tuberculosis in children under five years old.

Methods

A diagnostic study was conducted in Niger over a one-year period (January 1 to December 31, 2024), at multiple centers. Socio-demographic data and GeneXpert Ultra results from stool and gastric samples were collected in accordance with STARD guidelines. Data analysis was performed using Excel 2020 and JAMOVI 2.3.28 software to calculate sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results

This study included 401 patients with a median age of 18 months (interquartile range (IQR) 11–24 months) and a sex ratio of 1.37. The diagnostic performance of the test showed a sensitivity of 77.8% and a specificity of 99%, with an overall accuracy of 98.5%. The positive and negative predictive values were 63.6% and 99.5% respectively.

Conclusion

GeneXpert stool testing is a valuable complementary approach to gastric content examination for detecting tuberculosis. Integrating it into screening strategies, especially in resource-limited settings, can enhance pediatric tuberculosis detection and management.

Background

Tuberculosis is a significant infectious disease, causing 1.25 million deaths globally by 2023. It is a leading cause of morbidity and mortality, especially in resource-limited countries. The World Health Organization (WHO) estimated 10.8 million new TB cases globally in 2023, with 1.25 million cases in children (12%), particularly affecting those under five years old [1]. Sub-Saharan Africa has a high burden of tuberculosis, with nearly 25% of global cases. The region also faces high mortality rates among young people, raising concerns about the impact of the disease [1].

Tuberculosis is a significant public health issue in Niger, with an estimated incidence of 74 cases per 100,000 inhabitants in 2022 [1]. Children under five face particular challenges in diagnosing tuberculosis due to nonspecific symptoms, rapid disease progression, and difficulties in obtaining quality respiratory samples for testing.

Obtaining a quality respiratory sample from a child is challenging, making invasive techniques like gastric tubing and nasopharyngeal suctioning difficult to perform routinely [2]. Recent studies [2–4] have shown promising results using stool samples with this test for the diagnosis of pediatric tuberculosis. This non-invasive approach is more acceptable to children and parents due to its simplicity and minimal resource requirements.

It enhances pediatric tuberculosis detection and allows decentralized diagnosis in peripheral health facilities equipped with Cepheid’s GeneXpert system [3]. Several studies reported specificity and specificity values demonstrating good diagnostic performance (pooled specificity 98% [95%CI: 96–99], pooled sensitivity 57% [95%CI: 40–72]) [5]. This benefits the rural populations, the majority of whom are in Niger, by improving access to healthcare.

To date, no study has been conducted in Niger to assess the performance of this diagnostic approach for tuberculosis. This study aims to fill this gap and provide valuable data to improve diagnostic strategies in the country.

Methods

Aim

To study the analytical performance of stool compared to gastric content samples to diagnosis pulmonary tuberculosis in children using the GeneXpert MTB/RIF Ultra assay.

Study setting

The study was conducted at the National Reference Laboratory for Tuberculosis (NRL/TB) at Amirou Boubacar Diallo National Hospital in Niamey, as well as at the laboratory of Madarounfa health district, Maradi region, Niger.

Type and period of study

This multicenter cross-sectional diagnostic study was conducted from January 01 to December 31, 2024.

Study population and sampling

The study included patients with presumed pulmonary tuberculosis based on WHO criteria, regardless of gender. Sampling was done exhaustively.

Eligibility criteria

The study included all patients aged 0–5 years referred to the NRL/TB and Madarounfa health district laboratory for GeneXpert MTB/RIF Ultra examination with gastric content (GC) and stool samples.

Patients on anti-tuberculosis treatment for more than seven days or those with inadequate or poorly preserved samples were excluded.

Sample collection and biological analysis

Fresh stool samples were collected in a clean jar using the simple one-step (SOS) stool method from presumed pulmonary TB patients and transported according to biosafety standards. Gastric content (GC) samples were obtained by nasogastric probing and transferred to a collection jar for rapid transport to the laboratory.

At the laboratory, each patient’s sample was analyzed using the GeneXpert MTB/RIF Ultra system (Cepheid, USA)[6] in three steps (for both stool and GC):

1. Sample preparation and loading onto the GeneXpert cartridge.

2. Inserting the cartridge into the GeneXpert instrument for automated testing.

3. Interpretation of results generated by the GeneXpert system.

Stool samples were processed using a simple method involving dilution with buffer, vortexing, centrifugation to obtain sediment, and direct testing with GeneXpert Ultra cartridges, following manufacturer guidelines adapted for pediatric samples.

Data collection and analysis

A standardized survey sheet was used to collect data, including socio-demographic characteristics (age, gender, HIV status) and Xpert MTB/RIF Ultra test results for MTB, bacillary load, and rifampin resistance screening.

Data analysis was conducted using Excel 2020 and JAMOVI version 2.3.28. The STARD flow chart was used for patient inclusion data [7]. Analytical sensitivity, specificity, positive and negative predictive values were calculated to assess the performance of GeneXpert Ultra for diagnosing pulmonary tuberculosis from stool samples, with gastric content samples as the reference test. McNemar’s chi-square test with a 95% confidence interval was used to determine agreement between stool and gastric content sample results, with a p-value < 0.05 considered statistically significant.

Results

This study included 401 patients who underwent a total of 802 MTB/RIF Ultra tests on stool and gastric lavage samples (Fig. 1).

Fig. 1.

Study patient flow according to STARD recommendations

Basic characteristics of included patients

The median age of patients was 18 months (IQR 11–24 months). The 12- to 23-month age group had the highest number of patients. Fifty-seven point 8% of the patients were male, with a sex ratio of 1.37. The HIV status of 68.1% of patients was unknown (Table 1).

Table 1.

Baseline characteristics of patients

Characteristics	Numbers	Percentage
Age (months)
Median age 18 (IQR: 11–24)
0–11	111	27.7
12–23	164	40.9
24 months and over	126	31.4
Gender
Female	169	42.3
Male	231	57.8
HIV Statut
Unknown	273	68.1
Negative	128	31.9

HIV Human immunodeficiency virus

GeneXpert Ultra results on gastric content samples

Out of 401 gastric content samples tested with GeneXpert MTB/RIF, 9 (2.2%) were positive for Mycobacterium tuberculosis complex, mostly with low bacillary load (7 out of 9) (Table 2). Among the nine positive cases, only 1 was resistant to rifampin.

Table 2.

MTB/Rif Ultra results on GC samples

GeneXpert gastric content samples	Numbers	Percentage
Mycobacterium tuberculosis complex
Detected	9	2.2
Invalid	3	0.7
Not Detected	389	97.0
Bacillary load GC
Low	6	66.7
Very low	1	11.1
Trace	2	22.2
Rifampicin resistance
Detected	1	11.1
undetermined	3	33.3
Not detected	5	55.6

GC Gastric content

GeneXpert Ultra results on stool samples

Out of 401 stool samples tested with GeneXpert MTB/RIF, 11 (2.7%) were positive for Mycobacterium tuberculosis complex. Among these, the majority had a low bacillary load (n = 4; 40%). None of the 11 samples showed rifampicin resistance (Table 3).

Table 3.

MTB/RIF Ultra results on stool samples

GeneXpert stool samples	Numbers	Percentage
Mycobacterium tuberculosis complex
Detected	11	2.7
Invalid	11	2.7
Not detected	379	94.5
Bacillary load Stool
High	1	10.0
Medium	2	20.0
Low	2	20.0
Very low	1	10.0
Trace	4	40.0
Rifampicine resistance
undetermined	3	27.3
Not detected	8	72.7

Comparison of stool samples with gastric contents in terms of analytical performance

Stool samples demonstrated a sensitivity of 77.8% and a specificity of 99.0% for diagnosing pulmonary tuberculosis in children, compared to gastric contents (the Reference Standard test). The efficiency was 98.5%, with a positive predictive value of 63.6% and a negative predictive value of 99.5% (Table 4).

Table 4.

Correlation between stool and gastric contents

MTB STOOL		MTB GC			p value
		Detected	Invalid	Not Detected
Detected	Numbers	7	1	3	0,032
	%	77.8	33.3	0.8
Invalid	Numbers	2	1	8
	%	22.2	33.3	2.1
Not Detected	Numbers	0	1	378
	%	0.0	33.3	97.2
Total	Numbers	9	3	389
	%	100.0	100.0	100.0

MTB GC Gastric content GeneXpert result (reference Standard test), MTB Stool Stool GeneXpert result

Bacillary load reported for n = 9 positive GC samples; denominators reflect valid results excluding invalids.

The agreement between stool GeneXpert test and GC GeneXpert test was moderate, with Cohen’s Kappa of 0.546 ( IC95% = 0.36–0.73).

Stool samples compared with gastric content samples (Reference Standard test) had an AUC of 0.816 for diagnosing pulmonary tuberculosis in children (Fig. 2).

Fig. 2.

ROC of stool sample results compared with gastric contents

Discussion

Early diagnosis of tuberculosis in children is a significant challenge, especially in resource-limited countries like those in sub-Saharan Africa. Delayed diagnosis contributes to high morbidity and mortality rates in this population [7, 8]. A study conducted in Niger in 2024 at two centers (LNR-TB and DH Madarounfa) investigated the use of stool samples as an alternative to gastric tubing for diagnosing tuberculosis in children.

However, the study has limitations to consider. The wide confidence interval for sensitivity (40.0-97.2%) indicates variability, possibly due to the small positive sample size. This differs from a previous study by Mekkaoui et al. in 2021 [9], which reported a sensitivity of 91.1% (95% CI, 85.6–95.1). Larger-scale studies are needed to refine these estimates [10]. Additionally, including the culture of gastric contents, as suggested in previous literature, could improve the diagnosis of pulmonary tuberculosis [9–12].

This study found that the Xpert MTB/RIF Ultra test performed well on stool samples, with a sensitivity of 77.8% and a specificity of 99%. These results are consistent with recent literature [13]. Previous studies [11, 14] have reported lower sensitivity for the standard Xpert MTB/RIF test on stool samples in children. The improved performance of the Ultra version of GeneXpert may explain the higher sensitivity observed in this study [14, 15]. This advancement in technology reflects the ongoing evolution of diagnostic tools in tuberculosis testing [16].

The analysis of our study’s predictive values provides valuable insights for clinical practice. With a positive predictive value of 63.6% and a negative predictive value of 99.5%, our approach offers a reliable decision-making tool for clinicians. These results compare favorably with those reported by Ssengooba et al., who found a PPV of 50.0% and NPV of 95.9% for Xpert MTB/RIF Ultra [17]. The robust performance of molecular tests in various epidemiological contexts is evident in our study. The high NPV is particularly noteworthy given the low prevalence of tuberculosis in our pediatric population.

The high prevalence of low bacillary loads (82.1%) in our study mirrors the clinical reality of pediatric tuberculosis. This finding aligns with previous research [14] highlighting the paucibacillary nature of the disease in young children. The challenge of diagnosing tuberculosis in children under five, who often have difficulty producing good-quality sputum, underscores the need for a multifaceted diagnostic approach tailored to pediatric patients [18–23].

The study revealed a concerning 71.8% rate of unknown HIV status, highlighting the need for improved routine HIV screening in children with suspected TB. This issue is part of broader efforts to strengthen health systems and integrate services in Africa. The low rifampicin resistance rate of 3.6% aligns with regional data, with similar rates reported in the Sahel region [24–26]. However, countries like Guinea-Bissau and Guinea-Conakry have significantly higher resistance rates, emphasizing the importance of ongoing surveillance in the sub-region [7, 27].

Analysis of the area under the ROC curve (AUC) of stool samples on GeneXpert Ultra shows excellent performance in detecting Mycobacterium tuberculosis complex, with results consistent with previous studies [28–30]. The high AUC value of 0.97 indicates the test’s robustness, especially in cases with low bacterial load where traditional methods struggle. Our findings support the superiority of GeneXpert Ultra over previous versions, particularly in HIV-positive patients. The ROC curve analysis identifies an optimal operating point that balances sensitivity and specificity, aligning with literature on threshold optimization [30, 31].

Training healthcare staff is essential for the successful integration of new diagnostic approaches in tuberculosis control programs, as highlighted by WHO [32]. The use of non-invasive stool sampling shows for improving childhood tuberculosis diagnosis in resource-limited settings , especially in sub-saharan Africa [33].

We acknowledge several limitations in this study. First, GeneXpert performed on gastric aspirates was used as the reference Standard test, whereas culture remains the true gold standard for tuberculosis diagnosis. This methodological choice may have influenced the estimated performance of the evaluated tests. Second, only nine (n = 9) cases were positive according to the reference test, resulting in a wide confidence interval for sensitivity (40.0–97.0%) and therefore reduced precision. Furthermore, the positive predictive value (63.6%) may also be influenced by this positivity rate. The agreement between these tests (Cohen’s kappa with wide confidence interval IC95% = 0.36–0.73) may also be influenced by this rate.

Cross-study comparisons are limited by the lack of a uniform gold standard; here, gastric content Ultra was used as the reference standard comparator, consistent with pivotal pediatric studies.

These limitations highlight the need of validation by further studies with a larger cohorts (or meta-analyses) and using culture as the reference standard.

Conclusion

In this study, we compared the performance of the GeneXpert test on stool samples with gastric contents, as reference Standard test for diagnosing tuberculosis in this study. The findings indicate that GeneXpert stool test has good sensitivity but does not match the performance of the gastric contents test. However, its high specificity suggests that a positive stool result strongly indicates the presence of an infection. The positive and negative predictive values suggest that this test could be a valuable screening tool, especially in settings with limited access to invasive methods.

These results suggest that using stool samples as a diagnostic tool could improve tuberculosis screening, particularly for children and vulnerable patients. This could lead to earlier diagnosis and better case management. While stool GeneXpert Ultra shows good sensitivity, it serves as a valuable complementary, non-invasive approach to gastric content testing, with excellent specificity and NPV making it an ideal rule-out and initial screening tool in resource-limited settings. Integrating it into screening strategies, especially in resource-limited settings, could enhance tuberculosis detection and management.

Abbreviations

AUC

Area under the curve

GC

Gastric content

HD

Health District

HIV

Human immunodeficiency virus

IQR

Interquartile range

MTB/RIF

Mycobacterium tuberculosis and rifampicin resistance

NPV

Negative predictive value

NRL/TB

National Reference Laboratory for tuberculosis

PPE

Personal protective equipment

PPV

Positive predictive value

ROC

Receiver operating characteristic

SOS

Simple one-step

TB

Tuberculosis

USA

United States of America

WHO

World Health Organization

Authors’ contributions

Conceptualization: A.N.B, and Y.A; methodology: A.N.B, L.M.M and Y.A; validation: Y.A, B.S and S.M; investigation: A.N.B; writing-original draft preparation: A.N.B; writing-review and editing: A.N.B, Y.A, O.T, L.M.N, B.H, M.G.M, M.S.S, S.A, B.S, S.M; All authors have read and agreed to the published version of the manuscript.

Funding

This work was not supported by any research funding.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study received ethical approval from the Faculté des Sciences de la Santé, Université Abdou Moumouni institutional review board (IRB) (Reference No:000970/UAM/FSS/D). Moreover IRB was granted in each site: Hôpital National Amirou Boubacar Diallo (Reference No: 0000138/HNABD/DAF/GRH) and District Sanitaire de Madarounfa (Reference No: 44/DS/MDFA). Written informed consent was obtained from parents and legal guardians for their child’s participation in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.