This was a real-life study in the high TB and HIV prevalence setting where the MPT64 test was implemented in the routine EPTB diagnostics and used by the clinicians. The MPT64 test had a sensitivity of 92%, and performed far better than conventional tests such as Mtb culture and the WHO recommended GeneXpert in children with varied EPTB presentations. This is the second study where we have shown that the test is implementable in a basic pathology laboratory facility in a low-resource setting7. The results are consistent with results in the more controlled setting and shows that the MPT64 test can help in diagnosing pediatric EPTB5.
Our sample size is small with 30 EPTB patients in total. It is therefore difficult to make a good assessment of performance for different sites of disease. A recent study in Zanzibar found a sensitivity of the MPT64 test of 100% for pediatric TB adenitis7. In our study the risk for false negative MPT64 test results was low, which highlights its potential role as a confirmatory test for EPTB, adding sensitivity and timeliness to Mtb culture. As only 16 non-TB samples was assessed by the MPT64 test, the specificity (88%) could have been both under- or overestimated. Three non-TB samples had an uncertain MPT64 result, and were removed from the analysis of specificity. A conclusive result on these samples could have both increased or decreased the specificity of the MPT64 test.
The quality assessment performed by LS shows that many slides (7/20) were labelled as inconclusive by an outside pathologist from a Norwegian university hospital. 5/7 were labelled as uncertain due to poor quality of the slide. However, the pathologists in Mbeya Zonal Referral Hospital that had assessed over 250 MPT64 slides stained locally, did not report this ambiguity. This underlines that the result of the MPT64 test is dependent on the reader, and familiarity with local variations in staining patterns and artifacts.
Mtb culture performed poorer than the MPT64 test, but the assessment of its performance could have been affected by the number of contaminated samples. 10 Mtb culture samples were contaminated, of which 7 among TB patients. The overall sensitivity of Mtb culture on 22 samples was 45%. The transportation of samples to the Central Tuberculosis Reference Laboratory in Dar es Salaam could have affected the performance of Mtb culture both positively and negatively. However, being necessary for drug susceptibility testing, the long turn-around time and poor sensitivity of the test makes it less feasible as a stand-alone test. The poorer sensitivity of Mtb culture of 13% for EPTB samples is confirmed in our Zanzibar study7. Mtb culture remains an important diagnostic test despite its limitations and efforts should be made to improve the method through better laboratory facilities.
GeneXpert had a lower sensitivity of 31% as compared to Mtb culture (45%). Two samples (20%) were Mtb culture positive and GeneXpert negative. In previous studies GeneXpert has been shown to have a comparable, but slightly lower sensitivity than Mtb culture for various EPTB samples1,2. The advantage of GeneXpert compared to Mtb culture is the turn-around time, and in this study the reduced risk of inconclusive or contaminated results. A positive test result is very useful due to the high specificity, but the risk of false negatives is high, reducing its role as a stand-alone test for EPTB. The cost of GeneXpert when not subsidized is also a source for sustainability concerns20.
FNAC performed excellent in this study with a sensitivity of 90% and a specificity of 95%. These findings support the use of this simple test for diagnosis of EPTB. FNAC is simple, can be performed decentralized, slides can be prepared and sent to a central laboratory. However, due to the small sample size in our study (10 TB and 21 non-TB samples), the results should be interpreted with caution. FNAC has been shown to be unspecific in other settings, and diseases such as sarcoidosis, non-tuberculous mycobacteria and other granulomatous conditions can mimic TB3. The MPT64 antigen detection test can be used as a confirmatory test in cases where there is a suspicion for other granulomatous conditions. The pathologists did not report a suspicion of any non-TB granulomatous conditions in this study, which is similar to the results from Zanzibar. The MPT64 test had a correlation of 100% with positive FNAC results in our study. One FNAC sample did not show typical features of TB adenitis, but was MPT64 positive. The results from Zanzibar also supports a potential use of MPT64 as a confirmatory test, separating TB from non-TB related necrosis in lymphadenitis.
Cytological findings in cell smears of the effusions lacked both specificity and sensitivity to diagnose EPTB. The MPT64 test was positive in 14/16 TB cases that had effusions sampled, indicating the utility of the test in various types of samples. However, due to a small sample size, the results need to be validated in further studies.
The superior performance of MTP64 compared to Mtb culture and GeneXpert can be attributed to the paucibacillary nature of EPTB, and the potential role of antigen accumulation as a central part of EPTB pathogenesis. Antigen accumulation, rather than the bacillary load has been described as the central phenomenon in TB pathogenesis and tissue destruction21,22.
Accounting for the fact that Mtb culture is an imperfect gold standard, the MPT64 test was compared to a composite reference standard (CRS). We have included response to treatment to TB treatment as a part of the CRS to mimic real life practice. However, response to treatment is not conclusive as especially Rifampicin is fairly wide-spectred and would treat many conventional bacterial infections. The CRS accounted for this uncertainty by labeling patients with grades of certainty of their diagnosis; confirmed, probable, possible. The use of a CRS comes with a risk of misclassification bias. The MPT64 test was implemented and used by the clinicians, and some patients were started on TB treatment on the basis of their MPT64 result. If MPT64 had a tendency of being false positive in other infectious conditions, these patients could have been misclassified as TB cases. However, only two TB adenitis cases have been labelled as TB cases on the basis of response to treatment. One of these cases was MPT64 positive, and the other had a missing MPT64 test result.
In our study the prevalence of peritonitis (30%) and meningitis (20%) was higher than the usual distribution of pediatric EPTB sites. The severe presentations do not seem to be explained by a lack of BCG coverage, as BCG coverage was very good in our cohort. With concomitant malnutrition the severe presentations of TB peritonitis and meningitis is not surprising, but is not well documented in other studies23. Immunosuppression due to malnutrition, young age among the malnourished and recruitment as inpatients could have been contributing factors.
The prevalence of malnutrition was very high (29%). Malnutrition was more prevalent among the TB patients, as compared to the non-TB group, highlighting the need for a high level of clinical suspicion of not only pulmonary, but also extrapulmonary TB in this patient group. This is supported by a recent study in Ethiopia showing a higher incidence of malnutrition in adults with EPTB than in those with PTB24. From previous research it appears that malnutrition is a predictor of tuberculosis disease and is associated with worse outcomes23,25. In our study, 13/17 (76%) of the malnourished patients had EPTB, with peritonitis (4 patients) and meningitis (4 patients) being the most prevalent. Only 1/8 HIV infected patients had malnutrition, which reduces its potential impact on malnutrition in this study. Three of the malnourished patients had malignancies as their final diagnosis, which highlights that the malnourished patients in this study had serious concomitant diseases.
Both Mtb culture and GeneXpert performed poorer in the malnourished, with sensitivities of 30% and 25%, respectively. The MPT64 test had a sensitivity of 90%, but the results have to be interpreted with caution as only 10 samples were asessed. Also, surprisingly, the sensitivity of Mtb culture and GeneXpert did not improve in immunosuppressed patients (sensitivity of 38% and 28% respectively), where one would expect that the poor immune response of the host would lead to a greater bacterial load.
The mortality rate was equally high in the TB and non-TB group (23% vs 21%). However, these results could have been affected by the high number of patients lost to follow up in the non-TB group (13% vs 39%, TB versus non-TB group, respectively), masking a potential lower mortality rate in the non-TB group. When looking at mortality through the lens of malnutrition, one sees a trend towards increased mortality in this group with a mortality rate of 36%. A recent systematic review from Ethiopia found a double incidence of death in children < 5 y with severe acute malnutrition (SAM) and concomitant TB, and a high case fatality rate of SAM at 11.3%26. However, it is not specified if cases of EPTB are included in the review and their potential contribution to excess mortality. The contribution of TB to excess mortality in pediatric SAM is also reproduced in Zambia with a case fatality rate among children with SAM and TB of 56%. However, 94% of the TB cases in the Zambian study were pulmonary and the HIV prevalence was higher at 46.5% among SAM patients. The low number of EPTB cases in this study could point towards under detection27. Estimating the mortality rate of pediatric EPTB is challenging due to the lack of uniform case definitions and lack of standardized follow ups. Unfortunately, our small sample size reduces the strength of the analysis of contributing factors to mortality.
The trend of increased mortality was stronger in the group that included all causes of immunosuppression. In total, 10/61 patients died, of which 8 had immunosuppression. 6 patients that died had TB, 3 had malignancies. This highlights the need for a comprehensive evaluation of patients to ensure that various reasons of immunosuppression are considered.
Mbeya Zonal Referral Hospital is a tertiary referral hospital with laboratory services, radiology services and a histopathology lab. In this setting 51% of the presumptive EPTB cases were not classified as TB cases. In an even more resource constrained setting these patients could have been subject to a trial of TB treatment on suspicion of EPTB, leading to unnecessary treatment and diagnostic delay or even a wrong diagnosis. In our study the potential differential diagnosis ranged from benign conditions to life threatening conditions. 11/61 (18%) were diagnosed with malignancies and 10/61 (16%) with other infectious conditions. This highlights the need for diagnostic facilities and an awareness of differential diagnosis when examining patients with suspected pediatric EPTB. Under TB control programs only treatment for EPTB is free of cost, but all the diagnostic work-up incurs great financial burden on the patients. Diagnostic delay could have been a contributing factor to the high mortality. Early detection and timely referral of presumptive EPTB from local clinics to the tertiary hospital is probably needed to ensure a correct diagnosis and to reduce mortality.
In conclusion, the MPT64 test was implementable in the routine TB diagnostic setting in a tertiary hospital in a low-resource, high TB and HIV setting, and the test performance was better than routine diagnostic tests, including GeneXpert. However, this test cannot be implemented until a basic laboratory structure is available and health systems are strengthened. The method of immunostaining used in the MPT64 test is also useful for strengthening diagnosis of other diseases, especially cancers.