In 2011, 8.7 million new cases of tuberculosis (TB) (13% human immune-deficiency virus (HIV) co-infected) were found and 1.4 million people died from the disease. The proportion of TB-patients co-infected with HIV is highest in Africa resulting in a large proportion of sputum smear negative patients.
The worldwide case-detection rate has reached its peak at 60% and is unable to reach the target of 70% set by the WHO in 2000, in order to be able to halve the 1990 prevalence and mortality rates in 2015. The reasons for this stagnation are disorganized case finding and low sensitivity of the only widely available diagnostic tool, sputum smear microscopy. If the patient is identified as TB-case, a six month lasting treatment is started; monitored by repeated sputum smears for initially sputum smear positive and by weight gain for initially sputum smear negative patients. Both monitoring tools are shown not to be valuable in the prediction of failure (i.e. death or treatment failure) and an estimated 3% of HIV-uninfected and 9% of HIV infected TB-patients die while treated for TB.
While the search for better diagnostic tools continues, the need for an immediate improvement of case-finding is apparent and the WHO currently explores how to set up an effective screening for TB. Clinical scoring systems to enable triage and prioritize special interventions and focused follow-up have been developed for many diseases, but although highly demanded, there has only recently been proposed a potential monitoring tool for TB patients.
The original Bandim TBscore is a clinical scale measuring clinical severity of pulmonary TB (PTB), by assessing six signs (anemia, pulse>90 beats per minute, positive finding at lung auscultation, temperature >37°C (axillary), Body Mass Index (BMI) <18/<16 and mid-upper-armcircumference (MUAC) (<220mm/<200mm) and five self reported symptoms (cough, hemoptysis, dyspnea, chest pain and night sweats).
In this thesis we explored the TBscore, refined it and investigated its place in case-finding. Paper I is a study on the different signs´ and symptoms´ inter-observer agreement and a comparison with the Karnofsky Performance Index (KPS). We found the inter-observer variability of the TBscore variables to be slight to almost perfect. The grading of PTB patients into severity classes showed moderate agreement for both the TBscore (kw= 0.52, 95% CI 0.46 – 0.56) and the KPS (kw= 0.49, 95% CI 0.33 – 0.65). The intra-class correlation coefficient (ICC) was larger for the TBscore than for the KPS (0.822 vs 0.632), indicating that the variability between the scorings by the TBscore were due to a higher degree of variability among the scored patients than for the KPS. While the KPS was influenced by the HIV-status, the TBscore grated the patients into all three severity classes, unaffected by the HIV-status. However, the study revealed the need for revision of the score.
In paper II we attempted to refine the TBscore. To examine if the variables all were part of the same underlying construct we carried out an exploratory factor analysis. The result of this and the earlier inter-observer variation analysis guided the development of TBscoreII which was validated internally, in Bissau, Guinea-Bissau, and externally, in Gondar, Ethiopia. TBscoreII at treatment had the ability to predict failure during treatment in Bissau (p<0.001) and, though not significant, also in Gondar (p=0.208). A failure to decline ≥25% in score from baseline to two month of treatment was associated with subsequent mortality in Bissau (p=0.063) and Gondar (p<0.001).
Paper III is a study on the TBscore´s place in the screening of PTB suspects attending adult consultations at primary healthcare centers in Bissau, Guinea-Bissau. We assessed symptoms and signs in a cohort of 1089 patients, seeking care and having either a cough, reporting weight loss and/or expectoration. Of the included PTB suspects, 107 were diagnosed with PTB; 76.4% sputum smear positive and 25.4% HIV-infected. The area under the curve from the receivers operating characteristic (ROC) was 0.64 (95%CI 0.59 – 0.69) when assessing TBscoreII`s relation to PTB-diagnosis. The mean score for patients getting the PTB-diagnosis was 3.0 (95%CI 2.7 – 3.2), while it was 2.4 (95%CI 2.3 – 2.5) in patients not diagnosed with PTB. Applying a TBscoreII>2 as criteria to refer to diagnostic examination would have missed 9.3% of the by us found PTB-cases, while cough>2weeks missed 34.6%. TBscore/TBscoreII are shown to be useful monitoring tools for PTB-patients on treatment, they could fill the void which currently exists in risk-grading and guidance of a focused follow-up. The scores may have a role in future screening-routines for PTB, however, this requires that our findings are repeated in other settings.