Pulmonary tuberculosis case detection in a medium incidence middle-income country

Case detection is central to tuberculosis (TB) control. Missed or delayed diagnoses lead to longer periods of infectiousness that sustain TB transmission. Modelling indicates that detecting 70% of smear-positive pulmonary TB cases and curing 85% of them - the present World Health Organization targets - can decrease TB incidence by 10% per year. However, in 2014, worldwide, only 63% of the estimated smear-positive cases were detected and notified and TB incidence is declining at only 2% per year. Most research addressing case detection is biomedical and focuses on developing new diagnostic tests that perform better than the traditional ones. However, implementation of these tests has hardly had an impact on TB incidence. This thesis combined a health system and an operational perspective to investigate the broader process of case detection. We designed and conducted, in Lima Peru, six studies on pulmonary TB and MDR-TB case detection: from the criteria used in the selection of suspects to be tested, to their identification and the testing procedure itself, up to the start of treatment. The studies aimed to generate evidence on performance gaps at these steps and propose potential solutions for improvement. The study setting was San Juan de Lurigancho, a peri-urban district in Lima with over one million inhabitants where TB and MDR-TB prevalence are amongst the highest in the country. To augment the efficiency of the diagnosis of pulmonary TB with smear microscopy, we first evaluated the case definition of TB suspects– a person coughing for 14 days or more - and the compliance with this case definition. In a cross sectional survey among 4376 persons that had submitted a sputum sample for TB diagnosis, we determined the smear positivity rate and its association with the duration of cough and with patient and health service characteristics. Over half of the patients who had sputum tested (55.3%, 2418) reported cough for <14 days. Of them, 3.2% (78) were smear positive vs. 12.4% (243/1958) of those coughing ≥ 14 days. Cough for ≥ 14 days, being referred by health care staff, attending a secondary-level health care facility, male sex and age between 15 and 44 years were independent determinants of smear positivity. The inclusion of persons who were not TB suspects was a consequence of behavior driven by a performance target set by the TB program, which requires every health facility to perform smear microscopy in 5% of all adult patients that receive consultations. We recommend revising that target and continuing to use the current TB suspect definition, as it has a high smear positivity yield, but it is essential to ensure compliance with it. Subsequently, in a second study on TB diagnosis through smear microscopy, we modified the approach to sampling slides for the external quality assurance (EQA) of laboratory performance. In ten health facilities, during two years, instead of selecting diagnostic slides for EQA, we selected follow-up slides of patients that were on treatment. The false negative errors were more frequent in follow-up slides than in diagnostic slides: 25 (3.5%) vs. 3 (0.6%). Hence, the sensitivity to detect quality control problems was substantially increased when rechecking follow-up slides and the modified sampling scheme, which permitted a smaller sample size, proved very efficient and effective for identifying laboratories with substandard performance. However, we observed no actual improvements in performance in year two. This was a consequence of suboptimal feedback of EQA results to local laboratory technicians by the supervisor. We recommend implementing the EQA with our modified sampling strategy and strengthening the supervision and feedback process within the laboratory network. To enhance case detection in specific groups at increased risk of TB, we first conducted a longitudinal study in household contacts of recently diagnosed TB patients. We determined TB incidence within two years of exposure among 5466 household contacts of 1178 patients, and identified household, patient and contact characteristics that determine an increased risk. The TB incidence among household contacts was 1918 (95% confidence interval (CI) 1669-2194) per 100,000 person-years, more than ten times higher than in the general population. Incident TB occurred more than six months after the index case’s TB diagnosis in 121/205 (59.0%) household contacts. Bacillary load and time between symptoms and treatment initiation in the index case, as well as the kinship to the index case and sex of the contact were significantly associated with TB incidence in the contacts. We recommend extending the routine household contact follow-up from the current six months to two years after the index case’s diagnosis. Targeting household contacts on the basis of risk factors may increase efficiency, but the feasibility and acceptability for patients, contacts and TB staff needs to be determined. A second study on groups at increased risk focused on people living with HIV/AIDS. We determined the diagnostic accuracy of a commercial automated nucleic acid amplification test –Xpert®MTB/RIF- among HIV positive TB suspects. In 131 HIV positive patients with a high suspicion of pulmonary TB, detection of TB by Xpert®MTB/RIF was compared to a composite reference standard of Löwenstein-Jensen (LJ) and liquid culture. The sensitivity of Xpert MTB/RIF was 97.8% (95% CI 88.4–99.6) (44/45), specificity was 97.7% (95% CI 91.9–99.4) (84/86) and the positive predictive value was 95.7% (95% CI 85.5–98.8) (44/46). Xpert®MTB/RIF detected 13/14 smear-negative TB cases, outperforming smear microscopy (p = 0.0002). Detection of rifampicin resistance was also studied. The sensitivity of Xpert®MTB/RIF was 100% (95% CI 61.0–100.0) (6/6) and specificity was 91.0% (95% CI 76.4–96.9) (30/33). In summary, Xpert®MTB/RIF performed well for TB diagnosis and outperformed smear microscopy, but the specificity for detection of rifampicin resistance was low. We recommend the implementation of Xpert®MTB/RIF for TB and drug resistant TB diagnosis among HIV positive patients in replacement of the currently used underperforming nitrate reductase assay. To strengthen case detection and treatment initiation of MDR-TB, we first assessed the definition of MDR-TB suspect. We conducted a cross-sectional survey to determine the proportion of MDR-TB among TB patients who did not report any of the risk factors used by the TB program to perform a drug susceptibility test (known exposure to a TB patient whose treatment failed or who died or who was known to have MDR-TB; immunosuppressive co-morbidities; ex prison inmates; prison and health care workers; and alcohol or drug abuse). We found 6.3% (95%CI 4.4–8.3) (37/584) of MDR-TB among TB patients that would not be routinely tested for drug susceptibility. This suggests that testing patients for MDR-TB based on the presence of the currently used set of risk factors is insufficient in settings with high MDR-TB prevalence, as it misses many cases. We recommend to use a broader definition of MDR-TB suspects or, if possible, universal testing for MDR-TB in all TB patients. Universal testing will allow early starting of correct treatment regimens and appropriate management of contacts. Subsequently, to determine the delay between diagnosis and MDR-TB treatment initiation, we conducted a retrospective study in confirmed MDR-TB patients. We found that TB diagnosis did not translate into immediate treatment when MDR-TB was suspected or confirmed. Among 35% (13/37) of the patients directly started on a MDR-TB regimen, more than 30 days elapsed between the first positive smear result at the health facility and the initiation of treatment. In 27% (24/88) of patients switching to a MDR-TB regimen, more than 30 days passed between the last dose of the drug sensitive regimen and the first dose of the MDR regimen. These delays contribute to ongoing transmission of MDR-TB in the community. Operational processes and steps to design and make individualised regimens available in health facilities should be shortened. Throughout our studies, we obtained a comprehensive picture of the pulmonary TB case detection process in peripheral health facilities in a large district in eastern Lima. Our results led to recommendations that can enhance overall TB and MDR-TB case detection and speed up MDR-TB treatment initiation. A systematic understanding of the process of case detection can also facilitate the implementation of new tests, including future point-of-care tests and permit to reach their full potential. From a global perspective, our middle-income urban setting with a medium TB and MDR-TB burden and low HIV prevalence is particular. Our results may not directly extrapolate to different settings, but the approaches we used can be replicated elsewhere to conduct a systematic assessment of TB case detection.