This study found a high prevalence (3.9%, 3940 per 100,000) of previously undiagnosed TB among household contacts of newly diagnosed TB patients, only one-third lower than that (6075 per 100,000) observed in a similar contact tracing study in a peri-urban area with nearly three times the background incidence of TB [13]. The yield in our study of 8.5 new TB cases for every 100 index cases traced was substantially lower than that observed in the peri-urban area, however, where household contact tracing yielded 23 new TB cases/100 index cases traced. Household contact tracing in a peri-urban setting also resulted in a NHNS of 4.3, compared to 12 in our rural setting. Some of this discrepancy may be attributable to higher participation rates and/or larger household sizes in the peri-urban setting, where an average of 4 persons participated per household, compared to only 2 participants/household in our rural setting.
The sensitivity of smear for culture-confirmed active TB in this population was less than 10%, and despite the high prevalence of TB symptoms among index cases, household contacts with TB were no more likely to report symptoms than those without TB. This analysis demonstrates that, even in rural settings, household contact tracing can feasibly identify cases of active TB, but symptom screening and sputum smear microscopy are unhelpful in identifying TB cases. Thus, for contact tracing to have a meaningful impact in such settings, more expensive procedures (such as performing mycobacterial culture or Xpert MTB/RIF on all contacts), or radiography using digital chest X-rays for screening, will likely be required. This may be especially important for sub-populations, such as people living with HIV, for which sputum microscopy performs particularly poorly, and among which we found no cases of confirmed TB in our study.
While the World Health Organization (WHO) recommends TB screening for the household contacts of newly diagnosed TB patients (because of their elevated risk of TB disease), they do not recommend a specific algorithm [18]. Instead WHO provides a range of potential algorithms, such as testing only those persons with any cough, a cough of more than 2 weeks, or the presence of any TB symptoms (e.g. cough, fever, weight loss, night sweats, or lethargy) [18]. Had we used this symptom screen to identify household contacts for further testing, we would have missed 9 (82%) of the 11 undiagnosed prevalent TB cases – none of whom reported a cough. Options, such as screening using digital chest X-ray, may be a cost-effective approach to TB case-finding among household contacts, and should be explored in future research.
The fact that the majority of new TB cases identified by this study were asymptomatic and smear-negative indicates that household contact tracing in rural areas may identify cases of TB early, before substantial secondary transmission or TB-related morbidity or mortality can occur. Virtually all index cases recruited for our study were symptomatic, whereas household contacts with TB were no more likely to report symptoms than their family members without active TB. These findings suggest that TB cases captured by active contact tracing interventions are different than the cases captured through passive case detection, and that the diagnostic tests and screening algorithms required to detect them differ as well. While 73% of index cases were smear-positive, only one of 11 household contacts with prevalent active TB was positive on smear.
We identified a very high prevalence of non-TB mycobacteria (NTM) among household contacts, about 1.5 times as high as the prevalence of culture-confirmed TB. Other studies from South Africa have also identified high rates of NTM infection [19, 20]. This result raises questions regarding the timing of TB treatment initiation among culture-positive persons identified through contact tracing. Speciation takes approximately 5 additional days [21] from the time a culture positive result is returned. Avoiding treatment delays for those with active TB is paramount, but preventing unnecessary treatment should be an important consideration given the poor positive predictive value of mycobacterial culture for TB (38%) in this population. Testing household contacts with Xpert MTB/Rif rather than culture would allow for more rapid TB diagnosis, avoiding unnecessary TB treatment for NTMs and preventing treatment delays associated with active TB.
This study has a number of important limitations. First, although we screened more than 280 household contacts, our sample size of TB cases was small, leaving us without power to detect modest but potentially important differences between those with and without TB. Because we recruited index cases with and without laboratory-confirmed TB, it is possible that some of the participating index cases did not have active TB disease. However, this study sought to explore the feasibility and effectiveness of household contact tracing under operational conditions in which TB is not always bacteriologically confirmed. Finally, due to budgetary constraints in this small, pilot study, we were unable to perform HIV testing, chest X-ray, or TB testing with Xpert MTB/RIF for household contacts, or to perform genotyping to demonstrate transmission between the index case and household TB cases identified by the study. Previous genotyping studies in South Africa have found that a sizable portion of presumed case-pairs within households had TB strains that were genetically distinct [22, 23], suggesting that household TB transmission may not be responsible for all TB cases identified during household contact tracing. Further studies of TB contact tracing in rural settings could seek to expand the sample size, include additional data on room-sharing and contact duration, evaluate novel diagnostic tools including Xpert MTB/Rif and digital chest X-ray, study the cost-effectiveness of active contact tracing in this setting, and elucidate the relationships between HIV and TB status among household contacts.