Increasing the detection of cases of TB is a high priority for Vietnam. The most efficient strategies are those implemented in high-risk populations that can be readily tested and treated. This district clinic-based contact investigation study showed the feasibility of contact investigation within urban and rural clinics of the government TB control program. It found a substantial prevalence of previously undiagnosed TB among household contacts. Our findings suggest that household contacts are a high-yield population for efficient screening to enhance TB case detection.
A key feature of this intervention was its integration within the national TB control program. Our strategy was implemented by existing district clinic staff, working within a system that delivers the majority of TB care throughout Vietnam. If a contact investigation program is to be sustainable on a large scale, it is essential that it be closely integrated with the existing TB control program. This integrated strategy, based on strong national and local leadership, is highly relevant to tuberculosis control programs in similar high-prevalence settings.
Another important aspect of our study was engagement of index patients to recruit members of their own household. This approach had several practical advantages. First, the approach draws upon the existing relationship that patients have with health care staff to increase the likelihood of cooperation from other family members. Second, our patient-directed approach ensures that patient consent is a prerequisite for family involvement, which addresses some of the ethical concerns surrounding patient confidentiality that arise in TB management. 
The prevalence of disease among contacts in our study of 0.73% is similar to that of many published studies,
although the point estimate was below the mean prevalence of 3.1% (95% CI 2.1%–4.5%, I2
98.8%) reported in a recent meta-analysis of 71 studies from low-middle income countries 
. That review found considerable variability between settings in the prevalence of TB disease in household contacts.
The use of chest X-ray 
with double-reading of films will have enhanced the detection of TB suspects in the present study. On the other hand, the finding that only 54% of suspects were able to produce sputum may indicate that we have underestimated the prevalence of confirmed pulmonary TB disease. However, the majority of suspects who could not produce a sputum specimen were asymptomatic, suggesting that the observed radiological abnormalities may have been due to past, healed TB or non-tuberculous lung disease. In future case-finding programs, additional methods such as sputum induction should be considered as an adjunct to spontaneous sputum production in individuals who cannot produce sputum. 
This is likely to improve the yield of sputum samples, and improve the sensitivity of the screening approach, particularly in early disease.
It is also possible that we underestimated the number of incident cases owing to relatively low attendance at 6 and 12-month follow-up visits. We attempted to minimise default from follow-up by educating contacts about the key symptoms of disease at the initial visit, contacting them by telephone in advance of their 6 and 12 months appointments.
The incidence of TB remains high for several years after infection 
. Hence, periodic screening during two years after exposure may be considered as a part of a contact investigation program and this approach is routine in some high-income countries 
. In our study, we identified one new case at the six-month follow-up assessment. This is consistent with a recent meta-analysis that showed the incidence of TB among contacts was 1.5% (95% CI 0.9–2.4%, I2
96.3%) during the first year after exposure. 
Effective screening requires high rates of attendance in contacts. In our study we improved compliance by drawing upon the therapeutic relationship between the patient and health staff during the eight months of treatment. Staff incorporated contact visits into the routine visits of patients, which is a relatively effective and low-cost strategy, in combination with delivering education about the early symptoms of TB to contacts.
In addition to issues of feasibility, the effectiveness of contact investigation compared to usual passive case-finding is an important issue for policy-makers. 
Contact investigation aims to enable earlier identification of infectious patients, decreasing the period of infectiousness, thereby reducing ongoing transmission. As none of the contacts with TB in our study were known to have the disease at the time of diagnosis, it is likely that the intervention resulted in earlier diagnosis of the disease. However, we cannot predict what proportion would have eventually been diagnosed by ‘passive case finding’. In order to adequately determine effectiveness compared to usual care, it will be necessary to conduct a randomized controlled trial with matched control subjects that are not screened.
In order for active case-finding to reduce TB transmission, contacts not only need to be diagnosed, but also to complete treatment. 
All contacts diagnosed with the disease commenced a standard 8-month course of TB treatment, and all eventually completed treatment. However, one individual initially defaulted from treatment, and subsequently recommenced a second course of therapy after a six month delay. These outcomes are similar to the 91% completion rate achieved in Vietnam among patients diagnosed by passive case-finding. 
Molecular testing of paired M.tb
isolates in one index patient-contact pair in our study was consistent with direct transmission, based upon identical MIRU-VNTR patterns. In the second patient-contact pair, the isolate of the contact had three less bands in two loci compared to that of the index patient. While the genotype of M.tb
in a population can lose or add repeats over time, 
in a population with considerable strain heterogeneity, 
the loss of repeats at two separate loci probably indicates this is an independent strain which was acquired from another unrecognised patient. Other molecular epidemiology studies in low and high-prevalence settings have demonstrated that known index patients are not necessarily the source of infection in contacts. 
A study from a low-prevalence setting found that 70% (95% CI 56–82%) of isolates from index patient-contact pairs shared identical strains. 
In low prevalence settings, small differences in molecular typing can be useful marker of the presence of additional unknown source cases. 
However, in high-prevalence settings, such as Vietnam, the risk of an individual being infected by an unrecognised patient is much greater and reactivation of longstanding infection is also more likely. Hence, while household exposure contributes to the risk of TB, it may not be the only source of infection. 
Nonetheless, household contacts should be considered a suitable target population for screening, because they often share other risk factors in addition to exposure to the known index patient.
There are a number of limitations in interpreting this study. The small sample size of this pilot study means that confidence intervals for outcomes are wide, precluding a conclusion about the difference among contacts and the general population. We also noted a substantial drop-off in attendance for six and twelve month follow-up visits. This varied considerably between Districts, with the lowest rates occurring in a district that lacked radiology facilities, and required contacts to travel more than three kilometres for an X-ray. While follow-up rates may have been higher if our study staff conducted routine home visits to promote screening, this approach would add considerable cost to the existing program and contacts would still need to attend a health care facility to have a chest X-ray. Finally, although child contacts were enrolled in the study, screening identified no suspects under the age of 15 years. This may reflect the actual prevalence, however given that the diagnosis of childhood TB is often challenging our contact investigation may have missed some cases, particularly of extra-pulmonary disease. Other strategies in addition to chest X-ray may be required in order to identify more effective strategies of identifying early disease children. Contact investigation programs can also provide a suitable framework for concurrently implementing isoniazid prophylactic therapy, for which there is evidence of both efficacy and cost-effectiveness. 
The role of contact investigation as a strategy for TB control in resource limited settings remains a subject of debate. 
Until recently WHO recommendations have emphasised the role of passive, rather than active, case-finding in TB control programs 
. Recommendations for active case finding have been limited to contacts who are children and contacts of people living with HIV.
However, there is a growing interest in strategies that enhance case-finding for TB, reflected in new WHO guidelines for TB contact investigation in low and middle-income countries. 
A recent survey identified 65 countries with a national TB contact investigation policy, although it did not evaluate how widely these policies were implemented. 
National programs should consider introducing active case finding if they have a program that is able to deliver therapy reliably, has high treatment completion rates and available resources to invest in enhanced case detection. Therefore, analysis of cost-effectiveness is essential to justify adding or diverting resources for widespread contact investigation. Further research, involving a comparable control population, should be performed before committing substantial resources to contact investigation. Furthermore, it is important to recognise that contact investigation will still only identify a minority of undiagnosed cases in the whole population, and should be considered in combination with other active case-finding strategies. This study has demonstrated the importance of pilot testing of contact investigation, to learn lessons about contact investigation in the local context in advance of its widespread implementation.
TB remains a major public health challenge for Vietnam, despite a decade of high treatment completion rates. Enhancing case detection is now a key priority for the NTP, in order to reduce ongoing disease transmission in the community. This study shows that household contact investigation is a feasible strategy to achieve this goal. Further studies are required to assess its effectiveness and cost-effectiveness, particularly in resource limited settings.