Study rationale. While household socioeconomic position (SEP) may be clearly a root cause of tuberculosis\ud (TB), it is more challenging to understand how household SEP affects the risk of TB, how\ud this effect is mediated by risk factors that are on the causal pathway, how this effect differs at\ud different level of analysis and different pathogenic stages of TB, and how this evidence can\ud inform control strategies.\ud Objectives. The aim of this study was to: 1) explore the ecological correlation between community living\ud conditions and TB prevalence rates; 2) quantify the association between household SEP and\ud TB disease and TB infection and, 3) explore the causal mechanism underlying these\ud associations.\ud \ud Methods. This research project was nested within a population-based HIV-TB prevalence survey\ud conducted in 2005-2006 in two Zambian communities, one rural and one urban. Each\ud community is divided into smaller tracts called Census Advisory Areas (CSA). Prevalent TB\ud was diagnosed through culture and molecular testing of sputum samples collected from the\ud prevalence survey participants. The correlation between community living conditions and\ud TB prevalence was assessed through an ecological analysis conducted at CSA level. The\ud association between household SEP and TB disease and infection was assessed respectively\ud through a case-control and a cross-sectional study. CSA living conditions were explored\ud through poverty mapping. Variables accounting for four different domains of household SEP\ud were recorded (human resources; food availability; housing quality; and access to services)\ud and combined into a composite index using principal component analysis. In addition four\ud more SEP indices were developed: one for each household SEP dimension considered. The\ud analysis of the mediation pathway between household SEP and TB (disease and infection)\ud was driven by a pre-defined conceptual framework, including household SEP and individuallevel\ud risk factors. Adjusted Population Attributable Fractions (PAF) were estimated.\ud \ud Main findings. Ecological analysis - The overall cluster-adjusted TB prevalence rates was 870/100,000\ud (95%Cl: 570-1160/100,000) Urban CSA had both significantly higher TB prevalence rates\ud and wealthier living conditions compared to rural CSA. Although not significantly, TB\ud prevalence rates tended to increase with the worsening of urban CSA living conditions.\ud Case control study - At household level, prevalent TB was significantly associated with\ud lower household SEP [aOR = 6.2,95%Cl: 2.0-19.2 and aOR = 3.4,95%Cl: 1.8-7.6\ud respectively for low and medium household SEP compared to the baseline]. The food\ud availability domain was more strongly associated with prevalent TB than the other domains\ud considered [aOR = 4.2,95%Cl: 1.8-9.2]. At individual level, TB prevalent cases were significantly more likely to have a diet poor in proteins [aOR= 3.1,95%Cl: 1.1-8.7], to be\ud not BCG vaccinated [aOR = 7.7,95%CI: 2.8-20.8], to be HIV positive [aOR= 3.1,95%Cl:\ud 1.7-5.8], and to have migrated [aOR = 5.2,95%CI: 2.7-10.2] than controls. These\ud associations all persisted after controlling for household SEP. The association between\ud household SEP and TB appeared to be mainly mediated by inadequate nutrition. Before\ud mediation, PAF for household SEP was equal to 30%. The adjusted PAF for inadequate\ud nutrition and HIV infection was equal to 42% and 36%, respectively.\ud Cross sectional study - TB infection was associated with higher, rather than lower,\ud household SEP [aOR = 0.4,95%Cl: 0.2-0.9 and aOR= 0.4,95%Cl: 0.2-0.8 for high and\ud medium household SEP compared to low SEP]. This association was driven by the\ud household SEP domain on access to community services [aOR = 2.7,95% Cl: 1.0-7.1 ]. None\ud of the investigated risk factors appear to mediate the association between household SEP and\ud TB infection.\ud Conclusions. \ud In this setting, urban communities were wealthier but also had greater prevalence of TB. The\ud low power of the ecological analysis does not allow any conclusions; however - as for other\ud health indicators, even for TB it seems that the apparent advantage of living in an urban\ud setting can mask severe TB inequalities across smaller urban communities.\ud This study revealed an unexpected and counterintuitive result: low household SEP was\ud significantly associated with a higher risk of TB disease, but with a lower risk TB infection.\ud Higher household SEP may be associated with lifestyles increasing social mixing and thus\ud the risk of infection. Among the infected people, poorer individuals may be more likely to\ud progress to TB disease and, therefore, are more likely to be identified in prevalence surveys.\ud In the association with TB disease, household SEP seems to operate mainly through\ud inadequate nutrition. While both 11V and inadequate nutrition were important social\ud determinants of prevalent TB, these findings suggest that interventions addressing food\ud insecurity may prevent more TB cases in this setting.\ud Because of low study power and possible selection biases results this interpretation will have\ud to be confirmed by more powered and better designed studies. In the meantime, this study\ud suggests that interventions addressing household SEP, through food and financial support,\ud may effectively reduce TB prevalence and complement currently control strategies mainly\ud targeted to HIV positive people.\ud To evaluate the impact and cost-effectiveness of this approach is a priority.