SETTING: Zambian and South African TB and HIV Reduction (ZAMSTAR) cluster-randomised trial (CRT) communities, 2006-2009. OBJECTIVES: To develop TB stigma items, and evaluate changes in them in response to a household intervention aimed at reducing TB transmission and prevalence but not tailored to reduce stigma. DESIGN: TB stigma was measured at baseline and 18 months later among 1826 recently diagnosed TB patients and 1235 adult members of their households across 24 communities; 12 of 24 communities were randomised to receive the household intervention. We estimated the impact of the household intervention on TB stigma using standard CRT analytical methods. RESULTS: Among household members, prevalence of blame and belief in transmission myths fell in both study arms over time: adjusted prevalence ratios (aPRs) comparing the household intervention with the non-household intervention arm were respectively 0.61 (95%CI 0.26-1.44) and 0.77 (95%CI 0.48-1.25) at 18-month follow-up. Among TB patients, at baseline a low percentage experienced social exclusion and poor treatment by health staff and a relatively high percentage reported 'being made fun of', with little change over time. Disclosure of TB status increased over time in both study arms. Internalised stigma was less prevalent in the household arm at both baseline and follow-up, with an aPR of 0.85 (95%CI 0.41-1.76). Variability in stigma levels between countries and across communities was large. CONCLUSION: Robust TB stigma items were developed. TB stigma was not significantly reduced by the household intervention, although confidence intervals for estimated intervention effects were wide. We suggest that stigma-specific interventions are required to effectively address TB stigma.