Homestead food production (HFP) programs may improve diet and nutrition outcomes by increasing availability of nutrient dense foods such as vegetables and supporting livelihoods. We conducted a pair-matched cluster-randomized controlled trial to investigate whether vegetable home gardens could improve women's dietary diversity, household food security, maternal and child iron status, and the probability of women consuming nutrient-rich food groups. We enrolled 1,006 women of reproductive age (18-49 years) in ten villages in Pwani Region, Eastern Tanzania, matched the villages into pairs according to village characteristics, and randomly allocated villages to intervention or control. Households in the intervention villages received agricultural training, inputs to promote home production of nutritious crops, and nutrition and health education. Data were collected in 2016, 2017, and 2019 and analyzed using linear regression models with propensity score weighting adjusting for individual-level confounders, differential loss to follow-up, and fixed effects for village pairs to accommodate the pair-matched design. Results after one year of the intervention (previously published) found significant improvements in dietary diversity. However, three years after the start of the intervention the difference in dietary diversity disappeared, even though the number of women who grew at least one crop was significantly higher (75 percentage points, 95% CI: 72, 81) in treatment households compared to controls. Barriers to maintaining a home garden, including lack of irrigation opportunities and fencing materials, and social disruption may have precluded sustained impacts from home gardening in this context. Future home garden programs should carefully consider mechanisms and investments needed for sustained impact over time.