The risk of tuberculosis is greatest soon after infection, but Mycobacterium tuberculosis can remain in the body latently, and individuals can develop disease in the future, sometimes years later. However, there is uncertainty about how often reactivation of latent tuberculosis infection (LTBI) occurs. We searched eight databases (inception to June 25, 2019) to identify studies that quantified tuberculosis reactivation rates occurring more than 2 years after infection (late reactivation), with a focus on identifying untreated study cohorts with defined timing of LTBI acquisition (PROSPERO registered: CRD42017070594). We included 110 studies, divided into four methodological groups. Group 1 included studies that documented late reactivation rates from conversion (n=14) and group 2 documented late reactivation rates in LTBI cohorts from exposure (n=11). Group 3 included 86 studies in LTBI cohorts with an unknown exposure history, and group 4 included seven ecological studies. Since antibiotics have been used to treat tuberculosis, only 11 studies have documented late reactivation rates in infected, untreated cohorts from either conversion (group 1) or exposure (group 2); six of these studies lasted at least 4 years and none lasted longer than 10 years. These studies found that tuberculosis rates declined over time, reaching approximately 200 cases per 100 000 person-years or less by the fifth year, and possibly declining further after 5 years but interpretation was limited by decreasing or unspecified cohort sizes. In cohorts with latent tuberculosis and an unknown exposure history (group 3), tuberculosis rates were generally lower than those seen in groups 1 and 2, and beyond 10 years after screening, rates had declined to less than 100 per 100 000 person-years. Reinfection risks limit interpretation in all studies and the effect of age is unclear. Late reactivation rates are commonly estimated or modelled to prioritise tuberculosis control strategies towards tubuculosis elimination, but significant gaps remain in our understanding that must be acknowledged; the relative importance of late reactivation versus early progression to the global burden of tuberculosis remains unknown.