(A) For pairs of samples connected by a transmission event from i to j, transmission of Mtb is expected to have occurred on the terminal branch above j. Even though we do not know exactly when j went from latent TB to active TB, the latent period is included in the length of the terminal branch leading to j (see main text). We therefore use this branch length as an upwardly biased estimate for latency time. (B) For transmission pairs in the calculated transmission networks, the length (in years) of terminal branches leading to the recipient of the pairs (overestimated latency period) was extracted and stratified by HIV status of the recipient. To account for incomplete sampling, the analyses were performed on all 251 calculated transmission events as well as subsets including only the most likely transmission pairs (top 45, 35 and 25%). ***denotes p<0.001, *denotes p<0.05 as determined by unpaired t-test.