Well 74-44 is 30. Any chance those 30 odd patients we thought were non randomized and instead placed on the treatment arm are not charted? Would be strange and mean that the last patients enrolled weren't enrolled to the treatment arm only (as the last enrolled are highly likely to be those censors in the 19-22month range).
Forget the chart for a second. Can we agree 44 of 182 are alive at or beyond thirty-six months. Thus 64 must be alive at 36 months or less. Can we safely infer by taking 331 - 182 = 149. Therefore there are 149 people as of March 2017 that had surgery less than 36 months prior to that date. Thus 43% (aka: 64 / 149) of patients between 18 and 36 months post surgery were alive March 2017. Is that a safe presumption ignoring the chart?
Flipper, your math here is correct with one oversight... you have to remove the LTFU from the denominator as they are censored from both num/denominator. Therefore, the math is 64 / 137 = 46.7%