Weights, such as those applied to our control group (ie, the ReCORD-FL sample), can reduce the precision of statistical estimates such that the analyses behave as though we have a smaller-than-actual sample of controls. The ESS quantifies this reduction of precision in the controls. Thus, the larger the reduction from actual to effective sample size, the more statistical precision that is lost via weighting.

Auto-HSCT, autologous HSCT; ESS, effective sample size; FLIPI, Follicular Lymphoma International Prognostic Index; SD, standard deviation.

Enrolled patients are those who met inclusion or exclusion criteria and had a leukapheresis product accepted for manufacturing, regardless of infusion status (only 1 enrolled patient was not included owing to missing data).

Sample size after weighting (ie, sum of weights) was 99 for the ReCORD-FL study, and effective sample size was 95.

Because double refractoriness and last prior therapy refractory status both capture refractoriness status, only 1 prognostic factor (double refractoriness) was included in the propensity model. Furthermore, refractory status to last preceding therapy was already very well balanced (SMD < 0.25) before weighting, suggesting that the inclusion or exclusion of this prognostic factor has a limited impact on the analysis. FLIPI was excluded from the model on the basis of missingness (27 [19%] additional patients would be excluded from the analysis if FLIPI were included in the model). Considering that 3 of the 5 risk factors of the FLIPI score are already included in the model (age, number of nodal sites, and disease stage), which achieved excellent balance with absolute SMDs < 0.25, FLIPI was also well balanced between the cohorts before weighting, and thus, to conserve sample size, FLIPI was excluded.