Statistical considerations

No statistical hypotheses will be formally tested in this study. All data will be reported separately according to predecessor study. The safety and full analysis sets will comprise all patients who were randomized and who received at least one dose of tezepelumab or placebo in either of the predecessor studies, regardless of their enrolment into DESTINATION.

Subpopulations of patients who are enrolled in DESTINATION and who receive at least one dose of tezepelumab or placebo during this study are defined as the safety and full analysis LTE subsets. Given that the LTE subsets will comprise patients who completed the predecessor studies, they may be subject to selection bias. In particular, measures of asthma control may be affected because patients with less severe asthma may have been more likely to complete the predecessor study and enrol in the LTE study. These subpopulations will therefore be used for supportive analyses only, and the number of patients who discontinue treatment and patterns of discontinuation will be considered when interpreting the data.

The effect of tezepelumab after treatment cessation will be evaluated in an additional subpopulation comprising patients who were originally randomized in NAVIGATOR and who are in the safety analysis LTE subset, complete the LTE study and enter the extended 36-week follow-up period.

There are five possible treatment groupings across the predecessor and DESTINATION studies: 1, tezepelumab-tezepelumab; 2, tezepelumab-did not enrol; 3, placebo-tezepelumab; 4, placebo-placebo; 5, placebo-did not enrol (Fig. 4). The primary treatment groups for analyses are randomized tezepelumab (all patients originally randomized to tezepelumab in the predecessor studies, i.e. groups 1 and 2) and randomized placebo (all patients originally randomized to placebo in the predecessor studies, i.e. groups 3 [until switch to tezepelumab], 4 and 5).

Treatment groupings across the predecessor (NAVIGATOR or SOURCE) studies and the DESTINATION LTE study. LTE long-term extension

The supportive treatment groups are: tezepelumab predecessor plus tezepelumab LTE (all patients originally randomized to tezepelumab in the predecessor studies and re-randomized to tezepelumab in the LTE study, i.e. group 1) and placebo predecessor plus placebo LTE (all patients originally randomized to placebo in the predecessor studies and re-randomized to placebo in the LTE study, i.e. group 4) (Fig. 4). Other treatment groups will be used for exposure summaries and rare events, and to assess the durability of benefit and the effect after treatment cessation.

There will be two database locks in the DESTINATION study. The primary database lock will be conducted after the last patient completes week 104, and the final database lock will be conducted once all patients have completed the last follow-up visit at week 140. All statistical analyses of the primary and secondary efficacy outcomes will be performed based on the primary database lock data.

The baseline measurement for safety endpoints will be the last measurement on or before the first dose of study treatment in the predecessor study, and for efficacy endpoints it will be the last measurement on or before the date of randomization in the predecessor study. If there is no such value for either endpoint type, the baseline value will not be imputed, and the value will be described as ‘missing’. Statistical analyses of safety outcomes will be performed on the safety analysis set, and key safety analyses will be repeated using the safety analysis LTE subset.

AEs, AESIs and SAEs, which will be used to assess the primary objective, will be summarized for the safety analysis set over the 104-week study period using exposure-adjusted incidence (i.e. number of patients reporting events divided by person–time at risk) to account for the variability in follow-up. In addition, exposure-adjusted incidence will be explored over defined time periods to assess potential changes in risks with longer exposure to treatments. Observed and change-from-baseline values in laboratory data will be summarized descriptively.

Statistical analyses of efficacy outcomes will be performed primarily on the full analysis set; key efficacy analyses may be repeated using the full analysis LTE subset. The secondary objective will be assessed by comparing the AAER over 104 weeks in the tezepelumab group with that of the placebo group. This will be achieved by using a negative binomial model with the total number of asthma exacerbations experienced from baseline in the predecessor studies until week 104 in the LTE study as a response variable. Treatment, history of exacerbations and stratifying variables from the predecessor studies will be included as covariates in the model. The logarithm of the time at risk for exacerbation during the study will be used as an offset variable in the model.

Exploratory outcomes will be summarized descriptively; for the effect of tezepelumab after cessation of treatment, efficacy data will be presented using descriptive statistics by time period, and safety data will be summarized by time period using exposure-adjusted incidence to account for the variability in follow-up. Laboratory data will be summarized descriptively over time. Loss of clinical effect will be explored through assessments including initiation of another biologic treatment, increase in the OCS maintenance dose, increase in symptoms measured by the ACQ, asthma exacerbations and study withdrawal. Any additional statistical analyses required owing to COVID-19 will be pre-specified in the statistical analyses plan before sponsor unblinding and will be described when the results for this study are reported.