OR WAIT 15 SECS
Sponsors should be aware of the significant implications the Addendum is likely to have on clinical trial planning, conduct, statistical analysis, and interpretation.
With the announcement on September 4, 2017 that the E9 (R1) Addendum on Estimands and Sensitivity Analysis in Clinical Trials has reached Step 2b of the ICH process, the addendum has now entered the consultation period. The deadline for comments is February 28, 2018 for the EC, and April 30, 2018 for the U.S. But even before it becomes final, sponsors should be aware of the significant implications the Addendum is likely to have on clinical trial planning, conduct, statistical analysis, and interpretation.
The primary aim of the Addendum, which focuses on confirmatory clinical trials, is to provide a structured framework for linking trial objectives to suitable trial designs, estimation, and hypothesis testing. The establishment of a common framework is intended to facilitate the dialogue between all stakeholders-including clinical teams, regulatory agencies, payers, and patients-involved in all of these activities.
Confirmatory clinical trials are conducted to quantify the effects of a treatment and provide evidence of efficacy and safety to support regulatory decision-making. However, after start of the study treatment (as in real life), events may occur that complicate the description and interpretation of the study drug’s treatment effects. Such events are called ‘intercurrent events’ and include things like use of an alternative treatment or rescue treatment during the study, discontinuation from treatment or study dropout, treatment switch and sometimes also death. Defining efficacy and safety variables and methods for statistical analysis without first addressing the potential for intercurrent events can lead to ambiguity about the treatment effect being measured. It can also make communication about the trial and its results challenging for all involved-especially between sponsors and regulators, who may have differing interests when it comes to the objectives and scientific questions of interest in the study.
The concept of ‘estimand’ is central to the proposed framework (Fig. 1). An estimand is meant to be closely linked to the objective of the trial and clearly describe what is to be estimated based on the question of interest. The estimand is specified by four attributes:
· The population of interest;
· The variable or endpoint of interest;
· The measure of intervention effect, taking into account the potential impact of intercurrent events; and
· The population level summary for the respective variable.
In contrast to the estimand, the ‘estimator’ is the analytic approach to be used to compute an estimate from the observed clinical data. Inferences based on a particular estimand should be robust to limitations in the data and deviations from the statistical model assumptions used for the main estimator. This robustness is evaluated through sensitivity analyses aligned to the estimand.
A trial protocol should pre-specify the primary estimand that corresponds to the primary trial objective, the main estimator that is aligned with the primary estimand and leads to the primary analysis, and suitable sensitivity analyses. Additionally, estimands for secondary trial objectives that support regulatory decisions should be described in the protocol as well, each with a corresponding main estimator and a suitable sensitivity analysis.
The construction of the estimand should address each intercurrent event that may occur in the clinical trial and that might effect the interpretation of the results. The addendum defines five different strategies for addressing intercurrent events: 1) treatment policy, 2) composite, 3) hypothetical, 4) principal stratum, and 5) while on treatment. More than one strategy-but usually not all-may be relevant in a clinical trial.
For example, consider a randomized, placebo-controlled trial for a chronic, non-life-threatening disease, in which the full effect of the drug is expected to be observed at six months. However, placebo subjects are allowed to switch (before month six) to rescue medication due to lack of efficacy.
In this case, three different strategies for the estimand may be employed when taking the intercurrent event ‘switch to rescue medication’ into consideration:
· Treatment-policy strategy. This approach considers month six assessment for all subjects, independent of whether the subject has switched to rescue medication before month six or not. The missing month six information after switch to rescue medication would be imputed as defined in sensitivity analyses, e.g. by multiple imputations using data of subjects in the same treatment group.
· Hypothetical strategy. This approach asks, “What would the treatment effect be if no subjects had switched to rescue medication?” For subjects without switch to rescue medication, the month six assessment would be considered. For subjects with switch, only their data until the time point of switch would be considered, and month six values would be imputed using data of non-switched subjects in the same treatment group.
· While-on-treatment strategy. Using this approach, for all subjects with switch to rescue medication, only the data until the switch would be considered.
While noting that it “remains undisputed that randomization is a cornerstone of controlled clinical trials and that analysis should aim at exploiting the advantages of randomization to the greatest extent possible,” the addendum also makes clear that “the question remains whether understanding the effect of a treatment policy always targets the treatment effect of greatest relevance to regulatory and clinical decision making.”
The suggested new framework proposed in the ICH E9 (R1) addendum provides a basis for all stakeholders to align in the process on choices for trial design, data collection, and statistical analysis. The addendum envisions multiple ways to quantify treatment effects when considering such clinically relevant variables as tolerability, adherence, and whether or not additional medication is needed. Clinical teams may need to employ innovative endpoints and trial designs and endpoints along with new data collection and statistical methods to assess certain estimands. Greater harmonization on all of these matters from the beginning can reduce ambiguity about the magnitude and meaningfulness of clinical trial results, thus accelerating the clinical trial process.
Ülker Aydemir, Associate Director, Biostatistics Consultancy and Methodology, INC Research/inVentiv Health