Following on from his recent (August 2019) article, Darren Jolliffe provides an update on the finalised Estimands and Sensitivity analysis guidance.
The finalised addendum R1 to ICH E9 “Estimands and Sensitivity Analysis in Clinical Trials” was adopted on 20 November 2019. As explained in my earlier article, this guidance will have a direct impact on how protocols are written. See the earlier article for a description of the importance of relating clinical trial objectives to an appropriately defined estimand(s). This update will focus on considering the impact of and documenting relevant intercurrent events within the Estimand framework.
An Estimand is a precise description of the treatment effect which reflects the clinical question posed by a clinical trial objective. The attributes of an estimand are:
|1||Population of interest||Patients targeted by the scientific question|
|2||Variable (or endpoint) of interest||Measure(s) required to address the scientific question (to be obtained for each patient)|
|3||Treatment description||The treatment of interest, and if appropriate, the alternative treatment to which comparison will be made|
|4||Remaining intercurrent events||The specification of how to account for other intercurrent events (i.e. those relevant ones not specified in the endpoint definition) reflecting the scientific question of interest|
|5||Population level summary for the variable||Provides a basis for a comparison between treatments.|
An approach to choosing an Estimand can be based on the following steps:
- Consider the therapeutic setting and intent of treatment being given in determining a trial objective
- Identify (relevant, i.e. which will likely affect interpretation of the results) intercurrent events
- Discuss strategies (within the clinical team) to address these intercurrent events (and reasons for choices made)
- Construct the estimand
- Align the estimand with the study trial design, data collection and method of estimation
- Identify assumptions for the main analysis and suitable sensitivity analyses to investigate these assumptions
- Document the chosen estimand(s) within the trial protocol
It is important to note that writing the details of an estimand in the protocol is step 7 in the above list. A number of aspects need to be discussed first within the clinical team, before the estimand(s) can then be documented. There may be a desire for clinical personnel to first write the protocol, barring statistical input, and then ask the statistician to write the relevant statistical methods section (sometimes) in a short timeframe. Unless there has been upfront collaboration (including the statistician) and discussion around appropriate estimand specification (and documenting it in the protocol), then such an approach is likely going to lead to significant additional time being required to appropriately document estimand(s) before a protocol can be finalised.
Therefore, it is highly desirable to have the statistician involved early, following an approach like the steps listed above, to work on the protocol within a manageable timeframe. In addition, ICH E9(R1) does state that avoiding or over-simplifying the process of discussing and constructing an estimand risks misalignment between trial objectives, trial design, data collection and method of analysis.
Also, specifying an ‘Estimand’ is a newly proposed (ICH E9(R1)) approach, which will lead to a period of adjustment while companies familiarise themselves with the new process. For instance, it may take time to obtain relevant example estimand specifications that can be referred to/reused going forward (if applicable) in later studies.
Give yourself as much time as you can to work on the protocol, adhering to this new guidance. For those providing statistical support to “clients” it is probably a good idea to warn them that it would be best to involve the statistician early when developing a protocol and writing the statistical methods section.
Considerations around relevant intercurrent events
The definition of a clinically meaningful estimand needs to encompass all intercurrent events that are likely to occur, and which are clinically relevant in a given clinical trial setting. Intercurrent events are events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest. A good start-point would be for the clinical team to consider the following list of example questions.
Are any of the following intercurrent event aspects listed below clinically relevant to the trial, which would have a meaningful influence on the treatment effect to be estimated and reported:
|1||Are there any concomitant medications which are prohibited and/or likely to impact on patient results:|
(i) Relevant when taken any time during the trial?
(ii) Or to a specific time period of interest for instance shortly before or at the time of the efficacy variable assessment?
(iii) Specific type(s) of medication only of concern?
(iv) Only of concern if exceeding a specified duration of use or dose?
|2||Rescue medication (/alternative therapy):|
(i) Are patients allowed to take a specific type of rescue medication? Any type that may impact the treatment results?
|3||Study treatment compliance:|
(i) Minimum amount of time patient should adhere to taking study treatment?
(ii) Would there be an unacceptable deviation from the administration schedule which would be of concern?
(i) If a patient changes treatment/dose group during the study is this an issue that should be taken into account? How? Or is it suitable to analyse the patient as randomised?
(i) Are results after treatment discontinuation useable/valid or should they be excluded from consideration?
(ii) Are only treatment discontinuations for a specific reason an issue? E.g. Due to an adverse event or lack of efficacy?
(i) Are efficacy results at a specific scheduled protocol visit relevant for the efficacy variable and if so, how is this assessment post death to be handled? Are any measures prior to death instead relevant for use?
(ii) Is a patient dying (e.g. consider reason for death) in the trial relevant to interpreting the efficacy variable as having a negative outcome?
(iii) Any other relevant ‘terminal’ type events?
This should not be taken as an exhaustive list, as there may be other concerns specific to your trial that could potentially be relevant to consider. In addition, where such specific criteria are not expected to affect interpretation of the endpoint, they would not need to be addressed as intercurrent events. The clinical team (including the statistician) should discuss and agree (at the study design stage) which are relevant and therefore require expanding on within an estimand(s) definition in the protocol.
Also be aware of any unforeseen (and relevant) intercurrent events that occur as the trial progresses, which may then require a protocol amendment. In addition, note that it is not necessary to use the same ‘strategy’ (see August 2019 article for a defined list of strategies) to address all intercurrent events, and not all strategies will be equally acceptable for regulatory decision making.