Skip to main content

Table 1 Suggested strategies for defining estimands for core COVID-19 outcomesa

From: Treatment estimands in clinical trials of patients hospitalised for COVID-19: ensuring trials ask the right questions

Objective (outcome in bold). Objectives relate to the effect of treatment if introduced into a healthcare system

Treatment effect

Truncation by death

Treatment discontinuation

Evaluate the effect of treatment on mortality†

Difference in proportion dying by a specific time point (or risk ratio or odds ratio)

NA

Treatment policy strategyb

Evaluate the effect of treatment on the requirement for ventilation/oxygen/ICU as a measure of patient benefit

Difference in proportion affected by a specific time point (or risk ratio or odds ratio)

Composite strategy: death is set as failure

Treatment policy strategyb

Evaluate the effect of treatment on the requirement for ventilation/oxygen/ICU from a healthcare systems perspective

Difference in proportion affected by a specific time point (or risk ratio or odds ratio)

While-alive strategy: data from when the patient is alive is used (e.g. did they require ventilation prior to death?)

Treatment policy strategyb

Evaluate the effect of treatment on the number of days in hospital/on a ventilator/on oxygen/in ICU as a measure of patient benefit

Difference in means or restricted mean time

Composite strategy: outcome is defined as the number of days alive and out of hospital/off a ventilator/off oxygen/out of ICU within a given time period

Treatment policy strategyb

Evaluate the effect of treatment on the number of days in hospital/on a ventilator/on oxygen/in ICU from a healthcare systems perspective

Difference in means or restricted mean time

While-alive strategy: data from when the patient is alive is used (e.g. patients are counted as not on a ventilator from point of death)

Treatment policy strategyb

  1. aOther estimand aspects (treatment, population, other intercurrent events) also need to be specified in order to have fully defined estimands
  2. bCan be implemented using intention-to-treat analysis, where all randomised patients are included, and analysed according to their randomised group
  3. †Effect to individual patients or to healthcare systems as a whole on mortality