Main | What is an odds ratio in a logistic regression »
Wednesday
Jun122013

What’s the difference between a clinical objective and a clinical outcome?

Introduction

Within this tutorial we shall explore the differences between a clinical objective (a clinical research question) and a clinical outcome (a tangible variable to be measured). We will list the characteristics of a good choice for a clinical outcome, and introduce the notion of a surrogate outcome.

 

A clinical objective

A major pre-requisite for any clinical trial to be successful is a good choice of clinical objective and clinical outcomes, but what is a clinical objective and what is a clinical outcome?

A clinical objective is a qualitative goal that we are trying to achieve. Examples of a clinical objective are:

  • ·         Is this new treatment better than the standard treatment?
  • ·         Are there any side effects with this new treatment?
  • ·         How does the treatment affect a patient’s quality of life?

Keep in mind that as we draw up the objectives for our study then we will want to do more than just show the statistical evidence for these objectives, but we will also want a biological explanation for the results (eg. what is the biological explanation that one treatment is better than another).

Also keep in mind that within a clinical study we may very well have more than one clinical objective (or more than one question that we would like to answer). We often choose one or more objectives as the primary objectives, and the other objectives as secondary objectives. It is of high importance to answer primary objectives, and of lesser importance to answer secondary objectives.

 

A clinical outcome

Once we have determined the clinical objectives for a particular study then we need to determine how we will measure the levels of response in our study. For example in a cancer study we might measure the size of the tumours, and hence we would conclude that one treatment is better than another if there is a significant reduction in the tumour size. Within a good clinical trial there is a lot of thought put in to determine what would be the appropriate choice of outcomes that correspond to a particular objective.

We might also have more than one outcome for each particular objective (eg. if one treatment is better than another then we might measure tumour sizes or we might measure survival times following therapy). If we have more than outcome measure then we also need to keep in mind that each outcome measure will require a different sample size (some outcome measures will be more sensitive than others). In addition different outcome measures might indeed indicate that different study designs be used.

As we think about the different variables that we have within a study it is important to note that a variable might be an outcome variable within one research question, and it might be a predictive variable within another research question. For example in one question we might be interested in predicting a change in tumour size, and in another question we might be interested in knowing whether tumour size is a predictor for long-term survival.

 

Characteristics of a good outcome variable

There are a number of characteristics which are essential in the choice of a good outcome variable.

  • ·         Is the outcome valid?  (For example asking a person how many cigarettes they smoked in the last week might be a valid way of assessing a person’s smoking status)
  • ·         Is the outcome variable relevant to the research question being asked? (Cigarette smoking might have little relevance to whether a person contracts measles)
  • ·         Is the outcome variable quantifiable?  Keep in mind that an outcome variable could be a categorical variable (eg. “did the person die?”) or continuous (eg. size of tumour). Either way we must be able to measure the outcome.
  • ·         Is the outcome objective?  (Will different clinical staff record the same measurement for the same patient, or is some subjective interpretation required?)
  • ·         Is the outcome sensitive?  (A good outcome measure will reflect small changes in the treatment benefit)
  • ·         Is the outcome specific?  (A good outcome variable is not significantly affected by other, confounding variables)
  • ·         Is the outcome precise?  (Does it have small measurement error?)
  • ·         Is the outcome measure well-established and accepted by the scientific community? 

 

What is a surrogate outcome?

A surrogate outcome variable might sometimes be used in place of a more definitive clinical measure. For example in a cancer study we might choose to measure the size of tumors rather than the long-term survival rate. Often the surrogate variable is more focused on disease progression, while the definitive clinical measure is related to clinical benefit. There may be a number of benefits in measuring a surrogate outcome rather than the definitive measure. These include the time, cost, and technical difficulty involved in measuring the definitive measure.

The difficulty with using surrogate outcomes revolves around ensuring that these potential markers are indeed valid representations for our clinical measure. Such a marker needs to have a strong relationship with both (a) the clinical outcome and (b) the treatment benefits to the clinical outcome.

 

Conclusion

Any clinical study revolves around an appropriate choice for the clinical objectives and the clinical outcomes. One of the primary reasons for why some clinical trials have little benefit to the medical community is an inappropriate choice for the objectives and outcomes. A clinical trial can have a number of objectives, where each objective is a question that we would like to answer through the conduct of the clinical trial. A clinical outcome is a variable that we want to quantitatively measure during the process of the trial, where a trial can have one or more clinical outcomes. A surrogate outcome is an outcome that we will measure within a particular trial which we believe is indicative of a more ideal, long-term outcome.

 

If you would like to find out more:

Reader Comments

There are no comments for this journal entry. To create a new comment, use the form below.
Member Account Required
You must have a member account on this website in order to post comments. Log in to your account to enable posting.