Eventfree survival eventfree interval and cancerspecific survival

It is common to compare, analyse and report the time to an event which is not death. When doing this it is not always clear what to do with patients who have died before the event. For ease of discussion and without loss of generality we shall consider the event of disease progression to make it easier to understand and discuss the issues.

It is possible to include patients who have died before documented progressive disease by defining an event as progression of disease or death, using the phrase progression-free survival time to describe this outcome measure. Alternatively, it is possible to define an event as progression of disease only, censoring those patients who have died before their disease has progressed, using the phrase progression-free interval to describe this outcome. Which of these is most appropriate depends on the situation. In cancers where only a relatively small proportion of deaths are due to causes other than the cancer (less than 20 per cent, say), then it is safest (and conservative) to focus on the outcome measure progression-free survival. The reason for this is that a proportion of patients who have apparently died before documented progressive disease have probably died from their disease, but their disease has progressed too quickly to allow a formal assessment to be made. Further, a small proportion of patients may have died from toxicity of the treatment, and censoring of these patients could bias the analysis in favour of the more toxic treatment. Therefore, the noise introduced in the analysis by inclusion of patients who have died without documented evidence of progressive disease is likely to be relatively small. For example, in trials of women with advanced ovarian cancer, progression-free survival is widely used as an outcome measure, because the large majority of women with this disease will die from their disease (or its treatment), and only a small proportion will die from other causes.

In contrast, in cancers where a large proportion of deaths are due to causes other than the cancer (more than 50 per cent, say) then there maybe a stronger argument to focus on the outcome measure progression-free interval. The reason for this is that including deaths, a large proportion of which may be from causes other than the cancer under study, may plausibly dilute and hence obscure any difference between treatments and in some situations, through the play of chance, may enhance the difference. For example, in trials of men with early prostate cancer, progression-free interval is widely used as an outcome measure because many of the men with this disease will die from causes other than prostate cancer. As a safeguard against too much misclassification for this outcome measure, the information on cause of death is often examined so that patients who are reported as dying from this disease are included as 'event' even though they may previously have had no reported evidence of progressive disease. Although this maybe helpful, it is widely known that in nearly all countries cause of death is poorly classified.

There are similar difficulties when we consider the choice between the outcome measures overall survival and cancer-specific survival. The reasons for using cancer-specific survival rather than overall survival are the same as those given above for using progression-free interval, in that if there are a reasonable proportion of deaths from non-cancer causes, then this may obscure (or sometimes enhance) any difference between treatments. The principal difficulty with this outcome measure is that it is based on a reliable classification of the cause of death, which often cannot be guaranteed. This is a particular difficulty if unexpected toxicities occur. This problem was observed in the a meta-analysis of randomized trials [12] which compared surgery plus radiotherapy against surgery alone for women with early breast cancer. In this meta-analysis there was clear evidence of an effect of radiotherapy in improving cancer-specific survival. However, in contrast there was no evidence of a difference in overall survival. This was because although the radiotherapy had an impact on the disease and thus prevented some deaths, it was also the cause of some deaths from long-term toxicity from radiotherapy on the cardiovascular system, some of which is irradiated when the breast is being irradiated.

It is important that the approach to analysis is pre-specified in an analysis protocol and the reasons for the choice of outcome measure should be made clear. If possible disease and cause-specific measures should generally be avoided as the primary outcome measure. If they are chosen as the primary outcome measure, then it should be made clear how issues of possible bias and subjectivity are going to be addressed. As a rule, analysis of overall survival should always be performed, whatever the primary outcome measure. In rare circumstances where there is little information at the start of the trial and it is not clear which approach should be adopted it is probably best to perform both analyses. If they broadly agree, then for trials aiming to show a difference between treatments the least extreme result should be emphasized, and for trials aiming to show 'equivalence' between treatments the most extreme result should be emphasized. If the results of these two analyses disagree, then the reasons for the disagreement should be reported and explored. In these situations neither result should probably be emphasized above the other.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment