There he is, sitting in front of you, just been told, the dreaded “Cancer” word has been used. It is instinctive, and he desperately wants to know how to make preparations and when to say goodbyes – “how long more do I have to live?”
At this point a doctor has two choices, tell the truth that you cannot tell, or state a time. There is no situation more critical in medicine where data interpretation becomes a vital component of your communication. Having studied full time for a Masters in Biostatistics, my reply is “I know enough to know that I do not know the answer”.
It is intuitive to think that one understands the principals of stage adjusted survival, and hence can make a prediction of how long a patient can survive based on cancer stage, but exactly how wrong can you get?
An eminent Professor of Cancer in the UK was asked his opinion on how long al-Megrahi , the Lockerbie bomber (of Pan Am Flight 103 that exploded over Lockerbie in Scotland in 1988) had to live after developing prostate cancer. The assessment was that he had less than 3 months and the convicted bomber was given early release. Last week in the Sunday Times, the eminent Professor revised his estimate to state that al-Megrahi could live for another 10 years and that “…it’s embarrassing that he’s gone on for so long”.
The difficulties lie in the understanding median survival and the interpretation of a stage adjusted cancer survival plot. Firstly, they both apply to populations, so there is no statement, prediction or quantification of individual risk. If the median stage adjusted survival is one year, it means that 50% of the population is expected to die within a year, and 50% is expected to live longer than a year – how useful is that? When you look at a survival plot, it gives you the expected survival probability of a population at each time point, for example 80%, 40% and 2% survival at 6 months, 1 year and 10 years respectively. So your patient has stage IV disease, what makes you think that he won’t be in the 2% that survives past 10 years?
Most doctors do not appreciate these statistics, what more can you expect from the average patient with lung cancer, or authorities who request and expect a doctor to be able to predict the survival of an individual patient?
If you find this type of teaching useful and would like to learn more, I run an online statistics course for clinicians and researchers: