Survival probabilities (the Kaplan-Meier method)BMJ 1998; 317 doi: https://doi.org/10.1136/bmj.317.7172.1572 (Published 05 December 1998) Cite this as: BMJ 1998;317:1572
All rapid responses
Further to the Statistical Notes "Survival probabilities (the Kaplan-
Meier method)" by J.M. Bland and D.G. Altman (BMJ, 5 December 1998, Volume
317, page 1572), although I very much appreciated the attempt to explain
to non-statisticians (clinical researchers) the meaning and related
problems of "survival analysis" now very often used in medical research, I
have to draw attention to the figure showing the cumulative survival
probability estimate according to the Kaplan-Meier method.
I have spent much time and effort in explaining and correcting the
curves obtained by clinicians when using some statistical software, and so
I was disappointed to see that this curve does not stop at the time of the
last event (16 months, as shown in the table on the left) but at the time
of the last censored observation (24 months, as also shown in the table).
I think the authors should correct this wrong extrapolation of the
survival curve with sufficient emphasis to avoid the spread of this wrong
graphical representation, which I suspect has been obtained using a very
well known (and otherwise very good) statistical programme.
It is very important to counteract the fact that, more or less
consciously, clinicians tend to carry out the point estimate of the
survival probability curve until the last observation even if the correct
plot is shown, because they are not aware of the lack of precision of the
last part of the curve (only a limited number of patients are still at
risk of the event), or of the fact that it can drop to a value near to
zero if one event occurs or even to zero if the last observation is an
I also think that the authors should have placed greater emphasis on
the fact that survival curves have to be shown with the estimated (95%)
confidence intervals and their meanings (at least at some clinically
relevant times), and that is worth showing the number of patients still at
risk at the time at which each event occurs or at least at clinically
relevant follow-up times (e.g. every year or every six months).
I declare that there is no conflict of interest.
Dr. Bruno M. Cesana
Scientific Direction - Epidemiology Unit
Ospedale Maggiore di Milano, I.R.C.C.S.
Via F. Sforza, 28
20122 Milano, Italy.
Email address: Cesana@telemacus.it
Competing interests: No competing interests