ASA Connect

 View Only

  • 1.  Sample size for one sample surival with median time to achieve 95% CI

    Posted 09-12-2017 10:26

    Hello,

     

    I am helping my friend on this question.

     

    How do we calculate the sample size for a one sample survival data cohort if the median survival time is 44 months to achieve a 95% confidence interval within the range of 42-46 months (i.e. +/- 2 months), with varying accrual time and follow-up time.

     

    Does anyone have a sample size program or calculator to help on this please?

     

    Thanks

    Mary Tan



  • 2.  RE: Sample size for one sample surival with median time to achieve 95% CI

    Posted 09-13-2017 15:40

    Is one allowed to assume that the survival data follow a parametric distribution?

    Confidentiality Notice: This e-mail message, including any attachments, is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply e-mail and destroy all copies of the original message.



    Original Message


  • 3.  RE: Sample size for one sample surival with median time to achieve 95% CI

    Posted 09-14-2017 02:46
    If a parametric model cannot be assumed, the asymptotic results regarding the quantiles of the KM estimator in KF Cheng, Sankhya (1984) may be useful. It is nicely described in Chakraborti and Desu, Statistica Neerlandica (1986), who considered the CI for between-treatment difference in quantiles for 2 distributions using KM estimators.

    --
    David R. Bristol, PhD
    President, Statistical Consulting Services, Inc.
    1-336-293-7771



    Original Message


  • 4.  RE: Sample size for one sample surival with median time to achieve 95% CI

    Posted 09-14-2017 13:13

    A simple answer is to use an exponential distribution with a median of 44.  In addition, you need to specify a distribution for the censoring, e.g., another exponential according to the accrual and follow-up.  Based on such and a specification of the width of the 95% CI, the sample size can be computed (based on the sd of the estimator).  It may also be adjusted by the desired probability of obtaining a confidence interval width less than or equal to the value specified (c.f., SAS PROC POWER onesamplemeans), which will need some more work.  Without this adjustment, the sample size will still be reasonable and the computation is straightforward.

     




    Original Message


  • 5.  RE: Sample size for one sample surival with median time to achieve 95% CI

    Posted 09-14-2017 08:50
    There is no answer to your question: the variance of a median depends on the slope of the survival curve in the neighborhood of 0.5.  If the curve is dropping rapidly at that point then you won't need too many, but if it  has flattened out at .5 you would need a huge sample to get the precision you desire.  The theoretical variance for a median involves f(x), the density at that point, which is another way of saying the same thing.

    Terry Therneau

    ------------------------------
    Terry Therneau
    Assistant Professor, Head, Section of Biostatistics
    Mayo Clinic
    ------------------------------

    Original Message


Related Content