Year: 2007
Type: letter
Abstract: The issues described by Panageas et al. ( 1 ) are important refinements to interpreting progression-free survival results. Some additional considerations influencing the precision of the progression-free survival estimate include the proportion of missing data (missing assessments) and the conventions that may be chosen to impute a missing value. The price for missing an assessment should be the use of a conservative convention, such as censoring at the time of the last previous complete assessment. The authors’ idea of prespecifying assessment intervals based on the tumor type may have more value if the interval chosen has some relation to the “usual” progression expectations for the tumor type. The authors are correct in noting that the progression-free survival value is usually expressed in relation to the upper time limit of the assessment interval unless an interval censoring convention is prespecified. However, the most important requirement in drawing inferences is the availability of an appropriate concurrent comparator arm. Also, to minimize the potential bias that can be introduced by an asymmetric assessment interval between the two study arms, assessment timing should be performed symmetrically. This bias can magnify an apparent outcome difference if a systematic assessment difference of a few days occurs in a study of large sample size and small effect size. Other imbalances such as differing event times between arms for various elements comprising a composite endpoint can also weaken the conclusions.
DOI: 10.1093/jnci/djm040
Author: R. Kane
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W2156850453