Studies in variation associated with the measurement of solid tumors

Abstract

The existing system for solid tumor evaluation is criticized and revised criteria are proposed. The foundation for this analysis is a tumor stimulation experiment to determine the distribution associated with measurement of solid tumors. This tumor simulation experiment was conducted among 26 oncologists in order to assess the errors associated with the measurement of solid tumors by means of palpation. Tumor size, shape, and texture were included as study factors. Tumor texture did not appear to affect measurement outcome. Overall, the statistical law governing tumor area measurement is best described as a log-normal distribution. Under this assumption, the sufficient statistic for the evaluation of tumor response or progression is simply the ratio of the tumor areas at each evaluation. The data generalize to permit the construction of statistical power curves for the probability of declaring response or progression, based upon the theoretic tumor/area ratios. Response and progression criteria are related in logarithmic proportion, i.e., a 50% decrease for response corresponds to a 100% increase for progression. The use of more sensitive progression criteria (a 25% increase in solid tumor area), in light of the nature of tumor measurement error, can lead to premature rejection of any therapy that stabilizes or reduces disease. Furthermore, the use of specified cutoffs for progression and response introduce a significant error into clinical evaluations of solid tumors.