A simulation study to evaluate the impact of the number of lesions measured on response assessment

Abstract

The objectives of this study were to evaluate whether the number of lesions that are used to measure tumour burden affects response assessment and inter-rater variability. In order to accomplish this, a simulation study was conducted. Data were generated from a mixed-effects mixture model. Parameter values to input in the model were obtained from the analysis of real data. Response assessments based on 10, five, three, two and one lesion were evaluated. There was little difference between response assessments based on five lesions and response assessments based on 10 lesions. When fewer than five lesions were used to assess response, there were notable differences from the 10 lesion-based response assessment. Basing response assessment on a small number of lesions tends to overestimate response rates and leads to misclassification of patients' response status. Therefore, measuring five lesions per patient appears to sufficiently capture patients' response to therapy. Measuring fewer than five lesions results in the loss of information that may adversely affect clinical trial results as well as patient management.

Figure 1

Metastatic disease to the lungs. Note the smaller lesion (white arrows) has not changed in size from baseline to follow up, while the larger lesion (black arrows) has increased in size.

Figure 2

Flowchart of the simulation process

Figure 3

An example of 10 lesion measurements at baseline (shaded bar) and follow-up (light bar) for two sample patients who were classified as responders

Figure 4

An example of 10 lesion measurements at baseline (shaded bar) and follow-up (light bar) for a patient with progressive disease and a patient with stable disease.

Figure 5

Waterfall plots for a single simulated Phase II study with a low response rate and a single simulated Phase II study with a high response rate