Incorporating lesion-to-lesion heterogeneity into early oncology decision making

Abstract

RECISTv1.1 (Response Evaluation Criteria In Solid Tumors) is the most commonly used response grading criteria in early oncology trials. In this perspective, we argue that RECISTv1.1 is ambiguous regarding lesion-to-lesion variation that can introduce bias in decision making. We show theoretical examples of how lesion-to-lesion variability causes bias in RECISTv1.1, leading to misclassification of patient response. Next, we review immune checkpoint inhibitor (ICI) clinical trial data and find that lesion-to-lesion heterogeneity is widespread in ICI-treated patients. We illustrate the implications of ignoring lesion-to-lesion heterogeneity in interpreting biomarker data, selecting treatments for patients with progressive disease, and go/no-go decisions in drug development. Further, we propose that Quantitative Systems Pharmacology (QSP) models can aid in developing better metrics of patient response and treatment efficacy by capturing patient responses robustly by considering lesion-to-lesion heterogeneity. Overall, we believe patient response evaluation with an appreciation of lesion-to-lesion heterogeneity can potentially improve decision-making at the early stage of oncology drug development and benefit patient care.

Keywords: QSP model; RECIST v1.1; dissociated response; lesion-to-lesion heterogeneity; oncology clinical trials.

Figure 1

Example trajectories of patients with a RECISTv1.1 classification of Stable Disease. Each of the circles represents individual lesions in the patients. The aggregate response is represented by arrows. (A) Shows the standard visualization of aggregate response to therapy based on the Sum of Longest Diameters. (B) Shows a patient with a homogeneous response while (C) shows a patient with a heterogeneous response at the lesion level. When aggregate diameter changes by >20% (shown in labelled dotted line), the patient is classified as Progressive Disease. When aggregate diameter changes by < -30% (shown in labelled dotted line), the patient is classified as a Partial Responder. When all lesions have disappeared, the patient is classified as a Completer Responder.

Figure 2

Patients classified as Progressive Disease can be very variable in the response of their target, non-target & appearance of metastatic lesions. Each of the circles represents individual lesions in the patients (Target Lesions: Filled Circles, Non-target Lesions: Open Circles, New Metastatic Lesions:Filled with bold border). The aggregate response is represented by arrows. When aggregate diameter changes by >20% (A) (shown in labelled dotted line), the patient is classified as Progressive Disease. When aggregate diameter changes by < -30% (B) (shown in labelled dotted line), the patient is classified as a Partial Responder. When all lesions have disappeared, the patient is classified as a Completer Responder (C).

Figure 3

Patients with very different trajectories can be classified as Objective Responders. (A) Shows a prototypical Objective Responder. However, (B, C) show patients classified as Objective Responders who may not show such ideal trajectories (shown in red arrow). In (B) a patient who briefly shows reduction is classified as an Objective Responder as duration of response is not accounted for when RECIST response is assigned. In (C) the patient is classified as an Objective Responder, even though they dropped out at the first point due to an Adverse Event (AE). Others who arguably benefit may still not be classified as Objective Responders (shown in green arrows in bottom row). (D) Shows a patient whose tumor has stabilized just above the dSLD < -30% threshold. (E) Shows a patient who shows clear benefit from the therapy as tumor growth is inhibited but will be considered a non-responder as the lesion has not shrunk. (F) Shows a patient with a new metastatic lesion who will be classified as Progressive Disease even though that lesion may shrink on further treatment.

Figure 4

In these figures, patients are ordered from worst aggregate response (greatest dSLD, Patient#1) to the best aggregate response (least dSLD, Patient #200). In (A) we show the dSLD in solid coloring & the variability in individual lesion change in diameters is also show as dots. The red dotted lines represent standard deviation of the lesions change in per patient. In (B) we show the fraction of growing (change in diameter >20%), stable & shrinking (change in diameter < -20%) lesions in each of the patients. Reproduced (11) from with permission.

Figure 5

Hierarchical development of QSP model that provides a framework to incorporate lesion-to-lesion heterogeneity i. A single well-mixed lesion with interactions between tumor & immune system ii. Multiple target lesions within a single Virtual Patient tracked. The multiple target lesions have different growth rates, sizes etc. When ΔSLD>+20%, the patient is classified as PD iii. A stochastic model periodically predicts the probability of non-target driven PD (any one of non-target lesion growth or metastases or drop-out for other reasons). At this stage, the patient can be classified as PD when such an event occurs iv. A Virtual Population with such VPs that is calibrated to be consistent with reported clinical data – such as waterfall charts, RECIST scores, PFS curves.

Figure 6

Treatment beyond progression may control tumor burden after nontarget progression on pembrolizumab. Median tumor burden in patients from an N=1000 simulated trial receiving pembrolizumab beyond progression (red) or salvage chemotherapy (cyan). (A) Patients with target progression without non-target progression or new metastases. (B) Patients with non-target progression or new metastases without target progression. Solid lines represent medians, while shaded regions indicate interquartile ranges. Reproduced from with permission.