TL;DR
This paper highlights a critical ambiguity in the RECIST 1.1 guidelines for assessing tumor response in clinical trials: the selection of "equivocal" lesions at baseline. It argues that this subjectivity can lead to significant inter-reader variability and impact the sensitivity of trial endpoints, offering strategies to mitigate these risks.
What Problem Does This Paper Address?
Assessing tumor burden in oncology clinical trials is crucial for evaluating treatment efficacy. The Response Evaluation Criteria in Solid Tumors (RECIST) is the standard for this, but its 1.1 version, while improved, still leaves considerable room for interpretation, particularly in selecting which lesions to track at the start of a trial (baseline). Radiologists may encounter lesions whose malignancy is uncertain, leading to "equivocal" findings. This ambiguity can result in inconsistent lesion selection between different readers, potentially skewing trial results and affecting the assessment of treatment response. The paper aims to address the question of how radiologists should handle these uncertain lesions at baseline to ensure more reliable and sensitive trial outcomes.
Methods
The paper employs a multi-faceted approach:
- Literature Review and Expert Opinion: It synthesizes existing RECIST guidelines and discusses common practices and challenges faced by radiologists.
- Causal Analysis: It identifies and categorizes the factors contributing to equivocal lesions, distinguishing between uncertainty arising from radiological features and limitations in clinical information.
- Impact Analysis and Simulation: The core of the paper involves a mathematical simulation to quantify the impact of including non-malignant lesions (NMTL) within the "target lesion" (TL) pool on both response-related and progression-related endpoints. This simulation models how the presence of NMTL can alter the thresholds for determining stable disease (SD), partial response (PR), and progressive disease (PD).
- Strategy Development: Based on the analysis and simulations, the authors propose strategies for managing equivocal lesions, differentiating recommendations for phase II (response-focused) versus phase III (progression-focused) trials.
Datasets
Not discussed in the paper. The authors refer to a "Blinded Independent Central Review clinical trials with double reads using RECIST" database for illustrative examples but do not detail specific datasets used for quantitative analysis or simulation.
Key Insights and Ideas
- Equivocal lesions are a significant source of variability: The subjective nature of identifying malignancy at baseline is a major driver of inconsistent lesion selection among radiologists.
- Impact on endpoints is substantial: Including non-malignant lesions as target lesions can mask true treatment responses, leading to false negatives for PR and potentially preventing the identification of a complete response (CR).
- Phase of the trial matters: The optimal strategy for handling equivocal lesions differs between phase II trials (focused on response) and phase III trials (focused on progression).
- Conservative approach is key for response-focused trials: In phase II trials, a strictly conservative approach (avoiding equivocal lesions as targets) is recommended to avoid falsely indicating response.
- Non-target lesion classification is a viable strategy for progression-focused trials: In phase III trials, classifying equivocal lesions as non-target lesions (NTL) can help maximize the chances of detecting disease progression without biasing response assessment.
- Lymph node assessment has specific challenges: The criteria for measuring lymph nodes (10mm short axis) can be less specific, increasing the risk of including non-malignant nodes.
Key Results
- Simulation shows NMTL impact: The mathematical simulation demonstrates that even a relatively small proportion of non-malignant target lesions (e.g., 20% of the total sum of diameters) can significantly alter response and progression thresholds. For instance, it can prevent a complete response and make it harder to detect progression.
- Response-related endpoints are more sensitive to NMTL: The impact of non-malignant target lesions is more pronounced on response-related endpoints (like CR and PR) than on progression-related endpoints.
- Fewer targets amplify the risk: When patients have only a few target lesions, the inclusion of a single equivocal (and potentially non-malignant) lesion has a disproportionately larger impact on the overall assessment.
Limitations and Future Work
- Lack of empirical data on frequency: The authors acknowledge that no studies have specifically investigated the frequency of equivocal lesion inclusion in RECIST applications, relying instead on "radiologists' experience."
- Simulation assumptions: The mathematical simulation relies on assumptions about homogeneous lesion behavior and the impact of NMTL on the total burden.
- Future work: The paper implicitly suggests the need for clearer RECIST guidelines or standardized protocols for handling equivocal lesions at baseline. Further research could involve prospective studies to quantify the actual frequency of equivocal lesion inclusion and its real-world impact on clinical trial outcomes. The authors also highlight the importance of robust image acquisition protocols and adequate clinical information transfer in central reviews to minimize ambiguity.
Summary updated