Article Text

Download PDFPDF

276 Prognostic value of tumor size varies by treatment in a meta-analysis of 15 randomized clinical trials in advanced non-small cell lung cancer across immunotherapy, TKI, and chemotherapy regimens
  1. Jacqueline Buros1,
  2. Krzysztof Sakrejda1,
  3. Daniel Lee1,
  4. Eric Novik1 and
  5. Philip Jewsbury2
  1. 1Generable, Inc., Brooklyn, NY, USA
  2. 2AstraZeneca, Cambridge, UK


Background RECIST1 is commonly used to characterize intermediate outcomes for clinical trials in the context of solid tumors, and it is largely based on a standardized measure of tumor size known as the sum of longest diameters (SLD). In recent years, the FDA has granted accelerated approvals for several new compounds based on improvements in RECIST-based surrogate outcomes like overall response rate and progression-free survival.2 However, there are concerns regarding the robustness of these surrogate endpoints relative to overall survival (OS),3 4 and it is not known whether their prognostic value is similar across TKI, chemotherapy, and immunotherapy regimens.

Methods We have developed a Bayesian meta-analytic joint model for longitudinal SLD and OS in order to predict Phase III outcomes from early Phase II data. We validated this model in extensive simulation studies. The model utilizes a generalized Stein-Fojo equation5 to characterize SLD over time in terms of 3 parameters: f (proportion of tumor that is treatment-susceptible), ks (the decay rate among susceptible cells), and kg (the growth rate among resistant cells). Two quantities [tumor shrinkage (f * ks) and tumor regrowth ((1-f) * kg)] are then associated with survival in the context of a proportional-hazards survival model. We estimated this model using Stan6 on a dataset of >6,000 subjects in 15 randomized clinical trials in advanced non-small cell lung cancer.

Results Both tumor shrinkage and tumor regrowth were found to be associated with OS (HR for tumor shrinkage: median 0.51, 90% CrI 0.42 - 0.61; HR for tumor regrowth: median 1.24, 90% CrI 1.18 - 1.32). There is a stronger association between tumor shrinkage and OS among patients randomized to a PD-1/PD-L1 inhibitor, either as a monotherapy or in combination with a CTLA-4 inhibitor, than among patients in other trial arms (figure 1). By contrast, there were negligible differences across treatment classes in the association between tumor regrowth and OS.

Abstract 276 Figure 1

Hazard associated with SLD submodel parameters varies according to the class of treatment in a joint model for SLD and overall survival with varying association by assigned treatment regimen. The points represent posterior median values per treatment, with lines representing 90% posterior credible intervals (CrI). Two treatment classes demonstrated posterior probability greater than 90% of a non-zero treatment-specific effect for the response term: the combination PD-1/PD-L1 inhibitor + CTLA-4 inhibitor [interaction HR = 0.64 (90% CrI 0.39 - 1.00; posterior probability of HR<1: 95.2%)] and the PD-1/PD-L1 inhibitor alone [interaction HR = 0.62 (90% CrI 0.42 - 0.89; posterior probability of HR<1: 99.2%)].

Conclusions Our results suggest that not all reductions in tumor size are equal. A patient with a certain degree of tumor shrinkage on the PD-1/PD-L1 inhibitor will have lower mortality risk than a patient with a similar degree of shrinkage on the other regimens evaluated. More research is needed to determine whether the result is unique to this particular PD-1/PD-L1 inhibitor, to determine what mechanisms of action mediate these treatment-specific effects, and to develop improved surrogate measures of treatment efficacy.


  1. Eisenhauer EA, Therasse P, Bogaerts J, et al: New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45:228–247.

  2. Of Health USD, Services H, Others: Clinical trial endpoints for the approval of non-small cell lung cancer drugs and biologics: guidance for Industry, 2015

  3. Mandrekar SJ, An M-W, Meyers J, et al: Evaluation of Alternate Categorical Tumor Metrics and Cut Points for Response Categorization Using the RECIST 1.1 Data Warehouse. J Clin Orthod 32:841–850, 2014

  4. Blumenthal GM, Karuri SW, Zhang H, et al: Overall response rate, progression-free survival, and overall survival with targeted and standard therapies in advanced non-small-cell lung cancer: US Food and Drug Administration trial-level and patient-level analyses. J Clin Oncol 33:1008–1014, 2015

  5. Stein WD, Figg WD, Dahut W, et al: Tumor growth rates derived from data for patients in a clinical trial correlate strongly with patient survival: a novel strategy for evaluation of clinical trial data. Oncologist 13:1046–1054, 2008

  6. Carpenter B, Gelman A, Hoffman MD, et al: Stan: A probabilistic programming language [Internet]. jitc-2020-SITC2020.0277.pdf

This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.