From Bias to Confidence: Exploring Quantitative Bias Analysis’ (QBA’s) Role in Matching Adjusted Indirect Treatment Comparison
Author(s)
Isha Mol, MSc1, Yannan Hu, PhD1, Grace Hsu, MSc2, Thomas Leblanc, MD3, Patrick Hlavacek, MPH4, Joseph C. Cappelleri, MPH, MS, PhD5, Haitao Chu, PhD, MD4, Guido Nador, MD6, Isabel Perez Cruz, MD4, Bart Heeg, PhD1;
1Cytel Inc, Rotterdam, Netherlands, 2Cytel Inc, Waltham, MA, USA, 3Duke University School of Medicine, Division of Hematologic Malignancies and Cellular Therapy, Durham, NC, USA, 4Pfizer Inc, New York, NY, USA, 5Pfizer Inc, Groton, CT, USA, 6Pfizer Inc, Tadworth, United Kingdom
1Cytel Inc, Rotterdam, Netherlands, 2Cytel Inc, Waltham, MA, USA, 3Duke University School of Medicine, Division of Hematologic Malignancies and Cellular Therapy, Durham, NC, USA, 4Pfizer Inc, New York, NY, USA, 5Pfizer Inc, Groton, CT, USA, 6Pfizer Inc, Tadworth, United Kingdom
Presentation Documents
OBJECTIVES: With improved data access, external control arms are increasingly used for regulatory and payer submissions via matching adjusted indirect treatment comparisons (MAICs). Despite that, submissions with MAICs often receive critiques regarding the limitations especially in the context of the EU Joint Clinical Assessment. Quantitative bias analysis (QBA) is an analytical framework which quantifies the strength of biases on study estimates. This study illustrates how QBA can ascertain the robustness of MAIC findings despite missingness and confounding by estimating the bias strengths required to nullify study conclusions.
METHODS: A previously published MAIC between elranatamab and the real-world LocoMMotion study in patients with triple-class exposed multiple myeloma was used as a case study. In the study, elranatamab had significantly longer progression-free survival (PFS) and overall survival (OS) than the LocoMMotion treatment basket. However, among the clinically important variables listed by experts, two variables led to concerns, i.e., International Staging System (ISS) which had 14% missing and high-risk cytogenetic which was not reported and therefore excluded from the analysis. “Delta shift” imputation was conducted to assess the plausible missing assumptions with ISS. For the unmeasured confounder (i.e., high-risk cytogenetics), a range of clinically plausible percentages (20-40%) was tested to find a tipping point leading to insignificant results.
RESULTS: For PFS and OS across all ISS imputation scenarios, the MAIC results remained statistically significant. With all high-risk cytogenetic percentages (including the clinically implausible ones), the PFS results were still significant. For OS, only if the high-risk cytogenetic percentage reached 58%, which was regarded as clinically implausible, did the result become insignificant.
CONCLUSIONS: QBA can evaluate the impact of bias due to missing values or unmeasured confounders and assess the severity of these evidence gaps quantitatively. When the bias is estimated to be small, it enhances confidence in the conducted MAICs for decision-making.
METHODS: A previously published MAIC between elranatamab and the real-world LocoMMotion study in patients with triple-class exposed multiple myeloma was used as a case study. In the study, elranatamab had significantly longer progression-free survival (PFS) and overall survival (OS) than the LocoMMotion treatment basket. However, among the clinically important variables listed by experts, two variables led to concerns, i.e., International Staging System (ISS) which had 14% missing and high-risk cytogenetic which was not reported and therefore excluded from the analysis. “Delta shift” imputation was conducted to assess the plausible missing assumptions with ISS. For the unmeasured confounder (i.e., high-risk cytogenetics), a range of clinically plausible percentages (20-40%) was tested to find a tipping point leading to insignificant results.
RESULTS: For PFS and OS across all ISS imputation scenarios, the MAIC results remained statistically significant. With all high-risk cytogenetic percentages (including the clinically implausible ones), the PFS results were still significant. For OS, only if the high-risk cytogenetic percentage reached 58%, which was regarded as clinically implausible, did the result become insignificant.
CONCLUSIONS: QBA can evaluate the impact of bias due to missing values or unmeasured confounders and assess the severity of these evidence gaps quantitatively. When the bias is estimated to be small, it enhances confidence in the conducted MAICs for decision-making.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
MSR90
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference, Missing Data
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Oncology