JHJason C. Hsu

Jason C. Hsu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Jason C. Hsu has a total of 54 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Identifying the Best Predictive Biomarker in Pharmacogenomics Through Multiple Comparisons With the Best

    by Song Zhai, Judong Shen, Jason C. Hsu, Xinping Cui

    Single gene mutations are increasingly being adopted as clinical biomarkers for the optimal application of various therapeutic areas (such as cancer and cardiovascular disease). A single nucleotide polymorphism (SNP), the most common type of genetic variation in human populations, can affect the abundance and function of gene products at the molecular level. In practice, a single‐predictor model is usually built using the best single‐SNP (or single‐gene) biomarker with the most significant drug–SNP (or drug–gene) association in randomized clinical trials. Current statistical methods for the best single‐SNP biomarker selection rely on a variety of ranking procedures. However, these existing approaches (i) cannot accurately distinguish drug–SNP interaction effects (i.e., predictive effects) from SNP main effects (i.e., prognostic effects); (ii) do not necessarily yield the true ordering or provide a confidence interval for ranking. In this paper, we propose a novel method called Multiple comparisons with the best using Marginal Means (3M). Specifically, 3M first calculates unbiased estimates of SNPs’ predictive effects using marginal means (i.e., least squares means). To select the best predictive SNP (i.e., SNP with the largest unbiased drug–SNP interaction effect), 3M uses a nonparametric bootstrap method to construct constrained simultaneous confidence intervals under the framework of Multiple Comparisons with the Best. Simulation studies demonstrate that our proposed method is more reliable in distinguishing predictive effects from prognostic effects, and thus more powerful to identify the true best predictive SNP than existing methods. Finally, we applied our method to the IMPROVE‐IT (IMProved Reduction of Outcomes: Vytroin Efficacy International Trial) pharmacogenomics genome‐wide association study (GWAS) data to search for the best predictive SNP across the genome. One single best predictive SNP rs114462013 ( STAG1 ) on chromosome 3 was detected, with the association previously identified in the literature.


  • Open Access

    Counterfactual Uncertainty Quantification of Factual Estimand of Efficacy From Before‐and‐After Treatment Repeated Measures Randomized Controlled Trials

    by Xingya Wang, Yang Han, Yushi Liu, Szu‐Yu Tang, Jason C. Hsu

    This article quantifies the uncertainty reduction achievable for counterfactual estimand, and cautions against potential bias when the estimand uses Digital Twins. Posed by Neyman in 1923 (Statistical Science, 1990;5: 465–472) who showed unbiased point estimation from designed factual experiments is possible, counterfactual uncertainty quantification (CUQ) remained an open challenge for about 100 years. The counterfactual efficacy we focus on is the ideal estimand for comparing treatment with control , the expected outcome differential if each patient received both and . Enabled by our new statistical modeling principle called ETZ, we show CUQ is achievable in Randomized Controlled Trials (RCTs) with Before‐and‐After Repeated Measures, common in many therapeutic areas. The CUQ we are able to achieve typically has lower variability than factual UQ. We caution against using predictors with measurement error, which violates regression assumptions and can cause attenuation bias in estimating treatment effects. For traditional medicine and population‐averaged targeted therapy, counterfactual point estimation remains unbiased. However, in both Real Human and Digital Twin approaches, estimating effects in subgroups may suffer attenuation bias.


  • Open Access

    Special issue on multiple comparisons (MCP 2019)

    by Xinping Cui, Thorsten Dickhaus, Jason C. Hsu

  • Subgroups Analysis for Personalized and Precision Medicine Development

    by Yi Liu, Hong Tian, Jason C. Hsu

    Subgroup analysis occurs naturally in personalized medicine. This chapter describes them in the setting of Randomized Controlled Trials (RCTs) for targeted therapies. To personalize medicine is to compare the efficacy of treatment versus control in subgroups and their mixtures. Naturally occurring logical relationships among efficacy in subgroups and their mixtures should be respected. However, for binary and time-to-event outcomes, there has been an oversight in current practice of analyses of efficacy stratified on a biomarker, in that they do not respect such logical relationships. Causes of the illogical analyses are as follows: (a) the use of efficacy measures such as Odds Ratio and Hazard Ratio which are not logic-respecting and (b) incorrect mixing of efficacy measure such as Relative Response (RR) even when they are logic-respecting, due to ignoring the prognostic effect. We show that the path to achieve confident logical inference on efficacy in subgroups and their mixtures is as follows: (1) choose a logic-respecting efficacy measure, (2) model the data and adjust for imbalance using the Least Squares means technique, (3) apply the Subgroup Mixable Estimation principle to infer efficacy in subgroups and their mixtures, which automatically correctly takes the prognostic effect into account.


  • An Overview of Multiple Comparisons

    by Xinping Cui, Thorsten Dickhaus, Ying Ding, Jason C. Hsu

    In this chapter, we introduce some basic concepts and quantities, which will be referred to in several of the subsequent chapters. Furthermore, we provide an overview of the other chapters of this handbook and point the readers to online supplementary material.

  • Rejoinder for discussions on correct and logical causal inference for binary and time‐to‐event outcomes in randomized controlled trials

    by Yi Liu, Bushi Wang, Hong Tian, Jason C. Hsu

    Our paper differs from previous literature in two ways: We think in terms of clinical consequences, what benefits patients, what harms patients. Our main message is: using a not logic‐respecting efficacy measure can potentially harm patients in a randomized controlled trial (RCT), as we prove analytically, and demonstrate with the OAK blood‐based tumor mutational burden (bTMB) study. We follow nature, which mixes effects within each treatment arm. Our secondary message is that following nature to mix within each treatment arm first before calculating any efficacy measure between treatments resolves issues. For example, following natural mixing to prove ratio of time is logic‐respecting avoids the issue that weights of efficacy measures are implicit solution to an equation that depends on the unknown prognostic effect. More importantly, coding subgroup mixable estimation (SEM) by mixing within each treatment arm first and then calculating efficacy will make marginal and conditional efficacy agree, for logic‐respecting efficacy measures (be it a ratio or a difference), no matter the outcome is continuous, binary, or time‐to‐event. One does not have to choose between marginal and conditional.


  • Correct and logical causal inference for binary and time‐to‐event outcomes in randomized controlled trials

    by Yi Liu, Bushi Wang, Miao Yang, Jianan Hui, Heng Xu, Siyoen Kil, Jason C. Hsu

    Targeted therapies tend to have biomarker defined subgroups that derive differential efficacy from treatments. This article corrects three prevailing oversights in stratified analyses comparing treatments in randomized controlled trials (RCTs) with binary and time‐to‐event outcomes: Using efficacy measures such as odds ratio (OR) and hazard ratio (HR) can make a prognostic biomarker appear predictive, targeting wrong patients, because the inference is affected by a confounding/covert factor even with ignorable treatment assignment in an RCT. As shown analytically and with real immunotherapy patient level data, OR and HR cannot meet the causal Estimand requirement of ICH E9R1. Mixing efficacy in subgroups by prevalence, the prevailing practice, can give misleading results also, for any efficacy measured as a ratio. However, mixing relative response (RR) and ratio of median (RoM) survival times by the prognostic effect, the confounding/covert factor hiding in plain sight, will give causal inference in an RCT. Effects in subgroups should not be mixed on the logarithmic scale, because it creates an artificial Estimand for the whole population which changes depending on how the population is divided into subgroups. Current computer package implementations contain all these oversights. Probabilities, including survival curve probabilities, naturally average within each treatment arm by prevalence. The subgroup mixable estimation (SME) principle fixes the oversights by first averaging probabilities (not their logarithms) within each treatment arm, then computing simultaneous confidence intervals for ratio efficacy in subgroups and their mixtures based on rigorous mathematical derivation, to finally provide causal inference in the form of apps.


  • Open Access

    Statistical Design and Analysis of Diagnostic Tests for Mutating Viruses

    by Yang Han, Yujia Sun, Jason C. Hsu, Thomas House, Nick Gent, Ian Hall

    As the SARS-CoV-2 virus mutates, mutations harboured in patients become increasingly diverse. Patients classified into two strains may have overlapping non-variant-defining mutations. Mutation calling by sequencing is relative to a reference genome . As SARS-CoV-2 mutates, tracking emerging mutant strains may become increasingly problematic if the reference genome remains Wuhan-Hu-1, because the comparison then becomes indirect : current dominant strain relative to Wuhan-Hu-1 versus emerging strain relative to Wuhan-Hu-1. The original Thermo Fisher’s TaqPath PCR test, on which the UK has standardized national testing of SARS-CoV-2 primarily, targets Wuhan-Hu-1. PCR targets appear readily updated, as TaqPath 2.0 now targets both currently known and future SARS-CoV-2 mutations, probing the N gene and ORF1ab but not the S gene, with 8 probes instead of the original 3 probes. Going forward, our statistical method can more directly compare current wildtype versus emerging mutants, since our new method can use any pair of probes updated to probe the current wildtype and anticipated mutations. The fact that patients harbour mixtures of mutations allows our statistical methods to potentially catch emerging mutants. Given a PCR test which targets the current dominant strain (current wildtype), our statistical method can potentially directly differentiate the current wildtype from an emerging strain.