betaFeedback

WorksMathematics

RADISH: A Robust Covariate-Adaptive Trial Design with Historical Control Borrowing

Abstract

Incorporating historical controls into randomized clinical trials can improve statistical efficiency and alleviate enrollment challenges by reducing the need for concurrent controls, particularly when their recruitment is ethically or logistically challenging. These gains depend on comparability between the historical and concurrent controls. Many dynamic borrowing methods assess compatibility for the overall study populations and apply the same borrowing decision to every participant, even when comparability varies with participant covariates. We propose RADISH, Robust Adaptive Discrepancy-Informed Shrinkage for Historical Borrowing, a covariate-adaptive design that assesses compatibility at the individual participant level and incorporates robust borrowing into sequential randomization and final inference. Under the stated regularity conditions, the RADISH estimator is asymptotically normal and supports asymptotically valid Wald inference. Simulation studies, including semi-synthetic simulation studies based on the two real-world randomized trials, show that RADISH improves efficiency and assigns fewer participants to the control group when the historical and concurrent controls are compatible. As incompatibility increases, RADISH progressively reduces borrowing and approaches a concurrent-only design and analysis.

main17.pdf

This paper has been submitted for consideration for publication in Biometrics

Introduction

Randomized clinical trials provide a standard basis for evaluating treatment efficacy but can require substantial time and resources. Two complementary strategies can improve their efficiency. Covariate-adaptive randomization balances prognostic covariates across randomized arms and can improve estimation efficiency without additional enrollment [3]. Historical-control borrowing increases the amount of control information by reusing data from completed studies (Pocock, 1976; Viele et al., 2014). This strategy is particularly useful when concurrent-control recruitment is ethically or logistically difficult (Ghadessi et al., 2020). Integrating these strategies makes historical borrowing part of trial design, so the borrowed information guides both sequential allocation and final inference according to the degree of compatibility between historical and concurrent controls.

A central consideration in historical borrowing is systematic bias arising from differences in participant selection and trial conduct between historical and concurrent populations. Such bias can compromise Type I error control and the interpretation of the final analysis (Viele et al., 2014; Calderazzo, Wiesenfarth, and Kopp-Schneider, 2024). This concern has motivated a large literature on Bayesian dynamic borrowing where borrowing is regulated through one or more parameters that are prespecified, estimated empirically, or assigned a hyperprior within a fully Bayesian framework (Lesaffre et al., 2024). Prominent examples include the power prior, which scales the historical likelihood (Ibrahim et al., 2015); the commensurate prior, which captures between-trial heterogeneity through a precision hyperparameter (Hobbs et al., 2011); and the unit information prior, which measures borrowed information in single-observation Fisher-information units (Jin and Yin, 2021; Zhang and Yin, 2023; Gu and Yin, 2024). Other adaptive approaches, including self-adapting mixture and elastic priors, use data-driven compatibility functions to regulate cohort-level borrowing (Yang et al., 2023; Jiang, Nie, and Yuan, 2023). A common feature of these approaches is that they assess overall compatibility between historical and concurrent control cohorts and apply the resulting borrowing decision uniformly across participants.

The comparability of historical and concurrent control populations may vary substantially across clinically meaningful patient subgroups, however, even when the studies investigate the same disease and treatment. In osteoporosis trials, for example, vertebral fracture risk depends on prior fracture history, years since menopause, bone mineral density, and other indicators of underlying skeletal vulnerability (Black et al., 2007). Although patients can be matched on these observed prognostic characteristics, residual differences in disease severity, prior treatments, and other unmeasured risk factors may lead to varying degrees of comparability between historical and concurrent controls. Consequently, historical control information may be highly informative for some patient subgroups but less relevant, or potentially misleading, for others. This heterogeneity motivates a flexible borrowing strategy that accounts for subgroup-specific comparability rather than assuming uniform exchangeability across the study population.

When compatibility between historical and concurrent controls differs across participant groups, a global rule may borrow too little for participants whose historical and concurrent data are compatible and thereby lose efficiency. It may also borrow too much for participants whose data differ between the two sources, which can bias treatment-effect estimates. The Covariate-Adaptive Historical Borrowing (CAHB) design of Jin et al. (2023) integrates kernel-based covariate-adaptive design (KBCD) [3] with a Bayesian semiparametric model that allows compatibility to vary across the covariate space.

CAHB uses the estimated historical conditional mean as a fixed prior center, so the borrowing model does not propagate its sampling uncertainty. As a result, CAHB may borrow too much information from an imprecisely estimated historical mean, potentially biasing the treatment-effect estimate when historical and concurrent controls differ. To account for historical estimation uncertainty in local borrowing, we propose Robust Adaptive Discrepancy-Informed Shrinkage for Historical Borrowing (RADISH), a covariate-adaptive clinical trial design that integrates historical borrowing into sequential randomization and final inference. RADISH accounts explicitly for the sampling uncertainty of the estimated historical conditional mean and evaluates its discrepancy from the concurrent control mean relative to the standard errors of both estimates. This allows borrowing to depend on both the precision of historical information and its compatibility with concurrent outcomes. Motivated by two-sided conflict tail probabilities (Marshall and Spiegelhalter, 2007; Presanis et al., 2013), RADISH uses a covariate-specific conflict score to discount historical information as local disagreement increases, and the resulting information ratio guides sequential treatment allocation. At trial completion, a marginal conflict score provides an additional safeguard for borrowing in the final analysis to ensure its robustness. For final inference, we use a kernel-standardized estimator for the average treatment effect in the enrolled cohort, with a variance estimator based on kernel-regression residuals and linear-smoother weights. This approach provides analytical standard errors and confidence intervals. Under the stated regularity conditions, the centered and studentized RADISH estimator converges to a standard normal distribution, supporting asymptotically valid Wald inference.

The remainder of the paper is organized as follows. § 1.2 develops the RADISH trial design, covering the three-stage design and its large-sample properties. § 2 compares RADISH with KBCD and CAHB through simulation studies. § 2.2 presents semi-synthetic simulation studies based on data from the HORIZON pivotal fracture trial (Black et al., 2007) and fracture intervention trial (FIT) (Black et al., 1996; Cummings et al., 1998). § 2.3 concludes and outlines extensions to non-Gaussian endpoints, multi-source borrowing, and covariate overlap. Code is publicly available at https://github.com/BobZhangHT/HistInfoBorrow.

Methodology: The RADISH Design

Notation and Overview

RADISH is designed for a two-arm randomized clinical trial that sequentially enrolls participants from a target population. It uses an external historical control dataset to inform both treatment allocation and final inference. Let DH={(XjH,YjH)}j=1NH\mathcal{D}_H = \{(\mathbf{X}_j^H, Y_j^H)\}_{j=1}^{N_H} denote the historical control data, where XjH∈Rd\mathbf{X}_j^H \in\mathbb{R}^d is a baseline covariate vector and YjH∈RY_j^H \in\mathbb{R} is a continuous outcome. Let Dn={(Xi,Yi,Zi)}i=1n\mathcal{D}_n = \{(\mathbf{X}_i, Y_i, Z_i)\}_{i=1}^n denote the concurrent trial data observed up to interim step nn. The treatment indicator is Zi=1Z_i = 1 for the treatment arm and Zi=0Z_i = 0 for the concurrent control arm. The superscript HH distinguishes historical observations from concurrent ones throughout. We write Iz,n={i≤n:Zi=z}\mathcal{I}_{z,n} = \{i \leq n : Z_i = z\} for the index set of arm z∈{0,1}z \in\{0,1\} at step nn, and NzN_z for its final size at n=Nn = N, so that N0+N1=NN_0 + N_1 = N.

Let Yi(1)Y_i(1) and Yi(0)Y_i(0) denote the potential outcomes under treatment and control for participant ii in the current trial. We assume consistency, so that Yi=ZiYi(1)+(1−Zi)Yi(0)Y_i = Z_iY_i(1) + (1 - Z_i)Y_i(0), and write the conditional potential-outcome mean as μz(x)=E{Yi(z)∣Xi=x}\mu_z(x) = \mathbb{E}\{Y_i(z) \mid X_i = x\} for z∈{0,1}z \in\{0,1\}.

Our primary causal estimand is the average treatment effect standardized to the enrolled trial cohort,

ΔN=1N∑i=1N{μ1(Xi)−μ0(Xi)}.(1)\Delta_N = \frac{1}{N}\sum_{i=1}^{N}\{\mu_1(X_i) - \mu_0(X_i)\}. \tag*{(1)}

This is a sample-specific estimand defined by the covariate distribution of the participants enrolled in the trial. Conditional on X1,…,XN\mathbf{X}_1,\ldots,\mathbf{X}_N, ΔN\Delta_N is fixed, so the variance describes randomization and outcome uncertainty for the enrolled covariates.

The design proceeds in three stages. Stage I constructs a covariate-specific historical reference and quantifies its uncertainty. Stage II compares this reference with the accrued concurrent controls and uses the resulting local borrowing information to guide sequential randomization. After enrollment is complete, Stage III combines concurrent and discounted historical control information to estimate ΔN\Delta_N and conduct Wald inference. Supplementary Figure S1 and Algorithm 1 summarize this procedure.

For local borrowing, we adopt a Gaussian product kernel KhK_h with bandwidth vector h∈R+d\boldsymbol{h} \in\mathbb{R}_{+}^{d}. For a covariate profile x\boldsymbol{x}, define wjH(x)=Kh(XjH−x)w_j^H(\boldsymbol{x}) = K_h(\boldsymbol{X}_j^H - \boldsymbol{x}) and wi(x)=Kh(Xi−x)w_i(\boldsymbol{x}) = K_h(\boldsymbol{X}_i - \boldsymbol{x}), so that observations with covariates closer to x\boldsymbol{x} receive greater weight. The same kernel is used to construct the historical reference, summarize concurrent outcomes, determine local arm information in the randomization rule, and form the final kernel-standardized estimator. The bandwidth is computed at the planning stage and then fixed across all three stages. We use Scott’s rule of thumb (Scott, 2015), hk=NH−1/(d+4)σ^H,kh_k = N_H^{-1/(d+4)}\widehat{\sigma}_{H,k}, where σ^H,k\widehat{\sigma}_{H,k} is the historical sample standard deviation of covariate kk. If σ^H,k<10−12\widehat{\sigma}_{H,k} < 10^{-12}, replace it by one before calculating hkh_k. Throughout, the similarity weights are evaluated conditionally on the observed covariates, including the historical covariates {XjH}j=1NH\{\boldsymbol{X}_j^H\}_{j=1}^{N_H} and, after enrollment, the concurrent covariates {Xi}i≤n\{\boldsymbol{X}_i\}_{i\le n}. Thus, conditional on these covariates, the kernel weights are fixed, whereas treatment assignments and outcomes remain random.

Within a local neighborhood of a covariate profile x\boldsymbol{x} defined by the kernel weights, we adopt the working Gaussian model

Y∣Z=z, X≈x∼N{μz(x),σz2(x)},(2)Y \mid Z = z,\ X \approx x \sim\mathcal{N}\left\{\mu_z(x), \sigma_z^2(x)\right\}, \tag*{(2)}

where μz(x)\mu_z(\boldsymbol{x}) and σz2(x)\sigma_z^2(\boldsymbol{x}) denote the local conditional mean and variance in arm z∈{0,1}z \in\{0,1\}. For variance estimation and the large-sample theory, we further assume homoskedasticity within each concurrent arm, so that σz2(x)≡σz2\sigma_z^2(\boldsymbol{x}) \equiv\sigma_z^2 for z∈{0,1}z \in\{0,1\}.

Historical controls provide an external reference for the concurrent control mean μ0(x)\mu_0(\boldsymbol{x}). RADISH summarizes the historical data locally by a reference center θ(x)\theta(\boldsymbol{x}) and its estimation variance τH2(x)=Var⁡{θ(x)}\tau_H^2(\boldsymbol{x}) = \operatorname{Var}\{\theta(\boldsymbol{x})\}. The quantity τH2(x)\tau_H^2(\boldsymbol{x}) characterizes the uncertainty in the estimated historical reference, whereas σ02\sigma_0^2 characterizes the variability of individual concurrent control outcomes. The construction of θ(x)\theta(\boldsymbol{x}) and τH2(x)\tau_H^2(\boldsymbol{x}) is described in Stage I.

Stage I: Construction of the Local Historical Reference

Stage I summarizes the historical control data as a covariate-specific reference for the concurrent control mean. We assume that the historical outcomes are conditionally independent given their covariates and adopt the local-constant Gaussian working model YjH∣XjH≈x∼N{μH(x),σH2(x)}Y_j^H \mid X_j^H \approx x \sim\mathcal{N}\{\mu_H(x),\sigma_H^2(x)\}. Under this model, maximizing the kernel-weighted log-likelihood yields the local historical mean

θ(x)=∑j=1NHSH,j(x)YjH,SH,j(x)=wjH(x)∑ℓ=1NHwℓH(x)\theta(x) = \sum_{j=1}^{N_H} S_{H,j}(x)Y_j^H,\qquad S_{H,j}(x) = \frac{w_j^H(x)}{\sum_{\ell=1}^{N_H} w_\ell^H(x)}

Thus, θ(x)\theta(x) is the Nadaraya–Watson estimator of μH(x)\mu_H(x).

Under the working model, the reference variance is τH2(x)=σH2(x)∑jSH,j(x)2\tau_H^2(x) = \sigma_H^2(x)\sum_j S_{H,j}(x)^2. We use the weighted residual variance σ^H2(x)=∑jSH,j(x){YjH−θ(x)}2\widehat{\sigma}_H^2(x) = \sum_j S_{H,j}(x)\{Y_j^H-\theta(x)\}^2 and set τ^H2(x)=σ^H2(x)∑jSH,j(x)2\widehat{\tau}_H^2(x) = \widehat{\sigma}_H^2(x)\sum_j S_{H,j}(x)^2. For a fixed local outcome variance, broadly distributed weights give a more precise historical reference than weights concentrated on a few observations. After each calculation, local and residual variance estimates are floored at ϵv=10−12\epsilon_v = 10^{-12} for numerical stability. Algorithm S1 (Supplementary § 5.7) constructs the historical reference {θ(x),τ^H2(x)}\{\theta(x),\widehat{\tau}_H^2(x)\} used in subsequent stages.

Stage II: Randomization with Robust Borrowing

For each incoming covariate profile xx, Stage II discounts historical precision using a local historical–concurrent conflict score motivated by prior-data conflict diagnostics (Evans and Moshonov, 2006; Nott et al., 2020). Sequential updates use observed outcomes from previously enrolled participants. To assess compatibility, we calculate local concurrent-control summaries using the kernel function, conditional on the observed control covariates {Xi}i∈I0,n\{X_i\}_{i\in I_{0,n}}. Define the local kernel mass in arm z∈{0,1}z \in\{0,1\} by Mz,n(x)=∑i∈Iz,nwi(x)M_{z,n}(x) = \sum_{i\in I_{z,n}} w_i(x). For i∈I0,ni \in\mathcal{I}_{0,n}, write S0,n,i(x)=wi(x)/M0,n(x)S_{0,n,i}(x) = w_i(x)/M_{0,n}(x). The local summaries are

References

  1. [1]Andrews, D. W. K. (1995). Nonparametric kernel estimation for semiparametric models. Econometric Theory, 11(3):560–596.DOI
  2. [2]Härdle, W. (1990). Applied Nonparametric Regression. Number 19 in Econometric Society Monographs. Cambridge University Press, Cambridge.
  3. [3]Jiang, F., Ma, Y., and Yin, G. (2018). Kernel-based adaptive randomization toward balance in continuous and discrete covariates. Statistica Sinica, 28(4):2841–2856.DOI
  4. [4]Mack, Y. P. and Silverman, B. W. (1982). Weak and strong uniform consistency of kernel regression estimates. Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete, 61(3):405–415.DOI
  5. [5]Newey, W. K. (1994). Kernel estimation of partial means and a general variance estimator. Econometric Theory, 10(2):233–253.
  6. [6]Smith, R. L. (1984). Properties of biased coin designs in sequential clinical trials. The Annals of Statistics, 12(3):1018–1034.DOI
  7. [7]von Bahr, B. and Esseen, C.-G. (1965). Inequalities for the rth absolute moment of a sum of random variables, 1 ≤ r ≤ 2. The Annals of Mathematical Statistics, 36(1):299–303.DOI
  8. [8]Wand, M. P. and Jones, M. C. (1995). Kernel Smoothing, volume 60 of Monographs on Statistics and Applied Probability. Chapman & Hall, London, 1 edition.

Paper details

vlnc:W_v29q58fz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents