Tailoring treatments using treatment effect modification

Abstract

Running title: Tailoring treatment. Word count text: 3336 Word count abstract: 172 Number of references: 69Number of tables: 1 Number of figures: 2Keywords: Randomized controlled trial; nonrandomized study design; observational study design; statistics, effect modification, interaction, generalizability. 2 Conflict of interest statementNone of the authors of this paper have a financial or personal relationship with other people or organisations that could inappropriately influence or bias the content of the paper. Author contributionsAFS drafted the manuscript. OHK, MN, AB, AWH and RHHG provided guidance during initial planning of the paper and during critical revision. AcknowledgementsThis work was supported by Research Focus Areas funding of the Utrecht University, which is a collaboration between the faculties of medicine, science, and veterinary medicine. The funding body had no role in decisions on the design, writing or submission of the manuscript. We want to acknowledge and thank the two anonymous reviewers for their helpful suggestions which markedly improved the manuscript. Prior postings and presentationsThis study and its results have not been previously published, neither has it been presented at conferences. 3 Key Points Clinical studies are designed to provide evidence on average treatment effects. To tailor treatment towards individual patients, the presence or absence of treatment effect modification needs to be systematically elucidated. Generalizability of treatment effects can be tested within the framework of equivalence testing. The type of patient, the presence, the magnitude, and the number of effect modifiers determines whether no further analyses, univariable subgroup analyses, or multivariable subgroup analyses may need to be performed. 4 AbstractApplying results from clinical studies to individual patients can be a difficult process.Using the concept of treatment effect modification (also referred to as interaction), defined as a difference in treatment response between patient groups, we discuss whether and how treatment effects can be tailored to better meet patients' needs. First we argue that, contrary to how most studies are designed, treatment effect modification should be expected. Second, given this expected heterogeneity, a small number of clinically relevant subgroups should be a priori selected, depending on the expected magnitude of effect modification, and prevalence of the patient type. Third, by defining generalizability as the absence of treatment effect modification we show that generalizability can be evaluated within the usual statistical framework of equivalence testing. Fourth, when equivalence cannot be confirmed, we address the need for further analyses, and studies tailoring treatment towards groups of patients with similar response to treatment. Fifth, we argue that to properly frame, the entire body of evidence on effect modification should be quantified in a prior probability. 5

Pragmatic Trials in Genomic Medicine: The Integrating Pharmacogenetics In Clinical Care (I‐PICC) Study

“…Its findings may be less applicable to SLCO1B1 genetic testing for patients who are previous statin users or known statin‐intolerant, the subjects of a prior RCT . Whether the I‐PICC Study experience generalizes to patient populations outside VABHS involves other considerations, including patient characteristics, provider practice patterns, and organizational health system factors. Clinicians and policymakers should consider these elements in determining whether the trial's results will be applicable to their specific contexts …”

Sociodemographic and Health-Related Characteristics of a Safety-Net Patient Population

“…The differences in characteristics between safety-net populations and patients in the general population raise questions about definitions of safety-net institutions based on Medicaid or Medicare distribution. In addition, the transportability of findings from studies in which safety-net populations are unrepresented are questionable [29][30][31][32][33][34][35]. Despite a Federal Act in 1993 to improve representation of vulnerable populations in research [36], little improvement has been reported [37,38].…”

When drug treatments bias genetic studies: Mediation and interaction

“…Similarly, but in a distinct manner, treatment may bias γ * should treatment modify the effect of G on Y . For example, in the top right panel we depict two separate graphs for untreated D = 0 and treated D = 1 subjects, if we find that γ D = 0 ≠ γ D = 1 we say that there is a variant-by-treatment interaction, or equivalently, we say treatment modifies the effect of G on Y [7]. The traditional GWAS model (Eq 1) does not include such a variant-by-environment interaction, and therefore E [ γ *] ≠ y .…”