Giving and taking: ethical treatment assignment in controlled trials

The current version of the Declaration of Helsinki states that ‘the benefits, risks, burdens and effectiveness of a new intervention must be tested against those of the best current proven intervention(s) … ’. This wording implies that it is acceptable for patients to be assigned to receive an unproven new intervention and to be denied a best current proven intervention. We assert that patients being invited to participate in controlled trials cannot, ethically, be expected to forego proven beneficial forms of care. Patients being treated in controlled trials should not knowingly be disadvantaged compared with similar patients being treated in usual clinical care, where they have access to beneficial care. In this article, we have tried to separate for discussion ‘the withholding of effective care from trial participants’, ‘informed consent to treatment’, ‘blinding’ and ‘use of placebos’.

Durability of no evidence of disease activity-3 (NEDA-3) in patients receiving cladribine tablets: The CLARITY extension study

Background: No evidence of disease activity (NEDA-3) is a patient-centric outcome increasingly used as the goal of multiple sclerosis treatment. Objective: Determine treatment durability of cladribine tablets beyond 2 years considering the variable bridging interval of 0.1–116.0 weeks between CLARITY and CLARITY Extension. Methods: Between CLARITY and CLARITY Extension, patients transitioned from cladribine tablets 3.5 mg/kg to placebo (CP3.5 group, n = 98) or continued further treatment with cladribine tablets 3.5 mg/kg (CC7.0 group, n = 186). Treatment assignment was randomized and blinded in both CLARITY and CLARITY Extension. Results: The 2-year NEDA-3 in CLARITY Extension (encompassing both years of CLARITY Extension) was 29.6% in the CP3.5 group and 32.8% in the CC7.0 group. There was no evidence that treatment effect differed with varying bridging intervals. For patients in the CP3.5 group with a bridging interval of ⩽48 weeks, 1 year NEDA-3 (the first year of CLARITY Extension) was 44.4% (28/63) compared with 31.4% (11/35) in patients with a bridging interval of >48 weeks. Conclusion: Treatment with cladribine tablets in CLARITY, followed by either placebo or cladribine tablets in CLARITY Extension, produced sustained benefits for NEDA-3 and its constituent elements for a follow up period up to 6 years from CLARITY baseline.

Out of Control: The (Over)Use of Controls in Accounting Research

In the absence of random treatment assignment, the selection of appropriate control variables is essential to designing well-specified empirical tests of causal effects. However, the importance of control variables seems underappreciated in accounting research relative to other methodological issues. Despite the frequent reliance on control variables, the accounting literature has limited guidance on how to select them. We evaluate the evolution in use of control variables in accounting research and discuss some of the issues that researchers should consider when choosing control variables. Using simulations, we illustrate that more control is not always better and that some control variables can introduce bias into an otherwise well-specified model. We also demonstrate other issues with control variables including the effects of measurement error and complications associated with fixed effects. Lastly, we provide practical suggestions for future accounting research.

Modelling hospital outcome: problems with endogeneity

Background Mortality modelling in the critical care paradigm traditionally uses logistic regression, despite the availability of estimators commonly used in alternate disciplines. Little attention has been paid to covariate endogeneity and the status of non-randomized treatment assignment. Using a large registry database, various binary outcome modelling strategies and methods to account for covariate endogeneity were explored. Methods Patient mortality data was sourced from the Australian & New Zealand Intensive Society Adult Patient Database for 2016. Hospital mortality was modelled using logistic, probit and linear probability (LPM) models with intensive care (ICU) providers as fixed (FE) and random (RE) effects. Model comparison entailed indices of discrimination and calibration, information criteria (AIC and BIC) and binned residual analysis. Suspect covariate and ventilation treatment assignment endogeneity was identified by correlation between predictor variable and hospital mortality error terms, using the Stata™ “eprobit” estimator. Marginal effects were used to demonstrate effect estimate differences between probit and “eprobit” models. Results The cohort comprised 92,693 patients from 124 intensive care units (ICU) in calendar year 2016. Patients mean age was 61.8 (SD 17.5) years, 41.6% were female and APACHE III severity of illness score 54.5(25.6); 43.7% were ventilated. Of the models considered in predicting hospital mortality, logistic regression (with or without ICU FE) and RE logistic regression dominated, more so the latter using information criteria indices. The LPM suffered from many predictions outside the unit [0,1] interval and both poor discrimination and calibration. Error terms of hospital length of stay, an independent risk of death score and ventilation status were correlated with the mortality error term. Marked differences in the ventilation mortality marginal effect was demonstrated between the probit and the "eprobit" models which were scenario dependent. Endogeneity was not demonstrated for the APACHE III score. Conclusions Logistic regression accounting for provider effects was the preferred estimator for hospital mortality modelling. Endogeneity of covariates and treatment variables may be identified using appropriate modelling, but failure to do so yields problematic effect estimates.

Liquid biopsy testing: Impact on treatment assignment and survival in a community-based oncology practice—A real-world experience.

3027 Background: Liquid biopsy is a promising, rapid and minimally invasive genetic test examining circulating tumor DNA. It offers a significant potential in selecting signal-matched therapeutic options. Methods: A retrospective chart review was conducted on adult patients with advanced solid cancer whose tumors were tested with the Guardant 360® (Guardant Health) assay between December 2018 and 2019. A follow-up analysis (censor date: 01/06/2021) was carried to assess the actual impact of testing results on treatment assignment and survival. Results: A total of 178 patients underwent testing. Mean age at diagnosis was 65 years. Median (m) Karnofsky Performance Scale was 90% and the majority of patients (89.9%) had ≥ stage III-B disease. Lung (LCa; 50.56%), breast (BCa; 17.42%) and colorectal (CRCa; 7.87%) cancers were the most common cancer types. A positive test was reported in 140/178 patients (78.7%); of those, 105/140 (75%) had an actionable mutation, either with an FDA-approved target-matched therapy (n = 32/105; 30.5%) or with a therapy outside current FDA indication (n = 73/105; 69.5%). In patients with no actionable mutation (n = 35/140; 25%), 85.7% (n = 30/35) had a signal-based clinical trial opportunity. The actual overall signal-based matching rate was 17.8% (24/135; vs. 82.2% no-match rate). Within candidates for FDA-approved treatment, 50% (16/32) received targeted therapy while only 6.9% (5/73) were treated with targeted agents outside current FDA indication: mean matching score (number of matched drugs/number of actionable mutations) was 0.6 (range: 0.33-2) and 0.8 (range: 0.17-2), respectively. Only 10% (3/30) were referred to signal-based clinical trials. Survival analysis of LCa, BCa and CRCa patients with actionable mutations who actually received any therapy (n = 66) revealed post-testing survival advantage for target-matched therapy (n = 22) compared to unmatched therapy (n = 44): overall survival (OS) was longer in the matched cohort (mOS: 13.3 months; 95% CI: 11.8-14.8 vs. 10.7 months; 95% CI: 9-12.4 in unmatched) but did not reach statistical significance ( P = 0.09). Progression free survival (PFS) was significantly longer in patients who received matched therapy (mPFS: 11.3 months; 95% CI: 9.9-12.7 vs. mPFS: 6.8 months; 95% CI: 5.1-8.5 in unmatched; P < 0.05). Conclusions: Implementation of liquid biopsy testing is feasible in community practice and impacts therapeutic choices in patients with advanced malignancies. Receipt of liquid biopsy-generated signal-matched precision therapies conferred added survival benefit compared to unmatched therapy. Larger sample size studies are needed to validate these findings.

Effects of linagliptin vs glimepiride on cognitive performance in type 2 diabetes: results of the randomised double-blind, active-controlled CAROLINA-COGNITION study

Aims/hypothesis Type 2 diabetes, particularly with concomitant CVD, is associated with an increased risk of cognitive impairment. We assessed the effect on accelerated cognitive decline (ACD) of the DPP-4 inhibitor linagliptin vs the sulfonylurea glimepiride in individuals with type 2 diabetes. Methods The CAROLINA-COGNITION study was part of the randomised, double-blind, active-controlled CAROLINA trial that evaluated the cardiovascular safety of linagliptin vs glimepiride in individuals with age ≥40 and ≤85 years and HbA1c 48–69 mmol/mol (6.5–8.5%) receiving standard care, excluding insulin therapy. Participants were randomised 1:1 using an interactive telephone- and web-based system and treatment assignment was determined by a computer-generated random sequence with stratification by center. The primary cognitive outcome was occurrence of ACD at end of follow-up, defined as a regression-based index score ≤16th percentile on either the Mini-Mental State Examination (MMSE) or a composite measure of attention and executive functioning, in participants with a baseline MMSE score ≥24. Prespecified additional analyses included effects on ACD at week 160, in subgroups (sex, age, race, ethnicity, depressive symptoms, cardiovascular risk, duration of type 2 diabetes, albuminuria), and absolute changes in cognitive performance. Participants, caregivers, and people involved in measurements, examinations or adjudication, were all masked to treatment assignment. Results Of 6033 participants recruited from hospital and primary care sites, 3163 (38.0% female, mean age/diabetes duration 64/7.6 years, MMSE score 28.5, HbA1c 54 mmol/mol [7.1%]) represent the CAROLINA-COGNITION cohort. Over median 6.1 years, ACD occurred in 27.8% (449/1618, linagliptin) vs 27.6% (426/1545, glimepiride), OR 1.01 (95% CI 0.86, 1.18). Also, no differences in ACD were observed at week 160 (OR 1.07 [0.91, 1.25]), between treatments across subgroups, or for absolute cognitive changes. Conclusions/interpretation In a large, international outcome trial in people with relatively early type 2 diabetes at elevated cardiovascular risk, no difference in risk for ACD was observed between linagliptin and glimepiride over 6.1 years. Funding This study was sponsored by Boehringer Ingelheim. Trial registration ClinicalTrials.gov NCT01243424. Graphical abstract

Regression Discontinuity Designs with an Ordinal Running Variable: Evaluating the Effects of Extended Time Accommodations for English Language Learners

Regression discontinuity designs are commonly used for program evaluation with continuous treatment assignment variables. But in practice, treatment assignment is frequently based on discrete or ordinal variables. In this study, we propose a regression discontinuity design with an ordinal running variable to assess the effects of extended time accommodations (ETA) for English language learners (ELL). ETA eligibility is determined by ordinal ELL English proficiency categories of National Assessment of Educational Progress data. We discuss the identification and estimation of the average treatment effect, intent-to-treat effect, and the local average treatment effect at the cutoff. We also propose a series of sensitivity analyses to probe the effect estimates' robustness to the choices of scaling functions and cutoff scores, and unmeasured confounding.

Adaptive Treatment Assignment in Experiments for Policy Choice

Standard experimental designs are geared toward point estimation and hypothesis testing, while bandit algorithms are geared toward in‐sample outcomes. Here, we instead consider treatment assignment in an experiment with several waves for choosing the best among a set of possible policies (treatments) at the end of the experiment. We propose a computationally tractable assignment algorithm that we call “exploration sampling,” where assignment probabilities in each wave are an increasing concave function of the posterior probabilities that each treatment is optimal. We prove an asymptotic optimality result for this algorithm and demonstrate improvements in welfare in calibrated simulations over both non‐adaptive designs and bandit algorithms. An application to selecting between six different recruitment strategies for an agricultural extension service in India demonstrates practical feasibility.

A Tutorial on the Use of Differences-in-Differences in Management, Finance, and Accounting

ABSTRACT Context: natural experiments or quasi-experiments have become quite popular in management research. The differences-in-differences (DiD) estimator is possibly the workhorse of these techniques. Objective: the goal of this paper is to provide a tutorial that serves as practical guide for researchers considering using natural experiments to make causal inferences. Methods: we discuss the DiD advantages, concerns, and tests of validity. We also provide an application of the technique, in which we discuss the effect of government guarantees on banks’ degree of risk, using the 2008 financial crisis as a natural experiment. The database used, as well as the Stata and the R scripts containing the analyses, are available as online appendices. Conclusion: DiD may be used to tackle endogeneity concerns when treatment assignment is random.