DART: Design and Analysis of Randomised TrialsTelephone number: 01223 330385
Email Address: email@example.com
Since September 2020, I have been a Programme Leader Track (MRC Investigator) working as part of the Design and Analysis of Randomised Trials (DART) theme. My research aims to improve clinical trial design through the development of innovative methods that lie in the intersection between optimisation, machine learning and statistics. These methods may result in efficiency gains (i.e. smaller or faster trials) but face several practical barriers (e.g. a high computational cost) to be widely adopted. These innovations include patient-centric trials - i.e. those having an explicit goal of assigning more trial participants to superior treatments (e.g. an efficacious vaccine). My proposed work includes four main objectives: 1) developing computationally feasible innovative trial designs, 2) improving analysis methods of optimal, patient-centric adaptive trials (estimation and testing); 3) designing innovative trial designs in response of specific emerging challenges (including using adaptive experiments to enhance and personalise m-health apps) and 4) promoting update and appropriate application of these novel designs in practice.
I have a Ph. D. in Business Administration and Quantitative Methods at Universidad Carlos III de Madrid in July 2012, with a focus on Stochastic Dynamic Optimization. In 2013, I joined the MRC Biostatistics Unit (BSU) in Cambridge as part of a project on the design of multi-arm multi-stage clinical trials. In 2014, I was awarded the first ever Biometrika post-doctoral fellowship. I co-lead the Adaptive Designs Working Group part of the MRC-NIHR Trials Methodology Research Partnership (TMRP). I have lead a team of statisticians at the BSU and at Papworth Trials Unit Collaboration https://royalpapworth.nhs.uk/research-and-development/papworth-trials-unit-collaboration.I am the Department’s Academic Lead for EDI (Equality, Diversity and Inclusion) supporting the BSU's commitment to diversity.
I am the senior statistician in three national trials that include an adaptive or innovative element. These are two NIHR funded: The PIPAH trial (Positioning Imatinib for Pulmonary Arterial Hypertension) and The NOTACS trial (Nasal High-Flow Oxygen Therapy After Cardiac Surgery) and one MRC funded the STRATOSPHERE trial (Stratified adaptive therapeutic studies in pulmonary arterial hypertension caused by mutations in BMPR2).
The PIPAH trial
The goal of this study is to repurpose the anti-cancer drug Imatinib to treat a rare condition called idiopathic pulmonary arterial hypertension. This is an early stage open label single armed study that is composed of two parts, a dose selection stage and an efficacy stage. This will allow us to assess both the safety, tolerability and efficacy of Imatinib in the novel clinical setting. Both stages take advantage of adaptive designs to answer our hypotheses most efficiently with the expected number of patients. Our dose selection stage design is an example of the Bayesian continual reassessment method where we will combine prior information with observations of treatment toxicities at certain doses in a sequential manner to best identify the most tolerable dose. Our efficacy design follows a classical Simon’s two stage design that has a built-in interim analysis that will allow for early stopping due to futility. Such adaptive designs are essential in these settings where the rarity of the condition restricts our ability to conduct large scale controlled studies.
This trial is a collaboration between the MRC BSU Papworth Statistics Team (including Lauren Bell and Martin Law) and Imperial College London (Prof. Martin Wilkins).
The NOTACS trial
The aim of the NOTACS trial is to compare the efficacy, cost-effectiveness and safety of two types of oxygen therapy in patients at high risk of post-operative pulmonary complication after cardiac surgery. The study uses a composite primary outcome measure of days alive and at home within 90 days of surgery, which captures several important patient-centred outcomes (such as complications, prolonged hospital stay, discharge to any post-acute care nursing facility, hospital readmission, and early death after surgery) in a single measure.
The NOTACS trial is an adaptive, multicentre, parallel group, randomised controlled trial. Traditionally, clinical trials use a fixed sample size which is estimated up-front, before the trial starts. This can be problematic as we often find that there is little pre-existing data to confidently base our sample size calculations on. The adaptive design element of the NOTACS trial is an interim sample size re-estimation. After a fixed number of patients have been recruited and completed follow-up, we will repeat our sample size calculation using information from the data accumulated so far. This will allow us to make better informed decisions about the required sample size for the NOTACS trial to ensure that it is not over- or under-powered.
This trial is a collaboration between the MRC BSU Papworth Statistics Team (including Sarah Dawson as trial statistician).
The STRATOSPHERE Trial
Pulmonary arterial hypertension is a devastating life-limiting disease more likely to affect young women. Patients face daily symptoms, an early death and potentially lung transplantation. 1 in 4 patients has a genetic form of the disease. Treatments do not address the underlying genetic cause of the disease yet. Stratosphere is the first ever trial of treatments aimed at the genetic form of the disease, mutations in a protein called the bone morphogenetic type 2 receptor, using 2 drugs that have shown promise in improving function in cells taken from patients and in animal models; hydroxychloroquine and phenylbutyrate. The trial uses response-adaptive randomisation to assign drugs to patients-with a goal of allocating more patients to a working treatment while preserving the statistical power. As the trial progresses if either drug is showing a bigger effect, the trial will "adapt" to this new information by increasing the proportion of patients who get the drug. The trial will be available all across the UK running from all 7 of the nationally accredited pulmonary hypertension centres.
This trial is a collaboration between the MRC BSU Papworth Statistics Team (including Nina Deliu as trial statistician).
Selected PapersRobertson, D.S., Lee, K.M., Lopez-Kolkovska, B.C. and Villar, Sofia S. (To appear, with discussion)Response-adaptive randomization in clinical trials: from myths to practical considerations
Statistical Science :
Dawson, S.N., Chiu, Y.D., Klein, A.A., Earwaker, M. and Villar, Sofia S. (2022)Effect of high-flow nasal therapy on patient-centred outcomes in patients at high risk of postoperative pulmonary complications after cardiac surgery: a statistical analysis plan for NOTACS, a multicentre adaptive randomised controlled trial.
Trials 23(1): 1-8
Mavrogonatou, L., Sun, Y., Robertson, D.S. and Villar, Sofia S. (2022)A comparison of allocation strategies for optimising clinical trial designs under variance heterogeneity.
Computational Statistics & Data Analysis 176: p.107559
Chien I, Deliu N, Turner R, Weller A, Villar Sofia S., Kilbertus N. (2022)Multi-disciplinary fairness considerations in machine learning for clinical trials
In2022 ACM Conference on Fairness, Accountability, and Transparency 2022 Jun 21 (pp. 906-924). : 906-924
Grayling, M.J., Wason, J.M. and Villar, Sofia S. (2022)Response adaptive intervention allocation in stepped‐wedge cluster randomized trials.
Statistics in medicine 41(6): 1081-1099
Johnson, R., Jackson, C., Presanis, A., Villar, Sofia S. and De Angelis, D. (2022)Quantifying efficiency gains of innovative designs of two-arm vaccine trials for COVID-19 using an epidemic simulation model
Statistics in biopharmaceutical research 14(1): 33-41
Barnett Helen, Villar Sofia S., Geys Helena, Jaki Thomas. (2021)A novel statistical test for treatment differences in clinical trials using a response‐adaptive forward‐looking Gittins Index Rule
Biometrics : To appear
Shen C, Wang Z, Villar Sofia S., Van Der Schaar (2020)M. Learning for dose allocation in adaptive clinical trials with safety constraints
InInternational Conference on Machine Learning - PMLR 2020 Nov 21 : 8730-8740
Williamson SF, Villar Sofia S. (2020) A response‐adaptive randomization procedure for multi‐armed clinical trials with normally distributed outcomes.
Biometrics 76(1): 197-209.
Sofia S. Villar and William F. Rosenberger (2017)Covariate-adjusted response-adaptive randomization for multi-arm clinical trials using a modified forward looking Gittins index rule
Biometrics 74(1): 49-57
Williamson, Faye; Jacko, Peter; Villar, Sofia S.; Jaki, Thomas F. (2016)A Bayesian adaptive design for clinical trials in rare diseases
Computational Statistics and Data Analysis 113 : 136-153
Sofia S. Villar, Jack Bowden and James Wason (2015)Multi-armed Bandit Models for the Optimal Design of Clinical Trials: Benefits and Challenges
Statistical Science : Vol. 30, No. 2, 199–215
Sofia S. Villar, James Wason and Jack Bowden (2015)Response-adaptive Randomization for Multi-arm Clinical Trials using the Forward Looking Gittins Index rule
Biometrics Volume 71, Issue 4: Pages 969–978