Covariate adjustment in Bayesian adaptive randomized controlled trials.

In conventional randomized controlled trials, adjustment for baseline values of covariates known to be at least moderately associated with the outcome increases the power of the trial. Recent work has shown a particular benefit for more flexible frequentist designs, such as information adaptive and adaptive multi-arm designs. However, covariate adjustment has not been characterized within the more flexible Bayesian adaptive designs, despite their growing popularity. We focus on a subclass of these which allow for early stopping at an interim analysis given evidence of treatment superiority. We consider both collapsible and non-collapsible estimands and show how to obtain posterior samples of marginal estimands from adjusted analyses. We describe several estimands for three common outcome types. We perform a simulation study to assess the impact of covariate adjustment using a variety of adjustment models in several different scenarios. This is followed by a real-world application of the compared approaches to a COVID-19 trial with a binary endpoint. For all scenarios, it is shown that covariate adjustment increases power and the probability of stopping the trials early, and decreases the expected sample sizes as compared to unadjusted analyses.

Maternal exposure to pyrethroid insecticides during pregnancy and respiratory allergy symptoms among children participating in the Venda Health Examination of Mothers, Babies and their Environment (VHEMBE).

BACKGROUND: Pyrethroid insecticides use for indoor residual spraying (IRS) in malaria-endemic areas results in high levels of exposure to local populations. Pyrethroids may cause asthma and respiratory allergies but no prior study has investigated this question in an IRS area. METHODS: We measured maternal urinary concentrations of pyrethroid metabolites (cis-DBCA, cis-DCCA, trans-DCCA, 3-PBA) in samples collected at delivery from 751 mothers participating in the Venda Health Examination of Mothers, Babies, and their Environment (VHEMBE), a birth cohort study based in Limpopo, South Africa. At 3.5-year and 5-year follow-up visits, caregivers of 647 and 620 children, respectively, were queried about children's respiratory allergy symptoms based on validated instruments. We applied marginal structural models for repeated outcomes to estimate associations between biomarker concentrations and asthma diagnosis as well as respiratory allergy symptoms at ages 3.5 and 5 years. RESULTS: We found that a10-fold increase in maternal urinary cis-DCCA, trans-DCCA and 3-PBA concentrations were associated with more than a doubling in the risk of doctor-diagnosed asthma (cis-DCCA: RR = 2.1, 95% CI = 1.3, 3.3; trans-DCCA: RR = 2.1, 95% CI = 1.1, 3.9; 3-PBA: RR = 2.4, 95% CI = 1.0, 5.8) and an about 80% increase in the risk of wheezing or whistling in the chest (cis-DCCA: RR = 1.8, 95% CI = 1.1, 3.0; trans-DCCA: RR = 1.7, 95% CI = 1.1, 2.6; 3-PBA: RR = 1.8, 95% CI = 1.0, 3.3) and suspected asthma (cis-DCCA: RR = 1.8, 95% CI = 1.1, 3.1; trans-DCCA: RR = 1.8, 95% CI = 1.1, 2.8). We also observed that higher concentrations of cis-DBCA and 3-PBA were related to increases in the risks of dry cough at night (RR = 3.5, 95% CI = 1.3, 9.5) and seasonal rhinoconjunctivitis (RR = 2.0, 95% CI = 1.1, 3.9), respectively. CONCLUSION: Maternal exposure to pyrethroids may increase the risk of asthma and other respiratory allergy symptoms among preschool children from an IRS area.

Estimating individualized treatment rules in longitudinal studies with covariate-driven observation times.

The sequential treatment decisions made by physicians to treat chronic diseases are formalized in the statistical literature as dynamic treatment regimes. To date, methods for dynamic treatment regimes have been developed under the assumption that observation times, that is, treatment and outcome monitoring times, are determined by study investigators. That assumption is often not satisfied in electronic health records data in which the outcome, the observation times, and the treatment mechanism are associated with patients' characteristics. The treatment and observation processes can lead to spurious associations between the treatment of interest and the outcome to be optimized under the dynamic treatment regime if not adequately considered in the analysis. We address these associations by incorporating two inverse weights that are functions of a patient's covariates into dynamic weighted ordinary least squares to develop optimal single stage dynamic treatment regimes, known as individualized treatment rules. We show empirically that our methodology yields consistent, multiply robust estimators. In a cohort of new users of antidepressant drugs from the United Kingdom's Clinical Practice Research Datalink, the proposed method is used to develop an optimal treatment rule that chooses between two antidepressants to optimize a utility function related to the change in body mass index.

Statistics, philosophy, and health: the SMAC 2021 webconference.

SMAC 2021 was a webconference organized in June 2021. The aim of this conference was to bring together data scientists, (bio)statisticians, philosophers, and any person interested in the questions of causality and Bayesian statistics, ranging from technical to philosophical aspects. This webconference consisted of keynote speakers and contributed speakers, and closed with a round-table organized in an unusual fashion. Indeed, organisers asked world renowned scientists to prepare two videos: a short video presenting a question of interest to them and a longer one presenting their point of view on the question. The first video served as a "teaser" for the conference and the second were presented during the conference as an introduction to the round-table. These videos and this round-table generated original scientific insights and discussion worthy of being shared with the community which we do by means of this paper.

Prenatal exposure to insecticides and child cardiometabolic risk factors in the VHEMBE birth cohort.

As part of malaria control programs, many countries spray dichlorodiphenyltrichloroethane (DDT) or pyrethroid insecticides inside dwellings in a practice called indoor residual spraying that results in high levels of exposure to local populations. Gestational exposure to these endocrine- and metabolism-disrupting chemicals may influence child cardiometabolic health. Methods: We measured the serum concentration of DDT and dichlorodiphenyldichloroethylene (DDE) and urinary concentration of pyrethroid metabolites (cis-DBCA, cis-DCCA, trans-DCCA, 3-PBA) in peripartum samples collected between August 2012 and December 2013 from 637 women participating in the Venda Health Examination of Mothers, Babies and their Environment (VHEMBE), a birth cohort study based in Limpopo, South Africa. We applied marginal structural models to estimate the relationship between biomarker concentrations and child-size (height and weight), adiposity (body mass index [BMI], body fat percentage, waist circumference) and blood pressure at 5 years of age. Results: Maternal concentrations of all four pyrethroid metabolites were associated with lower adiposity including reduced BMI z-scores, smaller waist circumferences, and decreased body fat percentages. Reductions in BMI z-score were observed only among children of mothers with sufficient energy intake during pregnancy (ßcis-DCCA, trans -DCCA=-0.4, 95% confidence interval (CI) = -0.7,-0.1; pinteraction=0.03 and 0.04, respectively) but there was no evidence of effect modification for the other measures of adiposity. Maternal p,p'-DDT concentrations were associated with a reduction in body fat percentage (ß = -0.4%, 95% CI = -0.8,-0.0). Conclusions: Gestational exposure to pyrethroids may reduce adiposity in children at 5 years of age.

Generating community measures of food purchasing activities using store-level electronic grocery transaction records: an ecological study in Montreal, Canada.

OBJECTIVE: Geographic measurement of diets is generally not available at areas smaller than a national or provincial (state) scale, as existing nutrition surveys cannot achieve sample sizes needed for an acceptable statistical precision for small geographic units such as city subdivisions. DESIGN: Using geocoded Nielsen grocery transaction data collected from supermarket, supercentre and pharmacy chains combined with a gravity model that transforms store-level sales into area-level purchasing, we developed small-area public health indicators of food purchasing for neighbourhood districts. We generated the area-level indicators measuring per-resident purchasing quantity for soda, diet soda, flavoured (sugar-added) yogurt and plain yogurt purchasing. We then provided an illustrative public health application of these indicators as covariates for an ecological spatial regression model to estimate spatially correlated small-area risk of type 2 diabetes mellitus (T2D) obtained from the public health administrative data. SETTING: Greater Montreal, Canada in 2012. PARTICIPANTS: Neighbourhood districts (n 193). RESULTS: The indicator of flavoured yogurt had a positive association with neighbourhood-level risk of T2D (1·08, 95 % credible interval (CI) 1·02, 1·14), while that of plain yogurt had a negative association (0·93, 95 % CI 0·89, 0·96). The indicator of soda had an inconclusive association, and that of diet soda was excluded due to collinearity with soda. The addition of the indicators also improved model fit of the T2D spatial regression (Watanabe-Akaike information criterion = 1765 with the indicators, 1772 without). CONCLUSION: Store-level grocery sales data can be used to reveal micro-scale geographic disparities and trends of food selections that would be masked by traditional survey-based estimation.

General regression methods for respondent-driven sampling data.

Respondent-driven sampling is a variant of link-tracing sampling techniques that aim to recruit hard-to-reach populations by leveraging individuals' social relationships. As such, a respondent-driven sample has a graphical component which represents a partially observed network of unknown structure. Moreover, it is common to observe homophily, or the tendency to form connections with individuals who share similar traits. Currently, there is a lack of principled guidance on multivariate modelling strategies for respondent-driven sampling to address peer effects driven by homophily and the dependence between observations within the network. In this work, we propose a methodology for general regression techniques using respondent-driven sampling data. This is used to study the socio-demographic predictors of HIV treatment optimism (about the value of antiretroviral therapy) among gay, bisexual and other men who have sex with men, recruited into a respondent-driven sampling study in Montreal, Canada.

Formulating causal questions and principled statistical answers.

Although review papers on causal inference methods are now available, there is a lack of introductory overviews on what they can render and on the guiding criteria for choosing one particular method. This tutorial gives an overview in situations where an exposure of interest is set at a chosen baseline ("point exposure") and the target outcome arises at a later time point. We first phrase relevant causal questions and make a case for being specific about the possible exposure levels involved and the populations for which the question is relevant. Using the potential outcomes framework, we describe principled definitions of causal effects and of estimation approaches classified according to whether they invoke the no unmeasured confounding assumption (including outcome regression and propensity score-based methods) or an instrumental variable with added assumptions. We mainly focus on continuous outcomes and causal average treatment effects. We discuss interpretation, challenges, and potential pitfalls and illustrate application using a "simulation learner," that mimics the effect of various breastfeeding interventions on a child's later development. This involves a typical simulation component with generated exposure, covariate, and outcome data inspired by a randomized intervention study. The simulation learner further generates various (linked) exposure types with a set of possible values per observation unit, from which observed as well as potential outcome data are generated. It thus provides true values of several causal effects. R code for data generation and analysis is available on www.ofcaus.org, where SAS and Stata code for analysis is also provided.

Adaptive treatment strategies for chronic conditions: shared-parameter G-estimation with an application to rheumatoid arthritis.

Most estimation algorithms for adaptive treatment strategies assume that treatment rules at each decision point are independent from one another in the sense that they do not possess any common parameters. This is often unrealistic, as the same decisions may be made repeatedly over time. Sharing treatment-decision parameters across decision points offers several advantages, including estimation of fewer parameters and the clinical ease of a single, time-invariant decision to implement. We propose a new computational approach to estimation of shared-parameter G-estimation, which is efficient and shares the double robustness of the "unshared" sequential G-estimation. We use this approach to analyze data from the Scottish Early Rheumatoid Arthritis (SERA) Inception Cohort.

Previous incarceration impacts access to hepatitis C virus (HCV) treatment among HIV-HCV co-infected patients in Canada.

INTRODUCTION: The prevalence of hepatitis C virus (HCV) is far higher in prison settings than in the general population; thus, micro-elimination strategies must target people in prison to eliminate HCV. We aimed to examine incarceration patterns and determine whether incarceration impacts HCV treatment uptake among Canadian HIV-HCV co-infected individuals in the direct-acting antiviral (DAA) era. METHODS: The Canadian Co-Infection Cohort prospectively follows HIV-HCV co-infected people from 18 centres. HCV RNA-positive participants with available baseline information on incarceration history were included and followed from 21 November 2013 (when second-generation DAAs were approved by Health Canada) until 30 June 2017. A Cox proportional hazards model was used to assess the effect of time-updated incarceration status on time to treatment uptake, adjusting for patient-level characteristics known to be associated with treatment uptake in the DAA era. RESULTS: Overall, 1433 participants (1032/72% men) were included; 67% had a history of incarceration and 39% were re-incarcerated at least once. Compared to those never incarcerated, previously incarcerated participants were more likely to be Indigenous, earn <$1500 CAD/month, report current or past injection drug use and have poorly controlled HIV. There were 339 second-generation DAA treatment initiations during follow-up (18/100 person-years). Overall, 48% of participants never incarcerated were treated (27/100 person-years) compared to only 31% of previously incarcerated participants (15/100 person-years). Sustained virologic response (SVR) rates at 12 weeks were 95% and 92% respectively. After adjusting for other factors, participants with a history of incarceration (adjusted hazard ratio (aHR): 0.7, 95% CI: 0.5 to 0.9) were less likely to initiate treatment, as were those with a monthly income <$1500 (aHR: 0.7, 95% CI: 0.5 to 0.9) or who reported current injection drug use (aHR: 0.7, 95% CI: 0.4 to 1.0). Participants with undetectable HIV RNA (aHR: 2.1, 95% CI: 1.6 to 2.9) or significant fibrosis (aHR: 1.5, 95% CI: 1.2 to 1.9) were more likely to initiate treatment. CONCLUSIONS: The majority of HIV-HCV co-infected persons had a history of incarceration. Those previously incarcerated were 30% less likely to access treatment in the DAA era even after accounting for several patient-level characteristics. With SVR rates above 90%, HCV elimination may be possible if treatment is expanded for this vulnerable and neglected group.

Disparities in direct acting antivirals uptake in HIV-hepatitis C co-infected populations in Canada.

BACKGROUND: Direct acting antivirals (DAAs) have revolutionized hepatitis C (HCV) treatment with >90% cure rates even in real-world studies, giving hope that HCV can be eliminated. However, for DAAs to have a population-level impact on the burden of HCV disease, treatment uptake needs to be expanded. We investigated temporal trends in HCV treatment uptake and evaluated factors associated with second-generation DAA initiation and efficacy among key HIV-HCV co-infected populations in Canada. METHODS: The Canadian HIV-HCV Co-Infection Cohort Study prospectively follows 1699 participants from 18 centres. Among HCV RNA+ participants, we determined the incidence of HCV treatment initiation per year overall and by key populations between 2007 and 2015. Key populations were based on World Health Organization (WHO) guidelines including: people who actively inject drugs (PWID) (reporting injection drug use, last 6 months); Indigenous people; women and men who have sex with men (MSM). Multivariate Cox models were used to estimate adjusted hazard ratios (aHR) and 2-year probability of initiating second-generation DAAs for each of the key populations. RESULTS: Overall, HCV treatment initiation rates increased from 8 (95% CI, 6-11) /100 person-years in 2013 to 28 (95% CI, 23-33) /100 person-years in 2015. Among 911 HCV RNA + participants, there were 202 second-generation DAA initiations (93% with interferon-free regimens). After adjustment (aHR, 95% CI), active PWID (0.60, 0.38-0.94 compared to people not injecting drugs) and more generally, people with lower income (<$18 000 CAD/year) (0.50, 0.35, 0.71) were less likely to initiate treatment. Conversely, MSM were more likely to initiate 1.95 (1.33, 2.86) compared to heterosexual men. In our cohort, the population profile with the lowest 2-year probability of initiating DAAs was Indigenous, women who inject drugs (5%, 95% CI 3-8%). Not having any of these risk factors resulted in a 35% (95% CI 32-38%) probability of initiating DAA treatment. Sustained virologic response (SVR) rates were >82% in all key populations. CONCLUSION: While treatment uptake has increased with the availability of second-generation DAAs, marginalized populations, already engaged in care, are still failing to access treatment. Targeted strategies to address barriers are needed to avoid further health inequities and to maximize the public health impact of DAAs.

Model validation and selection for personalized medicine using dynamic-weighted ordinary least squares.

Model assessment is a standard component of statistical analysis, but it has received relatively little attention within the dynamic treatment regime literature. In this paper, we focus on the dynamic-weighted ordinary least squares approach to optimal dynamic treatment regime estimation, introducing how its double-robustness property may be leveraged for model assessment, and how quasilikelihood may be used for model selection. These ideas are demonstrated through simulation studies, as well as through application to data from the sequenced treatment alternatives to relieve depression study.

Linear growth trajectories in Zimbabwean infants.

BACKGROUND: Undernutrition in early life underlies 45% of child deaths globally. Stunting malnutrition (suboptimal linear growth) also has long-term negative effects on childhood development. Linear growth deficits accrue in the first 1000 d of life. Understanding the patterns and timing of linear growth faltering or recovery during this period is critical to inform interventions to improve infant nutritional status. OBJECTIVE: We aimed to identify the pattern and determinants of linear growth trajectories from birth through 24 mo of age in a cohort of Zimbabwean infants. DESIGN: We performed a secondary analysis of longitudinal data from a subset of 3338 HIV-unexposed infants in the Zimbabwe Vitamin A for Mothers and Babies trial. We used k-means clustering for longitudinal data to identify linear growth trajectories and multinomial logistic regression to identify covariates that were associated with each trajectory group. RESULTS: For the entire population, the mean length-for-age z score declined from -0.6 to -1.4 between birth and 24 mo of age. Within the population, 4 growth patterns were identified that were each characterized by worsening linear growth restriction but varied in the timing and severity of growth declines. In our multivariable model, 1-U increments in maternal height and education and infant birth weight and length were associated with greater relative odds of membership in the least-growth restricted groups (A and B) and reduced odds of membership in the more-growth restricted groups (C and D). Male infant sex was associated with reduced odds of membership in groups A and B but with increased odds of membership in groups C and D. CONCLUSION: In this population, all children were experiencing growth restriction but differences in magnitude were influenced by maternal height and education and infant sex, birth weight, and birth length, which suggest that key determinants of linear growth may already be established by the time of birth. This trial was registered at clinicaltrials.gov as NCT00198718.

Simulating sequential multiple assignment randomized trials to generate optimal personalized warfarin dosing strategies.

BACKGROUND: Due to the cost and complexity of conducting a sequential multiple assignment randomized trial (SMART), it is desirable to pre-define a small number of personalized regimes to study. PURPOSE: We proposed a simulation-based approach to studying personalized dosing strategies in contexts for which a therapeutic agent's pharmacokinetic and pharmacodynamics properties are well understood. We take dosing of warfarin as a case study, as its properties are well understood. We consider a SMART in which there are five intervention points in which dosing may be modified, following a loading phase of treatment. METHODS: Realistic SMARTs are simulated, and two methods of analysis, G-estimation and Q-learning, are used to assess potential personalized dosing strategies. RESULTS: In settings where outcome modelling may be complex due to the highly non-linear nature of the pharmacokinetic and pharmacodynamics mechanisms of the therapeutic agent, G-estimation provides for which the more promising method of estimating an optimal dosing strategy. Used in combination with the simulated SMARTs, we were able to improve simulated patient outcomes and suggest which patient characteristics were needed to best individually tailor dosing. In particular, our simulations suggest that current dosing should be determined by an individual's current coagulation time as measured by the international normalized ratio (INR), their last measured INR, and their last dose. Tailoring treatment only based on current INR and last warfarin dose provided inferior control of INR over the course of the trial. LIMITATIONS: The ability of the simulated SMARTs to suggest optimal personalized dosing strategies relies on the pharmacokinetic and pharmacodynamic models used to generate the hypothetical patient profiles. This approach is best suited to therapeutic agents whose effects are well studied. CONCLUSION: Prior to investing in a complex randomized trial that involves sequential treatment allocations, simulations should be used where possible in order to guide which dosing strategies to evaluate.