To What Extent Is the Association Between Race/Ethnicity and Fetal Growth Restriction Explained by Adequacy of Prenatal Care? A Mediation Analysis of a Retrospectively Selected Cohort

Race/ethnicity is associated with intrauterine growth restriction (IUGR) and small-for-gestational age (SGA) birth. We evaluated the extent to which this association is mediated by adequacy of prenatal care (PNC). A retrospective cohort study was conducted using US National Center for Health Statistics natality files for the years 2011-2017. We performed mediation analyses using a statistical approach that allows for exposure-mediator interaction, and we estimated natural direct effects, natural indirect effects, and proportions mediated. All effects were estimated as risk ratios. Among 23,118,656 singleton live births, the excess risk of IUGR among Black women, Hispanic women, and women of other race/ethnicity as compared with White women was partly mediated by PNC adequacy: 13% of the association between non-Hispanic Black race/ethnicity and IUGR, 12% of the association in Hispanic women, and 10% in other women was attributable to PNC inadequacy. The percentage of excess risk of SGA birth that was mediated was 7% in Black women, 6% in Hispanic women, and 5% in other women. Our findings suggest that PNC adequacy may partly mediate the association between race/ethnicity and fetal growth restriction. In future research, investigators should employ causal mediation frameworks to consider additional factors and mediators that could help us better understand this association.

Cost-Effectiveness of Housing First With Assertive Community Treatment: Results From the Canadian At Home/Chez Soi Trial

OBJECTIVE

The At Home/Chez Soi trial for homeless individuals with mental illness showed scattered-site Housing First with Assertive Community Treatment (ACT) to be more effective than treatment as usual. This study evaluated the cost-effectiveness of Housing First with ACT and treatment as usual.

METHODS

Between October 2009 and June 2011, a total of 950 homeless individuals with serious mental illness were recruited in five Canadian cities: Vancouver, Winnipeg, Toronto, Montreal, and Moncton. Participants were randomly assigned to Housing First (N=469) or treatment as usual (N=481) and followed up for up to 24 months. The intervention consisted of scattered-site Housing First, using rent supplements, with ACT. The treatment-as-usual group had access to all other services. The perspective of society was adopted for the cost-effectiveness analysis. Days of stable housing served as the outcome measure. Retrospective questionnaires captured service use data.

RESULTS

Most (69%) of the costs of the intervention were offset by savings in other costs, such as emergency shelters, reducing the net annual cost of the intervention to about Can$6,311 per person. The incremental cost-effectiveness ratio was Can$41.73 per day of stable housing (95% confidence interval=Can$1.96-$83.70). At up to Can$60 per day, Housing First had more than an 80% chance of being cost-effective, compared with treatment as usual. Cost-effectiveness did not vary by participant characteristics.

CONCLUSIONS

Housing First with ACT appeared about as cost-effective as Housing First with intensive case management for people with moderate needs. The optimal mix between the two remains to be determined.

Finite sample variance estimation for optimal dynamic treatment regimes of survival outcomes

Deriving valid confidence intervals for complex estimators is a challenging task in practice. Estimators of dynamic weighted survival modeling (DWSurv), a method to estimate an optimal dynamic treatment regime of censored outcomes, are asymptotically normal and consistent for their target parameters when at least a subset of the nuisance models is correctly specified. However, their behavior in finite samples and the impact of model misspecification on inferences remain unclear. In addition, the estimators' nonregularity may negatively affect the inferences under some specific data generating mechanisms. Our objective was to compare five methods, two asymptotic variance formulas (adjusting or not for the estimation of nuisance parameters) to three bootstrap approaches, to construct confidence intervals for the DWSurv parameters in finite samples. Via simulations, we considered practical scenarios, for example, when some nuisance models are misspecified or when nonregularity is problematic. We also compared the five methods in an application about the treatment of rheumatoid arthritis. We found that the bootstrap approaches performed consistently well at the cost of longer computational times. The asymptotic variance with adjustments generally yielded conservative confidence intervals. The asymptotic variance without adjustments yielded nominal coverages for large sample sizes. We recommend using the asymptotic variance with adjustments in small samples and the bootstrap if computationally feasible. Caution should be taken when nonregularity may be an issue.

Eliminating Structural Barriers: The Impact of Unrestricted Access on Hepatitis C Treatment Uptake Among People Living With Human Immunodeficiency Virus

BACKGROUND

High costs of direct-acting antivirals (DAAs) have led health-care insurers to limit access worldwide. Using a natural experiment, we evaluated the impact of removing fibrosis stage restrictions on hepatitis C (HCV) treatment initiation rates among people living with human immunodeficiency virus (HIV), and then examined who was left to be treated.

METHODS

Using data from the Canadian HIV-HCV Coinfection Cohort, we applied a difference-in-differences approach. Changes in treatment initiation rates following the removal of fibrosis stage restrictions were assessed using a negative binomial regression with generalized estimating equations. The policy change was then specifically assessed among people who inject drugs (PWID). We then identified the characteristics of participants who remained to be treated using a modified Poisson regression.

RESULTS

Between 2010-2018, there were a total of 585 HCV initiations among 1130 eligible participants. After removing fibrosis stage restrictions, DAA initiations increased by 1.8-fold (95% confidence interval [CI] 1.3-2.4) controlling for time-invariant differences and secular trends. Among PWID the impact appeared even stronger, with an adjusted incidence rate ratio of 3.6 (95% CI 1.8-7.4). However, this increased treatment uptake was not sustained. At 1 year following universal access, treatment rates declined to 0.8 (95% CI .5-1.1). Marginalized participants (PWID and those of indigenous ethnicity) and those disengaged from care were more likely to remain HCV RNA positive.

CONCLUSIONS

After the removal of fibrosis restrictions, HCV treatment initiations nearly doubled immediately, but this treatment rate was not sustained. To meet the World Health Organization elimination targets, the minimization of structural barriers and adoption of tailored interventions are needed to engage and treat all vulnerable populations.

Trajectories of Homeless Shelter Utilization in the At Home/Chez Soi Trial of Housing First

OBJECTIVE

Studies have shown that Housing First, a recovery-oriented housing intervention, is effective in reducing service utilization among homeless individuals with mental illness, but less is known about how Housing First affects patterns of service use over time and about characteristics associated with various utilization trajectories. This analysis aimed to explore latent class trajectories of shelter utilization in a randomized controlled trial of Housing First conducted across five Canadian cities.

METHODS

Data from the At Home/Chez Soi trial were analyzed (N=2,058). Latent class growth analysis was performed using days of shelter utilization to identify trajectories over 24 months. Multinomial logistic regression was used to determine which baseline variables, including treatment group, could predict class membership.

RESULTS

Four shelter use trajectories were identified: consistently low (N=1,631, 79%); mostly low (N=120, 6%); early temporary increase (N=179, 9%); and higher use, late temporary increase (N=128, 6%). Treatment group was a significant predictor of class membership. Those enrolled in Housing First had lower odds of experiencing higher shelter use trajectories (mostly low: odds ratio [OR]=0.50, 95% confidence interval [CI]=0.34-0.72; early temporary increase: OR=0.21, 95% CI=0.15-0.31; higher use, late temporary increase: OR=0.14, 95% CI=0.09-0.22). Other variables associated with trajectory classes included older age and longer time homeless, both of which were associated with higher shelter use.

CONCLUSIONS

Several participant characteristics were associated with different shelter use patterns. Knowledge of variables associated with more favorable trajectories may help to inform service planning and contribute to modeling efforts for homelessness.

Weighted regression analysis to correct for informative monitoring times and confounders in longitudinal studies

We address estimation of the marginal effect of a time-varying binary treatment on a continuous longitudinal outcome in the context of observational studies using electronic health records, when the relationship of interest is confounded, mediated, and further distorted by an informative visit process. We allow the longitudinal outcome to be recorded only sporadically and assume that its monitoring timing is informed by patients' characteristics. We propose two novel estimators based on linear models for the mean outcome that incorporate an adjustment for confounding and informative monitoring process through generalized inverse probability of treatment weights and a proportional intensity model, respectively. We allow for a flexible modeling of the intercept function as a function of time. Our estimators have closed-form solutions, and their asymptotic distributions can be derived. Extensive simulation studies show that both estimators outperform standard methods such as the ordinary least squares estimator or estimators that only account for informative monitoring or confounders. We illustrate our methods using data from the Add Health study, assessing the effect of depressive mood on weight in adolescents.

Adaptive Treatment Strategies With Survival Outcomes: An Application to the Treatment of Type 2 Diabetes Using a Large Observational Database

Sequences of treatments that adapt to a patient's changing condition over time are often needed for the management of chronic diseases. An adaptive treatment strategy (ATS) consists of personalized treatment rules to be applied through the course of a disease that input the patient's characteristics at the time of decision-making and output a recommended treatment. An optimal ATS is the sequence of tailored treatments that yields the best clinical outcome for patients sharing similar characteristics. Methods for estimating optimal adaptive treatment strategies, which must disentangle short- and long-term treatment effects, can be theoretically involved and hard to explain to clinicians, especially when the outcome to be optimized is a survival time subject to right-censoring. In this paper, we describe dynamic weighted survival modeling, a method for estimating an optimal ATS with survival outcomes. Using data from the Clinical Practice Research Datalink, a large primary-care database, we illustrate how it can answer an important clinical question about the treatment of type 2 diabetes. We identify an ATS pertaining to which drug add-ons to recommend when metformin in monotherapy does not achieve the therapeutic goals.

Precision medicine: Statistical methods for estimating adaptive treatment strategies

SERIES EDITORS' NOTE

The beauty of science is that all the important things are unpredictable. Freeman Dyson In the typescript which follows, Moodie and Krakow tackle the topical issue of precision medicine and statistical methods for estimating adaptive treatment strategies. This may be the most difficult typescript in our series so far for non-statisticians to understand. It even has equations! But please bear with the authors and give it a chance. One needs not to understand the equations to get the thrust of the strategy.Precision medicine as we discuss elsewhere, is misnamed. In statistics and mathematics precision refers to getting the same answer again and again. It does not mean getting the correct answer, the term for which is accuracy, not precision. However, precision is the current buzz word so there's no point trying to get this straight. When we think about precision we need to consider two elements, reproducibility and replicability. Reproducibility means you give me your data and computer code and I come to the same conclusion you did. Replicability is another matter. I try to replicate your experiment and hopefully reach the same conclusion. In medicine, replicability is obviously more important than reproducibility but things which cannot be reproduced are unlikely to be replicated.As the authors discuss, one can think about precision medicine as one does a family vacation. A best vacation depends on several co-variates: where you live, your prior travel experiences, advice from family and friends, online reviews, Wikitravel, cost, your travel budget, if you have kids and many other co-variates. Consequently, there is unlikely to be a best vacation for everyone. Yours might be a week at the Ritz Carlton Cancun with dinner at Careyes and ours, a week at the Pfister Hotel in Milwaukee with dinner at Mader's German Restaurant (bring simvastatin). Similarly, it is unlikely there is a best therapy of acute myeloid leukemia, a best donor, a best conditioning regimen, a best posttransplant immune suppressive regimen etc. and certainly no best combination of these co-variates for your patient.The question Moodie and Krakow tackle is how we can determine the best therapy or combination of therapies for someone receiving a haematopoietic cell transplant. Although the default answer is typically: randomized clinical trials are the gold standard, these inform us of the outcome of a cohort of subjects, not individuals. In many instances, although a new therapy may be shown to be better than an old one in a controlled randomized trial the benefit is not uniformly distributed. Some subjects in the experimental cohort may do worse with the new therapy compared with controls, others better. The question is who are the winners and losers? We cannot do a controlled randomized trial of one person. Moodie and Krakow discuss statistical tools to help us sort this out.Again, please do not be put off by the equations; forgetaboutit. The overriding message is not so complex, and important. We are always standing by on twitter @BMTStats to help. But don't confuse us with Match.com. And, by the way, Freeman Dyson was a professor at the Institute for Advanced Studies at Princeton but never got his PhD.Robert Peter Gale, Imperial College London, and Mei-Jie Zhang, Medical College of Wisconsin, Center for International Blood and Marrow Research (CIBMTR).

An Area-Level Indicator of Latent Soda Demand: Spatial Statistical Modeling of Grocery Store Transaction Data to Characterize the Nutritional Landscape in Montreal, Canada

Measurement of neighborhood dietary patterns at high spatial resolution allows public health agencies to identify and monitor communities with an elevated risk of nutrition-related chronic diseases. Currently, data on diet are obtained primarily through nutrition surveys, which produce measurements at low spatial resolutions. The availability of store-level grocery transaction data provides an opportunity to refine the measurement of neighborhood dietary patterns. We used these data to develop an indicator of area-level latent demand for soda in the Census Metropolitan Area of Montreal in 2012 by applying a hierarchical Bayesian spatial model to data on soda sales from 1,097 chain retail food outlets. The utility of the indicator of latent soda demand was evaluated by assessing its association with the neighborhood relative risk of prevalent type 2 diabetes mellitus. The indicator improved the fit of the disease-mapping model (deviance information criterion: 2,140 with the indicator and 2,148 without) and enables a novel approach to nutrition surveillance.

Cost-effectiveness of Housing First Intervention With Intensive Case Management Compared With Treatment as Usual for Homeless Adults With Mental Illness: Secondary Analysis of a Randomized Clinical Trial

IMPORTANCE

In the At Home/Chez Soi trial for homeless individuals with mental illness, the scattered-site Housing First (HF) with Intensive Case Management (ICM) intervention proved more effective than treatment as usual (TAU).

OBJECTIVE

To evaluate the cost-effectiveness of the HF plus ICM intervention compared with TAU.

DESIGN, SETTING, AND PARTICIPANTS

This is an economic evaluation study of data from the At Home/Chez Soi randomized clinical trial. From October 2009 through July 2011, 1198 individuals were randomized to the intervention (n = 689) or TAU (n = 509) and followed up for as long as 24 months. Participants were recruited in the Canadian cities of Vancouver, Winnipeg, Toronto, and Montreal. Participants with a current mental disorder who were homeless and had a moderate level of need were included. Data were analyzed from 2013 through 2019, per protocol.

INTERVENTIONS

Scattered-site HF (using rent supplements) with off-site ICM services was compared with usual housing and support services in each city.

MAIN OUTCOMES AND MEASURES

The analysis was performed from the perspective of society, with days of stable housing as the outcome. Service use was ascertained using questionnaires. Unit costs were estimated in 2016 Canadian dollars.

RESULTS

Of 1198 randomized individuals, 795 (66.4%) were men and 696 (58.1%) were aged 30 to 49 years. Almost all (1160 participants, including 677 in the HF group and 483 in the TAU group) contributed data to the economic analysis. Days of stable housing were higher by 140.34 days (95% CI, 128.14-153.31 days) in the HF group. The intervention cost $14 496 per person per year; reductions in costs of other services brought the net cost down by 46% to $7868 (95% CI, $4409-$11 405). The incremental cost-effectiveness ratio was $56.08 (95% CI, $29.55-$84.78) per additional day of stable housing. In sensitivity analyses, adjusting for baseline differences using a regression-based method, without altering the discount rate, caused the largest change in the incremental cost-effectiveness ratio with an increase to $60.18 (95% CI, $35.27-$86.95). At $67 per day of stable housing, there was an 80% chance that HF was cost-effective compared with TAU. The cost-effectiveness of HF appeared to be similar for all participants, although possibly less for those with a higher number of previous psychiatric hospitalizations.

CONCLUSIONS AND RELEVANCE

In this study, the cost per additional day of stable housing was similar to that of many interventions for homeless individuals. Based on these results, expanding access to HF with ICM appears to be warranted from an economic standpoint.

TRIAL REGISTRATION

isrctn.org Identifier: ISRCTN42520374.

Susceptibility to price discounting of soda by neighbourhood educational status: an ecological analysis of disparities in soda consumption using point-of-purchase transaction data in Montreal, Canada

Introduction

Price discounting is a marketing tactic used frequently by food industries and retailers, but the extent to which education modifies the effect of discounting on the purchasing of unhealthy foods has received little attention. We investigated whether there was a differential association of price discounting of soda with store-level soda purchasing records between 2008 and 2013 by store-neighbourhood education in Montreal, Canada.

Methods

Using data on grocery purchase transactions from a sample of supermarkets, pharmacies, supercentres and convenience stores, we performed an ecological time-series analysis, modelling weekly store-level sales of soda as a function of store-level price discounting, store- and neighbourhood-level confounders and an interaction term between discounting and categorical education in the neighbourhood of each store.

Results

Analysis by store type (n = 18 743, 12 437, 3965 and 49 533 store-weeks for superstores, pharmacies, supercentres and convenience stores, respectively) revealed that the effect measure modification of discounting by neighbourhood education on soda purchasing was lower in stores in the more educated neighbourhoods, most notably in pharmacies: -0.020 [95% confidence interval (CI): -0.028, -0.012] and -0.038 (95% CI: -0.051, -0.025), for middle- and high-education categories, respectively). Weaker effect modification was observed in convenience stores. There was no evidence of effect modification in supercentres or superstores.

Conclusions

Price discounting is an important environmental risk factor for soda purchasing and can widen education inequalities in excess sugar intake across levels of education. Interventions to regulate price discounting warrant further investigation as a public health strategy to improve population nutrition, particularly in lower-education neighbourhoods.

Evaluating Flexible Modeling of Continuous Covariates in Inverse-Weighted Estimators

Correct specification of the exposure model is essential for unbiased estimation in marginal structural models with inverse-probability-of-treatment weights. However, although flexible modeling is commonplace when estimating effects of continuous covariates in outcome models, its use is less frequent in estimation of inverse probability weights. Using simulations, we assess the accuracy of the treatment effect estimates and covariate balance obtained with different exposure model specifications when the true relationship between a continuous, possibly time-varying covariate Lt and the logit of the probability of exposure is nonlinear. Specifically, we compare 4 approaches to modeling the effect of Lt when estimating inverse probability weights: a linear function, the covariate-balancing propensity score, and 2 easy-to-implement flexible methods that relax the assumption of linearity: cubic regression splines and fractional polynomials. Using data from 2 empirical studies, we compare linear exposure models with flexible exposure models to estimate the effect of sustained virological response to hepatitis C virus treatment on the progression of liver fibrosis. Our simulation results demonstrate that ignoring important nonlinear relationships when fitting the exposure model may provide poorer covariate balance and induce substantial bias in the estimated exposure-outcome associations. Analysts should routinely consider flexible modeling of continuous covariates when estimating inverse-probability-of-treatment weights.

Bayesian estimation of the average treatment effect on the treated using inverse weighting

We develop a Bayesian approach to estimate the average treatment effect on the treated in the presence of confounding. The approach builds on developments proposed by Saarela et al in the context of marginal structural models, using importance sampling weights to adjust for confounding and estimate a causal effect. The Bayesian bootstrap is adopted to approximate posterior distributions of interest and avoid the issue of feedback that arises in Bayesian causal estimation relying on a joint likelihood. We present results from simulation studies to estimate the average treatment effect on the treated, evaluating the impact of sample size and the strength of confounding on estimation. We illustrate our approach using the classic Right Heart Catheterization data set and find a negative causal effect of the exposure on 30-day survival, in accordance with previous analyses of these data. We also apply our approach to the data set of the National Center for Health Statistics Birth Data and obtain a negative effect of maternal smoking during pregnancy on birth weight.

Should a propensity score model be super? The utility of ensemble procedures for causal adjustment

In investigations of the effect of treatment on outcome, the propensity score is a tool to eliminate imbalance in the distribution of confounding variables between treatment groups. Recent work has suggested that Super Learner, an ensemble method, outperforms logistic regression in nonlinear settings; however, experience with real-data analyses tends to show overfitting of the propensity score model using this approach. We investigated a wide range of simulated settings of varying complexities including simulations based on real data to compare the performances of logistic regression, generalized boosted models, and Super Learner in providing balance and for estimating the average treatment effect via propensity score regression, propensity score matching, and inverse probability of treatment weighting. We found that Super Learner and logistic regression are comparable in terms of covariate balance, bias, and mean squared error (MSE); however, Super Learner is computationally very expensive thus leaving no clear advantage to the more complex approach. Propensity scores estimated by generalized boosted models were inferior to the other two estimation approaches. We also found that propensity score regression adjustment was superior to either matching or inverse weighting when the form of the dependence on the treatment on the outcome is correctly specified.

Comparison of the predictive performance of adherence measures for virologic failure detection in people living with HIV: a systematic review and pairwise meta-analysis

A critical feature of an adherence assessment tool is its ability to predict virologic failure in people living with HIV (PLHIV). We, therefore, aimed to compare the predictive performance of commonly used adherence measures. We systematically searched MEDLINE, Embase and LILACS up to February 2018, to identify relevant observational studies comparing the effects of any two of the following adherence measurements on virologic outcomes: electronic monitoring, pill count, pharmacy refill, self-report and physician assessment. We analyzed data by pairwise meta-analyzes with a random-effects model. The proportion of virologic failures among non-adherent participants in each adherence measure was used to calculate the odds ratio (OR), with 95% Confidence Intervals (95%CI). Heterogeneity was assessed, with potential causes identified by sensitivity and subgroup analysis. We included 38 studies with individual patient data for 18,010 patients. All possible comparisons between pairs of the five adherence measures were considered and a total of nine comparison groups could be established. Meta-analysis suggested that self-report was a better predictor of virologic failure than pill count when the recall period was within one week (OR: 2.35, 95%CI: 1.07-5.18, p = 0.03). Physician assessment had higher odds of predicting virologic failure than did either self-report (OR: 2.63, 95%CI: 1.37-5.26, p < 0.01) or pharmacy refill (OR: 3.57, 95%CI: 1.69-7.14, p < 0.001). There was no difference in the predictive performance between any of the other measures that we were able to compare (p > 0.05). The combination of multiple measures did not increase the predictive value when compared to any of the measures alone. Low-cost and simple adherence measures such as self-report predict virologic failure better than or equally well as objective measures. Our results suggest that there is no need to use expensive or time-consuming adherence measures when the objective is to identify PLHIV at risk of treatment failure.

Real-world impact of direct acting antiviral therapy on health-related quality of life in HIV/Hepatitis C co-infected individuals

Clinical trial results of direct acting antivirals (DAAs) for the treatment of hepatitis C virus (HCV) have shown improvements in health-related quality of life (HR-QoL). However, the extent to which these results are broadly generalizable to real-world settings is unknown. We investigated the real-world impact of oral DAA therapy on HR-QoL among individuals coinfected with HIV/HCV. We used data from the Canadian HIV/HCV Co-Infection Cohort Study that prospectively follows 1795 participants from 18 centres. Since 2007, clinical, lifestyle, and HR-QoL data have been collected biannually through self-administered questionnaires and chart review. HR-QoL was measured using the EQ-5D instrument. Participants initiating oral DAAs, having at least one visit before treatment initiation and at least one visit after DAA treatment response was ascertained, were included. Successful treatment response was defined as a sustained viral response (SVR). Segmented multivariate linear mixed models were used to evaluate the impact of SVR on HR-QoL, controlling for pretreatment trends. 227 participants met our eligibility criteria, 93% of whom achieved SVR. Before treatment, the EQ-5D utility index decreased 0.6 percentage-point/y (95% CI, -0.9, -0.3) and health state was constant over time. The immediate effect of SVR resulted in an increase of 2.3-units (-0.1, 4.7) in patients' health state and 2.0 percentage-point increase (-0.2, 4.0) in utility index. Health state continued to increase post-SVR by 1.4 units/y (-0.9, 3.7), while utility trends post-SVR plateaued over the observation period. Overall using real-world data, we found modest improvements in HR-QoL following SVR, compared to previously published clinical trials.

New Challenges in HIV Research: Combining Phylogenetic Cluster Size and Epidemiological Data

An exciting new direction in HIV research is centered on using molecular phylogenetics to understand the social and behavioral drivers of HIV transmission. SPOT was an intervention designed to offer HIV point of care testing to men who have sex with men at a community-based site in Montreal, Canada; at the time of testing, a research questionnaire was also deployed to collect data on socio-demographic and behavioral characteristics of participating men. The men taking part in SPOT could be viewed, from the research perspective, as having been recruited via a convenience sample. Among men who were found to be HIV positive, phylogenetic cluster size was measured using a large cohort of HIV-positive individuals in the province of Quebec. The cluster size is likely subject to under-estimation. In this paper, we use SPOT data to evaluate the association between HIV transmission cluster size and the number of sex partners for MSM, after adjusting for the SPOT sampling scheme and correcting for measurement error in cluster size by leveraging external data sources. The sampling weights for SPOT participants were calculated from another study of men who have sex with men in Montreal by fitting a weight-adjusted model, whereas measurement error was corrected using the simulation-extrapolation conditional on covariates approach.