Can metformin prevent cancer relative to sulfonylureas? A target trial emulation accounting for competing risks and poor overlap via double/debiased machine learning estimators.

There is mounting interest in the possibility that metformin, indicated for glycemic control in type 2 diabetes, has a range of additional beneficial effects. Randomized trials have shown that metformin prevents adverse cardiovascular events, and metformin use has also been associated with reduced cognitive decline and cancer incidence. In this paper, we dig more deeply into whether metformin prevents cancer by emulating target randomized trials comparing metformin to sulfonylureas as first line diabetes therapy using data from Clinical Practice Research Datalink, a U.K. primary care database (1987-2018). We included individuals with diabetes, no prior cancer diagnosis, no chronic kidney disease, and no prior diabetes therapy who initiated metformin (N=93353) or a sulfonylurea (N=13864). In our cohort, the estimated overlap weighted additive separable direct effect of metformin compared to sulfonylureas on cancer risk at 6 years was -1% (.95 CI=-2.2%, 0.1%), which is consistent with metformin providing no direct protection against cancer incidence or substantial protection. The analysis faced two methodological challenges-poor overlap, and pre-cancer death as a competing risk. To address these issues while minimizing nuisance model misspecification, we develop and apply double/debiased machine learning estimators of overlap weighted separable effects in addition to more traditional effect estimates.

A Price Too High: Injury and Assault among Delivery Gig Workers in New York City.

The occupational health burden and mechanisms that link gig work to health are understudied. We described injury and assault prevalence among food delivery gig workers in New York City (NYC) and assessed the effect of job dependence on injury and assault through work-related mechanisms and across transportation modes (electric bike and moped versus car). Data were collected through a 2022 survey commissioned by the NYC Department of Consumer and Worker Protection among delivery gig workers between October and December 2021 in NYC. We used modified Poisson regression models to estimate the adjusted prevalence rate ratio associations between job dependence and injury and assault. Of 1650 respondents, 66.9% reported that food delivery gig work was their main or only job (i.e., fully dependent). About 21.9% and 20.8% of respondents reported being injured and assaulted, respectively. Injury and assault were more than twice as prevalent among two-wheeled drivers, in comparison to car users. Fully dependent respondents had a 1.61 (95% confidence interval (CI) 1.20, 2.16) and a 1.36 (95% CI 1.03, 1.80) times greater prevalence of injury and assault, respectively, than partially dependent respondents after adjusting for age, sex, race and ethnicity, language, employment length, transportation mode, and weekly work hours. These findings suggest that fully dependent food delivery gig workers, especially two-wheeled riders, are highly vulnerable to the negative consequences of working conditions under algorithmic management by the platforms. Improvements to food delivery gig worker health and safety are urgently needed, and company narratives surrounding worker autonomy and flexibility need to be revisited.

A formal causal interpretation of the case-crossover design.

The case-crossover design of Maclure is widely used in epidemiology and other fields to study causal effects of transient treatments on acute outcomes. However, its validity and causal interpretation have only been justified under informal conditions. Here, we place the design in a formal counterfactual framework for the first time. Doing so helps to clarify its assumptions and interpretation. In particular, when the treatment effect is nonnull, we identify a previously unnoticed bias arising from strong common causes of the outcome at different person-times. We analyze this bias and demonstrate its potential importance with simulations. We also use our derivation of the limit of the case-crossover estimator to analyze its sensitivity to treatment effect heterogeneity, a violation of one of the informal criteria for validity. The upshot of this work for practitioners is that, while the case-crossover design can be useful for testing the causal null hypothesis in the presence of baseline confounders, extra caution is warranted when using the case-crossover design for point estimation of causal effects.

Rejoinder: A formal causal interpretation of the case-crossover design.

We thank the discussants for their insightful commentary and questions. In our rejoinder, we extend our analysis to additional settings, control strategies, and sources of bias relevant to how case-crossover is often used in practice, as suggested by multiple discussants. In particular, we consider: control exposures that follow occurrence of events, settings with shared exposure trajectories (which are common in case-crossover studies of effects of air pollution), bias due to non-transient treatment effects, removing bias due to time trends in treatment using control subjects, and extending our results to the continuous time setting. We also take the opportunity to clarify an easily misinterpreted comment we made about collapsibility, which we thank Andersen and Martinussen for highlighting. Throughout, in the spirit of the discussants. contributions, we iterate between general bias formulas and simulations and numerical studies from illustrative simplified scenarios to build insight.

Systematically exploring repurposing effects of antihypertensives.

With availability of voluminous sets of observational data, an empirical paradigm to screen for drug repurposing opportunities (i.e., beneficial effects of drugs on nonindicated outcomes) is feasible. In this article, we use a linked claims and electronic health record database to comprehensively explore repurposing effects of antihypertensive drugs. We follow a target trial emulation framework for causal inference to emulate randomized controlled trials estimating confounding adjusted effects of antihypertensives on each of 262 outcomes of interest. We then fit hierarchical models to the results as a form of postprocessing to account for multiple comparisons and to sift through the results in a principled way. Our motivation is twofold. We seek both to surface genuinely intriguing drug repurposing opportunities and to elucidate through a real application some study design decisions and potential biases that arise in this context.

Estimating optimal dynamic treatment strategies under resource constraints using dynamic marginal structural models.

Methods for estimating optimal treatment strategies typically assume unlimited access to resources. However, when a health system has resource constraints, such as limited funds, access to medication, or monitoring capabilities, medical decisions must account for competition between individuals in resource usage. The problem of incorporating resource constraints into optimal treatment strategies has been solved for point exposures (1), that is, treatment strategies entailing a decision at just one time point. However, attempts to directly generalize the point exposure solution to dynamic time-varying treatment strategies run into complications. We sidestep these complications by targeting the optimal strategy within a clinically defined subclass. Our approach is to employ dynamic marginal structural models to estimate (counterfactual) resource usage under the class of candidate treatment strategies and solve a constrained optimization problem to choose the optimal strategy for which expected resource usage is within acceptable limits. We apply this method to determine the optimal dynamic monitoring strategy for people living with HIV when resource limits on monitoring exist using observational data from the HIV-CAUSAL Collaboration.

Fluid-limiting treatment strategies among sepsis patients in the ICU: a retrospective causal analysis.

OBJECTIVE: In septic patients, multiple retrospective studies show an association between large volumes of fluids administered in the first 24 h and mortality, suggesting a benefit to fluid restrictive strategies. However, these studies do not directly estimate the causal effects of fluid-restrictive strategies, nor do their analyses properly adjust for time-varying confounding by indication. In this study, we used causal inference techniques to estimate mortality outcomes that would result from imposing a range of arbitrary limits ("caps") on fluid volume administration during the first 24 h of intensive care unit (ICU) care. DESIGN: Retrospective cohort study SETTING: ICUs at the Beth Israel Deaconess Medical Center, 2008-2012 PATIENTS: One thousand six hundred thirty-nine septic patients (defined by Sepsis-3 criteria) 18 years and older, admitted to the ICU from the emergency department (ED), who received less than 4 L fluids administered prior to ICU admission MEASUREMENTS AND MAIN RESULTS: Data were obtained from the Medical Information Mart for Intensive Care III (MIMIC-III). We employed a dynamic Marginal Structural Model fit by inverse probability of treatment weighting to obtain confounding adjusted estimates of mortality rates that would have been observed had fluid resuscitation volume caps between 4 L-12 L been imposed on the population. The 30-day mortality in our cohort was 17%. We estimated that caps between 6 and 10 L on 24 h fluid volume would have reduced 30-day mortality by - 0.6 to - 1.0%, with the greatest reduction at 8 L (- 1.0% mortality, 95% CI [- 1.6%, - 0.3%]). CONCLUSIONS: We found that 30-day mortality would have likely decreased relative to observed mortality under current practice if these patients had been subject to "caps" on the total volume of fluid administered between 6 and 10 L, with the greatest reduction in mortality rate at 8 L.

The impact of aggressive and conservative propensity for initiation of neuromuscular blockade in mechanically ventilated patients with hypoxemic respiratory failure.

INTRODUCTION: Neuromuscular blockade (NMB) in ventilated patients may cause benefit or harm. We applied "incremental interventions" to determine the impact of altering NMB initiation aggressiveness. METHODS: Retrospective cohort study of ventilated patients with PaO2/FiO2 ratio < 150 mmHg and PEEP≥ 8cmH2O from the Medical Information Mart of Intensive Care IV database (MIMIC-IV version 1.0) estimating the effect of incremental interventions on in-hospital mortality and ventilator-free days, modifying hourly propensity for NMB initiation to be aggressive or conservative relative to usual care, adjusting for confounding with inverse probability weighting. RESULTS: 5221 patients were included (13.3% initiated on NMB). Incremental interventions estimated a strong effect on NMB usage: 5-fold higher hourly odds of initiation increased usage to 36.5% (CI = [34.3%,38.7%]) and 5-fold lower odds decreased usage to 3.8% (CI = [3.3%,4.3%]). Aggressive and conservative strategies demonstrated a U-shaped mortality relationship. 5-fold higher or lower propensity increased in-hospital mortality by 2.6% (0.95 CI = [1.5%,3.7%]) or 1.3% (0.95 CI = [0.1%,2.5%]) respectively. In secondary analysis of a healthier patient cohort, results were similar, however conservative strategies also improved ventilator-free days. INTERPRETATION: Aggressive or conservative initiation of NMB may worsen mortality. In healthier populations, marginally conservative NMB initiation strategies may lead to increased ventilator free days with minimal impact on mortality.

Effects of aggressive and conservative strategies for mechanical ventilation liberation.

BACKGROUND: The optimal approach for transitioning from strict lung protective ventilation to support modes of ventilation when patients determine their own respiratory rate and tidal volume remains unclear. While aggressive liberation from lung protective settings could expedite extubation and prevent harm from prolonged ventilation and sedation, conservative liberation could prevent lung injury from spontaneous breathing. RESEARCH QUESTION: Should physicians take a more aggressive or conservative approach to liberation? METHODS: Retrospective cohort study of mechanically ventilated patients from the Medical Information Mart for Intensive Care IV database (MIMIC-IV version 1.0) estimating effects of incremental interventions modifying the propensity for liberation to be more aggressive or conservative relative to usual care, with adjustment for confounding via inverse probability weighting. Outcomes included in-hospital mortality, ventilator free days, and ICU free days. Analysis was performed on the entire cohort as well as subgroups differentiated by PaO2/FiO2 ratio, and SOFA. RESULTS: 7433 patients were included. Strategies multiplying the odds of a first liberation relative to usual care at each hour had a large impact on time to first liberation attempt (43 h under usual care, 24 h (0.95 CI = [23,25]) with an aggressive strategy doubling liberation odds, and 74 h (0.95 CI = [69,78]) under a conservative strategy halving liberation odds). In the full cohort, we estimated aggressive liberation increased ICU-free days by 0.9 days (0.95 CI = [0.8,1.0]) and ventilator free days by 0.82 days (0.95 CI = [0.67,0.97]), but had minimal effect on mortality (only a 0.3% (0.95 CI = [-0.2%,0.8%]) difference between minimum and maximum rates). With baseline SOFA≥ 12 (n = 1355), aggressive liberation moderately increased mortality (58.5% [0.95 CI = (55.7%,61.2%)]) compared with conservative liberation (55.1% [0.95 CI = (51.6%,58.6%)]). INTERPRETATION: Aggressive liberation may improve ventilator free and ICU free days with little impact on mortality in patients with SOFA score < 12. Trials are needed.

Titration of Ventilator Settings to Target Driving Pressure and Mechanical Power.

PURPOSE: Driving pressure (ΔP) and mechanical power (MP) may be important mediators of lung injury in acute respiratory distress syndrome (ARDS) however there is little evidence for strategies directed at lowering these parameters. We applied predictive modeling to estimate the effects of modifying ventilator parameters on ΔP and MP. METHODS: 2,622 ARDS patients (Berlin criteria) from the Medical Information Mart for Intensive Care IV database (MIMIC-IV version1.0) admitted to the intensive care unit (ICU) at Beth Israel Deaconess Medical Center between 2008 and 2019 were included. Flexible confounding-adjusted regression models for time varying data were fit to estimate the effects of adjusting PEEP and tidal volume (VT) on ΔP, and adjusting VT and respiratory rate (f) on MP. RESULTS: Reduction in VT reduced ΔP and MP, with more pronounced effect on MP with lower compliance. Strategies reducing f, consistently increased MP (when VT was adjusted to maintain consistent minute ventilation). Adjustment of PEEP yielded a U-shaped effect on ΔP. CONCLUSIONS: This novel conditional modeling confirmed expected response patterns for ΔP, with the response to adjustments depending on patients' lung mechanics. Furthermore a VT -driven approach should be favored over a f -driven approach when aiming to reduce MP.

State of the Art Causal Inference in the Presence of Extraneous Covariates: A Simulation Study.

The central task of causal inference is to remove (via statistical adjustment) confounding bias that would be present in naive unadjusted comparisons of outcomes in different treatment groups. Statistical adjustment can roughly be broken down into two steps. In the first step, the researcher selects some set of variables to adjust for. In the second step, the researcher implements a causal inference algorithm to adjust for the selected variables and estimate the average treatment effect. In this paper, we use a simulation study to explore the operating characteristics and robustness of state-of-the-art methods for step two (statistical adjustment for selected variables) when step one (variable selection) is performed in a realistically sub-optimal manner. More specifically, we study the robustness of a cross-fit machine learning based causal effect estimator to the presence of extraneous variables in the adjustment set. The take-away for practitioners is that there is value to, if possible, identifying a small sufficient adjustment set using subject matter knowledge even when using machine learning methods for adjustment.

Blending Knowledge in Deep Recurrent Networks for Adverse Event Prediction at Hospital Discharge.

Deep learning architectures have an extremely high-capacity for modeling complex data in a wide variety of domains. However, these architectures have been limited in their ability to support complex prediction problems using insurance claims data, such as readmission at 30 days, mainly due to data sparsity issue. Consequently, classical machine learning methods, especially those that embed domain knowledge in handcrafted features, are often on par with, and sometimes outperform, deep learning approaches. In this paper, we illustrate how the potential of deep learning can be achieved by blending domain knowledge within deep learning architectures to predict adverse events at hospital discharge, including readmissions. More specifically, we introduce a learning architecture that fuses a representation of patient data computed by a self-attention based recurrent neural network, with clinically relevant features. We conduct extensive experiments on a large claims dataset and show that the blended method outperforms the standard machine learning approaches.

Finding Causal Mechanistic Drug-Drug Interactions from Observational Data.

Many adverse drug reactions (ADRs) are caused by drug-drug interactions (DDIs), meaning they arise from concurrent use of multiple medications. Detecting DDIs using observational data has at least three major challenges: (1) The number of potential DDIs is astronomical; (2) Associations between drugs and ADRs may not be causal due to observed or unobserved confounding; and (3) Frequently co-prescribed drug pairs that each independently cause an ADR do not necessarily causally interact, where causal interaction means that at least some patients would only experience the ADR if they take both drugs. We address (1) through data mining algorithms pre-filtering potential interactions, and (2) and (3) by fitting causal interaction models adjusting for observed confounders and conducting sensitivity analyses for unobserved confounding. We rank candidate DDIs robust to unobserved confounding more likely to be real. Our rigorous approach produces far fewer false positives than past applications that ignored (2) and (3).

G-Computation and Hierarchical Models for Estimating Multiple Causal Effects From Observational Disease Registries With Irregular Visits.

Huntington's Disease (HD) is a neurodegenerative disorder with serious motor, cognitive, and behavioral symptoms. Chorea, a motor symptom of HD characterized by abrupt involuntary movements, is typically treated with tetrabenazine or certain off-label antipsychotics. Clinical trial evidence about the impact of these drugs in the HD population is scant. However, multiple observational HD registries have recently been used with success to model HD progression1,2 and provide an opportunity to obtain effect estimates in the absence of clinical trials. We use a dataset integrated from four large-scale HD registries to generate evidence on the efficacy of chorea treatments for chorea as well as their impact on other aspects of HD progression. Clinical conclusions are meant only to illustrate our methodological approach. We employ parametric G-computation for causal inference to adjust for confounding and accommodate irregular visits and treatment patterns. We fit Bayesian hierarchical models to the results of multiple related analyses to share strength across studies and handle multiple comparisons concerns.