Identification and estimation of mediational effects of longitudinal modified treatment policies (preprint)

We demonstrate a comprehensive semiparametric approach to causal mediation analysis, addressing the complexities inherent in settings with longitudinal and continuous treatments, confounders, and mediators. Our methodology utilizes a nonparametric structural equation model and a cross-fitted sequential regression technique based on doubly robust pseudo-outcomes, yielding an efficient, asymptotically normal estimator without relying on restrictive parametric modeling assumptions. We are motivated by a recent scientific controversy regarding the effects of invasive mechanical ventilation (IMV) on the survival of COVID-19 patients, considering acute kidney injury (AKI) as a mediating factor. We highlight the possibility of "inconsistent mediation," in which the direct and indirect effects of the exposure operate in opposite directions. We discuss the significance of mediation analysis for scientific understanding and its potential utility in treatment decisions.

Effect of Paxlovid Treatment on Long COVID Onset: An EHR-Based Target Trial Emulation from N3C (preprint)

Preventing and treating post-acute sequelae of SARS-CoV-2 infection (PASC), commonly known as Long COVID, has become a public health priority. In this study, we examined whether treatment with Paxlovid in the acute phase of COVID-19 helps prevent the onset of PASC. We used electronic health records from the National Covid Cohort Collaborative (N3C) to define a cohort of 426,461 patients who had COVID-19 since April 1, 2022, and were eligible for Paxlovid treatment due to risk for progression to severe COVID-19. We used the target trial emulation (TTE) framework to estimate the effect of Paxlovid treatment on PASC incidence. Our primary outcome measure was a PASC computable phenotype. Secondary outcomes were the onset of novel cognitive, fatigue, and respiratory symptoms in the post-acute period. Paxlovid treatment did not have a significant effect on overall PASC incidence (relative risk [RR] = 0.99, 95% confidence interval [CI] 0.96-1.01). However, its effect varied across the cognitive (RR = 0.85, 95% CI 0.79-0.90), fatigue (RR = 0.93, 95% CI 0.89-0.96), and respiratory (RR = 0.99, 95% CI 0.95-1.02) symptom clusters, suggesting that Paxlovid treatment may help prevent post-acute cognitive and fatigue symptoms more than others.

Leveraging Neural Networks to Profile Health Care Providers with Application to Medicare Claims (preprint)

Encompassing numerous nationwide, statewide, and institutional initiatives in the United States, provider profiling has evolved into a major health care undertaking with ubiquitous applications, profound implications, and high-stakes consequences. In line with such a significant profile, the literature has accumulated a number of developments dedicated to enhancing the statistical paradigm of provider profiling. Tackling wide-ranging profiling issues, these methods typically adjust for risk factors using linear predictors. While this approach is simple, it can be too restrictive to characterize complex and dynamic factor-outcome associations in certain contexts. One such example arises from evaluating dialysis facilities treating Medicare beneficiaries with end-stage renal disease. It is of primary interest to consider how the coronavirus disease (COVID-19) affected 30-day unplanned readmissions in 2020. The impact of COVID-19 on the risk of readmission varied dramatically across pandemic phases. To efficiently capture the variation while profiling facilities, we develop a generalized partially linear model (GPLM) that incorporates a neural network. Considering provider-level clustering, we implement the GPLM as a stratified sampling-based stochastic optimization algorithm that features accelerated convergence. Furthermore, an exact test is designed to identify under- and over-performing facilities, with an accompanying funnel plot to visualize profiles. The advantages of the proposed methods are demonstrated through simulation experiments and profiling dialysis facilities using 2020 Medicare claims from the United States Renal Data System.

Detection and genetic characterization of enteric viruses in diarrhoea outbreaks from swine farms in Spain (preprint)

Background: The aim of this work was to study the prevalence and distribution of Porcine astrovirus (PAstV), Porcine kobuvirus (PKoV), Porcine torovirus (PToV), Mammalian orthoreovirus (MRV) and Porcine mastadenovirus (PAdV) as well as their association with widely recognized virus that cause diarrhoea in swine such as coronavirus (CoVs) and rotavirus (RVs) in diarrhoea outbreaks from Spanish swine farms. Furthermore, a selection of the viral strains was genetically characterized. Results: PAstV, PKoV, PToV, MRV and PAdV were frequently detected. Particularly, PAstV and PKoV were detected in almost 50% and 30% of the investigated farms, respectively, with an age-dependent distribution; PAstV was mainly detected in postweaning and fattening pigs, while PKoV was more frequent in sucking piglets. Viral co-infections were detected in almost half of the outbreaks, combining CoVs, RVs and the viruses studied, with a maximum of 5 different viral species reported in three investigated farms. Using a next generation sequencing approach, we obtained a total of 24 ARN viral genomes (>90% genome sequence), characterizing for first time the full genome of circulating strains of PAstV2, PAstV4, PAstV5 and PToV on Spanish farms. Phylogenetic analyses showed that PAstV, PKoV and PToV from Spanish swine farms clustered together with isolates of the same viral species from neighboring pig producing countries. Conclusions: Although further studies to evaluate the role of these enteric viruses in diarrhoea outbreaks are required, their wide distribution and frequent association in co-infections cannot be disregard. Hence, their inclusion into routine diagnostic panels for diarrhoea in swine should be considered.

Introducing longitudinal modified treatment policies: a unified framework for studying complex exposures (preprint)

This tutorial discusses a recently developed methodology for causal inference based on longitudinal modified treatment policies (LMTPs). LMTPs generalize many commonly used parameters for causal inference including average treatment effects, and facilitate the mathematical formalization, identification, and estimation of many novel parameters. LMTPs apply to a wide variety of exposures, including binary, multivariate, and continuous, as well as interventions that result in violations of the positivity assumption. LMTPs can accommodate time-varying treatments and confounders, competing risks, loss-to-follow-up, as well as survival, binary, or continuous outcomes. This tutorial aims to illustrate several practical uses of the LMTP framework, including describing different estimation strategies and their corresponding advantages and disadvantages. We provide numerous examples of types of research questions which can be answered within the proposed framework. We go into more depth with one of these examples -- specifically, estimating the effect of delaying intubation on critically ill COVID-19 patients' mortality. We demonstrate the use of the open source R package lmtp to estimate the effects, and we provide code on https://github.com/kathoffman/lmtp-tutorial.

Corticosteroids in COVID-19: Optimizing Observational Research through Target Trial Emulations (preprint)

Background: Observational research provides a unique opportunity to learn causal effects when randomized trials are not available, but obtaining the correct estimates hinges on a multitude of design and analysis choices. We illustrate the advantages of modern causal inference methods and compare to standard research practice to estimate the effect of corticosteroids on mortality in hospitalized COVID-19 patients in an observational dataset. We use several large RCTs to benchmark our results. Methods: Our retrospective data source consists of 3,293 COVID-19 patients hospitalized at New York Presbyterian March 1-May 15, 2020. We design our study using the Target Trial Emulation framework. We estimate the effect of an intervention consisting of 6 days of corticosteroids administered at the time of severe hypoxia and contrast with an intervention consisting of no corticosteroids administration. The dataset includes dozens of time-varying confounders. We estimate the causal effects using a doubly robust estimator where the probabilities of treatment, outcome, and censoring are estimated using flexible regressions via super learning. We compare these analyses to standard practice in clinical research, consisting of two main methods: (i) Cox models for an exposure of corticosteroids receipt within various time windows of hypoxia, and (ii) a Cox time-varying model where the exposure is daily administration of corticosteroids starting at the time of hospitalization. Results: The effect in our target trial emulation is qualitatively identical to an RCT benchmark, estimated to reduce 28-day mortality from 32% (95% confidence interval: 31-34) to 23% (21-24). The estimated effect from meta-analyses of RCTs for corticosteroids is an odds ratio of 0.66 (0.53-0.82)(1). Hazard ratios from the Cox models range in size and direction from 0.50 (0.41-0.62) to 1.08 (0.80-1.47) and all study designs suffer from various forms of bias. Conclusion: We demonstrate in a case study that clinical research based on observational data can unveil true causal relations. However, the correctness of these effect estimates requires designing and analyzing the data based on principles which are different from the current standard in clinical research. The widespread communication and adoption of these design and analytical techniques is of high importance for the improvement of clinical research based on observational data.

Characterization and cross-protection of experimental infections with SeCoV and two PEDV variants. (preprint)

Global emergence and re-emergence of Porcine epidemic diarrhea virus (PEDV), an Alphacoronavirus which causes a highly contagious enteric disease, have led to several studies addressing its variability. The aim of this study was to characterize the infection of weaned pigs with Swine enteric coronavirus (SeCoV) -a chimeric virus most likely originated from a recombination event between PEDV and Transmissible gastroenteritis virus, or its mutant Porcine respiratory coronavirus- , and two PEDV G1b variants, including a recently described recombinant PEDV-SeCoV (rPEDV-SeCoV), as well as to determine the degree of cross-protection achieved against the rPEDV-SeCoV. For this purpose, forty-eight 4-week-old weaned pigs were randomly allocated into four groups of 12 animals; piglets in groups B, C and D were orally inoculated with a PEDV variant (B and D) or SeCoV (C), while piglets in group A were mock inoculated and maintained as controls. At day 20 post-infection all groups were exposed to rPEDV-SeCoV; thus, group D was subjected to a homologous re-challenge, groups B and C to a heterologous re-challenge (PEDV/rPEDV-SeCoV and SeCoV/rPEDV-SeCoV, respectively) and group A was primary challenged (-/rPEDV-SeCoV). Clinical signs, viral shedding, microscopic lesions and specific humoral and cellular immune responses (IgG, IgA, neutralizing antibodies and IgA and IFN-γ-secreting cells) were monitored. After primo-infection all three viral strains induced an undistinguishable mild-to-moderate clinical disease with diarrhea as the main sign and villus shortening lesions in the small intestine. In homologous re-challenged pigs, no clinical signs or lesions were observed, and viral shedding was only detected in a single animal. This fact may be explained by the significant high level of rPEDV-SeCoV-specific neutralizing antibodies found in these pigs before the challenge. In contrast, prior exposition to a different PEDV G1b variant or SeCoV only provided partial cross-protection, allowing rPEDV-SeCoV replication and shedding in feces.

Causal survival analysis under competing risks using longitudinal modified treatment policies (preprint)

Longitudinal modified treatment policies (LMTP) have been recently developed as a novel method to define and estimate causal parameters that depend on the natural value of treatment. LMTPs represent an important advancement in causal inference for longitudinal studies as they allow the non-parametric definition and estimation of the joint effect of multiple categorical, numerical, or continuous exposures measured at several time points. We extend the LMTP methodology to problems in which the outcome is a time-to-event variable subject to right-censoring and competing risks. We present identification results and non-parametric locally efficient estimators that use flexible data-adaptive regression techniques to alleviate model misspecification bias, while retaining important asymptotic properties such as $\sqrt{n}$-consistency. We present an application to the estimation of the effect of the time-to-intubation on acute kidney injury amongst COVID-19 hospitalized patients, where death by other causes is taken to be the competing event.

Detection of respiratory viruses in cases of porcine respiratory disease in nurseries. (preprint)

Respiratory disease in weaned pigs is a common problem in the field, with a complex aetiology of both viruses and bacteria. In the present study, we investigated the presence of eleven viruses in nasal swabs collected from nurseries (fifty-five clinical outbreaks) under the suspicion of swine influenza A virus (swIAV) by cough and fever. The other ten viruses included influenza B (IBV) and influenza D viruses (IDV), Porcine reproductive and respiratory syndrome virus (PRRSV), Porcine respiratory coronavirus (PRCV), Porcine cytomegalovirus (PCMV), porcine circoviruses 2 (PCV2), 3 (PCV3) and 4 (PCV), Porcine parainfluenza 1 virus (PPIV1) and Swine orthopneumovirus (SOV). Twenty-nine swIAV-positive cases and twenty-six cases of swIAV-negative respiratory disease were primarily established. IBV, IBD, PCV4 and PPIV1 were not found in any case, while PRCV, SOV, and PCMV were more likely to be found in swIAV-positive nurseries with respiratory disease ( p <0.05) although, globally, PCV3, PRRSV, and PCMV were the most frequently detected agents on herd level. At an individual level, the prevalence of different viruses was: swIAV 48.6%; PRCV 48.0%; PRRSV 31.6%; SOV 33.8%; PCMV 48.3%, PCV2 36.0%; and PCV3 33.0%. Beyond that, it was common to find animals with low Ct values (< 30) for all agents except for PCV2 and PCV3. When analysed the association between different pathogens, PRCV was the one with the most associations. It positively interacted ( p < 0.05) with swIAV and SOV but was negatively associated ( p < 0.05) with PRRSV and PCVM. Besides these, swIAV and PRRSV were negatively related (p < 0.05). Further analysis of suckling pigs showed that circulation of PRCV, PCMV, SOV, and PCV3 started in the maternities, suggesting a role of the sows in the transmission. Overall, our data may contribute to a better understanding of the complex aetiology and the epidemiology of respiratory disease in weaners. This is the first report of SOV in Spain.

Optimizing Precision and Power by Machine Learning in Randomized Trials, with an Application to COVID-19 (preprint)

The rapid finding of effective therapeutics requires the efficient use of available resources in clinical trials. The use of covariate adjustment can yield statistical estimates with improved precision, resulting in a reduction in the number of participants required to draw futility or efficacy conclusions. We focus on time-to-event and ordinal outcomes. A key question for covariate adjustment in randomized studies is how to fit a model relating the outcome and the baseline covariates to maximize precision. We present a novel theoretical result establishing conditions for asymptotic normality of a variety of covariate-adjusted estimators that rely on machine learning (e.g., l1-regularization, Random Forests, XGBoost, and Multivariate Adaptive Regression Splines), under the assumption that outcome data is missing completely at random. We further present a consistent estimator of the asymptotic variance. Importantly, the conditions do not require the machine learning methods to converge to the true outcome distribution conditional on baseline variables, as long as they converge to some (possibly incorrect) limit. We conducted a simulation study to evaluate the performance of the aforementioned prediction methods in COVID-19 trials using longitudinal data from over 1,500 patients hospitalized with COVID-19 at Weill Cornell Medicine New York Presbyterian Hospital. We found that using l1-regularization led to estimators and corresponding hypothesis tests that control type 1 error and are more precise than an unadjusted estimator across all sample sizes tested. We also show that when covariates are not prognostic of the outcome, l1-regularization remains as precise as the unadjusted estimator, even at small sample sizes (n = 100). We give an R package adjrct that performs model-robust covariate adjustment for ordinal and time-to-event outcomes.

Inference for natural mediation effects under case-cohort sampling with applications in identifying COVID-19 vaccine correlates of protection (preprint)

Combating the SARS-CoV2 pandemic will require the fast development of effective preventive vaccines. Regulatory agencies may open accelerated approval pathways for vaccines if an immunological marker can be established as a mediator of a vaccine's protection. A rich source of information for identifying such correlates are large-scale efficacy trials of COVID-19 vaccines, where immune responses are measured subject to a case-cohort sampling design. We propose two approaches to estimation of mediation parameters in the context of case-cohort sampling designs. We establish the theoretical large-sample efficiency of our proposed estimators and evaluate them in a realistic simulation to understand whether they can be employed in the analysis of COVID-19 vaccine efficacy trials.

Risk of Ischemic Stroke in Patients with Covid-19 versus Patients with Influenza (preprint)

Importance: Case series without control groups suggest that Covid-19 may cause ischemic stroke, but whether Covid-19 is associated with a higher risk of ischemic stroke than would be expected from a viral respiratory infection is uncertain. Objective: To compare the rate of ischemic stroke between patients with Covid-19 and patients with influenza, a respiratory viral illness previously linked to stroke. Design: A retrospective cohort study. Setting: Two academic hospitals in New York City. Participants: We included adult patients with emergency department visits or hospitalizations with Covid-19 from March 4, 2020 through May 2, 2020. Our comparison cohort included adult patients with emergency department visits or hospitalizations with influenza A or B from January 1, 2016 through May 31, 2018 (calendar years spanning moderate and severe influenza seasons). Exposures: Covid-19 infection confirmed by evidence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in the nasopharynx by polymerase chain reaction, and laboratory-confirmed influenza A or B. Main Outcomes and Measures: A panel of neurologists adjudicated the primary outcome of acute ischemic stroke and its clinical characteristics, etiological mechanisms, and outcomes. We used logistic regression to compare the proportion of Covid-19 patients with ischemic stroke versus the proportion among patients with influenza. Results: Among 2,132 patients with emergency department visits or hospitalizations with Covid-19, 31 patients (1.5%; 95% confidence interval [CI], 1.0%-2.1%) had an acute ischemic stroke. The median age of patients with stroke was 69 years (interquartile range, 66-78) and 58% were men. Stroke was the reason for hospital presentation in 8 (26%) cases. For our comparison cohort, we identified 1,516 patients with influenza, of whom 0.2% (95% CI, 0.0-0.6%) had an acute ischemic stroke. After adjustment for age, sex, and race, the likelihood of stroke was significantly higher with Covid-19 than with influenza infection (odds ratio, 7.5; 95% CI, 2.3-24.9). Conclusions and Relevance: Approximately 1.5% of patients with emergency department visits or hospitalizations with Covid-19 experienced ischemic stroke, a rate 7.5-fold higher than in patients with influenza. Future studies should investigate the thrombotic mechanisms in Covid-19 in order to determine optimal strategies to prevent disabling complications like ischemic stroke.

SARS-CoV-2 amino acid substitutions widely spread in the human population are mainly located in highly conserved segments of the structural proteins (preprint)

The Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic offers a unique opportunity to study the introduction and evolution of a pathogen into a completely naive human population. We identified and analysed the amino acid mutations that gained prominence worldwide in the early months of the pandemic. Eight mutations have been identified along the viral genome, mostly located in conserved segments of the structural proteins and showing low variability among coronavirus, which indicated that they might have a functional impact. At the moment of writing this paper, these mutations present a varied success in the SARS-CoV-2 virus population; ranging from a change in the spike protein that becomes absolutely prevalent, two mutations in the nucleocapsid protein showing frequencies around 25%, to a mutation in the matrix protein that nearly fades out after reaching a frequency of 20%.

Improving Precision and Power in Randomized Trials for COVID-19 Treatments Using Covariate Adjustment, for Ordinal or Time to Event Outcomes (preprint)

SO_SCPLOWUMMARYC_SCPLOWTime is of the essence in evaluating potential drugs and biologics for the treatment and prevention of COVID-19. There are currently over 400 clinical trials (phase 2 and 3) of treatments for COVID-19 registered on clinicaltrials.gov. Covariate adjustment is a statistical analysis method with potential to improve precision and reduce the required sample size for a substantial number of these trials. Though covariate adjustment is recommended by the U.S. Food and Drug Administration and the European Medicines Agency, it is underutilized, especially for the types of outcomes (binary, ordinal and time-to-event) that are common in COVID-19 trials. To demonstrate the potential value added by covariate adjustment in this context, we simulated two-arm, randomized trials comparing a hypothetical COVID-19 treatment versus standard of care, where the primary outcome is binary, ordinal, or time-to-event. Our simulated distributions are derived from two sources: longitudinal data on over 500 patients hospitalized at Weill Cornell Medicine New York Presbyterian Hospital, and a Centers for Disease Control and Prevention (CDC) preliminary description of 2449 cases. We found substantial precision gains from using covariate adjustment-equivalent to 9-21% reductions in the required sample size to achieve a desired power-for a variety of estimands (targets of inference) when the trial sample size was at least 200. We provide an R package and practical recommendations for implementing covariate adjustment. The estimators that we consider are robust to model misspecification.