Publications

2022

1. Use of Machine Learning to Estimate the Per-Protocol Effect of Low-Dose Aspirin on Pregnancy Outcomes: A Secondary Analysis of a Randomized Clinical Trial

   Zhong, Yongqi, Brooks, Maria M.,  Kennedy, Edward H. and 2 more authors

   JAMA Network Open Mar 2022

   Abs HTML

   In randomized clinical trials (RCTs), per-protocol effects may be of interest in the presence of nonadherence with the randomized treatment protocol. Using machine learning in per-protocol effect estimation can help avoid model misspecification owing to strong parametric assumptions, as is common with standard methods (eg, logistic regression).To demonstrate the use of ensemble machine learning with augmented inverse probability weighting (AIPW) for per-protocol effect estimation in RCTs and to evaluate the per-protocol effect size of aspirin on pregnancy.This secondary analysis used data from 1227 women in the Effects of Aspirin in Gestation and Reproduction (EAGeR) trial, a multicenter, block-randomized, double-blind, placebo-controlled clinical trial of the effect of daily low-dose aspirin on pregnancy outcomes in women at high risk of pregnancy loss. Participants were recruited at 4 university medical centers in the US from June 15, 2007, to July 15, 2012. Women were followed up for 6 menstrual cycles for attempted pregnancy and 36 weeks of gestation if pregnancy occurred. Follow-up was completed on August 17, 2012. Data analyses were performed on July 9, 2021.Daily low-dose (81 mg) aspirin taken at least 5 of 7 days per week for at least 80% of follow-up time relative to placebo.Pregnancy detected using human chorionic gonadotropin (hCG) levels.Among the 1227 women included in the analysis (mean SD age, 28.74 [4.80] years), 1161 (94.6%) were non-Hispanic White and 858 (69.9%) adhered to the protocol. Five machine learning models were combined into 1 meta-algorithm, which was used to construct an AIPW estimator for the per-protocol effect. Compared with adhering to placebo, adherence to the daily low-dose aspirin protocol for at least 5 of 7 days per week was associated with an increase in the probability of hCG-detected pregnancy of 8.0 (95% CI, 2.5-13.6) more hCG-detected pregnancies per 100 women in the sample, which is substantially larger than the estimated intention-to-treat estimate of 4.3 (95% CI, −1.1 to 9.6) more hCG-detected pregnancies per 100 women in the sample.These findings suggest that a low-dose aspirin protocol is associated with increased hCG-detected pregnancy in women who adhere to treatment for at least 5 days per week. With the presence of nonadherence, per-protocol treatment effect estimates differ from intention-to-treat estimates in the EAGeR trial. The results of this secondary analysis of clinical trial data suggest that machine learning could be used to estimate per-protocol effects by adjusting for confounders related to nonadherence in a more flexible way than traditional regressions.ClinicalTrials.gov Identifier: NCT00467363

2021

1. 

   AIPW: An R Package for Augmented Inverse Probability Weighted Estimation of Average Causal Effects

   Zhong, Yongqi, Kennedy, Edward H,  Bodnar, Lisa M and 1 more author

   American Journal of Epidemiology Jul 2021

   Abs Bib HTML

   An increasing number of recent studies suggest doubly robust estimators with cross-fitting should be used when estimating causal effects with machine learning methods. However, existing programs that implement doubly robust estimators do not all support machine learning methods and cross-fitting, or provide estimates on multiplicative scales. To address these needs, we developed the AIPW package implementing the augmented inverse probability weighting (AIPW) estimation of average causal effects in R. Key features of the AIPW package includes cross-fitting and flexible covariate adjustment for observational studies and randomized trials (RCTs). In this paper, we use a simulated RCT to present the implementation of the AIPW estimator. We also perform a simulation study to evaluate the performance of the AIPW package compared with other doubly robust implementations including CausalGAM, npcausal, tmle, and tmle3. Our simulation shows that the xtbfAIPW package yielded comparable performance to other programs. Furthermore, we also found that cross-fitting substantively decreases the bias and improves the confidence interval coverage for doubly robust estimators fit with machine learning algorithms. Our findings suggest that the AIPW package can be a useful tool for estimating average causal effects with machine learning methods in RCTs and observational studies.

   @article{zhong_aipw_2021,
     author = {Zhong, Yongqi and Kennedy, Edward H and Bodnar, Lisa M and Naimi, Ashley I},
     doi = {10.1093/aje/kwab207},
     issn = {0002-9262},
     journal = {American Journal of Epidemiology},
     month = jul,
     number = {kwab207},
     shorttitle = {{AIPW}},
     title = {{AIPW}: {An} {R} {Package} for {Augmented} {Inverse} {Probability} {Weighted} {Estimation} of {Average} {Causal} {Effects}},
     urldate = {2021-07-16},
     year = {2021},
     bdsk-url-1 = {https://doi.org/10.1093/aje/kwab207},
   }

2. Brain 5-HT1A Receptor PET Binding, Cortisol Responses to Stress, and the Familial Transmission of Suicidal Behavior

   Melhem, Nadine M, Zhong, Yongqi,  Miller, Jeffrey M and 10 more authors

   International Journal of Neuropsychopharmacology Sep 2021

   Abs HTML

   The serotonin 1A (5-HT1A) receptor has been implicated in depression and suicidal behavior. Lower resting cortisol levels are associated with higher 5-HT1A receptor binding, and both differentiate suicide attempters with depression. However, it is not clear whether 5-HT1A receptor binding and cortisol responses to stress are related to familial risk and resilience for suicidal behavior. [ 11C]CUMI-101 PET imaging to quantify regional brain 5-HT1A receptor binding was conducted in individuals considered to be at high-risk for mood disorder or suicidal behavior on the basis of having a first or second degree relative(s) with an early onset mood disorder and history of suicidal behavior. These high-risk subjects were subdivided into: high-risk resilient having no mood disorder or suicidal behavior (HR-R, n=29); high-risk with mood disorder and no suicidal behavior history (HR-MOOD n=31); and high-risk with mood disorder and suicidal behavior (HR-SA/MOOD, n=25). Groups were compared to healthy volunteers without a family history of mood disorder or suicidal behavior (HV, n=34). Participants underwent the Trier Social Stress Task (TSST). All subjects were free from psychotropic medications at the time of the TSST and PET scanning. We observed no group differences in 5-HT1A receptor binding considering all regions simultaneously, nor did we observe heterogeneity of the effect of group across regions. These results were similar across outcome measures (BPND for all subjects and BPp in a subset of the sample), and definitions of regions of interest (ROIs; standard or serotonin system-specific ROIs). We also found no group differences on TSST outcomes. Within HR-SA/MOOD, lower BPp binding [β=-0.084, Standard Error or SE=0.038, p=0.048] and higher cortisol reactivity to stress [β=9.25, 95% CI (3.27,15.23), p=0.004] were associated with higher lethality attempts. There were no significant relationships between 5-HT1A binding and cortisol outcomes. 5-HT1A receptor binding in ROIs was not linked to familial risk or resilience protecting against suicidal behavior or mood disorder although it may be related to lethality of suicide attempt. Future studies are needed to better understand the biological mechanisms implicated in familial risk for suicidal behavior and how HPA axis function influences such risk.

3. Collaboration on evidence synthesis in Africa: a network study of growing research capacity

   Pan, Jiayi, Zhong, Yongqi,  Young, Sarah and 1 more author

   Health Research Policy and Systems Sep 2021

   Abs HTML

   Evidence-based practice in medicine and social policy relies heavily on evidence synthesis. To translate evidence into practical guidelines for low- and middle-income countries, local expertise is essential. The objectives of this study are to assess the change in capacity for conducting evidence synthesis in Africa and to identify key African institutions for regional capacity-building. We take on a network perspective, considering that the position of an institution in the African evidence ecosystem is one constituent of its research capacity.

4. Early Cortisol and Inflammatory Responses to Parental Cancer and Their Impact on Functional Impairment in Youth

   Hayes, Benjamin, Brent, Jacob,  Zhong, Yongqi and 8 more authors

   Journal of Clinical Medicine Jan 2021

   Abs HTML

   Purpose: Chronic stress is associated with increased risk for maladaptive psychological responses during childhood, adolescence, and young adulthood. Adults exposed to chronic stress during childhood exhibit dysregulation of hypothalamic-pituitary-adrenal (HPA) axis activity and inflammation. There are no studies examining the impact of stress on biological stress responses and functional impairment in adolescents and young adults early after the onset of a stressor. Methods: The sample consisted of 59 offspring, aged 11–25 years, 33 of parents diagnosed with cancer and 26 controls from families with no cancer or severe chronic illness in parents or siblings. Cancer patients and their families were recruited within an average of 62 days (SD = 35.9) and followed at 6 and 9 months later. Functional impairment was assessed and hair cortisol concentrations (HCC), salivary cortisol, and inflammatory markers were measured. Mixed regression analyses were conducted. Results: The stress group showed higher functional impairment (β = −5.5, 95% CI (−10.4, −0.06), p = 0.03, d= −0.40) and HCC (β = 10.5, 95% CI (−5.5, −0.50), p < 0.001, d = 1.43). However, HCC were reduced over time in the stress group (β= −0.3, 95% CI (−0.04, −0.01), p < 0.001, d = −1.08). Higher total cortisol output was associated with increased functional impairment over time (β = −3.0, 95% CI (−5.5, −0.5), p = 0.02, d = −0.60). Conclusions: Parental cancer is associated with early increase in cortisol, which was associated with increased functional impairment in offspring. Clinicians need to assess and monitor psychiatric symptoms and functioning in these offspring early on following parental cancer diagnosis.

5. Effect of Antithrombotic Therapy on Clinical Outcomes in Outpatients With Clinically Stable Symptomatic COVID-19: The ACTIV-4B Randomized Clinical Trial

   Connors, Jean M., Brooks, Maria M.,  Sciurba, Frank C. and 25 more authors

   JAMA Oct 2021

   Abs HTML

   Acutely ill inpatients with COVID-19 typically receive antithrombotic therapy, although the risks and benefits of this intervention among outpatients with COVID-19 have not been established.To assess whether anticoagulant or antiplatelet therapy can safely reduce major adverse cardiopulmonary outcomes among symptomatic but clinically stable outpatients with COVID-19.The ACTIV-4B Outpatient Thrombosis Prevention Trial was designed as a minimal-contact, adaptive, randomized, double-blind, placebo-controlled trial to compare anticoagulant and antiplatelet therapy among 7000 symptomatic but clinically stable outpatients with COVID-19. The trial was conducted at 52 US sites between September 2020 and June 2021; final follow-up was August 5, 2021. Prior to initiating treatment, participants were required to have platelet count greater than 100 000/mm3 and estimated glomerular filtration rate greater than 30 mL/min/1.73 m2.Random allocation in a 1:1:1:1 ratio to aspirin (81 mg orally once daily; n = 164), prophylactic-dose apixaban (2.5 mg orally twice daily; n = 165), therapeutic-dose apixaban (5 mg orally twice daily; n = 164), or placebo (n = 164) for 45 days.The primary end point was a composite of all-cause mortality, symptomatic venous or arterial thromboembolism, myocardial infarction, stroke, or hospitalization for cardiovascular or pulmonary cause. The primary analyses for efficacy and bleeding events were limited to participants who took at least 1 dose of trial medication.On June 18, 2021, the trial data and safety monitoring board recommended early termination because of lower than anticipated event rates; at that time, 657 symptomatic outpatients with COVID-19 had been randomized (median age, 54 years [IQR, 46-59]; 59% women). The median times from diagnosis to randomization and from randomization to initiation of study treatment were 7 days and 3 days, respectively. Twenty-two randomized participants (3.3%) were hospitalized for COVID-19 prior to initiating treatment. Among the 558 patients who initiated treatment, the adjudicated primary composite end point occurred in 1 patient (0.7%) in the aspirin group, 1 patient (0.7%) in the 2.5-mg apixaban group, 2 patients (1.4%) in the 5-mg apixaban group, and 1 patient (0.7%) in the placebo group. The risk differences compared with placebo for the primary end point were 0.0% (95% CI not calculable) in the aspirin group, 0.7% (95% CI, –2.1% to 4.1%) in the 2.5-mg apixaban group, and 1.4% (95% CI, –1.5% to 5.0%) in the 5-mg apixaban group. Risk differences compared with placebo for bleeding events were 2.0% (95% CI, –2.7% to 6.8%), 4.5% (95% CI, –0.7% to 10.2%), and 6.9% (95% CI, 1.4% to 12.9%) among participants who initiated therapy in the aspirin, prophylactic apixaban, and therapeutic apixaban groups, respectively, although none were major. Findings inclusive of all randomized patients were similar.Among symptomatic clinically stable outpatients with COVID-19, treatment with aspirin or apixaban compared with placebo did not reduce the rate of a composite clinical outcome. However, the study was terminated after enrollment of 9% of participants because of an event rate lower than anticipated.ClinicalTrials.gov Identifier: NCT04498273

6. Opioid use as a proximal risk factor for suicidal behavior in young adults

   Marengo, Laura, Douaihy, Antoine,  Zhong, Yongqi and 15 more authors

   Suicide and Life-Threatening Behavior Nov 2021

   Abs HTML

   Introduction: There is a concomitant rise in suicide rates with the prevalence of opioids involved in overdose deaths, especially among adolescents and young adults. However, there are limited studies on whether opioid use prospectively predicts suicidal behavior in youth. Methods: Our sample included 183 psychiatric patients (18—30 years) admitted for a suicide attempt (SA), have current suicidal ideation (SI), and psychiatric controls without ideation or attempt (PC). Suicidal behavior was assessed using the Columbia Suicide Severity Rating Scale. We also recruited a healthy control group (HC; n = 40). Patients and controls were followed over a year. ANOVA, regression, and cox regression were used. Results: Suicide attempt (β = 0.87, CI [0.1—1.6], p = 0.02) and SI [(β = 0.75, CI [0.03—1.5], p = 0.04) were significantly more likely than HCs to have used opioids in the past year at baseline. Opioid use was associated with increased anxiety symptoms (β = 0.75, CI [0.001—1.5], p = 0.05), PTSD symptoms (β = 3.90, CI [1.1—6.7], p = 0.01), and aggression (β = 0.02, CI [0.01—0.04], p = 0.02). Opioid use in the month prior to hospitalization predicted SA at 6 months (OR = 1.87, CI [1.06—3.31], p = 0.032). Conclusions: Opioid use is a proximal predictor for SA. These findings may help clinicians better identify patients at risk for suicidal behavior, allowing for more personalized treatment approaches.

7. The psychiatric sequelae of the COVID‐19 pandemic in adolescents, adults, and health care workers

   Murata, Stephen, Rezeppa, Taylor,  Thoma, Brian and 7 more authors

   Depression and Anxiety Nov 2021

   Abs HTML

   Background The COVID-19 pandemic is the most serious global public health crisis since the 1918 influenza pandemic. This study is the first to assess its mental health impact across the lifespan in the United States in adolescents, adults, and health care workers. Methods We recruited 4909 participants through an online survey advertising on Facebook and Instagram to assess exposure to COVID-19 and psychiatric symptoms from April 27 to July 13. We also recruited through the University of Pittsburgh, University of Pittsburgh Medical Center, and other health care systems around Pittsburgh. The primary outcomes were clinically significant depression, anxiety, and posttraumatic stress disorder (PTSD) symptoms, suicidal ideation or behavior, and grief reactions since COVID-19. Results Adolescents were significantly more likely to report moderate to severe symptoms of depression (55% vs. 29%; χ2 = 122, df = 1; p \textless .001), anxiety (48% vs. 29%; χ2 = 73; df = 1; p \textless .001), PTSD (45% vs. 33%; χ2 = 12; df = 1; p \textless .001), suicidal ideation or behavior (38% vs. 16%; χ2 = 117; df = 1; p \textless .001), and sleep problems (69% vs. 57%; χ2 = 26; df = 1; p \textless .001) compared to adults. The rates of intense grief reactions among those who lost someone to COVID-19 was 55%. Loneliness was the most common predictor across outcomes and higher number of hours spent on social media and exposure to media about COVID-19 predicted depression symptoms and suicidal ideation or behavior in adolescents. Conclusions The COVID-19 pandemic is associated with increased rates of clinically significant psychiatric symptoms. Loneliness could put individuals at increased risk for the onset of psychiatric disorders.

8. Therapeutic Anticoagulation with Heparin in Critically Ill Patients with Covid-19

   The REMAP-CAP, ACTIV-4a, and ATTACC Investigators,

   New England Journal of Medicine Aug 2021

   Abs HTML

   BACKGROUND Thrombosis and inflammation may contribute to morbidity and mortality among patients with coronavirus disease 2019 (Covid-19). We hypothesized that therapeutic-dose anticoagulation would improve outcomes in critically ill patients with Covid-19. METHODS In an open-label, adaptive, multiplatform, randomized clinical trial, critically ill patients with severe Covid-19 were randomly assigned to a pragmatically defined regimen of either therapeutic-dose anticoagulation with heparin or pharmacologic thromboprophylaxis in accordance with local usual care. The primary outcome was organ support–free days, evaluated on an ordinal scale that combined in-hospital death (assigned a value of −1) and the number of days free of cardiovascular or respiratory organ support up to day 21 among patients who survived to hospital discharge. The members of the executive writing committee and the block writing committee assume responsibility for the overall content and integrity of this article. The full names, academic degrees, and affiliations of the members of the executive writing committee and the block writing committee are listed in the Appendix. Address reprint requests to Dr. Zarychanski at the Sections of Hematology/Oncology and Critical Care, University of Manitoba, Winnipeg, MB, Canada R3E 0V9, or at -rzarychanski@c- ancercare .-mb.-ca. *The full list of investigators and collaborators is provided in the Supplementary Appendix, available at NEJM.org. RESULTS The trial was stopped when the prespecified criterion for futility was met for therapeutic-dose anticoagulation. Data on the primary outcome were available for 1098 patients (534 assigned to therapeutic-dose anticoagulation and 564 assigned to usual-care thromboprophylaxis). The median value for organ support–free days was 1 (interquartile range, −1 to 16) among the patients assigned to therapeuticdose anticoagulation and was 4 (interquartile range, −1 to 16) among the patients assigned to usual-care thromboprophylaxis (adjusted proportional odds ratio, 0.83; 95% credible interval, 0.67 to 1.03; posterior probability of futility [defined as an odds ratio \textless1.2], 99.9%). The percentage of patients who survived to hospital discharge was similar in the two groups (62.7% and 64.5%, respectively; adjusted odds ratio, 0.84; 95% credible interval, 0.64 to 1.11). Major bleeding occurred in 3.8% of the patients assigned to therapeutic-dose anticoagulation and in 2.3% of those assigned to usual-care pharmacologic thromboprophylaxis. Drs. Goligher, Bradbury, McVerry, Lawler, Berger, and Gong and Drs. Berry, McArthur, Neal, Hochman, Webb, and Zarychanski contributed equally to this article. This article was published on August 4, 2021, at NEJM.org. DOI: 10.1056/NEJMoa2103417 Copyright \copyright 2021 Massachusetts Medical Society. CONCLUSIONS In critically ill patients with Covid-19, an initial strategy of therapeutic-dose anticoagulation with heparin did not result in a greater probability of survival to hospital discharge or a greater number of days free of cardiovascular or respiratory organ support than did usual-care pharmacologic thromboprophylaxis. (REMAP-CAP, ACTIV-4a, and ATTACC ClinicalTrials.gov numbers, NCT02735707, NCT04505774, NCT04359277, and NCT04372589.)

2020

1. Signaling Networks in Inflammatory Pathways and Risk for Suicidal Behavior

   Rengasamy, Manivel, Zhong, Yongqi,  Marsland, Anna and 4 more authors

   Brain, Behavior, & Immunity - Health Aug 2020

   Abs HTML

   Background: Suicide is a leading cause of death in the young adult population, with few biological markers identified thus far to be associated with suicidality. Cytokines (including IL-6 and TNFα) may contribute to increased risk for depression and suicidality. Few studies have examined the associations of cytokine mRNA expression with depression and suicidal ideation and behavior. This study examines these associations and whether cytokine signaling networks differentiate suicide attempters (SA), suicide ideators (SI), and healthy controls (HC).

2019

1. Getting to precision psychopharmacology: Combining clinical and genetic information to predict fat gain from aripiprazole

   Oughli, H., Lenze, E. J.,  Locke, A. E. and 9 more authors

   Journal of Psychiatric Research Aug 2019

   Abs

   INTRODUCTION: All atypical antipsychotics are associated with some degree of weight gain. We applied a novel statistical approach to identify moderators of aripiprazole-induced fat gain using clinical and genetic data from a randomized clinical trial (RCT) of treatment resistant depression in older adults. MATERIALS AND METHODS: Adults aged ≥60 years with non-response to a prospective trial of venlafaxine were randomized to 12 weeks of aripiprazole augmentation (n = 91) or placebo (n = 90). Dual energy x-ray absorptiometry (DEXA) measured adiposity at baseline and 12 weeks. Independent moderators of total body fat gain were used to generate two combined multiple moderators, one including clinical data alone and one including both clinical and genetic data to characterize individuals who gained fat during aripiprazole augmentation. RESULTS: The value of the combined genetic + clinical multiple moderator (Mcg) was 0.57 [95% CI 0.46, 0.68] (effect size: 0.57), compared to the combined clinical moderator (Mc) value of 0.49 [0.34, 0.63] (effect size: 0.49). Individuals who gained adiposity in this study were more likely to be female and younger in age, have lower weight, fasting glucose and lipids at baseline and positive for the HTR2C polymorphism. DISCUSSION: These results demonstrate a combined multiple moderator approach, including both clinical and genetic moderators, can be applied to existing clinical trial data to understand adverse treatment effects. This method allowed for more specific characterization of individuals at risk for the outcome of interest. Further work is needed to identify additional genetic moderators and to validate the approach.