Arsenic Exposure-Related Hypertension in Bangladesh and Reduced Circulating Nitric Oxide Bioavailability

BACKGROUND

Hypertension is a major cause of death worldwide. Although arsenic exposure has been associated with the risk of hypertension, this association appears nonuniform due to inconsistent results from studies conducted in different populations. Moreover, hypertension is a complex condition with multiple underlying mechanisms and factors. One factor is impaired production and bioavailability of vascular nitric oxide (NO). However, the implications of the effects of arsenic exposure on circulating NO and its association with hypertension in humans are largely unknown.

OBJECTIVE

We investigated the dose-response relationship between arsenic exposure and hypertension with vascular NO levels as a potential mediator of arsenic-related hypertension in individuals exposed to a broad range of arsenic.

METHODS

A total of 828 participants were recruited from low- and high-arsenic exposure areas in Bangladesh. Participants' drinking water, hair, and nail arsenic concentrations were measured by inductively coupled plasma mass spectroscopy. Hypertension was defined as a systolic blood pressure (SBP) value of ≥ 140 and a diastolic (DBP) value of ≥ 90  mmHg . Serum NO levels reflected by total serum nitrite concentrations were measured by immunoassay. A formal causal mediation analysis was used to assess NO as a mediator of the association between arsenic level and hypertension.

RESULTS

Increasing concentrations of arsenic measured in drinking water, hair, and nails were associated with the increasing levels of SBP and DBP. The odds of hypertension were dose-dependently increased by arsenic even in participants exposed to relatively low to moderate levels (10 - 50 μ g / L ) of water arsenic [odds ratios (ORs) and 95% confidence intervals (CIs): 2.87 (95% CI: 1.28, 6.44), 2.67 (95% CI: 1.27, 5.60), and 5.04 (95% CI: 2.71, 9.35) for the 10 - 50 μ g / L , 50.01 - 150 μ g / L , and > 150 μ g / L groups, respectively]. Causal mediation analysis showed a significant mediating effect of NO on arsenic-related SBP, DBP, and hypertension.

CONCLUSION

Increasing exposure to arsenic was associated with increasing odds of hypertension. The association was mediated through the reduction of vascular NO bioavailability, suggesting that impaired NO bioavailability was a plausible underlying mechanism of arsenic-induced hypertension in this Bangladeshi population. https://doi.org/10.1289/EHP13018.

A plasma peptidomic signature reveals extracellular matrix remodeling and predicts prognosis in alcohol-related hepatitis

Alcohol-related hepatitis (AH) is plagued with high mortality and difficulty in identifying at-risk patients. The extracellular matrix undergoes significant remodeling during inflammatory liver injury that can be detected in biological fluids and potentially used for mortality prediction. EDTA plasma samples were collected from AH patients (n= 62); Model for End-Stage Liver Disease (MELD) score defined AH severity as moderate (12-20; n=28) and severe (>20; n=34). The peptidome data was collected by high resolution, high mass accuracy UPLC-MS. Univariate and multivariate analyses identified differentially abundant peptides, which were used for Gene Ontology, parent protein matrisomal composition and protease involvement. Machine learning methods were used on patient-specific peptidome and clinical data to develop mortality predictors. Analysis of plasma peptides from AH patients and healthy controls identified over 1,600 significant peptide features corresponding to 130 proteins. These were enriched for ECM fragments in AH samples, likely related to turnover of hepatic-derived proteins. Analysis of moderate versus severe AH peptidomes showed a shift in abundance of peptides from collagen 1A1 and fibrinogen A proteins. The dominant proteases for the AH peptidome spectrum appear to be CAPN1 and MMP12. Increase in hepatic expression of these proteases was orthogonally-validated in RNA-seq data of livers from AH patients. Causal graphical modeling identified four peptides directly linked to 90-day mortality in >90% of the learned graphs. These peptides improved the accuracy of mortality prediction over MELD score and were used to create a clinically applicable mortality prediction assay. A signature based on plasma peptidome is a novel, non-invasive method for prognosis stratification in AH patients. Our results could also lead to new mechanistic and/or surrogate biomarkers to identify new AH mechanisms.

Lay summary

We used degraded proteins found the blood of alcohol-related hepatitis patients to identify new potential mechanisms of injury and to predict 90 day mortality.

Imputation-based Q-learning for optimizing dynamic treatment regimes with right-censored survival outcome

Q-learning has been one of the most commonly used methods for optimizing dynamic treatment regimes (DTRs) in multistage decision-making. Right-censored survival outcome poses a significant challenge to Q-Learning due to its reliance on parametric models for counterfactual estimation which are subject to misspecification and sensitive to missing covariates. In this paper, we propose an imputation-based Q-learning (IQ-learning) where flexible nonparametric or semiparametric models are employed to estimate optimal treatment rules for each stage and then weighted hot-deck multiple imputation (MI) and direct-draw MI are used to predict optimal potential survival times. Missing data are handled using inverse probability weighting and MI, and the nonrandom treatment assignment among the observed is accounted for using a propensity-score approach. We investigate the performance of IQ-learning via extensive simulations and show that it is more robust to model misspecification than existing Q-Learning methods, imputes only plausible potential survival times contrary to parametric models and provides more flexibility in terms of baseline hazard shape. Using IQ-learning, we developed an optimal DTR for leukemia treatment based on a randomized trial with observational follow-up that motivated this study.

Symptoms of Liver Disease During Tenofovir Therapy With or Without Peginterferon: Results from the Hepatitis B Research Network Immune Active Trial

AIMS

Evaluate patient-reported liver symptoms during treatment for chronic hepatitis B viral (HBV) infection and associations between changes in symptoms and levels of alanine aminotransferase (ALT) and viral markers.

METHODS

Data from 200 participants in the Hepatitis B Research Network Immune Active Trial who completed symptom assessments were analyzed. Patients were treated with tenofovir, with or without peginterferon (TDF + PegIFN vs. TDF alone) for 192 weeks. Participants completed a Symptom Checklist at baseline and every 4-12 weeks. A total symptom score was created, ranging from 0 (none) to 40 (severe). The SF-36 was completed every 48 weeks. Associations of symptom scores with ALT and viral markers were evaluated at baseline and end of treatment.

RESULTS

Participants were 65% male, 83% Asian, with a mean age of 42. Baseline symptoms were mild (median = 2, range 0-25) and associated with baseline ALT, HBV DNA levels and HBeAg + status. Patients on TDF alone experienced a more rapid and greater improvement in symptoms, but by week 192, symptom improvement was similar in both groups (54% vs 36%). Symptom improvements correlated with ALT and HBV DNA, most markedly among those with symptoms at baseline. Most patients (4 out of 6) who achieved HBsAg loss experienced symptom improvements. Overall, SF-36 scores did not change with treatment.

CONCLUSIONS

Reduction in ALT and HBV DNA levels with therapy are associated with significant improvement in liver symptoms such as fatigue and pain over the liver, especially among those with higher ALT, HBV DNA, symptoms and HBeAg + status prior to treatment.

Withdrawal of Long-Term Nucleotide Analog Therapy in Chronic Hepatitis B: Outcomes From the Withdrawal Phase of the HBRN Immune Active Treatment Trial

INTRODUCTION

Withdrawal of nucleos(t)ide analog therapy is increasingly being evaluated in chronic hepatitis B infection as a strategy to induce hepatitis B surface antigen (HBsAg) loss. The Hepatitis B Research Network Immune-Active Trial evaluated treatment with tenofovir (TDF) for 4 years ± an initial 6 months of peginterferon-α (PegIFN) (NCT01369212) after which treatment was withdrawn.

METHODS

Eligible participants (hepatitis B e antigen [HBeAg]-/anti-HBe+, hepatitis B virus [HBV] DNA <10 3 IU/mL, no cirrhosis) who discontinued TDF were followed for at least 1 year with optional follow-up thereafter. Retreatment was based on predefined criteria.

RESULTS

Among 201 participants who received 4 years of treatment, 97 participants (45 TDF and 52 TDF + PegIFN arm, 79 Asian) discontinued TDF. HBsAg loss occurred in 5 participants, 2 within 25 weeks and 3 within 89-119 weeks postwithdrawal (cumulative rate 4.3% by 2 years). Alanine aminotransferase (ALT) flares (>5× upper limit of normal) after TDF withdrawal occurred in 36 (37.1%) participants and occurred more frequently and earlier in those HBeAg- compared with HBeAg+ at treatment initiation. ALT flares were associated with older age and higher HBV DNA pretreatment and at the visit before the flare. ALT flares were not significantly associated with HBsAg decline or loss but were associated with immune active disease at 1 year (70.6% vs 11.9%, P < 0.0001) and 2 years (66.7% vs 25.9%, P = 0.03) postwithdrawal. Treatment reinitiation was required in 13 (13.4%) participants, and 13 others remained in a sustained inactive carrier state by the end of the study follow-up. No criteria reliably predicted safe treatment withdrawal.

DISCUSSION

Results from this trial do not support TDF withdrawal as a therapeutic strategy. HBsAg loss was infrequent within 2 years of stopping long-term TDF. If withdrawal is considered, HBV DNA should be carefully monitored with reinitiation of therapy if levels rise above 4 log 10 IU/mL to reduce the risk of ALT flares, as they were not associated with subsequent HBsAg decline or loss.

Immunostaining for hepatitis B viral antigens in liver: Association with clinical, biochemical, and virologic features of disease

BACKGROUND AND AIM

Staining for hepatitis B viral antigens is often done in liver biopsies from patients with chronic hepatitis B, but its correlates with clinical phenotypes are not well described.

METHODS

Biopsies were collected from a large cohort of adults and children with chronic hepatitis B viral infection through the Hepatitis B Research Network. Immunohistochemical staining of sections was done for hepatitis B surface antigen (HBsAg) and hepatitis B core antigen (HBcAg) and then centrally read by the pathology committee. The degree of liver injury and pattern of staining were then correlated with clinical characteristics, including the clinical phenotype of hepatitis B.

RESULTS

Biopsies from 467 subjects were studied, including 46 from children. Immunostaining for HBsAg was positive in 417 (90%) with scattered hepatocyte staining being the most common pattern. HBsAg staining correlated best with serum levels of HBsAg and hepatitis B viral DNA; the absence of HBsAg staining was often a prelude to loss of HBsAg from serum. HBcAg staining was positive in 225 (49%), and, while cytoplasmic staining was more frequent than nuclear staining, both nuclear and cytoplasmic positivity were often seen in the same specimen. Staining for HBcAg correlated with both level of viremia and liver injury. No biopsies from inactive carriers had stainable HBcAg, while 91% of the biopsies from those with hepatitis B e antigen-positive chronic hepatitis B stained positively for HBcAg.

CONCLUSION

Immunostaining for hepatitis B viral antigens may yield helpful insights into liver disease pathogenesis but appears to add little to commonly used serological and biochemical blood tests.

Effect of Thromboprophylaxis on Clinical Outcomes After COVID-19 Hospitalization

BACKGROUND

Patients hospitalized with COVID-19 have an increased incidence of thromboembolism. The role of extended thromboprophylaxis after hospital discharge is unclear.

OBJECTIVE

To determine whether anticoagulation is superior to placebo in reducing death and thromboembolic complications among patients discharged after COVID-19 hospitalization.

DESIGN

Prospective, randomized, double-blind, placebo-controlled clinical trial. (ClinicalTrials.gov: NCT04650087).

SETTING

Done during 2021 to 2022 among 127 U.S. hospitals.

PARTICIPANTS

Adults aged 18 years or older hospitalized with COVID-19 for 48 hours or more and ready for discharge, excluding those with a requirement for, or contraindication to, anticoagulation.

INTERVENTION

2.5 mg of apixaban versus placebo twice daily for 30 days.

MEASUREMENTS

The primary efficacy end point was a 30-day composite of death, arterial thromboembolism, and venous thromboembolism. The primary safety end points were 30-day major bleeding and clinically relevant nonmajor bleeding.

RESULTS

Enrollment was terminated early, after 1217 participants were randomly assigned, because of a lower than anticipated event rate and a declining rate of COVID-19 hospitalizations. Median age was 54 years, 50.4% were women, 26.5% were Black, and 16.7% were Hispanic; 30.7% had a World Health Organization severity score of 5 or greater, and 11.0% had an International Medical Prevention Registry on Venous Thromboembolism risk prediction score of greater than 4. Incidence of the primary end point was 2.13% (95% CI, 1.14 to 3.62) in the apixaban group and 2.31% (CI, 1.27 to 3.84) in the placebo group. Major bleeding occurred in 2 (0.4%) and 1 (0.2%) and clinically relevant nonmajor bleeding occurred in 3 (0.6%) and 6 (1.1%) apixaban-treated and placebo-treated participants, respectively. By day 30, thirty-six (3.0%) participants were lost to follow-up, and 8.5% of apixaban and 11.9% of placebo participants permanently discontinued the study drug treatment.

LIMITATIONS

The introduction of SARS-CoV-2 vaccines decreased the risk for hospitalization and death. Study enrollment spanned the peaks of the Delta and Omicron variants in the United States, which influenced illness severity.

CONCLUSION

The incidence of death or thromboembolism was low in this cohort of patients discharged after hospitalization with COVID-19. Because of early enrollment termination, the results were imprecise and the study was inconclusive.

PRIMARY FUNDING SOURCE

National Institutes of Health.

Interim monitoring in sequential multiple assignment randomized trials

A sequential multiple assignment randomized trial (SMART) facilitates the comparison of multiple adaptive treatment strategies (ATSs) simultaneously. Previous studies have established a framework to test the homogeneity of multiple ATSs by a global Wald test through inverse probability weighting. SMARTs are generally lengthier than classical clinical trials due to the sequential nature of treatment randomization in multiple stages. Thus, it would be beneficial to add interim analyses allowing for an early stop if overwhelming efficacy is observed. We introduce group sequential methods to SMARTs to facilitate interim monitoring based on the multivariate chi-square distribution. Simulation studies demonstrate that the proposed interim monitoring in SMART (IM-SMART) maintains the desired type I error and power with reduced expected sample size compared to the classical SMART. Finally, we illustrate our method by reanalyzing a SMART assessing the effects of cognitive behavioral and physical therapies in patients with knee osteoarthritis and comorbid subsyndromal depressive symptoms.

HBV transcription and translation persist despite viral suppression in HBV-HIV co-infected patients on antiretroviral therapy

BACKGROUND AND AIMS

Liver injury may persist in patients with HBV receiving antiviral therapy who have ongoing transcription and translation. We sought to assess ongoing HBV transcription by serum HBV RNA, translation by serum hepatitis B core related antigen (HBcrAg), and their associations with hepatic HBsAg and HBcAg staining in patients coinfected with HBV and HIV.

METHODS

This is a cross-sectional study of 110 adults coinfected with HBV and HIV who underwent clinical assessment and liver biopsy. Immunohistochemistry (IHC) was performed for HBsAg and HBcAg. Viral biomarkers included quantitative HBsAg, HBV RNA, and HBcrAg.

RESULTS

Participants' median age was 49 years (male, 93%; Black, 51%; HBeAg+, 65%), with suppressed HBV DNA (79%) and undetectable HIV RNA (77%) on dually active antiretroviral therapy. Overall, HBV RNA and HBcrAg were quantifiable in 81% and 83%, respectively (96% and 100% in HBeAg+, respectively). HBcAg staining was detected in 60% and HBsAg in 79%. Higher HBV RNA was associated with higher HBcAg and HBsAg IHC grades (both p < 0.0001). The HBsAg membranous staining pattern was significantly associated with higher HBV-RNA and HBcrAg levels.

CONCLUSION

HBcAg and HBsAg IHC staining persisted despite viral suppression, and IHC grades and staining patterns correlated with markers of transcription (HBV RNA) and translation (HBcrAg). These data indicate that apparent HBV suppression is associated with residual transcription and translation that could contribute to liver pathology. Additional antiviral strategies directed to HBV protein expression may be useful to ameliorate liver injury.

Association between serum periostin levels and the severity of arsenic-induced skin lesions

Arsenic is a potent environmental toxicant and human carcinogen. Skin lesions are the most common manifestations of chronic exposure to arsenic. Advanced-stage skin lesions, particularly hyperkeratosis have been recognized as precancerous diseases. However, the underlying mechanism of arsenic-induced skin lesions remains unknown. Periostin, a matricellular protein, is implicated in the pathogenesis of many forms of skin lesions. The objective of this study was to examine whether periostin is associated with arsenic-induced skin lesions. A total of 442 individuals from low- (n = 123) and high-arsenic exposure areas (n = 319) in rural Bangladesh were evaluated for the presence of arsenic-induced skin lesions (Yes/No). Participants with skin lesions were further categorized into two groups: early-stage skin lesions (melanosis and keratosis) and advanced-stage skin lesions (hyperkeratosis). Drinking water, hair, and nail arsenic concentrations were considered as the participants' exposure levels. The higher levels of arsenic and serum periostin were significantly associated with skin lesions. Causal mediation analysis revealed the significant effect of arsenic on skin lesions through the mediator, periostin, suggesting that periostin contributes to the development of skin lesions. When skin lesion was used as a three-category outcome (none, early-stage, and advanced-stage skin lesions), higher serum periostin levels were significantly associated with both early-stage and advanced-stage skin lesions. Median (IQR) periostin levels were progressively increased with the increasing severity of skin lesions. Furthermore, there were general trends in increasing serum type 2 cytokines (IL-4, IL-5, IL-13, and eotaxin) and immunoglobulin E (IgE) levels with the progression of the disease. The median (IQR) of IL-4, IL-5, IL-13, eotaxin, and IgE levels were significantly higher in the early-and advanced-stage skin lesions compared to the group of participants without skin lesions. The results of this study suggest that periostin is implicated in the pathogenesis and progression of arsenic-induced skin lesions through the dysregulation of type 2 immune response.

A series of two-sample non-parametric tests for quantile residual life time

Quantile residual lifetime (QRL) is of significant interest in many clinical studies as an easily interpretable quantity compared to other summary measures of survival distributions. In cancer or other chronic diseases, treatments are often compared based on the distributions or quantiles of the residual lifetime. Thus a common problem of interest is to test the equality of the QRL between two populations. In this paper, we propose two classes of tests to compare two QRLs; one class is based on the difference between two estimated QRLs, and the other is based on the estimating function of the QRL, where the estimated QRL from one sample is plugged into the QRL-estimating-function of the other sample. We outline the asymptotic properties of these test statistics. Simulation studies demonstrate that the proposed tests produced Type I errors closer to the nominal level and are superior to some existing tests based on both Type I error and power. Our proposed test statistics are also computationally less intensive and more straightforward compared to tests based on the confidence intervals. We applied the proposed methods to a randomized multicenter phase III trial for breast cancer patients.

Acute Hepatitis B Virus Infection in North American Adults

BACKGROUND & AIMS

Acute hepatitis B virus (aHBV) is thought to be self-limited with clearance of hepatitis B surface antigen (HBsAg) within 6 months. There are limited reports of the presenting features and outcomes of adults with symptomatic aHBV in the United States.

METHODS

Demographics, clinical features, and 12-month outcomes of patients with adjudicated aHBV were captured prospectively and compared with a contemporaneous cohort of chronic HBV (cHBV) patients enrolled in the Hepatitis B Research Network.

RESULTS

Between 2011 and 2018, 60 adjudicated patients with aHBV were compared with 1534 cHBV untreated controls. Although similar in age, other features were dissimilar: aHBV patients were more often male (72% vs 51%), single (72% vs 30%), and non-Hispanic whites or blacks (75% vs 24%). They also were frequently genotype A₂ (65% vs 9%), having different risk factors: sexual exposure (75% vs 16%) or injection drug use (10% vs 2%), compared with the cHBV controls. In addition to higher serum aminotransferase and bilirubin levels, acute patients had higher HBV DNA levels (4.8 vs 3.6 log₁₀ IU/mL), whereas quantitative hepatitis B e antigen (HBeAg) levels were lower (1.4 vs 3.0 log₁₀ IU/mL), despite higher rates of HBeAg (73% vs 25%). The median time to HBsAg clearance was 27 weeks and to anti-HBs appearance, 41 weeks.

CONCLUSIONS

In the current era, caucasian men infected with genotype A₂ as a result of sexual exposure or injection drug use were the predominant group in aHBV, suggesting a potential strategy for adult vaccination in North America. Strikingly, only an estimated 36% of subjects cleared HBsAg by month 6; the definition of resolution in acute hepatitis B may need to be modified. ClinicalTirals.gov number NCT01263587.

Effect of HCV or HIV infection on mortality among hospitalized persons who used opioids, 2000-2010

Mortality due to opioid misuse and overdose has increased substantially in the United States over the past two decades. The study objective was to describe the causes of death among persons with opioid-related hospitalizations and examine survival by Hepatitis C virus (HCV) or HIV. Opioid-related hospitalization records in Pennsylvania from 2000 to 2010 were linked to death registry files to assess cause of death, and survival from first hospital discharge date to death date, or December 31, 2010. Accelerated failure time models were used to compare survival between persons with and without HCV or HIV diagnoses. Among the 136,416 individuals with an opioid-related hospitalization, 13.0% died over a median of 56 months of follow-up; the most common causes of death were circulatory diseases (26.4%) and drug overdose (23.5%). There were 27,122 (19.9%) and 3662 (2.7%) persons who had an HCV and HIV diagnosis, respectively. Among patients aged ≥20 years, those with HCV had shorter survival time compared to those without HCV, with discrepancies more pronounced at older ages. Patients with HIV also had shorter survival time (time ratio: 0.29 [95% CI: 0.26, 0.34]) compared to without HIV. These findings show that in a cohort of patients with opioid-related hospitalizations, those with HCV or HIV diagnoses have shorter survival. This has public health implications, providing further evidence that medical providers should educate patients who use opioids about the risks of HCV and HIV infection and focus prevention and treatment to decrease mortality among patients hospitalized for opioid use.

Hepatitis B RNA and Core-Related Antigen Provide Value Beyond DNA in Evaluating e But Not Surface Antigen Clearance

In chronic hepatitis B virus (HBV) infection, hepatitis B e antigen (HBeAg) and hepatitis B surface antigen (HBsAg) clearance are important milestones toward immune control.¹ A drop in HBV DNA is an established correlate of both HBeAg and HBsAg clearance.² We evaluated changes in HBV RNA and hepatitis B core-related antigen (HBcrAg) levels, markers of transcriptional activity of covalently closed circular DNA (cccDNA),^3,4 with HBeAg and HBsAg clearance, and compared them with changes in HBV DNA level among adult participants in the Hepatitis B Research Network (HBRN).

Elevated serum periostin levels among arsenic-exposed individuals and their associations with the features of asthma

Chronic exposure to arsenic via drinking water is a serious public health issue in many countries. Arsenic causes not only cancers but also non-malignant diseases, including asthma. We have previously reported that arsenic exposure increases the risk of Th2-mediated allergic asthma. The serum level of periostin, an extracellular matrix protein activated by Th2 cytokines, is recognized as a biomarker for Th2-mediated eosinophilic asthma and contributes to enhanced airway inflammation and remodeling. However, the role of periostin in arsenic-related asthma is unknown. Therefore, this study was designed to explore the associations of serum periostin levels with arsenic exposure and the features of asthma in 442 individuals in Bangladesh who participated in our previous study. Exposure levels of the participants were determined by measuring the arsenic concentrations in drinking water, hair, and nails through inductively coupled plasma mass spectroscopy. Periostin levels in serum were assessed by immunoassay. In this study, we found that serum periostin levels of the participants were increased with increasing exposure to arsenic. Notably, even the participants with 10.1-50 μg/L arsenic in drinking water had significantly higher levels of periostin than participants with <10 μg/L of water arsenic. Elevated serum periostin levels were positively associated with serum levels of Th2 mediators, such as interleukin (IL)-4, IL-5, IL-13, and eotaxin. Each log increase in periostin levels was associated with approximately eight- and three-fold increases in the odds ratios (ORs) for reversible airway obstruction (RAO) and asthma symptoms, respectively. Additionally, causal mediation analyses revealed that arsenic exposure metrics had both direct and indirect (periostin-mediated) effects on the risk of RAO and asthma symptoms. Thus, the results suggested that periostin may be involved in the arsenic-related pathogenesis of Th2-mediated asthma. The elevated serum periostin levels may predict the greater risk of asthma among the people living in arsenic-endemic areas.

Incidence and prediction of HBsAg seroclearance in a prospective multi-ethnic HBeAg-negative chronic hepatitis B cohort

BACKGROUND AND AIMS

Achieving HBsAg loss is an important landmark in the natural history of chronic hepatitis B (CHB). A more personalized approach to prediction of HBsAg loss is relevant in counseling patients. This study sought to develop and validate a prediction model for HBsAg loss based on quantitative HBsAg levels (qHBsAg) and other baseline characteristics.

METHODS

The Hepatitis B Research Network (HBRN) is a prospective cohort including 1240 untreated HBeAg-negative patients (1150 adults, 90 children) with median follow-up of 5.5 years. Incidence rates of HBsAg loss and hepatitis B surface antibody (anti-HBs) acquisition were determined, and a predictor score of HBsAg loss using readily available variables was developed and externally validated.

RESULTS

Crude incidence rates of HBsAg loss and anti-HBs acquisition were 1.6 and 1.1 per 100 person-years (PY); 67 achieved sustained HBsAg loss for an incidence rate of 1.2 per 100 PY. Increased HBsAg loss was significantly associated with older age, non-Asian race, HBV phenotype (inactive CHB vs. others), HBV genotype A, lower HBV-DNA levels, and lower and greater change in qHBsAg. The HBRN-SQuARe (sex,∆quantHBsAg, age, race) score predicted HBsAg loss over time with area under the receiver operating characteristic curve (AUROC) (95% CIs) at 1 and 3 years of 0.99 (95% CI: 0.987-1.00) and 0.95 (95% CI 0.91-1.00), respectively. In validation in another cohort of 1253 HBeAg-negative patients with median follow-up of 3.1 years, HBRN SQuARe predicted HBsAg loss at 1 and 3 years with AUROC values of 0.99 (0.98-1.00) and 0.88 (0.77-0.99), respectively.

CONCLUSION

HBsAg loss in predominantly untreated patients with HBeAg-negative CHB can be accurately predicted over a 3-year horizon using a simple validated score (HBRN SQuARe). This prognostication tool can be used to support patient care and counseling.

Prognostic accuracy for predicting ordinal competing risk outcomes using ROC surfaces

Many medical conditions are marked by a sequence of events in association with continuous changes in biomarkers. Few works have evaluated the overall accuracy of a biomarker in predicting disease progression. We thus extend the concept of receiver operating characteristic (ROC) surface and the volume under the surface (VUS) from multi-category outcomes to ordinal competing-risk outcomes that are also subject to noninformative censoring. Two VUS estimators are considered. One is based on the definition of the ROC surface and obtained by integrating the estimated ROC surface. The other is an inverse probability weighted U estimator that is built upon the equivalence of the VUS to the concordance probability between the marker and sequential outcomes. Both estimators have nice asymptotic results that can be derived using counting process techniques and U-statistics theory. We illustrate their good practical performances through simulations and applications to two studies of cognition and a transplant dataset.

A Prospective Study Evaluating Changes in Histology, Clinical and Virologic Outcomes in HBV-HIV Co-infected Adults in North America

BACKGROUND AND AIMS

Histological and clinical outcomes in HBV-HIV coinfection in the era of combination antiretroviral therapy (cART) are poorly defined.

APPROACH AND RESULTS

Adult patients co-infected with HBV-HIV from eight North American sites were enrolled in this National Institutes of Health (NIH)-funded prospective observational study (n = 139). Demographic, clinical, serological, and virological data were collected at entry and every 24 weeks for ≤ 192 weeks. Paired liver biopsies were obtained at study entry and at ≥ 3 years of follow-up. Biopsies were assessed by a central pathology committee using the modified Ishak scoring system. Clinical outcome rate and changes in histology are reported. Among participants with follow-up data (n = 114), median age was 49 years, 91% were male, 51% were non-Hispanic Black, and 13% had at-risk alcohol use, with a median infection of 20 years. At entry, 95% were on anti-HBV cART. Median CD4 count was 562 cells/mm³ and 93% had HIV < 400 copies/mL. HBeAg was positive in 61%, and HBV DNA was below the limit of quantification (< 20 IU/mL) in 61% and < 1,000 IU/mL in 80%. Clinical events were uncommon across follow-up: one hepatic decompensation, two HCC, no liver transplants, and one HBV-related deaths, with a composite endpoint rate of 0.61/100 person-years. Incident cirrhosis (n = 1), alanine aminotransferase flare (n = 2), and HBeAg loss (n = 13) rates were 0.40, 0.65, and 6.86 per 100 person-years, respectively. No participants had HBsAg loss. Paired biopsy (n = 62; median 3.6 years apart) revealed minimal improvement in Histologic Activity Index (median [interquartile range]: 3 [2-4] to 3 [1-3]; P = 0.02) and no significant change in fibrosis score (1 [1-2] to 1 [0-3]; P = 0.58).

CONCLUSIONS

In a North American cohort of adults with HBV-HIV on cART with virological suppression, clinical outcomes and worsening histological disease were uncommon.

Liver disease symptoms are associated with higher risk of adverse clinical outcomes: A longitudinal study of North American adults with chronic Hepatitis B

Background

Symptoms of chronic hepatitis B (CHB) are not well characterized.

Aims

To evaluate CHB symptoms and associations with disease activity and clinical outcomes.

Methods

Longitudinal data from 1,576 participants in the Hepatitis B Research Network Cohort Study who completed symptom assessments were analyzed. A composite symptom score was calculated using a Symptom Checklist (0=none to 40=extreme). Multivariable mixed models assessed variables associated with symptom change over time. Latent class symptom trajectories were evaluated. The cumulative probability of long-term clinical outcomes (new onset cirrhosis, hepatic decompensation, hepatocellular carcinoma, liver transplantation, death) was examined by baseline symptom groups.

Results

Participants median age was 42 (range:18-80), 51% were male, 75% Asian, (68% of whom were born outside North America) with a median follow-up of 4.2 years. On average, symptoms did not significantly change over time. The multivariable model identified several variables associated with higher symptoms during follow-up: being female, non-Asian, born in the US/Canada, lower education, higher AST, lower platelets, and more comorbidities. Two patient subgroups were identified based on longitudinal symptom trajectories: a low symptom group (92%, n=1,451) with symptom scores averaging 2.4 over time and a moderate symptom group (8%, n=125) with symptom scores averaging 11.5. During follow-up, 7.3% in the moderate symptom group, but only 3.2% of the low symptom group, developed adverse outcomes (p=0.02).

Conclusions

In this large cohort of CHB patients, symptoms were generally mild and stable over time. However, in some patients with moderate symptoms at baseline, deleterious clinical outcomes were more frequent in follow-up.

Adaptive randomization in a two-stage sequential multiple assignment randomized trial

Sequential multiple assignment randomized trials (SMARTs) are systematic and efficient media for comparing dynamic treatment regimes (DTRs), where each patient is involved in multiple stages of treatment with the randomization at each stage depending on the patient's previous treatment history and interim outcomes. Generally, patients enrolled in SMARTs are randomized equally to ethically acceptable treatment options regardless of how effective those treatments were during the previous stages, which results in some undesirable consequences in practice, such as low recruitment, less retention, and lower treatment adherence. In this article, we propose a response-adaptive SMART (RA-SMART) design where the allocation probabilities are imbalanced in favor of more promising treatments based on the accumulated information on treatment efficacy from previous patients and stages. The operating characteristics of the RA-SMART design relative to SMART design, including the consistency and efficiency of estimated response rate under each DTR, the power of identifying the optimal DTR, and the number of patients treated with the optimal and the worst DTRs, are assessed through extensive simulation studies. Some practical suggestions are discussed in the conclusion.

Precore and Basal Core Promoter Hepatitis B Virus (HBV) Variants Are Present From a Young Age and Differ Across HBV Genotypes

BACKGROUND AND AIMS

Hepatitis B virus (HBV) precore (PC) and dual basal core promoter (BCP) mutations halt and down-regulate hepatitis B e antigen (HBeAg) production respectively. PC mutation is rarely associated with HBV genotype A. We sought to examine the association of these variants with HBV genotypes, age, and HBeAg status in a racially diverse population in North America. Prospective study included 1,036 (808 adults, 228 children) participants in the Hepatitis B Research Network. PC and BCP variants were determined by Sanger sequencing, and dominant HBV species (>50%) were reported.

APPROACH AND RESULTS

Median age was 36.3 years (range, 2-80), 44.6% HBeAg(+), 74.2% Asians, 13.3% black, and 9.7% white. The dominant PC variant was present in 29.4% participants, including 20 with subgenotype A1 or A2. Seventeen of 20 participants with genotype A and PC had a compensatory C1858T mutation. In the HBeAg(+) cohort, the prevalence of PC and/or BCP variants increased from 14.4% in the first two decades to 51% after 40 years of age. Among those aged 2-18, 52% and 83% with dominant PC and BCP variants were HBeAg(+) compared to 3.8% and 29% in the >40 years age group. HBeAg clearance rates were significantly higher for those with dominant PC or BCP variants: 24.4 and 15.0 per 100 person-years compared to 6.0 in wild-type HBV (P < 0.0001).

CONCLUSIONS

PC variants can be present in HBV genotype A and are usually associated with C1858T, which preserves the pregenome encapsidation sequence. Selection of PC and BCP variants occurred at a young age, with increasing prevalence across age groups. HBeAg(+) participants with dominant PC and BCP variants progressed to the HBeAg(-) phase of chronic HBV infection significantly faster. This finding has potential clinical and therapeutic implications.

Highly multiplexed 2-dimensional imaging mass cytometry analysis of HBV-infected liver

Studies of human hepatitis B virus (HBV) immune pathogenesis are hampered by limited access to liver tissues and technologies for detailed analyses. Here, utilizing imaging mass cytometry (IMC) to simultaneously detect 30 immune, viral, and structural markers in liver biopsies from patients with hepatitis B e antigen⁺ (HBeAg⁺) chronic hepatitis B, we provide potentially novel comprehensive visualization, quantitation, and phenotypic characterizations of hepatic adaptive and innate immune subsets that correlated with hepatocellular injury, histological fibrosis, and age. We further show marked correlations between adaptive and innate immune cell frequencies and phenotype, highlighting complex immune interactions within the hepatic microenvironment with relevance to HBV pathogenesis.

Predictors of change in cardiovascular disease risk and events following gastric bypass: a 7-year prospective multicenter study

BACKGROUND

Change in short-term (i.e., 10-year) and lifetime risk of cardiovascular disease (CVD) following Roux-en-Y gastric bypass (RYGB) has significant heterogeneity.

OBJECTIVE

To identify predictors of change in CVD risk and cardiovascular events following RYGB.

METHODS

Between 2006-2009, 1625 adults without a history of CVD enrolled in a prospective cohort study and underwent RYGB at 1 of 10 U.S. hospitals. Participants were followed annually for a maximum of 7 years. Associations between presurgery characteristics (anthropometric, sociodemographic, physical and mental health, alcohol/drug use, eating behaviors) and 1) pre to postsurgery change in 10 year and lifetime atherosclerotic CVD (ASCVD) risk scores, respectively, and 2) having a CVD event (nonfatal myocardial infarction, stroke, ischemic heart disease, congestive heart failure, angina, percutaneous coronary intervention, coronary artery bypass grafting, or CVD-attributed death) as repeated measures (yr 1-7) were evaluated.

SETTING

Observational cohort study at ten hospitals throughout the United States.

RESULTS

Presurgery factors independently associated with decreases in both 10-year and lifetime risk scores 1-7 years post-RYGB were higher CVD risk score, female sex, higher household income, and normal kidney function. Additionally, Black race and having diabetes were independently associated with decreases in 10-year risk, while not having diabetes and a higher (better) composite mental health score were independently related to decreases in lifetime risk. A lower (worse) presurgery composite physical health score was associated with a higher CVD event risk (RR = 1.68, per 10 points).

CONCLUSION

This study identified multiple presurgery factors that characterize patients who may have more cardiovascular benefit from RYGB, and patients who might require additional support to improve their cardiovascular health.

Change in predicted 10-year and lifetime cardiovascular disease risk after Roux-en-Y gastric bypass

BACKGROUND

Long-term changes in cardiovascular disease (CVD) risk after bariatric surgery are not well characterized.

OBJECTIVE

To report sex-specific changes in CVD risk after Roux-en-Y gastric bypass surgery (RYGB).

SETTING

Observational cohort study at ten hospitals throughout the United States.

METHODS

Between 2006 and 2009, 1770 adults enrolled in a prospective cohort study underwent RYGB at 1 of 10 U.S. hospitals. Research assessments were conducted presurgery and annually postsurgery over 7 years. Sex specific-predicted 10-year and lifetime CVD risk were calculated using the Framingham10-year and lifetime risk scores, Framingham-body mass index, and atherosclerotic CVD scoring algorithms among participants with no history of CVD. Of 1566 eligible participants, 1234 (75.9%) with CVD risk determination pre- and postsurgery were included (1013 females, 221 males).

RESULTS

Based on the Framingham10-year and lifetime risk scores, the percentage of females with predicted high (>20%) 10-year CVD risk declined from presurgery (6.5% [95% confidence interval: 6.7-7.5]) to 1 year postsurgery (1.0% [95% confidence interval: .8-1.2]; P < .001), then increased 1 to 7 years postsurgery (to 2.8% [95% confidence interval: 1.6-3.3]; P = .003), but was lower 7 years postsurgery versus presurgery (P < .001). Time trends for percentage of high-risk participants and mean CVD risk scores were similar for both sexes and other evaluated CVD risk scores. For example, among males mean lifetime atherosclerotic CVD score declined from presurgery to 1 year postsurgery, then increased 1 to 7 years postsurgery. However, there was a net decline from presurgery (P < .001).

CONCLUSION

Among both females and males, predicted 10-year and lifetime CVD risk was substantially lower 7 years post RYGB than presurgery, suggesting RYGB surgery can lead to sustained improvements in short- and long-term CVD risk.

Cumulative incidence regression for dynamic treatment regimens

Recently dynamic treatment regimens (DTRs) have drawn considerable attention, as an effective tool for personalizing medicine. Sequential Multiple Assignment Randomized Trials (SMARTs) are often used to gather data for making inference on DTRs. In this article, we focus on regression analysis of DTRs from a two-stage SMART for competing risk outcomes based on cumulative incidence functions (CIFs). Even though there are extensive works on the regression problem for DTRs, no research has been done on modeling the CIF for SMART trials. We extend existing CIF regression models to handle covariate effects for DTRs. Asymptotic properties are established for our proposed estimators. The models can be implemented using existing software by an augmented-data approximation. We show the improvement provided by our proposed methods by simulation and illustrate its practical utility through an analysis of a SMART neuroblastoma study, where disease progression cannot be observed after death.

Evaluating Noninvasive Markers to Identify Advanced Fibrosis by Liver Biopsy in HBV/HIV Co-infected Adults

Noninvasive biomarkers are used increasingly to assess fibrosis in patients with chronic liver disease. We determined the utility of dual cutoffs for noninvasive biomarkers to exclude and confirm advanced fibrosis in hepatitis B virus (HBV)-human immunodeficiency virus (HIV) co-infected patients receiving combined antiretroviral therapy. Participants were anti-HIV/hepatitis B surface antigen-positive adults from eight clinical sites in the United States and Canada of the Hepatitis B Research Network. Fibrosis was staged by a central pathology committee using the Ishak fibrosis score (F). Clinical, laboratory, and vibration-controlled transient elastography (VCTE) data were collected at each site. Dual cutoffs for three noninvasive biomarkers (aspartate aminotransferase-to-platelet ratio index, Fibrosis-4 index [FIB-4], and liver stiffness by VCTE) with the best accuracy to exclude or confirm advanced fibrosis (F ≥ 3) were determined using established methodology. Of the 139 enrolled participants, 108 with a liver biopsy and having at least one noninvasive biomarker were included: 22% had advanced fibrosis and 54% had normal alanine aminotransferase. The median (interquartile range) of APRI (n = 106), FIB-4 (n = 106), and VCTE (n = 63) were 0.34 (0.26-0.56), 1.35 (0.99-1.89), and 4.9 (3.8-6.8) kPa, respectively. The area under the curve for advanced fibrosis was 0.69 for APRI, 0.66 for FIB-4, and 0.87 for VCTE. VCTE cutoffs of 5.0 kPa or less (to exclude) and 8.8 kPa or greater (to confirm) advanced fibrosis had a sensitivity of 92.3% and specificity of 96.0%, respectively, and accounted for 65.1% of participants. Among the 34.9% with values between the cutoffs, 26.1% had advanced fibrosis. Considering APRI or FIB-4 jointly with VCTE did not improve the discriminatory capacity. Conclusion: VCTE is a better biomarker of advanced fibrosis compared with APRI or FIB-4 in HBV/HIV co-infected adults on combined antiretroviral therapy. Using VCTE dual cutoffs, approximately two-thirds of patients could avoid biopsy to determine advanced fibrosis.

Patient-reported outcomes in a large North American cohort living with chronic hepatitis B virus: a cross-sectional analysis

BACKGROUND

Patient-reported outcomes (PROs) such as health-related quality of life (HRQoL) and symptoms associated with chronic hepatitis B viral (HBV) infection have not been well-described in North American cohorts.

AIMS

To evaluate several PROs and associations with HBV disease activity markers.

METHODS

Cross-sectional analysis including 876 adults who completed PRO measures during the Hepatitis B Research Network Adult Cohort Study. Participants on HBV treatment were excluded. Outcomes included: HRQoL using the SF-36 mental component summary and physical component summary scores; symptom burden using a 10-item Total Symptom Checklist and fatigue using an instrument from the Patient-Reported Outcomes Measurement Information System®. Covariates included laboratory markers of disease severity, virological status, comorbidities and medications.

RESULTS

Median age was 42 (range: 19-79), 51% were female, 73% Asian, 19% HBeAg (+), 2% had AST-platelet ratio index (APRI) ≥1.5 and 74% without comorbidities. Mean mental component summary T-score = 52, physical component summary T-score = 54 and PROMIS Fatigue T-score = 47. On a scale from 0 (none) to 40 (extreme), the mean Symptom Checklist score = 3 and 25% reported no symptoms. The most frequent symptoms were fatigue (60%), irritability (32%) and itching (32%). Most symptoms were 'a little bit' bothersome. In multivariable regressions, APRI ≥1.50 and more comorbidities were associated with worse patient-reported outcomes; virological markers were not. Adding the Total Symptom Checklist score to original regression models increased explanation of variation in the mental component summary score from 4% to 44% and the Physical Component Summary Score from 17% to 34%.

CONCLUSIONS

Untreated North American HBV patients with mild liver disease report favourable health-related quality of life and minimal symptoms. HBV does not impact health-related quality of life unless advanced liver disease or comorbidities are present. High symptom burden explains substantial variation in health-related quality of life. (CT.gov identifier: NCT01263587).

Alcohol, tobacco and coffee consumption and liver disease severity among individuals with Chronic Hepatitis B infection in North America

INTRODUCTION AND OBJECTIVES

The prevalence of alcohol, tobacco, and coffee use and association with liver health among North Americans with Chronic Hepatitis B (CHB) infection has not been well described.

MATERIALS AND METHODS

The Hepatitis B Research Network includes an observational study of untreated CHB adults enrolled at 21 sites in the United States and Canada. Alcohol use was categorized as none, moderate, and at-risk based on the definition from the National Institute on Alcohol Abuse and Alcoholism; tobacco use as never, current and former; coffee use as none, 1-2 cups/day, and ≥3 cups/day. Linear regression and linear mixed models were used to associate lifestyle behaviors with ALT and FIB-4 values.

RESULTS

1330 participants met eligibility: 53% males, 71% Asian and the median age was 42 years (IQR: 34-52). Median ALT was 33U/L (IQR: 22-50), 37% had HBV DNA <10³IU/mL, 71% were HBeAg negative, and 65% had a FIB-4 <1.45. At baseline, 8% of participants were at-risk alcohol drinkers, 11% were current smokers and 92% drank <3 cups of coffee/day. Current tobacco and 'at-risk' alcohol use, were significantly associated with elevated ALT levels in univariable analyses, however, these associations were not statistically significant when controlling for sociodemographic and HBV characteristics.

CONCLUSIONS

In this large diverse cohort of untreated CHB participants, at-risk alcohol use, current tobacco use and limited coffee consumption did not have an association with high ALT and FIB-4 values. In contrast, significant associations were found between the frequency of these lifestyle behaviors and sociodemographic factors.

Improving Patient Reported Outcomes and Preventing Depression and Anxiety in Older Adults With Knee Osteoarthritis: Results of a Sequenced Multiple Assignment Randomized Trial (SMART) Study

OBJECTIVE

Older adults with knee osteoarthritis (OA) and comorbid subsyndromal depressive symptoms are at elevated risk for incidental major depression or anxiety disorders. Using an indicated prevention paradigm, the authors conducted a sequenced multiple assignment randomized trial (SMART) to: 1) evaluate the effect of cognitive behavioral therapy (CBT) and physical therapy (PT), together with the temporal ordering of these interventions, on patient-reported global impression of change (P-GIC), mood, anxiety, and pain; and 2) compare the strategies' impact on incidence of common psychiatric disorders over 12-months.

METHODS

This intervention development trial compared four adaptive strategies delivered in two stages (each up to 8 weeks), contrasted with enhanced usual care (EUC). The strategies were CBT followed by an increased dose of CBT (CBT-CBT), CBT followed by PT (CBT-PT), PT followed by an increased dose of PT (PT-PT), and PT followed by CBT (PT-CBT). Participants (n = 99) were aged 60 years and older and met clinical criteria for knee OA and subthreshold depression. Response was defined as at least "much better" on the P-GIC. Participants were assessed quarterly for 12 months for incidence of psychiatric disorders.

RESULTS

Stage 1 response was higher for PT (47.5%) compared to CBT (20.5%). Non-responders receiving an additional dose of the same intervention experienced a response rate of 73%, higher than for switching to a different intervention. All strategies were superior to EUC (5%). Although not powered to detect effects on disorders, neither intervention strategy nor response status affected 12-month incidence of depression and anxiety disorders.

CONCLUSION

As response rates were similar for PT-PT and CBT-CBT, it may be dose and not type of these interventions that are necessary for clinical benefit. For non-responders, this finding may guide providers to stay the clinical course for up to 12 weeks before switching. These results support future trials of SMART designs in late-life depression prevention.

Silymarin in non-cirrhotics with non-alcoholic steatohepatitis: A randomized, double-blind, placebo controlled trial

The botanical product silymarin, an extract of milk thistle, is commonly used by patients to treat chronic liver disease and may be a treatment for NASH due to its antioxidant properties. We aimed to assess the safety and efficacy of higher than customary doses of silymarin in non-cirrhotic patients with NASH. This exploratory randomized double-blind placebo controlled multicenter Phase II trial tested a proprietary standardized silymarin preparation (Legalon^®, Rottapharm|Madaus, Mylan) and was conducted at 5 medical centers in the United States. Eligible adult patients had liver biopsy within 12 months showing NASH without cirrhosis with NAFLD Activity Score (NAS) ≥4 per site pathologist’s assessment. Participants were randomized to Legalon^® 420 mg, 700 mg, or placebo t.i.d. for 48 weeks. The primary endpoint was histological improvement ≥2 points in NAS. Of 116 patients screened, 78 were randomized. There were no significant differences in adverse events among the treatment groups. After 48–50 weeks, 4/27 (15%) in the 700 mg dose, 5/26 (19%) participants randomized to 420 mg, and 3/25 (12%) of placebo recipients reached the primary endpoint (p = 0.79) among all randomized participants, indicating no benefit from silymarin in the intention to treat analysis Review by a central pathologist demonstrated that a substantial number of participants (49, 63%) did not meet histological entry criteria and that fibrosis stage improved most in the placebo treated group, although not significantly different from other groups. Silymarin (Legalon^®) at the higher than customary doses tested in this study is safe and well tolerated. The effect of silymarin in patients with NASH remains inconclusive due to the substantial number of patients who entered the study but did not meet entry histological criteria, the lack of a statistically significant improvement in NAS of silymarin treated patients, and the unanticipated effect of placebo on fibrosis indicate the need for additional clinical trials.

Trial Registration: clinicaltrials.gov, Identifier: NCT00680407.

The Association Between Kidney Disease and Diabetes Remission in Bariatric Surgery Patients With Type 2 Diabetes

RATIONALE & OBJECTIVE

The association between bariatric surgery, type 2 diabetes, and chronic kidney disease (CKD) is poorly understood. We studied whether remission of type 2 diabetes induced by bariatric surgery influences markers of kidney disease, if CKD is associated with remission of diabetes after bariatric surgery, and if baseline levels of gut hormones and peptides modify these associations.

STUDY DESIGN

Prospective observational study.

STUDY PARTICIPANTS

737 bariatric surgery patients with type 2 diabetes who participated in a multicenter cohort study for up to 5 years.

PREDICTORS

Demographics, blood pressure, medications, type of bariatric surgery, anthropometrics, markers of kidney disease, and circulating levels of gut hormones and peptides.

OUTCOMES

Estimated glomerular filtration rate (eGFR), urinary albumin excretion, prognostic risk for CKD, and remission of diabetes.

ANALYTICAL APPROACH

Linear mixed models for eGFR; generalized linear mixed models with logit link for albuminuria, prognostic risk for CKD, and diabetes remission.

RESULTS

Remission of diabetes at 5 years post-bariatric surgery was not independently associated with eGFR but was associated with lower risk for moderate/severe increase in albuminuria (risk ratio, 0.66; 95% CI, 0.48-0.90) and stabilization in prognostic risk for CKD. These findings were modified by baseline ghrelin level. Lower preoperative eGFR and greater prognostic risk for CKD were independently associated with reduced likelihood of diabetes remission. The association with preoperative GFR was modified by C-peptide level. Higher baseline circulating ghrelin level was independently associated with a lower prognostic risk for CKD.

LIMITATIONS

A minority of participants had baseline CKD; lack of comparison group; no information on duration of diabetes, other clinical end points, or kidney biopsy results.

CONCLUSIONS

Remission of type 2 diabetes 5 years after bariatric surgery was associated with improvements in albuminuria and stabilized prognostic risk for CKD, but not with eGFR. Lower kidney function and greater prognostic risk at the time of bariatric surgery was linked to a lower likelihood of diabetes remission. These results highlight the need to identify the mechanisms through which bariatric surgery may delay the long-term progression of CKD in type 2 diabetes.

Spectrum of Liver Disease in Hepatitis B Virus (HBV) Patients Co-infected with Human Immunodeficiency Virus (HIV): Results of the HBV-HIV Cohort Study

BACKGROUND

Because most HBV/HIV co-infected patients on combination antiretroviral therapy (cART) have suppressed HBV DNA and normal liver enzymes, the histologic spectrum of liver disease in HBV/HIV coinfection is poorly defined. To address this gap in knowledge, we conducted a prospective study to comprehensively characterize liver disease severity assessed by liver biopsy in a well-defined cohort of HBV/HIV patients in North America receiving cART.

METHODS

Adult HIV/HBsAg positive patients on stable cART were recruited. Demographic, clinical, serological, and virological data were collected. Liver histology was assessed by a central pathology committee. The association of demographic, clinical, serologic, and virologic characteristics with liver histology was assessed using logistic regression.

RESULTS

In this cross-sectional analysis, the mean age of the cohort (N = 139) was 49 years; 92% were male, 51% were non-Hispanic black, 7% had at-risk alcohol use with a median duration of infections of 14 years. The median ALT was 28 IU/L and CD4 count was 568 cells/mm. Almost all (99%) were on cART. Three-fourths (75%) had undetectable HIV RNA (<20 copies/mL). HBeAg was positive in 62%, HBV DNA was below the limit of quantification (<20 IU/mL) in 57% and <1000 IU/ mL in 80%; 7% had incomplete viral suppression (HBV DNA ≥1000 IU/mL and HIV RNA <20 copies/mL). Liver histology (available in n = 114) showed significant periportal, lobular, and portal inflammation (scores ≥2) in 14%, 31%, and 22% respectively. Over a third (37%) had significant fibrosis (Ishak stage ≥2); 24% had advanced fibrosis (Ishak stage ≥3). Higher ALT (adjusted OR 1.19 per 10 IU/L; 95% CI [1.01, 1.41]; p = 0.03) and lower platelet count (adjusted OR 0.81 per 20,000 mm; 95% CI [0.67-0.97]; p = 0.02) but not HBV DNA were independently associated with advanced fibrosis.

CONCLUSIONS

In this cohort of patients with HBV/HIV coinfection receiving long-term cART with viral suppression, we observed significant fibrosis in more than one-third of patients.

Association of Obesity Subtypes in the Longitudinal Assessment of Bariatric Surgery Study and 3-Year Postoperative Weight Change

OBJECTIVE

The study objective was to empirically identify subgroups of patients with obesity and investigate their association with postoperative weight change.

METHODS

A longitudinal analysis of 2,458 adults in the Longitudinal Assessment of Bariatric Surgery (LABS) study was used. Baseline data were used to identify subgroups. The outcome was 3-year weight change after bariatric surgery.

RESULTS

We identified four classes (subtypes) of obesity, which could be characterized as diabetes with low rates of high-density lipoprotein (Class 1), disordered eating (Class 2), mixed (Class 3), and extreme obesity with early onset (Class 4). Approximately 98% of participants in Class 1 had diabetes compared with < 40% in the other classes. There were high rates of binge eating in Class 2, and more than 92% of those in this class reported eating when not hungry. Class 4 was characterized by a higher BMI at baseline. Adults in Class 4 lost an average of 25.0% (males) and 30.3% (females) of their baseline weight over 3 years. In contrast with participants in Class 1, those in Classes 2 and 3 had significantly larger 3-year weight losses than their peers in Class 4.

CONCLUSIONS

Obesity is a heterogeneous disease. Bariatric surgery may be most beneficial for adults with disordered eating.

Comparison of the Performance of Common Measures of Weight Regain After Bariatric Surgery for Association With Clinical Outcomes

IMPORTANCE

Estimates of weight regain following bariatric surgery vary widely.

OBJECTIVE

To describe weight regain after reaching nadir weight following Roux-en-Y gastric bypass (RYGB) surgery and compare weight regain measures for association with outcomes.

DESIGN, SETTING, AND PARTICIPANTS

Prospective cohort study of 2458 adults who underwent bariatric surgery at 10 hospitals in 6 US cities between March 2006 and April 2009. Assessments were conducted within 30 days' presurgery, at 6 months' postsurgery, and then annually until January 2015. Of the 1703 participants who underwent RYGB surgery, 1406 (83%) were followed up for 5 years or longer and had 5 or more weight measurements (excluding those who died or underwent surgical reversal).

EXPOSURES

Weight regain assessed by 5 continuous measures (weight in kilograms, body mass index [BMI], percentage of presurgery weight, percentage of nadir weight, and percentage of maximum weight lost) and 8 dichotomous measures (per established thresholds) were compared in relation to clinical outcomes based on statistical significance, magnitude of association, and model fit.

MAIN OUTCOMES AND MEASURES

Progression of diabetes, hyperlipidemia, and hypertension and declines in physical and mental health-related quality of life and satisfaction with surgery.

RESULTS

Among the 1406 participants who underwent RYGB surgery, the median age was 47 years (25th-75th percentile, 38-55 years) and the median BMI was 46.3 (25th-75th percentile, 42.3-51.8) prior to surgery. Most participants were female (80.3%) and white (85.6%). The median follow-up was 6.6 years (25th-75th percentile, 5.9-7.0 years). The median percentage of maximum weight loss was 37.4% (25th-75th percentile, 31.6%-43.3%) of presurgery weight and occurred a median of 2.0 years after RYGB surgery (25th-75th percentile, 1.0-3.2 years). The rate of weight regain was highest during the first year after reaching nadir weight, but weight regain continued to increase throughout follow-up (range, a median of 9.5% of maximum weight lost [25th-75th percentile, 4.7%-17.2%] to 26.8% of maximum weight lost [25th-75th percentile, 16.7%-41.5%] 1 to 5 years after reaching nadir weight). The percentage of participants who regained weight depended on threshold (eg, 5 years after nadir weight, 43.6% regained ≥5 BMI points; 50.2% regained ≥15% of nadir weight; and 67.3% regained ≥20% of maximum weight lost). Compared with other continuous weight regain measures, the percentage of maximum weight lost had the strongest association and best model fit for all outcomes except hyperlipidemia, which had a slightly stronger association with BMI. Of the dichotomous measures, 20% or greater of maximum weight lost performed better or similarly with most of the outcomes, and was the second best measure for hyperlipidemia (after ≥10 kg of weight) and hypertension (after ≥10% of maximum weight lost).

CONCLUSIONS AND RELEVANCE

Among a large cohort of adults who underwent RYGB surgery, weight regain quantified as percentage of maximum weight lost performed better for association with most clinical outcomes than the alternatives examined. These findings may inform standardizing the measurement of weight regain in studies of bariatric surgery.

Prospective evaluation of insulin and incretin dynamics in obese adults with and without diabetes for 2 years after Roux-en-Y gastric bypass

AIMS/HYPOTHESIS

In this prospective case-control study we tested the hypothesis that, while long-term improvements in insulin sensitivity (S_I) accompanying weight loss after Roux-en-Y gastric bypass (RYGB) would be similar in obese individuals with and without type 2 diabetes mellitus, stimulated-islet-cell insulin responses would differ, increasing (recovering) in those with diabetes but decreasing in those without. We investigated whether these changes would occur in conjunction with favourable alterations in meal-related gut hormone secretion and insulin processing.

METHODS

Forty participants with type 2 diabetes and 22 participants without diabetes from the Longitudinal Assessment of Bariatric Surgery (LABS-2) study were enrolled in a separate, longitudinal cohort (LABS-3 Diabetes) to examine the mechanisms of postsurgical diabetes improvement. Study procedures included measures of S_I, islet secretory response and gastrointestinal hormone secretion after both intravenous glucose (frequently-sampled IVGTT [FSIVGTT]) and a mixed meal (MM) prior to and up to 24 months after RYGB.

RESULTS

Postoperatively, weight loss and S_I-_FSIVGTT improvement was similar in both groups, whereas the acute insulin response to glucose (AIRglu) decreased in the non-diabetic participants and increased in the participants with type 2 diabetes. The resulting disposition indices (DI_FSIVGTT) increased by three- to ninefold in both groups. In contrast, during the MM, total insulin responsiveness did not significantly change in either group despite durable increases of up to eightfold in postprandial glucagon-like peptide 1 levels, and S_I-MM and DI_MM increased only in the diabetes group. Peak postprandial glucagon levels increased in both groups.

CONCLUSIONS/INTERPRETATION

For up to 2 years following RYGB, obese participants without diabetes showed improvements in DI that approach population norms. Those with type 2 diabetes recovered islet-cell insulin secretion response yet continued to manifest abnormal insulin processing, with DI values that remained well below population norms. These data suggest that, rather than waiting for lifestyle or medical failure, RYGB is ideally considered before, or as soon as possible after, onset of type 2 diabetes.

TRIAL REGISTRATION

ClinicalTrials.gov NCT00433810.

Effect of Bariatric Surgery on CKD Risk

Obesity is linked to the development and progression of CKD, but whether bariatric surgery protects against CKD is poorly understood. We, therefore, examined whether bariatric surgery influences CKD risk. The study included 2144 adults who underwent bariatric surgery from March of 2006 to April of 2009 and participated in the Longitudinal Assessment of Bariatric Surgery-2 Study cohort. The primary outcome was CKD risk categories as assessed by the Kidney Disease Improving Global Outcomes (KDIGO) consortium criteria using a combination of eGFR and albuminuria. Patients were 79% women and 87% white, with a median age of 46 years old. Improvements were observed in CKD risk at 1 and 7 years after surgery in patients with moderate baseline CKD risk (63% and 53%, respectively), high baseline risk (78% and 56%, respectively), and very high baseline risk (59% and 23%, respectively). The proportion of patients whose CKD risk worsened was ≤10%; five patients developed ESRD. Sensitivity analyses using year 1 as baseline to minimize the effect of weight loss on serum creatinine and differing eGFR equations offered qualitatively similar results. Treatment with bariatric surgery associated with an improvement in CKD risk categories in a large proportion of patients for up to 7 years, especially in those with moderate and high baseline risk. These findings support consideration of CKD risk in evaluation for bariatric surgery and further study of bariatric surgery as a treatment for high-risk obese patients with CKD.

Joint modeling of multiple ordinal adherence outcomes via generalized estimating equations with flexible correlation structure

Adherence to medication is critical in achieving effectiveness of many treatments. Factors that influence adherence behavior have been the subject of many clinical studies. Analyzing adherence is complicated because it is often measured on multiple drugs over a period, resulting in a multivariate longitudinal outcome. This paper is motivated by the Viral Resistance to Antiviral Therapy of Chronic Hepatitis C study, where adherence is measured on two drugs as a bivariate ordinal longitudinal outcome. To analyze such outcome, we propose a joint model assuming the multivariate ordinal outcome arose from a partitioned latent multivariate normal process. We also provide a flexible multilevel association structure covering both between and within outcome correlation. In simulation studies, we show that the joint model provides unbiased estimators for regression parameters, which are more efficient than those obtained through fitting separate model for each outcome. The joint method also yields unbiased estimators for the correlation parameters when the correlation structure is correctly specified. Finally, we analyze the Viral Resistance to Antiviral Therapy of Chronic Hepatitis C adherence data and discuss the findings.

Estimation of discrete survival function for error-prone diagnostic tests

The product limit or Kaplan-Meier (KM) estimator is commonly used to estimate the survival function in the presence of incomplete time to event. Application of this method assumes inherently that the occurrence of an event is known with certainty. However, the clinical diagnosis of an event is often subject to misclassification due to assay error or adjudication error, by which the event is assessed with some uncertainty. In the presence of such errors, the true distribution of the time to first event would not be estimated accurately using the KM method. We develop a method to estimate the true survival distribution by incorporating negative predictive values and positive predictive values, into a KM-like method of estimation. This allows us to quantify the bias in the KM survival estimates due to the presence of misclassified events in the observed data. We present an unbiased estimator of the true survival function and its variance. Asymptotic properties of the proposed estimators are provided, and these properties are examined through simulations. We demonstrate our methods using data from the Viral Resistance to Antiviral Therapy of Hepatitis C study.

Bayesian hierarchical joint modeling of repeatedly measured continuous and ordinal markers of disease severity: Application to Ugandan diabetes data

Modeling of correlated biomarkers jointly has been shown to improve the efficiency of parameter estimates, leading to better clinical decisions. In this paper, we employ a joint modeling approach to a unique diabetes dataset, where blood glucose (continuous) and urine glucose (ordinal) measures of disease severity for diabetes are known to be correlated. The postulated joint model assumes that the outcomes are from distributions that are in the exponential family and hence modeled as multivariate generalized linear mixed effects model associated through correlated and/or shared random effects. The Markov chain Monte Carlo Bayesian approach is used to approximate posterior distribution and draw inference on the parameters. This proposed methodology provides a flexible framework to account for the hierarchical structure of the highly unbalanced data as well as the association between the 2 outcomes. The results indicate improved efficiency of parameter estimates when blood glucose and urine glucose are modeled jointly. Moreover, the simulation studies show that estimates obtained from the joint model are consistently less biased and more efficient than those in the separate models.

Characteristics of US-Born Versus Foreign-Born Americans of African Descent With Chronic Hepatitis B

Hepatitis B virus (HBV) infection is more common in African Americans than in white Americans. We compared the epidemiologic, clinical, and virological characteristics of US-born African Americans (USAAs) to those of foreign-born African Americans (FBAAs) with chronic hepatitis B. The adult cohort study of the Hepatitis B Research Network enrolls patients with HBV infection from 21 clinical sites in the United States and Canada. A total of 237 (15%) of the adult participants with chronic HBV infection that were enrolled from January 20, 2011, to October 2, 2013, were of African descent, including 57 USAAs and 180 FBAAs (76%). Compared with FBAAs, USAAs were older and more likely to have acquired HBV through sexual exposure, to be HBeAg-positive, to have higher HBV DNA levels, and to be infected with HBV genotype A2. FBAAs from West Africa were more likely to have elevated serum alanine aminotransferase (72% vs. 50%; P < 0.01) and higher HBV DNA levels (median, 3.2 log10 IU/mL vs. 2.8 log10 IU/mL; P = 0.03) compared with East African FBAAs. The predominant HBV genotype among West African FBAAs was E (67%), whereas genotypes A (78%) and D (16%) were common in East African FBAAs. Significant differences were found between USAAs and FBAAs, highlighting the need for tailored strategies for prevention and management of chronic HBV infection for African Americans.

Assessing the Validity of Self-Reported Medication Adherence in Hepatitis C Treatment

Objective:

To assess the validity of self-reported medication adherence provided by individuals in treatment for hepatitis C virus (HCV) infection with a regimen of peginterferon and ribavirin.

Methods:

Adherence was evaluated prospectively among 196 African American and 205 white subjects enrolled in Virahep-C (Viral Resistance to Antiviral Therapy of Chronic Hepatitis C), a treatment study for genotype 1 HCV infection. Adherence to the prescribed dose was measured by 2 methods: self-report questions administered during multiple clinic visits, using a touch screen computer; and recordings of bottle openings, using an electronic monitor placed inside the cap of prescription containers. Self-reported responses were compared with the electronic monitor data. Nonparametric tests were used to test the association between adherence measures at 4, 12, 24, 36, and 48 weeks of treatment.

Results:

The estimated proportion of participants who were adherent prior to a given visit ranged from 85% to 97% (ribavirin) and 97% to 100% (peginterferon) by self report and from 69% to 90% (ribavirin) and 64% to 100% (peginterferon) by electronic monitors. For ribavirin, the percentage of cases in which the 2 measurement methods agreed varied from 68% to 90%; peginterferon agreement was from 84% to 100%. Overall, adherence was higher for peginterferon than for ribavirin but decreased over time for both medications. Self-reported adherence was usually higher than that assessed by electronic measures, and the level of discrepancy increased during the course of treatment.

Conclusions:

Adherence to peginterferon and ribavirin decreased gradually during therapy but remained relatively high. Simple self-reported measures can be used to screen for nonadherence to HCV drug therapy, but should be considered as overestimation of the actual amounts taken.

Bayesian Nonparametric Estimation for Dynamic Treatment Regimes with Sequential Transition Times

We analyze a dataset arising from a clinical trial involving multi-stage chemotherapy regimes for acute leukemia. The trial design was a 2 × 2 factorial for frontline therapies only. Motivated by the idea that subsequent salvage treatments affect survival time, we model therapy as a dynamic treatment regime (DTR), that is, an alternating sequence of adaptive treatments or other actions and transition times between disease states. These sequences may vary substantially between patients, depending on how the regime plays out. To evaluate the regimes, mean overall survival time is expressed as a weighted average of the means of all possible sums of successive transitions times. We assume a Bayesian nonparametric survival regression model for each transition time, with a dependent Dirichlet process prior and Gaussian process base measure (DDP-GP). Posterior simulation is implemented by Markov chain Monte Carlo (MCMC) sampling. We provide general guidelines for constructing a prior using empirical Bayes methods. The proposed approach is compared with inverse probability of treatment weighting, including a doubly robust augmented version of this approach, for both single-stage and multi-stage regimes with treatment assignment depending on baseline covariates. The simulations show that the proposed nonparametric Bayesian approach can substantially improve inference compared to existing methods. An R program for implementing the DDP-GP-based Bayesian nonparametric analysis is freely available at https://www.ma.utexas.edu/users/yxu/.

Effect of Wearable Technology Combined With a Lifestyle Intervention on Long-term Weight Loss: The IDEA Randomized Clinical Trial

IMPORTANCE

Effective long-term treatments are needed to address the obesity epidemic. Numerous wearable technologies specific to physical activity and diet are available, but it is unclear if these are effective at improving weight loss.

OBJECTIVE

To test the hypothesis that, compared with a standard behavioral weight loss intervention (standard intervention), a technology-enhanced weight loss intervention (enhanced intervention) would result in greater weight loss.

DESIGN, SETTING, PARTICIPANTS

Randomized clinical trial conducted at the University of Pittsburgh and enrolling 471 adult participants between October 2010 and October 2012, with data collection completed by December 2014.

INTERVENTIONS

Participants were placed on a low-calorie diet, prescribed increases in physical activity, and had group counseling sessions. At 6 months, the interventions added telephone counseling sessions, text message prompts, and access to study materials on a website. At 6 months, participants randomized to the standard intervention group initiated self-monitoring of diet and physical activity using a website, and those randomized to the enhanced intervention group were provided with a wearable device and accompanying web interface to monitor diet and physical activity.

MAIN OUTCOMES AND MEASURES

The primary outcome of weight was measured over 24 months at 6-month intervals, and the primary hypothesis tested the change in weight between 2 groups at 24 months. Secondary outcomes included body composition, fitness, physical activity, and dietary intake.

RESULTS

Among the 471 participants randomized (body mass index [BMI], 25 to <40; age range, 18-35 years; 28.9% nonwhite, 77.2% women), 470 (233 in the standard intervention group, 237 in the enhanced intervention group) initiated the interventions as randomized, and 74.5% completed the study. For the enhanced intervention group, mean base line weight was 96.3 kg (95% CI, 94.2-98.5) and 24-month weight 92.8 kg (95% CI, 90.6- 95.0) [corrected]. For the standard intervention group, mean baseline weight was 95.2kg (95%CI,93.0-97.3)and24-month weight was 89.3 kg (95%CI, 87.1-91.5) [corrected]. Weight change at 24 months differed significantly by intervention group (estimated mean weight loss, 3.5 kg [95% CI, 2.6-4.5} in the enhanced intervention group and 5.9 kg [95% CI, 5.0-6.8] in the standard intervention group; difference, 2.4 kg [95% CI, 1.0-3.7]; P = .002). Both groups had significant improvements in body composition, fitness, physical activity, and diet, with no significant difference between groups.

CONCLUSIONS AND RELEVANCE

Among young adults with a BMI between 25 and less than 40, the addition of a wearable technology device to a standard behavioral intervention resulted in less weight loss over 24 months. Devices that monitor and provide feedback on physical activity may not offer an advantage over standard behavioral weight loss approaches.

TRIAL REGISTRATION

clinicaltrials.gov Identifier: NCT01131871.

Type 2 Diabetes Remission Rates After Laparoscopic Gastric Bypass and Gastric Banding: Results of the Longitudinal Assessment of Bariatric Surgery Study

OBJECTIVE

The goals of this study were to determine baseline and postbariatric surgical characteristics associated with type 2 diabetes remission and if, after controlling for differences in weight loss, diabetes remission was greater after Roux-en-Y gastric bypass (RYGBP) than laparoscopic gastric banding (LAGB).

RESEARCH DESIGN AND METHODS

An observational cohort of obese participants was studied using generalized linear mixed models to examine the associations of bariatric surgery type and diabetes remission rates for up to 3 years. Of 2,458 obese participants enrolled, 1,868 (76%) had complete data to assess diabetes status at both baseline and at least one follow-up visit. Of these, 627 participants (34%) were classified with diabetes: 466 underwent RYGBP and 140 underwent LAGB.

RESULTS

After 3 years, 68.7% of RYGBP and 30.2% of LAGB participants were in diabetes remission. Baseline factors associated with diabetes remission included a lower weight for LAGB, greater fasting C-peptide, lower leptin-to-fat mass ratio for RYGBP, and a lower hemoglobin A1c without need for insulin for both procedures. After both procedures, greater postsurgical weight loss was associated with remission. However, even after controlling for differences in amount of weight lost, relative diabetes remission rates remained nearly twofold higher after RYGBP than LAGB.

CONCLUSIONS

Diabetes remission up to 3 years after RYGBP and LAGB was proportionally higher with increasing postsurgical weight loss. However, the nearly twofold greater weight loss-adjusted likelihood of diabetes remission in subjects undergoing RYGBP than LAGB suggests unique mechanisms contributing to improved glucose metabolism beyond weight loss after RYGBP.

Challenges and Solutions for Depression Prevention Research: Methodology for a Depression Prevention Trial for Older Adults with Knee Arthritis and Emotional Distress

OBJECTIVE

To describe the methodology, challenges, and baseline characteristics of a prevention development trial entitled "Reducing Pain, Preventing Depression."

METHODS

A sequential multiple assignment randomized trial (SMART) compared sequences of cognitive behavioral therapy (CBT) and physical therapy for knee pain and prevention of depression and anxiety. Participants were randomized to CBT, physical therapy, or enhanced usual care and followed for 12 months for new-episode depression or anxiety. Participants were age ≥ 60 with knee osteoarthritis and subsyndromal depression, defined as 9-item Patient Health Questionnaire (PHQ-9) score ≥1 (which included the endorsement of one of the cardinal symptoms of depression [low mood or anhedonia]) and no diagnosis of major depressive disorder per the Structured Clinical Interview for DSM-IV. Depression and anxiety severity and characterization of new episodes were assessed with the PHQ-9, Generalized Anxiety Disorder-7, and the PRIME-MD. Knee pain was characterized with the Western Ontario McMaster Arthritis Index. Response was defined as at least "Very Much Better" on a Patient Global Impression of Change.

RESULTS

At baseline there were 99 patients with an average age of 71; 61.62% were women and 81.8% white. The average PHQ-9 was 5.6 and average Generalized Anxiety Disorder-7, was 3.2. Most were satisfied with the interventions and study procedures. The challenges and solutions described here will also be used in a confirmatory clinical trial of efficacy.

CONCLUSION

A SMART design for depression and anxiety prevention, using both CBT and physical therapy, appears to be feasible and acceptable to participants. The methodologic innovations of this project may advance the field of late-life depression and anxiety prevention.

A determinant-based criterion for working correlation structure selection in generalized estimating equations

In generalized estimating equations (GEE), the correlation between the repeated observations on a subject is specified with a working correlation matrix. Correct specification of the working correlation structure ensures efficient estimators of the regression coefficients. Among the criteria used, in practice, for selecting working correlation structure, Rotnitzky-Jewell, Quasi Information Criterion (QIC) and Correlation Information Criterion (CIC) are based on the fact that if the assumed working correlation structure is correct then the model-based (naive) and the sandwich (robust) covariance estimators of the regression coefficient estimators should be close to each other. The sandwich covariance estimator, used in defining the Rotnitzky-Jewell, QIC and CIC criteria, is biased downward and has a larger variability than the corresponding model-based covariance estimator. Motivated by this fact, a new criterion is proposed in this paper based on the bias-corrected sandwich covariance estimator for selecting an appropriate working correlation structure in GEE. A comparison of the proposed and the competing criteria is shown using simulation studies with correlated binary responses. The results revealed that the proposed criterion generally performs better than the competing criteria. An example of selecting the appropriate working correlation structure has also been shown using the data from Madras Schizophrenia Study.

Teaching Missing Data Methodology to Undergraduates Using a Group-Based Project Within a Six-Week Summer Program

Missing data mechanisms, methods of handling missing data, and the potential impact of missing data on study results are usually not taught until graduate school. However, the appropriate handling of missing data is fundamental to biomedical research and should be introduced earlier on in a student's education. The Summer Institute for Training in Biostatistics (SIBS) provides practical experience to motivate trainees to pursue graduate training and biomedical research. Since 2010, SIBS Pittsburgh has demonstrated the feasibility of introducing missing data concepts to trainees in a small-group project-based setting that involves both simulation and data analysis. After learning about missing data mechanisms and statistical techniques, trainees apply what they have learned to a NIDDK/NIH-funded Hepatitis C treatment study, to examine how various hypothesized missing data patterns can affect results. A simulation is also used to examine the bias and precision of these methods under each missing data pattern. Our experience shows that under such project-based training, advanced topics, such as missing data, can be presented to trainees with limited statistical preparation, and ultimately, can further their statistical literacy and reasoning. The tools presented here are provided in the Appendix.