Discovering structure in multiple outcomes models for tests of childhood neurodevelopment

Bayesian model-based clustering provides a powerful and flexible tool that can be incorporated into regression models to better understand the grouping of observations. Using data from the Seychelles Child Development Study, we explore the effect of prenatal methylmercury exposure on 20 neurodevelopmental outcomes measured in 9-year-old children. Rather than cluster individual subjects, we cluster the outcomes within a multiple outcomes model. By using information in the data to nest the outcomes into groups called domains, the model more accurately reflects the shared characteristics of neurodevelopmental domains and improves estimation of the overall and outcome-specific exposure effects by shrinking effects within and between domains selected by the data. The Bayesian paradigm allows for sampling from the posterior distribution of the grouping parameters; thus, inference can be made about group membership and their defining characteristics. We avoid the often difficult and highly subjective requirement of a priori identification of the total number of groups by incorporating a Dirichlet process prior to form a fully Bayesian multiple outcomes model.

Estimation of group means using Bayesian generalized linear mixed models

Generalized linear mixed models (GLMM) are commonly used to model the treatment effect over time while controlling for important clinical covariates. Standard software procedures often provide estimates of the outcome based on the mean of the covariates; however, these estimates will be biased for the true group means in the GLMM. Implementing GLMM in the frequentist framework can lead to issues of convergence. A simulation study demonstrating the use of fully Bayesian GLMM for providing unbiased estimates of group means is shown. These models are very straightforward to implement and can be used for a broad variety of outcomes (eg, binary, categorical, and count data) that arise in clinical trials. We demonstrate the proposed method on a data set from a clinical trial in diabetes.

Creation and Piloting of a Model for Simulating a Minimally Invasive Myomectomy

Introduction: In the era of mandatory work hour restrictions for residency programs, the opportunity for mastery of complex surgical skills in the operating room (OR) has been compromised. All the while, gynecologic surgical techniques have continued to expand. Surgical simulation offers an adjuvant modality for helping young surgeons hone their surgical techniques. We sought to design, construct, and pilot a model for simulating a minimally invasive myomectomy procedure for the purpose of resident training. We undertook a preliminary evaluation of the model’s validity.

Methods: Gynecologic surgical simulation models were constructed from polyvinyl alcohol poured into 3D-printed injection molds. A total of 12 laparoscopic and 12 robot-assisted simulated myomectomies were performed using the models. Face and content validity were evaluated with post-simulation questionnaires. Construct validity was assessed by comparing procedural metrics (time to completion and estimated blood loss) between residents and attending surgeons.

Results: In the post-simulation survey, the majority of attending surgeons agreed the model was realistic (83.3%) and included the critical steps of a myomectomy (87.5%). Most residents agreed they would feel more prepared for a myomectomy if they practiced on the model beforehand (87.5%) and the majority of attending surgeons agreed they would feel comfortable giving a resident more operative autonomy if the resident had previously completed the simulation (71.4%). Procedural metrics were not significantly associated with expertise level.

Conclusion: We were able to successfully create a model for simulating a minimally invasive myomectomy. Initial simulations using the model were well received by participants. Further development and investigation of the model will be pursued to determine if this is a valid and useful tool for teaching and practicing a minimally invasive myomectomy.

Placental concentrations of essential, toxic, and understudied metals and relationships with birth outcomes in Chattanooga, TN

BACKGROUND

Comprehensive examinations of placental metal concentrations and correlations with infant parameters are under-investigated. Chattanooga, Tennessee's consistently high incidence of low birth weight and potential for metal exposure provides an ideal opportunity to investigate potential correlations.

OBJECTIVES

To investigate the associations between a wide variety of metals in placental tissue and multiple infant parameters.

METHODS

A total of 31 elements were screened via ICP-MS in 374 individual placental samples. Of those, 14 were quantifiable in > 86% of the samples. We examined correlations between metal concentrations and infant parameters (birth weight, gestational age, birth weight centile, placental weight, birth length and head circumference). We fit multivariable regression models to estimate the covariate-adjusted associations of birth weight with ln-transformed concentrations of each of the 14 metals and used generalized additive models to examine nonlinear relationships.

RESULTS

Some of the strongest relationships with infant parameters came from several lesser-studied metals. Placental rhodium concentrations were negatively correlated with almost all infant parameters. In the fully adjusted regression model, birth weight was significantly associated with several metals. On an IQR (25th to the 75th percentile) basis, estimated changes in birthweight were: for cobalt (82.5 g, IQR=6.05 µg/kg, p = 0.006), iron (-51.5 g, IQR = 171800 µg/kg, p = 0.030), manganese (-27.2 g, IQR=152.1 µg/kg, p = 0.017), lead (-72.7 g, IQR=16.55 µg/kg, p = 0.004) and rhodium (-1365.5 g, IQR = 0.33 µg/kg, p < 0.001). Finally, a generalized additive model showed significant nonlinear relationships between birth weight and concentrations of Co and Rh.

CONCLUSIONS

Our comprehensive examination of placental metals illustrate many strong associations between lesser-studied metals and infant parameters. These data, in combination with our correlations of well-studied metals, illustrate a need to consider in utero exposure to a broad array of metals when considering fetal development.

Obesity candidate genes, gestational weight gain, and body weight changes in pregnant women

OBJECTIVE

To examine the associations of two obesity-associated genes, FTO (rs9939609) and GNB3 (rs5443) single nucleotide polymorphisms (SNPs), with early pregnancy body mass index, gestational weight gain, and postpartum weight retention.

METHODS

Secondary data analysis of self-identified white (n = 580) and black (n = 194) women who participated in a randomized controlled trial (2009-2014) and provided a saliva sample of DNA. Bivariate relationships were assessed using analysis of variance. Multiple regression models assessed the relationship between outcomes and gene SNPs, controlling for income, parity, and smoking status.

RESULTS

FTO and GNB3 gene associations with pregnancy weight were different by racial group and early pregnancy body mass index. Obese black women homozygote for the FTO risk allele (AA) had a higher gestational weight gain compared with non-risk homozygotes (TT) (P = 0.006). GNB3 non-risk CC homozygotes tended to have a lower gestational weight gain compared with heterozygotes (P = 0.05). White GNB3 C carriers tended to be heavier in early pregnancy (P <0.1) and GNB3 homozygote (TT) overweight women tended to have lower postpartum weight retention than C carriers.

CONCLUSIONS

The FTO gene and possibly the GNB3 gene are associated with high gestational weight gain in obese black women. Obese carriers of the FTO risk allele gained 4.1 kg (AT) and 7.6 kg (TT) more than those without risk alleles. Overweight GNB3 heterozygotes (CT) gained 6.6 kg less than homozygotes (CC). Overweight or obese black women who have either risk variant are at risk for high gestational weight gain.

Bile salt receptor TGR5 is highly expressed in esophageal adenocarcinoma and precancerous lesions with significantly worse overall survival and gender differences

Bile acid reflux in the esophagus plays an important role in the carcinogenesis of esophageal adenocarcinoma (EAC). The G-protein coupled bile acid receptor (TGR5) has been associated with the development of gastrointestinal cancer. However, little is known regarding the role of TGR5 in esophageal carcinoma and precancerous lesions. We analyzed genomic DNA from 116 EACs for copy number aberrations via Affymetrix SNP6.0 microarrays. The TGR5 gene locus was amplified in 12.7% (14/116) of the EACs. The TGR5 protein expression was also assessed using immunohistochemistry from tissue microarrays, including Barrett’s esophagus (BE), low-(LGD) and high-grade dysplasia (HGD), columnar cell metaplasia (CM), squamous epithelium (SE), EAC and squamous cell carcinoma. The TGR5 protein was highly expressed in 71% of EAC (75/106), 100% of HGD (11/11), 72% of LGD (13/18), 66% of BE (23/35), 84% of CM (52/62), and 36% of SE (30/83). The patients with high expression of TGR5 exhibited significantly worse overall survival compared to the patients with nonhigh expression. TGR5 high expression was significantly increased in the males compared to the females in all cases with an odds ratio of 1.9 times. The vitamin D receptor (VDR) was significantly correlated with TGR5 expression. Our findings indicated that TGR5 may play an important role in the development and prognosis of EAC through a bile acid ligand. Gender differences in TGR5 and VDR expression may explain why males have a higher incidence of EAC compared to females.

MCM4 and MCM7, potential novel proliferation markers, significantly correlated with Ki-67, Bmi1, and cyclin E expression in esophageal adenocarcinoma, squamous cell carcinoma, and precancerous lesions

Minichromosomal maintenance (MCM) proteins are participants of DNA replication and may represent more accurate markers in determining the proliferative fraction within a tumor than proliferative marker Ki-67. Our study investigated the correlation between MCM4 and MCM7 expression and Ki-67, Bmi1, and cyclin E expression in esophageal adenocarcinoma, squamous cell carcinoma, and precancerous lesions. MCM4 and MCM7 expression had similar distribution as Ki-67 and Bmi1 expression in esophageal carcinoma and pre-cancerous lesions. The mean percentage of MCM4, MCM7, and Ki-67 expression increased from squamous epithelium (5.5%, 7.3%, and 5.9%, respectively), to columnar cell metaplasia (11.2, 13.5%, and 3.4%), Barrett's esophagus (27.7%, 35.3%, and 8.3%), low-grade dysplasia (42.6%, 52.2%, and 12.9%), high-grade dysplasia (63.2%, 77.7%, and 29.6%), adenocarcinoma (61.3%, 75.5%, and 24.5%), and squamous cell carcinoma (74.1, 85.4%, and 36.3%). The percentages of MCM4 and MCM7 expression were significantly higher than Ki-67 expression. Using univariate analysis we found a high percentage of MCM4 expression (>70%) to be significantly associated with lymph node metastasis and shorter survival in the adenocarcinoma group. We also demonstrated the percentage of MCM4 and MCM7 expression to be significantly correlated with Ki-67, Bmi1, and cyclin E expression in esophageal carcinoma and precancerous lesions. MCM4 and MCM7 may serve as more sensitive proliferative markers for the evaluation of esophageal lesions.