Mixture modeling for joint analysis of survival, discrete, and continuous data

Mixture modeling is commonly used to model categorical latent variables that represent subpopulations in which population membership is unknown but can be inferred from the data. In relatively recent years, the potential of finite mixture models has been applied in time-to-event data. However, the commonly used survival mixture model assumes that the effects of the covariates involved in failure times differ across latent classes, but the covariate distribution is homogeneous. The aim of this dissertation is to develop a method to examine time-to-event data in the presence of unobserved heterogeneity under a framework of mixture modeling. A joint model is developed to incorporate the latent survival trajectory along with the observed information for the joint analysis of a time-to-event variable, its discrete and continuous covariates, and a latent class variable. It is assumed that the effects of covariates on survival times and the distribution of covariates vary across different latent classes. The unobservable survival trajectories are identified through estimating the probability that a subject belongs to a particular class based on observed information. We applied this method to a Hodgkin lymphoma study with long-term follow-up and observed four distinct latent classes in terms of long-term survival and distributions of prognostic factors. Our results from simulation studies and from the Hodgkin lymphoma study demonstrated the superiority of our joint model compared with the conventional survival model. This flexible inference method provides more accurate estimation and accommodates unobservable heterogeneity among individuals while taking involved interactions between covariates into consideration.^

Longitudinal categorical data analysis: A continuous -time Markov chain approach

In this dissertation, we propose a continuous-time Markov chain model to examine the longitudinal data that have three categories in the outcome variable. The advantage of this model is that it permits a different number of measurements for each subject and the duration between two consecutive time points of measurements can be irregular. Using the maximum likelihood principle, we can estimate the transition probability between two time points. By using the information provided by the independent variables, this model can also estimate the transition probability for each subject. The Monte Carlo simulation method will be used to investigate the goodness of model fitting compared with that obtained from other models. A public health example will be used to demonstrate the application of this method. ^

GENETIC ANALYSIS OF THE FUNCTION OF THE DROSOPHILA DOUBLESEX-RELATED FACTOR dmrt93B

DMRT (Doublesex and Mab-3 related transcription factor) proteins generally associated with sexual differentiation in many organisms share a common DNA binding domain and are often expressed in reproductive tissues. Aside from doublesex, which is a central factor in the regulation of sex determination, Drosophila possesses three different dmrt genes that are of unknown function. Because the association with sexual differentiation and reproduction is not universal and some DMRT proteins have been found to play other developmental roles we chose to further characterize one of these Drosophila genes. We carried out genetic analysis of dmrt93B, which was previously found to be expressed sex-specifically in the developing somatic gonad and to affect testis morphogenesis in RNAi knockdowns. In order to disrupt this gene, the GAL4 yeast transcriptional activator followed by a polyadenylation signal was inserted after the dmrt93B start codon and introduced into the genome by homologous recombination. Analysis of the knock-in mutation as well as a small deletion removing all dmrt93B sequence demonstrate that loss of function causes partial lethality at the late pupal stage. Surprisingly, these mutations have no significant effect on gonad formation or male fertility. Analysis of GAL4-driven GFP reporter expression indicates that the dmrt93B promoter activity is highly specific to neurons in the suboesophageal and proventricular ganglion in larva and adult of both sexes suggesting a possible role in digestive tract function. Using the Capillary Feeder (CAFÉ) assay to measure daily food intake we find that reduction in this gene’s function leads to an increase in food consumption. These results suggest dmrt93 plays an important role in the formation or maintenance of neurons that affect feeding and support the idea that dmrt genes may not be restricted to roles in sexual differentiation.

Detailed analysis of sequence changes occurring during vlsE antigenic variation in the mouse model of Borrelia burgdorferi infection.

Lyme disease Borrelia can infect humans and animals for months to years, despite the presence of an active host immune response. The vls antigenic variation system, which expresses the surface-exposed lipoprotein VlsE, plays a major role in B. burgdorferi immune evasion. Gene conversion between vls silent cassettes and the vlsE expression site occurs at high frequency during mammalian infection, resulting in sequence variation in the VlsE product. In this study, we examined vlsE sequence variation in B. burgdorferi B31 during mouse infection by analyzing 1,399 clones isolated from bladder, heart, joint, ear, and skin tissues of mice infected for 4 to 365 days. The median number of codon changes increased progressively in C3H/HeN mice from 4 to 28 days post infection, and no clones retained the parental vlsE sequence at 28 days. In contrast, the decrease in the number of clones with the parental vlsE sequence and the increase in the number of sequence changes occurred more gradually in severe combined immunodeficiency (SCID) mice. Clones containing a stop codon were isolated, indicating that continuous expression of full-length VlsE is not required for survival in vivo; also, these clones continued to undergo vlsE recombination. Analysis of clones with apparent single recombination events indicated that recombinations into vlsE are nonselective with regard to the silent cassette utilized, as well as the length and location of the recombination event. Sequence changes as small as one base pair were common. Fifteen percent of recovered vlsE variants contained "template-independent" sequence changes, which clustered in the variable regions of vlsE. We hypothesize that the increased frequency and complexity of vlsE sequence changes observed in clones recovered from immunocompetent mice (as compared with SCID mice) is due to rapid clearance of relatively invariant clones by variable region-specific anti-VlsE antibody responses.

Detailed analysis of sequence changes occurring during vlsE antigenic variation in the mouse model of Borrelia burgdorferi infection.

Lyme disease Borrelia can infect humans and animals for months to years, despite the presence of an active host immune response. The vls antigenic variation system, which expresses the surface-exposed lipoprotein VlsE, plays a major role in B. burgdorferi immune evasion. Gene conversion between vls silent cassettes and the vlsE expression site occurs at high frequency during mammalian infection, resulting in sequence variation in the VlsE product. In this study, we examined vlsE sequence variation in B. burgdorferi B31 during mouse infection by analyzing 1,399 clones isolated from bladder, heart, joint, ear, and skin tissues of mice infected for 4 to 365 days. The median number of codon changes increased progressively in C3H/HeN mice from 4 to 28 days post infection, and no clones retained the parental vlsE sequence at 28 days. In contrast, the decrease in the number of clones with the parental vlsE sequence and the increase in the number of sequence changes occurred more gradually in severe combined immunodeficiency (SCID) mice. Clones containing a stop codon were isolated, indicating that continuous expression of full-length VlsE is not required for survival in vivo; also, these clones continued to undergo vlsE recombination. Analysis of clones with apparent single recombination events indicated that recombinations into vlsE are nonselective with regard to the silent cassette utilized, as well as the length and location of the recombination event. Sequence changes as small as one base pair were common. Fifteen percent of recovered vlsE variants contained "template-independent" sequence changes, which clustered in the variable regions of vlsE. We hypothesize that the increased frequency and complexity of vlsE sequence changes observed in clones recovered from immunocompetent mice (as compared with SCID mice) is due to rapid clearance of relatively invariant clones by variable region-specific anti-VlsE antibody responses.

STRUCTURAL ANALYSIS OF THE TRANSCRIPTION PRODUCTS OF A TEMPERATURE-SENSITIVE PHENOTYPIC MUTANT OF MOLONEY SARCOMA VIRUS--MUSV TS110

Cells infected with a temperature sensitive phenotypic mutant of Moloney sarcoma virus (MuSVts110) exhibit a transformed phenotype at 33('(DEGREES)) and synthesize two virus specific proteins, p85('gag-mos), a gag-mos fusion protein and p58('gag), a truncated gag precursor protein (the gag gene codes for viral structural proteins and mos is the MuSV transforming gene). At 39('(DEGREES)) only p58('gag) is synthesized and the morphology of the cells is similar to uninfected NRK parental cells. Two MuSVts110 specific RNAs are made in MuSVts110-infected cells, one of 4.0 kb in length, the other of 3.5 kb. Previous work indicated that each of these RNAs arose by a single central deletion of parental MuSV genetic material, and that p58('gag) was made by the 4.0 kb RNA and p85('gag-mos) from the 3.5 kb RNA. The objective of my dissertation research was to map precisely the deletion boundaries of both of the MuSVts110 RNAs, and to determine the proper reading frame across both deletion borders. This work succeeded in arriving at the following conclusions: (a) Using S-1 nuclease analysis and primer extension sequencing, it was found that the 4.0 kb MuSVts110 RNA arose by a 1488 base deletion of 5.2 kb parental MuSV genomic RNA. This deletion resulted in an out of frame fusion of the gag and mos genes that resulted in the formation of a "stop" codon which causes termination of translation just beyond the c-terminus of the gag region. Thus, this RNA can only be translated into the truncated gag protein p58('gag). (b) S-1 analysis of RNA from cells cultivated at different temperatures demonstrated that the 4.0 kb RNA was synthesized at all temperatures but that synthesis of the 3.5 kb RNA was temperature sensitive. These observations supported the data derived from blot hybridization experiments the interpretation of which argued for the existence of a single provirus in MuSVts110 infected cells, and hence only a single primary transcript (the 4.0 kb RNA). (c) Analyses similar to those described in (a) above showed that the 3.5 kb RNA was derived from the 4.0 kb MuSVts110 RNA by a further deletion of 431 bases, fusing the gag and mos genes into a continuous reading frame capable of directing synthesis of the p85('gag-mos) protein. These sequence data and the presence of only one MuSVts110-specific provirus, indicate that a splice mechanism is employed to generate the 3.5 kb RNA since the gag and mos genes are observed to be fused in frame in this RNA. . . . (Author's abstract exceeds stipulated maximum length. Discontinued here with permission of author.) UMI ^

Multiple imputation for missing continuous and dichotomous data: The effects of patterns of missingness and variable correlations on type I error with small samples

The purpose of this study is to investigate the effects of predictor variable correlations and patterns of missingness with dichotomous and/or continuous data in small samples when missing data is multiply imputed. Missing data of predictor variables is multiply imputed under three different multivariate models: the multivariate normal model for continuous data, the multinomial model for dichotomous data and the general location model for mixed dichotomous and continuous data. Subsequent to the multiple imputation process, Type I error rates of the regression coefficients obtained with logistic regression analysis are estimated under various conditions of correlation structure, sample size, type of data and patterns of missing data. The distributional properties of average mean, variance and correlations among the predictor variables are assessed after the multiple imputation process. ^ For continuous predictor data under the multivariate normal model, Type I error rates are generally within the nominal values with samples of size n = 100. Smaller samples of size n = 50 resulted in more conservative estimates (i.e., lower than the nominal value). Correlation and variance estimates of the original data are retained after multiple imputation with less than 50% missing continuous predictor data. For dichotomous predictor data under the multinomial model, Type I error rates are generally conservative, which in part is due to the sparseness of the data. The correlation structure for the predictor variables is not well retained on multiply-imputed data from small samples with more than 50% missing data with this model. For mixed continuous and dichotomous predictor data, the results are similar to those found under the multivariate normal model for continuous data and under the multinomial model for dichotomous data. With all data types, a fully-observed variable included with variables subject to missingness in the multiple imputation process and subsequent statistical analysis provided liberal (larger than nominal values) Type I error rates under a specific pattern of missing data. It is suggested that future studies focus on the effects of multiple imputation in multivariate settings with more realistic data characteristics and a variety of multivariate analyses, assessing both Type I error and power. ^

The socioeconomic status of families of hyperglycemic versus non-hyperglycemic children: A descriptive analysis

The proportion of children and adolescents living with type 2 diabetes mellitus (DM) is rising at an alarming rate. Studies have shown that poor dietary choices and sedentary behaviors account for progression of some of the most prevalent diseases in America, including obesity, heart disease and diabetes. Other studies have shown that genetics plays a role in the diabetic determination of an individual, although not very common. What are some of the differentiating factors between elevated and non-elevated fasting capillary glucose (FCG) levels in children of similar ages, knowing they spend a majority of their lives at home or at school? Why are some children acquiring diabetes while others are not? This study utilized an IRB-approved Family Demographic Survey to determine gender, family income, parent education levels, sedentary practices, and household size. Only those families who gave consent to take part in the study received a questionnaire. The statistical results were used to test the hypothesis that children living with elevated FCG levels are more likely to descend from families with lower incomes, and lower levels of education.^ With regard to household income and FCG status of non-hyperglycemic and hyperglycemic children (Table 4b), there are 10.4% more hyperglycemic children in the lower income bracket than non-hyperglycemic children in the same income bracket.^ With regard to maternal education and FCG status (Table 5b), there are 7.0% more hyperglycemic children in the high school or less maternal educational attainment level than non-hyperglycemic children in the same maternal educational level. The Pearson correlation of maternal education and FCG status showed a negative correlation value of -.035 (Table 5d). The higher the occurrence of hyperglycemia in a child, the lower the maternal educational status is. Household size ranges and averages are nearly identical in families of both hyperglycemic and non-hyperglycemic children. ^

Effects of early and late lesion of orbital frontal cortex on visual processing of faces in rhesus macaques (Macaca mulatta)

Many mental disorders disrupt social skills, yet few studies have examined how the brain processes social information. Functional neuroimaging, neuroconnectivity and electrophysiological studies suggest that orbital frontal cortex plays important roles in social cognition, including the analysis of information from faces, which are important cues in social interactions. Studies in humans and non-human primates show that damage to orbital frontal cortex produces social behavior impairments, including abnormal aggression, but these studies have failed to determine whether damage to this area impairs face processing. In addition, it is not known whether damage early in life is more detrimental than damage in adulthood. This study examined whether orbital frontal cortex is necessary for the discrimination of face identity and facial expressions, and for appropriate behavioral responses to aggressive (threatening) facial expressions. Rhesus monkeys (Macaca mulatta) received selective lesions of orbital frontal cortex as newborns or adults. As adults, these animals were compared with sham-operated controls on their ability to discriminate between faces of individual monkeys and between different facial expressions of emotion. A passive visual paired-comparison task with standardized rhesus monkey face stimuli was designed and used to assess discrimination. In addition, looking behavior toward aggressive expressions was assessed and compared with that of normal control animals. The results showed that lesion of orbital frontal cortex (1) may impair discrimination between faces of individual monkeys, (2) does not impair facial expression discrimination, and (3) changes the amount of time spent looking at aggressive (threatening) facial expressions depending on the context. The effects of early and late lesions did not differ. Thus, orbital frontal cortex appears to be part of the neural circuitry for recognizing individuals and for modulating the response to aggression in faces, and the plasticity of the immature brain does not allow for recovery of these functions when the damage occurs early in life. This study opens new avenues for the assessment of rhesus monkey face processing and the neural basis of social cognition, and allows a better understanding of the nature of the neuropathology in patients with mental disorders that disrupt social behavior, such as autism. ^

A new combined endpoint using U-statistic in analysis in clinical trials

Common endpoints can be divided into two categories. One is dichotomous endpoints which take only fixed values (most of the time two values). The other is continuous endpoints which can be any real number between two specified values. Choices of primary endpoints are critical in clinical trials. If we only use dichotomous endpoints, the power could be underestimated. If only continuous endpoints are chosen, we may not obtain expected sample size due to occurrence of some significant clinical events. Combined endpoints are used in clinical trials to give additional power. However, current combined endpoints or composite endpoints in cardiovascular disease clinical trials or most clinical trials are endpoints that combine either dichotomous endpoints (total mortality + total hospitalization), or continuous endpoints (risk score). Our present work applied U-statistic to combine one dichotomous endpoint and one continuous endpoint, which has three different assessments and to calculate the sample size and test the hypothesis to see if there is any treatment effect. It is especially useful when some patients cannot provide the most precise measurement due to medical contraindication or some personal reasons. Results show that this method has greater power then the analysis using continuous endpoints alone. ^

A novel method for evaluating an interaction effect of correlated continuous covariates in a linear model

Interaction effect is an important scientific interest for many areas of research. Common approach for investigating the interaction effect of two continuous covariates on a response variable is through a cross-product term in multiple linear regression. In epidemiological studies, the two-way analysis of variance (ANOVA) type of method has also been utilized to examine the interaction effect by replacing the continuous covariates with their discretized levels. However, the implications of model assumptions of either approach have not been examined and the statistical validation has only focused on the general method, not specifically for the interaction effect.^ In this dissertation, we investigated the validity of both approaches based on the mathematical assumptions for non-skewed data. We showed that linear regression may not be an appropriate model when the interaction effect exists because it implies a highly skewed distribution for the response variable. We also showed that the normality and constant variance assumptions required by ANOVA are not satisfied in the model where the continuous covariates are replaced with their discretized levels. Therefore, naïve application of ANOVA method may lead to an incorrect conclusion. ^ Given the problems identified above, we proposed a novel method modifying from the traditional ANOVA approach to rigorously evaluate the interaction effect. The analytical expression of the interaction effect was derived based on the conditional distribution of the response variable given the discretized continuous covariates. A testing procedure that combines the p-values from each level of the discretized covariates was developed to test the overall significance of the interaction effect. According to the simulation study, the proposed method is more powerful then the least squares regression and the ANOVA method in detecting the interaction effect when data comes from a trivariate normal distribution. The proposed method was applied to a dataset from the National Institute of Neurological Disorders and Stroke (NINDS) tissue plasminogen activator (t-PA) stroke trial, and baseline age-by-weight interaction effect was found significant in predicting the change from baseline in NIHSS at Month-3 among patients received t-PA therapy.^

Are anatomic brain locations important in determining the clinical outcomes in acute ischemic stroke? --- An exploratory analysis of the NINDS Stroke Trial data

Background and purpose. Brain lesions in acute ischemic stroke measured by imaging tools provide important clinical information for diagnosis and final infarct volume has been considered as a potential surrogate marker for clinical outcomes. Strong correlations have been found between lesion volume and clinical outcomes in the NINDS t-PA Stroke Trial but little has been published about lesion location and clinical outcomes. Studies of the National Institute of Neurological Disorders and Stroke (NINDS) t-PA Stroke Trial data found the direction of the t-PA treatment effect on a decrease in CT lesion volume was consistent with the observed clinical effects at 3 months, but measure of t-PA treatment benefits using CT lesion volumes showed a diminished statistical significance, as compared to using clinical scales. ^ Methods. We used the global test to evaluate the hypothesis that lesion locations were strongly associated with clinical outcomes within each treatment group at 3 months after stroke. The anatomic locations of CT scans were used for analysis. We also assessed the effect of t-PA on lesion location using a global statistical test. ^ Results. In the t-PA group, patients with frontal lesions had larger infarct volumes and worse NIHSS score at 3 months after stroke. The clinical status of patients with frontal lesions in t-PA group was less likely to be affected by lesion volume, as compared to those who had no frontal lesions in at 3 months. For patients within the placebo group, both brain stem and internal capsule locations were significantly associated with a lower odd of having favorable outcomes at 3 months. Using a global test we could not detect a significant effect of t-PA treatment on lesion location although differences between two treatment groups in the proportion of lesion findings in each location were found. ^ Conclusions. Frontal, brain stem, and internal capsule locations were significantly related to clinical status at 3 months after stroke onset. We detect no significant t-PA effect on all 9 locations although proportion of lesion findings in differed among locations between the two treatment groups.^

Pediatric asthma incidence in Texas and associations with ambient ozone levels in an urban area an analysis using Medicaid claims data

Few recent estimates of childhood asthma incidence exist in the literature, although the importance of incidence surveillance for understanding asthma risk factors has been recognized. Asthma prevalence, morbidity and mortality reports have repeatedly shown that low-income children are disproportionately impacted by the disease. The aim of this study was to demonstrate the utility of Medicaid claims data for providing statewide estimates of asthma incidence. Medicaid Analytic Extract (MAX) data for Texas children ages 0-17 enrolled in Medicaid between 2004 and 2007 were used to estimate incidence overall and by age group, gender, race and county of residence. A 13+ month period of continuous enrollment was required in order to distinguish incident from prevalent cases identified in the claims data. Age-adjusted incidence of asthma was 4.26/100 person-years during 2005-2007, higher than reported in other populations. Incidence rates decreased with age, were higher for males than females, differed by race, and tended to be higher in rural than urban areas. With this study, we were able to demonstrate the utility of MAX data for estimating asthma incidence, and create a dataset of incident cases to use in further analysis. ^ In subsequent analyses, we investigated a possible association between ambient air pollutants and incident asthma among Medicaid-enrolled children in Harris County Texas between 2005 and 2007. This population is at high risk for asthma, and living in an area with historically poor air quality. We used a time-stratified case-crossover design and conditional logistic regression to calculate odds ratios, adjusted for weather variables and aeroallergens, to assess the effect of increases in ozone, NO2 and PM2.5 concentrations on risk of developing asthma. Our results show that a 10 ppb increase in ozone was significantly associated with asthma during the warm season (May-October), with the strongest effect seen when a 6-day cumulative lag period was used to compute the exposure metric (OR=1.05, 95% CI, 1.02–1.08). Similar results were seen for NO2 and PM 2.5 (OR=1.07, 95% CI, 1.03–1.11 and OR=1.12, 95% CI, 1.03–1.22, respectively). PM2.5 also had significant effects in the cold season (November-April), 5-day cumulative lag: OR=1.11, 95% CI, 1.00–1.22. When compared with children in the lowest quartile of O3 exposure, the risk for children in the highest quartile was 20% higher. This study indicates that these pollutants are associated with newly-diagnosed childhood asthma in this low-income urban population, particularly during the summer months. ^

Age -period -cohort analysis: An extension of Cox's lifetable regression with time -dependent covariates

It is well known that an identification problem exists in the analysis of age-period-cohort data because of the relationship among the three factors (date of birth + age at death = date of death). There are numerous suggestions about how to analyze the data. No one solution has been satisfactory. The purpose of this study is to provide another analytic method by extending the Cox's lifetable regression model with time-dependent covariates. The new approach contains the following features: (1) It is based on the conditional maximum likelihood procedure using a proportional hazard function described by Cox (1972), treating the age factor as the underlying hazard to estimate the parameters for the cohort and period factors. (2) The model is flexible so that both the cohort and period factors can be treated as dummy or continuous variables, and the parameter estimations can be obtained for numerous combinations of variables as in a regression analysis. (3) The model is applicable even when the time period is unequally spaced.^ Two specific models are considered to illustrate the new approach and applied to the U.S. prostate cancer data. We find that there are significant differences between all cohorts and there is a significant period effect for both whites and nonwhites. The underlying hazard increases exponentially with age indicating that old people have much higher risk than young people. A log transformation of relative risk shows that the prostate cancer risk declined in recent cohorts for both models. However, prostate cancer risk declined 5 cohorts (25 years) earlier for whites than for nonwhites under the period factor model (0 0 0 1 1 1 1). These latter results are similar to the previous study by Holford (1983).^ The new approach offers a general method to analyze the age-period-cohort data without using any arbitrary constraint in the model. ^

Development of combined micro -preparation, separation, and mass spectrometry methods and application in the microdialysis study of neuropeptide metabolism

Two sets of mass spectrometry-based methods were developed specifically for the in vivo study of extracellular neuropeptide biochemistry. First, an integrated micro-concentration/desalting/matrix-addition device was constructed for matrix-assisted laser desorption ionization mass spectrometry (MALDI MS) to achieve attomole sensitivity for microdialysis samples. Second, capillary electrophoresis (CE) was incorporated into the above micro-liquid chromatography (LC) and MALDI MS system to provide two-dimensional separation and identification (i.e. electrophoretic mobility and molecular mass) for the analysis of complex mixtures. The latter technique includes two parts of instrumentation: (1) the coupling of a preconcentration LC column to the inlet of a CE capillary, and (2) the utilization of a matrix-precoated membrane target for continuous CE effluent deposition and for automatic MALDI MS analysis (imaging) of the CE track.^ Initial in vivo data reveals a carboxypeptidase A (CPA) activity in rat brain involved in extracellular neurotensin metabolism. Benzylsuccinic acid, a CPA inhibitor, inhibited neurotensin metabolite NT1-12 formation by 70%, while inhibitors of other major extracellular peptide metabolizing enzymes increased NT1-12 formation. CPA activity has not been observed in previous in vitro experiments. Next, the validity of the methodology was demonstrated in the detection and structural elucidation of an endogenous neuropeptide, (L)VV-hemorphin-7, in rat brain upon ATP stimulation. Finally, the combined micro-LC/CE/MALDI MS was used in the in vivo metabolic study of peptide E, a mu-selective opioid peptide with 25 amino acid residues. Profiles of 88 metabolites were obtained, their identity being determined by their mass-to-charge ratio and electrophoretic mobility. The results indicate that there are several primary cleavage sites in vivo for peptide E in the release of its enkephalin-containing fragments. ^