2007 Poster Competition Winners
Title: Analysis of Smoking Cessation Patterns Using a Stochastic Mixed Effects Model with a Latent Cured State
Background: Smoking is the leading preventable cause of death in the U.S. In the U.S. alone, 44.5 million adults, or 20.9% of the adult population were smokers in 2004; 440,000 annual premature deaths are attributable to smoking. Smoking is a major cause of a large number of diseases, e.g. cancers of the lung, larynx, and pharynx, etc. The slow reduction of adult smoking prevalence is partly due to high rates of relapse following quit attempts among smokers. A major problem when studying addiction behavior is that participants typically make several quit attempts before they successfully quit. Thus, for efficient development, targeting and evaluation of interventions, it is necessary to distinguish transient quitting (temporarily smoking-free but relapse later) from permanent quitting (lifelong smoking-free) and identify the risk factors associated with permanent quitting.
Objective: To identify and quantify baseline factors associated with success of permanent quitting and describe the full stochastic nature of the smoking addiction pattern using the Alpha-Tocopherol, Beta-Carotene (ATBC) Lung Cancer Prevention study data set.
Methods: The ATBC study is a large (29,133 participants) longitudinal cohort study and it contains unique information about smoking and health status of each participant during every 4-month interval throughout the follow-up period. We used the ATBC data set to model smoking cessation patterns using the proposed discrete-time stochastic mixed-effect model with three states: smoking, transient quitting and permanent quitting (absorbent state). Random participant specific transition probabilities among these states were used to account for participant-to-participant heterogeneity. Another important innovation in our research was to design computationally practical methods for dealing with the size of the data set and complexity of the models. This was achieved by integrating over the Beta distribution of random effects and obtaining the marginal likelihood as an explicit function of the model parameters. We thus avoided complicated model fitting techniques, e.g. numerical integration or Markov Chain Monte Carlo (MCMC). This provided large computational advantages over the more popular model using logistic multivariate normal random effects.
Results: Baseline age was positively associated with probability of making quit attempts (p<0.001). However, years of smoking, cigarette and alcohol consumption had inverse association with probability of making quit attempt (p<0.001). If the quit attempt was made, more cigarette consumption per day was associated with lower probability of relapsing (p=0.003), while more alcohol consumption per day was associated with higher probability of relapsing (p=0.004). Moreover, 5.1 years in age increased the odds of permanent quitting by 10.4% (95% CI: -0.40%~22.4%; p=0.06). Individuals with psychological symptoms were significantly less likely to be successful permanent quitters (p=0.03).
Conclusion: An important, previously unknown, but often debated, finding was that baseline risk factors have different effects on different transition probabilities. We provided mathematical quantification of the participant-to-participant variation in transition probabilities. Our modeling strategy enriched the statistical arsenal of cure modeling and provided new and valuable scientific insight into the smoking addiction behavior.
Title: A Validation Study on Smoking Questionnaire in the Shanghai Women's Health Study
In order to determine the potential exposure levels of secondhand smoke (SHS) at a population level and validate the self-reported smoking questionnaire in the Shanghai Women's Health Study, SWHS, we designed a cross-sectional validation study of urinary cotinine at baseline. A total of 571 urine samples were randomly selected by strata of smoking levels of neither active smoking nor SHS exposure (never SHS), former smokers, ever only exposure to SHS (only SHS), and current smokers. The geometric means (95% confidence intervals) for urinary cotinine were 2.8 (2.2-3.5), 3.4 (2.6-4.4), 5.2 (4.7-5.8) and 933.2 (551.4-1579.4) in ng/ml urine among women who reported never SHS, former smokers, only SHS, and current smokers, respectively. All women, including never SHS, had cotinine detectable in their urine samples. In the linear regression analysis, all the self-reported smoking measures at baseline were significantly correlated with urinary cotinine after adjusting for creatinine and other covariates. Urinary cotinine significantly increased by 0.370 ng/ml urine per one additional cigarette smoked per day by women and by 0.035 ng/ml urine per one additional cigarette smoked per day by their husbands. Spearman correlation coefficient, rs, with creatinine-adjusted urinary cotinine was 0.94 (p< 10-5) for cigarettes smoked per day by current-smoking women at baseline with a great variation of cotinine levels across individuals and 0.51 (p< 10-5) for cigarettes smoked per day by their husbands. The weighted Kappa coefficient was 0.95 (p< 10-5), indicating almost perfect agreement between self-reported smoking status and the urinary cotinine-indicated smoking status. Sensitivity for self-reported current smokers was 88% and specificity was 100%, when using the urinary cotinine-indicated smoking status as a gold standard. In conclusion, the smoking questionnaire of both active smoking and SHS in the SWHS is valid. However, the validity decreases slightly in quantitative measures of smoking due to the large variation of cotinine across individuals. The universal presence of cotinine in the urine samples suggests SHS is a severe public health issue in China.
Title: Applying alternative propensity score techniques to test whether marijuana problems cause depression
Increased marijuana use among adolescents and reports linking marijuana use to other psychiatric disorders has prompted research on the relationship between adolescent marijuana use and later depression. Using propensity score (PS) adjustment techniques we test the potential causal link between adolescent marijuana problems (abuse or dependence) and young adult depression (past-year major depression diagnosis). A cohort of 2,311 first-graders was interviewed over a 15-year period with 75% retention. We used both parametric and non-parametric PS estimation techniques and five alternative applications of the PS, including matching, subclassification, and weighting. The PS adjusted the final logistic regression of depression on marijuana problems by balancing all observed confounding covariates. Unadjusted, 24% of individuals with marijuana problems reported depression whereas only 13% of individuals without marijuana problems reported depression. The PS-adjusted analyses all resulted in increased odds of depression among those with marijuana problems [Odds Ratio range 1.12 (p>0.05) 1.88 (p<0.001)]. The variation in the magnitude of the effect and statistical significance does not directly suggest one of the PS techniques to be superior. We use statistical simulations and comparisons of the baseline covariate balance to determine if one PS method is most effective. This work suggests both methodological and substantive conclusions: non-parametric estimation of the PS may be an improvement over parametric estimation for these data, and adolescents with marijuana problems are at increased odds for young adult major depression, but the causal link is modest.
Title: The immunoglobulin superfamily in Anopheles gambiae: potential immune molecules in the malaria vector
Although falciparum malaria is one of the most devastating infectious diseases in the world, proper control of the parasite and its vector, the Anopheles gambiae mosquito, remain elusive. Because the mosquito vector is required for transmission and because vector control can be cost-effective and is unaffected by compliance of individuals, we are focused on finding a way to control malaria though molecular manipulation of the mosquito. The mosquito's immune system is active against the malaria parasite yet determination of the molecules and mechanisms involved is still in its infancy. Because immunoglobulin superfamily members are among the most important and best characterized molecules of the immune system and this family has not been well-studied in the context of insect immunity, we initiated a screen of mosquito proteins containing immunoglobulin domains to identify candidate immune molecules. First, a bioinformatics-based analysis was used to define the IgSF in Anopheles gambiae and select and test candidates for implication in immune functions. This screen identified 145 genes that have at least one immunoglobulin domain. From here, a gene expression-based analysis was performed in order to determine the infection responsiveness of individual exons representing each of the 145 IgSF genes. Expression profiles of adult female mosquitoes and mosquito cell lines challenged with Plasmodium parasites, Gram negative bacteria, Gram positive bacteria and fungi show that 237 exons from 84 genes are significantly regulated in response to at least one challenge, many responding to multiple challenges. We think that some of this regulation is attributable to immune defense. Data from both the computational and microarray screens were used to choose six genes as candidates for further functional analyses to establish possible immune relevance. Candidates were named Infection Responsive with Immunoglobulin Domain 1-6 (IRID1-6), and chosen for testing based on domain composition suggestive of immune relevance and significant regulation of at least two exons upon challenge. Each of the six IRID genes were silenced in adult female mosquitoes using RNAi and subsequently challenged with Gram positive bacteria, Gram negative bacteria or Plasmodium. Survival rates and oocyst loads from these assays suggest several of these candidates are involved in defense against these pathogens. One candidate, IRID3, seems to have a drastic effect on a mosquito's ability to survive after bacterial challenge and another, IRID6, has a modulatory effect on Plasmodium falciparum infection in the mosquito. From here, we can begin our characterization of IRID3 and IRID6 which may not only represent a novel class of innate immune molecules. IRID6 is of particular interest as a potential target for malaria control at the vector level.
Title: Regulatory T Cells Mediate Seoul Virus Persistence and Pathology in Norway Rats
Hantaviruses cause one of two diseases in humans, hantavirus pulmonary syndrome (HPS) or hemorrhagic fever with renal syndrome (HFRS). Human pathology is hypothesized to be caused by sustained, elevated levels of proinflammatory cytokines. In rodents, hantaviruses are transmitted horizontally and cause persistent infection in the absence of disease. The mechanisms mediating hantavirus persistence in rodents are unknown. Regulatory T cells contribute to persistence of several pathogens and act locally at sites of pathogen replication by suppressing proinflammatory responses. Following inoculation with Seoul virus, regulatory T cells (CD4+CD25+FoxP3+) are elevated in the lungs during the persistent phase of infection (i.e. at Day 30 p.i.). Additionally, expression of foxp3, a transcription factor that is unique to regulatory T cells, and TGF-, a cytokine produced by regulatory T cells, is elevated in the lungs of male Norway rats 30 days post-inoculation (p.i.), but remains unchanged in the spleen during infection. To test the hypothesis that regulatory T cells affect persistence of Seoul virus, male Norway rats were administered anti-CD25 mAb, to deplete regulatory T cells, or saline and were inoculated with Seoul virus or vehicle. Following depletion of regulatory T cells, Seoul virus RNA copies in the lungs and proportion of rats shedding virus in saliva are reduced during infection. Additionally, depletion of regulatory T cells significantly reduces pathology, including development of hemorrhage and edema, in the lungs during Seoul virus infection. Taken together, these data suggest that regulatory T cell responses are elevated during infection and may contribute to Seoul virus persistence and pathology in male Norway rats.
Title: Increased androgenic sensitivity during aging contributes to the prostate proliferative potential
Benign prostate hyperplasia (BPH) is a prevalent condition in aging men. Spontaneous epithelial cell hyperplasia in the dorsal and lateral lobes of old Brown Norway rats is analogous to BPH in humans. Therefore, the Brown Norway (B-N) rat represents an excellent model to investigate the mechanisms that regulate cell proliferation and prostate hyperplasia. The majority of cells in the adult prostate are quiescent in the G0 phase and in order for hyperplasia to occur, cells must escape from cell cycle arrest and enter S phase through G1/S restriction. In the androgen withdrawn-replenished Brown Norway rats, proliferation rate quantified by BrdU labeling index of all three lobes in response to androgens is time and dose dependent, and peak at day 3. Under the similar serum testosterone levels monitored by radioimmunoassay, aging dorsal and lateral prostates have higher proliferation rate than the young animals, while there is no age dependent difference in proliferation rate of ventral lobes. Immunoblots demonstrated the abundance of cyclinD1, cdk6, cyclin E, cdk2 was significantly upregulated, and P27Kip1 abundance was significantly downregulated, affected by androgens. Interestingly, aging dorsal and lateral prostates have higher cyclin D1 and cyclin E abundance than the young ones, no matter the intact control animals, or the 3days androgen replenished young and aging animals under the similar serum testosterone levels. Androgens can also affect Rb phosphorylation, shown by immunoprecipitation and immunoblots. Immunostaining demonstrated changing patterns of cdk4 subcelluar localization and probability of cyclinD1 nuclei localization corresponded to the time frame of cell proliferation affected by androgens. These data suggested increased androgenic sensitivity and its effect on the key cell cycle regulators (especially those related to G1/S transitions) may be one of the multiple mechanisms leading to BPH, which develops during aging while the serum testosterone levels decrease.