• The Korean Journal of Applied Statistics
• /
• v.29 no.7
• /
• pp.1459-1473
• /
• 2016

Both longitudinal data and survival data are collected simultaneously in longitudinal data which are observed throughout the passage of time. In this case, the effect of the independent variable becomes biased (provided that sole use of longitudinal data analysis does not consider the relation between both data used) if the missing that occurred in the longitudinal data is non-ignorable because it is caused by a correlation with the survival data. A joint model of longitudinal data and survival data was studied as a solution for such problem in order to obtain an unbiased result by considering the survival model for the cause of missing. In this paper, a joint model of the longitudinal zero-inflated count data and survival data is studied by replacing the longitudinal part with zero-inflated count data. A hurdle model and proportional hazards model were used for each longitudinal zero inflated count data and survival data; in addition, both sub-models were linked based on the assumption that the random effect of sub-models follow the multivariate normal distribution. We used the EM algorithm for the maximum likelihood estimator of parameters and estimated standard errors of parameters were calculated using the profile likelihood method. In simulation, we observed a better performance of the joint model in bias and coverage probability compared to the separate model.

• Korean Circulation Journal
• /
• v.48 no.12
• /
• pp.1148-1156
• /
• 2018

Background and Objectives: It is controversial that decreased left ventricular function could predict poststroke outcomes. The purpose of this study is to elucidate whether left ventricular ejection fraction (LVEF) can predict cardiovascular events and mortality in acute ischemic stroke (AIS) without atrial fibrillation (AF) and coronary heart disease (CHD). Methods: Transthoracic echocardiography was conducted consecutively in patients with AIS or transient ischemic attack at Soonchunhyang University Hospital between January 2008 and July 2016. The clinical data and echocardiographic LVEF of 1,465 patients were reviewed after excluding AF and CHD. Poststroke disability, major adverse cardiac events (MACE; nonfatal stroke, nonfatal myocardial infarction, and cardiovascular death) and all-cause mortality during 1 year after index stroke were prospectively captured. Cox proportional hazards regressions analysis were applied adjusting traditional risk factors and potential determinants. Results: The mean follow-up time was $259.9{\pm}148.8days$ with a total of 29 non-fatal strokes, 3 myocardial infarctions, 33 cardiovascular deaths, and 53 all-cause mortality. The cumulative incidence of MACE and all-cause mortality were significantly higher in the lowest LVEF (<55) group compared with the others (p=0.022 and 0.009). In prediction models, LVEF (per 10%) had hazards ratios of 0.54 (95% confidence interval [CI], 0.36-0.80, p=0.002) for MACE and 0.61 (95% CI, 0.39-0.97, p=0.037) for all-cause mortality. Conclusions: LVEF could be an independent predictor of cardiovascular events and mortality after AIS in the absence of AF and CHD.

• Asian-Australasian Journal of Animal Sciences
• /
• v.33 no.8
• /
• pp.1360-1368
• /
• 2020

Objective: The aim of our study was to determine the associations of heifer reproductive performance with survival up to the first calving, first-lactation milk yield, and the probability of being culled within 50 days after first calving. Methods: Data from 33 large Holstein-Friesian commercial dairy herds were gathered from the official milk recording database in Hungary. The data of heifers first inseminated between January 1, 2011 and December 31, 2014 were analyzed retrospectively, using Cox proportional hazards models, competing risks models, multivariate linear and logistic mixed-effects models. Results: Heifers (n = 35,128) with younger age at conception were more likely to remain in the herd until calving, and each additional month in age at conception increased culling risk by 5.1%. Season of birth was related to first-lactation milk yield (MY1; n = 19,931), with cows born in autumn having the highest milk production (p<0.001). The highest MY1 was achieved by heifers that first calved between 22.00 and 25.99 months of age. Heifers that calved in autumn had the highest MY1, whereas calving in summer was related to the lowest milk production (p<0.001). The risk of culling within 50 days in milk in first lactation (n = 21,225) increased along with first calving age, e.g. heifers that first calved after 30 months of age were 5.52-times more likely to be culled compared to heifers that calved before 22 months of age (p<0.001). Calving difficulty was related to higher culling risk in early lactation (p<0.001). Heifers that required caesarean section were 24.01-times more likely to leave the herd within 50 days after first calving compared to heifers that needed no assistance (p<0.001). Conclusion: Reproductive performance of replacement heifers is closely linked to longevity and milk production in dairy herds.

• Journal of Preventive Medicine and Public Health
• /
• v.41 no.1
• /
• pp.23-29
• /
• 2008

Objectives : We examined the association between alcohol consumption and incidence of colorectal cancer in elderly Koreans. Methods : The cohort members (n=14,304) consisted of 4,834 males and 9,470 females derived from the Korea Elderly Pharmacoepidemiologic Cohort (KEPEC), a population-based dynamic cohort. They were aged 65 years old or older and lived in Busan between 1993-1998; they were beneficiaries of the Korean Medical Insurance Corporation (KMIC). Baseline information was surveyed by a self-administered, mailed questionnaire. This study population was restricted to 14,304 participants who reported alcohol drinking habits on the questionnaire and had not been diagnosed with colorectal cancer at baseline. The adjusted hazard ratios (aHR) of status, type, frequency and daily average amount of alcohol consumption were computed with Cox's proportional hazard model, with the never-drinkers as a reference group and controlling for age and gender. Results : After 4.82 person-years of mean follow-up 112 cases of colorectal cancer occurred. The incidence densities of colorectal cancer were 161 (95% CI=123-200) for never-drinkers, 219 (95% CI=125-339) for ex-drinkers, and 137 (95% CI=84-189) for current-drinkers per 100,000 person-year. The status, type, frequency, and daily average amount of alcohol consumption were not significantly related to the incidence of colorectal cancer after controlling for age and gender. Conclusions : There was no significant association between alcohol consumption and colorectal cancer among elderly people after controlling for age and gender.

• Korean Journal of Radiology
• /
• v.24 no.6
• /
• pp.590-598
• /
• 2023

Objective: To investigate whether the levels of inflammation detected by 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) can predict disease relapse in immunoglobulin G4-related disease (IgG4-RD) patients receiving standard induction steroid therapy. Materials and Methods: This prospective study analyzed pretherapy FDG PET/CT images from 48 patients (mean age, 63 ± 12.9 years; 45 males and 3 females) diagnosed with IgG4-RD between September 2008 and February 2018, who subsequently received standard induction steroid therapy as the first-line treatment. Multivariable Cox proportional hazards models were used to identify the potential prognostic factors associated with relapse-free survival (RFS). Results: The median follow-up time for the entire cohort was 1913 days (interquartile range [IQR], 803-2929 days). Relapse occurred in 81.3% (39/48) patients during the follow-up period. The median time to relapse was 210 days (IQR, 140-308 days) after completion of standardized induction steroid therapy. Among the 17 parameters analyzed, Cox proportional hazard analysis identified whole-body total lesion glycolysis (WTLG) > 600 on FDG-PET as an independent risk factor for disease relapse (median RFS, 175 vs. 308 days; adjusted hazard ratio, 2.196 [95% confidence interval: 1.080-4.374]; P = 0.030). Conclusion: WTLG on pretherapy FDG PET/CT was the only significant factor associated with RFS in IgG-RD patients receiving standard steroid induction therapy.

• Asian Pacific Journal of Cancer Prevention
• /
• v.15 no.4
• /
• pp.1563-1566
• /
• 2014

Background: Esophageal squamous cell carcinoma (ESCC) accounts for most esophageal cancer in Asia, and is the sixth common cause of cancer-related deaths worldwide. Previous studies indicated HOXB7 is overexpressed in ESCC tissues, but data on prognostic value are limited. Methods: A total of 76 advanced ESCC cases were investigated. Immunohistochemistry (IHC) was used to detect the expression levels of HOXB7 and Kaplan-Meier curves and Cox regression models to determine prognostic significance. Stratified analysis was also performed according to lymph node (LN) status. Results: Kaplan-Meier curve analysis indicated that HOXB7 positive patients had significantly shorter overall survival (OS) than HOXB7 negative patients. Multivariate analysis using the Cox proportional hazards model indicated only TNM stage and HOXB7 expression to be independent predictors of overall survival of advanced ESCC patients. HOXB7 indicated poor OS in both lymph node negative (LN-) and lymph node positive (LN+) patients. Conclusion: HOXB7 predicts poor prognosis of advanced ESCC patients and can be applied as an independent prognostic predictor.

• Journal of Preventive Medicine and Public Health
• /
• v.47 no.6
• /
• pp.289-297
• /
• 2014

Objectives: This community-based cohort study was performed to investigate the incidence of and factors related to self-reported fragility fractures among middle-aged and elderly women living in rural Korea. Methods: The osteoporosis cohort recruited 430 women 40 to 69 years old in 1999, and 396 of these women were followed over 11 years. In 1999, questionnaires from all participants assessed general characteristics, medical history, lifestyle, menstrual and reproductive characteristics, and bone mineral density. In 2010, self-reported fractures and the date, site, and cause of these fractures were recorded. Cox proportional hazards models were used to calculate hazard ratios (HRs). Results: Seventy-six participants among 3949.7 person-years experienced fragility fractures during the 11-year follow-up. The incidence of fragility fractures was 1924.2 per 100 000 person-years (95% confidence interval [CI], 1491.6 to 2356.8). In the multivariate model, low body mass index (HR, 2.66; 95% CI, 1.13 to 6.24), a parental history of osteoporosis (HR, 2.03; 95% CI, 1.18 to 3.49), and postmenopausal status (HR, 3.50; 95% CI, 1.05 to 11.67) were significantly related to fragility fracture. Conclusions: Fracture prevention programs are needed among postmenopausal, rural, Korean women with a low body mass index and parental history of osteoporosis Korea.

• Communications for Statistical Applications and Methods
• /
• v.17 no.2
• /
• pp.231-239
• /
• 2010

In vaccine research, proportional hazards model including only first event have been widely used for estimating vaccine efficacy because it is easy to interpret and convenient. However, this method causes not only loss of information but also biased result when heterogeneity of study subject in exposure and susceptibility exists. Furthermore, it is hard to ignore the possibility that each event is correlated with each other in the repeated events. Therefore, we compare various statistical models to estimate vaccine efficacy under various situations with heterogeneity and event dependency.

• Journal of Preventive Medicine and Public Health
• /
• v.46 no.6
• /
• pp.319-328
• /
• 2013

Objectives: This study estimated the association of cardiovascular health behaviors with the risk of all-cause and cardiovascular disease (CVD) mortality in middle-aged men in Korea. Methods: In total, 12 538 men aged 40 to 59 years were enrolled in 1993 and followed up through 2011. Cardiovascular health metrics defined the following lifestyle behaviors proposed by the American Heart Association: smoking, physical activity, body mass index, diet habit score, total cholesterol, blood pressure, and fasting blood glucose. The cardiovascular health metrics score was calculated as a single categorical variable, by assigning 1 point to each ideal healthy behavior. A Cox proportional hazards regression model was used to estimate the hazard ratio of cardiovascular health behavior. Population attributable risks (PARs) were calculated from the significant cardiovascular health metrics. Results: There were 1054 total and 171 CVD deaths over 230 690 person-years of follow-up. The prevalence of meeting all 7 cardiovascular health metrics was 0.67%. Current smoking, elevated blood pressure, and high fasting blood glucose were significantly associated with all-cause and CVD mortality. The adjusted PARs for the 3 significant metrics combined were 35.2% (95% confidence interval [CI], 21.7 to 47.4) and 52.8% (95% CI, 22.0 to 74.0) for all-cause and CVD mortality, respectively. The adjusted hazard ratios of the groups with a 6-7 vs. 0-2 cardiovascular health metrics score were 0.42 (95% CI, 0.31 to 0.59) for all-cause mortality and 0.10 (95% CI, 0.03 to 0.29) for CVD mortality. Conclusions: Among cardiovascular health behaviors, not smoking, normal blood pressure, and recommended fasting blood glucose levels were associated with reduced risks of all-cause and CVD mortality. Meeting a greater number of cardiovascular health metrics was associated with a lower risk of all-cause and CVD mortality.

• Safety and Health at Work
• /
• v.8 no.3
• /
• pp.258-266
• /
• 2017

Background: Welders are exposed to many known and suspected carcinogens. An excess lung cancer risk among welders is well established, but whether this is attributable to welding fumes is unclear. Excess risks of other cancers have been suggested, but not established. We investigated welding cancer risks in the population-based Canadian Census Health and Environmental Cohort. Methods: Among 1.1 million male workers, 12,845 welders were identified using Standard Occupational Classification codes and followed through retrospective linkage of 1991 Canadian Long Form Census and Canadian Cancer Registry (1992-2010) records. Hazard ratios (HRs) were calculated using Cox proportional hazards models based on estimated risks of lung cancer, mesothelioma, and nasal, brain, stomach, kidney, and bladder cancers, and ocular melanoma. Lung cancer histological subtypes and risks by industry group and for occasional welders were examined. Some analyses restricted comparisons to blue-collar workers to minimize effects of potential confounders. Results: Among welders, elevated risks were observed for lung cancer [HR: 1.16, 95% confidence interval (CI): 1.03-1.31], mesothelioma (HR: 1.78, 95% CI: 1.01-3.18), bladder cancer (HR: 1.40, 95% CI: 1.15-1.70), and kidney cancer (HR: 1.30, 95% CI: 1.01-1.67). When restricted to blue-collar workers, lung cancer and mesothelioma risks were attenuated, while bladder and kidney cancer risks increased. Conclusion: Excess risks of lung cancer and mesothelioma may be partly attributable to factors including smoking and asbestos. Welding-specific exposures may increase bladder and kidney cancer risks, and particular sources of exposure should be investigated. Studies that are able to disentangle welding effects from smoking and asbestos exposure are needed.