Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/53730
Title: | 以統計模式評估台灣社區研究族群的酒精飲用與巴金森氏症多階段進展 Statistical models for evaluating the effect of alcohol consumption and risk of multistate process of Parkinson Disease using a community-based population study in Taiwan |
Authors: | Fu-Chen Liang 梁富珍 |
Advisor: | 陳秀熙(Hsiu-Hsi Chen) |
Keyword: | 統計模式,帕金森氏症多階段進展,飲酒, statistical model,parkinson disease,alcohol, |
Publication Year : | 2015 |
Degree: | 碩士 |
Abstract: | Background
In spite of numerous studies addressing the effect of alcohol intake on the risk of Parkinson disease (PD) and several systematic reviews, the results were still controversial and inconsistent. In addition to the selection of sample and study design issues, one of important issue related to long latent period may have strong influence on the causal relationship between alcohol intake and the risk of PD. The obstacle of considering latent period is the unknown latent period that precluded the conventional statistical methods from dealing with this issue. Aims By making use a longitudinal Parkinson disease (PD) screened or a community-based integrated screening cohort with complete information on alcohol intake with dichotomous, frequency and duration of alcohol intake, this thesis aimed (1) to investigate the effect of alcohol intake on the risk of PD in terms of dichotomous property (yes/no) or three categories (current drinkers, ex-drinkers, and non-drinkers), frequency of alcohol intake, and duration of alcohol intake with adjustment for demographic features, smoking, betel nuts chewing. (2) To investigate the effect of alcohol intake on the risk of PD making allowance for different latent periods with a naïve method with adjustment for demographic features, smoking, and betel nuts chewing. (3) to provide a formal assessment of the hazard of alcohol-induced process for the development of PD compared with that of non-alcohol-induced process with a Markov and semi-Markov model; (4) to estimate the possible latent period of alcohol intake and proportion of PD attributed to alcohol intake based on the parameters obtained from (3). Materials and Methods The study population comprised two cohorts. The first cohort was included participants in the Parkinson disease (PD) screening program in 2001. A total of 9,829 eligible subjects aged over 60 formed a PD screening cohort was enrolled. Among them, a total of 55 PD cases were detected in 2001 and other 208 PD cases was diagnosed in clinical practice. The PD screened cohort with complete information on age at exposure was composed of 5061 subjects. The second cohorts included participants aged over 60 in the Keelung community-based integrated screening program during the period between 2002 and 2004. A total of 18,414 subjects including 324 PD cases were enrolled in our analysis. The combined 18,414 subjects with 5061 subjects forms a complete cohort for the analysis (N=23,475). Information on age, gender, education level, smoking, betel nut chewing, and alcohol consumption habit on age at exposure, duration of exposure, current status (quit/current), frequency by per week, and quantity for each time were collected. The logistic regression model was used to assess the alcohol intake and other risk factors associated with Parkinson disease. To tackle the alcohol-induced process start to affect the development of PD, we constructed a three-state alcohol-induced and non-alcohol-induced PD model to estimate the parameters using the empirical data from the community-based KCIS screened cohort (Keelung community-based integrated screening program). Parametric methods of Markov and semi-Markov models and non-parametric method with piecewise constant hazards were applied. According to the lethality function, the proportion for PD caused by alcohol intake was calculated based on the estimated results from the proposed models. A five-state Markov model classified by Hoehn and Yahr staging scale (H&Y scale) was developed with further consideration of covariates. Results The effect of alcohol intake (aOR: 1.03, 95% CI: 0.73-1.46) on Parkinson disease (PD) after adjusting the confounding factors was found in PD screening cohort (N=9,829) but not statistically significant. The crude odds ratios of alcohol intake were 0.70 (95% CI: 0.40-1.23) and 0.69 (95% CI: 0.51-0.95) for PD screening cohort (N=5,061) and KCIS screened cohort (N=18,414). The alcohol intake was negative associated with PD in two combined cohorts (N=23,475). After adjusting for smoking and other demographic variables, a non-significant 0.84 (95%CI: 0.62-1.15) of adjusted odds ratio for alcohol intake was found. The adjusted odds ratios were 0.67 (95%CI: 0.38-1.16) and 1.08 (95%CI: 0.71-1.64) for low and high intensity of alcohol intake, respectively. The similar findings were found in frequency and duration of alcohol intake. The odds ratio increased from 0.84% (95%CI: 0.62-1.15) without adjustment for latent period and 1.20 (95%CI: 0.85-1.69) with the adjustment for latent period when making allowance for other confounding factors. In the three-state alcohol-induced PD Markov model, the annual rate of commencing alcohol intake was estimated as 3 (95% CI: 2.95, 3.13) per 1000 person-years. The alcohol-induced and non-alcohol-induced hazards were estimated as 37 (95% CI: 29, 46) and 26 (95% CI: 23, 29) per 100,000 person years, respectively, (hazard ratio (HR) =1.42). According the lethality function of alcohol intake, the proportion for PD caused by alcohol intake was 29.73%. The semi-Markov model also suggests that both alcohol- and non-alcohol-induced hazards (HR=1.63) increased with time. The proportion for PD caused by alcohol intake was 38.73%. According to semi-Markov model, it took around 40 years of alcohol intake to render the risk of PD for those with alcohol intake higher than that for those without alcohol intake. Smokers had around 8-fold risk of alcohol intake, but had protective effects on the hazard functions to develop PD on both alcohol-induced (RR=0.95, 95% CI: 0.54, 1.66) and non-alcohol induced (RR=0.52, 95% CI: 0.38, 0.72) PD. However, regardless of smoking status, alcohol-induced hazards were still higher than non-alcohol induced. Based on the piece-wise exponential Markov model, the proportion for PD caused by alcohol intake was 25.5% for those exposed to alcohol at 20 years-old whereas this figure was reduced to 6.8% if age at exposure starts 30 years. In the natural history model for PD taking Hoehn and Yahr staging scale (H&Y scale) and detection mode into account, we found that those who ever had alcohol intake had higher risk of developing Parkinson’s disease (RR=1.62; 95% CI:1.01-2.59). Conclusions We developed a series of novel Markov and semi-Markov models to estimate the precise effect of alcohol intake on the risk of Parkinson disease (PD) taking into latent period based on a large population- and community-based screened cohort with complete information on age at exposure to alcohol and intensity of alcohol intake. The latent period we found through such a formal assessment was 40 years on which we were based to estimate 1.60-fold risk for alcohol-induced PD versus non-alcohol induced process, accounting for 39% PD cases provided the duration of alcohol is sufficient longer than the latent period. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/53730 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 流行病學與預防醫學研究所 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-104-1.pdf Restricted Access | 5.24 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.