Increased Risk of Dementia in Patients Exposed to Nitrogen Dioxide and Carbon Monoxide: A Population-Based Retrospective Cohort Study

Background The air pollution caused by vehicular emissions is associated with cognitive decline. However, the associations between the levels of nitrogen dioxide (NO2) and carbon monoxide (CO) exposure and dementia remain poorly defined and have been addressed in only a few previous studies. Materials and Methods In this study, we obtained data on 29547 people from the National Health Insurance Research Database (NHIRD) of Taiwan, including data on 1720 patients diagnosed with dementia between 2000 and 2010, and we evaluated the risk of dementia among four levels of air pollutant. Detailed data on daily air pollution were available from January 1, 1998 to December 31, 2010. Yearly average concentrations of pollutants were calculated from the baseline to the date of dementia occurrence, withdrawal of patients, or the end of the study, and these data were categorized into quartiles, with Q1 being the lowest level and Q4 being the highest. Results In the case of NO2, the adjusted hazard ratios (HRs) of dementia for all participants in Q2, Q3, and Q4 compared to Q1 were 1.10 (95% confidence interval (CI), 0.96–1.26), 1.01 (95% CI, 0.87–1.17), and 1.54 (95% CI, 1.34–1.77), and in the case of CO, the adjusted HRs were 1.07 (95% CI, 0.92–1.25), 1.37 (95% CI, 1.19–1.58), and 1.61 (95% CI, 1.39–1.85). Conclusion The results of this large retrospective, population-based study indicate that exposure to NO2 and CO is associated with an increased risk of dementia in the Taiwanese population.


Introduction
Ambient air pollution includes solid and gaseous pollutants [1,2]. Most of the studies that have investigated the effects of pollutants on cognitive functions have examined the influence of solid pollutants [3][4][5][6][7][8]. However, exposure to ambient gaseous pollutants such as nitrogen dioxide (NO 2 ) is known to increase the risk of cerebrovascular and neurodegenerative diseases and ischemic stroke [9][10][11][12]. Cerebrovascular disease is the principal contributor to dementia [13,14], and Alzheimer's disease (AD) is the most common neurodegenerative disease. Moreover, a population-base study reported that dementia often developed after the occurrence of an ischemic stroke [15]. Several previous studies have suggested negative associations between NO 2 exposure and cognitive development in children, including preschool children [16][17][18], and animal studies have indicated that NO 2 exposure inhibits the recovery of nerve function after a stroke [19,20]. In addition, one animal study reported that nitration can induce beta-amyloid aggregation and plaque formation [21]; beta-amyloid aggregation is a pathologic hallmark of AD. However, a literature search indicated that only a few studies have been conducted to address the link between NO 2 exposure and cognitive function in adults. In a recent study conducted on 1496 middle-aged people living in Los Angeles, no statistically significant correlation was detected between the level of NO 2 exposure and cognitive functions [22]. Therefore, we conducted a retrospective cohort study to determine the association between NO 2 and dementia risk. Furthermore, in this study, we evaluated the influence of carbon monoxide (CO), because acute CO poisoning may cause headache, nausea, malaise, and fatigue [23], and chronic CO exposure has been linked to depression, confusion, memory loss, and cognitive decline [24,25]. Comparison between this study with other environmental study of Taiwan NHRID, the main difference is the residential area definition. In previous studies, the residential area is as the insurance area [26]. In present study, we defined the residential areas as the location of clinics which subjects sought treatment for acute upper respiratory infections.

Data sources and study population
In March 1995, the Taiwan National Health Insurance (NHI) program, which is a single-payer, compulsory social insurance system that has provided insurance coverage to almost every citizen in Taiwan, was established. The NHI covered approximately 99% of the 22.96 million citizens in Taiwan at the end of 2007 [27]. To protect patient privacy, the data on patient identities are encrypted in the National Health Insurance Research Database (NHIRD), and the database is accessible to researchers and the public in Taiwan. In this study, we used a subset of the NHIRD data containing comprehensive health-care data, including files on ambulatory care claims, inpatient claims, and prescriptions received by 1000000 people who were randomly selected from all insured beneficiaries. These data files can be linked through an encrypted but unique personal identification number and, thus, provide a longitudinal medical history of each patient. The health status of each person was identified according to the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM).

Exposure assessment
Across Taiwan, 74 ambient air quality monitoring stations are located based on population density. Air quality data are maintained by Taiwan Environmental Protection Administration. [28]. A database containing daily NO 2 and CO concentrations measured at the monitoring stations was available for the period from January 1, 1998 to December 31, 2010. The people included in this study were assigned pollutant-exposure values based on the data obtained from the monitoring station present in the residential district in which the clinic where the people most frequently sought treatment for acute upper respiratory infection was located (ICD-9-CM Code 460). Yearly average concentrations of pollutants were calculated from the baseline to the date of dementia occurrence, the withdrawal of patients, or the end of the study period, and the data were categorized into quartiles.

Study patients
We identified 29547 people who were aged 50 years or older and for whom estimable air pollution data were available, but who did not present a history of head injury (ICD

Ethics statement
Because identification numbers of patients had been encrypted, patient consent was not required for this study. This study was approved by the Research Ethic Committee at China Medical University (CMU-REC-101-012). The committee waived the requirement for consent.

Statistical analysis
We used x 2 tests to examine the distributions of sex, monthly income (New Taiwan 6, 241.7-296.9, .296.9). A one-way analysis of variance (ANOVA) was performed to compare the age among the quartiles of NO 2 and CO concentrations. We calculated the incidence density rates of dementia in person-years in each quarter stratified according to sex. The incidence rate ratio (IRR) was estimated using a Poisson regression. Univariate and multivariate Cox proportional hazard regression analyses were performed to calculate the hazard ratios (HRs) and 95% confidence intervals (CIs) of the risk of dementia in association with pollutant levels. Multiple models were tested by controlling for age, sex, monthly income, DM, HT, IHD, COPD, alcoholism, and urbanization. Plots of the Kaplan-Meier analysis were used to determine the probability of people remaining without dementia, and the logrank test was used to evaluate the differences among quartiles of pollutant concentrations. All analyses were performed using SAS 9.2 software (SAS Institute Inc., Cary, NC, USA), and the Kaplan-Meier survival curve was plotted using the Statistical Package for the Social Sciences (Version 15.1; SPSS Inc, Chicago, IL, USA). All tests were considered statistically significant when two-tailed P values were ,.05.

Results
We obtained a total of 29547 and 29537 data on daily NO 2 and CO exposure, respectively. Dementia was not present at the baseline (2000), and 1720 people developed dementia after followup (yearly CO data were available for 1718 people). We categorized the NO 2 and CO levels into quartiles, with Q1 being the lowest level and Q4 being the highest. The people included in this study had a mean age of 61.4 years (SD 8.5 y). In both the NO 2 and CO groups, the highest level of the quartiles was associated with the people being slightly younger, more frequently earning a high monthly income, and living in a highly urbanized residential area, but less frequently exhibiting IHD and COPD compared with other quartiles (Tables 1 and 2). Table 3 shows the associations between the gaseous pollutant levels and the risk of dementia. Among the quartiles Q1, Q2, Q3, and Q4 of NO 2 in all patients, the IRRs in Q2, Q3, and Q4 compared with that in Q1 were 1.05, 0.90, and 1.35, and the adjusted HRs of dementia were 1.10 (95% CI, 0.96-1.26), 1.01 (95% CI, 0.87-1.17), and 1.54 (95% CI, 1.34-1.77), respectively. Among men, we determined that the IRRs in Q2, Q3, and Q4 compared with that in Q1 were 1.08, 0.79, and 1.28, and the adjusted HRs were 1.16 (95% CI, 0.95-1.43), 0.89 (95% CI, 0.71-1.11), and 1.52 (95% CI, 1.23-1.88), respectively. Among women, the IRRs in Q2, Q3, and Q4 compared with that in Q1 were 1.05, 1.11, and 1.56, and the adjusted HRs were 1.05 (95% CI, 0.87-1.27), 1.11 (95% CI, 0.92-1.35), and 1.56 (95% CI, 1.29-1.87), respectively. When the data on sex were stratified or merged for analysis, statistically significant correlations of IRRs and adjusted HRs were measured in Q4 compared with those in Q1.
Among the quartiles of CO concentration, the IRRs in Q2, Q3, and Q4 compared with that in Q1 were 0.96, 1.  Figures 1 and 2 show the Kaplan-Meier curves of freedom that were calculated for dementia and are separated according to pollutant level. Statistically significant differences in the occurrence of dementia were observed among the quartiles of NO 2 and CO concentrations (log-rank test, P,.001).

Discussion
The major finding of previous animal study was that nitration was highly correlated with beta-amyloid aggregation and plaque formation, and beta-amyloid aggregation is a pathologic hallmark of AD [21]. Another animal study indicated that NO 2 expose can exacerbate the ultra structural impairment of synapses in stroke rats, and induce neuronal damage in healthy rats [29]. The apolipoprotein E (APOE) e4 allele was a well know genetic risk factor or AD, and a randomized clinical trial has found CO poisoning can induce APOE e4 carriers suffer greater morbidity [30].
The major finding of our study was that increased exposure to NO 2 (Q4) is associated with an enhanced risk of dementia in men and women. The probability of dementia occurrence was increased by 52%-56% in Q4 compared with Q1. A similar trend was observed in the CO group, and the results collectively showed that increasing levels of the 2 pollutants increased the risk of dementia in a dose-dependent manner.
This study was a national population-based investigation on ambient air pollution and dementia. Therefore, collecting individual exposure data was not feasible. To obtain exposure data associated with study patients, previous studies have identified the residential areas of patients by employing a GIS-based system. To protect the privacy of patients, the NHIRD does not provide patients' addresses. Therefore, we identified the residential areas of the patients based on the location of the clinic at which the patients most frequently sought treatment for acute upper respiratory tract infection. In the United States, upper respiratory tract infections are the most common type of infectious disease, and each adult experiences approximately 3 respiratory infections annually [31]. Identifying residential areas in the accessible medical resources, as we did in this study, is more accurate than listing patients according to insurance area [32,33].
Previous studies have suggested that smoking and drinking alcohol are highly correlated with the risk of AD [34][35][36][37][38][39][40]. Because of the limitations of the NHIRD, we could not obtain data on the smoking or drinking status of the patients. Therefore, we performed multivariate analysis with COPD and alcoholism adjusted in accordance with previous studies that indicated that smoking is a major causative factor in the development of COPD, and in which alcoholism was diagnosed based on drinking patterns and the attitudes of patients [41][42][43]. In Taiwan, women are not encouraged to smoke or drink alcohol, as reflected in the low prevalence of these behaviors among women (3% and 1%, respectively) [44,45]. We were able to overcome this limitation by stratifying and adjusting the data according to sex [46].
We adjusted for urbanization in the multivariate analysis. The level of urbanization was determined according to population density (number of people/km 2 ), the population ratio of people with a college-level education or higher, the population ratio of people aged over 65 years, the population ratio of agricultural workers, and the number of physicians per 100000 people [47]. The 359 communities in Taiwan were classified into 7groups: highly urbanized area, moderately urbanized area, boomtown, general town, aging town, agricultural town, and remote town. This classification method has been used in several studies [48][49][50].
In addition, we obtained results contrasting those related to dementia, as shown in Tables 1 and 2: the frequency of IHD and COPD were low at the highest level of the pollutants. These results agree with the explanation provided by previous studies suggesting that patients who are highly educated and earn a high monthly income live in areas where the level of air pollutants is high [6,22].
The strengths of this study are the following. First, this study was based on a long follow-up period, which allowed the possible occurrence of dementia to be assessed. Second, Taiwan launched a national health insurance (NHI) in 1995, operated by a singlebuyer, the government. All insurance claims should be scrutinized by medical reimbursement specialists and peer review. The diagnoses of dementia were based on the ICD-9 code determined by qualified clinical neurology physicians under strict audit in the reimbursement process. Therefore the diagnoses and codes for dementia should be accurate and reliable. Third, this study was conducted using a large population derived from the NHIRD. In Taiwan, the government is the only compulsory social insurance provider; approximately 99% of the 23.74 million citizens of Taiwan are enrolled in the NHI program. Because this was a nationwide study, we considered urbanized towns throughout Taiwan. Lastly, in this study, cerebrovascular and cardiovascular diseases were considered and the association between pollutants and dementia was evaluated. We excluded subjects with cardiovascular before the index date in this study because cardiovascular was a widely known predictor for dementia. IHD increased 27% risk for dementia in both model 1 and model 2. (Table S1).
Certain limitations of this study should be considered. First, the evidence derived from a retrospective cohort study is generally lower in statistical quality than that obtained from randomized trials because, in such retrospective studies, potential biases exist that are related to the adjustment of confounding variables. Despite our meticulous study design and the measures adopted to control for confounding factors, bias resulting from unknown confounders may have affected our results. Second, all data in the   NHIRD are anonymous. Thus, relevant clinical variables, such as imaging results and pathology findings, were unavailable for the patient cases included in this study. Third, the participants were assigned to residential districts based on the clinic where they most frequently sought treatment for acute upper respiratory infection. Therefore, the resident who has no acute upper respiratory infection during study period had being excluded in this study. In our opinion, the resident without respiratory infection related medical record exposed to low level air pollutants. It might under the estimated risk of dementia. Nevertheless, the data on air pollutants and dementia diagnoses were reliable.

Conclusions
Understanding the regional distribution of human health statuses can facilitate the investigation of the spread of diseases and the related risk factors as well as the assessment of medical resources and the planning of the use of these resources. In future research, animal studies can be conducted to further examine the association between air pollutants and neurological disorders.

Supporting Information
Table S1 Adjusted hazard ratio for dementia and dementia-associated risk factors.