高雄市2015年登革熱流行特徵與2016年流行後社區血清流行病學研究

Epidemiological Characteristics of Dengue Epidemics in 2015 in Kaohsiung and Post-Epidemic Community-based Seroepidemiological Studies in 2016

陳柏堯(Po-Yao Chen)

指導教授 ： 金傳春

共同指導教授 ： 方啟泰(Chi-Tai Fang)

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803739

摘要

南台灣的高雄市在2014與2015年分別由第一型與第二型登革病毒所導致史無前例的兩波登革熱大流行[各有14999與19723確定病例(27與158死亡)]；然而2015年流行幅度更大而嚴重之開始失守因素為何？又欲知未來的流行趨勢必須先掌握抗體之分布。因此，本研究的主要目的有二：（一）描述2015年登革熱大流行的重要時空流行病學特徵，並探討此流行前期的危險因子（如人口密度、空屋率與病媒蚊列管點等）及（二）在此流行後以血清流行病學明瞭高雄某高危險區各年齡層的抗登革病毒IgG血清盛行率與IgM血清發生率，並找出IgG抗體陽性者的相關危險/保護因子，藉以提供未來社區衛生教育的重點方向及在登革熱流行「前期」的最佳防疫策略。
本研究有兩部分：(一)自2015年7月12日至2016年2月9日，流行共31週，依流行曲線而定義為三時期: (1)流行第1至第11週為「流行前期」（每週個案數緩升階段），(2)第12至第19週為「流行中期」（每週個案數陡升至高峰），及(3)第20至31週為「流行晚期」（每週個案數由高峰降至零）；先探討確定病例的個案年齡與性別比例隨流行時間的變化，續分析登革熱個案的空間分布，並了解登革熱死亡的危險因素，最後再探察登革熱流行前期的因素與時空分析。為探討此流行前期的危險因子，病媒蚊列管點資料(蚊蟲陽性水溝、髒亂空屋、地下室積水)由高雄市衛生局提供，用電量採台灣電力公司數據；在流行前期的舊高雄市與鳳山區共528里，先以皮爾森相關係數（Pearson’s correlation coefficient）找出高雄市2015年以里為單位的登革熱發生率(Li-specific incidence rate)與各解釋變項之間的關係，再用地理加權迴歸（Geographically Weighted Regression, GWR）模型，探討在2015年登革熱流行前期不同解釋變項在不同空間的危險性，並與簡單線性回歸(ordinary least squares, OLS)模型比較GWR地理相關性所增加的解釋度。(二)在2016年12月下旬對高雄市鳳山區的一所國小（467位）、國中（653位）、高中（618位）之共1738位學童及社區共123位成人與老人，進行抗登革病毒IgG抗體的血清流行病學探究。實驗法採用DENV-IgG DxSelect™試劑，初篩血清內的抗登革抗體，為了避免抗日本腦炎套膜蛋白（envelope protein, E）與登革病毒套膜蛋白的交叉反應而可能造成偽陽性，再用台灣疾管署開發的間接酵素連結免疫吸附法(indirect ELISA)，同時測試抗登革病毒和抗日本腦炎病毒之非結構性第一蛋白(non-structural protein 1, NS1) IgG抗體，若anti-DENV/anti-JEV NS1 IgG OD比值 > 2代表為登革病毒感染，因此可排除抗日本腦炎病毒抗體造成的偽陽性。再對此IgG陽性者，以DENV DetectTM IgM Capture ELISA (InBiOS)試劑，測量抗登革病毒IgM抗體，以得知2016年登革病毒感染的血清發生率(seroincidence)。據實驗所得的登革病毒IgG抗體陽性與陰性者，進行相關因子的單變項分析，並進一步將其統計顯著的因子進行多變項分析。
2015年登革熱大流行的流行病學特徵上，在不同流行期的年齡分布，流行前六週主由21-35歲者較多(25.1% 65/259)；中期由51-65歲者較高(26.9%, 3254/12075)；而流行最後五週又轉為21-35歲年齡者居多(23.4%, 37/158)；在性別上，前期與晚期以男性居多，中期以女性較多。在登革熱致死率的分析，發現年齡與性別為死亡的危險因子，年齡愈大的死亡風險愈高(p<0.001)，且男性死亡風險為女性的1.62倍(p = 0.0036)。空間分析發現第一週病例集中於左營區，且自第三週始病例往周圍擴散，且前3週的登革熱個案集中於主要交通路線，自第4週後開始社區傳染；相關係數分析顯示在流行前期528里的登革熱發生率和人口密度無顯著相關，但在中、晚期登革熱病例數多時，其與人口密度有具統計意義的低度正相關[相關係數(ρ)=0.163, p<0.001和0.11, p<0.01]，同時在流行前期登革熱發生率與用電量極低(似空屋)的住宅戶數率有正相關（ρ=0.107, p<0.05），並與病媒蚊列管點的髒亂空屋數(ρ=0.185, p<0.001)、蚊蟲陽性水溝數(ρ=0.166, p<0.001)和地下室積水數(ρ=0.234, p<0.001)均具統計意義的正相關；尤其2015年七月中高雄開始流行第二型登革病毒的頭三週有病例處的前兩週，可見個案發生地靠近過去兩週的蚊蟲陽性水溝。在統計模型中GWR模型的解釋力明顯地比OLS模型高，且顯示用電量低的住戶率勝算比（OR）最大為1.10，主分布於舊高雄市北部；髒亂空屋數之最大OR為9.76分布在前鎮區；蚊蟲陽性水溝數最大OR為1.83位於苓雅區與旗津區，此三因素的地理風險分布不同。
血清流行病學調查，顯示高雄市鳳山區的6-19歲學童與23-80歲社區成人的抗登革病毒-IgG抗體之血清盛行率各為3.05% (53/1738)與33.33% (41/123)，又此抗體在學童的年級間沒有統計差異，但社區成人血清盛行率隨年齡層增而有具統計意義的上升(p = 0.0015)。單變項分析發現與血清登革病毒-IgG抗體陽性勝算比較高而具統計顯著的危險因子有：(1)學童過去有印尼旅遊(OR = 5.9, p<0.001)，(2)廚房有蚊子者較室外庭院為高（OR=3.33, p=0.01）；另具統計顯著的保護因子有：(1)學童母親的學歷越高（高中:OR = 0.32, p<0.01；大學:OR = 0.25, p<0.01），(2)清除室外積水容器的頻率愈高 (偶爾清除:OR = 0.57, 時常清除: OR = 0.16, p = 0.068)。另發現有62% (31/51)的學童與60% (17/28)的社區成人並不知道自己過去曾經得過登革熱。多變項分析，發現廚房有蚊蟲(OR=2.82)與赴東南亞旅行(OR=2.1)為登革病毒IgG抗體陽性的危險因子；而母親高教育程度(OR=0.48)為保護因子。至於抗登革病毒-IgM抗體之血清發生率在學童與社區成人兩研究群均為0%。
本研究提供登革熱發病、死亡與人群感染的個人與群體之科學實證危險因子，以提供未來防疫的策略建議，包括：（1）病媒蚊列管點調查如髒亂空地、病媒蚊孳生之水溝可作為未來先行防治的參考指標；（2）藉由衛教來提升民眾對於登革熱的認知度，並提升他們對居家環境的滅蚊與防蚊行動力；（3）加強在登革熱流行季時對旅遊業者的登革熱講習與來自東南亞旅客的偵測。未來研究仍待努力處為：（1）針對抗登革病毒IgG陽性受試者做病毒分型試驗，以了解可能感染的型別；（2）應有較完整與連續的蚊蟲密度調查，以提供建模參數更妥當的預測。

關鍵字

登革熱 ； 血清盛行率 ； 時空趨勢 ； 病媒蚊列管 ； 地理加權迴歸 ； 公共衛生政策 ； 台灣

並列摘要

The two unprecedented largest epidemics of dengue caused by dengue virus serotype 1 and 2 (DENV-1, 2) in 2014 and 2015, respectively occurred in Kaohsiung of Taiwan. The aims of this study were: (1) to describe the temporally and spatially epidemiological characteristics of the 2015 epidemic, (2) to investigate the risk factors of fatal dengue cases, (3) to find out factors involved in the early stage of this epidemic; (4) to measure the age-specific seroprevalence and seroincidence rates of DENV-IgG and IgM antibodies, and (5) to search for the risk/protective factors of those with seropositive for providing scientific-based public health recommendations.
This study had two parts. Part (I) the 2015 dengue epidemic in Kaohsiung started form July 12, 2015 [1st DENV-2 case confirmed] to February 9, 2016 covering 31 weeks (wks). The 3 stages of the epidemic were: (1) 1st to 11th wk as the early stage, (2)12th to 19th wk as the middle stage and (3) 20th to 31st wk as the late stage. Temporal distributions of the confirmed cases’ age and gender at different epidemic stages and their spatial distributions were analyzed. Using the 528 Lis with dengue cases at the early epidemic stage, and information on mosquito-monitoring sites from Kaohsiung Dept. of Health, and electricity usage data from Taiwan Electricity Power Co., Pearson’s correlation coefficient was used to screen the associations among case numbers, incidence rates and risk factors. Additionally, geographically weighted regression (GWR) was used to analyze the effects of spatial closeness. Then, each factor was evaluated by both GWR and OLS regression. Part (II) seroepidemiological studies in 1738 schoolchildren and 123 community adults with their blood samples collected in late Dec., 2016 were conducted to test for anti-DENV-IgG. The seropositive ones were verified by ELISA testing for anti-DENV NS1 IgG and anti- JEV-NS1 IgG simultaneously (to avoid cross-reaction by anti-E). Then, those anti-DENV-IgG seropositive were compared with seronegative ones for univariate and multivariate analyses on subject’s travel history, past clinical symptoms and mosquito breeding environment. All the anti-DENV-IgG seropositive serum samples were detected for anti-DENV IgM.
In epidemiological characteristics, age distributions varied in different stages of the epidemic [21-35 years (25.1% 65/259), 51-65 years (26.9%, 3254/12075), and 21-35 years (23.4%, 37/158) accounted most for the first 6 wks of the early, middle, and last 5 wks of late stages, respectively]. In addition, males were higher in the early and late stages. lder age had higher fatality rate of dengue (p<0.001) and males had 1.62-fold risk of dengue deaths (p=0.0036). Spatial analysis showed that dengue cases cumulated in Zuoying District at 1st week (wk); then started spreading to other districts after the 3rd wk. Meanwhile, dengue cases in the first 3 wks were close to the main transportation lines and stated to spread in communities after the 4th wk. In correlation analyses, incidence of dengue did not have correlation with population density at the early stage but such correlations were present at the middle and late stages [coefficient (ρ) = 0.163, p<0.001; ρ = 0.11, p<0.01]. Evaluation of mosquito-monitoring sites, dengue incidence at the early epidemic stage had positive correlations with proportion of empty households with low electricity consumption (LEC, ρ=0.107, p<0.05), numbers of dirty and vacant houses (DVH, ρ=0.185, p<0.001), numbers of ditches with mosquitoes (DWM, ρ=0.166, p<0.001), and numbers of basement with standing water (BSW, ρ=0.234, p<0.001). Especially, dengue cases were close to mosquito-breeding ditches at the first 2 wks. GWR model demonstrated better explanation percentage than the OLS method and also identified geographical variations in different mosquito monitoring sites [OR of 9.76 for DVH in Cianjhen District, OR of 1.83 for DWM in LinYa and Chijin Districts].
The seroepidemiological study revealed that the seroprevalence rates of anti-DENV-IgG in 6-19-year-old schoolchildren and 23-80-year-old community adults at Fengshan district were 3.05% (53/1738) and 33.33% (41/123). Such rates increased significantly with age (p=0.0015). Univariate analyses identified both significant risk factors [travel to Indonesia (OR=5.9, p<0.001) and mosquitoes appearing in kitchen (OR=3.33, p=0.01)] and significant protective factors [higher mother’s education levels (OR: senior high = 0.32, p<0.01; college = 0.25, p<0.01), and higher frequency of cleaning water containers (OR: occasionally = 0.57, frequently = 0.16, p=0.068)]. Multivariate analysis identified kitchen with mosquitoes (OR=2.82), traveling to SE Asia (OR=2.1) were risk factors whereas mother’s higher education levels was protective factor (OR=0.48). Furthermore, 62% (31/51) of schoolchildren and 60% (17/28) of community adults with anti-DENV-IgG positive did not know they had ever been infected with DENV. Additionally, the seroincidence rates of anti-DENV-IgM in both schoolchildren and community adults were 0%.
In summary, this study provided scientific evidences on risk factors associated with dengue cases, deaths and DENV infection at both individual and population levels. Future prevention strategies include: (1) continuous monitoring on mosquito indices; (2) promoting residents’ knowledge on dengue and their action in preventing and controlling mosquitoes; (3) providing workshops on dengue for travel agencies and enhancing surveillance for travelers coming from Southeast Asia. Future research efforts need to target on serotyping of anti-DENV IgG-positive serum samples and etting up a more comprehensive mosquito density data for better prediction.

並列關鍵字

Dengue ； seroprevalence ； temporal/spatial trends ； mosquito monitoring sites ； GWR ； public health policies ； Taiwan

參考文獻

1. WHO. Dengue and severe dengue (Fact sheet N°117). World Health Organization, Geneva, Switzerland, 2018. (2018) Available from: http://www.who.int/en/news-room/fact-sheets/detail/dengue-and-severe-dengue

2. Taiwan CDC. Government Website Open Information Announcement. Daily laboratory-confirmed dengue cases since 1998. (2018) Available from:

https://data.gov.tw/dataset/73178

3. Stanaway JD et el. The Global Burden of Dengue: an analysis from the Global Burden of Disease Study 2013. Lancet Infectious Disease 16(6): 712–723. (2016)

4. Kyle JL and Harris E. Global spread and persistence of dengue. Annual Review of Microbiology 62:71-92. (2008)