空氣污染改善對第二型糖尿病預防之效益：遺傳易感性的修飾作用

Abstract

研究目的：本研究旨在探討空氣污染物年變化與第二型糖尿病（Type 2 diabetes, T2D）發生風險之關聯，並進一步評估血糖相關遺傳易感性是否修飾此關聯。 研究方法：本研究採前瞻性世代研究設計，使用臺灣生物資料庫（Taiwan Biobank）資料，共納入基線時無第二型糖尿病且完成追蹤之27,452位受試者。第二型糖尿病事件定義為追蹤期間自述醫師診斷糖尿病、空腹血糖（fasting plasma glucose）≥126 mg/dL或糖化血色素（HbA1c）≥6.5%。依據受試者居住地，以機器學習土地利用迴歸模式（machine learning-based land-use regression model）估算PM2.5、CO、NO2、SO2及O3之年度平均暴露濃度，並將污染物年變化作為時間相依暴露（time-varying exposure）納入分析。空腹血糖遺傳分數（fasting glucose genetic score, FGS）及HbA1c多基因風險分數（polygenic risk score, PRS）採用clumping and thresholding方法建立。利用時間相依Cox比例風險模型（time-varying Cox proportional hazards model）估計危險比（hazard ratio, HR）及95%信賴區間（confidence interval, CI），並校正人口學特徵、生活型態及心血管代謝相關共變項。 研究結果：追蹤期間共發生1,281例第二型糖尿病。PM2.5、CO、NO2及O3年濃度下降與第二型糖尿病風險降低具有顯著關聯，而SO2則未達統計顯著。於雙污染物模式中，PM2.5與第二型糖尿病之關聯仍具穩定性。HbA1c PRS與第二型糖尿病呈現明顯劑量反應關係，而FGS則呈現較弱之正向關聯。空氣污染物年變化與遺傳易感性間之乘法交互作用皆未達統計顯著；然而，限制立方樣條（restricted cubic spline, RCS）分析顯示，不同HbA1c PRS族群具有相似之暴露－反應趨勢，高遺傳風險族群於整個PM2.5改善範圍內皆維持較高之第二型糖尿病估計風險。 研究結論：空氣污染濃度逐年改善，尤其PM2.5濃度下降，與第二型糖尿病風險降低具有顯著關聯，且此關聯於不同遺傳易感性族群中大致一致。本研究支持改善空氣品質作為第二型糖尿病族群預防策略之潛力，並顯示結合華人族群多基因風險分數，可望提升第二型糖尿病之精準預防與風險分層能力。 關鍵詞: 第二型糖尿病、空氣污染、PM₂.₅、多基因風險分數、遺傳易感性、臺灣生物資料庫、時間相依Cox比例風險模型

Objective: This study aimed to investigate the associations between annual changes in ambient air pollutant exposures and incident T2D and to examine whether glycemic genetic susceptibility modifies these associations in a Taiwanese population. Methods: We conducted a prospective cohort study using data from the Taiwan Biobank. A total of 27,452 participants free of T2D at baseline and with follow-up data were included. Incident T2D was defined by self-reported physician diagnosis, fasting plasma glucose ≥126 mg/dL, or HbA1c ≥6.5% during follow-up. Annual average exposures to PM2.5, CO, NO2, SO2, and O3 were estimated using machine learning-based land-use regression models according to participants’ residential locations. Annual changes in air pollutant concentrations were incorporated as time-varying exposures. Polygenic risk scores (PRSs) for fasting glucose (FGS) and HbA1c were constructed using the clumping and thresholding approach. Time-varying Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs), adjusting for demographic characteristics, lifestyle factors, and cardiometabolic comorbidities. Results: A total of 1,281 participants developed incident T2D. Greater annual reductions in PM2.5, CO, NO2, and O3 concentrations were associated with lower risks of incident T2D, whereas no significant association was observed for SO2. The association for PM2.5 remained robust after adjustment for co-pollutants. HbA1c PRS demonstrated a clear dose–response relationship with incident T2D, whereas FGS showed a comparatively weaker positive association. No statistically significant multiplicative interactions were observed between annual changes in PM2.5 exposure and glycemic genetic susceptibility. However, restricted cubic spline analyses demonstrated similar exposure–response patterns across HbA1c PRS quartiles, with higher genetic risk groups consistently exhibiting higher estimated hazard ratios throughout the exposure range. Conclusions: Greater annual reductions in ambient air pollution, particularly PM2.5, were associated with lower risks of incident T2D. These associations were generally consistent across different levels of genetic susceptibility. Our findings support the potential benefits of improving ambient air quality as a population-level strategy for T2D prevention and suggest that integrating ancestry-specific genetic information into precision prevention and risk stratification may further reduce the future burden of T2D. Keywords: Type 2 diabetes; Air pollution; PM2.5; Polygenic risk score; Genetic susceptibility; Taiwan Biobank; Time-varying Cox model.