遺傳與臨床因子預測台灣兒童氣喘之風險

Abstract

背景： 氣喘是目前最常見的兒童慢性疾病，其受到許多的基因因素與環境因素的影響。雖然，許多基因變異已經被確認與氣喘發生及臨床表現極為相關，但是至今尚未有一個合併基因因素與臨床因素的危險評估模式來衡量疾病發生的可能性。 方法： 我們從3個台灣的世代研究(TCHSI，GBCA，與TCCAS)的個案中，分析8個與氣喘相關的基因，分析了10個單核甘酸多形性 (SNPs)。所有參與者的父母親都接受訪談，並完成與氣喘危險因子相關的問卷調查。利用改良式的氣喘預估指標(mAPI)，以及由10 SNPs構成的基因危險分數(GRS)建構成一個預測模式，來評估與氣喘的相關性。利用模式區別能力(discrimination)與模式再分類能力(reclassification)的統計方法，以及模式校對法(calibration)來衡量預測模式的預測能力。再進行leave-one-out 交叉驗證法，驗證預測模式。 結果： 我們一共收集了640個氣喘兒童與1921對照組兒童。經由羅吉斯迴歸分析調整了年齡，性別，及身體質量指數(BMI)後，兩個環境因素，包括潮溼與牆壁受潮，與兒童氣喘有相關，這些因素被納入基本模式(baseline model)作分析。由基本模式合併改良式的氣喘預估指標(mAPI)與基因危險分數(GRS)組成綜合模式(combined model)，其氣喘預測能力，比由基本模式合併改良式的氣喘預估指標(mAPI)組成的臨床模式，有較好的預測能力，可見到曲線下面積(AUC)由0.748增加到0.780 (p<0.0001)。同時也發現，綜合模式對疾病的區分能力(IDI: 0.022, p<0.0001)與再分類的能力(continuous NRI: 0.344, p<0.0001)也都達到統計學上的意義。考量兒童氣喘的危險機率為18%，綜合模式對氣喘的預測能力能達到78.8%的敏感性，59.4%的特異性，39.5%陽性預測值，與89.3%的陰性預測值。 結論： 我們成功的建構一個由基因因素與臨床因素構成的預測模式衡量兒童氣喘的危險評估模式。我們發現在臨床因子為主的預測模式加上基因危險分數(GRS)能有效的提升兒童氣喘發生率的預測能力。

Background Asthma is the single most common chronic childhood disease affected by multiple genetic and environmental factors. Although many genetic variants have been noted to be associated with childhood asthma, up to the present, no risk assessment models that incorporate genetic and clinical predictors for asthma occurrence are currently available. Methods We analyzed 10 single nucleotide polymorphisms (SNPs) identified from 8 asthma-associated genes among subjects who participated in three children’s cohort studies in Taiwan (TCHS, GBCA and TCCAS). The parents of all participants were interviewed regarding the information of asthma risk factors. Modified Asthma Predictive Index (mAPI) and genetic risk scores (GRSs) of 10 SNPs were used in prediction model assessment. The performance of prediction ability was assessed by discrimination and reclassification statistics, and calibration. Cross validation with leave-one-out cross validation method was also conducted. Results In total, 640 asthma cases and 1921 control subjects were included in current study. After controlling for age, sex and BMI, two environmental factors, moist and water damage of wall, were significantly associated with childhood asthma and all these covariates were incorporated into baseline model. The prediction ability of childhood asthma improved greatly in combined model composed of GRSs and mAPI compared with clinical model, demonstrating an increased AUC from 0.748 to 0.780 (p<0.0001). Besides, the improvement in discrimination (IDI: 0.022, p<0.0001) and reclassification (continuous NRI: 0.344, p<0.0001) also showed while comparing clinical model and combined model. Considering the 18% general risk of asthma in children, the combined model concluded sensitivity, specificity, positive predictive value, and negative predictive value as 78.8%, 59.4%, 39.5%, and 89.3%, respectively. Conclusion We have successfully constructed a prediction model composed of genetic and clinical factors in asthma risk assessment. Adding GRS into clinical factor-based prediction model would increase the prediction ability substantially in predicting the occurrence of childhood asthma.