ارائه روش باز نمونه گيري بوت استرپ در تحليل منحني ROC به منظور ارزيابي آزمون تشخيصي T4 و TSH سرمي در كم كاري تيروئيد نوزادان

Presentation Bootstrap Resampling Method in ROC Curve Analysis to Assessment Diagnostic Tests of TSH Serum in Neonatal Hypothyroidism

زﻣﯿﻨﻪ و ﻫﺪف: ﮐﻢﮐﺎري ﻣﺎدرزادي ﺗﯿﺮوﺋﯿﺪ از ﺷـﺎﯾﻊﺗﺮﯾﻦ ﺑﯿﻤﺎريﻫﺎي ﻏﺪد درونرﯾﺰ و ﻣﺘﺎﺑﻮﻟﯿﮏ ﮐﻮدﮐﺎن و از ﻣﻬﻢﺗﺮﯾﻦ ﻋﻠﻞ ﻋﻘﺐﻣﺎﻧﺪﮔﯽ ذﻫﻨﯽ ﻧﻮزادان اﺳــﺖ. ﮐﻢﮐﺎري اﯾﻦ ﻏﺪه ﻣﯽﺗﻮاﻧﺪ در ﻋﻠﻞ ﺛﺎﻧﻮﯾﻪ ﺑﻪ ﺷــﮑﻞ ﻧﻘﺺ ﮐﺎرﮐﺮد در ﻫﯿﭙﻮﻓﯿﺰ ﯾﺎ ﻫﯿﭙﻮـﺗﺎﻻﻣﻮس و ﮐﻤﺒﻮد TSHﺑـﺎﺷـــﺪ. ﻋﻼﺋﻢ ﺑـﺎﻟﯿﻨﯽ ﮐﻢﮐـﺎري ﺗﯿﺮوﺋﯿـﺪ در روزﻫـﺎي اول زـﻧﺪﮔﯽ ﮐﻢ و ﻏﯿﺮـﻗﺎـﺑﻞ ﺗﺸـــﺨﯿﺺ اﺳـﺖ؛ ﺑﻨﺎﺑﺮاﯾﻦ ﺗﻨﻬﺎ راه ﺗﺸـﺨﯿﺺ و درﻣﺎن ﻣﺆﺛﺮ اﻧﺠﺎم ﻏﺮﺑﺎﻟﮕﺮي ﻣﯽﺑﺎﺷـﺪ. در اﯾﻦ ﻣﻄﺎﻟﻌﻪ ﺑﺎ اﺳـﺘﻔﺎده از ﻣﻨﺤﻨﯽ ROCو روش ﺑﻮت اﺳﺘﺮپ، ﻗﺪرت ﺗﺸﺨﯿﺺ آزﻣﺎﯾﺶ ﻏﺮﺑﺎﻟﮕﺮي TSH و T4 ارزﯾﺎﺑﯽ ﺷﺪ. ﻣﻮاد و روشﻫﺎ: در اﯾﻦ ﻣﻄﺎﻟﻌﻪ، روش ﺑﻮت اﺳﺘﺮپ ﺑﺮاي ﺗﺮﺳﯿﻢ ﻣﻨﺤﻨﯽ ROCو ﺗﻌﯿﯿﻦ ﻧﻘﻄﻪ ﺑﺮش TSH و T4 ﺑﺮاﺳﺎس ﺷـﺎﺧﺺ Youden ﭘﯿﺸـﻨﻬﺎد ﺷـﺪ. اﺑﺘﺪا زﯾﺮﻧﻤﻮﻧﻪﻫﺎي ﻣﺨﺘﻠﻒ، ﺗﻮﻟﯿﺪ و ﺑﺮاي ﻫﺮ ﻧﻤﻮﻧﻪ ﻧﻘﻄﻪ ﺑﺮش و ﻧﻤﻮدار ROC ﻣﺤﺎﺳـﺒﻪ ﺷﺪ. روش ﺑﺎﻻ 500 ﺑﺎر ﺗﮑﺮار ﺷﺪ و در ﻫﺮ ﺗﮑﺮار ﻧﻘﻄﻪ ﺑﺮﺷﯽ اﻧﺘﺨﺎب ﺷﺪ ﮐﻪ داراي ﻣﺎﮐﺰﯾﻤﻢ ﺷﺎﺧﺺ Youden ﺑﻮد. ﺑﺮاي ﻫﺮ ﻧﻤﻮﻧﻪ ﻣﯿﺎﻧﮕﯿﻦ 500 ﺗﮑﺮار ﻧﻘﻄﻪ ﺑﺮش ﻣﺤﺎﺳـﺒﻪ ﺷـﺪ و ﻧﻘﻄﻪ ﺑﺮﺷـﯽ ﺑﺎ ﻣﯿﻨﯿﻤﻢ ﺧﻄﺎي دﺳـﺘﻪﺑﻨﺪي را داﺷـﺖ ﺑﻪﻋﻨﻮان ﻧﻘﻄﻪ ﺑﺮش ﺑﻬﯿﻨﻪ اﻧﺘﺨﺎب ﺷﺪ. ـﯾﺎﻓﺘـﻪﻫـﺎ: ﺑﺮاﺳـــﺎس ﻣﻨﺤﻨﯽ ROCﻧﻘﻄـﻪ ﺑﺮش ﺑﻬﯿﻨـﻪ ﺑﺮاي ﻣﺘﻐﯿﺮ TSHﺑﺮاﺑﺮ 10ﻣﯿﻠﯽ واﺣـﺪ در ﻟﯿﺘﺮ و ﺑﺮاي ﻣﺘﻐﯿﺮ T4 ﺑﺮاﺑﺮ 100 ﻣﯿﮑﺮوﮔﺮم در دﺳﯽﻟﯿﺘﺮ ﺑﻮد. ﺑﺎ اﺳﺘﻔﺎده از روش ﺑﻮت اﺳﺘﺮپ ﻣﻘﺪار دﻗﯿﻖ ﻧﻘﻄﻪ ﺑﺮش ﺑﺮاي ﻣﺘﻐﯿﺮ TSH، ﻣﻘﺪار 9.54 m/l و ﺑﺮاي ﻣﺘﻐﯿﺮ T4، ﻣﻘﺪار mg/dl 101.5 ﺑﻪدﺳﺖ آﻣﺪه ﺑﻮد. ﻧﺘﯿﺠﻪﮔﯿﺮي: ﻧﻘﻄﻪ ﺑﺮش ﺑﻬﯿﻨﻪاي ﮐﻪ از روش ﺑﻮت اﺳـﺘﺮپ ﺑﻪدﺳـﺖ آﻣﺪ ﺧﻄﺎي دﺳـﺘﻪﺑﻨﺪي ﮐﻤﺘﺮي ﻧﺴـﺒﺖ ﺑﻪ ﻣﻨﺤﻨﯽ ROC دارد. در ﺣﺠﻢ ﻧﻤﻮﻧﻪﻫﺎي ﮐﻮﭼﮏ ﻧﯿﺰ ﺧﻄﺎي دﺳـﺘﻪﺑﻨﺪي ﭘﺎﯾﯿﻨﯽ ﻧﺴـﺒﺖ ﺑﻪ ﻣﻨﺤﻨﯽ ROCﻧﺴـﺒﺖ ﺑﻪ ﺣﺠﻢ ﻧﻤﻮﻧﻪ ﺑﺰرگ ﻣﺸﺎﻫﺪه ﺷﺪ.

Congenital Hypothyroidism is one of the most prevalence disease of endocrinology and metabolic infant and important caused of mental deficiency. Hypothyroidism in the secondary causes represented by problem in pituitary or hypothalamus and shortage of TSH. Clinical symptom of Hypothyroidism at the first days of birth are indiscernible so only way of diagnosis and treatment of that, is screening. In this study we used the bootstrap method to evaluation the ROC curve of diagnostic screening TSH test. Materials and Methods: In this study a bootstrap method is proposed to plot ROC curve and determinate the cut point of TSH & T4 based on YOUDEN method. at the first generated subsample and estimated ROC and cut-off point for each sample. The procedure run for 500 replicate. For each replicate selected cut-off where have maximum YOUDEN. The cut-off point calculated for 500 replications from each sample and then selected mean of 500 replicate. The cut-off which have minimum misclassification selected optimal cut-off point. Results: Based ordinary method to estimate the ROC curve the optimal cut-off point for TSH variable is 10 mlU/L and for T4 variable is 100 μg/dL which based the bootstrap method respectively optimal cut-off point for TSH is 9.54 and for T4 is 100.07. Conclusion: The optimal cut-off point where obtained from bootstrap method have minimum misclassification rate versus ordinary method. Based the small sample bootstrap have minimum misclassification rate toward ordinary method with large sample.