探討乳癌手術病患手術前後生活品質與腫瘤分類預測模式之研究

研究背景與目的: 乳癌是目前全球女性最高發生率的癌症，乳癌每年造成全球46萬名婦女死亡，對台灣婦女的健康更遭受到嚴重的威脅，依據行政院衛生署資料，2010年癌症死因統計，計有8,136人被診斷罹患為乳癌，因女性乳癌死亡者有1,706人，該年標準化死亡率為十萬人口之11.0%，在女性癌症死亡原因位居第四，標準發生率也有逐年升高的趨勢。透過正規治療乳癌5年整體存活率可高達85%。因此，其健康生活品質狀況開始受到關注，研究如何增進癌症患者治療後的生活品質是一項不可或缺的議題。本研究目的如下: 目的一、探討乳癌病患手術前後健康相關生活品質(HRQoL)之長期改變趨勢。目的二、探討乳癌疾病別與整體性測量工具評估HRQoL之差異性。目的三、探討影響乳癌手術病患術後HRQoL改變之相關影響因素。目的四、比較類神經網路(ANN)與複迴歸分析(LR)對於預測乳癌腫瘤分類之準確性。研究方法: 1.本研究設計屬於前瞻性縱貫性研究，取樣選擇南部兩家醫學中心，進行乳癌手術前面談與手術後電話訪談病患其健康相關生活品質，共四個間點(術前、術後6個月、術後一年、術後兩年)。而乳癌病患特性、醫療機構特性及醫療照護品質特性等訊息的獲得經由術後六個月之病歷審查(Chart Review)，整體而言，本研究樣本有366位，滿足四個時間點之追蹤總樣本數為275位。以廣義估計方程式(Generalized estimating equation, GEE)控制研究時間點變項，探討乳癌病患手術前後間的健康相關生活品質的長期改變趨勢及影響乳癌病患術後生活品質改變之相關重要因素。 2.以病歷審查方式獲得670位良性腫瘤個案與275位惡性腫瘤個案之病患特性資料，以總樣本數共945位，分為70%做訓練樣本(Training)、20%做測試樣本(Testing)、10%做為驗證樣本(Validation)，並建立乳癌病患特性變項之腫瘤分類預測模式，比較複迴歸與類神經網路模式對於預測腫瘤分類之準確性。研究結果: 1.手術前後病患的健康相關生活品質的長期改變趨勢分析: 以術前分數為參考基準，術後六個月分數低於術前，大部分構面分數皆有顯著性下降的趨勢(P<0.05)；以術後六個月分數為參考基準，術後一年分數高於術後六個月，大部分構面分數皆有顯著性上升的趨勢(P<0.05)；而術後一年與術後兩年的分數比較，大部分構面沒有顯著差異。 2.疾病別QLQ-C30與整體性測量工具SF-36評估生活品質之差異性: 結果顯示出在QLQ-C30量表在功能與症狀領域之地板與天花板效應在手術介入後生活品質分數的改變，QLQ-C30相較SF-36能更能反應病患術後所獲得生活品質分數改善之人數增加的趨勢。 3.影響乳癌病患術後生活品質改變之相關影響因素: 研究顯示病患特性之手術年齡、是否停經、有無抽菸、先前有無乳房疾病、有無乳癌家族病史、病理分化、病理分期，和醫療機構特性之手術方式、ASA 分數、有無化學治療、有無賀爾蒙治療、有無放射治療，和醫療照護品質特性之術後住院天數、總住院天數、30日內有無再入院，以及術前功能評估生活品質分數等等因素皆會顯著影響病患術後生活品質(P<0.05)。 4.類神經網路模式與複迴歸模式對於預測腫瘤分類之準確性比較: 類神經網路模式的敏感度(68.59% vs. 31.94%)、陰性預測值(87.26% vs. 76.87%)、準確率(81.87% vs. 74.47%)、AUC曲線的預測值(85.67% vs. 61.83%)，皆比複迴歸模式來得好、預測也較準確，所以類神經網路預測模式與複迴歸分析的比較，類神經網路對於預測腫瘤分類之準確性略勝一籌。結論與建議: 本研究乳癌手術病患健康相關生活品質的改變趨勢，提供病患與其家屬作為手術後的心理調適與應對，減少手術前的恐懼與了解手術後會遭遇的情況，提早做準備，其改變趨勢可讓衛教政策作為參考依據，對於醫療資源能夠依據不同的癌症患者做有效應用並針對術後衛教及其復健認知，有效提升醫療照護品質，也提供給未來研究癌症健康相關生活品質方面的議題做參考。另外，依據本研究預測模式類神經網路有良好的準確性，可與醫學影像系統與醫學技術結合，提供醫療人員腫瘤分類之參考依據，加以互補。

關鍵字

乳癌；健康相關生活品質；類神經網路；複迴歸

並列摘要

Background & Purpose: Breast cancer is the highest incidence in femal in Taiwan and causes 460,000 deaths around the world. According to the Department of Health data in 2010, there were 8,136 people been diagnosed with breast cancer, including 1,706 deaths, and the standardized mortality rate was 11.0 of the 100,000 population. And 5-year overall survival through the treatment of breast cancer can be as high as 85%. Therefore, this study examined longitudinal change in each subscale of a generic QOL scale for cancer patients and a breast cancer-specific QOL scale and explored their relationships to effective predictors of QOL outcomes in breast cancer surgery patients. The purpose of this study as follows: I.To evaluate the long-term change trend of health-related quality of life (HRQoL) in breast cancer patients before and after surgery. II.To compare the effectiveness of the breast cancer disease and general questionnaires. III.To explore the related factors of HRQoL for breast cancer surgical patients. IV.To compare the accuracy of the Artificial Neural Network (ANN) and Multiple Regression Analysis (LR) for predicting breast tumors. Methods: This study design is a prospective and longitudinal study, there are 275 breast cancer surgical patients from two medical centers in Southern Taiwan , and competed the health-related life quality of questionnaires at preoperative, postoperative 6 months, one year after surgery, two years after surgery. To obtain the characteristics of breast cancer patients, the characteristics of medical institutions and the characteristics of medical health care quality after postoperative 6 months by Chart Review. To explore the long-term change trend of health-related quality of life (HRQoL) in breast cancer patients before and after surgery and the related factors of HRQoL change for breast cancer surgical patients by Generalized estimating equation control variables of the time. There are 670 benign cases and 275 malignant cases by Chart Review. The total 945 patients to divided into 70% of the training samples (training groups), 20% of test samples (testing groups), 10% of the validation sample (Validation).Then compare the accuracy of the Artificial Neural Network (ANN) and Multiple Regression Analysis (LR) for predicting breast tumors. Results: I.The scores of Preoperative as the reference standard, the scores of postoperative six months lower than the preoperative, most of dimensions are significantly (P <0.05); The scores of postoperative six months as the reference standard, the scores of postoperative a year higher than the scores of postoperative six months, most of dimensions are significantly (P <0.05);To compare the scores of postoperative 1 year and the scores of postoperative 2 year, most of dimensions are nonsignificantly (P <0.05). II.The breast cancer disease and general questionnaire tool to assess HRQoL differences, the results show the floor and ceiling effects in function and symptoms, the QLQ-C30 is better response than SF-36. III.The related factors of HRQoL change for breast cancer surgical patients ,the results show that patient characteristics,like age,menopause, smoking, breast disease history,family history of breast cancer, stage of breast cancer and so on, the characteristics of medical institutions, like ASA score, chemotherapy, hormonal therapy, radiation therapy, and characteristics of medical health care quality,like length of stay, 30 days with or without re-admission, and the functional of preoperative,these factors are significantly (P <0.05). IV.To compare the accuracy of the Artificial Neural Network (ANN) and Multiple Regression Analysis (LR) for predicting breast tumors, the results show that sensitivity of the Artificial neural network higher than LR(68.59% vs. 31.94%),the negative predictive value higher than LR (87.26% vs. 76.87%), accuracy (81.87% vs. 74.47%), the AUC curve (85.67% vs. 61.83%) are better than the LR, Artificial neural networks is better to predict breast tumor. Conclusions: When evaluating HRQoL after breast cancer surgery, several factors other than the surgery itself should be considered. Patients should also be advised that their postoperative HRQoL might depend not only on the success of their operations, but also on their preoperative functional status. In addition, based on Artificial neural network of this study have good accuracy, and combined with medical imaging systems and medical technology to provide the reference of the breast tumors.