The Establishment And Evaluation Of Scoring System For Predicting The Risk Of Breast Cancer-related Lymphedema

Objectives To screen the predictor factors of breast cancer-related lymphedema andthen establish a scoring system for predicting the risk of breast cancer-relatedlymphedema.To evaluate the effectiveness of the scoring system and find the optimalthreshold value to determine high-risk groups of lymphedema.Methods The Social Demographic Information(such as age, height et al),DiseaseInformation(such as tumor staging, histology et al),Treatment Information(such as typeof surgery, level of axillary lymph node dissection et al),Behavior Information(such asfunction of upper limb, Lymphedema Risk-Reduction Behavior et al) of breast cancerpatients who underwent breast-conserving surgery/mastectomy with axillary lymphnode dissection（ALND）were analyzed retrospectively from one breast disease center inHefei Anhui from Sep.2011to Dec.2012. Lymphedema was defined by circumferentialmeasurement.By comparing the data between lymphedema groups andnon-lymphedema, the variables with statistically significant difference between twogroups were selected out for further multiple stepwise logistic regression, Logistic scoresystem and Additive score system were then established based on regressioncoefficients and OR values of variables which selected into regression model.Receiveroperating characteristic curve (ROC) was plotted to evaluate the predictive efficiency ofthe score system and find the optimal threshold value to determine high-risk groups oflymphedema. Results A total of346breast cancer patients were eligible for analysis. By comparisontwo groups, the statistical difference was found in location of tumor, incision type,axillary lymph node status, level of ALND, number of axillary lymph nodes removed,neo-adjuvant chemotherapy, neo-adjuvant chemotherapy in ipsilateral arm, radiotherapy,hormonal therapy, Lymphedema Risk-Reduction Behavior:“don’t ignore lymphedema,avoid strenuous exercise, don’t lift heavy, avoid injury, avoid damage when cuttingnails,wearing gloves when doing housework, keep reasonable weight”, function ofipsilateral arm (P<0.05). Those candidates were selected out for further multiplestepwise logistic regression.Finally, location of tumor, incision type, level of ALND,radiotherapy, Lymphedema Risk-Reduction Behavior:“don’t ignore lymphedema, avoidstrenuous exercise, avoid injury” were selected into regression model. Logistic scoresystem and Additive score system were then established based on regressioncoefficients and OR values of variables which selected into regression model. ForLogistic score system: ROC curve showed a sensitivity of84.7%and specificity of70.5%in the ability to predict the risk of breast cancer-related lymphedema at a cut-offof35points, and the area under the curve (AUC) was0.836(P<0.001，95%CI：0.784～0.888). For Additive score system: ROC curve showed a sensitivity of87.0%andspecificity of68.2%in the ability to predict the risk of breast cancer-relatedlymphedema at a cut-off of35points, and the area under the curve (AUC) was0.834(P<0.001，95%CI：0.782～0.886).Conclusions The scoring system can be used to predict the risk of breast cancer-relatedlymphedema and screen high-risk groups of lymphedema.