Risk Grade Prediction Model For Breast Cancer-Related Lymphedema

BackgroundWith the increase of the incidence of breast cancer,breast cancer-related lymphedema,as one of the serious complications after breast cancer surgery,has attracted increasing attention.Because the pathological process of breast cancer-related lymphedema shows progressive development,its prevention and control should focus on early prevention.Therefore,it is particularly important to establish the risk factors of and a risk grade prediction model for breast cancer-related lymphedema,screening out patients with breast cancer-related lymphedema risk as early as possible,and providing individualized preventive measures according to different risk grade.In this study,a retrospective study was conducted to establish a risk level prediction model for breast cancer-related lymphedema,so as to provide references for early and individualized prevention of breast cancer-related lymphedema,improve the cost effectiveness of early prevention and reduce the incidence rate.Objectives1.Describe the occurrence of BCRL in breast cancer patients 3 years after surgery,and analyze the risk factors of BCRL.2.Construct the risk prediction model A of breast cancer-related lymphedema and draw the nomogram of model A.Then the risk gradeprediction model B for breast cancer-related lymphedema was developed,which provided a basis for screening breast cancer-related lymphedema risk patients and distinguishing different risk grades,so as to provide a basis for the early implementation of targeted and individualized preventive measures.Methods1.A questionnaire on risk factors of breast cancer-related lymphedema was designed based on relevant literatures and pilot investigation.Patients from a top three general hospital who underwent modified breast cancer between June 2013 and December 2015 were retrospectively investigated.The relevant data of patients from June 2013 to December 2018 were collected and the occurrence of BCRL was counted.2.Univariate Cox analysis was used to determine the potential risk factorsof breastcancer-relatedlymphedema(P<0.1),andstepwise multivariate Cox regression was performed to establish the independent risk factors and risk prediction model A of breast cancer-related lymphedema(P<0.05).Software R 3.0 was used to draw the nomogram of model A,we calculate the comprehensive score and risk probability of each patient's risk of breast cancer-related lymphedema.The risk scores were divided into different grade according to the optimal scale and the suggestions of clinical experts,and finally the risk grade model B for breast cancer-related lymphedema was developed.Harrell's c-index tests the discriminability of the model,and c-index>0.7 indicates that the discriminability is good.Hosmer-lemeshow(h-l)fitting test was used to test the calibration degree of the model,and the P>0.05 test was passed.Outcomes1.A total of 383 cases of breast cancer patients were included,71 cases developed breast cancer-related lymphedema.The cumulative incidence at 1,2,and 3 years after surgery was 7.57%,15.67%,and 18.53%,respectively.2.Risk factors for breast cancer-related lymphedema include weight gain(?2.5kg),hypertension,preoperative labor,number of positive lymph nodes(>10),number of lymph node dissection(>7),radiotherapy,blood sampling from affected limb,intravenous injection of affected limb,and knowledge of lymphedema.Independent risk factors for breast cancer-related lymphedema include radiotherapy,postoperative weight gain(?2.5kg),number of lymph node dissections((>7)),and knowledge of lymphedema.3.Risk prediction model A incorporate four independent risk factors,including radiotherapy,postoperative weight gain,number of lymph node dissection,and knowledge of lymphedema.The model met the equal proportion risk(x~2=5.46,P=0.24),had good model differentiation(c-index=0.77),and passed the calibration test(x~2=8.24,P=0.509).4.Risk prediction model B contained four predictors,including radiotherapy,postoperative weight gain,number of lymph node dissection,and knowledge of lymphedema.Risk grades were divided into three.The patient's comprehensive score was?160 is low-risk,the risk probability of BCRL after one year was less than 5%and less than 10%after 2-3 years;the patient's comprehensive score was 160-240 is moderate-risk,the risk probability of BCRL after one year was5%-25%and 10%-50%after 2–3years;the patient's comprehensive score was?240 is high-risk,the risk probability of BCRL after one year was?25%and?50%after 2–3 years.The model's c-index=0.74 and the calibration test results x~2=3.14,P=0.207.ConclusionsThe risk grade pridiction model B has a good degree of discrimination and calibration.It can help medical staff screen patients with breast cancer-related lymphedema risk and and distinguish different risk grades,providing a basis for early individualized prevention of breast cancer-related lymphedema,improving the cost effectiveness of early prevention and declining the incidence of breast cancer-related lymphedema.