| Background and objective: Chronic inflammatory diseases of the non-lactating breasts cause considerable difficulty in diagnosis and treatment.There is a spectrum of aetiological factors ranging from infection to autoimmune disorders.The disease causes considerable morbidity and psychological distress in relatively young females.Non-lactating mastitis predominantly occurs in premenopausal women and the clinical symptoms have often been misjudged as breast cancer.So the differential diagnosis of non-lactating mastitis and breast cancer is of great significance in clinical practice.The study aimed to collect and analyze a wide spectrum of non-lactating mastitis and breast cancer diseases and establish a diagnostic modal to predict them based on the binary Logistic regression which was evaluated by calculating the area under the ROC curve.Methods: The study included 68 female breast disease patients with histological confirmation from our hospital.The patients were categorized as NLM group and BC group based on histological findings.There are 30 non-lactating mastitis patients in the NLM group and 38 breast cancer cases in the BC group.Retrospective analysis was conducted including the general details of patients,clinical feature and ultrasonic findings.First of all,we chose some possible factors to study by the single factor analysis in statistics,including age,place of residence,smoking history,the number of reproduction and abortion,lesion distribution,lesion pain,swelling,nipple discharge,nipple retraction,whether if the axillary lymph nodes touched,lesion diameter,ascularization,and the mobility,calcification,shape and boundary of the nidus.As independent variables,statistically significant factors were bring into Logistic regression equation to establish differential diagnosis model and calculate the probability predictive value P.Then the optimal cut-off point of probability predictive value was obtained by choosing MedCalc software to make ROC curve.Including Sensitivity,Specificity,Position likelihood ratio,Negative likelihood ratio and Yuden index,these indicators based on the cut-off point can evaluate the diagnosis veracity of the modal.Results: For the single factor analysis,patients whose mean age was 38.5±12.1 in the NLM group were younger than ones whose mean age was 55.5±11.7 yeas old in the BC group(t=1.137,p<0.001).It was significance of difference in statistics between two groups that for BC group 31 cases(31/38)live in the city but only 14 people(14/30)in the urban for NLM group(X2=9.129,P=0.003).According to the data,painful lesions were more likely chronic inflammatory diseases than those without ache(66.7% versus26.3%,respectively;X2=21.820,P<0.001).In addition,there were obviously more people suffering with non-lactating mastitis whose lesions were located around the areola area than those suffering with breast cancer(60% versus 21.1%,respectively;X2=10.768,P=0.001).Under the ultrasound,the lesions with calcification were at 45.6percent of all(31/68).Of 31 cases,ultrasound revealed 24 patients with calcification in the BC group,but only 7 cases in the NLM group(X2=10.719,P=0.001).Including smoking history,the number of reproduction and abortion,swelling,nipple discharge,whether if the axillary lymph nodes touched,nipple retraction,and the mobility,diameter,shape and boundary of the lesion,ascularization,these variables weren`t statistically significant different.Then we used these significant different factors to build the differential diagnostic modal by the Logistic regression equation.These variables were brought into the modal which were age(OR=1.120;95%CI:1.027~1.222),place of residence(OR=7.322;95%CI: 1.009~53.116),lesion pain(OR=7.771;95%CI: 1.265~47.719),lesion distribution(OR=9.523;95%CI:1.313~69.065),calcified nidus(OR=9.632;95%CI: 1.503~61.740).The multivariate analysis showed that the patients more easily suffered from breast cancer who were older,living in the city,not feeling painful,not areola area lesions and calcified but conversely those were included to get non-lactating mastitis who are younger,not living in the city,feeling painful,areola area lesions and not calcified.We worked out the formula: P(the probability of breast cancer)= exp(-9.704+0.114 age+2.254 diseased region +2.05 pain+2.265 calcification+1.991 domicile)/[1+ exp(-9.704+0.114age+2.254 diseased region+2.05 pain+2.265 calcification+1.991 domicile)].The cut off point of the modal is 0.4696 which reveals the area under the ROC curve is 0.952,the sensitivity is 92.11%,the specificity is 86.7%,the position likelihood ratio is 6.91,the negative likelihood ratio is 0.091,the Youden index is 78.78,the agreement rate is91.18.Conclusion: The differential diagnostic modal of non-lactating mastitis and breast cancer was satisfied with the higher sensitivity and specificity and helpful to diagnose two diseases. |
PDF Full Text Request