A Chinese study showed that abortion does not increase the risk of breast cancer, and the latest meta-analysis had not found a relationship between abortion and breast cancer risk[12]. A study by Ilic et al.[28] suggested that even short pregnancies ending in abortion add to the protection against breast cancer. Therefore, the covariates of this study included only the number of pregnancies instead of the number of deliveries.
Egan et al.[29] drawn conclusion that breast size is a positive predictor of postmenopausal breast cancer limited to those who were especially lean as young women from a population-based case-control study of women aged 50 to 79 years. Williams et al.[30] deemed baseline bra cup size was the strongest predictor of breast cancer mortality. A prospective study by Kusano et al. suggested that for women with a BMI below 25 kg/m2, those with a bra cup size of "D or larger" had a significantly higher incidence of breast cancer than women who reported "A or smaller" (covariate adjusted HR = 1.80; 95% CI = 1.13–2.88; p trend = 0.01), though this association was limited to leaner women[31]. Then some experts pointed out the shortcomings of this study, they thought that the use of cup size alone without taking rib cage circumference into account was poorly rigorous[32]. In addition, cup size labeling was not standardized and different brands of brassieres differ in their labeling of cup size for the same breast volume. Breast size as measured by self-reported bra cup size was the biggest drawback of these studies.
For the measurement of breast volume, the gold standard is the water displacement method[33]. However, its operation is complicated and it is difficult for patients to cooperate. Three-dimensional ultrasound (3-D US)[33] is a relatively close to the water replacement method, if it is not expensive and requires professional cooperation. 3D scanning is a new and more advanced method[34]. However, for women with larger breast volume, 3D scanning is not accurate[35], and the technical and cost requirements are higher. Choppin et al.[36] considered the highest accuracy of magnetic resonance imaging (MRI) scan after comparing 3D scanning, mammography, MRI, CT examination, model casting and other methods. Itsukage et al.[37] also believed that MRI is more accurate in measuring breast volume.However, it requires data analysis software and is more expensive. The breast measurement BREAST-V formula used in this study is the first unified, more effective and reliable breast volume prediction formula designed by Longo.[27] It is the most common method used by the researcher's unit and can be used for volume assessment of breasts of different sizes and is easy to operate without higher requirement for the measurement technique. The data is subjectively less affected, and it is more accurate for measuring the breast volume.
In the present study, there was no significant difference in the age between case group and control group before PSM. The reason may be that we excluded the lower age population in the process of collecting case group data, and we found that except for the unit physical examination, the age of physical examination population has a distribution of 40 to 70, which is similar to the high-risk age of onset of breast cancer. The reason for the lack of significant difference in BMI between two groups may be due to insufficient sample size, in addition does not rule out the existence of Berkson bias.
The advantage of this study is that our breast volume data was obtained from a simple and convenient method based on linear measurement of the breast; on the other hand, PSM was used to balanced potential confounding between case group and control group to ensure the comparability between two groups to a certain extent. Compared with the traditional multivariate regression model, the over-fitting was reduced to some degree since the propensity value score was applied to all the selected patients and finally matched according to the tendency value score. In the sample size, considering the problem of unsuccessful matching, the control group collected more data than the case group, which guaranteed the matching success rate to some extent.
The limitation of this study is that although the hospitals selected have a wide range of radiation, they are limited after all, and some patients with double mastectomy cannot obtain breast size data. For the exposure rate in the population can not be accurately estimated. According to a large number of purchase data analysis, women with C cups and above accounted for at least 30%. This study requires the measurer to communicate with the person being measured, blindness cannot be established, so the existence of suspected bias is not ruled out. Young women who were breastfeeding and not breastfeeding have differences in breast volume, while older women will reduce their breast volume due to fat reduction. Therefore, although this study used PSM, imbalance cannot be completely eliminated, which is one of the shortcomings of this research.