Introduction

Non-puerperal mastitis (NPM), defined as a general term for non-specific inflammatory mastitis diseases that occur during the non-puerperal period, primarily involves granulomatous lobular mastitis, ductal dilation, and periductal mastitis1. As the largest subgroup in chronic mastitis, NPM usually occurs in young and middle-aged women who are not lactating or pregnant, and its aetiology remains unclear2,3.

The typical clinical symptom of NPM is breast lumps presenting with pain and erythema. Due to the concomitant symptoms and similar image appearance in radiology, patients in the early stage of NPM with milder symptoms could be misdiagnosed as breast cancer4,5,6. Primary treatments of NPM include oral treatment and surgery. For oral treatment, the main pharmacological treatments are antibiotics, steroids7,8, methotrexate, and traditional Chinese medicine9. Traditional Chinese encompasses various therapeutic approaches, including both systemic and topical administration of missing data imputation.. However, as the pathogen of NPM is not clear, the efficacy of most oral treatments remains unsupported by robust evidence. The lack of validated oral treatments and accurate diagnosis potentially lead to the progression of the disease’s severity. As a result, most patients have clinically developed pus and abscesses that rupture successively in different areas of the breast, which could eventually result in mammary fistulas, sinus tracts, and ulcers1. Consequently, surgery becomes a more efficacious approach for NPM patients with severe symptoms10.

Various surgical procedures are available for treating NPM, yet none is considered optimal due to the long healing time of surgery incisions11,12,13,14. This challenge is compounded by the difficulty in diagnosing and assessing the severity of NPM through radiology methods4,5,15. Furthermore, the inability to accurately gauge the severity of NPM before surgery leads to a high risk of recurrence after NPM surgery, posing an emerging concern both for patients and clinical experts16,17. Previous studies have shown that the long-lasting condition of NPM and the potential repeated breast surgeries could negatively affect the physical and mental health of patients, as well as incur higher medical costs18. Hence, identifying factors that could help assess the severity before surgery is becoming increasingly important.

Although there is no consensus on the factors related to disease severity under NPM terminology, numerous discussions have explored the risk factors associated with the onset and recurrence of NPM-related diseases19,20. Previous studies focused on plasma cell mastitis and granulomatous mastitis have identified several factors for disease onset and recurrence, including obesity21, nipple deformity7,11, lifestyle, specific medication utilization, dietary patterns, and occupation7,22. Other potential factors include histories of ductal dilatation, nipple discharge, and breast and cranial trauma23. In addition, blockage of milk ducts and the onset of acute mastitis during pregnancy and lactation, are relevant to the higher risk of NPM as well14. However, no previous study demonstrated solid evidence of the causal association between these factors and NPM.

In terms of laboratory tests, some evidence illustrated that increasing prolactin (PRL) levels were highly related to the occurrence24 and recurrence25,26of NPM. This is likely attributable to the fact that hyperprolactinemia can trigger a local hypersensitivity reaction within the mammary ducts27. However, few studies focused on the association between PRL level and NPM severity of first-time onset or further investigated risk factors among NPM patients who need surgical treatment. Most of the previous studies are case reports and are limited in sample size, data quality, valid study design, and analysis3, and they primarily described the characteristics of the NPM27,28.

This observational study intends to explore an inclusive understanding of NPM by applying electronic medical record (EMR) data. The primary objective is to map the characteristics of NPM patients who are first-time onset and need surgical treatment. The secondary objective is to investigate the association between PRL level and NPM severity and clarify if there were other potential determinants related to NPM severity. We hypothesize that higher PRL level increases the risk of NPM severity compared to normal to low PRL levels.

Method

Data source and sample

This observational study included the diagnosis information and medical history data from EMRs of NPM patients admitted to the Department of Breast Surgery, Longhua Hospital Affiliated with Shanghai University of Traditional Chinese Medicine for operation from Jan.2016 to Dec.2020. All included NPM patients received surgery after they were diagnosed with mammary inflammatory granuloma, mammary duct fistula, plasma cell mastitis, and mammary duct ectasia, and additionally with the following specific texture in pathology diagnosis report: granuloma formation, duct dilatation, periductal inflammation, and fistula. All data are collected at the time of admission. For those inpatients with multiple admissions during the period, this study only counted them one time and used the data at the first admission. 7 patients who have two-side onset are excluded. To ensure the validity of missing data imputation, 8 patients who had missing values in glucose (GLU), triglycerides (TG), total cholesterol (tCHOL), testosterone (TE), and follicle-stimulating hormone are excluded. These values are entirely absent for these patients at the beginning of the study period, raising concerns that the missingness may not be random and could violate to random missing assumption in the context of missing data imputation. This study was approved by the Medical Ethics Committee of Longhua Hospital affiliated with the Shanghai University of Traditional Chinese Medicine (approval no.2021LCSY047). Written informed consent for personal data collection and utilization was obtained from all individual participants included in the study. All information collected was kept private and confidential. Everyone who participated had the option to withdraw at any moment. This study adhered to the declaration of Helsinki. All methods were performed following relevant guidelines and regulations. The study registration number is ChiCTR2000035929.

Measures

Exposure: prolactin (PRL) level

We used the blood test lab result of PRL at the admission of each patient as primary exposure. The clinical normal level of PRL is from 108.8~557.1 mIU/ L. We define patients with PRL higher than 557.1 mIU/ L as the high-level PRL group, and the rest of them as the low-normal PRL group.

Outcome: NPM severity assessment score

Currently, there is no official diagnostic standard for the severity of NPM in China. Hence, we created a unified scoring measurement for daily clinical practice to assess NPM severity through clinical manifestations of patients, the imaging results of the ultrasound, and breast enhancement Magnetic Resonance Imaging(MRI). The lesion of a quadrant of the breast and main mammary accumulates to the total score for each patient, with a scale of 1 to 5 points (Attached file: The scoring scale of NPM severity). Each patient was assessed back-to-back by two attending physicians engaged in clinical work for more than five years. If the scores of the two attending physicians are different, a chief physician with more than 10 years of working experience will conduct the re-assessment to reach an agreeable conclusion.

Covariates/confounders

This study included demographics, physical risks, lab tests, and medical history indicators in the multivariable logistic regression model based on the conclusions from previous studies1,2,11,18,22. In this study, demographics included age and smoking status (have a smoking history or not).

Physical risks include two variables: nipple deformity and external force. The definition of nipple deformity is congenital nipple invagination (NI), in that the nipple is born with not protruding from the areola plane, or even contracting and rolling into a crater shape, which is also known as a crater nipple. The information about external forces is collected through the inquiry. The questions include “Was there an accidental impact on the breast”, “Is the patient used to wearing a tight bra”, and “Is there an external pressure check (i.e. mammography)”. We further asked the patient to investigate whether the external force was relevant to the latest episode or worsening the disease.

Lab test and medical history indicators include four variables: GLU level (High level if GLU higher than 6.1 mmol/L), Alanine aminotransferase (ALT) level(High level if ALT higher than 45 IU/L), TE level(High level if TE higher than 1.67 nmol/L), and hyperlipidemia(HLP) status (Yes if TG > = 1.7 mmol/L or tCHOL > = 5.72 mmol/L or positive in the previous history of hyperlipidemia or fatty liver disease).

Statistical analysis

Missing data imputation

This study assumed that the missing completely at random (MCAR) and applied Multiple Imputation by Chained Equations (MICE) for missing data imputation. We conducted predictive mean matching (pmm) in MICE since variables that need imputation are quantitative and with missing data of less than 5% in total. The algorithm was set to run for 50 iterations and presented us with 5 imputations for each missing datum. GLU is the variable that has the highest missing proportion (31 missing in 674 patients).

Data analysis

The current study used the t-test for numerical variables and the ANOVA tests for categorical variables and reported the distribution of demographics, physical risks, lab tests, medical history indicators, and other characteristics stratifying patients into with and without nipple deformity.

This study conducted logistic regression among NPM patients without pituitary tumour history. We defined the NPM severity assessment score into reference level (1–3 points) and higher lever (4–5 points). The binomial regression model was applied separately in patients with and without nipple deformity. The current study defined the significant level at a p-value less than 0.05 and reported all the results in both regression models. We used R studio version 4.0.3 (2020-10-10) for data analysis and conducted missing data imputation with the mice package29.

Ethics, consent and permissions statement

The Medical Ethics Committee of Longhua Hospital affiliated to Shanghai University of Traditional Chinese Medicine approved this research(2021LCSY047).

Consent to publish

We have obtained consent to publish from the participant (or legal parent or guardian for children) to report individual patient data.

Result

This study included 674 NPM inpatients admitted from 2016 to 2020 in demographic analysis. Table 1 shows the characteristics of NPM patients by PRL level between nipple deformity subgroups. The majority of NPM inpatients are not born with nipple inversion. Among NPM inpatients without normal breasts, those who have high PRL levels are more likely to have a pituitary tumour history (p = 0.016).

Table 1 Comparison of NPM patient characteristics between PRL level stratified by breast deformity.
Full size table

Table 2 implied that the severity degree and the characteristics are different between NPM patients with and without nipple deformity. The distribution of patients in severity level varied between NPM patients with normal nipple structure and those are nipple deformity (p < 0.001). NPM patients with normal breasts are less likely to have a pituitary tumour (p = 0.001). In addition, NPM patients with normal breast structure are more likely to be younger than those with congenital nipple deformity on average (p = 0.002 in Table 2), which is not significant within the corresponding nipple structure subgroups (p = 0.211) among those with NI and p = 0.614 among without NI in Table 1). However, the interaction effect between PRL level and nipple deformity in the logistic regression model is not statistically significant at 0.05 level (Table S1).

Table 2 Comparison of NPM patients with and without breast deformity (only show the covariant with p < 0.05).
Full size table

As previously mentioned, the distribution of NPM patients with a history of pituitary tumours varies across patients differ in breast deformity. To better control for potential confounding, we included only the 653 NPM patients without a history of pituitary tumours in the logistic regression analysis. Additionally, testosterone levels were excluded as a co-variable in the regression model for patients with breast deformity, as the number of patients with higher testosterone levels was too small after excluding those with a history of pituitary tumours.

The odds of NPM severity for patients with higher PRL levels varied in unadjusted and adjusted OR (Table 3) between the stratified nipple deformity subgroups. Among NPM patients without congenital nipple inversion, those with higher PRL levels had significantly higher odds of more severe NPM episodes both in unadjusted (OR: 2.93; 95%CI:1.91–4.47; p < 0.001) and adjusted (OR: 2.91; 95%CI:1.88–4.52; p < 0.001). Besides, age is a protective factor for NPM severity (OR:0.95; 95%CI: 0.91–0.99; p = 0.012) while smoking history increased the odds of more severe NPM symptoms (OR: 2.22; 95%CI:1.22–4.05; p = 0.011) in patients with normal breast structure. For NPM patients with congenital nipple inversion, there is no statistical difference in the risk of NPM severity between the patients who had higher PRL or other potential risk factors.

Table 3 Crude and adjusted OR stratified by breast structure. PRL Low-normal is the reference level.
Full size table

Discussion

This study discovered that the risk of NPM severity is higher among patients with higher PRL levels. NPM severity distribution and the effects of risk factors varied between NI and non-NI patients, implying the potential interaction effect between PRL level and nipple deformity.

This is the first study that investigated the relationship between PRL level and NPM severity considering confounders including nipple deformity, age, smoking, external force, and autoimmune with a feasible sample size. The conclusion is agreeable to previous studies that the effect of these factors differed from the breast structure14,30,31. The results illustrate that NPM patients with nipple deformity are more likely to receive milder assessment results (p < 0.001), which is consistent with previous studies assuming that the reason for restricted lesions among these patients is associated with squamous metaplasia of lactiferous ducts that link to subareolar abscesses32.

However, since this is an observational cohort study, the sample size of the non-NI subgroup is about 2.5 times of the NI subgroup. The restriction and imbalance in patient size in the analysis could lead to several potential biases.

First, we failed to include other potential risk factors in the regression model because of the restricted patient number. Previous evidence demonstrated that the incidence of relapse condition is higher among patients who received simple segmental mastectomy33. Other evidence showed that hormonal contraceptive utilization, breastfeeding, abortion, hormone supplements, thyroid abnormalities, and mental health conditions7,19,22,34 are also potential confounders related to NPM. This study did not include these factors in the analysis because the patient number of corresponding potential confounders was too small to conduct a valid analysis. For example, there are only 2 patients who have a hormonal contraceptive use history and 13 patients with a breastfeeding history. Other confounders that were excluded in the analysis were listed in Table 1 as in the “Other” category.

Second, this study was not able to clarify the interaction effect between PRL and nipple deformity associated to NPM severity because of the imbalance sample size between nipple deformity subgroups. Although we found the difference in OR after stratification which implied that nipple deformity may be an effect measure modification (EMM), the result was not statistically significant in OR when directly using interaction terms in the logistic model. Because of the disagreement between the two approaches of EMM testification, in addition to that the causal association between breast structure and NPM severity is not clear, further study is needed to verify the interaction effect between PRL and nipple deformity, and whether the interaction is additive or multiplicative.

This study only observed the effects of age and smoking among NPM patients without NI, both of which agree with NPM clinical features and previous studies although those studies did not stratify the NPM patients by breast duct structure7,35.

Younger age is protective in NPM patients with nipple deformity groups, and the same tendency was observed in NPM patients without nipple deformity group as well, although the result was not statistically significant at 0.05 level. We inferred that the protective effect could relate to the lower PRL level since the PRL level will decrease as age increases, which further influences the NPM severity. The reason for different statistical significance could be associated with the imbalance in patient size between nipple deformity subgroups.

Another concern of the current study is the validity of the NPM severity assessment scoring system. Although the measurement of NPM severity is based on imaging results, clinical manifestations and physician judgement36,37, it is still limited in potential inter-observer variability and scoring system reliability, and the measurement form is not widely used as a guideline1,38. Although the current study re-categorized the severity level into two groups for comparison to increase the internal validity, further investigation is needed to assess the validity of the scoring system and how the limitation of the scoring system influences the result.

Finally, this cohort study only demonstrated a snapshot of the un-causal relationship between PRL level and NPM severity among patients with severe cases who need surgical treatment. Future research that emphasizes longitudinal studies and randomized clinical studies is needed, to clarify the causal relationship between relevant factors regarding different NPM outcomes.

Conclusion

As an umbrella term for non-specific inflammatory mastitis diseases occurring during the non-puerperal period without a well-defined aetiology, it is critical to investigate risk factors that influence NPM severity, as they provide insights into its pathogenesis. This study identified an association between NPM severity and the interaction of breast structure and PRL level. Patients with higher PRL levels exhibited an increased risk of NPM severe episodes, with the association being more pronounced in those with normal breast structure. This finding implied different pathogenesis between NPM patients with varied breast anatomy. Further discussion on refining the current NPM severity assessment is warranted. Finally, inconsistencies in the validation of the interaction effect highlight potential areas for improvement in the current study.