Breast cancer is one of the most diagnosed types of cancer and the leading cause of cancer-related mortality among women worldwide
[1]. Recent advances in breast cancer diagnosis and treatment have significantly improved long-term survival
[2], but postoperative complications are still a major concern
[3]. Up to 20% of breast cancer survivors are affected by breast cancer-related lymphedema (BCRL)
[4], one of the most common and troublesome complications of breast cancer treatment
[5]. BCRL is marked by an abnormal buildup of lymphatic fluid in the parts of the body that are affected
[6-7]. This can cause long-term physical and mental symptoms that can make a patient’ s quality of life much worse
[8-10]. In addition, breast cancer survivors face a lifetime risk of lymphedema, and there is currently no cure for BCRL. Thus, strategies to prevent the onset of BCRL and improve the quality of life for breast cancer survivors are urgently required.
Existing risk factors for development of lymphedema include axillary lymph node dissection (ALND)
[11], regional lymph node radiation (RLNR)
[12], the number of positive lymph nodes
[13], a higher body mass index (BMI)
[14], cellulitis
[3], and low-level limb volume changes
[15]. All of these factors contribute to an increased risk for the development of lymphedema and have been studied extensively. However, there are still many unanswered questions regarding lymphedema risk factors. Multiple studies have examined the relationship between BCRL and chemotherapy
[16-18]. The results of these studies have been inconclusive, indicating that much more research is required in this area.
One of standard treatments for breast cancer is taxane-based adjuvant chemotherapy, which has been shown to considerably increase patients’ overall survival and progression-free survival
[19-20]. Taxane-based adjuvant therapy, particularly docetaxel, promotes fluid retention in the extremities because of an increase in extracellular fluid
[21]. To prevent or postpone the onset of fluid retention caused by taxanes, patients frequently receive corticosteroids during treatment
[22]. The long-term effects of arm edema from taxane-based adjuvant chemotherapy are not yet known, though. There is conflicting evidence regarding the relationship between taxane-based adjuvant chemotherapy and the emergence of lymphedema in breast cancer survivors
[16-17, 23]. These contradictory findings imply that further study is required to better comprehend the link between taxane-based adjuvant chemotherapy and arm edema. These previous results likely varied based on demographic factors, tumor characteristics, surgical procedures, and treatment protocols. Moreover, these current studies had clear faults, such as the fact that none of them addressed the combined impact of many hazards while examining a single risk. In addition, these studies were conducted on Western populations, and there is a dearth of information on Chinese breast cancer patients. Further research is needed that accounts for a variety of characteristics and patient populations to correctly draw conclusions about the association between taxane-based chemotherapy and BCRL.
Prior studies
[16, 23] mainly relied on conventional regression models. A traditional regression model may be biased due to unmeasured or residual confounding given the abundance of potential confounding variables, whereas adding more variables may lead to model overfitting and potentially prevent the identification of the association between the relevant exposure and the outcome
[24-25]. The propensity score matching (PSM) method is widely used for controlling confounding in observational research. It is especially beneficial in studies where there are numerous covariates that may confound a rare outcome, where there may be potential confounding by indication, and where there are resource limitations that preclude the conduct of randomized clinical trials
[26-27]. By balancing the distribution of covariates between exposed and unexposed cases, the PSM approach can help reduce possible bias from confounding, and when implemented correctly, it can produce reliable results even in observational studies. This study aims to investigate the association between taxane-based adjuvant chemotherapy and BCRL in Chinese using PSM.
1 Materials and Methods
1.1 Ethics statements
The study design and use of clinical data were approved by the institutional review board of the Third Xiangya Hospital of Central South University (Approval number: 2023-S527). Due to the retrospective nature of the study, the requirement for informed consent was waived, and all patient data were anonymized.
1.2 Subjects
A retrospective study was conducted. We examined the medical records of individuals who underwent breast cancer surgery in the Third Xiangya Hospital, Central South University from January 2013 to September 2023. Inclusion criteria were: Female patients aged 18 years or older, with a confirmed diagnosis of breast cancer, and undergoing surgical treatment for breast cancer. Individuals who met any of the following criteria were excluded: 1) Had metastasis, bilateral or recurrent breast cancer; 2) had upper limbs or the neck with a history of major surgery; 3) lacked clinical data and follow-up information; 4) had chemotherapy prior to breast cancer surgery; 5) had less than 2 courses of adjuvant chemotherapy or had neoadjuvant chemotherapy. BCRL group included patients who suffered postoperative BCRL, while control group included patients who did not experience postoperative BCRL during the follow-up period.
Figure 1 depicts the patient selection process flowchart.
1.3 Data extraction
Medical and nursing records uploaded to the electronic medical record system were mined for information. We received patient information, tumor information, and therapy information from the database. Patient information included age, gender, height, weight, educational level, marital status, arm lymphedema swelling severity, and comorbidities. Tumor information included the stage of cancer, tumor stage, lymph node (N) stage, and lymph node metastasis status. Treatment information, namely the type of therapy, included the number of lymph nodes removed, ALND, chemotherapy with paclitaxel, docetaxel, or other regimens, radiotherapy, and endocrine therapy.
1.4 Identification of chemotherapy
Paclitaxel, albumin paclitaxel, and docetaxel are all components of taxane-based chemotherapy. Dexamethasone premedication was given before chemotherapy as per each institution’s practice
[28]. Neoadjuvant chemotherapy recipients were not included in this analysis. If a patient changed chemotherapy regimens during adjuvant therapy, the final regimen prevailed.
1.5 Diagnosis and measurements of BCRL
The diagnosis of BCRL was made with an objective measurement of limb circumference. The physiatrist measured the circumference of both arms at the metacarpophalangeal joint, the wrist, 10 cm distal to the antecubital fossa of the elbow, the elbow, and 10 and 20 cm proximal to the elbow
[29], after feeling the skin and subcutaneous tissue texture of the entire arm. Patients were diagnosed with lymphedema if they had a circumferential difference (CD) of less than 1 cm between their arms, or if they had a CD of less than 1 centimeter in conjunction with pitting edema or hardening of the subcutaneous tissue. If the CDs of any measured site were less than 2 cm, more than 2 cm but less than 3 cm, or greater than 3 cm, the lymphedema was classed as mild, moderate, or severe, respectively
[30]. Patients diagnosed with BCRL were given care and monitored in special clinics.
1.6 PSM
PSM is an efficient statistical technique that can reduce the influence of confounding variables
[31], such as a high BMI, more invasive surgery, advanced-stage cancer, radiation, and so on, that may contribute to outcome bias. The propensity score was assigned using a multivariable Logistic regression model and estimated based on the probability that a patient would develop arm lymphedema. A one-to-two nearest neighbor matching technique with a caliper width of 0.02 was implemented. Age, BMI, TNM stage, ALND, lymph node metastases, number of lymph nodes resected, sentinel lymph node biopsy, surgical type, and radiotherapy were used to create the propensity score.
1.7 Reduction of bias
We intended to address potential sources of bias by doing the following. All cases and matched controls were drawn from the same cohort. We also matched the propensity scores of cases and controls to reduce bias arising from confounding and other factors. Lastly, the variables were obtained from trained postgraduate students who did not participate in clinical management.
1.8 Statistical analysis
SPSS 25 (IBM) and R 4.0 statistical analyses (R Project for Statistical Computing) were used to analyze the data. Quantitative data were represented using mean±standard deviation and the independent samples t test for normally distributed data. M (P25, P75) and the Mann-Whitney U test were used for skewed data. The ratio, or composition ratio and frequency, were used for count data, and the chi-square test or Fisher’s exact test was employed. The association between taxane-based chemotherapy and BCRL was assessed using logistic regression and expressed as an odds ratio (OR) and 95% confidence interval (CI). Model 1 was not modified. Adjustments were made to Model 2 to account for BMI, ALND, sentinel lymph node biopsy, radiation, and TNM stage. Model 3 was modified for variables in Model 2 plus additional factors, including age, lymph node metastases, lymph nodes that were removed, and breast surgery. To limit the influence of confounding variables, we applied matching strategies based on propensity scores. Standardized mean difference (SMD) was calculated to evaluate the efficacy of PSM between the 2 groups. P<0.05 (2-sided) was considered statistically significant.
2 Results
2.1 Characteristics of patients before PSM
A total of 1 494 participants were enrolled in this study (
Figure 1). According to the diagnostic criteria for BCRL, 124 (8.3%) of the participants had lymphedema, while 1 370 (91.7% of the subjects) did not. The median age of the total study population was 56 (50.0, 62.0) years. The features of the baseline are reported in
Table 1. There were no statistically significant variations in age between the BCRL and control groups (
P=0.445). Patients in the BCRL group had a substantially higher BMI than those in the control group [24.0 (21.9, 25.9) vs 23.0 (21.2, 25.2),
P=0.021]. Patients in the BCRL group had a higher TNM stage (
P=0.010), more lymph node metastasis (54.8% vs 38.6%,
P=0.001), more ALND (30.6% vs 19.1%,
P=0.003), more radiotherapy (28.2% vs 13.9%,
P<0.001), and more chemotherapy (90.3% vs 80.6%,
P=0.011), but fewer sentinel lymph node biopsy (12.9% vs 25.1%,
P=0.003).
A multivariable Logistic regression analysis was done to determine the potential causes of BCRL occurrences (
Table 2). The included parameters were as follows: BMI, TNM stage, sentinel lymph node biopsy, ALND, radiotherapy, chemotherapy, and lymph node metastasis. Variables that showed a significant influence on this event were BMI (
OR=0.938, 95%
CI 0.882 to 0.998;
P=0.042) and radiotherapy (
OR=0.581, 95%
CI 0.362 to 0.946;
P=0.026).
2.2 Characteristics of patients after PSM
By using a 1:2 PSM, 121 BCRL group subjects were finally matched with 242 control group subjects. We compared the characteristics between the 2 groups and found that these differences were well controlled after PSM (
Table 1).
2.3 Relationship between taxane-based chemotherapy and the risk of BCRL
Table 3 presents the frequency distribution of adjuvant taxane chemotherapy in the BCRL and control groups before and after PSM. The unadjusted, marginally adjusted, and completely adjusted models are showed in
Table 4. The link between taxane-based chemotherapy and the risk of BCRL was not significant in the primary population (
OR=0.676, 95%
CI 0.418 to 1.055;
P=0.097). The result was not statistically significant after further adjustment (
OR=0.979, 95%
CI 0.587 to 1.580;
P=0.932). A multivariable Logistic regression analysis of the PSM population showed that there was no statistically significant difference between the BCRL and control groups (
OR=0.959, 95%
CI 0.540 to 1.672;
P=0.885).
3 Discussion
Taxane-based chemotherapy is now used to treat patients with node-positive and high-risk node-negative breast cancer because it has been shown to lower mortality
[32]. Taxanes are being given to more patients in the early stages of breast cancer as the use of anthracycline-alone therapies declines
[33]. Therefore, a comprehensive assessment of the quality of life and long-term effects of taxanes is required. Given the increase in extracellular fluid volume, adjuvant taxane-based therapy enhances fluid retention in the extremities. Due to the similarities between the mechanism of fluid retention and the onset of lymphedema in breast cancer patients
[34-35], the association between taxane-based chemotherapy and lymphedema requires more investigation. The link between BCRL and taxane-based therapy has been described but not exhaustively studied.
This retrospective analysis observed no significant association between taxane-based adjuvant chemotherapy and an elevated incidence of lymphedema compared with no adjuvant chemotherapy or non-taxane chemotherapy. Our findings appear to contradict the findings of prior research
[36-37], which suggested that taxane-based adjuvant treatment has a considerable association with BCRL. This disparity may be partially explained by design and approach differences. In a larger retrospective cohort research by Cariati, et al
[16], the diagnosis of BCRL was based on changes in the arm’s appearance and/or volumetric measures between the ulna styloid process and the anterior axillary fold of both arms, which may be less reliable than the quantitative test. In a similar study by Jung, et al
[23], which looked at patients who had ALND, taxane-based chemotherapy was found to be an independent risk factor for lymphedema. Localized forearm or hand edema might have gone unnoticed because they were only measured through the circumference of the upper arm. What’s more, because the studies were not randomized, it was challenging to balance variables between the 2 groups. Since the PSM methods offered a control group that was very similar to what could be found in a randomized controlled trial, our work was not limited in this way. In this study, the PSM method diminished the disparities between the BCRL and control groups. Consequently, the relationships between taxane-based adjuvant chemotherapy and BCRL were independent of the confounding factors evaluated in this investigation, including patient demographics, treatment, and clinicopathologic characteristics.
Our results are consistent with a major prospective cohort research
[17] undertaken at the Massachusetts General Hospital in the United States. They studied the link between taxane-based chemotherapy and the incidence of lymphedema in 1 221 breast cancer patients who were followed for 2 years. In a multivariate analysis, patients who received taxane-based chemotherapy were not significantly different from those who did not. These may be due to the low cumulative incidence of lymphedema across the cohort and instability of multivariate analysis in lymphedema risk factors (e.g. ALND) identification. In this study, the use of PSM removed the effect of other risk variables on BCRL, which may explain our findings. A study
[38] from the 2022 San Antonio Breast Cancer Symposium also showed that adjuvant chemotherapy, including taxane-based chemotherapy, was not associated with a significant increase in the risk of BCRL. To determine adjusted hazard ratios, a Cox regression model was employed in the study. The results were adjusted for BMI at baseline, ALND, RLNR, age at diagnosis, and smoking history. The findings demonstrated that adjuvant chemotherapy is not an independent risk factor for BCRL. Our results were similar in that we conducted a multifactorial analysis of the primary population before PSM, and after adjusting for other factors, the results also indicated that chemotherapy was not a risk factor for BCRL.
Previous study
[35] has shown that exposure to taxanes can induce endothelial inflammation, leading to abnormal capillary permeability. Other researchers
[36] have concluded that taxanes cause progressive protein accumulation in the interstitial space due to abnormal capillary permeability, thereby enlarging the interstitial extracellular fluid compartment and often leading to fluid retention in edema. Although the mechanism of this fluid retention shares certain similarities with the pathogenesis of lymphedema in breast cancer patients, the findings of this study suggest that drug-induced edema and lymphedema may represent two distinct pathophysiological processes. Edema caused by taxane drugs is typically reversible and cyclical, whereas lymphedema is a chronic, progressive condition resulting from damage to the lymphatic system. The findings of this study suggest that clinicians should be vigilant against misinterpreting transient edema during chemotherapy as an early manifestation of lymphedema. Given the well-established survival benefit of taxane-based therapies, these results further support their value in adjuvant breast cancer treatment. For patients, this discovery may also alleviate concerns that chemotherapy could exacerbate lymphedema, thereby improving treatment adherence and quality of life.
This is the first study to our knowledge that employs PSM to analyze and evaluate the relationship between taxane-based treatment and the risk of BCRL. Since the propensity score was applied to all the selected patients and the final matching was based on the propensity score, overfitting in this study was substantially mitigated compared to normal multivariate regression. Our results fill a gap in the research on the relationship between lymphedema and taxane-based chemotherapy among Chinese breast cancer patients. Moreover, our findings will improve patient education about the potential association between BCRL and adjuvant chemotherapy, particularly taxane-based chemotherapy.
Several limitations should be carefully noted in this investigation. The first systematic error source is recall bias, a well-known cause of erroneous associations between exposures and the outcome of interest. Second, the retrospective study design made it impossible to make causal inferences on the association between taxane-based treatment and BCRL. Therefore, additional prospective trials are required to assess whether taxane-based chemotherapy increases the incidence of BCRL. In addition, our findings cannot be extended to the broader Chinese population because all subjects were selected from the same province.
This study used a PSM analysis to indicate that taxane-based chemotherapy does not increase the incidence of BCRL beyond that of other known risk factors. However, more study is needed to assess the association of taxane-based treatment with BCRL risk. It is suggested that prospective study designs, large samples, and various centers be used in future research.