Lymphoedema After Breast Cancer Treatment is Associated With Higher Body Mass Index: A Systematic Review and Meta-Analysis

Background: Excess body weight has been identified as an important risk factor for lymphoedema following breast cancer treatment, however it remains unclear how much risk increases as weight increases. We conducted a meta-analysis to assess the relationship between body mass index (BMI) and risk of lymphoedema in breast cancer patients, and to estimate the level of risk by BMI category. Methods: We conducted a systematic search of all articles published through May 2018 in PubMed and the Cochrane library. Studies that included data on BMI and lymphoedema in breast cancer patients were included in the meta-analysis. We compared risk of lymphoedema in BMI groups as: BMI<25 versus BMI≥25, BMI<25 versus BMI≥30, BMI≥25 to <30 versus BMI≥30, BMI<30 versus BMI≥30, BMI<25 versus BMI≥25 to BMI<30. Results: After exclusion of ineligible studies, 57 studies were included in the meta-analysis. The mean difference in BMI between patients with lymphoedema compared to those without lymphoedema was 1.7 (95% CI, 1.3–2.2). Compared to patients with a BMI<25, risk of lymphoedema was higher in those with a BMI >25 to <30 (odds ratio [OR] 1.3; 95% CI, 1.2 to 1.5), a BMI≥25 (OR 1.7; 95% CI, 1.5 to 1.9), or a BMI≥30 (OR 1.9; 95% CI, 1.6 to 2.4). Compared to patients with a BMI of >25 to <30, risk of lymphoedema was higher in patients with a BMI>30 (OR 1.5; 95% CI,1.4 to 1.8). Conclusion: Excess body weight is a risk factor for lymphoedema following treatment of breast cancer, with the magnitude of risk increasing across higher categories of BMI.


INTRODUCTION
L ymphoedema of the upper limb is a complication of breast cancer treatment, especially mastectomy, radiation therapy and chemotherapy. 1 It results from reduced lymphatic drainage and stasis of fluid in the extremities, 2 and can occur during treatment or develop years after treatment has been completed. 3 Estimates of the prevalence of lymphoedema following breast cancer treatment are imprecise due to inconsistencies in the definition of lymphoedema. [4][5][6][7][8] However, 1 systematic review found that more than 1 in 5 women who survive breast cancer developed lymphoedema. 9 Several clinical factors have been associated with increased risk of lymphoedema, including: breast surgery, axillary lymph node dissection, sentinel lymph node dissection, radiation therapy, and postoperative infections. 10,11 Obesity has been identified as the primary demographic factor associated with increased risk of lymphoedema of the upper limb following breast cancer treatment.
A number of studies have examined the relationship between obesity and development of arm lymphoedema after breast cancer treatment, with the majority finding that. However, most studies do not report on the frequency of lymphoedema within strata of women who are normal weight, overweight or obese, thus precise estimates on the level of risk associated with each weight strata are lacking. Given the high frequency of overweight and obesity among breast cancer patients, clarification of the level of risk of lymphoedema after breast cancer treatment in overweight or obese women is needed to enhance clinical management of breast cancer in this patient subgroup.
We sought to address this knowledge gap by conducting meta-analyses to assess 1) whether body mass index (BMI, defined as weight in kilograms divided by height in metres squared) differs in breast cancer patients with and without lymphoedema after breast cancer treatment, and 2) risk of lymphoedema after breast cancer treatment in subgroups of BMI.

Search Strategy
A systematic search of all articles published in the English language up to 23 May 2018 was conducted on PubMed and the Cochrane library, using MeSH key words: "breast cancer and lymphoedema". All references resulting from the MeSH search were imported into Endnote X8, and were examined by 2 independent reviewers. During their first round of review, each reviewer evaluated study titles; those that did not contain the targeted search terms were excluded. During the second round of review the full-text of retained study were evaluated to determine if it was potentially eligible for inclusion in the meta-analyses. Discrepancies between reviewers were resolved via discussion.

Inclusion and Exclusion Criteria
The inclusion criteria were: publication in English in a peer-reviewed science or medical journal; assessment of BMI (as a continuous or categorical variable) and lymphoedema in female breast cancer patients; and a period of follow-up less than or equal to 10 years. No published abstracts were included. Included and excluded studies are summarised in Figure 1.

Data Extraction
The following variables were extracted from the published papers (Table): authors, year of publication, study design, patients, type of data, duration of the study, proportion of lymphoedema, lymphoedema evaluation, country where the study was conducted and the definition of lymphoedema. Where available, we extracted patient BMI as a continuous variable; means and ranges were adjusted into mean and standard deviation using the method described by Wan X. 12 Data on BMI were also extracted as a categorical variable, and where appropriate, regrouped to represent the following categories: BMI<25, BMI≥25, BMI≥ 25 to <30, and BMI≥30.

Quality Assessment of Studies
We performed quality assessment of studies included in this meta-analysis using 2 tools: the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies, a 14item inventory; and the Quality Assessment of Case-Control Studies, a 12-item inventory. 13 If an observational cohort or cross-sectional study had more than 8 positive items, or a case-control study had more than 7 positive items, the study was deemed to be of high quality.

Statistical Analysis
In studies with continuous data for BMI, we calculated the mean difference and 95% confidence interval of BMI between patients with lymphoedema and those without lymphoedema. We used data on the number of lymphoedema events among patients within each BMI category to calculate odd ratios of the association between BMI category and lymph-

Study Characteristics
We Fourteen studies were classified as low quality and 43 were classified as high quality. The proportion of lymphoedema in prospective studies included in the meta-analysis ranged between 3% and 71.4% (Table). In total, 5,407 participants from 20 studies contributed data for analysis of mean differences in BMI between patients with and without lymphoedema.

Differences in BMI between Patients With and Without Lymphoedema
In meta-analysis of 20 studies the overall mean difference in BMI between breast cancer patients with and without lymphoedema was 1.7 (95% confidence interval [CI], 1.3 to 2.2); heterogeneity among studies was nonsignificant (I2=28%). In subgroup analysis by study design, in all study subgroups BMI was higher in breast cancer patients with lymphoedema compared to those without. However, the mean difference in BMI was higher in retrospective studies 2.5 (95% CI, 1.4 to 3.6), compared to prospective 1.6 (95% CI, 0.9 to 2.3), cross-sectional 1.9 (95% CI, 1.0 to 2.7) and case-control studies 1.1 (95% CI,0.1 to 2.1). Heterogeneity in prospective studies was moderate (I2=53%) and nonsignificant (I2=0%) in retrospective and cross-sectional studies.

Odds of Lymphoedema by BMI Category
Breast cancer patients with a BMI in the overweight or obese range more frequently developed lymphoedema than those with a BMI<25, with risk rising across higher BMI categories.

FIGURE 2. Comparison of BMI Between Patients With and Without Lymphoedema
Compared to patients with a BMI<25, risk of lymphoedema was higher in those with a BMI in range of 25 to less than 30 (odds ratio [OR] 1.3; 95% CI, 1.2 to 1.5), a BMI≥25 ( OR 1.7; 95% CI, 1.5 to 1.9), or a BMI≥30 (OR 1.9; 95% CI, 1.6 to 2.4). Even among overweight or obese patients, higher BMI was associated with a greater frequency of lymphoedema. Compared to patients with a BMI between 25 and less than 30, odds of lymphoedema was 50% higher patients with a BMI>30 (OR 1.5; 95% CI,1.4 to 1.8). Heterogeneity of OR estimates across studies was moderate in overall analyses comparing patients with BMI<25 to those with BMI>25 (I2=53%) and comparing patients with BMI<25 to those with BMI>30 (I2=49%). Cross-study heterogenity was also moderate in subgroup analysis of prospective studies comparing patients with BMI<25 to those with BMI>25 (I2=53%) and comparing patients with BMI<25 to those with BMI>30 (I2=49%). Study heterogeneity was only substantial in subgroup analysis of cross-sectional studies comparing patients with BMI between 25 and 30 to those with BMI>30 (I2=75%). In all other analyses heterogeneity was nonsignificant.
In subgroup analyses based on study design comparing patients with a BMI<25 to patients with a BMI>25, mean ORs were higher in cross-sectional studies (OR 2.9; 95% CI, 1.7 to 5.3) and case-control studies (OR 2.4; 95% CI, 1.6 to 3.7) compared to prospective studies (OR 1.7; 95% CI, 1.5 to 2.1), and retrospective studies (OR 1.3; 95% CI, 1.1 to 1.5). In contrast, in subgroup analyses comparing patients with a BMI of 25 to less than 30 to patients with a BMI>30, the mean OR was higher in prospective studies (OR 1.6; 95% CI, 1.4 to 1.8) compared to retrospective studies (OR 1.3; 95% CI, 0.8 to 1.9) and cross-sectional studies (OR 1.2; 95% CI, 0.1 to 13.6).In subgroup analyses comparing patients with a BMI<25 to patients with a BMI>30, mean ORs by study types ranged from 1.9 to 2.5 in cross-prospective, cross-sectional and case control studies, with an overall OR of near 2 (OR 1.9; 95% CI, 1.6 to  (Figure 4). A wider range of mean ORs was observed in subgroup analyses comparing patients with a BMI<25 to patients with a BMI>25 and less than 30. The OR was lowest in the retrospective study subgroup, which represented a single study (OR 1; 95% CI, 0.6 to 1.7), moderate in the prospective study subgroup (OR 1.3; 95% CI, 1.2 to 1.5), and highest in the cross-sectional study subgroup (OR 1.9; 95% CI, 0.7 to 5.5) and the case-control study subgroup (OR 1.9; 95% CI, 1.2 to 3.2), which also represented a single study ( Figure 6).

DISCUSSION
In this meta-analysis, we found strong associations between BMI and lymphoedema in female breast cancer patients. Mean differences in BMI were significantly elevated in lymphoedema patients compared with those who did not develop lymphoedema. Further, compared to a reference BMI value of <25 (that is, at or below normal weight), ORs for lymphoedema increased in magnitude across higher categories of BMI, within the range of 1.3 to 1.9. This finding reflects a trend of increasing risk of lymphoedema with increasing weight reported in individual studies included in this meta-analysis. While ORs of the association of BMI category and lymphoedema from individual studies ranged from 0.3 62 to 7.1, 64 only 5 studies reported an OR below 1, reflecting the robustness of our overall estimate. Further, study heterogeneity was moderate to nonsignificant in most analyses and 75% of included studies were of high quality. We observed some variability in the magnitude of ORs by subgroup of study design type, however, subgroup ORs were largely consistent with overall ORs. Strikingly, we found that even among overweight and obese cancer patients, higher BMI increased risk of lymphoedema. In particular, our analysis estimated that risk of lymphoedema was 50% higher in patients with a BMI>30 compared to those with a BMI in the range of 25 to less than 30. This finding is supported by a recent meta-analysis of BMI and risk of lymphoedema, which reported an 39% increased risk of breast cancer-related lymphoedema in obese patients compared to overweight patients. 81 However, lymphoedema is more noticeable, and thus potentially more readily diagnosed in patients with a high BMI compared to those with a normal BMI. While the contribution of diagnostic bias to the observed association between higher BMI and increased risk of lymphoedema is unknown, the observed dose-response relationship between excess body weight and increased risk of lymphoedema suggests a biological link between the 2.
In prospective studies that were included in this meta-analysis, we found a high proportion of lymphoedema, ranging from 3% to 67.7%, with a mean of 24.19%. A similarly high proportion of lymphoedema has been reported in other studies. Based on insurance claim data, 10% of patients had lymphoedema within 2 years of treatment of newly diagnosed breast cancer. 85 A prospective cohort study of breast cancer survivors reported that within 5 years of treatment 43% to 94% of patients had lymphoedema, with estimates varying depending upon how the lymphoedema was defined. 86 These incidence estimates are derived from overall patient populations, and may be even higher in subgroups of overweight and obese women in whom risk of lymphoedema is elevated.
The process through which higher BMI may lead to the development of lymphoedema remains unclear but several mechanisms have been proposed. In particular, lipid accumulation throughout the body may impede lymphatic transport of fluids, in a process driven in part by chronic inflammatory responses. 82 In a mouse model, lymphoedema in obese mice was found to impair lymphatic function, associated with increased subcutaneous adipose deposition, a higher frequency of CD45+ and CD4+ inflammatory cells, and fibrosis without any change in the number of lymphatic vessels. 83

Limitations
Our meta-analysis has some limitations, which should be considered. Firstly, methods used to diagnose lymphoedema were not consistent across the studies included in this meta-analysis, and some studies did report on how diagnosis was conducted. In some studies, BMI was not a primary variable of interest, and thus may not have been carefully recorded. Further, a variety of study populations were represented across studies, including. While this may improve the overall generalizability of our findings, it may also have resulted in wider confidence intervals around our pooled estimates. Our study did not include 191 non-English citations identified by our MeSH search, which could contain important data not considered in this study. Moreover, the majority of studies included in our meta-analysis were conducted in the USA (52.3%) or in Europe, thus our results may not reflect the impact of BMI on risk of lymphoedema in geographic areas not included in the analysis. The publication biases assessment has been summarised in Figure 7 by using funnel plots. The significant asymmetry was found in the funnel plots referring to Figure 3. This should be caused by heterogeneity within studies.

CONCLUSION
This meta-analysis showed that being overweight or obese is an important risk factor for developing lymphoedema of the upper limb following breast cancer treatment. Lymphoedema is more noticeable, and thus potentially more readily diagnosed in patients with a high BMI compared to those with a normal BMI. However, our finding that the magnitude of risk of lymphoedeoma rises across higher categories of BMI supports a biological link between being overweight and developing lymphoedeoma. To further clarify the relationship between excess body weight and risk of lymphoedema, future studies should detail methods used to diagnose lymphoedema and report the frequency of lymphoedema in BMI subgroups from patient populations representing a range of BMI levels.