Abstract
Breast cancer-related lymphedema (BCRL) is a serious chronic condition after breast cancer (BC) surgery and treatment. It is unclear if BCRL risk varies by race/ethnicity. In a multiethnic prospective cohort study of 2953 BC patients, we examined the association of self-reported BCRL status with self-reported race/ethnicity and estimated genetic ancestry. Hazard ratios (HR) and 95 % confidence intervals (CI) were calculated by multivariable Cox proportional hazards models, with follow-up starting 6 months post-BC diagnosis. Estimates were further stratified by body mass index (BMI). By 48 months of follow-up, 342 (11.6 %) women reported having BCRL. Younger age at BC diagnosis, higher BMI at baseline, and lower physical activity were associated with greater BCRL risk. African American (AA) women had a 2-fold increased risk of BCRL compared with White women (HR = 2.04; 95 % CI 1.35–3.08). African genetic ancestry was also associated with an increased risk (HR = 2.50; 95 % CI 1.43, 4.36). Both risks were attenuated but remained elevated after adjusting for known risk factors and became more pronounced when restricted to the nonobese women (adjusted HR = 2.31 for AA and HR = 3.70 for African ancestry, both p < 0.05). There was also evidence of increased BCRL risk with Hispanic ethnicity in the nonobese women. Nonobese AA women had a higher risk of BCRL than White women, which cannot be fully explained by known risk factors. This is the first large-scale, prospective study demonstrating differences in BCRL risk according to race/ethnicity as assessed by both self-report and genetic ancestry data, with a potential ancestry–obesity interaction.
Keywords: Breast cancer-related lymphedema, Racial disparity, Ancestry, Genetics, Obesity, Interaction
Introduction
Breast cancer-related lymphedema (BCRL) is a serious chronic condition characterized by lymphatic fluid retention and subsequent tissue swelling, typically in one or both arms [1, 2]. Other symptoms include pain, heaviness, discomfort, and reduced mobility and function [3]. BCRL can occur in about 20 % of patients after breast cancer (BC) surgery and treatment [1, 4]. Elevated risk is associated with younger age, higher body mass index (BMI), and less physical activity [5].
Only a few prospective studies have examined racial/ethnic differences in BCRL risk [6–8], including our group which found a nearly twofold elevated risk among African Americans (AA) [6]. However, the other studies found no association after 40 [7] and 123 months of follow-up [8] with multivariable adjustment accounting for BMI, chemotherapy, and hypertension. To our knowledge, studies of patients' genetic background with BCRL risk have not been conducted. Given that race/ethnicity is a complex construct of sociodemographic, cultural, and genetic factors, the investigation of genetic heterogeneity in BCRL risk may not only affirm the differences in BCRL risk by self-reported race, but also disentangle potential genetic from nongenetic contributing factors.
We examined the associations of race/ethnicity and genetic ancestry with self-reported BCRL status in a cohort of 2953 breast cancer patients with up to four years of follow-up, and investigated interactions of self-reported race/ethnicity and genetic ancestry with obesity on BCRL status.
Methods
Study population
The Pathways Study is a prospective cohort of women with newly diagnosed, invasive breast cancer at Kaiser Permanente Northern California (KPNC). Recruitment was from January 2006 to May 2013 through rapid case ascertainment procedures [9]. Briefly, cases were identified through daily scanning of computerized pathology records for new patients with recently diagnosed invasive breast cancer. Participation was restricted to KPNC female members of at least 21 years of age who had no previous history of malignancy other than nonmelanoma skin cancer and spoke English, Spanish, Cantonese, or Mandarin. The study was approved by the Institutional Review Boards of KPNC and Roswell Park Cancer Institute (RPCI), and written informed consent was obtained from all participants.
Analyses presented here include 2953 women who completed the baseline interview, provided blood and/or saliva samples, and were eligible for the 48-month follow-up questionnaire as of September 2, 2014. Mean (standard deviation, SD) time from diagnosis to enrollment was 2.0 (±0.7) months.
Data collection
To determine BCRL status, women were asked the following question on the 12-month follow-up questionnaire: “In the last 6 months, have you had lymphedema (Swelling of the arm and/or hand on the same side where you had your breast cancer surgery)?”, and the date of first being aware of the condition was recorded. Women were asked the same question, but referencing the previous 24 months, on the 48-month follow-up questionnaire.
At the baseline interview, sociodemographic and lifestyle information including race/ethnicity (White, AA, Hispanic, Asian, and Other) was asked by a trained interviewer. Height and weight were collected to calculate BMI (normal < 25 kg/m2; overweight 25–29 kg/m2; obese ≥ 30 kg/m2). Physical activity was assessed using a semiquantitative questionnaire to calculate metabolic equivalent (MET) hours/week of moderate–vigorous activity [10, 11].
Tumor and clinical characteristics were obtained from the KPNC Cancer Registry [12]. Information on breast surgery, adjuvant chemotherapy, radiation therapy, and hormonal therapy were collected from the KPNC Cancer Registry and KPNC electronic medical record.
History of comorbidities and hypertension at the time of breast cancer diagnosis was obtained from the KPNC electronic medical record and other data sources. Comorbidities were calculated using the Charlson comorbidity index [13].
Genetic ancestry
Germline DNA was available from blood or saliva samples collected around the time of the baseline interview the biospecimens were shipped to RPCI and processed under the auspices of the Data Bank and Biorepository, an RPCI Cancer Center Support Grant Shared Resource. Global genetic ancestry was estimated based on a validated panel of 124 ancestry informative markers (AIMs) selected and optimized by Kosoy et al. to distinguish continental origin in the major U.S. populations [14]. The AIMs were designed as part of a customized Illumina GoldenGate 1152-plex chip and genotyped at the RPCI Genomics Shared Resource. The ADMIXTURE program [15] was used to infer genetic ancestry, with ancestral genotype data from the same study that first developed the panel as controls in the estimation.
BCRL outcome
BCRL was considered as any event self-reported by the participant that was at least 6 months or more after the breast cancer diagnosis. This criterion was applied in order to minimize misclassification of BCRL as transient postoperative swelling due to the definitive breast cancer surgery.
Statistical analysis
Univariate analyses of patient and clinical characteristics between those with and without BCRL were conducted using Pearson Chi-square test for categorical variables or Satterthwaite t test for continuous variables. Associations of BCRL risk with patient factors were estimated by hazard ratios (HR) and 95 % confidence intervals (CI) from multivariable Cox proportional hazards models, with follow-up until the date of first self-reported BCRL, or the date of last follow-up questionnaire up to 48 months for those without BCRL. The estimated proportions of African, Asian, and Native American ancestry for each patient were entered in the models as continuous variables, each with a range of 0–1, with European ancestry set as the default reference. Thus, the HRs for each of the three ancestries can be interpreted as the hazards of a “genetically pure” individual of the respective ancestry relative to a “genetically pure” individual of European ancestry.
Models were adjusted initially for education, household income, employment, American Joint Committee on Cancer (AJCC) stage, breast and lymph node surgery, number of lymph nodes examined, chemotherapy, and radiation therapy. The race/ethnicity and ancestry models were further adjusted for age at breast cancer diagnosis, baseline BMI, baseline physical activity, comorbidity, and hypertension, which are the recognized risk factors for BCRL. Finally, stratified analyses were conducted by obesity status (BMI < 30 vs. BMI ≥ 30 and BMI < 25 vs. BMI ≥ 25) to examine whether obesity modified the BCRL racial disparities. To test whether there was possible residual confounding by BMI after stratification, BMI was included in the multivariable models.
Based on the AIM genotypes, we identified outliers in each of the self-reported racial/ethnic populations using EIGENSOFT software [16, 17]. An individual who exceeded four standard deviations along the top three principal components was flagged as an outlier. Outliers were then excluded in sensitivity analyses for the genetic ancestry models.
Results
Mean (SD) time from enrollment to end of follow-up was 38.1 (±17.0) months. A total of 342 (11.6 %) women reported having BCRL up to 48 month follow-up, with 59.7 % White, 10.8 % AA, 12.0 % Hispanic, and 14.3 % Asian (Table 1). Compared to patients with no BCRL, those with BCRL were more likely to be AA (10.8 vs. 6.6 %) and had a nonsignificantly higher proportion of African ancestry (0.80 vs. 0.77). Other factors associated with greater BCRL risk included younger age at breast cancer diagnosis (mean 56 vs. 60 years), higher BMI at baseline (mean 30.6 vs. 28.0 kg/m2), and less moderate–vigorous physical activity at baseline (26.8 vs. 30.3 MET-hours/week). Similar to the BMI pattern, increasing waist-to-hip ratio (WHR) was associated with higher BCRL risk, although the trend across WHR categories was not statistically significant. BCRL was not associated with socioeconomic status (SES) factors, such as education, marital status, household income, and being employed at breast cancer diagnosis, or with lifestyle factors such as smoking and alcohol intake.
Table 1. Patient characteristics of the Pathways Study cohort by self-reported BCRL status at 12 and 48 months postbreast cancer diagnosis (n = 2953).
BCRL from self-reporta | |||
---|---|---|---|
| |||
No n = 2611 (88.4 %) | Yes n = 342 (11.6 %) | p valueb | |
N (%) | N (%) | ||
Race/ethnicity | 0.0093 | ||
White | 1765 (67.6) | 204 (59.7) | |
African American | 173 (6.6) | 37 (10.8) | |
Hispanic | 312 (11.9) | 41 (12.0) | |
Asian | 302 (11.6) | 49 (14.3) | |
Other | 59 (2.3) | 11 (3.2) | |
Estimated ancestry, median (range) | |||
% European ancestry among whites | 0.88 (0.03–1.00) | 0.88 (0.46–1.00) | 0.38 |
% African ancestry among African Americans | 0.77 (0.06–0.98) | 0.80 (0.01–0.95) | 0.13 |
% Asian ancestry among Asians | 0.83 (0.01–1.00) | 0.81 (0.02–1.00) | 0.14 |
% Native American ancestry among Hispanics | 0.23 (0-0.83) | 0.23 (0–0.74) | 0.88 |
Age at breast cancer diagnosis | <0.0001 | ||
<50 years | 550 (21.1) | 108 (31.6) | |
50–59 years | 765 (29.3) | 126 (36.8) | |
60–69 years | 768 (29.4) | 68 (19.9) | |
≥70 years | 528 (20.2) | 40 (11.7) | |
Mean (SD) | 60.0 (11.9) | 56.0 (11.1) | <0.0001 |
Education at baseline | 0.15 | ||
High school or less | 416 (15.9) | 43 (12.5) | |
Some college | 916 (35.1) | 122 (35.7) | |
College graduate | 699 (26.8) | 108 (31.6) | |
Postgraduate | 580 (22.2) | 69 (20.2) | |
Marital status at baseline | 0.52 | ||
Married/marriage-like | 1598 (61.3) | 215 (63.1) | |
Single/separated/widowed | 1011 (38.7) | 126 (36.9) | |
Household income ($) at baseline | 0.57 | ||
<25,000 | 259 (11.1) | 28 (9.2) | |
25,000–49,999 | 516 (22.1) | 61 (20.1) | |
50,000–89,999 | 743 (31.8) | 99 (32.7) | |
>90,000 | 820 (35.0) | 115 (38.0) | |
Employment status at breast cancer diagnosis | 0.97 | ||
Not working | 1369 (53.0) | 176 (52.9) | |
Working | 1215 (47.0) | 157 (47.1) | |
BMI (kg/m2) at baseline | <0.0001 | ||
<25 (normal weight) | 963 (37.2) | 87 (25.6) | |
25–29 (overweight) | 814 (31.4) | 99 (29.1) | |
≥30 (obese) | 815 (31.4) | 154 (45.3) | |
Mean (SD) | 28.0 (6.4) | 30.6 (7.9) | <0.0001 |
Waist-to-hip ratio (WHR) at baseline | 0.04 | ||
Tertile 1 (<0.832) | 894 (35.0) | 98 (29.5) | |
Tertile 2 (0.832–0.926) | 850 (33.2) | 107 (32.2) | |
Tertile 3 (>0.926) | 813 (31.8) | 127 (38.3) | |
Mean (SD) | 0.87 (0.08) | 0.88 (0.08) | 0.072 |
Moderate-to-vigorous physical activity (MET-hours/week) at baseline | 0.03 | ||
Quartile 1 (<8.2) | 644 (24.8) | 96 (28.2) | |
Quartile 2 (8.2–21.1) | 649 (25.0) | 81 (23.8) | |
Quartile 3 (21.2–42.5) | 641 (24.7) | 99 (29.0) | |
Quartile 4 (>42.5) | 664 (25.5) | 65 (19.0) | |
Mean (SD) | 30.3 (31.4) | 26.8 (27.9) | 0.046 |
Smoking status at baseline | 0.14 | ||
Never | 1433 (54.9) | 210 (61.4) | |
Former | 1047 (40.1) | 119 (34.8) | |
Current | 130 (5.0) | 13 (3.8) | |
Alcohol intake at baseline (g/day) | 0.09 | ||
Never | 569 (28.0) | 91 (34.6) | |
Below (or equal) median | 415 (20.5) | 47 (17.9) | |
Above median | 1043 (51.5) | 125 (47.5) | |
Mean (SD) | 6.5 (11.7) | 5.6 (10.8) | 0.20 |
Self-reported BCRL from the 12-month or 48-month Health Status Updates
From Pearson Chi-square test (categorical variables) and Satterthwaite t-test (continuous variables)
Not reported, missing, unknown are as follows: Marital status n = 3, Household income n = 312, Employment status n = 36, BMI at baseline n = 21, WHR at baseline n = 64, Moderate-to-Vigorous physical activity n = 14, Alcohol intake at baseline n = 663
Several tumor and clinical characteristics were associated with BCRL risk (Table 2). More advanced stage at diagnosis, mastectomy, greater number of lymph nodes examined, greater number of positive lymph nodes, and chemotherapy were each associated with elevated risk (p < 0.05). Women who had no radiation therapy (69.3 %) compared with those who had radiation therapy (30.7 %) were more likely to report having BCRL (p < 0.05). Among patients who reported BCRL compared with those who did not, sentinel lymph node biopsy only (8.4 vs. 17.0 %) was less common, whereas axillary lymph node dissection only (19.3 vs. 16.8 %) or both (54.2 vs. 49.9 %) were more common (p = 0.0013).
Table 2. Clinical characteristics of the Pathways Study cohort by self-reported BCRL status at 12 and 48 months postbreast cancer diagnosis (n = 2953).
BCRL from self-reporta | |||
---|---|---|---|
| |||
No n = 2611 (88.4 %) | Yes n = 342 (11.6 %) | p valueb | |
N (%) | N (%) | ||
AJCC stage at breast cancer diagnosis | <0.0001 | ||
I | 1475 (56.5) | 89 (26.0) | |
II | 883 (33.8) | 161 (47.1) | |
III | 215 (8.2) | 90 (26.3) | |
IV | 38 (1.5) | 2 (0.6) | |
Breast surgeryc | <0.0001 | ||
Lumpectomy only | 1628 (62.4) | 161 (47.1) | |
Mastectomy | 961 (36.8) | 177 (51.7) | |
None | 21 (0.8) | 4 (1.2) | |
Lymph node surgery | 0.0013 | ||
Axillary lymph node dissection | 404 (16.8) | 62 (19.3) | |
Sentinel lymph node biopsy | 409 (17.0) | 27 (8.4) | |
Both | 1199 (49.9) | 174 (54.2) | |
none | 389 (16.2) | 58 (18.1) | |
Number of lymph nodes examined | <0.0001 | ||
Mean [SD (range)] | 5.6 (6.3 [0–50]) | 11.2 (8.4 [0–44]) | |
Number of positive lymph nodes | |||
Mean [SD (range)] | 1.0 (3.1 [0–49]) | 2.6 (4.3 [0–32]) | <0.0001 |
Chemotherapy | <0.0001 | ||
No | 1401 (53.9) | 72 (21.1) | |
Yes | 1198 (46.1) | 270 (78.9) | |
Radiation therapy | <0.0001 | ||
No | 1494 (57.2) | 237 (69.3) | |
Yes | 1117 (42.8) | 105 (30.7) | |
Hormonal therapy | 0.09 | ||
No | 640 (24.7) | 99 (28.9) | |
Yes | 1954 (75.3) | 243 (71.1) | |
Comorbidity history at BC diagnosis | 0.22 | ||
No | 2239 (85.9) | 302 (88.3) | |
Yes | 369 (14.2) | 40 (11.7) | |
Hypertension history at BC diagnosis | 0.60 | ||
No | 1419 (54.3) | 191 (55.9) | |
Yes | 1192 (45.7) | 151 (44.2) |
Self-reported BCRL from the 12-month or 48-month Health Status Updates
From Pearson Chi-square test (categorical variables) and Satterthwaite t test (continuous variables)
Bilateral surgery, n = 146 (4.94 %), Unilateral surgery, n = 2767 (93.70 %), No surgery, n = 26 (0.88 %), NOS/Unknown, n = 14 (0.47 %)
Not reported, missing, unknown are as follows: Lymph node surgery n = 231, Chemotherapy n = 12, Hormonal therapy n = 17, Comorbidity history n = 3
In the unadjusted model, AA women had a twofold increased risk of BCRL compared with White women (HR = 2.04; 95 % CI: 1.35, 3.08) (Table 3), also shown in cumulative incidence curves (Fig. 1a). The increased risk was attenuated (HR = 1.60, 95 % CI 1.04, 2.47) in the base model after adjusting for standard covariates. Upon further adjustment for BCRL risk factors including age at BC diagnosis, baseline BMI, baseline physical activity, comorbidity, and hypertension, the association was further attenuated (HR = 1.47; 95 % CI 0.94, 2.30). Identification with other racial/ethnic groups was not associated with BCRL risk in the adjusted models, except for an increased risk in Hispanics (HR = 1.47; 95 % CI 1.00, 2.15), which was substantially attenuated after adjustment for covariates (HR = 1.18; 95 % CI 0.78, 1.77).
Table 3. Associations of self-reported race/ethnicity with self-reported BCRL status.
n | # Eventsa | Unadjusted model | Base Modelb | Further adjusted for age BC Dx, baseline BMI, baseline PA, comorbidity, HTN | |
---|---|---|---|---|---|
HR (95 % CI) | HR (95 % CI) | HR (95 % CI) | |||
Race/ethnicity from self-report | |||||
White | 1969 | 133 | Ref | Ref | Ref |
African American | 210 | 27 | 2.04 (1.35, 3.08) | 1.60 (1.04, 2.47) | 1.47 (0.94, 2.30) |
Hispanic | 351 | 33 | 1.47 (1.00, 2.15) | 1.28 (0.86, 1.90) | 1.18 (0.78, 1.77) |
Asian | 353 | 30 | 1.31 (0.88, 1.95) | 1.18 (0.78, 1.79) | 1.24 (0.81, 1.92) |
Other | 70 | 7 | 1.57 (0.73, 3.36) | 1.20 (0.55, 2.59) | 1.21 (0.56, 2.61) |
Genetic ancestryc | |||||
European ancestry | – | – | Ref | Ref | |
African ancestry | – | – | 2.50 (1.43, 4.36) | 1.93 (1.08, 3.43) | 1.73 (0.96, 3.14) |
Native American ancestry | – | – | 2.70 (0.79, 9.23) | 1.91 (0.52, 7.07) | 1.33 (0.32, 5.51) |
Asian ancestry | – | – | 1.23 (0.73, 2.05) | 1.14 (0.66, 1.94) | 1.21 (0.69, 2.11) |
Follow-up starting at 6 months postbreast cancer diagnosis until BCRL self-report or date of last patient contact (12- or 48-month follow-up), whichever occurred first
Cox proportional hazards models, adjusted for education, household income, employment, AJCC stage, breast and lymph node surgery, number of lymph nodes examined, chemotherapy, radiation therapy
The estimated percentages of African, Asian and Native American ancestry for each patient were entered in the models as continuous, each with a range of 0–1 and European ancestry set as the default reference. Thus, the HRs with each of the three ancestries should be interpreted as the hazards of a “genetically pure” individual of the respective ancestry relative to a genetically pure individual of European ancestry
Fig. 1. Kaplan–Meier curve showing cumulative incidence of BCRL in the entire cohort (a), nonobese only (b) and obese only (c) by self-reported race/ethnicity.
Consistent with self-reported race, African genetic ancestry was associated with an increased risk of BCRL in the unadjusted model (HR = 2.50; 95 % CI 1.43, 4.36), which was attenuated in the base model (HR = 1.93; 95 % CI 1.08, 3.43), relative to European ancestry. After further covariate adjustment, the association was again less striking (HR = 1.73, 95 % CI 0.96–3.14) (Table 3). Native American and Asian ancestries were not associated with BCRL risk compared with European ancestry.
In stratified models by obesity status at baseline (BMI < 30 kg/m2 vs. ≥30 kg/m2), AA women who were nonobese had a statistically significant elevated risk of BCRL in all models (fully adjusted HR = 2.20; 95 % CI 1.16, 4.19) (Table 4; Fig. 1b). However, there was no association with BCRL among obese AA women (HR = 1.21; 95 % CI 0.65, 2.26) (Fig. 1c). Consistent with self-reported AA race, African ancestry was associated with an increased BCRL risk among the nonobese (fully adjusted HR = 3.51; 95 % CI 1.50, 8.21), but not among the obese (fully adjusted HR = 1.14; 95 % CI 0.49, 2.65). Similar to AA women but to a lesser extent, Hispanic women who were nonobese had a nearly twofold increased risk of BCRL (HR = 1.96; 95 % CI 1.22, 3.15) (Table 4; Fig. 1b), even after adjustment for covariates (HR = 1.59; 95 % CI 0.95, 2.66). No such association was observed in their obese counterparts (HR = 0.70; 95 % CI 0.35, 1.43). In unadjusted models among the nonobese women, Native American ancestry was associated with increased risk of BCRL. However, due to small sample size and low percentage of Native American ancestry present in the study population, the CI was very wide and the risk estimate was nonsignificant after adjustment for covariates. Neither self-reported Asian race/ethnicity nor Asian ancestry was associated with BCRL in either the obese or nonobese group.
Table 4. Associations of self-reported race/ethnicity and genetic ancestry with self-reported BCRL status by BMI (nonobese vs. obese).
NonObese (BMI < 30) n = 1810 | Obese (BMI ≥ 30) n = 930 | |||||||||
---|---|---|---|---|---|---|---|---|---|---|
|
|
|||||||||
n | # Eventsa | Unadjusted model | Base modelb | Further adjusted for age BC Dx, baseline BMI, baseline PA, comorbidity, HTN | n | # eventsa | Unadjusted Model | Base modelb | Further adjusted for age BC Dx, baseline BMI, baseline PA, comorbidity, HTN | |
HR (95 % CI) | HR (95 % CI) | HR (95 % CI) | HR (95 % CI) | HR (95 % CI) | HR (95 % CI) | |||||
Race/ethnicity from self-report | ||||||||||
White | 1188 | 72 | Ref | Ref | Ref | 593 | 52 | Ref | Ref | Ref |
African American | 94 | 12 | 2.33 (1.27, 4.29) | 2.33 (1.25, 4.36) | 2.20 (1.16, 4.19 | 114 | 15 | 1.62 (0.92, 2.88) | 1.22 (0.66, 2.27) | 1.21 (0.65, 2.26) |
Hispanic | 199 | 21 | 1.96 (1.22, 3.15) | 1.73 (1.05, 2.84) | 1.59 (0.95, 2.66) | 139 | 9 | 0.86 (0.44, 1.69) | 0.73 (0.36, 1.48) | 0.70 (0.35, 1.43) |
Asian | 292 | 22 | 1.29 (0.80, 2.07) | 1.12 (0.68, 1.84) | 1.07 (0.64, 1.81) | 52 | 8 | 1.93 (0.92, 4.06) | 1.80 (0.83, 3.90) | 1.75 (0.80, 3.85) |
Genetic Ancestryc | ||||||||||
European ancestry | – | – | Ref | Ref | Ref | – | – | Ref | Ref | Ref |
African ancestry | – | – | 3.63 (1.63, 8.06) | 3.67 (1.61, 8.39) | 3.51 (1.50, 8.21) | – | – | 1.52 (0.69, 3.32) | 1.27 (0.43, 3.77) | 1.14 (0.49, 2.65) |
Native American ancestry | – | – | 6.66 (1.49, 29.7) | 4.10 (0.80, 21.0) | 3.02 (0.56, 16.25) | – | – | 0.26 (0.02, 3.29) | 0.19 (0.01, 2.63) | 0.16 (0.01, 2.54) |
Asian ancestry | – | – | 1.35 (0.74, 2.46) | 1.24 (0.66, 2.34) | 1.20 (0.62, 2.33) | – | – | 1.47 (0.50, 4.32) | 1.27 (0.43, 3.77) | 1.32 (0.43, 4.04) |
Follow-up starting at 6 months postbreast cancer diagnosis until BCRL self-report or date of last patient contact (12- or 48-month follow-up), whichever occurred first
Adjusted for education, household income, employment, AJCC stage, breast and lymph node surgery, number of lymph nodes examined, chemotherapy, radiation therapy
The estimated percentages of African, Asian and Native American ancestry for each patient were entered in the models as continuous, each with a range of 0-1 and European ancestry set as the default reference. Thus, the HRs with each of the three ancestries should be interpreted as the hazards of a “genetically pure” individual of the respective ancestry relative to a genetically pure individual of European ancestry
When models were stratified by overweight status at baseline (BMI < 25 kg/m2 vs. ≥25 kg/m2), the associations had a similar pattern of increased BCRL risk with self-reported AA race/ethnicity and African ancestry among the nonoverweight women (data not shown). When genetic ancestry outliers were excluded from the models of genetic ancestry (n = 30), the results remained the same (data not shown).
Discussion
In this prospective study of BCRL in 2953 women with breast cancer, we confirm the risk factors for BCRL from previous studies, including younger age at breast cancer diagnosis, higher BMI, and lower physical activity. Further, we report increased risks of BCRL among patients who self-reported their race as AA, and a similar increased risk associated with estimated African genetic ancestry, both of which could not be entirely explained by known BCRL risk factors. Intriguingly, these increased risks of BCRL appeared only among the nonobese women at breast cancer diagnosis. No associations were observed among the corresponding obese groups. To our knowledge, this is the first study to comprehensively examine potential racial/ethnic differences in risk of BCRL using both self-reported race/ethnicity and genetic ancestry information, and to investigate the interaction between race and obesity.
We previously reported a twofold increased risk of BCRL among AA women in our study cohort [6]. However, that analysis included only the first 997 enrolled women, with mean follow-up time of 20.9 months (0.7–31.8) after breast cancer diagnosis. Our current study builds upon the original study by increasing the sample size (n = 2953), having a longer mean follow-up of 38.5 months since breast cancer diagnosis, and including genetic ancestry markers. Our previous study also differed from the current analysis in ascertainment of BCRL. Previously, a BCRL diagnosis was determined by clinical indication from outpatient or hospitalization diagnosis codes, outpatient procedure codes, and durable medical equipment orders. In the present study, we focused on BCRL as a patient-reported outcome because of potential limitations in accurately classifying BCRL diagnoses by diagnosis and treatment codes using electronic data sources [18]. We conducted a validation analysis of the patient's self-report and compared with medical chart review in our analytic cohort (n = 200). While sensitivity and negative predictive value (NPV) were 92 and 98 %, respectively, specificity and positive predictive value (PPV) were lower at 56 and 22 %, respectively (unpublished data). When comparing ICD-9 diagnosis codes for lymphedema against chart review, the sensitivity, specificity, PPV, and NPV were 67, 52, 16, and 92 %, respectively.
A principal strength of our current study is the investigation of genetic ancestry in parallel with self-reported race/ethnicity. For all risk estimates, those associated with African genetic ancestry were consistently stronger and more robust to adjustment for known BCRL risk factors than those with self-reported African American race/ethnicity. These findings not only confirm the elevated risk observed in self-reported AA women, but also suggest that the this racial disparity in BCRL risk might be due, in part, to biological differences.
Obesity might contribute to some of these biological differences, given that it is an established BCRL risk factor with differences in prevalence among AA and White women, 45 and 31 % in our study, respectively. There are also reported differences in the relationship of BMI with body composition between AA and White women. Specifically, AA women are more likely to have higher lean mass and lower fat [19], and lower visceral adipose tissue and higher subcutaneous adipose tissue for a given BMI compared with White women [20]. Furthermore, differences in levels of obesity-related circulating adipokines and inflammatory biomarkers have been reported in AA and White women after adjusting for BMI, with AA women having higher levels of leptin, C-reactive protein and interleukin-6, and lower levels of adiponectin [21]. Given that the lymphatic system is critical in management of inflammatory, immune function, and obesity processes [22], one could hypothesize that these varying levels of obesity-related biomarkers could be differentially affecting the lymphatic vessels of AA and White women, contributing to our observed elevated BCRL risk in AA women.
However, when we examined the associations separately among the nonobese and obese women, BCRL risk was higher in the nonobese, but not in the obese AA women and those with African ancestry. This is counterintuitive to the fairly established association of greater BMI, especially ≥30 kg/m2, being associated with increased risk of BCRL [5, 23–25]. Nevertheless, our findings are conceptually similar to an earlier study of smoking-related lung cancer, which demonstrated a significantly higher risk in AAs than in Whites only among light and moderate smokers, but not among heavy smokers [26]. Specifically, the race- and ancestry–environment interaction might be explained by a hypothesis that any racial differences due to genetic susceptibility, such as the aforementioned differences in obesity-related adipokines and inflammation among AA and White women, may only be apparent without the presence of strong environmental risk factors, such as heavy smoking in the lung cancer study, or obesity in our current BCRL study. However, when such risk factors are present and become the major causal factors, differences in genetic susceptibility may be saturated by those exposures. These findings emphasize the importance of gene–environment interactions in health-related disparities and suggest striking disparities could exist among populations considered to be at low risk (e.g., nonobese women for BCRL). Further research is needed to understand the underlying mechanisms of these processes.
However, obesity is unlikely the only factor contributing to the observed BCRL racial disparity. In a recent study, Togawa et al. showed that a 1.6-fold increased risk of BCRL in AA women was almost entirely eliminated after adjusting for BMI, hypertension, and treatment-related factors [8]. However, in our analyses, controlling for these and other risk factors attenuated but did not completely explain the elevated risk in AA women or with African ancestry. The elevated risk became even more pronounced among nonobese women when adjusting for all other risk factors.
The proportion of women reporting BCRL in our study (11.6 %) is somewhat lower than that reported in the literature (8.4–21.4 %) [4]. Misclassification of BCRL status is certainly possible; however, our analyses excluded any reported BCRL event within the first 6 months of breast cancer diagnosis to minimize misclassification due to postsurgical swelling. Although we assessed self-reported BCRL only up to 48 months after cancer diagnosis, it was recently reported that BCRL incidence primarily occurs within the first five years [8]. While we fully acknowledge that self-report should ideally be validated against physical measures such as perometry, bioimpedance, or arm circumference measures, our proportion of women reporting BCRL was in the range of what has been reported in the literature. Alternatively, given that sentinel lymph node biopsy (SLNB) use is associated with lower BCRL risk [4, 5, 27, 28], this difference could be attributed to more SLNB being done in our contemporary cohort compared with older cohorts prior to the full endorsement of SLNB as the preferred method of axillary staging from 2005 to 2007 [29–31]. Indeed, women in our cohort were diagnosed with breast cancer from 2005 to 2013, and 77 % had SLNB with or without axillary lymph node dissection, compared with 4 % [32], 48 % [33], and 64 % [25] in other studies.
Interestingly, notable racial/ethnic differences in use of SLNB were recently reported in a large Surveillance, Epidemiology, and End Results Program (SEER)—Medicare study from 2002 to 2007 [34]. AA breast cancer patients were less likely to undergo SLNB compared with their White counterparts (odds ratio = 0.67; 95 % CI 0.60–0.75), which contributed to a statistically significant difference in 5-year cumulative risk of BCRL in Whites (8.2 %) vs. AAs (12.3 %) (p < 0.001). However, in our analysis, we did account for the type of lymph node surgery received, and the increased risk of BCRL remained among AA and with African ancestry.
To conclude, in this prospective cohort study of 2953 women with breast cancer, AA women had a greater risk of self-reported BCRL compared to White women. Moreover, African genetic ancestry was also linked with an increased risk of BCRL compared to European ancestry. This increased risk appears to be most pronounced for those who were nonobese at the time of breast cancer diagnosis. When identifying who might be at higher BCRL risk in breast cancer survivors, race/ethnicity might also be considered, along with the more established risk factors of age, BMI, and surgery, to direct appropriate surveillance or teach risk reduction interventions to these higher risk groups.
Acknowledgments
The authors thank the office and field staff for data collection, processing, and preparation. The authors thank all the Pathways Study participants for their numerous contributions to this study. This work was supported by the National Cancer Institute at the National Institutes of Health (Grant Number: R01 CA105274 to L.H.K) and the American Cancer Society (Grant Number: RSG-06-209-01-LR to M.L.K). The Roswell Park Cancer Institute Data Bank and Biorepository and Genomics Shared Resource are supported by a National Cancer Institute Cancer Center Support Grant (Grant Number: P30 CA16056). The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the funding agencies.
Abbreviations
- AA
African American
- AIM
Ancestry informative marker
- AJCC
American Joint Committee on Cancer
- BC
Breast cancer
- BCRL
Breast cancer-related lymphedema
- BMI
Body mass index
- CI
Confidence interval
- HR
Hazards ratio
- KPNC
Kaiser Permanente Northern California
- MET
Metabolic equivalent
- RPCI
Roswell Park Cancer Institute
- SD
Standard deviation
- SEER
Surveillance, epidemiology, and end results program
- SES
Socioeconomic status
- SLNB
Sentinel lymph node biopsy
- WHR
Waist-to-hip ratio
Footnotes
Compliance with ethical standards: Conflict of Interest The authors declare that they have no conflict of interest.
References
- 1.Erickson VS, Pearson ML, Ganz PA, Adams J, Kahn KL. Arm edema in breast cancer patients. J Natl Cancer Inst. 2001;93(2):96–111. doi: 10.1093/jnci/93.2.96. [DOI] [PubMed] [Google Scholar]
- 2.Lawenda BD, Mondry TE, Johnstone PA. Lymphedema: a primer on the identification and management of a chronic condition in oncologic treatment. CA Cancer J Clin. 2009;59(1):8–24. doi: 10.3322/caac.20001. [DOI] [PubMed] [Google Scholar]
- 3.Pusic AL, Cemal Y, Albornoz C, Klassen A, Cano S, Sulimanoff I, Hernandez M, Massey M, Cordeiro P, Morrow M, Mehrara B. Quality of life among breast cancer patients with lymphedema: a systematic review of patient-reported outcome instruments and outcomes. J Cancer Surviv. 2013;7(1):83–92. doi: 10.1007/s11764-012-0247-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.DiSipio T, Rye S, Newman B, Hayes S. Incidence of unilateral arm lymphoedema after breast cancer: a systematic review and meta-analysis. Lancet Oncol. 2013;14(6):500–515. doi: 10.1016/S1470-2045(13)70076-7. [DOI] [PubMed] [Google Scholar]
- 5.Paskett ED, Dean JA, Oliveri JM, Harrop JP. Cancer-related lymphedema risk factors, diagnosis, treatment, and impact: a review. J Clin Oncol. 2012;30(30):3726–3733. doi: 10.1200/JCO.2012.41.8574. [DOI] [PubMed] [Google Scholar]
- 6.Kwan ML, Darbinian J, Schmitz KH, Citron R, Partee P, Kutner SE, Kushi LH. Risk factors for lymphedema in a prospective breast cancer survivorship study: the Pathways Study. Arch Surg. 2010;145(11):1055–1063. doi: 10.1001/archsurg.2010.231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Meeske KA, Sullivan-Halley J, Smith AW, McTiernan A, Baumgartner KB, Harlan LC, Bernstein L. Risk factors for arm lymphedema following breast cancer diagnosis in Black women and White women. Breast Cancer Res Treat. 2009;113(2):383–391. doi: 10.1007/s10549-008-9940-5. [DOI] [PubMed] [Google Scholar]
- 8.Togawa K, Ma H, Sullivan-Halley J, Neuhouser ML, Imayama I, Baumgartner KB, Smith AW, Alfano CM, McTiernan A, Ballard-Barbash R, Bernstein L. Risk factors for self-reported arm lymphedema among female breast cancer survivors: a prospective cohort study. Breast Cancer Res. 2014;16(4):414. doi: 10.1186/s13058-014-0414-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kwan ML, Ambrosone CB, Lee MM, Barlow J, Krathwohl SE, Ergas IJ, Ashley CH, Bittner JR, Darbinian J, Stronach K, Caan BJ, Davis W, Kutner SE, Quesenberry CP, Somkin CP, Sternfeld B, Wiencke JK, Zheng S, Kushi LH. The Pathways Study: a prospective study of breast cancer survivorship within Kaiser Permanente Northern California. Cancer Causes Control. 2008;19(10):1065–1076. doi: 10.1007/s10552-008-9170-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Staten LK, Taren DL, Howell WH, Tobar M, Poehlman ET, Hill A, Reid PM, Ritenbaugh C. Validation of the Arizona activity frequency questionnaire using doubly labeled water. Med Sci Sports Exerc. 2001;33(11):1959–1967. doi: 10.1097/00005768-200111000-00024. [DOI] [PubMed] [Google Scholar]
- 11.Sternfeld B, Weltzien E, Quesenberry CP, Jr, Castillo AL, Kwan M, Slattery ML, Caan BJ. Physical activity and risk of recurrence and mortality in breast cancer survivors: findings from the LACE study. Cancer Epidemiol Biomark Prev. 2009;18(1):87–95. doi: 10.1158/1055-9965.EPI-08-0595. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Oehrli MD, Quesenberry CP. Annual Report on Trends, Incidence, and Outcomes, 2013. Kaiser Permanente, Northern California Cancer Registry; 2013. [Google Scholar]
- 13.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Diseases. 1987;40(5):373–383. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 14.Kosoy R, Nassir R, Tian C, White PA, Butler LM, Silva G, Kittles R, Alarcon-Riquelme ME, Gregersen PK, Belmont JW, De La Vega FM, Seldin MF. Ancestry informative marker sets for determining continental origin and admixture proportions in common populations in America. Hum Mutat. 2009;30(1):69–78. doi: 10.1002/humu.20822. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Alexander DH, Novembre J, Lange K. Fast model-based estimation of ancestry in unrelated individuals. Genome Res. 2009;19(9):1655–1664. doi: 10.1101/gr.094052.109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Patterson N, Price AL, Reich D. Population structure and eigenanalysis. PLoS Genet. 2006;2(12):e190. doi: 10.1371/journal.pgen.0020190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet. 2006;38(8):904–909. doi: 10.1038/ng1847. [DOI] [PubMed] [Google Scholar]
- 18.Yen TW, Laud PW, Sparapani RA, Li J, Nattinger AB. An algorithm to identify the development of lymphedema after breast cancer treatment. J Cancer Surviv. 2014;9:161–171. doi: 10.1007/s11764-014-0393-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Flegal KM, Shepherd JA, Looker AC, Graubard BI, Borrud LG, Ogden CL, Harris TB, Everhart JE, Schenker N. Comparisons of percentage body fat, body mass index, waist circumference, and waist-stature ratio in adults. Am J Clin Nutr. 2009;89(2):500–508. doi: 10.3945/ajcn.2008.26847. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Katzmarzyk PT, Bray GA, Greenway FL, Johnson WD, Newton RL, Jr, Ravussin E, Ryan DH, Smith SR, Bouchard C. Racial differences in abdominal depot-specific adiposity in white and African American adults. Am J Clin Nutr. 2010;91(1):7–15. doi: 10.3945/ajcn.2009.28136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Morimoto Y, Conroy SM, Ollberding NJ, Kim Y, Lim U, Cooney RV, Franke AA, Wilkens LR, Hernandez BY, Goodman MT, Henderson BE, Kolonel LN, Le Marchand L, Maskarinec G. Ethnic differences in serum adipokine and C-reactive protein levels: the multiethnic cohort. Int J Obes. 2014;38(11):1416–1422. doi: 10.1038/ijo.2014.25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Tammela T, Alitalo K. Lymphangiogenesis: molecular mechanisms and future promise. Cell. 2010;140(4):460–476. doi: 10.1016/j.cell.2010.01.045. [DOI] [PubMed] [Google Scholar]
- 23.Ahmed RL, Schmitz KH, Prizment AE, Folsom AR. Risk factors for lymphedema in breast cancer survivors, the Iowa Women's Health Study. Breast Cancer Res Treat. 2011;130(3):981–991. doi: 10.1007/s10549-011-1667-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dominick SA, Madlensky L, Natarajan L, Pierce JP. Risk factors associated with breast cancer-related lymphedema in the WHEL Study. J Cancer Surviv. 2013;7(1):115–123. doi: 10.1007/s11764-012-0251-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Jammallo LS, Miller CL, Singer M, Horick NK, Skolny MN, Specht MC, O'Toole J, Taghian AG. Impact of body mass index and weight fluctuation on lymphedema risk in patients treated for breast cancer. Breast Cancer Res Treat. 2013;142(1):59–67. doi: 10.1007/s10549-013-2715-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Haiman CA, Stram DO, Wilkens LR, Pike MC, Kolonel LN, Henderson BE, Le Marchand L. Ethnic and racial differences in the smoking-related risk of lung cancer. N Engl J Med. 2006;354(4):333–342. doi: 10.1056/NEJMoa033250. [DOI] [PubMed] [Google Scholar]
- 27.Miller CL, Specht MC, Skolny MN, Horick N, Jammallo LS, O'Toole J, Shenouda MN, Sadek BT, Smith BL, Taghian AG. Risk of lymphedema after mastectomy: potential benefit of applying ACOSOG Z0011 protocol to mastectomy patients. Breast Cancer Res Treat. 2014;144(1):71–77. doi: 10.1007/s10549-014-2856-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Tsai RJ, Dennis LK, Lynch CF, Snetselaar LG, Zamba GK, Scott-Conner C. The risk of developing arm lymphedema among breast cancer survivors: a meta-analysis of treatment factors. Ann Surg Oncol. 2009;16(7):1959–1972. doi: 10.1245/s10434-009-0452-2. [DOI] [PubMed] [Google Scholar]
- 29.Lyman GH, Giuliano AE, Somerfield MR, Benson AB, 3rd, Bodurka DC, Burstein HJ, Cochran AJ, Cody HS, 3rd, Edge SB, Galper S, Hayman JA, Kim TY, Perkins CL, Podoloff DA, Sivasubramaniam VH, Turner RR, Wahl R, Weaver DL, Wolff AC, Winer EP. American Society of Clinical Oncology guideline recommendations for sentinel lymph node biopsy in early-stage breast cancer. J Clin Oncol. 2005;23(30):7703–7720. doi: 10.1200/JCO.2005.08.001. [DOI] [PubMed] [Google Scholar]
- 30.Lyman GH, Temin S, Edge SB, Newman LA, Turner RR, Weaver DL, Benson AB, 3rd, Bosserman LD, Burstein HJ, Cody H, 3rd, Hayman J, Perkins CL, Podoloff DA, Giuliano AE. Sentinel lymph node biopsy for patients with early-stage breast cancer: American Society of Clinical Oncology clinical practice guideline update. J Clin Oncol. 2014;32(13):1365–1383. doi: 10.1200/JCO.2013.54.1177. [DOI] [PubMed] [Google Scholar]
- 31.NCCN clinical practice guidelines in oncology. [Accessed 13 Mar 2015]; http://www.nccn.org/store/login/login.aspx?ReturnURL=http://www.nccn.org/professionals/physician_gls/pdf/breast.pdf.
- 32.Paskett ED, Naughton MJ, McCoy TP, Case LD, Abbott JM. The epidemiology of arm and hand swelling in premenopausal breast cancer survivors. Cancer Epidemiol Biomark Prev. 2007;16(4):775–782. doi: 10.1158/1055-9965.EPI-06-0168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Norman SA, Localio AR, Kallan MJ, Weber AL, Torpey HA, Potashnik SL, Miller LT, Fox KR, DeMichele A, Solin LJ. Risk factors for lymphedema after breast cancer treatment. Cancer Epidemiol Biomark Prev. 2010;19(11):2734–2746. doi: 10.1158/1055-9965.EPI-09-1245. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Black DM, Jiang J, Kuerer HM, Buchholz TA, Smith BD. Racial disparities in adoption of axillary sentinel lymph node biopsy and lymphedema risk in women with breast cancer. JAMA Surg. 2014;149(8):788–796. doi: 10.1001/jamasurg.2014.23. [DOI] [PMC free article] [PubMed] [Google Scholar]