Lymphedema Signs, Symptoms, and Diagnosis in Women Who Are in Minority and Low-Income Groups and Have Survived Breast Cancer

Abstract

Background

Breast cancer–related lymphedema (BCRL) is a well-known side effect of cancer and its treatment with wide-ranging prevalence estimates.

Objective

This study describes associations between breast cancer–related lymphedema (BCRL) signs, symptoms, and diagnosis for women who were African American, white, or had a low income and survived breast cancer.

Design

This is a cross-sectional, observational study that used a computer-assisted telephone interview.

Methods

Women who had survived breast cancer were queried on the presence of 5 lymphedema signs and symptoms (edema in the breast, axilla, arm, and/or hand; tissue fibrosis; pitting; hemosiderin staining; heaviness) and whether they had a diagnosis of BCRL. Relationships between signs/symptoms and diagnosis for each group were evaluated with kappa and chi-square statistics.

Results

The study sample included 528 women who had survived breast cancer (266 white and 262 African American), with 514 reporting complete data on household income; 45% of the latter reported an annual household income of ≤$20,000. Women who were African American or had a low income were nearly twice as likely as women who were white to have any of 8 signs/symptoms of BCRL. Regardless of race and income, >50% of women with all BCRL signs and symptoms reported that they were not diagnosed with BCRL.

Limitations

The main limitations of our study are the lack of medical chart data and longitudinal design.

Conclusions

Women who were African American or had a low income and had survived breast cancer had a greater burden of BCRL signs and symptoms than women who were white. The lack of a strong association between BCRL signs, symptoms, and diagnosis suggests that BCRL may be underdiagnosed. These findings suggest that more rigorous screening and detection of BCRL—especially for women who are African American or have a low income—may be warranted. Cancer rehabilitation programs may be able to fill this gap.

Breast cancer–related lymphedema (BCRL) is a well-known side effect of cancer and its treatment with wide-ranging prevalence estimates.^1,2 There is no general consensus on prevalence rates of BCRL and previous studies have found very different prevalence rates across populations.^1–7 To illustrate, 1 study shows a prevalence of 0% to 65% among those receiving radiation and concluded that radiation contributes to a higher risk of developing BCRL.³ Other research shows prevalences anywhere between 13% and 65%,⁴ depending on the type of surgical resection, axillary versus sentinel node biopsy, and measurement method.^5–7 Two other studies showed prevalence rates from 8% to 31% and a wider range of 3% to 87%.^8,9 The fact that none of these studies had large sample sizes or included minorities or people of low socioeconomic status makes generalization difficult.

BCRL is a clinical diagnosis based on patient reports of signs and symptoms and physical examination. In addition to the presence of edema, skin texture and color changes, pitting, and feelings of heaviness in the affected area are also diagnostic hallmarks of lymphedema.^4,10,11 Although several diagnostic criteria exist to arrive at a diagnosis of lymphedema, the most widely accepted criteria are those put in place by the National Lymphedema Network (NLN) and the International Society of Lymphology (ISL). Both organizations recommend the combination of volumetric measurements, patient signs and symptoms, and physical examination to diagnose and then stage breast cancer related lymphedema.^12–15 These criteria are firmly rooted in earlier research and care standards developed by Földi and Földi.¹¹ Onset of BCRL typically occurs within the first 3 years following treatment,^3,7 lasts a lifetime, and requires lifelong management.^12–15

With the exception of latent (stage 0 or 1a) lymphedema, the condition involves volume change. However, the stage or severity of lymphedema is determined in part by volume (minimal [<20% increase in volume compared to baseline]; moderate [20%–40% increase], and severe [>40% increase])¹⁰ and also by whether tissue fibrosis, hemosiderin staining, and other skin changes have occurred.^10,13,15 Stage 1 lymphedema is defined as an increase in volume regardless of amount that is easily reduced with elevation; no tissue changes, such as fibrosis, are present. Stage 2 involves increased volume that is not reversible with elevation, and pitting is present. Progression to stage 3 is characterized by lymphostatic elephantiasis, advanced tissue fibrosis, and adipose deposition. Also occurring at this stage are trophic changes to the skin, such as thickening, darkening (ie, hemosiderin staining), and cystic and warty growths. This stage often results in deformity.^10,13,15 In this study, we focused on the criteria recommended by the ISL and NLN that are conducive to survey research: presence of edema, pitting, feeling of heaviness, tissue texture changes (hardening of the tissue that arises from fibrosis), and skin color changes due to hemosiderin staining.^10,13,15,16

Risk factors for BCRL include obesity and multimodality cancer treatment. Obesity, measured by a body mass index (BMI) of >25 kg/m², is considered an independent risk factor for the development BCRL.^17–20 This risk increases once BMI exceeds 30 kg/m².^18–20 To illustrate, Meeske et al showed that women with a BMI of >25 kg/m² had a 2-fold increase for arm lymphedema, whereas women with a BMI of >30 kg/m² had a 3-fold increase.²⁰ Some research suggests that obesity is higher among people who are African American and have survived breast cancer, placing them at higher risk of developing BCRL than people who are white.^18,20 Meeske et al also showed that 20% to 28% of African Americans but 12% to 21% of whites were affected by BCRL²⁰; another study found that 77% of African Americans but 39% of whites who survived breast cancer developed BCRL, with no explanation provided for why this difference may have occurred.²¹ Multimodal breast cancer treatment is also associated with varying levels of risk for BCRL.²² For example, breast conserving surgery and, more importantly, sentinel lymph node biopsy have been hailed as substantially reducing both the incidence and severity of BCRL.⁶ Additional risk for BCRL has been attributed to radiation therapy^3,9 and chemotherapy.^9,18 A more recent study with a large sample size found that hand edema was the biggest predictor of BCRL.²³ Unlike our study, all of these studies included samples that were largely white or had a high socioeconomic status or the socioeconomic status was not reported.

Socioeconomic status is a well-known social determinant of cancer outcomes with larger effects than race or ethnicity.^21,24 Health disparities result from a complex set of factors such as race, ethnicity, income, education, and other social determinants that interact to deepen inequalities.²⁴ Persistent cancer-related side effects, such as BCRL, serve as constant and distressing reminders of the cancer experience long after treatment has ended.²⁵ Despite breast cancer being one of the most common forms of cancer,²⁶ people who have survived breast cancer are often unprepared for physical impairments such as BCRL.^27–29 With the overall cancer survivor population expected to exceed 18 million by 2020,³⁰ we lack systematic examinations of how BCRL affects specific subpopulations, including minorities and those with low incomes. The purpose of this study was to examine whether disparities exist in terms of prevalence of self-reported lymphedema signs, symptoms, and whether breast cancer survivors report that they have been diagnosed with BCRL. We examined disparities across race (African Americans and white) and household income levels.

Methods

Study Participants

All participants were women who survived breast cancer and were part of the Southern Community Cohort Study (SCCS). The SCCS is a longitudinal, prospective, epidemiologic cohort study that enrolled 85,689 adults who were 40 to 79 years old and were from 12 southern states (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Virginia, and West Virginia) during 2002 to 2009.³¹ Eligibility criteria included English-speaking and no treatment for any cancer within the year preceding SCCS enrollment. Ninety-four percent of SCCS enrollees reported that their race/ethnicity is either African American or white.³¹ It is from this subgroup that we drew our sample of breast cancer survivors. Most (86%) SCCS participants were recruited from community health centers that serve residents with a low income and limited health insurance. The remainder of the SCCS cohort was recruited from the general population via various publicly available registries, including voting registries, motor vehicles registries and commercial registries. A description of the SCCS recruitment and survey methods is located online at www.southerncommunitystudy.org³² and in Signorello et al.³¹

We identified 1109 potentially eligible women in the SCCS who reported having been diagnosed with breast cancer prior to SCCS enrollment. After institutional review board approval (Vanderbilt University and Northeastern University), we made attempts to contact all women who had survived breast cancer and were thought to be alive for recruitment into The Southern Community Cancer Survivorship Study—a spin-off of the SCCS. An introductory recruitment packet was sent that included a letter of introduction, 2 copies of the consent document, a stamped/addressed return mailer, and sets of answer choices (similar to hand cards used in face-face interviews) with illustrations that helped to direct participants to answer choices during the telephone interview. The illustrations corresponded to sets of answer choices for different sets of questions and were used to cue participants on the answer choices to ensure completeness of data. Verbal consent was obtained per institutional review board guidelines. After the recruitment packets were mailed, we then made follow-up telephone calls (up to 15 attempts) to consent to participating in our study and to either conduct a computer-assisted telephone interview at the time of the recruitment or schedule an interview. Professional interviewers were trained on informed consenting, interview procedures, and computer-assisted telephone interviews for this study. The professional interviewers collected all study data and were overseen by the principal investigator (A.M.F.).

As noted in a previous publication,³³ we completed 577 interviews. Of nonparticipants, 337 could not be reached by telephone, 13 never had breast cancer, 68 refused, 65 could not be reached, and 49 died per National Death Index verification. Of the responding 577 women who had survived breast cancer, those missing race (n = 42) or those not identified as African American or white (n = 7) were excluded, yielding a sample of 528 women who had survived breast cancer for all analyses comparing racial groups. The sample size was reduced to 514 for income comparisons because 14 additional participants had missing data, refused to answer, or reported that they did not know their household income. Nearly all (~92%) women who had survived breast cancer had some form of health insurance.

Survey Instrument

All written materials for this study, including the survey, had a readability level no higher than the eighth-grade level. All questions were determined to be culturally sensitive and relevant per methods described by Bailey et al³⁴ and Erwin et al.^35,36 The survey questionnaire included items from the published literature about lymphedema signs and symptoms and are part of the ISL criteria for lymphedema diagnosis. We also asked each participant whether she had been diagnosed with BCRL by her doctor. The survey queried participants about sociodemographic characteristics (age, race, ethnicity, socioeconomic status, education level, household income, health insurance status), years since diagnosis, location of SCCS enrollment (community health center or general population), medical history, breast cancer diagnosis and extent of breast cancer treatment (surgery, chemotherapy, radiation, and type of lymph node biopsy [sentinel lymph node biopsy versus axillary lymph node dissection]), and comorbidity. We relied on the literature and review of the survey by experts in lymphedema, breast cancer, health disparities, and health literacy to be sure we were asking questions that would be asked in the clinic to help a determine a diagnosis of lymphedema. This allowed us to achieve content and face validity of our survey. We relied on self-reports as we could not realistically conduct a chart review on each of our participants because this would require dozens of institutional review board approvals, trips to hospitals in all states included in the SCCS, and costs that exceeded the limited budget for this study.

Lymphedema Signs, Symptoms, and Diagnosis

Based on diagnostic criteria developed by Földi and Földi¹¹ as well as the ISL and NLN,^13,15,16 participants were first asked, “Have you ever had any swelling like this in your arms, breast, chest, armpit, or hand after your breast cancer?” For those who answered “Yes” to this question, we also asked whether their health care provider had diagnosed the participant with lymphedema related to their breast cancer and this was measured as yes (=1 diagnosed with lymphedema) or no (=0 not diagnosed with lymphedema). Regardless, we also inquired about whether they had any skin-pitting edema related to their breast cancer and/or cancer treatment. During the interview, participants were instructed to press their finger into the affected area and report whether the skin bounces back up (normal), stays down for a while like a little pit, or was too firm to yield. Abnormal results from the skin pitting test are indicated if the skin leaves a pit that slowly refills or does not yield because the skin is too taut from the underlying edema. Participants were then asked whether they had any of the following signs and symptoms as a result of their breast cancer and/or treatment with yes (1)/no (0) response choices: breast edema, axillary edema, arm edema, or hand edema; feelings of heaviness of the involved area(s); hardening (ie, fibrosis) of the skin in the involved area(s); darkening (hemosiderin staining) of the skin in the involved area(s). Participants were allowed to determine the presence of their own skin color change in the edematous area. We explained that this color change would look darker, brownish or reddish-brownish, and be localized in the edematous area. We reminded participants that we were interested in knowing whether the skin color change was a result of edema in the involved area. Participants were allowed multiple answers to whether edema was present in the breast, axilla, arm or hand. We summed the number of lymphedema signs and symptoms present with the highest possible score being 8 (1 point each for the presence of the following in the involved area: pitting; breast edema; axillary edema; arm edema; hand edema; heaviness; skin hardening; skin darkening). Each sign and symptom was treated equally in our summation because no weighting system is included in the literature nor recommended by the ISL. We evaluated whether disparities exist for each sign and symptom by race and income level. The summed score of lymphedema signs and symptoms allowed us to evaluate whether disparities exist between extent of BCRL signs and symptoms and a diagnosis of lymphedema by race and income level.

Sociodemographic Characteristics

Age is measured as the number of years reported by the participant. We categorize race into 2 categories: 1 = African American; 2 = white. Socioeconomic status was measured by level of education (<9 years, 9 to 11 years, high school/GED, vocational/technical, some college/junior college, college graduate, or higher) and annual household income (<$10 K, $10 K–$20 K, $21 K–$40 K, and >$40 K.) Our income categories were determined by approximating the 2011 US Census Bureau official federal poverty level ($22,811 for a family of 4).^37,38 This poverty threshold was selected because the average number of children in a household in the southern states from which we drew our sample ranges between 2 and 3.³⁷ Thus, our categories of <$10 K and $10 K–$20 K represented “low income.”^37,38 Insurance status was classified by type of health insurance carrier (Medicare, Medicaid, private carrier, CHAMPUS, other, no insurance). We allowed participants to select more than 1 type of insurance. The first insurance selected was considered the primary insurance. The number of years since breast cancer diagnosis was calculated as the year reported for the breast cancer diagnosis subtracted from the date of the interview. Location of SCCS enrollment was controlled in this study and measured as a binary variable (community health center or general public).

Cancer Treatment History

Participants were asked about their breast cancer experience with a series of yes (1)/no (0) questions regarding the number of cancer treatment modalities (mastectomy, lumpectomy, sentinel lymph node biopsy, axillary lymph node dissection, chemotherapy, radiation, and long-term oral medication). We summed the number of treatment modalities to create a continuous variable to represent the extent of cancer treatment with a minimum of 0 and a maximum of 7.³⁹ We included the option of having more than 1 type of surgical procedure because reexcision and repeat surgery are not uncommon; reexcision ranged from 13% to 41%, whereas repeat surgery ranged from 28% to 41%.⁴⁰

BMI was calculated as self-reported current weight (kilograms) divided by height (meters) squared from the most recent SCCS survey (either baseline or follow-up survey). Participants were asked whether they had any of the following medical conditions using a yes (1) or no (0) response choice: high blood pressure, heart attack (myocardial infarct), diabetes mellitus, stroke (cerebrovascular accident), emphysema, depression, osteoarthritis, congestive heart failure, HIV/AIDS, memory problems, and paralysis. We used the same comorbidity categories as the SCCS in the baseline and follow-up surveys. The sum of these conditions is an acceptable method representing the number of comorbid conditions among women who had survived breast cancer.^41,42

Data Analysis

We stratified our sample by race (African American or white) and annual household income categories (≤$10 K, $10,001–$20 K, $20,001–$40 K, and >$40,001) to compare lymphedema signs and symptoms with BCRL diagnosis. We present frequencies and percentages for categorical variables and means and standard deviations for continuous variables to characterize our participants in terms of sociodemographics, breast cancer medical history, comorbidity, and prevalence of lymphedema signs, symptoms, and BCRL diagnosis. Differences by race were assessed with chi-square tests or Fisher’s exact tests for categorical variables and independent sample t tests for continuous variables. Income differences were assessed with chi-square tests or Fisher exact tests for categorical variables. Statistical significance was determined with P values of < .05, and all analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC, USA).

Role of the Funding Source

The funder—U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute (R21 CA137483 [PI: A.M. Flores; Co-I: W.J. Blot] and R01 CA092447 [PI: W.J. Blot; Co-I: A.M. Flores])—played no role in the design, conduct, or reporting of this study.

Results

Demographic characteristics are presented in Table 1. Overall, the mean age of our sample was 64 years, with African Americans being significantly younger than whites (African American = 62.2 years; white = 64.9 years; t test P = .0002) and with no differences by education level, years since their breast cancer diagnosis (survival for ~ 13 years), and source of SCCS enrollment. Slightly more than half (57%) of our sample was enrolled into our study from SCCS participants originally recruited from a community health center. BMI was high overall (mean BMI = 31); however, African Americans had a significantly higher average BMI than whites (African American = 32.9; white = 29.5; t test P < .0001). African Americans reported significantly lower household incomes (58% of African Americans earned $20 K or less) than whites (36% earned $20 K or less; chi-square P < .0001); however, income did not vary much across age. Those reporting household incomes of >$40 K were, on average, 61 years old (range = 46–82 years), whereas those with lower incomes had an average age varying between 63 and 64 years (range = 45–85 years). African Americans were also more likely to be on Medicaid than whites (chi-square P = .0006), whereas whites were more likely to have private health insurance than African Americans (chi-square P = .0041). In terms of medical/surgical management of breast cancer, whites were more likely to be taking a long-term oral medication (antineoplastic) than African Americans (chi-square P = .044) but otherwise there were no significant differences in breast cancer management. African Americans were also more likely than whites to have high blood pressure and diabetes mellitus (chi-square P < .0001, respectively), whereas whites were more likely to have experienced menopause than African Americans (99% of whites and 94% of African Americans; chi-square P = .0056).

Table 1.

Demographic Characteristics by Race^a

Characteristic	White (n = 266) b	African American (n = 262) b	P c
Age, y, mean (SD)	64.9 (8.4)	62.2 (8.1)	.0002
Formal education			.4872
<High school	41 (15.4)	46 (17.6)
High school	78 (29.3)	70 (26.7)
Some college or vocational training	73 (27.4)	78 (29.8)
College graduate	39 (14.7)	36 (13.7)
Graduate school	34 (12.8)	27 (10.3)
Missing/refused to provide	1 (0.4)	5 (1.9)
No. of years since BC diagnosis, mean (SD)	12.7 (7.4)	12.5 (7.2)	.7397
BMI, mean (SD)	29.5 (6.5)	32.9 (7.3)	<.0001
Household income, $d			<.0001
≤10 K	29 (11.2)	85 (33.3)
10,001–20 K	64 (24.7)	63 (24.7)
20,001–40 K	79 (30.5)	57 (22.4)
>40,001	87 (33.6)	50 (19.6)
Location of enrollment			.128
Community health center	144 (54.1)	159 (60.7)
Public recruitment	122 (45.9)	103 (39.3)
Insurance statuse
Medicaid	30 (11.3)	59 (22.5)	.0006
Medicare	152 (57.1)	138 (52.7)	.3019
Private	160 (60.2)	125 (47.7)	.0041
CHAMPUS	13 (4.9)	11 (4.2)	.7040
Other insurance	27 (10.2)	23 (8.8)	.5904
No insurance	20 (7.5)	19 (7.3)	.8970
Missing/refused to provide	0	0
BC treatment
BC surgery	262 (98.5)	259 (98.9)	.7186
Mastectomy	160 (60.2)	158 (61.0)	.9879
Lumpectomy	128 (48.1)	112 (43.2)	.1988
Sentinel node biopsy or axillary node dissection	163 (61.3)	138 (52.7)	.1057
Chemotherapy	122 (45.9)	129 (49.2)	.4132
Radiation	126 (47.4)	127 (48.5)	.8308
Long-term oral medication	139 (52.3)	113 (43.1)	.0435
No. of treatment modalities, mean (SD)	3.4 (1.5)	3.2 (1.3)	.0914
Comorbidityf
High blood pressure	105 (39.5)	179 (68.3)	<.0001
Myocardial infarct (yes only)	10 (3.8)	13 (5.0)	.5044
Diabetes mellitus	43 (16.2)	89 (34.0)	<.0001
CVA	11 (4.1)	12 (4.6)	.7950
Hepatitis	3 (1.1)	9 (3.4)	.0764
Emphysema	23 (8.6)	21 (8.0)	.8135
Depression	67 (25.2)	56 (21.4)	.2999
Osteoarthritis	149 (56.0)	159 (60.7)	.2763
Menopause	262 (98.5)	246 (93.9)	.0056

^aBC = breast cancer; BMI = body mass index; CVA = cerebrovascular accident; K = thousand.

^bValues are reported as numbers (percentages) of women unless otherwise indicated.

^c P values for differences between racial groups were from independent sample t tests for continuous variables and χ² tests for categorical characteristics.

^dFive hundred fourteen cases had complete data for household income by race as follows: White (n = 259) and African American (n = 255).

^eSurvey participants were asked to select all insurance types that applied. Several combinations of insurance types were discovered during the survey: Medicaid and Medicare (n = 39); Medicaid, Medicare, and other (n = 13); Medicaid and other (n = 8); and Medicare and other (n = 152).

^fThe following comorbidities were excluded because they had no statistically significant effect on our sample: congestive heart failure, HIV/AIDS, memory problems, and paralysis.

Prevalence of Lymphedema Signs and Symptoms by Race and Income Level

Of the 528 women who had survived breast cancer and were interviewed, 29.5% (n = 156) reported lymphedema signs or symptoms. With the exception of skin pitting, African Americans were more likely to report having each of the remaining 7 signs/symptoms than whites: swelling in breast (P = .0016), axilla (P = .0037), arm (P = .0008), and hand (P = .0346); skin hardening (P = .0012); skin darkening (P < .0001); and feelings of heaviness in the affected area (P = .0005) (Tab. 2). The percentages shown in Table 2 were calculated using participants who responded “yes” or “no” to the survey question. A small number for each response may have been excluded because of missing data or participants responding “do not know.” Similarly, those with incomes of $10 K or less were the most likely to report having each of the lymphedema symptoms and signs with the exception of skin pitting and edema in the arm and hand which were not significantly different across income groups (Tab. 3).

Table 2.

Prevalence of Lymphedema Symptoms and Diagnosis by Race

Lymphedema Signs and Symptoms	White (n = 266) a	African American (n = 262) a	χ 2 P b
Self-reported swelling
Breast	17 (6.4)	39 (14.9)	.0016
Axilla	31 (11.7)	55 (21.0)	.0037
Arm	61 (22.9)	95 (36.2)	.0008
Hand	39 (14.7)	57 (21.8)	.0346
Skin hardening	22 (8.3)	47 (17.9)	.0012
Skin pitting			.3131
Stays down like a pit	20 (7.5)	29 (11.1)
Comes right back up	233 (87.6)	224 (85.5)
Skin does not yield	3 (1.1)	5 (1.9)
Skin darkening	13 (4.9)	52 (19.8)	<.0001
Heaviness from edema	46 (17.3)	81 (30.9)	.0005
Lymphedema diagnosisc	30 (57.7)	42 (40.4)	.0409

^aValues are reported as numbers (percentages) of women unless otherwise indicated.

^b P values for differences between racial groups were from χ² tests for categorical characteristics.

^cAmong participants reporting lymphedema signs or symptoms (white = 52; African American = 104).

Table 3.

Prevalence and Association of Lymphedema Symptoms and Diagnosis by Household Income

	Household Income, $ a
Lymphedema Signs and Symptoms	≤10 K (n = 114)	10,001–20 K (n = 127)	20,001–40 K (n = 136)	≥40,001 (n = 137)	P b
Self-reported edema
Breast	24 (21.1)	10 (7.9)	9 (6.6)	11 (8.0)	.0074
Axilla	25 (21.9)	21 (16.5)	21 (15.4)	15 (10.9)	.0218
Arm	38 (33.3)	36 (28.3)	43 (31.6)	36 (26.3)	.2987
Hand	22 (19.3)	27 (21.3)	25 (18.4)	19 (13.9)	.1542
Skin hardening	29 (25.4)	16 (12.6)	12 (8.8)	9 (6.6)	<.0001
Skin pitting					.0599
Stays down like a pit	18 (15.8)	8 (6.3)	12 (8.8)	9 (6.6)
Comes right back up	90 (78.9)	114 (89.8)	116 (85.2)	125 (91.2)
Skin does not yield	4 (3.5)	1 (0.8)	1 (0.7)	2 (1.5)
Skin darkening	22 (19.3)	22 (17.3)	13 (9.6)	7 (5.1)	.0001
Heaviness from edema	40 (35.1)	27 (21.3)	34 (25.0)	24 (17.5)	.0062
Lymphedema diagnosisc	16 (32.7)	22 (56.4)	24 (66.7)	24 (85.7)	.0194

^aValues are reported as numbers (percentages) of women unless otherwise indicated. K = thousand.

^b P values for differences between income groups were from Cochran-Armitage tests of trend or the Fisher exact test, when appropriate, for categorical characteristics.

^cAmong participants reporting lymphedema signs or symptoms (≤10 K, n = 49; 10,001–20 K, n = 39; 20,001–40 K, n = 36; and ≥ 40,001, n = 28).

Likelihood of BCRL Diagnosis

With the exception of pitting edema, women who were African American were more likely to report the remaining signs and symptoms of BCRL (Tab. 2). Our participants were asked about a diagnosis of BCRL only if they reported breast cancer–related swelling. Women who were African American with BCRL signs and symptoms were less likely to report having been diagnosed with BCRL than women who were white (P = .0409) (Tab. 2). Being diagnosed with BCRL was significantly associated with having a higher income among participants reporting lymphedema signs or symptoms (P = .0194) (Tab. 3). More than 50% of those who reported BCRL signs and symptoms reported that they were not diagnosed with BCRL. For the analysis in Table 4, 326 women who had survived breast cancer (White = 189; African American = 137) did not report any swelling and were not asked about a lymphedema diagnosis. Among the remaining 202 participants, 7 (White = 5; African American = 2) had missing information on lymphedema signs and symptoms and were excluded from the analysis of agreement. Among the remaining 195 participants (White = 72; African American = 123), 39 (White = 20; African American = 19) reported no lymphedema signs and symptoms. No significant relationship was observed between whether any signs and symptoms of BCRL existed and self-report of a physician diagnosis of BCRL (Tab. 4). In separate analyses, we also examined relationships between different levels of reported symptoms and diagnosis but these too were not significant.

Table 4.

Agreement Between Lymphedema Symptoms and Diagnosis of Lymphedema by Race

		Lymphedema Signs/Symptoms
Race	Lymphedema Diagnosis	Yes	No	Total No. of Participants	P	Kappa a
White	Yes	30	9	39	.33	0.11
	No	22	11	33
	Total	52	20	72
African American	Yes	42	5	47	.25	0.06
	No	62	14	76
	Total	104	19	123

^aA kappa value of 0 indicated agreement equivalent to chance; a kappa value of 1 indicated perfect agreement.

Discussion

This cross-sectional study was designed to examine whether disparities in self-reported signs, symptoms, and diagnosis of BCRL existed among women who were African American, white, or had a low income and who had survived breast cancer. This study shows that African Americans and those with low income levels in our sample were more likely to have nearly all patient-reported signs and symptoms of lymphedema—upper quarter edema, tissue hardening (ie, fibrosis), and skin darkening (ie, hemosiderin staining)—but less likely to be diagnosed with BCRL when compared to those who are white and/or with higher incomes. A recent case series studied a small sample of breast cancer survivors (n = 9) at risk of developing BCRL suggested that hand edema may be a risk factor for the development of BCRL. Our results suggest that more research is needed on the differential role of edema in the axilla, breast, arm and/or hand as risk factors for development of BCRL.²³ One study shows a prevalence of 0% to 65%, with those receiving radiation being at a slightly higher risk of developing BCRL,³ whereas other research shows a prevalence anywhere between 13% and 65%,⁴ depending on the type of surgical resection, axillary versus sentinel node biopsy, and measurement method.^5–7 Still other research shows prevalence rates of 8% to 31%, whereas others show a wider range (3%–87%).^8,9 Recall that none of these studies have large sample sizes or include minorities or people of low socioeconomic status, which makes generalization difficult.

Minorities and those who are socioeconomically disadvantaged have higher levels of mortality due to breast cancer,^26,31 and our study points to another cancer disparity issue—the potential underdiagnosis of BCRL in women who are in minority and socioeconomically disadvantaged groups and have survived breast cancer. Our study supports recent research by Brunelle et al²³ and goes even farther by shedding light on disparities in BCRL. A low percentage of our participants report ever having been diagnosed with BCRL by their health care provider, despite multiple BCRL signs and symptoms, suggesting that early detection of BCRL might not have occurred. Patient self-reports of signs and symptoms of BCRL are purportedly taken seriously, and long-term effects and complications (eg, cellulitis) of persistent lymphedema are well known.^{11,19,43–46} The racial and income patterns in BCRL point to potential bias that may deepen health care disparities in cancer survivorship. Dean at al point to lower education and income as proxies for access to health care resources. Although our sample did not differ in terms of education, our participants did differ in terms of income and type of insurance with African Americans more likely to report being enrolled in Medicaid.⁴⁷ A lack of access to health care resources with expertise in lymphedema detection, referral and management may partially explain our results.⁴⁷ Regardless, our study expands the idea that the burden of cancer may also include undiagnosed BCRL and lack of access to highly specialized care despite having insurance.

Our findings support literature that shows consistent trends of the poor bearing the highest burden of cancer mortality, advanced stage cancers, and health care related financial debt regardless of race or ethnic background.^21,31 Living in an economically impoverished neighborhood appears to be associated with lower physical functioning,⁴⁵ diminished physical activity,^48,49 and high body mass index.^20,50 Our study suggests that those with low incomes might need more targeted and programmatic approaches regarding monitoring and risk reduction for BCRL.

Despite the fact that previous studies have shown a relationship between BMI and lymphedema^9,29 we did not find that BMI was related to whether a participant reported having been diagnosed with breast cancer related lymphedema (results not shown). Some researchers suggest that women who are African American and have survived breast cancer experience higher rates of lymphedema because of a higher average body mass index (BMI) than their counterparts who are white.¹⁸ Interestingly, some of the highest average body mass index scores are found in the southeastern United States,⁴⁸ and our sample is no different. All of our participants resided in the southern United States and all were, on average, overweight or obese with little variation. Nevertheless, our participants who were African American had a significantly higher BMI than their counterparts who were white.

Because our study was administered via computer-assisted telephone interviews from participants in many different states, we were unable to obtain medical chart data. This means we did not have data on number of lymph nodes removed, pathological involvement of lymph nodes, and tumor characteristics such as tumor subtype, stage, grade, hormone sensitivity, etc. Recall that many of these participants are long-term survivors of their breast cancer. This long-term survival might present recall bias with respect to their experiences with BCRL.

A well-known risk factor for lymphedema development is axillary lymph node dissection versus sentinel lymph node biopsy, which we measured in this study. Regardless of how we accounted for these variables—as separate or aggregated variables—they were not statistically associated with signs, symptoms, and diagnosis of lymphedema (results not shown). A limited number of studies describe the importance of the number of lymph nodes removed (not necessarily the type of surgical lymph node resection) and pathological involvement of the removed nodes in the development of BCRL.^49,51,52 Although the number of removed lymph nodes vary widely in the development of BCRL, to date, there is no set threshold on the number of nodes that will predict the development of BCRL.¹⁶

A recent, large, prospective cohort study showed that multi-modal cancer treatment increased the risk of BCRL development.⁵³ Most importantly, the highest risk of developing BCRL was found among those with axillary lymph node dissection plus radiation therapy and/or anthracycline/cyclophosphamide plus taxane-based chemotherapy.⁵³ We included variables that represented the type and number of medical/surgical modalities used for our survivors, but we could not include data representing taxane-based chemotherapy agents because we were unable to conduct medical chart reviews.

For those who reported that they had been diagnosed with BCRL, clinical data on edema volume and date of BCRL diagnosis were also not available. It is possible that the duration of BCRL signs and symptoms could be responsible for our findings. Although these variables may not adequately account for the duration of signs and symptoms, they may lend clues about the length of exposure to the development of BCRL across the southern United States. Unfortunately, given our study reach across many southern states, it was simply not realistic to conduct medical chart reviews. Even though we specifically asked each participant if her doctor had ever diagnosed her with breast cancer-related lymphedema, we might have had some individuals that could not recall whether such a diagnosis had been made. Perhaps inquiring about whether the participant had ever received compression bandaging or a compression garment (both of which require a physician’s referral and are parts of complete decongestive therapy—the gold standard treatment approach for managing lymphedema) might have lent more validity to our diagnosis data. Since we know that BCRL is most likely to develop within 3 years after breast cancer treatment, our results might have differed between those diagnosed with BCRL for ≤3 years versus >3 years prior to enrollment. However, we did not have information on participants’ date/year of their BCRL diagnosis.

As noted earlier, different schemas to arrive at a diagnosis and stage of BCRL exist; however, the ISL and NLN statements are considered to provide definitive diagnostic and staging criteria that are based on the original criteria (of Földi and Földⁱ¹¹) of patient-reported signs and symptoms, skin inspection, and palpation of the affected area to test for pitting. We relied on the criteria that were amenable to survey research. It is important to appreciate the difference between diagnosing BCRL and staging BCRL. A diagnosis for treatment referral relies on the presence of obvious edema in the affected area and physical examination of the affected area by the clinician. Staging of lymphedema, on the other hand, is determined by the combination of signs, symptoms, observation, and physical examination. Since this is a survey study with participants all over the southern United States, we were unable to conduct a complete physical examination of the participant to definitively diagnose and stage BCRL. Instead, we incorporated instructions on how to conduct a skin pitting test to assess tissue response. We did not have the advantage of being able to observe the participant so we also specifically queried participants about whether any skin changes (tissue texture and discoloration) were present. Despite these study limitations, it is important to understand the clinical reality that referral for treatment of BCRL requires a diagnosis of BCRL by a referring health care provider. It is often the case that the referring provider diagnoses the presence of BCRL but the lymphedema therapist to whom the patient is referred stages the lymphedema. The stage of lymphedema is most helpful to determine prognosis and select appropriate techniques and supplies (eg, lymphatic surgery, manual lymph drainage, compression, therapeutic exercise, skin care, compression bandaging, foam inserts and chips, night time compression, compression garments) to manage BCRL.

Provider bias against people who are in minority and socioeconomically disadvantaged groups and have survived breast cancer might explain the racial and income patterns of BCRL signs, symptoms, and diagnosis. Racial prejudice and stigmatization of poverty are well-known causal factors of breast cancer disparities.⁵⁴ It is totally plausible that similar mechanisms might be at play with regard to BCRL. It may also be that providers are not prioritizing BCRL detection, referral and management in their clinical practice or providing patient education for early detection and self-management. Moreover, patient-provider communications imbued with provider bias might result in women who have survived breast cancer downplaying, disregarding, or misrepresenting BCRL signs and symptoms and feeling resigned to “live with it.” Future research should incorporate interviews with health care providers and medical chart reviews to explore BCRL detection, patient education, and referral patterns and clinical pathways.

Since we relied on survey methods, we did not include newer diagnostic medical procedures to test for the presence and extent of lymphatic structures involved such as lymphangioscintigraphy, indocyanine green tracing, and other radiological imaging techniques. Such medical procedures for this study would have been unrealistic to perform and cost-prohibitive given the geographic range of our participants. Moreover, these procedures are not routinely done on patients diagnosed with BCRL with the exception of cases for which lymphatic surgery will be performed.¹³

Body mass index is cited as a risk factor for the development of BCRL.^55–62 By using BMI, our results might be comparable across other studies. But none of the studies cited evaluated the relationships between BCRL and BMI in a racially and socioeconomically diverse sample such as ours. Although BMI was significantly higher among African Americans, our sample had a high average BMI overall and may have had less variation in BMI than those reported for other samples. This may explain why we did not find a relationship between BMI and a BCRL diagnosis. Moreover, all of our participants are from the southern United States—an area with higher BMI rates across all groups than the national average.⁴⁸ Alternatively, the lack of longitudinal and contextual information (ie, weight gain/loss patterns since breast cancer treatment) might have also explained why we did not find a significant association. Previous studies assume that high BMI precedes BCRL, thus making BMI a risk factor for the development of BCRL. Obesity is associated with lymph vessel leakage which may place obese individuals at higher risk of lymphedema development.⁶³ However, newer lymphedema research in mouse models and other research in fibroproliferative diseases (eg, cirrhosis, pulmonary fibrosis, scleroderma) suggest that lymphedema and its associated tissue fibrosis might actually trigger subcutaneous fat generation and deposition.^63–65 Taken together, it might be that obesity is likely both a risk factor and an outcome of BCRL. More longitudinal studies utilizing survey, clinical, and biological data in human models are needed to understand the relationships between obesity and BCRL.

Like obesity, vessel leakage is also part of the pathophysiology of diabetes and uncontrolled hypertension.¹¹ Although we collected data on the presence of diabetes and hypertension, we did not collect data on the length of time since diagnosis of these conditions nor did we have information of whether these conditions were controlled by medication and/or lifestyle changes. It is possible that the combination of obesity, diabetes, and hypertension causes greater vessel leakage and microangiopathy, thereby increasing the risk of developing BCRL.¹¹ However, an analysis of the pathophysiology of BCRL and comorbidity is beyond the scope of this article. The development of indocyanine green and near-infrared imaging to evaluate lymphatic backflow from vessel damage and leakage^12,13 may be helpful to identify those at highest risk for developing BCRL. In addition, newer research suggests that a genetic predisposition to the development of BCRL might exist for those with African American ancestry who are not obese.⁶⁶ Specifically, 2 single nucleotide polymorphisms in the gap junction protein alpha 4 gene have recently been identified as potential biomarkers for BCRL risk.⁶⁷ Future studies that include survey measures (like ours) combined with clinical outcome assessments (ie, indocyanine green to evaluate lymphatic vessel leakage and backflow) alongside genotyping might contribute to precise risk-stratification models for early identification of BCRL.

With cancer survivors having a high cost of health care, the survivorship phase—or “continuing care phase”—is the costliest, particularly for women who have survived breast cancer.⁶⁸ Schmitz et al estimated that the treatment for BCRL has total cost of $5,636 per woman over 18 months following breast cancer surgery.⁶⁹ In addition, these costs were nearly 4 times higher for those with 4 or more adverse treatment effects during the same 18 months after surgery.⁶⁹ Moreover, Basta et al shows that those with complicated lymphedema (ie, lymphedema requiring hospitalization), have a substantial increase in the cost of care during the first 2 years after a BCRL diagnosis compared with those without BCRL—$58,088 and $31,819, respectively.⁷⁰ Again, none of these studies included a sample as diverse as ours, especially with regard to race and income. Cancer survivors already incur a great amount of treatment-related debt, and this economic demand is particularly burdensome for those with low incomes, such as those in our study.^{19,21,68–70} BCRL, particularly if not diagnosed and treated, will cause pain and persistent, serious, and life-threatening problems such as skin breakdown, infections (eg, cellulitis) that may require hospitalization, severe edema, tissue fibrosis, skin discoloration, chylous cysts, and disability resulting in greater health care spending^{4,6,11,21,41,69,70} adding to the financial toxicity of cancer. Our results suggest the possibility that there may be an underreporting of BCRL signs, symptoms, and perhaps even an underdiagnosis of BCRL for long-term breast cancer survivors. This has downstream implications for the delivery and cost of care to treat BCRL. Future research might explore relationships between functional and economic impacts of BCRL with comparisons by race and income between those with and without BCRL as well as those who have not been treated for BCRL.

Our study suggests that women who are African American or have a low income and have survived breast cancer disproportionately report BCRL signs and symptoms and diagnosis. However, BCRL diagnosis did not seem related to the presence of signs and symptoms of BCRL. Whether a similar relationship exists in terms of lymphedema treatment utilization is not known at this time and is the next step of our research. Our study suggests that self-reported signs and symptoms of BCRL may be more prevalent than previously thought among African Americans and those with low income levels. Broadly speaking, it seems that African Americans, and especially those with low incomes, might bear a much greater burden of breast cancer and its side effects than previously known.

Author Contributions and Acknowledgments

Concept/idea/research design: A. M. Flores, A. P. Sander, W.J. Blot

Writing: A. M. Flores, J. Nelson, L. Sowles, R. G. Stephenson, K. Robinson, A. Cheville, A. P. Sander, W. J. Blot

Data collection: A. M. Flores

Data analysis: A. M. Flores, J. Nelson, L. Sowles, W. J. Blot

Project management: A. M. Flores, L. Sowles

Fund procurement: A. M. Flores, W.J. Blot

Providing participants: A. M. Flores, W.J. Blot

Providing facilities/equipment: A. M. Flores

Providing institutional liaisons: A. M. Flores, W.J. Blot

Clerical/secretarial support: A. M. Flores, L. Sowles

Consultation (including review of manuscript before submitting): J. Nelson, L. Sowles, R.G. Stephenson, K. Robinson, A. Cheville, A. P. Sander, W.J. Blot

This work was supported by the National Institutes of Health (R21 CA137483 to A.M.F. and R01 CA092447 to W.J.B.).

Funding

The authors received funding from the U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute (R01 CA092447 [PI: W.J. Blot; Co-I: A.M. Flores] and R21 CA137483 [PI: A.M. Flores; Co-I: W.J. Blot]).

Disclosure and Presentations

The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no conflicts of interest.

A portion of this paper was presented at the Annual Combined Sections Meeting of the American Physical Therapy Association in February 2018, New Orleans, Loiusiana (USA).