Supportive Care in Cancer

, Volume 21, Issue 7, pp 2017–2024

Lymphedema: experience of a cohort of women with breast cancer followed for 4 years after diagnosis in Victoria, Australia


    • Women’s Health Program, School of Public Health and Preventive MedicineMonash University
  • Penelope J. Robinson
    • Women’s Health Program, School of Public Health and Preventive MedicineMonash University
  • Raychel Barallon
    • Women’s Health Program, School of Public Health and Preventive MedicineMonash University
  • Pamela Fradkin
    • Women’s Health Program, School of Public Health and Preventive MedicineMonash University
  • Max Schwarz
    • Department of Medicine, Central Clinical SchoolMonash University
    • Alfred Health
  • Susan R. Davis
    • Women’s Health Program, School of Public Health and Preventive MedicineMonash University
Original Article

DOI: 10.1007/s00520-013-1763-1

Cite this article as:
Bell, R.J., Robinson, P.J., Barallon, R. et al. Support Care Cancer (2013) 21: 2017. doi:10.1007/s00520-013-1763-1



The aim of this work was to study the incidence and prevalence of self-reported lymphedema in breast cancer survivors between 2 and 4 years following diagnosis, the factors associated with the development of lymphedema and the impact of lymphedema on psychological well-being.


We assessed self-reported lymphedema in the BUPA Health Foundation Health and Wellbeing After Breast Cancer Study, a questionnaire-based study of 1,683 women newly diagnosed with their first episode of invasive breast cancer in Victoria, Australia. Psychological well-being was assessed using the Psychological General Well-being Index.


Two years after diagnosis, nearly 20 % of women reported lymphedema and this proportion remained above 18 % 2 years later. However, self-reported lymphedema was a dynamic phenomenon, with the condition resolving in some women and others reporting onset for the first time up to 4 years from diagnosis. Lymphedema 2 years from diagnosis was positively associated with the number of nodes removed at initial surgery, although this variable only explained a small proportion of the likelihood of reporting lymphedema. The presence of lymphedema was associated with lower psychological general well-being.


Lymphedema after breast cancer treatment frequently has a dynamic pattern and may emerge as an issue for women several years after their initial treatment. It is associated with a lower level of general well-being.


LymphedemaBreast cancerPrevalenceIncidenceResolutionDeterminants


Secondary lymphedema of the arm due to disruption of the lymphatic channels by either surgery or radiotherapy commonly occurs after treatment for breast cancer [1]. If left untreated, lymphedema can persist for months to years and can predispose to infection which can be life-threatening [1]. It can also impair quality of life [2] as it can be disfiguring, be associated with pain and tenderness and interfere with motor skills such as writing [3].

Although modern treatment for breast cancer incorporates approaches that will minimise the risk of the development of lymphedema, including the use of sentinel node biopsy to minimise the need for axillary dissection [4], lymphedema following breast cancer treatment has not been eliminated. A recent Australian review concluded that one in five women will experience lymphedema following treatment for breast cancer with the prevalence of the condition increasing with longer follow-up; however, 70–80 % of women were reported to have developed the condition within 12 months of treatment [5]. This review also suggested that large prospective studies were needed to improve understanding of the patterns of incidence of the condition. Lymphedema has been assessed using a variety of objective methods. A gold standard volumetric method is lacking and different volumetric assessments have been shown to provide discordant results [6]. Furthermore, the repeated use of volumetric measures to assess a large study sample would be logistically challenging. Most importantly, volumetric measures do not necessarily identify the women actually bothered by the condition, which is what is the most important clinically.

The aim of this study was to determine the incidence and prevalence of lymphedema in the BUPA Health Foundation Health and Wellbeing After Breast Cancer Study (BUPA study), a questionnaire-based prospective cohort of 1,683 women with their first diagnosis of invasive breast cancer recruited across the state of Victoria, Australia between 2004 and 2006 [7]. Lymphedema was documented by self-reporting through a series of questions. This allowed both the inclusion of all women in the study at the time of completion of each questionnaire and enabled us to track both the incidence and resolution of the lymphedema over time. Critically, lymphedema was not defined or graded by specific measurements but was determined from the perspective of the experience of the individual.


Study population

Women in the BUPA study were recruited mainly (78 %) through the Victorian Cancer Registry (VCR) with the remaining women recruited directly to our research unit. A detailed account of recruitment to this study has been published [7]. For recruitment through the VCR, we were only denied access to 5 % of potential recruits on the recommendation of the managing clinician. Only 16 % of women approached through the VCR declined to participate. The women recruited to the study are representative of all women diagnosed with invasive breast cancer in the state of Victoria in terms of their age distribution, the size of their tumours at diagnosis and their location of residence [7].

Study procedures

The study was conducted entirely by mail. All participants completed an enrolment questionnaire (EQ) within 12 months of their diagnosis and then completed an annual follow-up questionnaire thereafter (FQ1, FQ2 and FQ3). Time between diagnosis and the completion of the enrolment questionnaire was on average 0.8 years (90 % of questionnaires were completed in the interval 0.2 to 1.1 years from diagnosis). Subsequent questionnaires were completed at 12-month intervals and so were completed on average 1.8 (90 % confidence interval (90 %CI) 1.2 to 2.2), 2.8 (90 %CI 2.1 to 3.2) and 3.8 (90 %CI 3.1 to 4.3) years from diagnosis.

The details of the pathology for each participant, which included the number of axillary lymph nodes removed at surgery, were provided to our study by the VCR. The EQ included questions about demographics and initial treatment [7]. Subsequent questionnaires included questions about lymphedema as well as questions about the development of further active disease (evidence of recurrence either local or distant as well as another primary breast cancer) and further treatment including surgery, radiotherapy and adjuvant therapy.

In FQ1, women were asked whether they had experienced lymphedema in the previous 12 months and were asked to rate how disabling the condition was (not at all, a little, moderately so to severely disabling). To be considered free of lymphedema, women needed to answer “no” to the question about whether or not they had experienced lymphedema in the past 12 months. To be classified as having lymphedema, a woman had to answer “yes” to the question and also give an indication of the degree to which she was bothered by it.

The questions about lymphedema were preceded by an explanatory statement: “Lymphedema occurs if the lymphatic system, which normally maintains fluid balance in the body, is impaired. This causes swelling of certain body parts (e.g. arm). Lymphedema may develop following certain medical treatments or procedures.” The questions about lymphedema were repeated in FQ2 and FQ3.

Each of the follow-up questionnaires included the Psychological General Well-being Index [8] (PGWB). We have reported on its use in our cohort previously [9] as well as in healthy women in the community [10]. This is a 22-item questionnaire with six domains: anxiety, depressed mood, positive well-being, self-control, general health and vitality. The six domains of the PGWB are created by summing the scores of groups of questions. The answers to some questions are reverse-scored so that for the individual domains and for the score as a whole, a higher score indicates a higher level of well-being. Each of the domain scores takes values up to 15, 20 or 25, depending on the number of questions in the domain and the highest total score possible is 110. The PGWB has been shown to demonstrate validity and reliability, and data from its use in the National Health and Nutrition Examination Survey have demonstrated that it can be generalised to adults aged 25 to 74 years [8].

The data were initially explored using tables, cross tabulations and bar graphs. Comparisons of medians were made using non-parametric statistics where data were not normally distributed.

Statistical analyses were performed using the statistical software SPSS 19.0 for Microsoft Windows (SPSS Inc., Chicago, IL, USA). The characteristics of women who did and did not report lymphedema at FQ1 were included in a forward logistic regression analysis to describe which variables contributed independently to the likelihood of reporting lymphedema at FQ1 and estimate what proportion of the likelihood of reporting lymphedema could be explained by the differences observed between groups. Forward multiple linear regression was used to explore factors both positively and inversely associated with the PGWB total score.


At the time of completion of the FQ1 when the questions about lymphedema were asked for the first time, 1,588 participants remained in the study. Women were lost to the study either because they did not survive long enough to complete a subsequent questionnaire, they elected not to continue in the study or they (rarely) were lost to follow-up. Eleven women who completed at least one study questionnaire were excluded from the analysis because they did not answer the question about lymphedema. The final numbers of women completing each questionnaire and included in the analysis who did and did not report lymphedema in each of FQ1, FQ2 and FQ3 are shown in Fig. 1.
Fig. 1

Numbers of women who did and did not report lymphedema in FQ1, FQ2 and FQ3 amongst the total number of women completing each of the questionnaires (with 11 women excluded from the analysis)

The prevalence of lymphedema at the time of the FQ1 was the highest reported in the study at 19.7 %. There was a marginal reduction in the prevalence of lymphedema from that point to 19.5 % in the subsequent questionnaire (FQ2) nearly 3 years from diagnosis and 18.2 % nearly 4 years from diagnosis (FQ3).

Despite the relative consistency in the prevalence of lymphedema at each time point, there was a substantial movement of women between the groups reporting presence or absence of the condition with about one in four women who reported the presence of lymphedema in FQ1 or FQ2 reporting no lymphedema in the subsequent questionnaire and about 1 in 20 women who reported “no lymphedema” in FQ1 or FQ2 moving into the “lymphedema” group in the next questionnaire. Forty-three women reported lymphedema for the first time nearly 4 years from diagnosis.

By the time of the FQ3, only 163 women had reported the presence of lymphedema in each of the three follow-up questionnaires and represent 11 % of the 1,435 women who remained in the analysis by FQ3. The total number of women who reported no lymphedema in each of the follow-up questionnaires was 1,044 and represents 73 % of the women remaining in the analysis by FQ3. Some women reported developing either recurrent disease or a new BC (active disease) in FQ2 and FQ3. However, this phenomenon did not explain the development of new lymphedema in the majority of cases as between both FQ1 to FQ2 and FQ2 to FQ3, only two women reported new active disease.

The majority of women reporting lymphedema at each point reported that they were only minimally bothered by it. At FQ1, 229/311 (74 %) reported being only minimally bothered with the remaining 26 % (82/311) reporting being moderately or severely bothered. The proportions at FQ2 and FQ3 reporting that they were minimally bothered were 73 and 72 %, respectively.

Women who reported lymphedema at FQ1 were more likely to be beyond stage 1 at diagnosis (p = 0.001), have been treated with chemotherapy by FQ1 (p < 0.001), live outside the metropolitan area (p = 0.007), report being on an aromatase inhibitor rather than tamoxifen at FQ1 (p = 0.003), be younger (by about 2 years) (p = 0.002), have a higher body mass index (by about 1 unit of BMI) (p < 0.001) and have had more lymph nodes removed (about four more) (p < 0.001) (Table 1). Nonetheless, Fig. 2 illustrates the large variation in the number of nodes removed in both groups.
Table 1

A comparison of women who did and did not report lymphedema at FQ1


No lymphedema 1,266 (80.3 %)

With lymphedema 311 (19.7 %)

Total 1,577


N (%)

N (%)

N (%)


p value

Age at diagnosis (median, range)

57.2 (26.7, 88.7)

55.1 (29.0, 84.9)

56.7 (26.7, 88.7)



Body mass index (median, range)

25.4 (16.3, 54.2)

26.5 (18.6, 54.0)

25.6 (16.3, 54.2)



Number of nodes removed (median, range)

9.0 (0, 36)

13.0 (0, 40)

10.0 (0,40)



Stage at diagnosis

Stage 1

632 (51.8)

126 (41.2)

758 (49.7)



>Stage 1

588 (48.2)

180 (58.8)

768 (50.3)

Type of surgery by FQ1


371 (29.4)

107 (34.4)

478 (30.4)




892 (70.6)

204 (65.6)

1,096 (69.6)

Reports active disease

Active disease by FQ1

37 (2.9)

8 (2.6)

45 (2.9)



No active disease by FQ1

1,229 (97.1)

303 (97.4)

1,532 (97.1)

Radiotherapy at any point up to FQ1

Treated with radiotherapy

967 (76.4)

239 (76.8)

1,206 (76.5)



Not treated with radiotherapy

299 (23.6)

72 (23.2)

371 (23.5)

Chemotherapy at any point up to FQ1

Treated with chemotherapy

603 (47.6)

187 (60.1)

790 (50.1)



Not treated with chemotherapy

663 (52.4)

124 (39.9)

787 (49.9)

Endocrine therapy


360 (28.4)

87 (28.0)

447 (28.3)




332 (26.2)

110 (35.4)

442 (28.0)

Tamoxifen (reference group)

574 (45.3)

114 (36.7)

688 (43.6)


Year 12 or below

691 (54.7)

164 (52.7)

855 (54.3)



Post secondary school

573 (54.3)

147 (47.3)

720 (45.7)

Location of residence at EQ

Metropolitan area

879 (69.4)

191 (61.4)

1,070 (67.9)



Non-metropolitan area

387 (30.6)

120 (38.6)

507 (32.1)

Partnership status at FQ1

Has a partner at FQ1

696 (76.5)

241 (77.7)

1,210 (76.8)



No partner at FQ1

297 (23.5)

69 (22.3)

366 (23.2)

Smoking status at FQ1

Current smoker

117 (9.2)

22 (7.1)

139 (8.8)



Non- or past smoker

1,149 (90.8)

289 (92.9)

1,438 (91.2)

Consumes alcohol at FQ1


930 (73.7)

238 (77.0)

1,168 (74.3)




332 (26.3)

71 (23.0)

403 (25.7)

There were 1,588 women who completed FQ1, but 11 women are not included as they did not complete the question about lymphedema; not all totals add to 1,577 due to small numbers of missing data points

aMann–Whitney test
Fig. 2

Number of women with and without lymphedema in each category determined by the number of nodes removed at surgery

Lymphedema was not more likely amongst women who were treated with mastectomy compared with breast-conserving surgery and was also not more likely amongst those who reported treatment with radiotherapy.

The single most important factor contributing to the likelihood of lymphedema at FQ1 was the number of lymph nodes removed (p < 0.001), although this variable alone only explained about 7 % of the likelihood of reporting lymphedema (Nagelkerke R2 0.067) (Table 2). Other variables we found to be positively associated with the presence of lymphedema were body mass index (p < 0.001), living outside of the metropolitan area (p = 0.024) and being treated with an aromatase inhibitor rather than tamoxifen (p = 0.004). Age was inversely associated with the likelihood of reporting lymphedema (p = 0.005). Our most complete model only explained 10 % of the total likelihood of reporting lymphedema (Nagelkerke R2 = 0.10).
Table 2

Logistic regression modelling factors associated with lymphedema at FQ1 (forward regression (n = 1,500, R2 = 0.10))

No. added


Nagelkerke R2

Odds ratio

p value


Number of nodes removeda





Body mass index (kg/m2)





Age at diagnosis





Endocrine therapy




None vs. tamoxifen




Aromatase inhibitor vs. tamoxifen





Metro vs. country




aData provided by the Victorian Cancer Registry

The total well-being score at FQ1 for women reporting lymphedema was lower by about six points compared with women who did not report lymphedema (Table 3). The scores for each of the domains of the PGWB (apart from the domain of self-control) were also lower in the women reporting lymphedema. In a multi-linear regression model, lymphedema was associated with a lower total PGWB score at FQ1 (p < 0.001) as were smoking (p < 0.001), the presence of active disease (p < 0.001), being educated beyond school (p = 0.001) and having a higher body mass index (p = 0.009) (Table 4). Being older (p < 0.001), having a partner (p = 0.001) and being on tamoxifen compared with either being on no endocrine therapy or an aromatase inhibitor (p < 0.001) were each associated with a higher total PBWB score at FQ1.
Table 3

A comparison of PGWB domain and total scores for women who did and did not develop lymphedema at FQ1 (n = 96 women missing as they withdrew prior to FQ1, n = 11 women missing because their lymphedema status was missing)


No lymphedema 1,266 (80.3 %)

With lymphedema 311 (19.7 %)

Total 1,577

Mann–Whitney M-W test


Median, mean (SD)

Median, mean (SD)

Median mean (SD)


p value

PGWB subscale—anxiety

19.0, 18.3 (4.8)

17.5, 16.8 (5.3)

19.0, 18.0 (4.9)



PGWB subscale—depressed mood

13.0, 12.7 (2.7)

13.0, 12.0 (3.1)

13.0, 12.5 (2.8)



PGWB subscale—positive well-being

14.0, 13.2 (3.6)

13.0, 12.4 (3.9)

14.0, 13.1 (3.7)



PGWB subscale—self-control

9.0, 9.3 (1.6)

9.0, 9.2 (1.6)

9.0, 9.3 (1.6)



PGWB subscale—general health

10.0, 10.0 (2.8)

9.0, 9.0 (3.0)

10.0, 9.8 (2.9)



PGWB subscale—vitality

13.0, 12.3 (3.8)

11.0, 11.0 (4.1)

12.0, 12.1 (3.9)



PGWB total

78.0, 75.9 (15.8)

72.0, 70.2 (17.5)

77.0, 74.7 (16.3)



SD standard deviation

Table 4

Multiple linear regression modelling for factors associated with the total PGWB score (forward regression, n = 1,521, R2 = 0.104)

Number added



R2 change

Beta coefficient

p value


Current smoker at FQ1






Age at diagnosis






Lymphedema at FQ1






Active disease at FQ1






Partnered at FQ1






Tamoxifen (vs. none/AI)






Education beyond year 12












Our study has confirmed that about one in five women reported lymphedema by 2 years post diagnosis of invasive breast cancer and that this proportion remains relatively constant until nearly 4 years from diagnosis. Although the likelihood of lymphedema 2 years post diagnosis was associated with removal of more lymph nodes from the axilla at the time of initial surgery for breast cancer, this variable explained only a small proportion of the likelihood of reporting lymphedema.

The majority of women with lymphedema in our study report were only mildly bothered by it, although the mean psychological well-being of women reporting lymphedema was lower than that of unaffected women, even when other factors known to be associated with reduced well-being were taken into account. The magnitude of the difference in the overall well-being score was of a similar order of magnitude as we have documented with other chronic health conditions such as urinary incontinence [11] and back pain [12].

A novel finding is that self-reported lymphedema fluctuates over time. Some women who reported lymphedema early did not continue to report it. This could be because it resolved spontaneously or no longer bothered them enough, or it could have responded to treatment. Although we asked women about the use of massage therapy in our questionnaires, this question was not specifically related to the treatment of lymphedema so we are unable to assess the impact of therapy on resolution of the condition. Some women reported lymphedema for the first time nearly 4 years from diagnosis, possibly as a consequence of an event such as a skin infection which highlighted the presence of the condition which had previously gone unnoticed. There has been a recent report of a woman with a past history of breast cancer developing lymphedema for the first time after a vaccination [13].

More conservative axillary surgery has been associated with reduced likelihood of morbidity from lymphedema [14, 15] and our data did show a significant association between lymphedema and the number of axillary nodes removed. However, lymphedema has been reported in women whose only axillary surgery has been sentinel lymph node biopsy and who have only had on average three lymph nodes removed [4].

Our estimate of one in five fits well with the recent Australian review which included studies using a number of different objective measurements to define the presence of lymphedema [5]. Our results suggest that the proportion of women who reported lymphedema at some point after their treatment is likely to be more than one in five. By the time of completion of the FQ3 nearly 4 years from diagnosis, about 27 % of women had reported lymphedema at some point after treatment but only 11 % had reported it in each questionnaire. Consistent with this picture, a study which ascertained the presence of lymphedema in women for 20 years after mastectomy and axillary dissection has shown that 49 % reported lymphedema at some point [16].

Many factors will affect the findings of studies of lymphedema including the characteristics of the population recruited, particularly stage at diagnosis (which will influence treatment including management of the axilla) and also the distribution of age and body mass index [17, 18]. The estimated prevalence of lymphedema will also be affected by whether women are assessed on only one occasion (and the time from diagnosis at which that assessment is performed) or whether they are followed over time. If women are followed over time, the findings will be affected by whether women are lost to follow-up between assessments (as in our study) in which case each assessment of prevalence is performed on a slightly different group of women each time or whether the repeated assessments are restricted to women available for all assessments (in which case the study is restricted to a relatively low risk group as all women must have survived to complete all the assessments). The statistic chosen for reporting is also important. Most studies report a point prevalence or a series of point prevalences while others report a cumulative incidence [1921]. Where a cumulative incidence is reported, once a woman has developed lymphedema, she will always be counted in the lymphedema group. This approach is limited in terms of assessing a condition which our results would suggest is quite dynamic.

The feature of lymphedema studies which has received the most attention has been the method of assessing the presence of lymphedema. Objective methods used include the following: differences in arm circumference (either before or after studies or comparing one arm with the other), water volume displacement studies, perometry (where an array of optoelectronic sensors is used to assess limb volume) and bioimpedance spectroscopy to assess the volume of extracellular fluid [22]. Studies which have used different objective methods in the same women have demonstrated that they give different results [6] and studies in which subjective assessment has been included have shown that it compares favourably with objective approaches [23, 24]. Although the apparent attraction of objective methods is reproducibility, they miss the point in terms of the impact of lymphedema on the affected women and this is what is critical from a clinical care perspective.

The strength of our study is that it is large and generalisable to women diagnosed with their first episode of invasive breast cancer in the southern Australian state of Victoria. Another strength is that we assessed women on more than one occasion between 2 and 4 years from the time of diagnosis and we did not assume that once a woman had reported the presence of lymphedema, she would always be classified as having lymphedema. Although assessing the presence of lymphedema using self-report could be considered a weakness [5], self-report has been found to be a sensitive measure of the impact of lymphedema on the individual and objective measures have arbitrary cut-off values and do not agree well with each other [6]. Another weakness could be that we only asked women about the presence of lymphedema and did not ask about feelings of “heaviness” or problems with rings not fitting or difficulty writing or other symptoms and signs that could have indicated the presence of lymphedema [19]. However, we did explain what lymphedema was in a preamble to the question.

In conclusion, we have confirmed that self-reported lymphedema is experienced by at least 27 % of breast cancer survivors by the time they are 4 years from diagnosis, with most women being only mildly bothered by the condition. Despite this, lymphedema is independently associated with lower self-reported psychological general well-being. It is important for health-care providers to understand that the phenomenon is a dynamic one and that some women may report lymphedema for the first time as long as 4 years from diagnosis. As the survival of women with breast cancer continues to improve [25], lymphedema remains a clinical priority.


The authors wish to thank the study participants and the members of our Study Advisory Group: Dr Jacquie Chirgwin, A/Professor John Collins, Professor Graham Giles, Mr Peter Gregory, Mr Stewart Hart, Ms Suzanne Neil and Mrs Avis McPhee. The authors also wish to thank the members of the research team of the Health and Wellbeing After Breast Cancer study, without whose hard work this large cohort study would not be possible (Maria La China and Jo Bradbury). Finally, we thank Ms Helen Farrugia, Director of Information Systems and Professor Graham Giles, Director, of the Victorian Cancer Registry, for their ongoing support of this study.

Funding sources

This work was supported by the BUPA Health Foundation (previously the Medical Benefits Fund of Australia Limited Foundation) (to SRD and RJB), the National Health and Medical Research Council of Australia (grant nos. 219279 to SRD and RJB, 490938 to SRD), Novartis Oncology Australia, the L.E.W. Carty Trust, the Jack and Robert Smorgon Families Foundation and Connie and Craig Kimberley and Roy Morgan Research (all to SRD and RJB). This research project was also supported by the Victorian Government through a Victorian Cancer Agency Research Fellowship (to RJB).

Conflict of interest

None of the authors consider that they have any conflict of interest that could inappropriately influence or bias this work. None of the funding agencies had any role in determining study design; in the collection, analysis and interpretation of data; and in the writing of the manuscript or in the decision to submit the manuscript for publication.

Ethical approval

The study is being carried out with the approval of the Ethics Committee of the Cancer Council of Victoria and the Human Ethics Committee of Monash University and all participants have provided written informed consent.

Copyright information

© Springer-Verlag Berlin Heidelberg 2013