Cohort study of association of risk of breast cancer with cyst type in women with gross cystic disease of the breastBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7085.925 (Published 29 March 1997) Cite this as: BMJ 1997;314:925
- Paolo Bruzzi, heada,
- Luigi Dogliotti, professor of oncologyb,
- Carlo Naldoni, deputy headc,
- Lauro Bucchi, epidemiologistc,
- Massimo Costantini, epidemiologista,
- Alessandra Cicognani, research fellowc,
- Mirella Torta, research assistantb,
- Gian Franco Buzzi, directorc,
- Alberto Angeli, professor of internal medicined
- a Unit of Clinical Epidemiology and Trials National Institute for Cancer Research Largo Rosanna Benzi 10 16132 Genova Italy
- b Department of Clinical and Biological Sciences Medical Oncology University of Turin at St Luigi Hospital 10043 Orbassano (Torino) Italy
- c Cancer Prevention Center St. Maria delle Croci Hospital Via Missiroli 10 48100 Ravenna Italy
- d Department of Clinical and Biological Sciences Internal Medicine University of Turin at St. Luigi Hospital 10043 Orbassano (Torino) Italy
- Correspondence to: Dr Bruzzi
- Accepted 16 January 1997
Objective: To assess correlation between type of breast cyst and risk of breast cancer in women with gross cystic disease of the breast.
Design: Cohort study of women with breast cysts aspirated between 1983 and 1993 who were followed up until December 1994 for occurrence of breast cancer.
Setting: Major cancer prevention centre.
Subjects: 802 women with aspirated breast cysts.
Main outcome measures: Type of breast cyst based on cationic content of cyst fluid: type I (potassium:sodium ratio >1.5), type II (potassium:sodium ratio <1.5), or mixed (both types). Subsequent occurrence and type of breast cancer.
Results: After median follow up of six years (range 2-12 years) 15 cases of invasive breast cancer and two ductal carcinomas in situ were diagnosed in the cohort: 12 invasive cancers (and two carcinomas in situ) among the 417 women with type I cysts, two cancers among the 325 women with type II cysts, and one among the 60 women with mixed cysts. The incidence of breast cancer in women with type I cysts was significantly higher than that in women with type II cysts (relative risk 4.62 (95% confidence interval 1.26 to 29.7)). These results were confirmed after adjustment for several risk factors for breast cancer (relative risk 4.24 (1.12 to 27.5)).
Conclusions: The increased risk of breast cancer of women with breast cysts seems to be concentrated among women with type I breast cysts.
Several studies have shown that women with palpable cysts in their breasts are at increased risk of breast cancer
Two types of breast cyst can be identified–type I cysts, with low concentrations of sodium and high concentrations of potassium ions, and type II cysts, with opposite characteristics
We investigated the correlation between cyst type and risk of breast cancer in 802 women with aspirated breast cysts
After median follow up of six years, the women had a relative risk of breast cancer of 1.69 compared with the general population, and those with type I cysts had a risk four times higher than those with type II cysts
The excess risk of breast cancer of women with breast cysts seems to be concentrated among women with type I cysts, but the size and duration of this increased risk are still to be assessed
Gross cystic disease of the breast, defined on the basis of the presence of palpable cysts, is reported to occur in 7% of women in the Western world.1 Overall, the available evidence suggests that women with breast cysts have a moderately increased risk of breast cancer.1 2 3 4 5 6 7 8 9 10 11 It should be noted that breast cysts are not considered to be premalignant lesions but are simply markers of an increased risk affecting the whole organ.1
It has been observed that the fluid in breast cysts consistently shows a bimodal distribution in the concentration of many substances, including cations,12 13 14 15 16 17 hormones,12 13 14 15 16 17 18 19 and growth factors.15 19 20 Thus, two major types of cysts can be identified21: type I cysts have high concentrations of potassium and low concentrations of sodium and chloride, high concentrations of androgen and oestrogen conjugates, and high concentrations of epidermal growth factor; type II cysts have an electrolyte composition more similar to that found in plasma (high concentrations of sodium and chloride and low concentrations of potassium) and lower concentrations of sex hormones and epidermal growth factor. It has been suggested that women with type I breast cysts have an increased risk of developing breast cancer.10 21
In 1983 we started a cohort study to evaluate the relation between the risk of breast cancer and the cationic composition of breast cysts in women with gross cystic disease. We previously reported the distribution of epidemiological and mammographic risk factors for breast cancer and the rates of cyst recurrence among the women of this cohort according to their type of cyst(s).22 23 24 25 We present here the incidence of breast cancer by cyst type.
Between 25 February 1983 and 15 February 1993, 1323 women aged 30-69 underwent aspiration of one or more breast cysts at the Cancer Prevention Center of Ravenna, Italy. Of these, we excluded 229 women who did not meet the eligibility criteria for our study–38 had had a previous diagnosis of cancer of any site including breast, eight had breast cancer diagnosed at the initial examination, 22 were not living in the local area, and 161 had provided samples of cyst fluid of less than 1 ml. For a further 292 women, the fluid of their breast cyst was not assayed for cationic content–either because, in accordance with the original study protocol, electrolyte concentrations were not measured for women enrolled before 1990 if less than 3 ml of fluid was aspirated from their cyst or because of accident. As a consequence, our present report concerns 802 women.
Classification of cysts
Sodium and potassium concentrations in cysts fluid were determined by flame photometry after 15 minutes of centrifugation and appropriate dilution. We calculated the ratio of potassium to sodium concentrations and, according to the original protocol,23 used a cut off value of 1.5 in the ratio to divide breast cysts into two groups–type I cysts with a potassium:sodium ratio of 1.5 and type II cysts with a ratio of <1.5.
Detection of breast cancer
All the women in the cohort were included in a follow up programme for detecting breast cancer based on yearly clinical and ultrasound examination and biennial mammography. We also searched for the women's names in the files of the Romagna Cancer Registry.26 After 31 December 1994, we questioned all the women by telephone about their medical history. Thus far, only seven women have been lost to follow up: one emigrated abroad after 122 months; five could not be traced after 24, 31, 49, 104, and 113 months of follow up respectively; and one could not be traced after the initial aspiration of her cyst.
We classified the women according to the type of their first cyst that had been assayed for cationic composition. Those presenting with multiple cysts of the same type were assigned to the corresponding group, but if they had cysts of both types they were assigned to a third group (mixed). We calculated the time of follow up for each woman from the date of aspiration of the index cyst to 31 December 1994 or to the date of diagnosis of breast cancer or death from any cause, whichever occurred first. The follow up times for the seven women who were lost to follow up and for one woman who underwent prophylactic bilateral mastectomy were censored at the time of the last follow up examination.
We calculated age standardised ratios of incidence of invasive breast cancer to compare the rates in the three groups of women in our cohort with those of the general population living in the local area (derived from the age specific rates reported by the Romagna Cancer Registry from 1986 to 198826). We calculated exact 95% confidence intervals by direct exploration of the profile likelihood function in a Poisson regression model,27 and, similarly, used a multivariate Poisson regression model to estimate relative risks and exact 95% confidence intervals when we compared the three groups directly, both including and excluding cases of ductal carcinoma in situ.
Our study was designed to detect a fourfold increase in risk of breast cancer in women with type I breast cysts relative to women with other types of cysts, assuming that the incidence in the whole cohort was twice that observed in the general population of the Romagna Cancer Registry.23 The analysis was planned after 4400 person years of observation. By 31 December 1994, the 802 women in the cohort had contributed 5112 person years of observation, and the median follow up time was 2219 days (range 684-4358).
Table 1) shows the baseline characteristics of the 802 women in the study cohort: 417 (52%) had type I breast cysts, 325 (41%) had type II, and 60 (7%) had mixed cysts. Compared with those with type II and mixed cysts, the women with type I cysts were younger (P=0.047) and reported fewer births (P=0.005).
By 31 December 1994, 17 cases of breast cancer had been diagnosed in the cohort–15 cases of invasive breast cancer (14 epithelial cancers and one malignant phyllodes tumour) and two cases of ductal carcinoma in situ. Fourteen of the cancers (including the two ductal carcinomas in situ) occurred in the women with type I cysts, two occurred in those with type II cysts, and one in those with mixed cysts. The median time from aspiration of the index cyst to diagnosis of cancer was 38.1 months (range 2.3-129.0). In nine cases the cancer was found in the same breast as the index cyst had been, in five cases it was found in the other breast, and in three cases bilateral cysts had been aspirated.
The incidence of invasive breast cancer in the whole cohort was higher than that expected from the rates reported by the Romagna Cancer Registry (age standardised incidence ratio 1.69 (95% confidence interval 0.97 to 2.70)) (table 2). The increased risk was concentrated in the women with type I cysts (12 cases observed v 4.58 expected, age standardised incidence ratio 2.62 (1.40 to 4.40)), while no excess was seen among the women with type II and mixed cysts.
Table 3) shows the results of direct comparison among the three groups of women: the risk of invasive breast cancer in the women with type I cysts was significantly higher than in the women with type II cysts both in univariate analysis (relative risk 4.62, (95% confidence interval 1.26 to 29.7)) and in multivariate analysis (4.24 (1.12 to 27.5)). The association was stronger when the two cases of ductal carcinoma in situ were included in the analyses (relative risks 5.41 (1.51 to 34.4) and 5.06 (1.38 to 32.5) respectively). The women with multiple cysts at enrolment were not at increased risk of breast cancer compared with women with solitary cysts.
Our study shows that among 802 women with gross cystic disease the risk of breast cancer was associated with the cationic content of fluid from their breast cysts. Despite the small number of cases of cancer, the association was significant, and various considerations suggest that it was not due to chance, bias, or confounding from known risk factors for breast cancer.
Validity of results
The primary aim of our study was to compare the incidence of breast cancer in women with type I cysts with that in women with type II cysts, and the analysis was conducted at a time specified in advance. The choice of a cut off value of 1.5 in the ratio of potassium to sodium concentrations for classifying the cysts was already indicated in the original study protocol.23 The appropriateness of this cut off value was supported by our subsequent analysis of a subset of women from the same cohort in which we used an independent criterion–the rate of recurrence of cysts.22 Indeed, the association between type of cyst and risk of breast cancer would have been even stronger if we had used the cut off value of 0.33 for the potassium:sodium ratio that was proposed by Dixon et al13 and used in several studies since one of the women with invasive breast cancer had had an index cyst with a potassium:sodium ratio of 1.45 (see fig 1).
In our series of 17 cancers two (12%) were ductal carcinomas in situ, a higher proportion than that reported in the same age groups by the Romagna Cancer Registry (5.3%). This suggests that our comparison with the registry rates might be biased due to overdiagnosis. However, this problem does not apply to the internal comparisons among the different groups of women within the cohort since compliance to follow up was virtually complete and was independent of cyst type. Incidentally, the age standardised incidence ratio of breast cancer in our cohort was only 1.69 compared with the general population of the area, close to the relative risk reported in another Italian study,4 which was the lowest of those reported from studies of similar cohorts.28
Finally, the relative risk of breast cancer remained virtually unchanged after we had adjusted for several known risk factors for breast cancer. Therefore, our observed association seems to have been a real one, which supports the view that breast cysts are not a homogeneous entity but should be separated into two major populations.
Comparison with other studies
Our observation of an association between type I breast cysts and risk of breast cancer is supported by other studies showing an increased risk of breast cancer among women with multiple aspirations of cysts2 9 and among women with apocrine cysts,11 since an increased rate of cyst recurrence has been observed among women with type I cysts22 29 and these are more often lined by apocrine epithelium than type II cysts.13
Miller et al reported that, among 18 women who subsequently developed breast cancer, the distribution of cyst type was clearly different from that observed in a group of women who did not develop breast cancer.21 In contrast, however, Ebbs and Bates failed to find any relation between the potassium:sodium ratio in fluid from breast cysts and breast cancer in a consecutive series of 101 women with breast cysts,30 but methodological problems possibly invalidate their report.29
Explanation of results
There is no clear explanation for the increased risk of breast cancer among women with type I breast cysts. Breast cysts are not precursor lesions of cancer, and the occurrence of a cancer within a cyst is quite rare. In our study, in agreement with previous reports,1 the risk of cancer was not limited to the breast from which a cyst had been aspirated. Thus, a high potassium:sodium ratio in cyst fluid seems to be a marker of an increased risk affecting the whole breast tissue.
The bimodal distribution of cations in breast cysts' fluid is mirrored by the distribution of other substances, such as steroid sex hormones12 13 14 18 19 and growth factors.15 19 20 The concentration of mitogenic growth factors such as epidermal growth factor is substantially higher in type I cysts,15 whereas transforming growth factor ß–which is claimed to have an inhibitory effect on growth of tumour cells of epithelial origin–is preferably accumulated in type II cysts.31
Our study suggests that the increased risk of breast cancer among women with gross cystic disease, previously observed in several cohort studies, is limited to a subgroup of these women. Such women can be identified on the basis of a high ratio of potassium to sodium ions in the fluid aspirated from breast cysts. The potential clinical implications of this finding are considerable in view of the simplicity and worldwide availability of these measurements. From on our data, however, it is not possible to estimate with acceptable precision the size of the increased risk among women with type I cysts, nor to assess its duration. These issues need to be further explored by longer follow up of our cohort and in larger, independent cohort studies.
Presented in part at the 31st Annual Meeting of the American Society of Clinical Oncology, Los Angeles CA, USA, 20-24 May 1995, and at the 4th Cherasco Meeting on Gross Cystic Disease of the Breast, Cherasco, Italy, 23 June 1995. We thank Licia Tavolazzi (Chemical Pathology Laboratory, Santa Maria delle Croci Hospital, Ravenna), Antonio Mazzotti (RIA Laboratory, Santa Maria delle Croci Hospital, Ravenna), and Luca Boni (Unit of Clinical Epidemiology and Trials, National Institute for Cancer Research, Genova) for their participation in the study.
Funding: This study was supported by a grant from the Istituto Oncologico Romagnolo, Forlì, Italy (contract No 90133.1); a grant from the Associazione Italiana per la Ricerca sul Cancro (AIRC), Italy (Family history of cancer, inheritance, and molecular mechanisms in benign and malignant breast diseases); and a 60% grant from MURST, Italy.
Conflict of interest: None.