| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Addenbrooke's Hospital, University of Cambridge,
Cambridge, United Kingdom.
2 Cambridge Breast Unit, Addenbrooke's Hospital, University of Cambridge, Hills
Rd., Cambridge, CB2 2QQ, United Kingdom.
3 Strangeways Research Laboratory, Institute of Public Health, University of
Cambridge, Cambridge, United Kingdom.
4 Department of Radiology, Norfolk and Norwich University Hospital, Norwich,
United Kingdom.
5 Clinical Gerontology Unit, Addenbrooke's Hospital, University of Cambridge,
Cambridge, United Kingdom.
Received March 2, 2005;
accepted after revision April 26, 2005.
Address correspondence to M. Kataoka
(mk435{at}cam.ac.uk).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. The study population, nested within the European Prospective Investigation of Cancer-Norfolk (EPIC-Norfolk) cohort study, consisted of 1,590 women older than 55 years, not taking hormone replacement therapy, and with available screening mammograms. Mammograms were coded by three radiologists for presence or absence of BAC. History of coronary heart disease (CHD), stroke, and diabetes and risk factors for cardiovascular disease (including smoking status, body mass index [BMI], blood pressure, diabetes, and glycosylated hemoglobin [HbA1c]) were independently measured from health examinations in the EPIC study.
RESULTS. The prevalence of BAC was 16.0%. Women with BAC were significantly older than those without it. BAC was associated with prevalent CHD, but not stroke. The odds ratio of having CHD was 2.54 (95% confidence interval, 1.03-6.30). The sensitivity and specificity were 32.4% and 85.5%, respectively. Except for smoking, which showed an inverse association, there was no consistent significant association of BAC with cardiovascular disease risk factors including BMI, diabetes, HbA1c, or lipids.
CONCLUSION. BAC found on mammograms was associated with prevalent CHD after adjustment for age, but with low sensitivity. BAC may provide additional information toward identifying cardiovascular disease risk among otherwise healthy women.
Keywords: breast • breast arterial calcification • cardiovascular disease • mammography
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Breast arterial calcification (BAC) is a mammographic finding unrelated to cancer and frequently observed among elderly women [9, 10]. BAC is classified as medial arterial calcification, or Mönckeberg calcification, which is different from intimal calcification [11]. Several studies suggest that BAC is associated with cardiovascular disease, diabetes, or hypertension and may be used as a marker of arterial disease or cardiovascular disease [10, 12-16]. Although a single case-control study refuted the association between BAC and CHD [17], two cohort studies showed increased risk of cardiovascular disease [10, 15]. The inconsistency among the studies may be attributed to the variations in the study population including the age distribution, disease prevalence, or disease risk (Table 1).
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Within this population were 1,590 women who had participated in the National Health Service Breast Screening Programme (NHSBSP) and who, therefore, had available screening mammographies. Their mammograms were retrieved to review for a mammography density study. BAC was also recorded as part of the interpretations.
Mammogram Data
The mammograms of the 1,590 women were collected and digitized for
analysis. Because all available mammograms were collected, more than one
mammogram was available for some of the women. The mammograms were taken by
screen-film mammography between 1985 and 2001. During that period, the film
brand, optical density, screens, and equipment related to image processing
underwent considerable change in 1994.
Interpretations of the mammograms were recorded independently by three experienced radiologists who had reviewed mammograms in their practices for 17, 8, and 7 years. The original interpretations were performed primarily for assessing mammographic density. BAC was evaluated as a secondary outcome. When more than one mammogram from a woman was available, the mammograms were evaluated sequentially on the same reporting session. BAC on a mammogram was defined as the presence of two parallel calcific lines in a railroad track configuration, which is distinct from calcium within breast ducts [9, 11, 12] (Fig. 2). Some of the mammograms have two views (mediolateral oblique and craniocaudal); in these, the two views of each breast were evaluated together for BAC. BAC was counted as positive if it was found in either the right or left breast, in both sides, in either of the two views. BAC was evaluated as present or absent— the severity of calcification was not evaluated.
|
For analysis, one representative BAC status for a woman was summarized from mammograms taken from several occasions. Because the prevalence of BAC increased with advancing age [9, 10], using the data from the latest mammography likely maximized the prevalence and thus maximized the statistical power. However, some of BAC might be missed because it was not associated with cancer and not the primary outcome. Therefore, we classified women as BAC-positive if any of their mammograms were coded as positive.
The breast density recorded by the three reviewers was used for the current analysis to explore the effect of breast density on the prevalence of BAC. Interpretation was based on six density categories by Boyd et al. (22), and the midpoints of the density categories in the form of percentages were used as point values. Results for the both right and left breasts were averaged to give a summary density value for an individual on a particular screening date.
Measurements from the EPIC-Norfolk Study
The disease status and risk factors were obtained from the EPIC-Norfolk
study. Participants underwent a survey between 1993 and 1997 by completing a
health and lifestyle questionnaire, including the question: "Has a
doctor ever told you that you have any of the following?" Below the
question, conditions including diabetes, heart attack (coronary heart
disease), and stroke were listed. Smoking history was also obtained, and
participants were examined by trained nurses. Body mass index (BMI) was
calculated as weight in kilograms divided by height in meters squared.
Systolic and diastolic blood pressures (BP) were measured after 5 minutes'
resting; the two interpretations were averaged for analysis. Blood samples
were obtained to assess serum total and LDL cholesterol level. From November
1995, an additional ethylenediaminetetraacetic acid (EDTA)-anticoagulated
blood sample was taken for measurement of glycosylated hemoglobin
(HbA1c).
Statistical Analysis
The analysis included all 1,590 women who participated in the EPIC-Norfolk
cohort study and whose mammograms from the NHSBSP were available. The
interobserver variability of all three reviewers for all 4,407 mammograms was
calculated using Cohen's kappa statistics. Because reviews were conducted per
participant in a sequential way, not per film, the interpreting results for
mammograms from one specific subject might not be independent. Therefore, data
from the first mammogram, which was unlikely to be affected by the following
mammogram interpretations, were also calculated. The results from all three
reviewers were combined (coded positive if at least one reviewer recorded as
positive) and used as one reference value.
To explore the possible effect of change in image quality, the prevalence of BAC among films before 1994 (until December 31, 1993) were compared with the BAC among films after January 1, 1994, with adjustment for age. For this analysis, it is assumed that the baseline prevalence of BAC was stable over the study period. Similarly, the effect of breast density on the prevalence of calcification was explored by comparing the prevalence of BAC in dense breasts with BAC in fatty breasts. The median value of the density among all mammograms was 26.7, and 27 was used as a cutoff point for fatty and denser breasts.
The coded information of BAC status in each participant was compared with the data of cardiovascular disease status or risk factors. To examine the first hypothesis (BAC is associated with cardiovascular disease), BAC status was considered as exposure and cardiovascular disease status was considered as outcome. Analysis was performed for CHD and stroke separately and for the two combined as "cardiovascular disease." Logistic regression was used to determine the relation between BAC and cardiovascular disease. The odds ratio (OR) of having disease when BAC was positive versus when BAC was negative was calculated with 95% confidence intervals (CIs). Because age was considered a strong determinant of BAC [9], the age group from whom the mammograms were obtained was regarded as a confounder. Based on the age when their last or first mammograms were taken, women were classified into 5-year age categories starting from 50 and younger, 51-55, 56-60, 61-65, 66-70, and 71 and older.
For examining the second hypothesis (established risk factors are associated with BAC status), risk factors were treated as exposure and BAC was regarded as outcome. For continuous variables such as BMI (kg/m2), total cholesterol level (mmol/L), LDL cholesterol level (mmol/L), systolic and diastolic BP (mm Hg), HbA1c level (%), data from BAC-positive women were compared with data from BAC-negative women and Student's t tests were performed. Again, age was considered as a confounder, and age-adjusted values were calculated using a univariate linear model. In addition to the categoric risk factors such as smoking status and diabetes mellitus (self-report), newly diagnosed diabetes mellitus was defined using HbA1c of more than 7% as a cutoff point. Hypertension was defined as systolic BP more than 160 mm Hg or diastolic BP more than 100 mm Hg, based on moderate hypertension defined by the 1999 World Health Organization (WHO)-International Society of Hypertension (ISH) hypertension guidelines [23]. For the analysis of categoric values, logistic regression was used to determine the relation between the risk factor and BAC, with OR estimated for having positive BAC if the risk factor was present versus absent.
The sample size was calculated for the difference in proportions of cardiovascular diseases between the BAC-positive group versus the BAC-negative group at a power of 80% and significance of 5%. Given the older age of the study population, the prevalence of BAC was estimated to be 15%. Because the study population was community-based with women who attended breast screening, disease prevalence was estimated to be 2.5%. To detect a difference of disease prevalence of more than 5%, the sample size needed was estimated as 1,550, slightly below the size of the available population.
All the statistical analyses were performed using ststistical software (SPSS for Windows version 11.0, SPSS and Stata version 8.0, Stata).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The prevalence of BAC among the 1,590 women according to positive BAC at the first mammogram, positive BAC at the last mammogram, and positive BAC at least once among all available mammograms (used for the main analysis) was calculated. The prevalence of BAC using these definitions was 7.5% (120 of 1,590), 15.5% (246 of 1,590), and 16% (254 of 1,590), respectively. Age at last mammogram was 63.2 years on average, and age at the first mammogram was 57.6 years on average. Among 1,331 participants with more than one mammogram, 134 were women who were documented as having appearance of BAC (10.1%) whereas only nine were women who were documented as having disappearance of BAC during the course of mammogram follow-up (0.7%). (We could not confirm whether the disappearance of BAC in the nine women was true- or false-negative because of underreporting.)
The recorded interpretation data of the three independent reviewers were compared. For all 4,407 mammograms, 506 mammograms were coded as BAC-positive when all three reviewers' results were combined. Each reviewer coded 414 (82%), 278 (55%), and 254 (50%) mammograms as positive, respectively. For interobserver variability for detecting BAC at all mammograms, the kappa value ranged from 0.603 to 0.614, suggesting moderate agreement. The analyses were repeated for the first mammograms (total number, 1,590; of which positive mammograms were 120 based on combined results). Each interpreter coded 97 (81%), 52 (43%), and 48 (40%) mammograms as positive, respectively, with the kappa value ranging from 0.474 to 0.566.
The ratio of positive BAC at mammograms before 1994 was 8.1% (130 of 1,600), whereas after 1994 the ratio was 13.4% (376 of 2,807). The difference was not significant after adjustment for age at which mammograms were taken (crude OR 1.75, 95% CI, 1.42-2.16; age-adjusted OR 1.18, 95% CI, 0.95-1.47). Although there might be a modest increase in detection rate related to improved image quality, no correction was made in the following analyses for the year the mammograms were obtained because of the small effect and potential complexity.
Density value was missing in one mammogram, giving a total of 4,406 mammograms for analysis. The ratio of positive BAC in mammograms in fatty breasts and dense breasts was 12.2% (274 of 2,253) and 10.7% (231 of 2,153), respectively. Although BAC was less frequent in dense breasts, the difference was not significant and the crude and age-adjusted ORs of BAC-positive mammograms were 0.87 (95% CI, 0.72-1.04) and 1.03 (95% CI, 0.85-1.25), respectively.
Breast Arterial Calcification and Cardiovascular Disease and Risk Factors
The average age at health check was 61.6 years, which was 1.6 years younger
than the average age at last mammogram. Disease prevalence was low with CHD
(1.5%), stroke (0.9%), cardiovascular disease (CVD) (2.3%), and diabetes
mellitus (1.9%). Because measurement of HbA1c started from 1995,
36% of the participants lacked the HbA1c data. Among 1,018 women
with available HbA1c measures, 26 (2.6%) had HbA1c of
more than 7%.
Table 2 shows the relationship between BAC status and CVD. The number of participants with known disease was low (24 for CHD, 15 for stroke). BAC-positive women showed a significantly raised OR of having CHD and CVD (CHD and stroke) with the age-adjusted OR of 2.54 (95% CI, 1.03-6.30) and 2.43 (95% CI, 1.19-5.19), respectively.
The association between established risk factors and BAC was examined. Table 3 compares established risk factors between BAC-positive and BAC-negative groups. Age differed significantly between BAC-positive and BAC-negative groups. For categoric values such as smoking status or prevalent diabetes, the ORs of having these conditions in the BAC-positive group versus the BAC-negative group are shown in Table 4. The OR for current smokers was lower among the BAC-positive group, whereas that of former smokers was nonsignificant after adjustment for age. Differences in the prevalence of diabetes mellitus (self-report), high HbA1c (> 7%), diabetes mellitus (self-report and high HbA1c combined), and hypertension were not significant before and after adjustment for age.
|
|
Similar analysis was conducted using BAC at first mammograms instead of any mammograms (data not shown). Because of the low prevalence of BAC (7.5%), the number of participants with disease among BAC-positive women was smaller (six for CHD, three for stroke, two for diabetes mellitus) among BAC positive women, making meaningful analysis difficult.
The sensitivity, specificity, and positive and negative predictive value of BAC found on mammograms for predicting prevalent CVD were calculated. Among 37 women with CVD, 12 women had positive BAC at mammography. Among 1,533 women without CVD, 1,311 women had negative BAC at mammography. Thus, sensitivity, specificity, and positive and negative predictive values were 32.4%, 85.5%, 4.7%, and 98.1%, respectively.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Before discussing this association, we present the quality of the measurements. Moderate agreement occurred among the three reviewers, with a kappa value of 0.5-0.6. Considering that BAC is a benign finding and was not the primary outcome at original recording, disagreement would be caused by underestimation. The prevalence of BAC was 16% when all three reviewers' results were combined. Considering the older age distribution and exclusion of HRT users, this value agrees with those reported in previous studies. Screening mammography has undergone continuous improvement over the decades [24]. Because the mammograms for this analysis were obtained between 1987 and 2001 and considerable film-screen system changes were implemented in 1994, we were concerned that the quality change may affect the apparent BAC prevalence. However, the results showed no significant difference before and after 1994. This may be because BAC is more distinct and easier to recognize than the calcifications associated with breast cancer, which are often minute and obscure. Similarly, the prevalence of BAC was not affected by breast density, probably reflecting the distinct nature of BAC at mammography.
The association between BAC and prevalent CHD shown here is consistent with previous studies in the general population [14, 16], patients with coronary angiography [12], and with cardiovascular mortality in the general population [10, 15]. Although the case-control study by Henkin et al. [17] did not support the association between BAC and CHD diagnosed by angiography, studies based on the general population more consistently report an association between BAC and CHD. The observed increased risk is approximately twofold, which is comparable to the increased risk reported by other studies and higher than the risk associated with aortic calcification (4-8, 10). BAC at a younger age might have a stronger association than BAC in older age, as indicated by Moshyedi et al. [12]. Subgroup analysis limiting mammograms taken at no more than 60 years old might give some clue, but the small sample size and higher age distribution in this population did not allow meaningful analysis. Using a first mammography only might be another option, yet the small number of BAC-positive women with disease would be a limit for this analysis.
The association between BAC and stroke was not significant. This may be because of the small sample size and low prevalence of stroke in this population. The low prevalence of stroke cases might be explained by the devastating nature of stroke, which prevents women from attending breast screening or any other health check. A cohort study examining mortality from stroke reported a nonsignificant relationship with BAC [15], although another study of similar sample size and design showed a significant increase of strokes among BAC-positive women [10]. The association between BAC and stroke appears less consistent than that between BAC and CHD.
Some previous studies reported that the morbidity and mortality from CVD were increased in BAC-positive women when it coexisted with diabetes mellitus [12, 15]. A recent study examining BAC status among diabetic patients suggested that BAC tends to be found with patients suffering from diabetes longer than 8 years, with severe complications [25]. It is possible that BAC in diabetic patients indicates the duration and severity of diabetes, which affects risk of cardiovascular events. Because of the small sample size and low prevalence of diabetes among this population, it was impossible to examine the effect of coexisting diabetes on the association between BAC status and coronary artery disease.
Although BAC may be associated with prevalent CHD, there are some issues to consider in using BAC as a CHD marker. In this study population, the sensitivity of identifying prevalent CVD was only 32.4%, and the positive predictive value of BAC in predicting CHD was below 5%, whereas negative predictive value was 98%. Thus, using BAC as the sole screening method would not be a good idea. Also, the increase in risk is relatively small compared with that by other established cardiovascular risk factors such as smoking, hypertension, and diabetes [26]. Some studies in large populations showed the contribution of BAC as an independent risk factor of CVD by multivariate analysis, adjusting for other established risk factors [10, 14, 15], and showed the increase in risk of 1.3-1.8. Because of the small sample size, multivariate analysis in the current study is unlikely to give meaningful results and thus was not conducted.
No significant association between cardiovascular risk factors and BAC status was found in this analysis. This may be because of lack of stastical power. Incidences of both diabetes mellitus and current smoking were low. It is possible that women who were willing to take part in the EPIC-Norfolk cohort and mammography screening were more health conscious than the general population. The inverse relationship between smoking status and BAC was an unexpected observation. This was partly explained by age confounding because age adjustment attenuated the relationship. Interestingly, this paradoxical relationship was reported in several other studies [10, 14, 15], and the association remained significant after adjustment for age and other factors [10]. Although consistent with several studies, this association seems difficult to explain. BAC might be a better marker of coronary artery disease among nonsmokers.
The underlying mechanism of BAC, a form of medial arterial calcification [11], and its relation to CVD was not well understood. Kemmeren et al. [15] argued that the association of medial arterial calcification with cardiovascular mortality could be the explained by the concomitant presence of intimal diseases. Alternatively, they suggested the association could be explained by medial arterial sclerosis and the occurrence of complications of diabetes mellitus found by Everhart et al. [27]. The latter theory might explain the higher mortality rate among diabetes patients with BAC, which is also supported by the result reported by Fuster et al. [25].
The strength in the current analysis includes the availability of multiple reviewers. BAC was not a primary outcome in the original interpretations, which may be a disadvantage for accurate measurement but would lead to random rather than biased errors. This also indicates the accuracy of BAC as used in the actual screening setting where cancer detection is the most important objective and BAC is likely to be recorded incidentally. This recording is less liable to systematic bias introduced by investigators themselves because their interpretation was not aimed at detecting calcification. In addition, having three independent reviewers made it possible to obtain analyses on interobserver variability. The use of HRT is a strong potential confounder affecting the prevalence of BAC [28] and cardiovascular outcome, so exclusion of HRT users in the selection of the current population ensures a more homogeneous group. The comprehensive measurement of established cardiovascular risk factors is another strength in our analyses. In contrast to some studies, which lack data on major risk factors, health check measurements from the EPIC-Norfolk examination provided standardized measurements on important risk factors (BMI, blood pressure, lipid profile, HbA1c, information on smoking, and history of prevalent diseases).
Among the limitations in this study is the study design itself, which was essentially cross-sectional, making it difficult to explore the temporal relationship between risk factors, BAC, and CVD. Moreover, participants with prevalent disease, such as CHD and diabetes, may have changed their lifestyles, thus affecting the relationship between risk factors and BAC. Another major limitation is the defined sample size, which is too small to make a robust analysis of all related questions. The prevalence of BAC was relatively high compared with other study designs, but the prevalence of CVD and diabetes mellitus was low. Using incidence data might be more informative to explore the association. However, it was impossible to conduct meaningful analyses for incident cases because of the small number of cases of disease. The possible positive effect of coexisting diabetes mellitus with BAC might be an interesting association to look at, but, again, low prevalence of diabetes mellitus made such subgroup analysis difficult.
BAC was not a primary outcome in the original interpretations and thus the accuracy of measurement may be lower than that in more structured studies, in which BAC was the main outcome and standardized criteria were prepared [9]. Nevertheless, the degree of BAC may be subject to error. One method to compensate for the inaccuracy of measurement would be to validate the interpretations of a subset of the mammograms using BAC as a primary outcome, comparing results, and then estimating true prevalence. Unfortunately, the interpretations were completed 2 years ago and the original mammograms are not easily accessible, making it impossible to conduct a validation review.
Disease status was based on self-reports and not confirmed by hospital records or any independent diagnostic system. Therefore, subclinical, undiagnosed, and unreported disease status would have been missed in the analyses, whereas overreporting of the disease status might be possible. Similar concerns may be raised for smoking status. Subjects may have underreported smoking in the health questionnaire, so the true prevalence of current smoking would have been underestimated.
Despite the study's limitations, this analysis may have some implications. The advantage of using BAC for predicting disease status in a population-based project is the use of an already-established screening system covering large numbers of otherwise healthy women. Although the association between BAC and CHD shown here is not strong enough to recommend using BAC as a sole screening method, BAC can play a supplementary role along with conventional risk factors to estimate cardiovascular risk, as suggested by two recent multivariate analyses [10, 16].
To overcome the limitations of this analysis, the prospective analysis of populations with sufficient cases for analyzing incidence for cardiovascular outcomes—possibly in a population with high prevalence of diabetes mellintus and including younger women— might be a next step to provide more solid evidence for using BAC in the context of CVD.
In conclusion, BAC on mammography was found in 16% of women in the current study population. Women with BAC tended to be older than those without BAC, but none of the other known cardiovascular risk factors, including blood pressure and diabetes, showed significant difference between the two groups after adjustment for age. BAC status was associated with prevalent CHD, with a more than twofold increase in risk.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  V. M. Miller and S. P. Duckles Vascular Actions of Estrogens: Functional Implications Pharmacol. Rev., June 1, 2008; 60(2): 210 - 241. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
