Population-based mammographic service screening was introduced in Sweden in the late 1980s and early 1990s following recommendations from the National Board of Health and Welfare in 1986.1, 2 This was based on results from randomized trials, one of which was the Malmö Mammographic Screening Trial, MMST.3, 4, 5 This trial was followed by a service screening program, Malmö Mammographic Service Screening Program, MMSSP, which started in 1990.
To be efficient, a screening program requires high attendance rates. Attendance in Swedish population based service screening programs is known to vary by region, type of population, and by age.6, 7 In Malmö, women in areas with less favorable socio-economic circumstances seemed to be less willing to participate.8 Socio-economic factors have been shown to influence attendance in the mammography program in the county of Uppsala in Sweden.9 Furthermore, subsequent breast cancer in non-attenders is relatively advanced on average.4, 10, 11 Patterns of attendance and stage distribution among non-attenders might change over time. Our study assesses changes in attendance, staging and survival in breast cancer among non-attenders in MMSSP compared to non-attenders and the control group in MMST. Furthermore, non-attenders are compared to attenders in the MMSSP in terms of socio-economic factors, incidence of breast cancer and stage at diagnosis.
MATERIAL AND METHODS
MMST was a population based individually randomized trial conducted from 1976–90. Women (45–69 years of age) were randomly allocated either to invitation to screening or to a control group, comprising 21,088 and 21,195 women, respectively. The trial has been described in detail elsewhere.3, 4, 12
The trial terminated in 1990 and was followed by MMSSP. All women 50–69 years of age were invited to mammography every 18 or 24 months depending on parenchymal pattern (dense or non-dense). Also women from MMST who were <50 years were invited. A computerized screening register was established containing personal identification number, name, address, and screening results. This register is continuously updated from the population register of the city.
The invitation strategy in MMSSP differed somewhat from most screening programs. First a letter was sent out to all eligible women asking whether they would be interested in attending the mammographic screening program. Those answering “yes” were subsequently invited, but those answering “no” or did not answer after having received a reminder were regarded as not interested and were hence not invited. Those having expressed interest got an appointment by letter that could be changed. Eventually some women did not come for the examination despite having expressed an interest. In our study, the latter group together with women not interested and hence not invited were considered as non-attenders.
All women who had received a letter from the start in May 1990 until 31 December 1993 were identified.
Cancer, mortality and socio-economic information
Breast cancer cases in MMST had already been collected in a database. In MMSSP breast cancer cases were collected from the date of invitation until the end of follow-up, which was 31 December 1999. Cases were retrieved from the Swedish Cancer Register. A regional cancer register administered by the Southern Swedish Breast Cancer Group provided additional information such as stage at diagnosis based on the TNM-system.13 The date and cause of death among breast cancer patients was retrieved from the Swedish Cause of Death Register.
Data linkage to the 1990 Swedish Population Census provided information on marital status, ethnic background, education, occupation, employment status and size of household. Household income was added from the Income Register. Ethnical background was defined as: born in Sweden, yes/no. Marital status was divided into 4 groups: married/cohabiting, unmarried, divorced and widowed women. Level of education was categorized according to the Swedish school system into 3 categories with 9 years of comprehensive school, 2–3 years of college and university studies. Women were assigned to one of the groups according to their highest degree of education. Socio-economic groups were based on employment status and were divided into 5 groups: manual, non-manual, self-employed, other (could not be classified) and not currently employed. The last group includes retired and unemployed women. Crowded housing conditions was defined as more than 1 person/room in a household. Household income per person was the total household income per year in 1,000 SEK divided by the number of persons living in the household. Women not in the population census, those who died or moved or reached an age >70 years and those with breast cancer before invitation were excluded from the analysis (Fig. 1).
One of 2 private clinics that offered mammography in the city of Malmö provided information from their registers. We could identify women in the service screening register, both attenders and non-attenders that had undergone examination with mammography at the private clinic at least once during the study period 1990–93.
Attendance in MMSSP compared to MMST.
Attendance rates in 5-year-age groups were computed, including women with an earlier diagnosis of breast cancer, for women invited 1990–93 to MMSSP and for the first screening round for the study group of MMST. Differences in attendance rates between MMST and MMSSP were assessed by the χ2 test. All tests in our study were two-tailed and a p-value <0.05 was considered statistically significant.
Non-attenders in MMSSP compared to non-attenders and control group in MMST.
For the analysis of stage distribution and survival from breast cancer, breast cancer cases among non-attenders in MMSSP diagnosed after the invitation were followed from the date of diagnosis of breast cancer until date of death (all causes) or until end of follow up, 31 December 1999. Information from registers on breast cancer cases and causes of death was only available through 1999 at the time of the analyses. To assess a similar follow-up time for comparison, women in the control group of MMST and non-attenders in the first screening round of MMST were followed from date of randomization until date of death (all causes) or until 31 December 1985.
Breast cancer cases were classified, according to stage at diagnosis, as Stage 0–1 or Stage II+, the latter comprising Stage II, III and IV. Two cases (0.6%) with missing stage were excluded in the control group in MMST and 52 (6.1%) in MMSSP (30 among attenders and 22 among non-attenders). Cases with missing stage in MMSSP were not possible to trace in medical records because the identification numbers were removed after record linkage with the census. The proportion of Stage II+ breast cancers and 95% confidence intervals (CI) of all breast cancers among incident cases by age (<55 years and 55 years and older) was calculated in non-attenders in MMSSP, in the control group and among non-attenders in MMST. Differences in the proportions were tested with the χ.2
For the survival analysis all cause mortality was calculated in the 3 groups. Cox's proportional hazards analysis was used to estimate relative risks (RR) with 95% CI for death after breast cancer diagnosis in the different groups. The analysis was adjusted for age and stage at diagnosis.
Non-attenders vs. attenders in MMSSP with regard to socio-economic factors.
Univariate unconditional logistic regression analyses were carried out to analyze the odds of socio-economic variables, age and former participation in MMST in non-attender vs. attenders in MMSSP yielding odds ratios (OR) with 95 % CI. The statistical significance was tested with the likelihood ratio test. The analyses were repeated, adjusted for age at invitation. Finally all variables, i.e., age at invitation, participation in trial, ethnic background, marital status, level of education, socio-economic group, crowded housing conditions and household income per person, were tested in a multivariate logistic regression model.
Incidence of breast cancer and Stage II+ breast cancer in non-attenders vs. attenders in MMSSP.
The incidence of breast cancer and the incidence of Stage II+ cancer per 1,000 person-years were calculated among non-attenders and attenders. Women were followed from the date they received the inquiry, which could be between May 1990 and December 1993, until December 31, 1999, until she was diagnosed with breast cancer or died. Cox's proportional hazards analysis with a 95% CI was used to estimate RR for breast cancer diagnosis and Stage II+ breast cancer in the different groups. The analysis was also adjusted for age at invitation to the screening program.
To assess whether socio-economic factors were associated with large palpable tumors among cancer cases in first-time-invited non-attenders in MMSSP, tumors in this group were divided by size into quartiles. The factors tested were age at invitation, ethnic background, marital status, level of education, socio-economic group, crowded housing conditions and household income per person. Differences between the quartiles in terms of socio-economic factors were tested with the χ2 test and variations in age at invitation, age at diagnosis and household income per person were tested with Kruskal-Wallis test.
Screening outside the service screening program.
We calculated the proportions among attenders and non-attenders respectively in MMSSP, that had undergone mammography at one of the private clinics at least once during the study period 1990–93. Differences in proportions were tested with the χ2 test.
Attendance in MMSSP compared to MMST
In both trial and program, attendance was lowest in the oldest age groups. Overall, 74% of the women in the study group in MMST attended the first screening round, whereas attendance was significantly lower, 65%, among women invited to MMSSP (Table I).
Cases with former diagnosis of breast cancer included.
Non-attenders in MMSSP compared to non-attenders and control group in MMST
The proportions of Stage II+ breast cancers are shown in Table II. The proportion was 27 percentage points higher among non-attenders in MMST compared to the control group in MMST (p = 0.07) in women <55 years of age. In women 55 years of age or older it differed 7 percentage points (p = 0.32). Among non-attenders in MMSSP, the proportion was 9 percentage points higher compared to the control group of MMST in women <55 years of age (p = 0.46). There was no difference between the mentioned groups in women 55 years of age or older. Non-attenders in MMSSP compared to non-attenders in MMST had 18 percentage points lower proportion of advanced cancers in women <55 years of age (p = 0.31) and this difference was 6 percentage points in women 55 years of age and older (p = 0.49).
Table II. Proportion With Stage II+ Breast Cancers
Survival after breast cancer diagnosis adjusted for stage and age at diagnosis in the 3 groups is shown in Table III and Figure 2. The RR was 0.80 (0.49–1.30) for non-attenders in MMSSP compared to non-attenders in MMST, and the survival almost approached that of the control group in MMST. In MMST the RR was 1.37 (0.87–2.17) when comparing non-attenders and the control group.
Table III. All Cause Mortality Among Breast Cancer Patients
Non-attenders vs. attenders in MMSSP with regard to socio-economic factors
Women 55 years of age or more were slightly more likely to be non-attenders at invitation (Table IV). More than 50% of the attenders had been in the study group and were participants in MMST and had been examined with mammography at least once before compared to 39% of the non-attenders. Non-attenders in MMSSP were less likely to have been participants in MMST. Non-attenders were also more likely to have been born abroad, being unmarried, divorced or widowed and being not currently employed or having missing information on education. Non-attenders were more often living in cramped housing accommodation and had a lower income.
Table IV. Distribution of Socio-Economic Factors, Odds Ratios and Confidence Intervals for Non-Attenders and Attenders in MMSSP
Non-attenders n = 11,314
Attenders n = 21,418
Non-attenders vs. attenders OR (95% CI)
Age at invitation
Participation in trial
Born in Sweden
O-level college (≤9 years)
A-level college (≤12 years)
Not currently employed
Crowded housing conditions
Missing (cannot be calculated)
Household income/person (1000 SEK)
The odds for non-attendance by screening trial participation, ethnic background, marital status, socio-economic group, and cramped housing accommodation and household income/person remained fairly stable in the multivariate model (Table IV). Women 55 years of age or more at invitation were not significantly associated with non-attendance when adjusted for the other variables. Women with higher education (university/university college) had a higher risk of non-attendance in the multivariate model.
Incidence of breast cancer and Stage II+ breast cancer in non-attenders vs. attenders in MMSSP
Breast cancer cases (232) were detected among non-attenders (88,966 person-years) and 573 cases among attenders (176,324 person-years). The breast cancer incidence among non-attenders, 2.61/1,000 person-years, was, as expected, lower than for attenders, 3.25/1,000 person-years and the age adjusted RR was 0.80 (0.69–0.93). One hundred thirty two breast cancers were Stage II+ among non-attenders and 187 among attenders. Among non-attenders the incidence of Stage II+ cancers was 1.48/1,000 person-years, which was higher than for attenders, 1.06/1,000 person-years, corresponding to an age adjusted RR of 1.74 (1.34–2.17).
Tumor size among breast cancer cases in first time invited non-attenders in MMSSP
Women born abroad and married or cohabiting women tended to have smaller tumors. Women with lower education also seemed to have smaller tumors at diagnosis, but women with lower household income per person seemed to have larger tumors at diagnosis. The differences in socio-economic factors between the 4 size quartiles were, however, not statistically significant.
Screening outside the service screening program
In MMSSP 9.5% of non-attenders had undergone mammography at one of the private clinics at least once during the period 1990–93, compared to 4.5% of attenders (p < 0.01).
Our study found that attendance in mammographic screening has decreased over time whereas stage distribution and survival among non-attenders seem to have improved. Socio-economic factors that predict non-attendance have been identified and non-attenders have been shown to have a higher risk for advanced breast cancer compared to attenders.
The attendance rate is one of the factors deciding the outcome of a screening program. Different recruitment strategies seem to give diverse results, which has been illustrated in both trials and evaluations of service screening programs.14 Two different ways of invitation have been used in Malmö, one in MMST and another in the following service screening program, MMSSP. Attendance rates were significantly lower in MMSSP than in MMST. In MMST, women received information and simultaneously an appointment for examination. When MMSSP started, a letter was first sent out asking whether women were interested in being invited. Only a month or two later did those interested get an appointment. The rationale behind the letter was to give women an opportunity to seek information on screening to make an informed decision whether to participate or not. Over time, the information on the pros and cons of screening mammography has accumulated. At times the professional and media debate has been skeptical against screening, which might have influenced the attitudes of women in a more “negative” direction. The initial letter might give more time for reflection and maybe more women decided against screening or postponed the decision as a result. One cannot ignore the fact that having to take an additional initiative when responding to the invitation also could be a major barrier.
Our study indicates that the odds of having advanced stage breast cancer at diagnosis are lower among non-attenders in the service screening program today than among the non-attenders in the trial. Increased awareness of breast cancer in the population might influence women to seek help earlier nowadays. Also access to mammography has increased during the years. Cohort effects in terms of risk factors for breast cancer such as use of first generation oral contraceptives with high oestrogen levels, reproductive factors and other hormonal factors also might be other explanations that we could not account for in our study.
Accordingly, non-attenders in MMSSP seem to have a slightly better survival than non-attenders in MMST. In addition to a more favorable staging, improvements in therapy over time could also account for a better survival. Interestingly, both non-attenders in MMST and MMSSP seemed to have worse prognosis than the essentially unscreened population in the control group of MMST, which indicates that non-attendance is not a random phenomenon.
The differences in proportions of advanced breast cancer and in survival did not reach statistical significance. But the tendency is that the outcome seems to be more favorable for non-attenders in MMSSP than in MMST. With regard to the size of the material, the lack of statistical significance is probably a question of power.
What are the main reasons for not attending a screening program? The present findings confirm, on an individual level, the pattern seen on area level in one of our recent studies where non-attendance and socio-economic circumstances co-varied in an inverse fashion in the city of Malmö.8
Higher age at invitation has been shown to be a predictor for non-attendance in several studies.15, 16, 17 We found a weak association similar to another recent study,9 whereas others have found no association.18, 19
Women in our study population who had attended at least one screening round in MMST were more likely to proceed being an attender in MMSSP. Positive experience in the trial could be one reason. Ever having had a mammogram before seem to promote attendance20, 21, 22, 23, 24 that could be due to better information on the benefits of screening, or a history of breast problems and a positive experience of mammography and hence an interest in having regular check-ups. Having had a recent mammography examination outside the screening program or a diagnostic mammogram because of symptoms from the breasts shortly before an invitation to screening, however, can also be a reason for non-attendance.25, 26
Foreign-born women were more likely to be non-attenders. Similar results have been seen elsewhere in Sweden,9 Australia27 and in the U.S.24 This might reflect a language barrier or different attitudes due to cultural or religious factors.
Married or cohabiting women were less often non-attenders compared to divorced, unmarried and widowed women. This is in accordance with another Swedish study.9 Marital status could be considered as a proxy for social support. Similar results have been confirmed in several other studies18, 24, 26, 28, 29, 30 whereas another study did not find any link between marital status and non-attendance.23
Educational level did not seem to affect participation in our study, but in the multivariate analysis women with higher educational level had a tendency toward higher non-attendance rate compared to the lowest educational level. One reason could be that women with higher education use private mammography services to a greater extent. Various results have been seen in other studies; no association,19, 23, 24 somewhat lower attendance rate for women of the lowest and highest educational level9 and level of education inversely correlated with attendance.28 Employment status is maybe a better measure of socio-economic status; among socio economic groups based on employment status no difference was seen between manual and non-manual, but women not currently employed were more likely not to attend. The latter is in line with other studies.16, 17, 31, 32 In the group not currently employed unemployed women, women with disablement pension or old age pension are included. It was not possible to distinguish further between the subgroups. We believe, however, that employed/not employed status is a relevant classification for our purpose. Living in crowded housing conditions is a measure of socio-economic status and was a strong predictor for non-attendance in our study, even though this group was very small.
As another measure of the economic situation we chose household income/person. This is probably a more fair measure of the economic situation than total income for the single woman. The lowest quartile of household income/person and those with missing information on income were more prone to be non-attenders than the 3 higher quartiles. The out-of-pocket cost for a screening examination is low, about $13 U.S., but for women with a scarce economic situation a screening examination might not be given high priority. The cost has been shown to be a barrier in some studies,27, 33 and to be of no importance in another.6
Questionnaires and interviews are used frequently in studies concerning socio-economic circumstances and non-attendance at service screening with mammography. A major problem with these methods is that non-attenders in a screening program also do not tend to respond to questionnaires and interviews. Thus, the representativity of such statistics can often be questioned. In our study it was possible to obtain information on socio-economy on almost all women invited to the screening program because the study was based on data from the Swedish Population Census. Less than 1% of our study population did not take part of the census, which minimizes the risk for selection bias. The risk for misclassifications is small because the register data is of high quality. The fact that our study involved an urban population, with a demography different from rural areas, make the results applicable primarily to other urban populations.
Declining attendance rates threatens the outcome of a screening program. By identifying factors that predict non-attendance one can make efforts to promote breast cancer screening in certain groups. In Malmö, where non-attendance is known to vary between areas of residence, mobile mammography units and campaigns could be one solution. The attendance rates should also be monitored and further follow-up is hence needed.
Outside the public service screening program in Malmö there are other options in the city to get a screening mammogram. Two private clinics in the city offer mammography primarily to symptomatic women although a certain proportion of their patients are asymptomatic. The cost is the same as in the public service screening program. The 2 private clinics existed already during MMST. The lower attendance rate in Malmö compared to other screening programs in Sweden,7 maybe be explained in part by this fact. In our study, 9.5% of the non-attenders had undergone mammography at one of the private clinics at least once during the years 1990–93, compared to 4.5% of the attenders. It was not possible to distinguish between diagnostic and screening mammograms in the material from the private clinic. In Finland, Aro et al.25 have studied subgroup differences of the non-attenders, real non-attenders and others. Those who did not attend because of a mammogram taken elsewhere were not considered as real non-attenders and were often urban, well-to-do women with an interest in their own health.25 The same pattern was seen in a study from Italy.28 This is partly in line with the findings in our study where Swedish-born, married, non-manually working women with higher income were over-represented among the non-attenders having had an examination at one of the private clinics during the study period (data not shown). Taking this into account, the importance of socio-economic factors for non-attendance is probably underestimated in our study. Because we only had information from one of the two private clinics, though the largest, not all women having had a mammogram taken outside the screening program have been traced. It can anyhow illustrate the socio-economical characteristics for women using private mammography options. We have no reason to believe that women having had a mammogram at the other private clinic should differ, in terms of socio-economy.
The incidence of breast cancer during the period 1990–99 was significantly higher among attenders than among non-attenders. This is expected due to a lead time bias in the screening detected tumors. The detection rate was, however, lower than in the first round of the MMST and most other studies. This can be explained by the fact that the first screening round of MMSSP was not a pure prevalence round as a substantial proportion of those invited had participated in the previous trial.
We also found a significantly higher incidence of Stage 2+ cancers among the non-attenders compared to attenders in MMSSP, which has been seen in other studies as well.4, 10, 11 Because stage at diagnosis is one of the most important determinants of prognosis, the non-attenders may have an important impact on the overall efficiency of a screening program in terms of breast cancer mortality. Exposure for screening or not is apparently not the only reason for presenting with a large tumor. We could neither account for established risk factors for breast cancer such as parity and family history of breast cancer, nor could we control for use of oral contraceptives, hormone replacement therapy, obesity, alcohol consumption and smoking; factors that have been discussed as potential risk factors or modifiers.
Among first-time invited women who chose not to attend MMSSP, some certainly had palpable tumors already when they received the invitation. A tumor in the breast can be palpable at the size of about 10 mm. Women in the 3 higher size quartiles probably knew that they had a breast lump, but did not want to participate anyway, which could illustrate denial as a coping mechanism or patient's delay. In our material we could not find any statistically significant differences in socio-economic factors that accounted for differences in tumor size at diagnosis. Coping mechanisms and patient's delay are maybe also related to psychosocial situation, which we could not control for. Several studies have investigated determinant factors for diagnostic delay but many of them were of insufficient quality according to a review article.34 Hence there is need for further research on this issue.
In conclusion attendance in mammographic screening has decreased over time, which might be related to changes in invitation mode, to the fact that women use private options or simply due to lower enthusiasm for mammography. This remains to be further evaluated. The distribution of advanced carcinomas among non-attenders in service screening, however, seems to shift toward less advanced in comparison with non-attenders during the trial. This might indicate that awareness of breast cancer and the ability to seek advice earlier have increased over the years. Furthermore we could identify several socio-economic groups that were more likely to be non-attenders. The risk for advanced carcinoma at diagnosis was higher among non-attenders.
We are indebted to The South Swedish Breast Cancer Group, Lund University Hospital, for kindly providing information on TNM. We also thank Dr. U. Nilsson for his contribution to the data material. Financial support to our study was received from the Council for Medical Health Care Research in South Sweden.