Conflict of Interest: R. Boer has recently started working for Pfizer, Inc., which markets and develops products for treatment of early and late stage breast cancer, the success of which might, to some extent, depend on the results of this research.
Department of Public Health, Erasmus MC, 3000 CA Rotterdam, The Netherlands
Breast cancer screening programmes are effective in reducing mortality from breast cancer.1, 2 By detecting malignancies earlier, relatively smaller tumours are diagnosed and reductions in late stage disease are expected to be found. Consequently, at repeated screening, tumour stage distribution will be more favourable than at prevalent screening. In practice, stage distribution is not, or only slightly, more favourable.3–6 “False reassurance” is one of the explanations put forward for this lack of improvement at repeated screening relative to first screening rounds.6 False reassurance is defined as diagnostic delay due to having participated in screening. It occurs after a negative screen result, when a patient or doctor, perceiving the risk of developing cancer to be small, is consequently less alert to present symptoms or the need for further evaluation. Prolonged patient delays, in literature often defined as intervals of more than 12 weeks, have been associated with increased tumour sizes,7, 8 more positive lymph nodes7 and with decreased long-term survival.7, 9 False reassurance may therefore influence tumour stage and prognosis. Although various authors have suggested the possibility of false reassurance as a negative consequence of screening,10–12 as far as we know, no study has actually examined the phenomenon of delayed symptom presentation after mammography screening.
In our study, the impact of breast cancer screening on the time of presentation of breast cancer symptoms and the moment of breast assessment was examined. The length of the period between the first symptom(s), first medical consultation and breast assessment is compared between a group of breast cancer patients who were regularly screened and a group of patients who were not. Predictive factors that could underlie a prolonged delay are further investigated.
Study participants were recruited from the breast clinics at hospitals in Rotterdam (The Netherlands). Recruitment initially took place at the Erasmus MC–Daniel den Hoed Cancer Clinic, but was later extended to the breast clinics at 4 teaching hospitals in Rotterdam. In Rotterdam, all women with breast complaints suspicious for breast cancer are referred to a breast clinic. The main function of the breast clinics is to evaluate breast abnormalities and determine the course of treatment. During 20 months at the Erasmus MC-Daniel den Hoed Cancer Clinic, and during 13 months at the teaching hospitals, all consecutive women with breast abnormalities highly suspicious for breast malignancy who visited the breast clinic for the first time were invited to participate in the false reassurance study. Within this time frame, a study population large enough to assess the role of false reassurance in breast cancer screening with reasonable certainty was expected to be included. However, because false reassurance was not examined before, it was not known how many screened breast cancer patients would be symptomatic before diagnosis, and how long these women would delay to present symptoms. We were therefore unable to make well-founded assumptions about the sample size that would be needed to obtain sufficient power.
After their first visit to the breast clinic, patients were given written and oral information about the study. Patients who had cognitive impairments, who did not speak Dutch, or who had had breast cancer in the past were excluded from participation. Eligible patients, who gave their informed consent to participate in the study, were asked to complete a questionnaire at home, after the first visit to the breast clinic. Based on the answers to questionnaire items regarding participation in the nation-wide breast cancer screening programme in the past 5 years, women were either included in a screening group or a control group of women who had not been screened (Fig. 1). The “screening group” consisted of women who were referred to the breast clinic after a recent positive mammography and who had participated in at least 1 screening round prior to that (“screen detected group”), and women who discovered abnormalities in the interval between 2 screening rounds (“interval group”). Patients regularly undergoing mammography for surveillance purposes outside the screening programme were not included in the analysis. Because the main objective of our study was to assess the length of the period between noticing the first symptom(s) and seeking medical advice, only those patients who themselves detected breast abnormalities were further analysed. In the “screen detected group”, this meant that prior to the detection of an abnormality by mammography, the patient had already noticed (a) symptom (-s). The screen detected group and the interval group together composed the “symptomatic screened group”. The “control group” included women who either fell outside the age limits for a screening invitation (<50 or =75 years old) or had never attended screening. Women who were, at a subsequent visit to the breast clinic, not diagnosed with breast cancer were retrospectively excluded from the analysis. Only those patients whose symptoms were most likely related to the breast cancer diagnosis were further analysed.
The questionnaire focused on breast abnormalities and their discovery, seeking help, diagnosis and start of treatment. We asked participants whether they had observed 1 or more of the following breast abnormalities, changes or symptoms during the past 12 months: a lump in the breast, nipple discharge, scaling/eczema/retraction of the nipple, dimple in the breast/skin retraction, pain in the breast, other abnormalities or no symptoms. The date on which the (first) abnormality was detected, the subsequent visit to the general practitioner (GP) and referral and first visit to the breast clinic were recorded. The nature of present and past breast-related symptoms was enquired, as well as general health and preventive behaviour patterns and demographic characteristics. The second part of the questionnaire covered a number of psychological aspects involved, including the attitude towards screening and the reasons women decide, or decide not to consult a GP. Knowledge and beliefs about breast cancer and screening were also measured.
Women were further asked to participate in a telephone interview that was designed to validate the data mentioned in the questionnaire. The interview lasted half an hour and was carried out by a qualified researcher (EvA). The questionnaire data were verified on the basis of hospital records for women who allowed 30 days or more to elapse between detection of a breast abnormality and visiting the doctor, women with a relatively long period between the GP- and breast clinic consult, women with missing questionnaire data and women with screen-detected cancer.
Before the study was initiated, a small pilot survey among breast cancer patients at the Erasmus MC-Daniel den Hoed was conducted. The Medical Ethical Committee of the Erasmus Medical Centre approved of the study protocol (MEC 185.919/1999/193).
To compare demographic characteristics and the frequency of breast symptoms between the symptomatic screened and the control group, the chi-square and Mann–Whitney U test were used. The percentage of patients who delay to present symptoms for ≥30 days and ≥90 days was compared between both groups, using a chi-square test. Within the control group, differences in the time of presenting symptoms between women who were not screened because they fell outside the age limits for screening and women who did not attend screening were assessed, using a Student's t-test. With a Kaplan–Meier survival model and 2-sided log-rank test, the median time between the first symptom(s) and the first visit to the GP (“symptom–GP period”) was compared between the symptomatic screened and control patients. Because the end point in this analysis was the moment of seeking medical care, detection by screening (in the “screen detected group”) was considered a censoring event. Furthermore, the hazard ratio and confidence interval of the symptom–GP period between control and screened patients was calculated, and the influence of potential covariates on the symptom–GP period was estimated using a Cox regression model. The length of time between the first GP- and first breast clinic visit (“GP–breast clinic period”) was also calculated, using a Kaplan–Meier survival model and log rank testing. A p-value of 0.05 was considered to be significant. Additionally, a 2-sided non-parametric bootstrap procedure was used to calculate 95% confidence intervals around the median symptom–GP and GP–breast clinic period in the 2 groups. A total of 1,000 resamples were therefore randomly drawn with replacement from the original dataset, and the 2.5 and 97.5 percentiles were used to obtain a bootstrap confidence interval.
Results were analysed using SPSS statistical software, version 11.0 and S-Plus, version 6.
Of the 210 patients who were invited to participate in the study, 167 women (79.5%) gave their informed consent and completed the questionnaire. The most common reason cited for not participating in the study was “having too much on my mind”. Forty-seven patient record reviews and 113 patient interviews were additionally conducted. Twelve patients (7.2%) were retrospectively excluded because they had benign abnormalities (n = 10) or had had cancer in the past (n = 2). A total of 155 eligible breast cancer patients were thus further analysed.
Women who were screened for surveillance (n = 15) and women with screen-detected cancer who participated in the nation-wide screening programme for the first time (n = 14) were not included in the analysis. The total study population thus consisted of 126 patients; 42 of them not attending screening (“control group”) and 84 who did participate in screening (“screening group”) (Fig. 1). The women in the screening arm were significantly older (Mean age in screening group was 61.0 vs. 49.3 in the control group, Table I) and less well educated than the control patients. The 2 groups did not differ with regard to marital status and insurance (Table I).
Table I. Demographic Characteristics of Participants
Participated in screening program (“screening group”) (n = 84)
Symptomatic screened group (n = 32)
Not attending screening program (“control group”) (n = 42)
p-Value of difference between control group and screening group
p-Value of difference between control group and symptomatic screened group
n = Total number included.
Mann–Whitney U test.
Age in years (M, SD and range) atthe time of completing the questionnaire
Out of the 84 patients in the screening group, 53 were detected by screening (“screen detected group”) and 31 were diagnosed in the interval between 2 screening examinations, after their previous screening test (<5 years ago) was negative (“interval group”). Of those women with screen-detected cancer, 51 (96%) reported not to have noticed any breast cancer abnormalities prior to the positive mammogram. These patients were not further analysed. Two women with screen-detected cancer (3.8%) did have previous symptoms. Of the 31 patients who were diagnosed in the interval between 2 screening rounds, 1 was not further analysed because her symptoms were not self-detected, but initially discovered by a GP (Fig. 1). The remaining 30 patients and the 2 symptomatic patients with screen-detected cancer together composed the “symptomatic screened group”. The symptomatic screened women differed significantly from the control patients with regard to age (The mean age of symptomatic screened women was 59.0, Table I).
The median symptom–GP period in the symptomatic screened group, after censoring patients who were referred after a screening visit, was 7.0 days (Kaplan–Meier, 95% C.I. 0.0–15.3), whereas this period was 13.5 days (Kaplan–Meier, 95% C.I. 7.3–19.7) in the control group (Fig. 2, Table II). The log-rank test demonstrated that the 2 groups did not differ significantly regarding the length of time between discovery of the (first) breast abnormality and the first visit to a GP (p = 0.9). The hazard ratio of the symptom–GP period between control and screened patients was 1.03 (95% C.I. 0.62–1.71). The differences in age and education were not associated with the period between the appearance of the first symptoms and visiting a GP (Table III).
Table II. Symptoms and the Time Between Discovery and Visiting a Doctor
Symptomatic screened group (n = 32)
Control group (n = 42
n = Total number included.
Time in days between discovery of the (first) symptom and the first GP visit
Two women in the symptomatic screened group had relatively long symptom–GP periods of 890 and 1,453 days. They presented with retraction of the nipple and breast pain. These were the 2 patients with screen-detected cancer who indicated having noticed abnormalities prior to undergoing screening. No abnormalities were found during the previous screening round 2 years earlier. The estimated tumour stage at diagnosis of these patients was T1C (n unknown) and T1CN+ (a tumour size of 13 and 16 mm).
In the control group, 34 patients did not participate in screening because they fell outside the age limits for breast cancer screening; 8 women were eligible for screening but did not attend. The groups did not significantly differ with regard to the median symptom–GP period (Median symptom–GP period of women outside the age limits: 13.5 days, median symptom–GP period of non-attendees: 23.0 days, p = 0.96). Neither did the group of non-attendees differ significantly from the symptomatic screened group (p = 0.6).
Of the symptomatic screened patients, 31.2% (n = 10) had a symptom–GP period of 30 or more days, and 12.5% (n = 4) had a symptom–GP period of 90 or more days. Of the control patients, 31.0% (n = 13) allowed 30 or more days to elapse between first symptoms and visiting a GP, and 19.0% (n = 8) had a symptom–GP period of 90 or more days. No significant differences were found between symptomatic and control patients in the percentages of patients with symptom–GP periods of ≥30 and ≥90 days (chi-square test: p = 0.9 and p = 0.4). Tumour size differed borderline significantly between women with symptom–GP periods <30 or ≥30 days (p = 0.06). No statistically significant tumour size differences were found among patients waiting ≥90 days from the time of finding a first symptom to the moment of consulting a GP, as compared to women with a symptom–GP period of <90 days (p = 0.4) (Table IV).
Table IV. Symptom–GP Period and Tumour Size
≤ 20 mm (n, %)
>20 mm (n, %)
p = 0.06 (10 missing values; of 2 women the symptom–GP period was not known and of 8 women the tumour size was not available).
p = 0.74 (10 missing values; of 2 women the symptom–GP period was not known and of 8 women the tumour size was not available).
Period between observing(a) first symptom(s) and (first) GP consultation1
Period between observing (a) first symptom(s) and (first) GP consultation2
The length of the period between the first GP and first breast clinic visit did not differ significantly between both study arms (Fig. 3, log-rank: p = 0.9). In the symptomatic screened group, the median GP–breast clinic period was 7.0 days (Kaplan–Meier, 95% C.I. 5.9–8.1). In the control group, this was 6.0 days (Kaplan–Meier, 95% C.I. 4.0–8.0). The percentage of women who had a GP–breast clinic period of 10 days or more was 21.9% in the symptomatic screened group (n = 7) and 26.2% in the control group (n = 11) (Table II). The bootstrap 95% confidence intervals around the median symptom–GP and GP–breast clinic period were comparable with the intervals calculated in the Kaplan–Meier survival analysis.
The most commonly cited reason for patients not to immediately see a physician was: “my complaints will disappear spontaneously” (17.6%) (Table V). This resulted in a median symptom–GP period of 42.0 days. For 6 (18.8%) of the symptomatic women who had participated in screening, previous negative screening results were cited as a reason to postpone a visit to a GP, with a median resulting symptom–GP period of 34.5 days (range 15–48 days). Still, a majority of women (78.1%) indicated that a prior negative screen did not influence the decision to consult a GP for the current symptom(s) (data not shown). In addition, only 2 out of the 74 participants indicated that an upcoming screening invitation resulted in a delay of the GP-consult, by 15.0 and 23.0 days, respectively. The main reasons for ultimately deciding to see a GP were “worrying” and “wanting to know whether it was something serious, or not”. For 47.3 and 43.2 % of all patients, these motives played a role in the decision to consult a GP. Five women eventually presented their symptoms while consulting the physician for other problems. This resulted in the longest symptom–GP period, with a median of 42 days (Table V).
Table V. Reasons to Visit a GP and Symptom–GP Period
Frequency (n, %)
Median time between discovery of the symptom and first GP visit in days (Median, SD)
Percentages were based on the study group of 74 women with self-observed breast abnormalities.
Percentage was based on the 32 women in the symptomatic screened group.
When the 74 patients with self-reported breast abnormalities were asked how they would estimate the general influence of screening programmes on the moment of visiting a doctor and the awareness of breast abnormalities, opinions varied. Almost half (46.0%) of the participants disagreed with the statement that women with negative mammography results tend to postpone a visit to a GP after the discovery of a change in the breast. On the other hand, 35.1% of all respondents indicated that they considered the notion of false reassurance to be more or less plausible. One-third (33.8%) of the women believed that the screening programme contributed to a certain extent to decreasing the attention paid in general by women to possible changes in the breast. In contrast, almost 50% of the women (47.3%) disagreed emphatically and less emphatically with this statement. The majority (86.5%) of women agreed that breast self-examination was useful, and should continue to be performed, along with the screening programme. Screened and non-screened women did not differ in their perception of the influence of screening programmes.
Knowledge about breast cancer and screening was reasonable. For instance, most participants (86.5%) knew that only women aged 50–75 years are invited for screening, and that mammography can detect changes in the breast that are not yet observable (75.7%). A majority of women (87.8%) also knew that breast cancer could occur without symptoms or without feeling ill, and that lumps or changes in the breast do not necessarily indicate malignancy (75.7 and 78.4%). Half of the respondents with self-detected abnormalities knew that the lifetime risk of developing breast cancer is 1 out of 10. Fifty per cent of the patients (58.1%) were not familiar with the fact that approximately three-fourth of all Dutch breast cancer patients are older than 50, but 63.5% knew that breast cancer is the most common cancer among women in this age category in the Netherlands. Approximately 20% of the respondents assumed that the nation-wide screening programme takes place once every 5 years, perhaps confusing breast cancer screening with the cervical cancer-screening programme.
A majority of the women in our study indicated that previous or upcoming screening visits did not affect their decision to present current symptoms. The analysis showed for the first time that the period between initially observing abnormalities, the first GP visit and the first breast clinic consult was similar between screened and non-screened women. Additionally, only 2 out of 53 patients (3.8%) with screen-detected cancer reported having noticed a breast abnormality prior to the recent positive screening. It is possible that their long symptom–GP period of 890 and 1,453 days is attributable to false reassurance, but other causes cannot be ruled out. If their delay is indeed attributable to false reassurance, these results indicate that false reassurance plays, at the most, only a minor role in breast cancer screening, and probably does not affect the tumour stage distribution found at repeated screening rounds. Nevertheless, a relatively high percentage (18.8%) of women in our study population indicated that the last negative screening result constituted a consideration to postpone a visit to the GP. Their median symptom–GP period was 34.5 days (range, 15–48 days), indicating that false reassurance generally does not lead to long delays in presenting symptoms. However, 31% of all symptomatic patients waited more than 30 days to consult a doctor about their symptoms, and 12.5% of the symptomatic screened and 19.0% of the control patients had a period of more than 90 days between experiencing the first symptoms and visiting a GP.
A delay in presenting symptoms might partially be attributed to wrong assumptions about the risk of breast cancer and the benefits and disadvantages of screening. Our survey indicated that the knowledge about screening was generally realistic, probably because all Dutch women who are invited for screening receive detailed information about breast cancer and screening. Other studies, on the contrary, have shown that, in other countries, knowledge about screening may be limited. For example, a high number of women assume that mammography prevents or reduces the risk of contracting breast cancer.11, 13 Others were not familiar with the fact screening may have false negative results.11 Some screened women may therefore underestimate the risk of developing breast cancer or have fewer fears about cancer, which has been associated with delay.14–16 A lack of knowledge about breast cancer symptoms may also cause delay.11, 17 Most women are familiar with breast lumps as a potential sign of breast cancer, but knowledge about other symptoms may be limited.15 The long delay after non-lump symptoms, which was illustrated in our analysis and has been demonstrated in previous studies,8, 14, 16, 18, 19 could have resulted from this. Providing women with valid information about breast cancer and mammography screening, including the risk of false-negative screening outcomes and the possibility that symptoms develop in the period between 2 screening examinations, is needed to minimize the risk of delay and advanced disease.
Some factors could have influenced our results. Recall bias was a risk in our particular study design. In some cases, first symptoms were observed long ago, which made accurate timing difficult. Moreover, screened patients may well differ from non-screened individuals in (not) attributing certain symptoms to cancer. For instance, after a negative screening result, a patient may forget about a symptom that was present before screening. However, after a positive screening, symptoms may be reported that would not have led the woman to consult her GP had she not been screened. Our approach, in which we interviewed the women about their symptoms, next to examining patient hospital records to check for the presence of breast symptoms, has led us to feel relatively confident that we have succeeded in obtaining fairly reliable results. Moreover, to avoid having the women feel compelled to give “socially correct” answers, we downplayed the notion (in both the questionnaire and patient information) that a screening programme could possibly lead to delay. General terms were used when referring to symptoms and time periods. Another limitation of the analysis was the age and education difference: non-screened women were significantly younger and better educated than women who were screened. Although age7, 17, 20 and education14, 21 have been associated with delay, there was no evidence that the age and education differences affected our results.
The backwards design of our study also constituted some limitation. The alternative, a follow-up study with registration of the symptoms, would be a complex and expensive method to assess the role of false reassurance in breast cancer screening. Such an approach, however, might be recommended in future research.
The study population of 155 consecutive patients, 74 of whom reporting symptoms, was considered to be of sufficient magnitude to detect significant differences between the 2 study groups. The hazard ratio of the symptom–GP period between control and screened patients of 1.03 showed that, at each point in time, there is a 3% difference between the 2 groups in the likeliness that they will consult a doctor for their symptoms. Only a small proportion of the screened women in our study postpone seeking medical consultation for breast symptoms for a long period. However, the 95% confidence interval of the hazard ratio of is 0.62–1.71, indicating uncertainty in the analysis. Nevertheless, the fact that only 2 out of the 53 patients with screen-detected breast cancer had symptoms before the positive mammogram indicates that the role of false reassurance in breast cancer screening is small. The extent to which false reassurance plays a role in breast cancer screening remains to be studied in greater depth. Questionnaires that are directed at assessing the presence of symptoms before a screening visit and detailed interviews about psychological responses to screening should provide more insight into the role of false reassurance in delay.
Patients who participated in breast cancer screening did not have a significantly longer period between noticing (a) first breast abnormality (-s) and the first time presenting this to a GP and breast clinic than patients who did not undergo screening. Our data show that false reassurance plays, at most, only a minor role in breast cancer screening.
The authors are grateful to Mrs. E. Boot, Dr. J.H. Driebeek-van Dam, Dr. W.S. Meyer, Dr. J.C.J. Wereldsma and Dr. M.A. Paul for their help in collecting the data. They also thank Mr. C.W. Looman for statistical advice and Mrs. N. van der Zouwe for her advice in designing the psychological questionnaire.