Prevalence, Long-term Development, and Predictors of Psychosocial Consequences of False-Positive Mammography among Women Attending Population-Based Screening

One side effect of breast cancer screening is false-positive mammography among healthy women; that is, findings on a screening mammogram leading to additional breast examination(s) but where the woman eventually is considered free from breast cancer (1). It has been estimated that for every 1,000 women in Europe participating in 10 biennial breast cancer screening rounds, 200 women will have a false-positive screening result (2).

Experiencing false-positive screening mammography yields short-term psychosocial consequences; that is, consequences that occur after receiving a recall letter for additional breast examination up to approximately 3 months following the diagnostic work-up (3–5). Long-term consequences although may occur and/or persist up to the subsequent screening mammography (4, 5). In previous studies, the long-term consequences of false-positive screening mammography have been investigated at various time intervals from approximately 1 month after false-positive screening mammography up to 3 years later (4). Thus, there is no consensus regarding the definition of long-term consequences in terms of either measurement interval or time point. Outcomes of long-term consequences also vary; some studies reported significant distress, intrusive thinking, perceived likelihood of breast cancer, or increased frequency of breast self-examination, whereas others found either no psychosocial consequences or reported mixed findings (4–6). Depression and general anxiety were investigated in some studies, but generally no evidence of such long-term effects was found (4). A meta-analysis found that long-term psychosocial consequences of false-positive mammographic screening are limited to breast cancer–related outcomes such as anxiety and distress about breast cancer (7). However, these outcomes have mainly been studied by means of single questions or questionnaires whose adequacy for the assessment of psychosocial consequences of false-positive mammographic screening has been questioned (5). Therefore, the Consequences of Screening-Breast Cancer questionnaire (COS-BC) was developed to target such consequences (8, 9). A recent COS-BC–based study confirmed adverse effects up to 3 years following false-positive screening mammography among women in Denmark (10). Mapping the prevalence and longitudinal development of these consequences in other countries using the COS-BC would add new insights regarding the impact of screening programs across countries.

Identifying sociodemographic and psychological predictors for experiencing such consequences is necessary to offer interventions to women at risk. For example, it has been found that a low level of education (11, 12), living in a high density urban area (12), dissatisfaction with client–provider communication (13), and distress at screening and diagnostic work-up (11–13) were potential predictors of long-term consequences. Having invasive breast examinations (fine needle/core biopsy) following an abnormal screening mammogram or being scheduled for early recall (subsequent mammography performed with a shorter interval than routine screening round) have also been identified as potentially influencing the psychosocial consequences of false-positive screening mammography (14). Consequently, it seems reasonable to provide resources to those who are in circumstances of a compromised ability to overcome psychosocial consequences of false-positive screening mammography. Sociodemographic predictors such as low socioeconomic status or living alone might indicate who may benefit from interventions, for instance, intensified information and personalized face-to-face communication, or counseling. However, previous observations warrant further investigation, particularly when the problem is addressed using the COS-BC. This study investigates the prevalence, longitudinal development, and predictors of psychosocial consequences of the false-positive mammographic screening among women attending a population-based breast cancer screening program in Sweden.

The study sample consisted of 848 Swedish-speaking women attending mammographic screening between September 2009 and December 2010 in 3 municipalities in southern Sweden, including the city of Malmö (Fig. 1). The facility has been providing population-based mammographic screening and diagnostic work-up for almost 30 years. In 2009 to 2010, 47,872 women (40–74 years of age) were screened at the facility and the average recall rate was 3.1%.

Batch-reading is practiced so that the women go home after their mammogram. The radiologic findings are categorized according to a classification scale with 5 levels: 1, negative/benign; 2, a discrete lesion having benign characteristics; 3, an abnormality present of indeterminate significance; 4, features suspicious of malignancy; and 5, malignant features (15). Approximately 2 weeks after the screening mammogram, each woman receives either written information that breast cancer was not found (negative screening mammography; scores 1–2) or a recall letter to have an additional examination within 1 week (scores 3–5). At the recall visit, a clinical mammogram is usually conducted along with ultrasound breast examination, which provides sufficient diagnostic work-up for most women considered free from breast cancer. In some cases, supplemental mammography projections are acquired, such as spot compression views or magnification views. Women are informed at the visit by the radiologist that additional examinations did not show breast cancer. Some women need further breast examinations, such as fine needle/core biopsy and follow-up by the surgical breast clinic. Women who eventually are considered free from breast cancer following these additional procedures are invited to subsequent screening mammography according to the standard practice. In some cases, early recall for subsequent mammography following the initial diagnostic assessment (i.e., a clinical mammogram, ultrasound breast examination, fine needle/core biopsy, and follow-up by the surgical breast clinic) is recommended. Early recall is thus a clinical mammography with a shorter interval, usually 4 to 6 months, contrary to routine screening of 18 to 24 months (15).

Group I consisted of 399 recalled women [mean (SD; range) age, 53 (9; 40–75)] enrolled consecutively immediately after receiving a negative diagnosis (free from breast cancer) following a recall examination. Some of these women also had supplementary diagnostic work-up such as fine/core needle biopsy and follow-up by the surgical breast clinic and/or were scheduled for early recall. Women who reported having felt a lump are routinely recalled and were therefore excluded.

Group II consisted of 449 women [mean (SD; range) age, 53 (9; 40–75)] with negative screening mammography. This group was matched with Group I women regarding age (±5 years) and timing of screening mammography. To compensate for anticipated nonparticipation, 2 women for each participating woman in Group I were invited.

The original questionnaire has been adapted to Swedish (Supplementary Table S1) and psychometrically tested among the same women with false-positive screening mammography as included in this study (16, 17). The Swedish COS-BC parts 1 (COS-BC 1) and 2 (COS-BC 2) consist of 6 scales each and 7 single items. The COS-BC 1 is intended to represent psychosocial consequences of an abnormal screening mammogram before the final diagnosis (8, 9). However, these consequences may persist and some may not occur until after the diagnosis (10). Therefore, the COS-BC 1 was used for mapping both short- and long-term consequences. The COS-BC 2 assesses psychosocial consequences of false-positive mammographic screening in the long-term, that is, consequences occurring after the final diagnosis (8). Responses (not at all, a bit, quite a bit, a lot; scored 0–3) to part 1 items are summed for each scale, where higher scores denote more negative consequences (8, 9). In part 2, responses (much less, less, the same as before, more, much more) are first recoded into 3 categories (“the same as before” is unchanged, “less” and “more” becomes “less/more,” and “much less” and “much more” becomes “much less/much more”) scored 0 to 2 and then summed for each scale; higher scores represent higher degree of change (regardless of direction) of psychosocial consequences (8, 10). Four COS-BC 1 scales (Sense of dejection, Anxiety, Behavioral, Sleep) and one COS-BC 2 scale (Existential values) have supported psychometric properties for cross-sectional and longitudinal group level assessments in a Swedish setting (16) and were therefore used here. Internal and external construct validity was supported by sufficient Rasch model fit of the data and empirical observations were in accordance with a priori validity–related hypotheses (16). Cronbach α was >0.86 for all 5 outcomes, and test–retest reliability (as measured by the intraclass correlation coefficient) was ≥0.75 for each outcome (16).

Variables considered potential predictors were based on previous studies (11–14) and collected for Group I women. Sociodemographic register data corresponding to 1 year before study entry were linked from Statistics Sweden. History of malignant cancer diagnoses and screening history and recent supplementary examination(s) were collected from the Swedish Cancer Registry and clinical records, respectively. Self-reported questions (Table 1) developed in the context of attendance patterns in Swedish mammographic screening (18) were also used. The Nottingham Health Profile (NHP) questionnaire (19) was applied to assess baseline health status among women in Group I.

Sociodemographic and cancer register data were also collected for Group II women to control for potential group differences at study entry. To perform a nonparticipation analysis, aggregated data from Statistics Sweden were used for Group I women who declined participation.

Women in Group I completed the COS-BC 1 (i.e., assessment of short-term consequences) shortly after they had been informed that no malignancy was found, referred to as baseline (Fig. 1). They were instructed to respond according to their experiences before the final diagnosis. They also responded to the questions addressing potential predictors and to the NHP. These women were followed-up by the COS-BC 1 and 2 regarding their experiences 6 and 12 months later (i.e., assessment of long-term consequences). Women scheduled for early recall responded to the follow-up questionnaire immediately after their appointments at approximately 6 and 12 months after baseline.

Women in Group II received the COS-BC questionnaires at the same time points as women in Group I and were instructed to respond according to their current experiences.

The study was approved by the Regional Ethical Review Board (no. 373/2008), Statistics Sweden (no. 218299/871210-1 and 220843/8721813), and the Swedish National Board of Health and Welfare (no. 59062/2012). All participants voluntarily signed the informed consent before entering the study.

Independent t tests or χ² tests were applied to test for group differences at study entry and potential differences between Group I women and those who declined participation.

Total scores of the COS-BC scales (Sense of dejection, Anxiety, Behavioral, Sleep, and Existential values) were dichotomized; women scoring >0 were considered experiencing psychosocial consequences. No imputation of the scores was applied. Missing item responses were randomly distributed and total scale scores could be computed for >95% of respondents across all data administrations. Between-group differences addressing the prevalence of psychosocial consequences were analyzed by χ² tests. Changes of the prevalence over time for Group I women were assessed by McNemar test. Logistic regression analysis was conducted to investigate the likelihood of each outcome at all time points depending on the diagnosis (i.e., false-positive vs. -negative screening), controlled for age.

Control for bias due to potential systematic drop-out in Group I was performed; that is, baseline outcomes for responders and nonresponders at the 6- and 12-month follow-ups were compared (χ² test).

To identify potential predictors of long-term psychosocial consequences of false-positive screening mammography, the following approach was applied: (i) a univariate logistic regression of each variable was conducted to predict each outcome at both follow-ups in Group I and (ii) variables with P < 0.25 were included in a multivariate forward stepwise likelihood ratio logistic regression to define a final prediction model of each outcome. The <0.25 P-value threshold was chosen to avoid that a variable alone might associate weakly with the outcome but become significant when analyzed together with other potentially important predictors (20). The multivariate regression models were controlled for age and reported baseline consequences.

Data were analyzed using IBM SPSS statistics, version 20. Two-tailed P ≤ 0.05 was considered significant.

There were no sociodemographic or cancer history differences between Group I and Group II women (Table 1). In Group I, 135 women (34%) attended their first screening mammography. Fine/core needle biopsy was conducted among 117 women (29%), of whom 45 needed follow-up by the surgical breast clinic. Eighty-seven women (22%) were scheduled for early recall. Women who declined participation did not differ from Group I participants by sociodemographic variables (P ≥ 0.097) except for 2; being foreign-born (P < 0.001) and being unemployed (P < 0.042), indicating an underrepresentation of these groups.

For all 5 COS-BC I and II scales and at all time points (baseline and 6 and 12 months after baseline), the reported prevalence of consequences was consistently higher (P < 0.001) among women who had experienced false-positive screening mammography (Group I) than among controls (Group II; Fig. 2). The prevalence of psychosocial consequences in Group I decreased (P < 0.001) according to all 5 scales between baseline and 6 months later but remained stable between 6 and 12 months (P ≥ 0.136). Experiencing false-positive screening mammography was associated with higher OR for psychosocial consequences (P < 0.001) than a negative diagnosis was for all outcomes and time points (Table 2). The ORs were more than 5 times higher at baseline (OR, 5.47–15.52) and more than twice as high (OR, 2.75–4.32) at 6 and 12 months later. The same pattern was observed when women having early recall were excluded (data available on request).

There were no differences (P ≥ 0.087) at baseline between responders and nonresponders in Group I to the 6- and 12-month follow-ups, except for behavioral consequences among responders and nonresponders to the 6-month follow-up [65% (n = 215) and 80% (n = 43), respectively; P = 0.038].

All studied predictor variables except age, marital status, and first time breast cancer screening attendance met the P < 0.25 criterion for inclusion in multivariate analysis. Previous cancer diagnosis and not trusting the health care system were not included due to low prevalence.

The final multivariate models for the 5 outcomes are reported in Tables 3 and 4. Early recall demonstrated the largest prediction for several outcomes at both 6 (OR, 6.25–10.31) and 12 (OR, 3.67) months. Susceptibility (OR, 1.50) and worry about breast cancer (OR, 1.28–1.47) were also predictors for consequences at 6 and 12 months. Dissatisfaction with information at recall (OR, 2.28–2.56) and lack of social support (OR, 1.13–1.25) were also identified as predictors for most outcomes. Among potential sociodemographic predictors, being foreign-born (OR, 2.35–3.71), was the most evident predictor at both follow-ups.

This 1-year follow up cohort study shows a higher prevalence of psychosocial consequences among women who had experienced false-positive screening mammography compared with those who had negative breast cancer screening. The odds for experiencing short- and long-term psychosocial consequences regarding sense of dejection, anxiety, impact on behavior, sleep, and existential values were at least twice as high for the recalled women compared with controls at all study assessments. We also identified several potential predictors of long-term psychosocial consequences after having been considered free from breast cancer. In particular, early recall profoundly influenced the prevalence of psychosocial consequences. Additional predictors included dissatisfaction with information at recall, lack of social support, and being foreign-born.

These results confirm previous evidence of short-term psychosocial consequences among women with false-positive screening mammography and add further knowledge about the prevalence and prediction of long-term effects. Importantly and in contrast to all but one study, assessments were conducted by means of a questionnaire specifically targeting psychosocial consequences of false-positive mammographic screening and developed among women with such experiences. It can therefore be argued that the outcomes are to be considered relevant from women's perspective but also for the providers of screening. Furthermore, the study provides new data in a setting of established population-based screening, which is of importance as the majority of previous studies were conducted in the 1990s at the initiation of screening.

Our results are in line with a Danish study, the only previous study using the COS-BC (10). For example, the study reported significant psychosocial effects at 6- and 18-month assessments following false-positive screening mammography. Although the consequences decreased over time, they were consistently higher than the experiences of controls according to both our own and the Danish results. This implies a need to prevent or minimize consequences of screening among women who do not necessarily benefit from the program.

In a comparable screening setting, a previous Swedish study found significant distress among women 6 months following false-positive screening mammography (12). The Psychological Consequences of Screening (PCQ) questionnaire (21) for measuring consequences of attendance at breast cancer screening was used. However, in another PCQ-based study, no consequences of a false-positive result were found at 8-month follow-up in an Australian screening program (22). It seems unlikely that the 2-month longer follow-up time would explain the difference, as our findings demonstrated generally unchanged prevalence of psychosocial consequences at 6 and 12 months. A British study, also applying the PCQ, found significant consequences at 5 months among women who underwent invasive diagnostic work-up or who were scheduled for early recall (14). The PCQ seems to provide somewhat inconsistent results and its adequacy of the assessment of long-term consequences has been questioned (5). In other words, the inconsistencies across the findings in Sweden and other countries could be due to the survey instruments, as well as by the reason of potential cultural differences in terms of health practices and values, beliefs, and attitudes toward breast cancer and mammographic screening.

Findings in other studies of breast cancer–related thoughts after false-positive screening mammography are somewhat difficult to compare with. For instance, it has been found that women experience mammography-related anxiety (23) and intrusive thinking (24) at 3 and up to 12 months after false-positive screening mammography, respectively. On the other hand, no effects on breast cancer worry and fear were found up to 6 months following a false-positive result (25, 26). Noteworthy, the assessment of these long-term consequences was conducted by the use of various single questions. Therefore, conflicting findings in these studies seem to strengthen the argument of the adequacy of using an assessment tool that is developed and validated in the target population.

Early recall demonstrated the strongest prediction for experiencing long-term psychosocial consequences. This differs somewhat from previous studies, which have found that although women scheduled for early recall experienced long-term distress, the prevalence of their experiences did generally not differ from those having fine/core needle biopsy (6, 13, 14). However, early recall seems to create confusion and doubt about the diagnosis (14, 27). This fact, together with our observations, supports the European guidelines to avoid early recall (target <1% of screened women; ref. 15, which corresponds to 0.18% in our setting). In addition, our findings that fine/core needle biopsy was unlikely to predict the outcomes argue against the impact of invasive diagnostic work-up on long-term consequences. This finding seems reasonable, as the gold standard in assessment of suspicious breast cancer is invasive procedures, combined with imaging and clinical examination (i.e., triple assessment; ref. 15), which probably provide confidence in the diagnosis for most women.

Dissatisfaction with information at recall also appeared to influence most study outcomes, which is in line with findings of another study (13). This highlights the significance of information in enabling women to cope with their situation. Interviews with women who had experienced false-positive screening mammography have found that recalled women look for adequate and individualized information (27). Personalized information and communication therefore appear both preferable and needed to prevent or lessen psychosocial consequences of false-positive mammographic screening. The importance of communicating information, also emphasized in the European guidelines for breast cancer screening (15), becomes particularly apparent in multicultural populations with potentially diverse needs and attitudes toward screening. Noteworthy is also the fact that women's cultural background influenced the prevalence of long-term psychosocial consequences of false-positive mammographic screening. This might further argue for the existence of different communication needs within this group but also warrant further research as such a prediction has not been found before (11).

Altogether, the findings presented here suggest that the occurrence of long-term psychosocial consequences of false-positive breast cancer screening should be acknowledged and reported to the consumers and providers of breast cancer screening. Delivery of screening implies a duty to introduce interventions to minimize adverse consequences and to provide support for women with a compromised ability to overcome them. Intervention provision might not only be of value from the women's perspective but might also contribute to the cost-effectiveness of screening. For example, reducing the number of those subjected to early recall not only has the potential to decrease the prevalence of psychosocial consequences of false-positive mammographic screening but also to reduce the costs of diagnostic follow-up, particularly as the positive predictive value of early recall is considered to be low (15). However, it is suggested that this should be investigated further. Furthermore, women invited to attend screening should receive clear and balanced information about the benefits and harm of screening to promote informed decision-making about attendance and to increase the awareness of the program, which respects the autonomy of each individual.

We observed a relatively high number (44%) of nonconsenting women among controls, which might be a potential limitation of the study. However, we anticipated lower participation among these women, as it had been observed that psychosocial consequences among women with a negative screening result are scarce (3). Importantly, we found in general no evidence of selection bias among women with false-positive screening mammography and observed no sociodemographic differences between both study groups, which we believe indicates sufficient representativeness of the study sample in relation to the target population. Furthermore, with one exception, there was no sign of systematic drop-out over time, which supports integrity of the results and conclusions of the study. However, women experiencing behavioral consequences at baseline were underrepresented in the follow-up assessments, which might indicate underestimation of these consequences in our results.

According to the NHP, the sample represented generally healthy women. It may therefore be assumed that at study entry women were in general equally influenced by their thoughts about breast cancer. However, we do not have prospective COS-BC data addressing women's experiences before the screening mammography or diagnostic work-up. On the other hand, assessment of psychosocial consequences of an abnormal screening mammogram, that is, before the final diagnosis, might have influenced women's well-being. Consequently, because of ethical considerations, we decided to collect retrospective baseline data, being aware of this potential limitation.

We dichotomized total scores of the COS-BC scales at the threshold ≥1. One argument for such an approach is that women who experience false-positive screening mammography are generally healthy people, and if they experience psychosocial consequences, they have been affected by screening regardless of the magnitude of their experiences. In addition, qualitative interviews with women scoring as low as 2 on the entire COS-BC, have revealed excruciating experiences following false-positive screening mammography (27).

Delivery of population-based screening service requires continuous monitoring of the program (28). It also raises the issue of responsibility for the impact of false-positive screening mammography on the psychosocial well-being of healthy women. Our results show that psychosocial consequences of false-positive mammograms are common and can persist over time. This argues for an intervention to prevent the development of such consequences. In particular, early recall demonstrated the potential to increase the prevalence of long-term psychosocial consequences. We therefore stress that this procedure should be applied cautiously, as it seems to create confusion and maintain psychosocial distress, whereas its potential advantage of providing reassurance can be questioned (27). The importance of finding cancers through early recall should of course be balanced against the drawbacks. However, evidence of the diagnostic yield of early recall is limited, and the general opinion is that the frequency of early recall should be kept at a minimum (15, 29).

We also emphasize the importance of providing information, in particular personalized face-to-face communication. Interviews with recalled women have shown that support from health care professionals induced comfort and security (27), and counseling has been associated with a lower risk of psychosocial distress (30). However, additional studies are needed to identify effective methods for facilitating communication and support for diverse populations participating in mammography screening.

No potential conflicts of interest were disclosed.

Conception and design: A. Bolejko, P. Hagell, C. Wann-Hansson, S. Zackrisson

Development of methodology: A. Bolejko, P. Hagell, C. Wann-Hansson, S. Zackrisson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Bolejko

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Bolejko, P. Hagell, S. Zackrisson

Writing, review, and/or revision of the manuscript: A. Bolejko, P. Hagell, C. Wann-Hansson, S. Zackrisson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Bolejko

Study supervision: P. Hagell, C. Wann-Hansson, S. Zackrisson

The authors thank the participating women and the staff where the study participants were enrolled for their cooperation. They also give their gratitude to Eva Prahl and Gunbrith Borgström for assistance with data collection.

The research was supported by Department of Medical Imaging and Physiology, Skåne University Hospital Malmö (to A. Bolejko), and by governmental funding of clinical research within the National Health Services, Lund University (ALF-grant; to S. Zackrisson).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.