2.1 Campaign overview

The third national BCW70 campaign ran from 22 February 2018 to 31 March 2018 in England. A communication framework, based on a modified version of the Government Communications Standards, was designed by PHE Marketing and Wavemaker, an external media planning agency (Wavemaker, 2018). The framework consisted of three stages: Alert, Motivate and Support (see Figure 1). Further details on the campaign can be found on the BCoC website (Public Health England, 2020b).

2.2 Data sources and metrics

All data used in this study were anonymous and acquired from external sources. No primary data collection or participant recruitment was undertaken by the study team. Participants in the symptom awareness survey were sampled from a cohort of a market and social research agency (Kantar Ltd), and all had given prior consent to be approached for such surveys and to have their responses shared with third parties. All remaining data were acquired from publicly owned, anonymised datasets. As such, ethical approval for this study was not sought.

The main data sources used for the metrics were The Health Improvement Network (THIN), National Cancer Waiting Times Monitoring Data Set and Diagnostic Imaging Dataset. Data were collected for a series of metrics covering the patient pathway from symptom awareness, attending a GP practice with relevant symptoms, urgent GP referral through to diagnosis of cancer and stage at diagnosis. The analysis for each metric compared the period during/shortly after the campaign (analysis period) to the same period in the previous year (comparison period). The analysis and comparison periods are slightly different for each metric due to when we might expect an impact of the campaign to occur (National Cancer Registration and Analysis Service, n.d.). For example, a cancer diagnosis would not be expected until at least 2 weeks after the first presentation at the GP (see Table 1).

2.3 Statistical analysis

Analyses included women aged 70 years and over. The outcome metrics including weekly or monthly counts (except early-stage at diagnosis) were analysed using Poisson regression or negative binomial regression with one binary variable for the comparison and analysis period. Results are presented as total counts in each period, the estimated rate ratio (analysis period compared to the comparison period) with 95% confidence intervals and associated p-values. For GP attendances, the number of GP practices contributing data into THIN (The Health Improvement Network, 2021) each week was added as an offset to the model; therefore, the results are presented as count per practice. For symptom awareness/attitude to help-seeking and early-stage at diagnosis, counts were aggregated, and proportions calculated for the comparison and analysis periods. The percentage change between the two periods was calculated, and p values estimated using a two-sample test of proportions. Trends in counts and proportions over time were analysed using visual displays, with data covering 2017 to 2018. All statistical tests were two-sided, and p value < 0.05 was considered statistically significant. Analysis was conducted in Stata 16 (StataCorp, 2019).

3.1 Symptom awareness/attitude to help-seeking

There was a significant 6.0% (95% CI: 1.1% to 10.9%) increase in the proportion of respondents who identified that one in three BCs are diagnosed in women over 70 each year (p value = 0.019). There was a significant 9.3% (95% CI: −16.8% to −1.8%) decrease in the proportion of respondents who identified ‘pain in the breast’ as a warning sign of breast cancer (p value = 0.017).

There was no post-campaign increase in the proportion of respondents correctly identifying that a ‘lump or thickening in the breast or armpit’, ‘change(s) to the nipple(s)’, ‘change(s) to the skin of the breast(s)’, ‘change in the shape or size or feel of the breasts’ or ‘nipple discharge’ could be a symptom of BC (all p > 0.05). However, awareness of these symptoms was high pre-campaign, ranging from 65.5% for ‘change(s) to the skin of the breast(s)’ to 84.5% for ‘lump or thickening in the breast or armpit’.

Before the campaign, most patients believed that they would be too embarrassed to talk to their GP about BC symptoms (87.4%), were worried about wasting the doctor's time (85.3%) or scared to know that they have BC (48.6%). These proportions were not significantly different after the campaign. The proportion of women who did not believe that their ‘GP/doctor would be difficult to talk to about the signs and symptoms of breast cancer’ increased significantly by 8.3% (95% CI 1.5% to 15.1%) from 21.9% before the campaign to 30.3% after the campaign (p = 0.019; see Table 3).

3.2 GP attendances

The rate of GP attendances for symptoms of BC was 19% higher during the post-campaign analysis period compared to the comparison period (RR: 1.19, 95% CI: 1.01–1.39, p value = 0.033), a significant increase (see Table 4). The rate during the analysis period was not statistically different to that in the comparison (rate ratio [RR]: 1.15, 95% CI: 0.96–1.38; p value = 0.113).

3.3 Urgent referrals for cancer and breast symptoms, and cancers diagnosed from an urgent referral

The rate of cancer referrals was 23% higher during the analysis period compared to the comparison period (RR: 1.23, 95% CI: 1.09–1.38; p value = 0.001). The rate of breast tumours (malignant and in situ) diagnosed from an urgent cancer referral was 8% higher during the analysis period compared to the comparison (RR: 1.08, 95% CI: 1.01–1.15; p value = 0.03).

There were no significant changes in the rate of referrals for symptoms (RR: 1.03, 95% CI: 0.86–1.25; p value = 0.730) or breast tumours diagnosed from a referral for breast symptoms (RR: 1.02, 95% CI: 0.84–1.24; p value = 0.842) between the two periods.

3.4 Cancer diagnoses in the Cancer Waiting Times (CWT) database

The diagnosis rates for all BCs, malignant and in situ BC in the CWT database were higher by 14%, 12% and 50%, respectively, in the analysis period compared to the comparison (all BC RR: 1.14, 95% CI: 1.03–1.26; p value = 0.015; malignant BC RR:1.12, 95% CI: 1.01–1.24; p value = 0.035; in situ RR: 1.50, 95% CI: 1.27–1.87; p value = <0.001).

3.5 Cancers diagnosed in the National Cancer Registration Dataset

There were 468 more cases of malignant BCs and 63 more cases BC in situ diagnosed in the analysis period which represent 11% and 21% increases respectively (malignant BC RR: 1.11, 95% CI: 1.02–1.22; p value = 0.017; in situ RR: 1.21, 95% CI: 1.04–1.41; p value = 0.015).

3.6 Early-stage at diagnosis

The proportion of early-stage (TNM stage 1–2) cancers diagnosed in analysis period (82.4%) was not statistically different from that in the comparison period (80.8%) (p = 0.065).

3.7 Diagnostics in secondary care

The rate of mammograms and ultrasounds conducted was 14% higher in the analysis period compared to the comparison period (RR: 1.14, 95% CI: 1.04–1.26; p value = 0.008).

3.8 Comparison to previous campaigns

Table S1 provides a comparison of symptom awareness and attitude to help-seeking metrics for all three national campaigns. There was a significant increase in the proportion of respondents who had seen, heard or read any adverts regarding cancer in the last couple of months for the first campaign only.

There was a positive impact on symptom awareness for the first campaign only. In contrast to the third campaign, the first and second campaign did not increase the accuracy of awareness of BC risk, in terms of the number of correct answers to the statements about breast cancer risk, for women over 70 specifically. The first and second campaigns had greater impact on respondents' attitudes towards help-seeking, with more respondents disagreeing with unfavourable attitudes. However, there was a decrease in respondents strongly disagreeing.

Table S2 provides a comparison of clinical metrics (all metrics apart from symptom awareness and attitudes to help-seeking) for all three national campaigns. The first campaign appeared to have the greatest impact on all metrics analysed, except for breast carcinoma in situ diagnosed; the RR was highest for the third campaign. The impacts of the second and third campaigns were attenuated but prevailed for most metrics.

The impact on urgent GP referrals for breast symptoms and cancers diagnosed from an urgent GP referral for breast symptoms was no longer statistically significant after the first two campaigns. There were no increases in the proportion of cancers diagnosed at an early stage for any campaign.

4.1 Clinical metrics

The third BCW70 campaign influenced urgent GP referrals for suspected BC, BC diagnosed from an urgent GP referral, BCs diagnosed and mammograms and ultrasounds performed. GP attendance for breast symptoms increased after the analysis period (during the post-campaign analysis period), suggesting a delayed, but unexpected, effect of the campaign on this metric. The main outcome of the campaign was an increase in urgent cancer referrals and diagnosed BCs.

There was an increase in urgent cancer referrals, suspected BCs diagnosed from such referrals and mammograms and ultrasounds performed across all three national campaigns, probably because more women were referred. However, increases observed after the first campaign diminished over subsequent campaigns. The reduced impact of the third campaign may be due to a lower media spend and no public relations activity, and diminished returns in the wake of the previous two campaigns. An increase in referrals (Bethune et al., 2013; Broggio & Francis, 2015; Kabir & Khoo, 2016; Lai et al., 2020) and cancers diagnosed from them (Broggio & Francis, 2015; Lai et al., 2020) have also been found for other BCoC campaigns. Increases in mammograms and ultrasounds were most likely due to increased referrals (and therefore attendance) to secondary care.

Unlike urgent cancer referrals, urgent referrals for breast symptoms were unaffected by the third campaign, despite increases after the first and second campaigns. This may be related to overall downward trends for these referral types, through changes in the referral guidelines from 2015 (National Institute for Health and Care Excellence, 2021); furthermore, the two referral types are related, with a decision having to be made between the suspected BC and breast symptoms routes. Thus, increases in the referral route for suspected BC may partly stem from cases which might previously have been referred as breast symptoms. It is therefore perhaps unsurprising that there was no effect on GP referrals for more general breast symptoms in the final campaign, despite an increase following the first and second campaigns.

As delays in presenting to the GP with symptoms are associated with lower survival (Richards et al., 1999), it is important to examine how campaign messaging should be tailored to improve GP attendance. A mixed-methods approach combining quantitative with qualitative methods would allow deeper exploration than survey data alone. Nonetheless, cancers diagnosed from urgent GP referrals for suspected BC increased over all three campaigns, though this diminished with successive campaigns. This suggests that the campaigns influenced GP referral patterns even though there was only a delayed increase in GP attendance rates after the campaign had ended. The increase in referrals indicates that GPs may have been receptive to the BCoC campaign. Thus, the campaign may have led to changes in GPs' behaviour, rather than more women coming forward for assessment. As much of the campaign took place in GP surgeries and was intended to alert to BC risk, this is understandable. Future research could explore this by assessing awareness of the campaign and its messages in GPs, before and after the campaign.

BC diagnoses increased across all three campaigns, in line with other BCoC campaigns (Lai et al., 2020). Breast carcinoma in situ diagnoses increased following the third campaign but not following the first or second campaigns. This may be because of the increase in mammography in the older age group. There was no evidence that the proportion of early-stage malignant BC diagnoses was affected by any of the three campaigns.

4.2 Symptom awareness and attitudes to help-seeking

Several changes were observed in relation to awareness of BC risk, symptoms and attitudes towards help-seeking from one's GP for BC symptoms. Regarding BC risk and age, following the third campaign, there was a significant increase in women who accurately estimated that one in three women over 70 were at risk of developing BC. Thus, the third campaign improved women's risk perception in this age group, especially when compared to the first and second campaigns, which did not influence risk perception.

In relation to recognising warning signs of BC, a significant decrease in recognition was observed for pain in the breast—not normally a symptom of BC, and rarely a cause for referral in the elderly—following the third campaign, but not for other symptoms. However, symptom awareness before the campaign was already high.

Attitudes towards help-seeking from the GP were unaffected by the third campaign, although as attitudes were positive to begin with for all campaigns, this finding is unsurprising. Given the lack of change in symptom awareness, the observed increase in GP attendances warrants further investigation. Concerns about talking to the GP about signs and symptoms of BC improved after the third campaign, and fewer women stated they had these concerns. Moreover, fewer women felt that they would not want to know, or be scared to know, if they had BC, indicating that there had been a slight increase in fear and/or denial of symptoms. Overall, attitudes towards help-seeking from GPs were still overwhelmingly positive, with 80%–90% of respondents not being worried about wasting their GP's time and not being embarrassed or afraid to see their GP for symptoms. However, denial, fear and negative attitudes towards GPs can act as barriers towards help-seeking for BC symptoms (Bish et al., 2005; O'Mahony et al., 2011). As GP attendances for breast symptoms only increased following the second campaign, further research is needed to determine alternative methods of encouraging women to see their GP, particularly for those exhibiting fear, denial and/or concerns over contacting their GPs.

4.3 Strengths and limitations

The BCW70 campaign represents a substantial effort to raise public awareness and effect behaviour change. The campaign was designed in consultation with GPs and a panel of expert representatives from public health, primary and secondary care, the charity sector and academic research (Public Health England, 2020b).

The campaign affected several clinical metrics, as well as some aspects of public awareness of BC in women over 70. However, several limitations need to be acknowledged. First, the evaluation used existing routinely-collected datasets not specifically designed to address the research aims. Only two time points were considered, so the results must be interpreted with this in mind. Future campaigns may benefit from expert clinical advisory groups during conception and planning, to identify outcomes of most value, how to measure these reliably and how to assess changes in them. Second, clinical metrics may have been affected by clinical trials occurring at the same time as the campaign. For example, the NHS-led AgeX trial (Moser et al., 2011), which extended the age group of women invited for BC screening to include women aged 47–49 and 71–73, may have influenced BC diagnoses in the period covered by the three campaigns, as some of these women were perhaps diagnosed through screening who previously would have presented symptomatically to the GP. Third, behavioural factors may also have affected the results. Symptom awareness does not necessarily lead to (timely) GP attendance (Bish et al., 2005; O'Mahony et al., 2011). For example, downplaying symptoms or prioritising life events over seeing a physician can lead to older women delaying symptom presentation to healthcare (Petrova et al., 2020). A BC diagnosis does not always result in treatment, as this may not always be clinically appropriate, or even wanted by the patient (Frenkel, 2013). A decision to treat should involve balancing of risk and benefit, considering relevant co-morbidities. Finally, no data were available on the duration of BC symptoms before seeing the GP, and whether this appointment was motivated by the campaign. If symptoms were present long before the campaign, there would have been no opportunity for an early diagnosis attributable to the campaign. However, even for women with BC, earlier treatment through being motivated by the campaign to seek help may have at least resulted in benefits such as improved quality of life—again, there are currently no data which would allow an investigation of this possibility.

4.4 Impact across all three campaigns

Examining the impact across all three campaigns, for some metrics—GP attendances, urgent GP referrals and suspected breast cancers diagnosed from these referrals—effects diminished with each successive campaign. This is corresponds to other BCoC campaigns (National Cancer Registration and & Analysis Service, 2018; National Cancer Registration and Analysis Service, 2020). Effects on other metrics—cancers diagnosed from urgent GP breast symptom referrals, and some of the cancer diagnoses in CWT database—diminished in the second campaign, only to increase again in the third campaign, although they did not return to post-first campaign levels. While the diminished effects of the third campaign may be partly attributable to the reduced budget and reduced reach, any effect the first campaign had on individuals would not be repeated for the same individual with each subsequent campaign. Those most receptive to the campaign messages are likely to have been influenced by the earlier campaigns, leaving fewer people for successive campaigns to influence, leading to diminishing returns due to saturation. More time between campaigns could have led to these reaching a new audience who had no breast symptoms when the previous campaign had run or had moved into the target age group between campaigns. Future work should therefore identify why some women were less receptive to the BCoC campaign messages than others.

As discussed above, denial, fear and negative attitudes towards GPs are potential factors for diminished effect (Bish et al., 2005; O'Mahony et al., 2011), but there may be others. Future campaigns should consider different methods to influence those who are more difficult to reach, and how to sustain the effects of the campaigns. Finally, while the positive findings are encouraging, these findings may have been influenced by factors that were not measured or assessed. Further research, using mixed methods to allow deeper exploration, is needed to determine causal relationships between campaigns such as the present one, and clinical and public awareness outcomes.