Part I: Our one-size-fits-all breast cancer screening programme continues to miss many cancers and keep women in the dark about breast-tissue density. Could AI usher in a more tailored approach that saves lives?
Two years ago, Jill Jackson walked away from her mammogram with a sense of relief when it came back showing no sign of breast cancer. But six months later, the then 44-year-old found a lump. She wasn’t particularly concerned, though, because her mammogram had been clear. What the Auckland-based administrator didn’t know was that she had dense breasts, making her mammogram more difficult to read – any early sign of that cancerous lump might have been there but it was probably missed.
Talking to the Listener from her partner’s house in Tauranga, she apologises for her “scrambled brain’', which she blames on the oestrogen blocker she has to take to reduce the risk of her cancer returning.
She’s alive, yes, but she’s one of about 440 women who were diagnosed last year with an “interval cancer” – cancers found after a mammogram, before their next screening is scheduled.
Breast density is recognised as a leading cause of breast cancer: extremely dense-breasted women (7-12% of the population, although New Zealand numbers have never been measured or calculated) have double the chance of getting breast cancer compared with an average woman.
Breast density not only increases the risk a woman will get breast cancer, but also makes cancer harder to detect on a mammogram – like looking for a snowflake in a snowstorm. Though premenopausal women are more likely to have dense breasts, hormone replacement therapy and alcohol also increase density, according to recent studies. The US Mayo Clinic’s research suggests density can also be inherited.
The risk that goes unmentioned
Over the past decade, there have been calls to tell the 550,000 or so women aged 45-69 who get a free mammogram every two years about their breast density. BreastScreen Aotearoa (BSA) is consulting on whether breast density measuring and reporting should be included in our $65 million-a-year nationwide screening programme. After 25 years using the same software, a new ICT system is being rolled out early next year. An announcement on whether density will be screened will come in “due course’', says Dr Alana Ewe-Snow, prevention director at the National Public Health Service.
Internationally, there are moves to try to detect breast cancer much earlier and to offer more tailored screening rather than the current one-size-fits-all approach of our screening programme. There are calls for that here, too – the Breast Cancer Foundation thinks we’ll eventually move away from two-year screening to a system where women are screened according to risk with a broader suite of tools: abbreviated breast MRI and ultrasound for extremely dense-breasted women, for example, and 3D mammograms (tomosynthesis) for women with breasts that are heterogeneously dense (a mix of dense and fatty tissue).
Mammograms are increasingly thought to be imperfect: about four in 10 women probably need additional screening. A game changer could come from artificial intelligence (AI) – but there is caution. Overseas, AI is being used both to detect breast cancers and to assess the risk of developing breast cancer – one programme that will be trialled here if funded can spot a lesion up to five years before it shows up in the breast. AI pilots are happening in breast screening in Australia, and one to measure density has just been given ethical approval here.
Jackson had a lumpectomy last year. But two weeks later, when she was still nursing her wound, she noticed another breast lump that had not been detected on either her mammogram or the follow-up ultrasound. She was whisked in for another biopsy, which confirmed more extensive cancer, and rushed through for a double mastectomy – she wanted to get her good breast off too, so anxious was she that density in her cancer-free breast might mask further disease. Asked how it had been missed in her first screening, she says, “They told me that it wasn’t procedure to do an ultrasound at the same time as a biopsy and I also had just one ultrasound, which highlighted only the lump I could feel.’’
She finished chemotherapy last year and is getting used to her short, cropped hair with spikes of grey. “I was pretty angry at the time and wanted someone to blame. Only when I started looking at images of dense-breast mammograms did I realise how difficult it was to spot.
“I think the idea that women shouldn’t be told about density because it might scare them is not fair. I’d rather have known than not be aware of it.’’
In the United States, a Food and Drug Administration ruling that breast density should be both measured and reported by all breast clinics took effect last month, a move celebrated by clinicians. European Union countries report breast density and the EU suggests D-density (extremely dense) women get an MRI every two to four years, or an ultrasound as a second option.
In a significant – some say late – development, the Royal Australian and New Zealand College of Radiologists declared last December that density should be measured and reported in mammograms here and in Australia, so women – and their GPs – are informed and more vigilant about checking their breasts, understanding that an all-clear mammogram is not always accurate.
About 40% of women are in the heterogeneously dense (C) or extremely dense (D) categories and the college says they have up to twice the risk of getting breast cancer. The standard 2D mammogram is 57-71% effective for D breasts compared with 81-93% for fatty breasts. In laywoman’s terms, that means at least a third of extremely dense-breasted women who have a standard 2D mammogram shouldn’t rely on a clear finding as a definite result.
But in our current screening programme, women with extremely dense breasts don’t even know this health information, which the Breast Cancer Foundation says is unethical. These women don’t know they should be more vigilant unless they go to a private clinic or a mammogram technician tells them after looking at the images.
In a submission to BSA’s density review obtained under the Official Information Act, the foundation urged women to be told about their breast density – those with heterogeneously dense and extremely dense breasts should have that information recorded. “The current [mammogram] letter says ‘no abnormality detected’ without any explanation that this may not actually mean no abnormalities or that the woman could be high risk,” the foundation wrote. “This approach erodes trust as it’s inaccurate, which is a problem given the importance of rescreening, and is a missed opportunity to educate women and their providers on what action is required.’’
Targeted screening
Until the review, Te Whatu Ora’s national screening unit maintained that BSA’s move to digital mammography had lessened the risk of a cancer being missed. There was not enough evidence that extra monitoring of women with dense breasts would reduce deaths, a spokesperson told the Listener in 2022, and the risks of harm of extra imaging via ultrasound or MRI might outweigh the benefits for women of average risk.
But Fay Sowerby, chair of the research-focused Breast Cancer Cure charity and secretary of the Breast Cancer Aotearoa Coalition, disagrees, saying we need to identify women who are most at risk and seek to detect breast cancer earlier using a broader range of screening tools. First off, women need to know their density and so do their GPs.
While writing her submission to the review, Sowerby asked women whether they wanted to be told about density, and they did. She likens BSA’s argument that density information could make women more anxious to “not being told a road is dangerous to cross so you don’t know you have to be cautious’'.
Breast Cancer Foundation head of research and strategic programmes Adele Gautier has just returned from two European conferences and says our mammogram programme is good enough – but after 25 years of operating there are ways to improve it. She is adamant the programme should measure and report density and that screening should become more personalised. This would mean stretched resources going to detecting cancer in women at higher risk. “There’s no doubt that everybody [globally] wants to do something for women with dense breasts and everyone is under similar pressures around resources and the cost of these technologies.’’
Under a personalised screening programme, some women might not get a mammogram every two years, with the timing pushed out because they’re considered low risk. Women with heterogeneously dense breasts might be better served with a 3D/tomosynthesis mammogram. For the very densely breasted, an abbreviated MRI every so often might be more effective – those women might not need a mammogram at all.
We’re at a revolutionary moment in the world of breast cancer screening because of AI.
Another screening tool, contrast-enhanced mammography, used by Dr Monica Saini at her private Breast Institute New Zealand clinic in Lower Hutt, is also effective, and cheaper than an MRI ($700 at Saini’s clinic).
“There is interest in making screening more personalised for everyone so resources go where they’re needed,” says Gautier. “Everyone’s pretty clear that this is where screening needs to go and you put your money where your risk is. But that might involve saying to women who have very fatty breasts that you don’t need screening every two years. Saying, ‘Actually, you only need to come every five years’ would be quite a mind shift but you’re going to have to do that if you’re going to put your money where the risk is.’’
The US gets radiographers to report density by reading mammograms, which is cheaper and which we could do here, she says.
The foundation is about to pilot a blood test screening tool that might help women who don’t get mammograms, younger women and densely-breasted women. Blood tests are taking off globally and showing promise. “A screening blood test would basically work the same way as a mammogram,” says Gautier. “It would say, there’s something there that needs further investigation.’’
Extra tools are needed, especially if the screening age drops to 40 for high-risk women. One issue is a shortage of radiologists and equipment, so it’s prudent to look for other screening tools.
At a Scottish conference she attended, Gautier also heard a lot about AI in breast screening detection and diagnosis. Scotland has been a leader in using AI in breast cancer. But, she says, AI is still being treated with caution and it needs to be tested on local populations. The UK reported that the infrastructure for AI in breast-risk assessment, screening and diagnosis was complex.
Clinics in Victoria are about to start a trial in which AI will be used as a second reader of mammograms (currently two radiologists read each). At a Radiology Across Borders conference the Listener attended online, Associate Professor Helen Frazer, clinical director of St Vincent’s BreastScreen in Melbourne, told participants earlier trials had shown AI picked up more cancers than a radiologist. But it also reported false positives, which is why it is going to be a trial. “We’re at a revolutionary moment in the world of breast cancer screening because of AI,” says Frazer. “One of the most exciting things is not just AI reading a mammogram, but also the ability of AI to predict an individual’s future risk of developing breast cancer, helping us to personalise the programme.’’
Professor John Shepherd, chief scientific officer at the University of Hawai’i Cancer Centre, told the Listener a key advantage of AI was in lessening the workload of radiologists without reducing accuracy. “For areas with shortages of radiologists, that’s an important finding.” The thinking, though, is that AI should never read a mammogram on its own – only as an adjunct.
If BreastScreen Aotearoa responds to the call to measure and record density, experts say any AI software used must be proven to be accurate on New Zealand women. Monica Saini says AI can fail: it shouldn’t be relied on as the main reader and definitely needs to be trialled here. The overseas software was not developed for women of Māori, Pacific and Asian ethnicities.
Dr Nicholas Knowlton, a senior research fellow at the University of Auckland’s department of obstetrics, gynaecology and reproductive sciences, has ethical approval to run up to a decade of mammograms through an AI programme to assess density and density changes. Funded through Breast Cancer Cure, his team expects to assess 4-5 million mammograms.
Knowlton says about 30 AI programmes globally measure breast density – the Mirai model his team will use is as effective at measuring density in mammograms as an experienced mammographer.
Another commercial AI programme, Volpara, measures density in mammograms in countries including Qatar and Iceland, and is also used at many private clinics here, including Mercy Breast Clinic in Auckland and Lower Hutt’s Breast Institute NZ. Volpara was established by Wellington-based density-screening advocate Ralph Highnam, who has since sold the company. Chief executive Teri Thomas has told BreastScreen Aotearoa it would cost about $250,000 to add its density-reading software to our mammography machines. Thomas says that’s the same cost as treating a woman with metastatic breast cancer. “This is a deep discount because we are headquartered in NZ and we want the best care for our communities.’’
[Personalised screening] might involve saying to women who have very fatty breasts that you don’t need screening every two years.
Volpara wants density to be reported for all women, whether or not its software is adopted. “Breast density reporting must happen no matter the method,” says Thomas. “We have to prioritise data collection and have personalised screening in New Zealand so we can start closing the gaps in breast-cancer care and outcomes.’’
After Knowlton’s team retrospectively analyse their sample mammograms, he hopes to get funding and approval to put them through AI software to assess breast cancer risk up to five years in advance. That’s the kind of future that excites advocates such as Sowerby, who likes the idea of an AI risk-assessment tool. “That would enable us to know in advance who is at greater risk and who needs screening and the frequency with which they may need to be screened and modality of that screening.’’
New Zealand is one of just 22 countries to fund population-based mammograms, and our take-up is relatively good: Gautier says close to 70% of women get a biennial mammogram here, compared with 55% in Australia. But breast cancer is still the leading cause of cancer deaths in women. And critics of our programme talk about the haves and have-nots. The technology is increasingly being described as imperfect, especially for the densely breasted, other high-risk women and those with a family history of breast cancer.
Women with medical insurance and those who can afford to fund more specific screening tests in a private clinic have an advantage. In the state-funded clinics, 3D mammography – 100-odd images of each breast – is used only as a second check if something suspicious is found.
In its submission to the review, the Breast Cancer Foundation says density notification would be more equitable, ensuring that “all women, regardless of whether they have private funds or insurance, are being informed of the same thing. This means all women can understand that their mammogram may have been less likely to detect cancer and encourage vigilance to act on suspicious changes.’’
Cancers missed
Mercy Breast Clinic radiologist Dr Sugania Reddy points to the college of radiologists’ recent recommendation for mandatory density reporting and says, “Breast density is so important because it’s common – almost 40% of our patients have high-risk density.’’
Mercy clinic ran a pilot study of 2400 women who came for screening from May 2022 to June 2023. Women with dense breasts and high risk were offered supplementary screening with breast tomosynthesis, ultrasound and MRI. In preliminary findings, an additional 18 breast cancers missed on mammograms were detected – found only through ultrasound or an MRI. Supplementary MRI also found more cancers, known as multifocal disease, in women in whom one cancer was found either on a mammogram or ultrasound (such as Jackson’s scenario). Of the 18 women with breast cancer, 11 had high breast density – they were Cs or Ds. Half the patients had a significant family history.
Since then, screening at Mercy has become more personalised. High-risk women – with density, family history or both – get additional screening, while breast tomosynthesis (3D) is now the standard for all women at the clinic.
In an ideal world, dense-breasted women with increased risk would get a screening contrast like MRI, but medical insurers currently don’t fund this.
Public awareness of breast density, and its risk, continues to be almost zero.
Reddy describes breast cancer as complex – different types of cancers present in different ways and their detection is not the same for different breast types, so screening should not be one-size-fits-all because cancers get missed. She pulls up an image of a heterogeneously dense breast (C): the mammogram showed no sign of cancer but an MRI showed 6cm of cancer. “It was reported as a normal mammogram. But when we did her MRI we found she had six cancers in the left breast.
“High breast density can reduce the sensitivity of a standard mammogram and this can result in a false negative, so those women really benefit from supplementary screening.’’
HRT can make breasts denser – one study just released in the US found HRT and alcohol both increase density. Online breast cancer risk assessment programmes also list HRT as a risk factor. Mercy Breast Clinic patients with high breast density who are on HRT are advised to get annual screening.
Reddy sees the future of AI as “exciting” and Mercy is about to run trials using AI to detect cancer on a mammogram and to support radiologists.
Another submission to the density review came from Nikki Slade Robinson, co-ordinator of Facebook group the Aotearoa NZ Breast Cancer Community. The Eastern Bay of Plenty-based breast cancer survivor dedicated it to a woman who died last year after her aggressive triple negative cancer spread to her brain, a story included among 50 others in the long submission of women whose cancer was missed in mammograms. Slade Robinson wrote: “Public awareness of breast density, and its risk, continues to be almost zero, so no one knows to ask for this information at their screening mammograms. We are united in our desire to see the screening programme updated to better protect women yet to walk our path. We have daughters, sisters, mothers, other female relatives and friends, as will you who make screening policy decisions.’’
Of the college of radiologists’ ruling that density should be measured and reported, she says: “That’s great, but we now need to see some action in health policy.’’
Ethnic differences
Māori wāhine have higher breast density and are also more likely to get breast cancer before they hit 45 – the publicly funded screening age – than their Pākehā/European counterparts, notes the Breast Cancer Coalition’s Fay Sowerby. But Māori women are less likely to get a mammogram: in August, 62% of Māori wāhine in the eligible screening age group were screened, compared to an overall 69.2%.
Dr Annette Lasham, a senior research fellow in the University of Auckland’s molecular medicine and pathology department, has completed a study: One Size Does Not Fit All. Part of the Breast Cancer Foundation-funded Helena McAlpine Young Women’s Breast Cancer Study, Lasham’s team are studying breast cancer in younger women, which is more aggressive and has a poorer prognosis. What they found was cancer rates among pre-screening-age women varied depending on ethnicity: Māori, Pacific and Asian women have higher cancer rates proportionately than Pākehā women before they reach 45.
Lasham is researching the cost of extending breast screening to Māori, Pacific and Asian women from age 40, and says: “Our mammograms are inequitable for Māori, Pacific and Asian women who are more likely to develop cancer before the screening age.’’
Lasham is an advocate of density reporting, but she does say something the Listener heard from others: if you give women that information, is it going to be difficult if BSA doesn’t offer extra free screening for densely and heterogeneously densely breasted women?
“I think that’s why there’s pushback in New Zealand, because to make it equitable, you’re going to have to offer something beyond a 2D mammogram for certain groups of women, such as Māori and Asian women.’’
But Lasham also argues that knowledge is power and it might help women to be more vigilant about their breast health.
The foundation’s Gautier wants to encourage women to keep getting mammograms, describing them as an important screening tool. Screening does save lives – a 2015 study found mammograms had reduced breast cancer deaths by 34% compared with women with breast cancer who were never screened.
But looking ahead to the changes expected in screening and diagnosis, Gautier says: “You wouldn’t want to be planning that in 15 years’ time everyone is still getting a two-yearly mammogram.’’