Skip to main content

Together we are beating cancer

Donate now
  • For Researchers

Can shopping data help diagnose cancer earlier?

The Cancer Research UK logo
by Cancer Research UK | In depth

28 January 2025

0 comments 0 comments

loyalty cards

Bringing big data to bear on cancer early detection can sometimes mean thinking outside the box… and indeed, say James Flanagan and Sima Toopchiani, the idea of shopping behaviour as an early signal of ovarian cancer is certainly that. So… could it work?

Have you ever had one of those ideas that made you think that’s so crazy, it just might work?

Well, in December 2015, I participated in a Cancer Research UK -led Innovation sandpit event aiming to encourage ideas on how we could leverage big data for earlier detection of cancer. Our team proposed the idea of using shopping data to see what ovarian cancer patients were buying to manage their symptoms before they went to the doctors or were diagnosed in the cancer clinic. It had never been done before.

It was compelling – and different – enough that the committee bought into the idea, and our pilot study was funded.

Starting the CLOCS

Fast forward to 2019, with further funding from a CRUK project grant, the team set about testing this idea in an appropriately sized case-control study.

Ovarian cancer patients were buying products like pain and indigestion medication up to 8 months before their diagnosis date.

This expanded study was called The Cancer Loyalty Card Study (CLOCS). Sadly, the Covid pandemic got in the way of our recruitment, but in the end, we still managed to recruit 153 ovarian cancer patients and 120 control participants, from whom we could request purchasing data from loyalty card programs at two well-known retailers.

This provided data that went back up to 6 years prior to enrolment in the study, so we were able to look at what women were buying from 2015 to 2021 prior to their ovarian cancer diagnosis.

The results were incredibly interesting. The main result showed us that ovarian cancer patients were buying products like pain and indigestion medication up to 8 months before their diagnosis date. Their first GP visits were, on average, three and a half months before their diagnosis.

What this means is that patients were potentially managing their symptoms with over-the-counter medications about four to five months before they first visited their GP. This also provides empirical evidence for when women start to experience symptoms of ovarian cancer.

Our research also included a large survey study that asked the public whether they would be willing to share their loyalty card data. It is important to understand the reasons why people choose, or choose not to, participate in research. This survey found that 52% of people were willing to share their shopping data for health research. Our job in the next few years is to convince more people of the value of this research so they will feel more comfortable contributing to it.

Loyalty cards

 Concerns and limitations

The most frequently asked question we receive from other scientists, grant panels, journal reviewers and the public is “but what if they are buying products for other people?”. And it’s true, this is a limitation of the study.

We cannot confirm from the data that just because someone bought a product that they consumed it. They might have bought it for someone else in the household or bought it and not consumed it. However, with careful study design we can take this into account by assuming that any such events would happen equally in control participants as much as with the cancer patients, so it evens out across the whole study cohort.

It is very important that everyone understands that we cannot do anything with data without the explicit consent of the people involved in the study.

Another concern is that people are worried about privacy and the security of their data. It is very important that everyone understands that we cannot do anything with data without the explicit consent of the people involved in the study. Participating in CLOCS is an opt-in study. We are not allowed to ask the retailers for people’s data without their signed consent forms.

What next?

For our new study, CLOCS-2, in addition to ovarian cancer, we are also focusing on other cancer types –  oesophageal, pancreatic and colorectal cancer, to name a few – aiming for about 250 patients with each of these cancer types. We will also aim to recruit about 1,000 healthy control participants to match for the analysis.

The selection of these cancer types was based on the fact that they may also have symptoms you might treat with over-the-counter medications. As such, that the hope is we might be able to improve earlier diagnosis for these types of cancers. For some of these, like pancreatic cancer, any opportunity for earlier diagnosis is likely to greatly impact patient outcomes.

Authors

Professor James Flanagan and Sima Toopchiani

James is a Professor of Cancer Informatics at Imperial College London’s Department of Surgery & Cancer

Sima is a PhD candidate at Imperial College London within the Ageing Epidemiology Research Unit – School of Public Health. She is also Project Manager for Cancer Loyalty Card Study 2

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.