FAQs

The study lasts for three years

Yes, you can email the CLOCS-2 research team at clocs@imperial.ac.uk to express your desire to withdraw consent.

No, we will need access to the shopping data from your loyalty card only

No, but your contributions could inform early cancer detection methods.

When we recruit a participant, we can request consumer data that goes back up to 6 years before recruitment, so if they have been diagnosed recently with cancer we can see what they were buying before they were diagnosed. We analyse loyalty card purchase data to identify patterns in the purchase of over-the-counter products commonly used to manage non-specific symptoms (such as pain relief or indigestion remedies). By comparing purchasing behaviours before diagnosis, we can investigate whether increases or changes in these products may reflect self-care for symptoms prior to seeking medical advice. 

This comes down to the study design. CLOCS uses a retrospective case–control design, comparing individuals with a cancer diagnosis (cases) to those without (controls). They are carefully matched on age, and on how many people are in their household. If specific purchasing patterns are significantly more frequent in the case group than in matched controls, and after adjusting for known risk factors, this suggests the pattern is unlikely to be due to chance. Any findings that might turn out to be coincidental would be less likely to be statistically significant in the analysis. 

We use a combination of statistical approaches, including descriptive analyses, Fisher’s exact tests, and regression models (e.g. conditional logistic regression) to compare purchasing patterns between cases and controls. These methods allow us to estimate the strength and timing of associations, supported by confidence intervals and p-values. CLOCS‑2 will also explore predictive AI modelling approaches to assess the potential utility of these behaviours for early detection.

CLOCS‑1 and CLOCS‑2 are observational studies and do not involve returning individual results to participants or intervening in care. Participants are not contacted about potential health findings, as this is outside the scope of the current ethical approvals. Future studies would be needed to evaluate whether and how any alert-based system could be implemented safely in practice.

The study’s aim to identify when purchasing behaviours begin to differ between cases and controls prior to diagnosis. Previous findings (from CLOCS-1) suggest changes may occur several months (8-9 months) before diagnosis, but precisely how early this could be used for reliable prediction is still under investigation and requires further validation.

Public attitudes are mixed. Research conducted alongside CLOCS indicates that many people are comfortable sharing commercial data for health research when there are clear public benefits and strong safeguards. However, some individuals remain cautious, particularly around privacy and data use, which highlights the importance of transparency and robust governance.

All data are handled within secure, ISO27001-certified environments at Imperial College London (e.g. secure enclaves). Identifiable data are separated from research data and pseudonymised for analysis. Access is strictly controlled and limited to authorised researchers. Commercial partners do not have access to participants’ health data, and participants’ identities are not shared beyond what is necessary to securely obtain purchase records.

CLOCS‑1 provided proof-of-concept evidence that certain over-the-counter medication purchases (e.g. pain and indigestion treatments) increased in the months prior to ovarian cancer diagnosis, with statistically significant differences observed between cases and controls. However, these findings are observational and require validation in larger and broader studies such as what we are intending in CLOCS‑2. 

A key aim of CLOCS‑2 is to explore whether thresholds in purchasing behaviour could act as potential “alerts” for earlier investigation. Any real-world implementation (such as prompting individuals to seek medical advice) would require further prospective studies, validation, and careful consideration of ethics, accuracy, and clinical pathways before being introduced

CLOCS‑2 is expanding the approach beyond ovarian cancer to investigate ten different cancer types. This will help determine whether purchasing behaviour patterns associated with self-management of symptoms are consistent across multiple cancers, and whether the approach has broader applicability. Beyond cancer, it is likely that consumer data could be important in predicting early signs of other diseases such as cardiovascular disease or diabetes.