Dying is a natural process. We want to recognise it in time.

Adults with cancer who are being cared for in a hospital or hospice taking part in the study may be invited, at any stage of their illness. We need people at every stage, including people who are well and expect to live for a long time, so being invited tells you nothing about how long you or your relative may have.

A small sample of urine, and sometimes blood, usually once a week while you are staying in hospital or a hospice. The research team also notes some information from your records, such as your medicines and how you are eating. No extra visits are needed.

No. Your care carries on exactly as it would have done. The samples are analysed later, for research, and are not used to make any decisions about your treatment.

Taking part is completely up to you. You can say no, or stop at any time, without giving a reason. It will not affect your care in any way.

When you agree to take part, it's your own decision. If later on you become too unwell to make decisions for yourself, the law says we can't rely on that earlier agreement on its own. Instead, we would ask someone close to you, usually a family member, for their advice on whether you would want to continue. They don't have to say yes, and samples are only taken if they are comfortable each time. We explain this at the start so that nothing comes as a surprise.

No. The samples are studied together, in large numbers, to find patterns. A single sample does not tell us anything reliable about one person, so individual results are not shared.

Urine is easy to collect, painless, and is often already being collected as part of care. It carries many of the small molecules that reflect what is happening throughout the body. That makes it a good choice for a future test that could be used gently, at the bedside.

When we first started this work back in 2014, our only inclusion criterion was a diagnosis of cancer. As METEL rolled out to other sites, colleagues kept asking us to be more specific. I'd always said I wanted solid tumours, because people with haematological cancers die differently. The process is just different.

So we went back to the ethics committee and asked to change "cancer" to "solid tumour". They pushed back: we'd said this was about cancer, so why change it now? On reflection, I think they were right. There is very little haematological disease in hospices, so most of what we collect will be from people with solid tumours anyway. If a hospital colleague wants to recruit someone with a haematological cancer who they think is in the last months of life, that's fine. We'll analyse those participants as a separate subgroup.

For the main scientific question, proving there is a process to dying in solid tumours, we will only include solid tumours, and we'll be upfront about that.

Please don't use an upper limit. We want to know whether the test works across the last three months of life, and to answer that we also need samples from people who are more than six months from death. Those participants act as the comparison group: they show us what is different about the last months of life. So we need people at every stage of their cancer journey. Several samples from the same person over time are a bonus, but they aren't the reason to recruit early.

Once a week for as long as someone is an inpatient is great. Twice weekly is welcome if the participant is happy, but it isn't necessary.

When we started in 2014 we knew almost nothing. It was a purely exploratory study, and I was particularly interested in the last days of life, so we collected as often as we could, sometimes up to three times a week. We now know enough to be confident that weekly sampling gives us what we need.

All of them. In our Communications Medicine paper we describe a model of seven metabolites that predicts the last 30 days of life, and it works at time points throughout those 30 days.

The paper also describes broader metabolic changes that point to different processes going on as people approach death:

• rapid muscle loss in the last three weeks of life
• bone loss
• falling food and caffeine intake
• increasing mitochondrial dysfunction, meaning the cell's energy supply is working less well
• possibly reduced DNA and protein synthesis

We now want to understand how all of these change over the last days, weeks and months of life.

This is exactly what we originally proposed when we wrote the study in 2012–13. The Research Ethics Committee told us it isn't permitted under the Mental Capacity Act 2005: advance consent to research can't carry forward once someone loses capacity. They advised the consultee process we use now, which follows the law. So the current approach stays.

A consultee's role only begins if a participant loses capacity. Until then, there's nothing for them to agree to. It's fine to explain the process to families at the start so it doesn't come as a surprise later.

Date of death is what matters most, because it tells us where each sample sat on that person's illness trajectory. Please keep a local log linking participant numbers to identifiable information. That log stays with you and is never shared with the study team. Check the records (the NHS Spine) three to six months after discharge to see whether the person has died, and enter the date of death in REDCap.

For cause of death, we'll assume they died from their cancer. If someone wasn't an inpatient when they died, we appreciate you won't know what their dying looked like, so record that as unknown.

There are no rigid criteria. I don't need surgical history or past episodes. I need conditions affecting the person's body right now, such as COPD, heart failure or diabetes. A pneumonia two years ago isn't relevant. You don't need to record start or end dates.

We ask about medication because we can detect many drugs and their metabolites in urine. So all I need is a record of what the person has had in the 24 hours before each sample: whether they're on morphine, aspirin and so on. Start dates aren't needed, although they are helpful for a recently started antibiotic.

For morphine, please give the total oral morphine equivalent over the previous 24 hours. For example:

That tells us roughly how much morphine is in someone's system when the sample is taken, so we can look for it in the urine. Other opioids, such as oxycodone and fentanyl, should be recorded as given. There's no need to convert them. For syringe drivers and subcutaneous doses, again, just record what was given in the previous 24 hours.

Alongside the samples, we collect a few pieces of clinical information. Each one helps us make sense of what we see in the urine.

• Phase of illness. Is the person stable, unstable, deteriorating or dying? This is well recognised in palliative care and gives us a clinical anchor for each sample.
• Food intake. Is it normal, reduced, minimal or none? These broad categories are deliberately simple, and we find them particularly useful. Falling food intake shows up clearly in the metabolites.
• Fatigue. We grade it in four groups: no fatigue; some fatigue, but still able to do normal activities; fatigue that limits the person to self-care only; and fatigue that stops them managing their own self-care. Grading it this way lets us ask questions about fatigue when we analyse the samples.
• How someone died, if you know. If someone dies in the hospice and you have a clinical suspicion of the cause, please tell us. Examples are a sudden death that might suggest a PE, death from infection, bowel obstruction, or a primary brain tumour driving the symptoms. These are unusual, but worth recording.

The pattern of dying. This is the one I'm most interested in. Everyone in palliative care knows that people don't all die the same way, yet it isn't described anywhere in the literature. I think there are three obvious groups:

If you know which group fits, please tell us. If someone doesn't fit any of them, record "not applicable". If you don't know, just say so; there's no need to guess.

We will report that we collected this information. We'll then use it in our analysis to see whether the test can tell these different patterns apart.