For the first time, a trial randomised the decision itself

🎧 Listen: Weekly · Nov 9, from 01:27.

Every nephrologist has met this patient. She is 84. Her eGFR is 12 and falling. She uses a walker, takes eleven medicines, and spends half her day in a chair. The clinic's default is to start preparing her for dialysis: the modality talk, the vein mapping, the fistula. And nobody in the room knows whether that default serves her.

Until now there was no randomized evidence at all. A Cochrane review in 2025 looked for trials comparing conservative kidney management with dialysis in older adults and found none, only 23 non-randomized studies of very low certainty. Prepare for Kidney Care is the first.

What was randomized

This is the part that is easy to get wrong, and it matters for every conclusion that follows. The trial did not randomize people to dialysis or no dialysis. It randomized two preparation strategies, which is the actual decision a clinic faces years before the first needle.

One arm prepared for what the investigators call responsive management: a protocolised version of conservative kidney care, with a comprehensive assessment including advance care planning, usually done at home, then monthly phone calls and annual home visits, extra visits when needed, and palliative care. The other arm prepared for dialysis in the ordinary way: information, modality choice, access surgery and scans, dialysis if the disease progressed, then palliative care. Both arms were cared for by the patient's own team, and if someone in the conservative arm needed dialysis, they got it.

Participants had an eGFR below 15 confirmed on a second reading and not from AKI, plus either age 80 or over, or age 65 to 79 with frailty or multimorbidity. Frailty here was WHO performance status 3 or worse, meaning in a bed or chair at least half the waking day. Multimorbidity was a Davies comorbidity score of 2 or more. Anyone the clinical team judged unfit for dialysis was excluded, as was anyone expected to start within four weeks.

The run was long and the funnel was narrow. Thirty-one of the 67 adult kidney centers in the UK recruited from July 2017 to August 2024, with follow-up to August 2025. Of 6,120 people screened, 2,457 were eligible, 1,261 were approached, and 445 were randomized: 228 to responsive management, 217 to dialysis preparation. That is 7% of those screened, which tells you something about how hard this question is to study and something about who was willing to leave the choice to chance.

The two arms were well matched. Median age 82 in both, three quarters aged 80 or over, 31% women, 91% White, median Davies score 2, median eGFR 13.5 and 13.0. Importantly, about two thirds in each arm had no clear treatment plan before randomization.

The primary outcome, and why they chose it

Quality-adjusted life-years, from EQ-5D-5L collected every four months combined with survival. The trial's patient co-applicant argued that length of life and quality of life mattered equally, and the design followed that. The prespecified minimum clinically important difference was 0.345 QALYs, which is roughly six months of life at the utility value of UK haemodialysis. The working hypothesis, stated up front, was that responsive management would come out ahead.

What they found

It did not. QALYs were 1.354 (95% CI 1.201 to 1.508) with responsive management and 1.472 (1.288 to 1.655) with dialysis preparation. The adjusted difference was −0.147 QALYs (95% CI −0.350 to 0.056), p = 0.16. The authors translate that as 78 days at a utility of 0.69, well inside the 0.345 they had called clinically important, and the confidence interval crosses zero. The result held across prespecified subgroups and in per-protocol and sensitivity analyses.

Mortality was no different: 29.13 versus 26.44 deaths per 100 patient-years, adjusted HR 1.10 (0.87 to 1.40), p = 0.42. By the end, 69% and 63% of each arm had died. Place of death, hospital-free days (median 603 versus 640), symptoms on IPOS-Renal, capability on the ICECAP measures, treatment burden, timed up and go and grip strength were all similar.

Two numbers are worth carrying out of the secondary analyses. The repeated-measures EQ-5D-5L utility score favored dialysis preparation by a small margin that reached significance when censored at death (−0.039, 95% CI −0.074 to −0.005, p = 0.02), largest in the pain and discomfort domain; the authors judge it unlikely to be clinically important. And dialysis access complications, added post hoc during peer review, ran at 3.74 versus 11.33 per 100 patient-years, adjusted IRR 0.41 (0.19 to 0.88), p = 0.023. That is 19 events in 13 patients against 55 events in 23 patients: peritonitis, declots and fistula repairs, radiological procedures, access sepsis, bleeding needing admission.

The number that reframes the whole trial

Over four and a half years, only 28% of the dialysis-preparation arm ever started dialysis. In the responsive-management arm, 18% did. Meanwhile annual mortality ran at 25 to 30%.

Read that again, because it is the practical finding. For this population the risk of dying far outweighed the chance of ever needing the access you were building. Most of the preparation was for an event that never arrived. The authors say the result should challenge a standard of care in which preparing for dialysis, surgery included, is the default for frail or multimorbid older people. An economic evaluation is still to come.

How to hold it

This was a superiority trial that found no superiority. It was not designed or powered for equivalence or non-inferiority, so "the two are the same" is not what it says. What it says is that the hypothesis that conservative preparation gives frail older patients more good time was tested properly and was not supported, and that the reverse was not established either.

The other limits are real. About a quarter of each arm came off their allocated strategy, at a median of 13 and 11 months, which is what you would expect when preferences meet lived experience. Only 9% of participants were non-White, against 17 to 25% among age-matched UK incident dialysis patients. The sex finding, that men did worse with responsive management, is post hoc and underpowered; do not act on it. Dialysis preparation was deliberately not protocolised, so the comparator is UK usual care. And analyzing from randomization can hide differences at the high-risk transitions themselves.

One piece of context the authors supply is worth knowing when you read this from outside the UK. In Canada and Australia roughly half of people with kidney failure receive dialysis or a transplant; in the United States it is about 90%. A UK discrete choice experiment found that older people with kidney failure wanted an absolute survival benefit somewhere between 8% and 59% before they would accept dialysis. The default and the preference are not in the same place.

What changes on Monday is not that you stop offering dialysis. It is that for the frail 84-year-old, preparing for dialysis is now a choice you have to justify rather than a step you take automatically, and you can say honestly that the trial found no quality-adjusted advantage either way, and fewer access complications when you did not build the access.

The source: Caskey FJ, Worthington J, Liu Y, et al. Lancet. Published online November 4, 2026. Read the original

🎧 Listen: Weekly · Nov 9, from 01:27.

Physician-led, AI-assisted. Dr. Wael Hussein chooses every item NephSpace covers and reviews and edits every story before it is published. The first draft was written in our own words with AI (Claude, by Anthropic) from the full text of the paper and checked against our reading notes; Dr. Hussein then reviewed and edited it. For education and information only, not medical advice. Please talk to your own physician or care team about your health. Clinicians: read the original before changing practice. See our disclaimer and disclosures. Spotted an error? Tell us.