The slippery slope is the most quoted claim in the assisted dying debate and one of the least measured. It appears in parliamentary committee rooms, in op-eds, in the leaflets pushed through letterboxes before a free vote: permit euthanasia for the dying, and within a generation you will be ending the lives of people with dementia and people with depression, in numbers nobody intended. Belgium, which legalised euthanasia in 2002, is the exhibit most often held up. It is also, conveniently, a country that has been counting. Every case must be reported to the Federal Commission for the Control and Evaluation of Euthanasia, and two researchers, Jacques Wels of University College London and the Universite libre de Bruxelles and Natasia Hamarat of the ULB, have now run the whole file.
Their study, posted as a preprint on medRxiv in January 2025 under the blunt title "Addressing the Slippery Slope Argument" and published in peer-reviewed form in PLOS Medicine on 19 November 2025, covers 33,592 reported cases from 2002 to 2023. Of those, 427 cases (1.27 percent) were justified by psychiatric disorders and 310 (0.92 percent) by dementia. In 2023, the single busiest year in the series, there were 3,423 reported cases in total, of which 48 involved psychiatric disorders and 41 involved dementia. As the authors put it, these conditions appear in the register roughly a hundred times less often than cancer and multimorbidity. This is a registry analysis of reported cases, not a trial and not a study of people who asked and were refused.
What makes it more than a tally is the denominator. Raw counts of anything rise in an ageing, growing population, which is why descriptive trend lines have been able to carry the slippery slope argument for twenty years without ever quite testing it. Wels and Hamarat modelled the series with zero-inflated negative binomial regression and then re-ran it with a population offset, using Belgian statistical office figures by age group, sex and region, so that the question becomes whether these cases grew faster than the population and faster than everything else in the register. For psychiatric disorders the answer was no: the trend ran level with other reasons for euthanasia (count ratio 1.00, 95 percent confidence interval 0.98 to 1.03; rate 1.02, 0.99 to 1.04). For dementia it was a qualified yes: a modest but real excess (count 1.03, 1.00 to 1.06; rate 1.04, 1.01 to 1.07). Roughly three to four percent a year more than the rest of the practice, compounding quietly over two decades.
That second number deserves to be read without flinching, because it is the part a campaigner on either side will want to lose. The honest summary is not that nothing changed. It is that the two categories behaved differently, and that neither behaved like a collapse of restraint. The same work found the gaps inside Belgium narrowing rather than widening: psychiatric cases were initially more common among women, with men catching up over time, and the Dutch-speaking population has the higher overall rate while the French-speaking population is increasing faster. The authors' preprint went further than the journal version in tone, concluding that the Belgian experience shows such laws can operate "without evidence of significant misuse". The published paper is cooler, and families reading the coverage should notice which version they are being shown.
For a household living with a dementia diagnosis, the practical meaning is narrower than the headlines on either side suggest. Belgian law turns on whether death is expected in the foreseeable future, taken as within twelve months, a threshold most people with dementia do not meet for years. The register also shows that dementia rarely stands alone in these records: among cases logged with a multimorbidity profile, 4.9 percent included dementia or memory disorders and 11.3 percent included psychiatric disorders, which is a reminder that the neat categories in a table are messier in a consulting room. A trend line cannot tell you whether any individual request was properly assessed, whether a person felt like a burden, or whether better care would have changed the question. It can only tell you that the numbers did not do what the argument said they would do.