Pin the Lancet Commission's list on a clinic wall and it reads like a universal instruction: fourteen things to watch, forty-five per cent of the world's dementia burden theoretically in play. Carry that same poster from a memory clinic in Boston to one in rural Bihar or Guangxi, though, and the boxes stop matching the people in front of you. That is the plain finding of a comparative cross-sectional study led by Emma Nichols at the University of Southern California, published in The Lancet Healthy Longevity on 12 July 2026 and presented the same week at the Alzheimer's Association International Conference in London. The team pooled harmonised data on 214,251 respondents aged 50 and older from 11 nationally representative ageing studies covering 14 countries and regions: Ireland, the USA, England, Northern Ireland, four regions of continental Europe, South Korea, Mexico, China, Malaysia, Brazil and India, drawn from the most recent waves with refresher samples collected between 2009 and 2023.
The spread is not subtle. Low formal education, the single factor the Commission weights most heavily, was recorded in 85.6% of older adults in the Chinese cohort (95% CI 84.8 to 86.5) and in 12.0% of the American one (95% CI 11.3 to 12.7). High body mass index ran the other way, reaching 44.9% in the United States against 13.3% in India, according to the study and the accompanying release from USC. In other words, two countries can work from the same fourteen-item list and be looking at almost inverted problems. A programme designed around school attainment in one place is aimed at a tenth of the population in another; a weight-management service that makes sense in Ohio addresses a much smaller slice of older adults in Uttar Pradesh.
What makes the comparison worth trusting is the unglamorous part: harmonisation. The Gateway to Global Aging Data team recodes each partner survey into identically defined variables before pooling, so that a difference in prevalence is less likely to be an artefact of how a country worded its question. That process also sets the study's first honest boundary. Only 12 of the Commission's 14 factors could be included. Traumatic brain injury and air pollution were left out because the data were not available across the studies. Education, smoking, physical activity and alcohol consumption were self-reported. Anyone reading a national ranking here should read it as a ranking of what these harmonised instruments can see.
The surprise, the authors say, was not the divergence but the company these factors keep. Across settings, certain factors kept turning up in the same individuals: cardiovascular risks such as high LDL cholesterol and hypertension, and behavioural ones such as smoking and drinking. "I was less surprised by the differences and more surprised by some of the similarities," Nichols said in a statement. The practical reading is that a health system does not need twelve separate programmes. A diabetes service can be widened to catch the cardiometabolic cluster around it. The authors are explicit that dedicated cluster-informed trials would be needed to show whether treating a cluster actually improves outcomes.
Here is the sentence a family should hold onto, and it comes from the paper's own discussion rather than from any commentary about it: apart from low education, the prevalence rankings in this study differed from rankings based on population attributable fractions, because attributable fraction calculations also account for how strongly each factor raises risk, while this study measured prevalence alone. A very common risk factor with a modest relative risk and a rarer one with a steep relative risk can trade places depending on which arithmetic you run. Prevalence tells you how many people carry something. It does not tell you how much dementia would be averted by removing it.
And because the design is cross-sectional, snapshots taken at a single point in each cohort, it cannot say whether anything changed, or would change, if a country acted. It measures the distribution of risk, not the movement of incidence. For a reader whose parent has just been diagnosed, the study offers no new instruction; nothing here alters what any individual should do, which remains a conversation with their own clinician. For a health minister deciding where the first pound or rupee of a prevention budget goes, it offers something rarer: comparable numbers, gathered the same way, about their own population rather than someone else's.