A single cohort produced three estimates of the link between a MIND diet score and mortality. In the first model, people with a high score had a lower estimated hazard of death than those with a low score. By the third model, the point estimate was close to 1. Reading the models side by side shows how much the result depends on which participant characteristics enter the comparison.
- The all-cause mortality analysis covered 15,427 adults from the Czech Republic and Poland. The cognition analysis covered 10,415.
- For high versus low diet scores, the all-cause mortality hazard ratio moved from 0.88 to 0.94 to 1.01 as the authors added variables.
- The verbal fluency association persisted in the most adjusted model. Cognitive testing gives a snapshot of performance in this analysis.
- Paper date
- 20 November 2026
- Mortality sample
- 15,427 adults in the Czech Republic and Poland
- Cognitive sample
- 10,415 adults
- Recorded deaths
- 3,792 over approximately 20 years of follow-up
Original publication: 20 Nov 2026 · The date above refers to this brief.
The two HAPIEE analyses used different sample sizes
people
Three questions with different evidence
Were diet scores linked with later deaths?
Compare the same mortality outcome under the three models.
How did people score on cognitive tests?
Read the one-time testing results and the smaller denominator.
Did assigned diets change cognitive test scores?
Read the three-year randomized comparison in its own study population.
What a higher score captured
The MIND score combines food groups drawn from the Mediterranean and DASH dietary patterns. Participants reported what they had eaten during the previous three months using questionnaires at recruitment. The researchers grouped the resulting scores as low (below 6), moderate (6 to 7) and high (above 7). Age at recruitment ranged from 45 to 69. Diet groups already differed in characteristics including education, deprivation and smoking history. The researchers used that information to build successively more adjusted comparisons.
Source 1 ↗Follow the same mortality comparison across three models
For high versus low scores, the hazard ratio was 0.88 after accounting for age, sex, country cohort and energy intake. After adding education, marital status and deprivation, it moved to 0.94. After additionally including smoking, alcohol intake, recreational sports activity and body mass index, it reached 1.01. The fully adjusted 95% confidence interval was 0.93 to 1.10 and includes 1, the value for equal hazards. The estimates come from the same group of 15,427 adults. The 0.13 change between the first and third hazard ratios describes the model estimates and cannot assign the difference to any one variable.
Source 1 ↗The cognition result uses a different clock
The mortality comparison followed registry outcomes after enrolment. For cognition, the authors related a single set of tests to the dietary score in a cross-sectional analysis. The sample for that analysis was smaller by 5,012 people, about 32.5% of the mortality analysis count. The higher score remained associated with better verbal fluency in the fully adjusted analysis, by 0.058 standardised test units for high versus low scores. This finding compares test performance at a point in time. It leaves both the direction of the association and the rate of cognitive decline open.
Source 1 ↗What an earlier randomized study adds
Barnes and colleagues assigned 604 older adults to follow the MIND diet or a control diet, both with mild caloric restriction. After three years the difference in change in global cognition was 0.035 standardised units, with a 95% confidence interval from −0.022 to 0.092. That trial compares assigned diets in selected older people and measures change over time. HAPIEE describes an association with verbal fluency in a different population. The two designs help map what each result can answer and what remains uncertain.
Source 1 ↗Source 2 ↗A follow-up could track changes in diet and test performance
Repeated diet questionnaires alongside repeated cognitive tests would show whether changes in reported diet tend to precede changes in test scores. Specifying verbal fluency in advance would give researchers a clearer way to test whether that one finding recurs. Recording the reasons for missing tests would help readers understand who each estimate represents.
Source 1 ↗Read the adjustment and the timing before interpreting the diet score
Publisher full paper. Checked recruitment and analytic samples, the timing of the outcomes, Table 3 hazard ratios, the verbal fluency result and the study limitations.
01Was there a robust mortality association across models?
- What was observed
- The high-versus-low hazard ratio rose from 0.88 in Model 1 to 1.01 in Model 3.
- Where the conclusion stops
- The most adjusted interval spans equal hazards and is compatible with small differences in either direction.
Source 1 · Table 3, all-cause mortality
02Does the verbal fluency result show slower decline?
- What was observed
- The fully adjusted coefficient was +0.058 standardised test units.
- Where the conclusion stops
- This analysis uses a one-time assessment. It cannot measure change in verbal fluency over time.
Source 1 · Table 4, verbal fluency; Discussion
Numbers you can inspect
| Measure | Value & unit | Origin & method |
|---|---|---|
| Adults in mortality analysis | 15427 people | Reported Analytic cohort for Table 3. Source 1 · Methods and Table 3 |
| Adults in cognitive analysis | 10415 people | Reported Analytic sample used in Table 4. Source 1 · Table 4 |
| Difference between mortality and cognition counts | 5012 people | Calculated Subtract 10,415 from 15,427. Source 1 · Tables 3 and 4 |
| Difference as share of mortality sample count | 32.5 percent | Calculated Divide 5,012 by 15,427 and multiply by 100, rounded to one decimal. Source 1 · Tables 3 and 4 |
| Model 1 mortality, high versus low | 0.88 hazard ratio | Reported Age, sex, country cohort and energy intake. CI 0.81 to 0.96. Source 1 · Table 3, all-cause mortality |
| Model 2 mortality, high versus low | 0.94 hazard ratio | Reported Adds education, marital status and deprivation. CI 0.87 to 1.02. Source 1 · Table 3, all-cause mortality |
| Model 3 mortality, high versus low | 1.01 hazard ratio | Reported Also includes lifestyle variables and BMI. CI 0.93 to 1.10. Source 1 · Table 3, all-cause mortality |
| Change in point estimate from Model 1 to Model 3 | 0.13 hazard ratio units | Calculated Subtract 0.88 from 1.01. Describes two fitted estimates. Source 1 · Table 3, all-cause mortality |
| Verbal fluency, high versus low | 0.058 standardised test units | Reported Fully adjusted coefficient for a single cognitive assessment. Source 1 · Results and Table 4, verbal fluency |
| Randomized trial, three-year cognition change contrast | 0.035 standardised test units | Reported Difference in change under two assigned diets. CI −0.022 to 0.092. Source 2 · Barnes et al., Results, cognitive function |
Compare the actual experiments
These studies answer different questions. Read the unit and endpoint before comparing results.
| Study | Unit & setting | Readout | Interpretation boundary |
|---|---|---|---|
| HAPIEE mortality and cognitive analyses Source 1 · Methods; Tables 3 and 4 | Two analytic samples in the same observational study | Later registry deaths and one-time cognitive scores | Longer registry follow-up does not give the study repeated cognition measurements. |
| HAPIEE and the randomized MIND diet trial Source 2 · Barnes et al., Methods and Results | Czech and Polish adults and selected older trial participants | Verbal fluency level and global cognition change | These outcomes come from different designs and cannot share a numerical effect estimate. |
The export includes claims, available numbers, methods and source locations. It contains our reading notes and published summaries; it is not raw participant data or an independent reanalysis.
Evidence update · 22 Nov 2026
First publication. Any calculations in this brief use reported figures. The reading notes identify possible follow-up questions. AI source check; no human editorial or clinical review.
The same mortality outcome under three models
| Model | Hazard ratio for high vs low score | Variables included |
|---|---|---|
| Model 1 | 0.88 (95% CI 0.81 to 0.96) | Age, sex, country cohort and energy intake. |
| Model 2 | 0.94 (95% CI 0.87 to 1.02) | Adds education, marital status and deprivation. |
| Model 3 | 1.01 (95% CI 0.93 to 1.10) | Also adds smoking, alcohol, sports activity and body mass index. |
All-cause mortality from Table 3. Each row compares the same score categories in the same analytic sample.
Your questions, answered
Does a hazard ratio of 0.88 mean 12 fewer deaths per 100 people?
A hazard ratio compares estimated rates at a point during follow-up among people still at risk. It does not specify an absolute difference in the total number of deaths.
Why did the estimate move from 0.88 to 1.01?
The models held more measured characteristics constant. The shift shows sensitivity to those model choices. It does not identify which one factor produced the change.
Did the study measure how quickly memory declined?
The HAPIEE analysis linked the diet score to cognition measured at one point. It reported test performance, including verbal fluency, without estimating change over time.
How did the randomized trial differ?
The trial assigned two diets, supported both groups and followed changes in global cognitive performance over three years. It studied selected older adults.
Limits of this interpretation
- Diet was self-reported at recruitment, with no repeated measurements of diet in this analysis.
- The cognition analysis is cross-sectional and uses a smaller analytic sample.
- The adjusted estimates depend on measured factors and the authors’ model choices. The food patterns were observed.
Sources & transparency
- Wong, Yau, Kozela et al. (2026) — Association of the MIND diet with cognition and total and cardiovascular mortality in HAPIEE
Publisher full paper. Checked recruitment and analytic samples, the timing of the outcomes, Table 3 hazard ratios, the verbal fluency result and the study limitations. · Accessed 22 Nov 2026
DOI: 10.1038/s41430-026-01855-7 - Barnes et al. (2023) — Trial of the MIND Diet for Prevention of Cognitive Decline in Older Persons
Same research question approached through a randomized diet comparison. Checked group assignment, the three-year global cognition result and its interval. · Accessed 22 Nov 2026
DOI: 10.1056/NEJMoa2302368
Prepared and source-checked with AI for Press-news Team. The source notes identify the material available during preparation. The article received no human editorial or clinical review. The source list supports the reported results and the calculations shown.
Source check: AI source check — primary research and stated source limits
Clinical review: Not applicable to this educational guide
Suggest a correction