chatGPT(6AstraMaxPaid):
This is a useful observational study of familial longevity, but its claims about slower biological ageing and clinical screening exceed the evidence. Its most distinctive finding, a smaller rise in the Alzheimer’s-related biomarker pTau217, is tentative. Several claims of novelty also overlook earlier research.
The paper is Seidu et al., Parental longevity and polygenic longevity scores in relation to ageing-related factors in a population of 70-year-olds followed over six years, published in GeroScience in September 2026.
The study compares parental lifespan and genetic longevity scores with health at age 70 and subsequent changes.
The researchers examined 1,126 people from the population-based Gothenburg H70 cohort in Sweden. Participants were divided into:
- High parental longevity: both parents survived to 85; 194 participants.
- Medium parental longevity: one parent survived to 85; 488 participants.
- Low parental longevity: neither parent survived to 85; 444 participants.
They also calculated two polygenic longevity scores from previously published genetic associations: one including the APOE region and one excluding it. These scores capture part of inherited susceptibility to longevity.
Participants underwent extensive clinical, cognitive, social and biochemical assessments. Approximately six years later, 800 were re-examined.
People with two long-lived parents generally had more education, better socioeconomic circumstances, less smoking, less hypertension, lower BMI and homocysteine, and higher cognitive scores. They also had higher total, HDL and LDL cholesterol.
Selected results from Tables 3 and 5 illustrate both the findings and their uncertainty:
| Outcome | High versus low parental longevity | Reported corrected p-value |
|---|---|---|
| Cognitive performance, measured by MMSE | 0.329 standard deviations higher | 0.0003 |
| BMI | 0.283 standard deviations lower | 0.002 |
| Homocysteine | 0.252 standard deviations lower | 0.007 |
| Hypertension | Odds ratio 0.549 | 0.001 |
| Current smoking | Odds ratio 0.391 | 0.009 |
| LDL cholesterol | 0.200 standard deviations higher | 0.039 |
| CRP | 0.195 standard deviations lower | 0.051 |
| IL-6 | 0.190 standard deviations lower | 0.061 |
| Change in pTau217 over follow-up | Smaller increase; standardised coefficient -0.226 | 0.065 |
The odds ratios describe differences in conditions or behaviours present at baseline. They do not measure reductions in future disease risk.
The inflammation and pTau217 findings did not meet the conventional 0.05 threshold after the authors’ multiple-testing correction. This does not establish that the associations are absent, but it makes them preliminary.
Higher genetic longevity scores were associated mainly with education, household income and lower odds of previous myocardial infarction. Table 4 also reports small positive associations with HDL cholesterol. The scores were not significantly associated with parental longevity in this sample, or with changes in the measured outcomes during follow-up.
There was no convincing evidence that participants with longer-lived parents experienced slower cognitive decline over the six years.
The main novelty is the combination of detailed measurements in one narrowly defined age group, particularly repeated pTau217 measurements.
The broad finding that offspring of long-lived parents have favourable health characteristics is well established. This paper adds a detailed comparison of family history and genetic scores in the same cohort.
Its specific novelty claims need qualification:
-
Parental longevity and pTau217 change: This appears to be the most distinctive contribution. However, it is a biomarker association requiring replication, with no demonstrated reduction in subsequent Alzheimer’s disease or dementia.
-
Familial longevity and lower inflammation: The general association is not new. Arai et al. reported a lower composite inflammation score in offspring of centenarians in 2015. That differs from this study’s exact parental-age definition and individual-marker analyses, but it limits the broader novelty claim. Seidu et al. actually cite that earlier paper. :chatgpt-content-reference{index=“0”}
-
Genetic longevity scores and myocardial infarction: The claim that this had not previously been reported is particularly difficult to defend. Don et al. examined 11 longevity scores in UK Biobank and reported associations with lower odds of heart attack in 2024. This paper is also cited by Seidu et al. :chatgpt-content-reference{index=“1”}
-
Familial longevity and homocysteine: The research question had already been investigated. The 2011 Leiden Longevity Study measured homocysteine in 1,907 participants and found no difference between offspring of long-lived families and their partners. Seidu et al. provide a positive finding in a different, older population, which should be discussed alongside that earlier null result. :chatgpt-content-reference{index=“2”}
The study’s strengths are its population-based recruitment, extensive measurements and comparison of family history with genetic scores.
The narrow age range reduces confounding by chronological age. Clinical examinations and laboratory measurements provide more reliable information than a questionnaire alone. The genetic discovery dataset did not include these participants, reducing concerns about testing a score in the same people used to develop it.
The authors also report confidence intervals, corrected p-values and several limitations. These allow a more cautious interpretation than the abstract and graphical summary suggest.
The principal weaknesses are:
-
Limited adjustment leaves major alternative explanations unresolved.
The main cross-sectional models adjusted for age and sex, with additional ancestry adjustment for genetic analyses. They did not adequately separate parental longevity from education, socioeconomic circumstances, smoking, adiposity, medication use and other relevant factors.
For example, participants with two long-lived parents averaged 14.2 years of education, compared with 12.2 years in the low-longevity group. Education could explain some of the cognitive difference. Smoking and adiposity could explain some of the inflammatory difference.
Some of these factors may mediate familial advantages rather than merely confound them. Distinguishing those possibilities requires an explicit causal framework and sequential models, which this paper does not provide.
-
The treatment of multiple testing is too permissive for the strength of the conclusions.
The investigators tested many outcomes, two parental-longevity contrasts, two genetic scores, longitudinal changes and sex interactions. Having plausible hypotheses does not remove the chance of false-positive findings across this collection.
Moreover, the adjusted p-values appear consistent with correction within individual outcomes across the two contrasts or scores, rather than across the full set of outcomes. That is an inference from the tables; the correction family is not clearly specified.
The inflammatory and pTau217 findings therefore deserve replication before being used to support broad claims about biological ageing.
-
The six-year follow-up provides little evidence of generally slower ageing.
Most positive findings are differences measured at age 70. A healthier starting point does not establish a slower subsequent rate of deterioration.
No measured longitudinal association survived the reported correction. Inflammation was not measured at follow-up, so the study cannot establish slower inflammatory ageing.
Attrition also matters: only about 71% of the original analytical sample returned, and fewer had the necessary pTau217 measurements. People who died or did not return may have differed systematically from those retained.
-
Table 5 contains apparent reporting errors.
In eight of the nine high-parental-longevity rows, the uncorrected p-value exactly duplicates the upper confidence limit.
For example, the alcohol-change estimate is -0.160, with a confidence interval from -0.356 to 0.037, yet the listed p-value is also 0.037. Under the usual corresponding two-sided regression test, a confidence interval spanning zero is inconsistent with that p-value.
This appears to be a copying or typesetting error. It does not demonstrate that the underlying analyses are wrong, but the table needs correction. The pTau217 row does not exhibit this particular duplication.
-
The genetic results cannot establish that parental longevity mainly reflects non-genetic influences.
A polygenic score captures only part of genetic variation relevant to longevity. A non-significant association in approximately 1,067 genotyped participants could reflect limited power, imperfect score construction or differences between the discovery population and this cohort.
Parental lifespan also combines inherited biology, shared environment, behaviour, healthcare and chance. Comparing the number of significant associations for parental lifespan and a genetic score does not quantify the relative contributions of genes and environment.
-
The mitochondrial inheritance argument is technically misplaced.
The discussion suggests that maternal lifespan should correlate with the genetic score because mitochondrial genes are maternally inherited.
However, the score was constructed from autosomal variants, not mitochondrial DNA. Nuclear genes affecting mitochondrial function are generally inherited from both parents. Maternal inheritance of mitochondrial DNA therefore does not justify the expected association with this particular score.
The study provides no direct test of mitochondrial inheritance or mitochondrial function.
-
The cholesterol interpretation is speculative.
Higher LDL cholesterol in people with long-lived parents does not establish that higher LDL promotes longevity. Medication use, underlying illness, nutritional status and selection of survivors can influence this association.
The authors’ finding of no interaction with lipid-lowering treatment does not establish the absence of confounding by treatment. An interaction test asks whether associations differ between groups; it does not resolve whether treatment explains part of an observed association.
-
Clinical screening value remains untested.
The paper proposes parental longevity as a screening measure for age-related disorders, but does not demonstrate predictive accuracy, calibration or added value beyond conventional risk factors.
It also does not analyse whether parental longevity predicts the participants’ mortality during follow-up, despite recording 69 deaths. Nor does it validate parental longevity against a direct measure of biological ageing.
A stronger follow-up would test incident disease, disability and mortality; compare prediction with and without parental lifespan; account for attrition; and replicate the pTau217 finding in an independent cohort.