iiinova — where your digital self keeps evolving

Jennifer Weuve @epidancer.bsky.social · 30/06/2025

Additional gratitude to The @tejenson.bsky.social for orchestrating the session, which included the best marketing prose on an epi/EH methods seminar invite in recent memory.

100

Jennifer Weuve @epidancer.bsky.social · 30/06/2025

What was esp helpful about Amanda's talk was getting a firmer sense of seeing how ITB plays out with exposures distributed on a continuum (an expansion on the examples many of us have seen that involve dichotomous exposures).

110

Jennifer Weuve @epidancer.bsky.social · 30/06/2025

In the true-story example she chronicled, effects estimated using a design that induced ITB were impressively large (one might call a HR of 51 gargantuan). Upon correction, they all shrank. & though some remained large (that HR of 51 became a HR of 5.5), some previously adverse findings became null.

110

Jennifer Weuve @epidancer.bsky.social · 30/06/2025

A big holler-y grateful shout-out to **Amanda Gassett**, Research Scientist at University of Washington, for presenting clearly and constructively on dodging immortal time bias in studies of air pollution's health effects. #EnviroBrain #CONFAB @busph.bsky.social

130

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Patterns of earnings trajectories (age 25 thru 60) differed by sex and a bit by birth cohort. Cognitive decline was faster among ppl w earnings trajectories that cycled and that were consistently low. Those w consistently high earnings had the slowest mean decline. @melodem.bsky.social

040

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

@ccolvin94.bsky.social presents meticulously thought-thru & generated results on life course earnings trajectories & cognitive decline, by birth cohort. The #MELODEM2025 equivalent of a walk-off HR will launch us into deep thoughts, conversations, and collaborations over the coming weeks and months.

130

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Proposed setting/Discovery cohort: WHICAP Offspring Study (many wmn had not entered menopause as of baseline)

010

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

More enlightening lightning (talks) at #MELODEM2025: Justina Avila-Rieger dares to wade thru the thicket that is reproductive status (and aging) in relation to dementia risk. Idea: exploit DNA methylation, which can reflect changes in the endocrine system and hormonal status. @melodem.bsky.social

120

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Exposure to fluoride was assoc'd w higher achievement in education (e.g., math). Not at all assoc'd w cogn at age 60.

010

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

So ... how about cogn later on in life? Setting: High School & Beyond Exposure: Based on location of pt's secondary school (>1000 high schools are represented here) and the fluoridation status of water there. Sufficient for none, part, or all of childhood.

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Fluoride. Great for oral health and stunning teeth. Does it exposure to it in childhood related to cogn later in life? @ #MELODEM2025, Rob Warren (U MN) points out the extraordinarily high exposures in studies of kids' IQs: not relevant to what's in our public water supply. @melodem.bsky.social

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Trend 1: cardiometabolic RF prevalence has been increasing Trend 2: dementia incidence and prevalence have been decreasing But, in the absence of trend 1, trend 2 (the decreases in dementia inc and prev) would have been steeper.

020

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Now up @ #MELODEM2025, @astokespop.bsky.social teases out the contribution of secular trends in midlife cardiometabolic health in relation to trends in dementia incidence and prevalence in the US. @melodem.bsky.social

120

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Cool innovation here: this is about the cog effects of incident trauma, not the effects of prevalent (historical) trauma.

010

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Our next lightning bolt @ #MELODEM2025 is from Katrina Kezios, presenting research on incident traumatic experiences in relation to cog change in older adults. Upshot: these exp were assoc'd w cog decline, esp among those w worse financial well-being. @melodem.bsky.social

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Iris Strangmann issues the first bolt of a lightning talk in today's #MELODEM2025. Does being multilingual offer cognitive benefits in older age? She is using data from LASI-DAD to answer this q, evaluating assns within strata of urbanicity and SES.

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Younger age, male (v fem), being White (not Latinx, Black), higher mid-life income, lower cog scores, more disability, and more comorbidities. (Others coming to a journal near you.)

010

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Carrying the #MELODEM2025 baton now is Melinda Power (GW). Among ppl who have #dementia, a large % have not been diagnosed. What predicts who gets diagnosed in a timely fashion? @melodem.bsky.social

120

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

A shout-out to yesterday afternoon's and todays moderators--Suzanne Judd, @ermayeda.bsky.social, @relajoie.bsky.social, & @miguelarce.bsky.social--for shrewdly ensuring that talks run on time and discussions are inclusive. @melodem.bsky.social

030

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

A key ingredient to the appeal and success of #MELODEM2025 (as with other #MELODEM mtgs) comes from the generous blocks of time dedicated to discussion. And a key ingredient of that discussion is excellent moderating.

120

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Broader context: A4 was viewed as innovative bc it started earlier in the course of disease. It was 5 years long. Is longer treatment duration needed? Or maybe a different agent operating via a different mechanism?

010

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

little evidence of cognitive benefit in spite of amyloid reduction. So not much in way of mediation.

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

... and adopted an IV approach that is a complement to mediation analysis. 1. % of treatment effect mediated by the surrogate (amyloid) 2. Assess treatment effect (on cogn) per change in surrogate

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Now at #MELODEM2025 Is reducing/removing amyloid from the brain a reasonable surrogate for slowing/stopping cognitive decline in dementia? @sfackley.bsky.social uses data from the A4 RCT (of solanezumab) to answer this q, ... @melodem.bsky.social

131

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

In Sim 2, the expected regression bias occurred, but ESPECIALLY wrt p-tau (blood). Are these errors responsible for weaker findings wrt blood biomarkers? Is it better to recruit a smaller # of ppl and do imaging, vs use blood biomarkers and recruit far more ppl (needed to detect an effect).

010

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Sim 1: age-->amyloid (imaging) and age-->ptau-181 (blood) Sim 2: amyloid (imaging)-->MMSE and ptau-181 (blood)-->MMSE

110

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

(Wild: in simulating measured centiloids of amyloid, there was not much published data to reference on noise in measuring this entity.)

100

Jennifer Weuve @epidancer.bsky.social · 25/06/2025

Why care about error in the measure of AD-related blood biomarkers? @ #MELODEM2025, Eleanor Hayes-Larson examines situations when a biomarker is the "exposure" and when it is the "outcome," using ADNI-informed simulations. @melodem.bsky.social

120

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Less dramatic diffs between AoU and NHANES in associations of a range of factors w mortality, except wrt clinical factors.

010

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Lots of substantial differences, incl in age (AoU older), college ed (AoU higher %), alc consumption (AoU more), missing data on clin factors (AoU higher %). Age-specific mortality rate: AoU lower (possibly due to better health or missing data)

110

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

He chose the 1999-initiated continuous #NHANES (w its sampling wts) as a gauge for national representativeness. Some observations:

120

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Closing out a day jammed w thoughtful talks @ #MELODEM2025 on methods challenges in #dementia research is Jingxuan Wang, presenting on the #AllOfUs research study and the extent to which it represents the US population. With implications for #research on #dementia. @melodem.bsky.social

140

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

And no race-ethnic diffs in assns of the biomarkers w cog fn.

010

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

New data from High School & Beyond: No difs between Black & White and Latinx & non-Lat White pts in 4 biomarkers. One exception: Black pts had lower pTau-181 than did White pts, but this comp did not survive multiple-comps correction.

110

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Community-based studies seem to be more likely to find, between race+ethn groups few diffs in biomarker levels and few diff in the assn of those biomarkers w clinical outcomes. But clinical samples have reported large diffs.

120

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Poor attention to inclusion dynamics in biomarker studies can produce all range of strange results and inferences.

100

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

The @manlyepic.bsky.social draws on #inclusionscience to show how to carry out meaningful #dementia #biomarker research in populations that carry that highest #dementia risk, e.g., Black and Latinx populations in the US. #MELODEM2025 @melodem.bsky.social

140

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

In the end, the perceived risk of mis-estimating the weights seemed worse than the perceived risk of bias from differential selection.

010

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

In addition, the 2 sampling stages seemed to draw different types of people, at least with respect to cognitive risk. The RSS process favored people at higher cognitive risk. But ppl in this group who had MRIs had lower cognitive risk.

120

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Inverse probability-of-participation seems like a great idea on the surface. He shows the tough decisions that have to be made in operationalizing those weights. E.g., who is the target population in the context of an open cohort and ongoing study cycles.

100

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

At bat @ #MELODEM2025 is @ryanmandrews.bsky.social who reported on his journey through understanding the potential for selection bias in a brain MRI study in which the participant pool resulted from a complex 2-stage sampling process: (1) stratified RS; (2) volunteering.

110

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Know your estimand before turning the crank.

020

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

In context of cog function and cog change, estimands include: controlled direct effect and, newly, conditional separable effects (Stensrud 2023 JASA). Choose based on interpretations and assumptions that are deemed acceptable.

110

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

(Similarly, in studies of #cognitive function, death is a truncation event.)

110

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

In studies of incident #dementia, death is a competing event. In studies of #cognitive change, death is a truncation event.

100

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

@palolili23.bsky.social argues (4 #MELODEM2025) that causal estimands have 4 elements + a 5th: intercurrent events. Ie, events that prevent us from observing the exposure or outcome, e.g., death. @melodem.bsky.social

121

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

In LASI-DAD, more formal education is assoc'd w more cognitive decline. Emma Nichols broke down 3 candidate non-causal explanations: practice effects, differential mortality, & regression to the mean. #MELODEM2025 @melodem.bsky.social

020

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

Many thanks to @statszach.bsky.social for his deft (and appropriately comical) moderating.

010

Jennifer Weuve @epidancer.bsky.social · 24/06/2025

... and modeling cognitive trajectory.

010