Peter Attia MD197 - The science of obesity & how to improve nutritional epidemiology | David Allison, Ph.D
CHAPTERS
- 0:02 – 9:01
David Allison's path from psychology to statistics and obesity
Peter Attia introduces David Allison and asks about his background. Allison explains how a habit of asking for evidence took him from psychology to statistics, and how a Vassar course on Stanley Schachter's work drew him to obesity.
- •Allison trained as a psychologist and became a statistician because he wanted to understand the evidence
- •He argues body fat obeys the same laws of probability and physics as any other variable
- •He compares lay opinions on obesity with the deference given to diabetologists on beta cells
- •Schachter's creative experiments on external cues and eating hooked him as an undergraduate
- •He likes that obesity can be studied from many angles, with no single cause
- 9:01 – 17:11
The NHANES III wake-up call and the shift to public health advocacy
Allison describes how obesity was viewed in the 1980s and how the early-1990s NHANES III data changed that. Public health and environmental framing brought funding and attention but, in his view, also less rigor and more advocacy.
- •NHANES III showed a big jump in obesity and the word epidemic took hold
- •Pharmaceuticals and surgery were stigmatized before the 1990s
- •Kelly Brownell and others promoted a toxic environment framing
- •Public health people entered without knowing the prior literature, and rigor diluted
- •Allison calls the energy balance statement correct but not explanatory, using a triangle analogy
- 17:11 – 20:36
Genetics of obesity: twin and adoption studies
Attia asks about concordance in identical twins reared apart. Allison reports BMI correlations of about .9 and explains how Mickey Stunkard's Nordic twin and adoption studies established the genetic component around 1983.
- •Selective breeding of animals is prima facie evidence of a genetic component
- •BMI correlation for identical twins reared apart is close to that for twins reared together, about .9
- •Child rearing in the home has some influence but not a huge one
- •Stunkard's Swedish and Danish twin and adoption studies in the New England Journal of Medicine settled the question
- 20:36 – 27:47
Behavioral treatment, fen-phen, and the rise of obesity pharmacotherapy
Attia asks why genetic understanding did not lead to medical treatment sooner. Allison covers behavior therapy tools, the fen-phen craze, the valvulopathy discovered by a Mayo Clinic pathologist, and how the episode pushed obesity toward being seen as a serious medical disorder.
- •Most treatment tools aim to help people reduce energy intake, which works but is hard
- •Behavioral approaches have improved only incrementally
- •Fen-phen combined phentermine and fenfluramine and showed better benefit than expected
- •Fenfluramine was withdrawn after a valvulopathy was found; phentermine is still used
- •The episode helped make obesity a credible medical field
- 27:47 – 34:08
Bariatric surgery and the Swedish Obese Subjects study
Attia asks when surgery moved from fringe to mainstream. Allison credits Lars Sjöström's Swedish Obese Subjects study, a controlled but non-randomized trial, with showing reduced mortality, and discusses effect sizes and the hypertension finding.
- •Surgery was controversial even among physicians and scientists
- •Sjöström studied whether surgery prolongs life, after being told it was obvious
- •The study used matched controls because the IRB would not allow randomization
- •Across studies, mortality reduction is often around 50%, with heterogeneity
- •Diabetes improvement persisted, but hypertension came back
- •Allison stresses that you must do the experiment rather than assume
- 34:08 – 41:50
The obesity paradox and the shifting BMI nadir
Allison explains that the term obesity paradox covers two ideas: the U-shaped BMI–mortality curve and better survival of heavier people who are sick or injured. He describes a model, with a Grand Canyon fat-suit analogy, in which the lowest-mortality BMI rises with age, and gives sample BMI values.
- •The phrase is loosely defined and used to mean at least two things
- •The mortality curve is bathtub-shaped, with higher mortality at very low and very high BMI
- •The Grand Canyon fat-suit analogy illustrates why the optimal BMI could rise with age
- •The nadir roughly tracks average weight gain of about a pound a year, reaching low 30s by age 80
- •Allison says the data fit his model but do not prove it
- 41:50 – 52:08
Confounding in BMI and mortality data
Attia asks about confounders. Allison calls confounding a nightmare of observational research, describes how excluding smokers and early deaths was thought to fix it, and reports his team's proofs that excluding early deaths does not reliably help. He argues the key question is the effect of interventions.
- •Allison recommends triangulating across multiple kinds of studies
- •His team showed mathematically that excluding early deaths can fail or make confounding worse
- •Likely confounders include smoking, socioeconomic status, and stigma
- •The lowest-mortality BMI has risen over calendar time, perhaps because what is stigmatized changes
- •Following Don Rubin, the right question is what the intervention is and compared to what
- •Surgery, SGLT2 inhibitors and GLP-1 agonists appear to prolong life
- 52:08 – 1:02:31
BMI's limits: body composition, ethnicity, and better tools
Allison defends BMI as a tool for epidemiology while conceding its limits for individuals. He discusses ethnic differences in the BMI–mortality curve and the lack of a BMI–mortality association in Hispanic Americans. Attia describes DEXA metrics he prefers, and Allison describes 3D photography and other tools, stressing fit for purpose.
- •BMI uses mass over height squared, an empirical choice dating to Quetelet
- •BMI is useful for epidemiology but not a perfect clinical tool
- •Risk appears at lower BMIs in people of Middle Eastern and East Asian descent
- •Allison's group found no association between BMI and mortality in Hispanic Americans
- •Attia prefers visceral fat, bone density and lean mass index from DEXA
- •3D photography, DEXA and other methods may replace BMI; the tool should fit its purpose
- 1:02:31 – 1:06:55
What actually reduces obesity: evidence for clinical versus public health approaches
Attia raises the so-what question of acting on population data. Allison says surgery, drugs, and some individualized or group behavioral treatments work, but community and school public health measures lack evidence, and he says what he would fund with $10 million.
- •Surgery, pharmaceuticals, and individualized behavioral treatment work to varying degrees
- •Allison sees virtually no evidence that community, school, or public health interventions reduce obesity
- •He distinguishes goals: consciousness raising, showing care, or actually reducing obesity
- •With $10 million he would fund bariatric surgery for a subset of people rather than farmers markets
- 1:06:55 – 1:18:45
Why maintaining weight loss is hard: thrifty gene, drifty gene, and other evolutionary ideas
Asked why weight loss maintenance is so hard, Allison critiques the thrifty gene hypothesis and presents John Speakman's drift hypothesis and other evolutionary explanations. He suggests multiple factors combine, drawing an analogy to theories of sexual reproduction and the Red Queen hypothesis.
- •The thrifty gene idea assumes humans were always near starvation, which is not clearly supported
- •Speakman's drifty hypothesis proposes freedom from predation relaxed selection for satiety genes
- •Other proposals involve food safety, mating competition, and social order
- •Several partial explanations may combine, as with theories for the evolution of sexual reproduction
- •The Red Queen hypothesis ties sex to keeping up with fast-evolving microbes
- 1:18:45 – 1:35:12
Nutritional epidemiology: abolition versus reformation and the mouse study
Attia sets out the camps on nutritional epidemiology. Allison favors reformation over abolition, arguing many findings do not hold up in trials and presenting a mouse study in which the observational association reversed the causal effect. He also argues for stronger designs between randomized trials and weak observational studies.
- •Allison describes abolitionists on one side and defenders of the status quo on the other
- •He says many nutritional epidemiology findings have not held up in randomized trials
- •In his mouse study, mice choosing to eat more lived longer, while assigned higher calories shortened life
- •Co-twin controls and natural or non-randomized interventions are stronger designs
- •Attia criticizes food frequency questionnaires and small hazard ratios
- •Allison calls for honesty about limits and epistemic humility
- 1:35:12 – 1:42:07
Katherine Flegal, ad hominem attacks, and why nutrition science is so heated
Attia asks about Katherine Flegal's paper, and Allison describes the media frenzy and personal attacks that followed. He argues that data, methods, and logic are what count in science and explains that everyday experience with food makes people opinionated.
- •Flegal's 2005 meta-analysis found overweight-range BMI not consistently tied to higher mortality
- •The finding was not new, but it was met with vociferous criticism
- •Allison says three things matter in science: data, methods, and the logic connecting them
- •Attacks and retraction attempts have also targeted other nutrition researchers
- •Everyday experience with food, culture, and emotion makes people feel like experts
- 1:42:07 – 1:53:29
Path forward on obesity: treatments, stigma, and cluster randomized trials
Asked whether nutrition will progress like surgery and drugs, Allison questions whether any diet reliably produces sustained weight loss. He supports better access to surgery and drugs, progress on stigma as a moral issue, and criticizes the school and community public health literature, including flawed cluster randomized trial analyses.
- •Allison questions the premise that a diet reliably causes sustained weight loss
- •He backs more funding and access for surgery and pharmaceuticals
- •He says shaming people about body habitus is a moral issue, not an empirical one
- •He calls public health, school and community approaches off the path, with distorted evidence
- •Cluster randomized trials must account for clustering, and flawed analyses have led to retractions
- 1:53:29 – 1:57:46
Most promising interventions: education, housing, and four funding pots
Asked to speculate about promising interventions, Allison points to general education, citing the Abecedarian study and the Moving to Opportunity study. He outlines how he would split a federal budget into surgery, pharmaceuticals, general education and well-being, and basic research.
- •Provocative evidence suggests general education, especially for girls and women, is linked to lower obesity and diabetes
- •The Abecedarian early-education study found less obesity in women decades later
- •Moving to Opportunity found less obesity and diabetes among families moved to less poor neighborhoods
- •His four budget pots are surgery, pharmaceuticals, general education and well-being, and basic research
- 1:57:46 – 2:08:11
Reproducibility, normative versus non-normative errors, and the rhetoric of science
Discussing his National Academies reproducibility panel, Allison says there is no crisis but no cause for complacency. He distinguishes normative from non-normative errors with historical examples and argues scientists should be willing to lose battles rather than use rhetorical tricks.
- •The panel's phrase was no crisis but no cause for complacency
- •Allison believes science is more rigorous than ever but must still improve
- •Galileo's light-speed experiment is a normative error; the thymus sampling error is non-normative
- •Misanalysis of cluster trials and causal use of food frequency questionnaires are non-normative errors
- •Scientists should avoid ad hominem and exaggeration and promote asking simple questions about evidence
- 2:08:11 – 2:21:41
Rebuilding trust: scientist versus advocate hats
Attia reflects on COVID and public trust. Allison argues that people should say whether they speak as scientists or advocates, that trust in science overall is not down, and that trust in nutrition scientists specifically is low. They close on a two-hat idea and fatherly examples.
- •Attia says COVID communication has eroded faith in science
- •Allison says his sacred duty as a scientist is to pursue and communicate truth
- •Public health professionals may prioritize improving health over strict truth-telling
- •Allison says people should state whether they speak as scientists or advocates
- •Pew surveys suggest trust in science is high, but trust in nutrition scientists is low
- •The two-hat idea is illustrated with parenting
