What are the opening hours?

Monday 9am - 5pm Wednesday 9am - 5pm Friday 9am - 5pm Saturday 9am - 5pm

Anil Bajnath, MD, 7030 Hi Tech Drive Suite 101, Hanover, MD (2026)

06/08/2026

A question that every clinician using epigenetic age testing should be asking: was that blood draw fasted or fed?

A preprint published June 2026 by Seale, Dwaraka, Giosan, Mendez and Smith investigated how an overnight fast followed by acute refeeding affects 24 epigenetic clocks across three cohorts — a within-person paired design of 15 participants, a cross-sectional validation cohort of 2,895 individuals, and technical replicates to establish the measurement noise floor. The results have direct implications for how clinicians and researchers collect samples, choose clocks, and interpret results.

PC-based clocks and SystemsAge organ-system clocks shifted between 0.5 and 3 years in the fasted versus refed state. Mortality-trained clocks — GrimAge V1, GrimAge V2, and OMICmAge — showed no detectable acute fasting effect in the paired cohort. DunedinPACE was similarly stable. Critically, these shifts were not random noise.

Fasting induced measurable immune cell redistribution within the same individuals: CD4+ naive T cells increased by 54.8% and T regulatory cells by 42.7%. When immune cell composition was added to the statistical models, 74 to 100% of the significant fasting effects were attenuated, and no clock retained a significant fasting association after immune adjustment in either cohort. The immune-insensitive IntrinClock showed no fasting effect in either dataset, providing a clean negative control for the proposed mechanism.

One of the paper's most important contributions is the reframing of intraclass correlation coefficients as a reliability metric. Raw ICCs remained 0.93 to 0.98 across clocks — comparable to technical replicates — even as group-level fasting shifts were clearly present. The authors demonstrate that high ICC and systematic directional shifts can coexist because a 1–3 year fasting effect is small relative to the large age-driven between-person variance that dominates the ICC denominator.

ICC alone cannot characterize a clock's behavior under a specific physiological perturbation. Reliability is not a fixed property of a clock; it depends on the perturbation, the population, and the adjustment applied.

The practical implications are concrete. In an intervention trial, a participant who provides a fasted sample at baseline and a fed sample at follow-up could show a spurious epigenetic age increase of 1–3 years — potentially obscuring a real treatment effect. The authors recommend three changes: standardizing fasting status at all collection timepoints, reporting both age-adjusted and immune-adjusted clock values, and interpreting ICC-based reliability on a perturbation-specific and clock-specific basis benchmarked against technical replicates.

This paper is essential reading for anyone designing, conducting, or interpreting epigenetic aging research or clinical longevity panels.

06/04/2026

I want to share a paper I co-authored that was just published in the Journal of Translational Medicine — and I want to share it the way it was written: with honesty about what we know, what we do not yet know, and what still needs to be built.

P4 medicine — predictive, preventive, personalized, participatory — has been a compelling vision for two decades. The core idea is right: molecular profiling should detect disease before symptoms appear, computation should guide prevention, care should be tailored to the individual, and patients should be genuine stakeholders in their own health data.

What has been harder is translating that vision into clinical systems that actually work at scale. The most ambitious early attempt, the Pioneer 100 / Arivale program, ultimately ceased operations — a reminder that scientific ambition without sustainable digital infrastructure cannot scale.

In this perspective, my colleagues and I argue that each pillar of P4 medicine needs to be anchored to a specific, implementable technology.

We propose a tiered multi-omics framework organized by measurement frequency and temporal dynamics, because understanding when to measure each layer is as important as knowing what to measure.

We share preliminary metabolomic data from 2,072 individuals as an illustration of what frequent metabolic profiling can reveal about aging biology — and we are careful to frame it as exactly that: a proof of concept from a cross-sectional dataset on a single platform, not yet replicated externally, and not yet ready for clinical use. The limitations are named in full because the field is not served by inflated claims.

We also write honestly about the challenges that remain unsolved: how to validate predictions that are individualized by design, how to build regulatory pathways for integrated systems rather than isolated components, how to ensure that multi-omic tools reach patients across economic and demographic divides, and how to protect privacy in longitudinal molecular datasets that are resistant to true de-identification.

The core tools described in this paper exist in prototype or early deployment. The integrated system does not. We offer this framework as one honest step toward closing that gap — with full awareness of how much further there is to go.
I would be glad to hear the perspectives of colleagues working on any piece of this problem.

06/02/2026

Most people who develop lung cancer were never eligible for screening in the first place.

A comprehensive review just published in Nature Reviews Clinical Oncology by Zhang, Park, Pandya, Shum, and an international team from MD Anderson, Mount Sinai, the Francis Crick Institute, UCLA, and collaborating centres confronts this reality directly — and maps what a transformed lung cancer screening and interception paradigm could look like.

The numbers are sobering. Despite clear mortality reductions of 20–24% demonstrated in the NLST and NELSON trials, fewer than 16% of eligible individuals in the USA underwent LDCT screening in 2024. More fundamentally, modelling studies project that nearly half of all newly diagnosed lung cancers occur in individuals who never meet eligibility criteria based on age and smoking history.

Lung cancer in never-smokers has become the fifth leading cause of cancer death globally. In Taiwan, it constitutes two-thirds of all lung cancers. Across Asia, it accounts for 63% of cases. Risk in this population is driven by PM2.5 air pollution, germline EGFR variants, familial clustering, and other factors that current US guidelines do not account for.

The review documents the emerging biomarker landscape designed to broaden and sharpen screening. AI-driven deep-learning models applied to LDCT — such as Sybil — have demonstrated AUCs of 86–94% for predicting lung cancer risk over 6 years, outperforming radiologist Lung-RADS assessment at comparable sensitivity. Liquid biopsy platforms integrating cfDNA methylation, fragmentomics, and miRNA panels are advancing, though most still show limited positive predictive value at screening-population prevalence.

Volatile organic compound breath analysis offers a non-invasive, scalable adjunct, with pooled sensitivity and specificity of 85% and 86% in a meta-analysis of 25 studies — though standardization across platforms remains a critical unresolved challenge.

The most forward-looking section of the review addresses precancer interception: the therapeutic targeting of persistent pulmonary nodules before they progress to invasive disease. These nodules — many representing atypical adenomatous hyperplasia, adenocarcinoma in situ, or minimally invasive adenocarcinoma — typically do not regress spontaneously. Pioneering research has shown that immunoediting and immune checkpoint upregulation begin at the precancerous stage, not at invasion.

The phase II Can-Prevent-Lung trial showed that anti-IL-1β therapy with canakinumab produced measurable shrinkage in 11 of 15 patients, with a median volume reduction of 7% per year versus 15% growth per year in matched controls.

The phase II IMPRINT-Lung trial of pembrolizumab demonstrated nodule shrinkage in 18 of 34 evaluable patients, with a toxicity profile strikingly more favourable than that observed in advanced-stage disease.

Preclinical data confirm that IL-1β blockade is effective only at the precancerous stage — not once invasive tumours are established — underscoring that interception must occur within a defined and finite therapeutic window.

For clinicians working at the intersection of precision, functional, and integrative medicine, this review reframes a critical clinical question: lung cancer does not have to be detected late. The tools to find it earlier, and perhaps to stop it before it arrives, are being built now.

Read the full paper.
DOI: https://doi.org/10.1038/s41571-026-01131-4

06/01/2026

There is a risk factor for chronic kidney disease that most clinicians are not asking about in the clinical encounter: the air their patients breathe every day.

A landmark review just published in Nature Reviews Nephrology by Ahn, Yun, Kim, Al-Aly, Bell, and Lee brings together the epidemiological and experimental evidence linking atmospheric stressors — ambient air pollution, wildfire smoke, and extreme temperatures — to acute kidney injury, chronic kidney disease, and kidney failure. The scope of the evidence is striking.

Multiple large cohort studies, including analyses of over two million US veterans and over 60 million Medicare enrollees, have reported consistent associations between long-term exposure to PM2.5 and a range of adverse kidney outcomes: incident CKD, eGFR decline, progression to kidney failure, and AKI-related hospitalization and death. These associations have been replicated across Europe, Asia, and North America, and observed at concentrations below current WHO air quality guidelines. Gaseous pollutants including nitrogen dioxide, ozone, and carbon monoxide have also been independently associated with kidney-related outcomes across multiple study designs and populations.

The mechanisms are not speculative. Inhaled PM2.5 generates reactive oxygen species directly and through NADPH oxidase activation, sensitizes proximal tubular epithelial cells to ischaemia-reperfusion injury, activates the NLRP3 inflammasome driving renal inflammation and fibrosis, and contributes to intraglomerular hypertension and glomerulosclerosis through both haemodynamic and non-haemodynamic pathways.

Extreme heat adds further insult through dehydration-driven fructokinase activation, uric acid generation, and extracellular volume depletion — all converging on tubular injury and GFR decline.

Wildfire smoke, with its higher free-radical and carbonaceous particulate content compared to urban ambient pollution, represents an emerging exposure of growing concern. A 2025 study of haemodialysis patients during the 2023 Canadian wildfires found an 18% increase in same-day all-cause mortality associated with smoke plume presence, and a 139% increase associated with a 10 µg/m³ increase in wildfire-related PM2.5.

The review also identifies populations at amplified risk: the elderly, children, pregnant women, and those with diabetes, hypertension, obesity, or pre-existing kidney disease. These are precisely the patients already in functional medicine and nephrology practices — and their environmental exposures are rarely part of the clinical picture.

This is the paper that makes the case for including atmospheric stressor assessment in the nephrology workup. Read it in full.

DOI: https://doi.org/10.1038/s41581-026-01073-1

05/26/2026

The gut microbiome has never been more clinically relevant — but not for the reason most practitioners assume.

The prevailing conversation about the microbiome focuses on which bacterial species are present or absent. A comprehensive review just published in Nature Reviews Gastroenterology & Hepatology argues that this framing misses the mechanism entirely.

The real determinants of microbiome–host crosstalk are microbial enzymes — thousands of them — that actively modify bile acids, steroid hormones, neurotransmitters, immunoglobulins, nutrients, and the drugs your patients take every day.

The review introduces and systematically develops the concept of microbiota–host isozymes: gut microbial enzymes that perform the same biochemical function as host enzymes but differ in protein structure, remain unaffected by drugs targeting their human counterparts, and vary substantially between individuals.

The flagship example is Bacteroides-derived DPP4, which degrades the incretin hormone GLP-1 in the gut — exactly as host DPP4 does — yet sitagliptin, the standard human DPP4 inhibitor, cannot inhibit the microbial version. The review's authors, writing from Peking University, argue this may explain why some patients with type 2 diabetes respond robustly to gliptin therapy while others do not.

The clinical reach of microbial enzymes extends further. Bile salt hydrolase activity has been mechanistically linked to polycystic o***y syndrome and to the metabolic benefits of metformin. Gut bacterial tyrosine decarboxylase converts levodopa into dopamine before it reaches the brain, reducing the efficacy of Parkinson's disease treatment. Microbial β-glucuronidase reconverts an inactivated chemotherapy metabolite back to its toxic form in the intestinal lumen, causing the severe diarrhea associated with irinotecan.

Microbial enzymes synthesizing trimethylamine from dietary choline are mechanistically implicated in cardiovascular disease risk. The same microbiota converts glucosinolates from cruciferous vegetables into chemoprotective isothiocyanates.

The review also maps the full landscape of how these enzymes are discovered — from classical biochemistry and functional metagenomics to AI-assisted platforms now capable of predicting novel bile acid metabolic enzymes at scale — and outlines therapeutic strategies ranging from targeted small-molecule inhibitors to engineered live biotherapeutics.

For those practicing at the intersection of functional and precision medicine, this paper provides the mechanistic foundation for what many clinicians have observed empirically: the microbiome shapes drug response, hormone metabolism, and metabolic health through molecular machinery that compositional testing alone cannot capture.

This is the paper to read.
DOI: https://doi.org/10.1038/s41575-026-01195-8

05/21/2026

A paper just published in Nature Aging answers a question that has sat unresolved in immunology for years: can macrophages undergo true, irreversible cellular senescence — and if so, do they matter clinically?

The answer, based on multi-omic characterization of primary mouse and human macrophages, appears to be yes on both counts. Salladay-Perez and colleagues identified a population of p21⁺TREM2⁺ senescent macrophages that are mechanistically distinct from classical inflammatory or anti-inflammatory macrophage states. These cells exhibit permanent cell-cycle arrest, a robust senescence-associated secretory phenotype, lipid droplet accumulation, and elevated type I interferon signaling driven by cytosolic mitochondrial DNA — a process the authors traced to upregulation of CMPK2 and activation of the cGAS-STING pathway.

One of the most clinically significant findings is that p16, the marker most commonly used to identify senescent cells in aging research, was paradoxically downregulated or unchanged in senescent macrophages in both mouse and human models. p21 proved to be the more reliable marker. This distinction matters enormously for any clinical or research application that relies on p16-based reporter systems to detect or target senescent immune populations.

The study also establishes cholesterol as a driver of macrophage senescence. Excess cholesterol ester loading via acetylated LDL was sufficient to induce a senescent, foam cell-like state characterized by TREM2 and p21 upregulation, SA-β-galactosidase activity, cell-cycle arrest, and SASP activation — without any exogenous DNA damage. In aged mouse livers, p21⁺F4/80⁺ macrophages rose from roughly 5% to 50% of the total macrophage pool. In human cirrhotic liver tissue, the macrophage senescence gene signature was most strongly enriched in TREM2⁺ scar-associated macrophages.

The therapeutic arm of the study is particularly striking. The senolytic ABT-263 selectively eliminated senescent macrophages — sparing M1 and M2 populations — and reduced hepatic steatosis, systemic inflammatory markers, and liver NAD⁺ decline in both aged mice and a diet-induced metabolic liver disease model. Liver NAD⁺ levels increased by 30% following senolytic treatment, consistent with the known role of CD38-expressing senescent cells in tissue NAD⁺ degradation.

For anyone working at the intersection of metabolic health, biological aging, and chronic inflammation, this paper offers one of the most mechanistically rigorous accounts yet of how the immune system becomes a source of its own pathology with age. The data suggest that macrophage senescence is not incidental to metabolic liver disease — it may be central to it.

The full paper is worth reading carefully, particularly the multi-omic datasets and the human cirrhosis single-cell analysis.

DOI: https://doi.org/10.1038/s43587-026-01101-6

05/20/2026

If you carry APOE4, your risk of Alzheimer's disease is two to three times higher than average — up to twelvefold if you carry two copies.

If you carry APOE2, your risk is significantly lower, and your age of onset is delayed even in families with high-penetrance mutations. The question that has remained largely unanswered is why — and whether the two alleles work through the same biological system in opposite directions, or through fundamentally different mechanisms.

A landmark study published in Nature Aging (Lu et al., May 2026) answers that question with the most comprehensive cross-platform proteomic analysis of APOE isoforms ever conducted. Spanning five cohorts — GNPC, BioFINDER-2, ADNI, UK Biobank, and PPMI — and more than 11,000 individuals across plasma and cerebrospinal fluid, the study systematically mapped how APOE2 and APOE4 shape the circulating proteome before and after the onset of amyloid pathology and clinical disease.

The APOE2 findings are particularly striking. Over 73% of APOE2-associated protein changes were already detectable in cognitively unimpaired individuals, stable across age groups from early adulthood to late life, and largely independent of AD diagnosis.

The key upstream mediators of APOE2's protective effect — SNAP23, APOB, WARS2, and PCLAF — point to mechanisms involving endocytic trafficking, lipid metabolism, mitochondrial translation, and DNA repair. Together they suggest that APOE2 establishes a constitutive biological resilience program early in life that actively suppresses stress and inflammatory signaling and resists the pathological remodeling that characterizes Alzheimer's disease progression.

APOE4 tells a different story. While some APOE4-associated protein changes appear early and independently of amyloid, the majority of the APOE4 proteomic signature is shaped by downstream pathology. Only a small number of proteins — including SPC25, S100A13, and TBCA — showed evidence of upstream mediation, meaning APOE4 itself drives their changes rather than inheriting them from amyloid accumulation. The rest largely reflect vascular, glial, immune, and proteostatic dysfunction that emerges as the disease unfolds.

Critically, APOE2 and APOE4 are not simply mirror images of each other. They operate through largely non-overlapping sets of mediator proteins. A small group of proteins — VPS29, PHGDH, and FOXO1 — are oppositely regulated by the two alleles and may represent molecular switch nodes that drive divergent biological trajectories toward resilience or vulnerability. These allele-specific upstream proteins are detectable before amyloid positivity, supported by Alzheimer's GWAS genetic evidence, and show spatial co-expression with APOE in the human brain — making them credible candidates for early biomarker development and allele-targeted preventive intervention.

For anyone thinking about brain longevity and Alzheimer's prevention, this paper reframes the conversation. Genotype-informed prevention will require separate therapeutic logic for APOE2 carriers and APOE4 carriers — not simply more or less of the same approach.

Read the full paper here: https://doi.org/10.1038/s43587-026-01123-0

05/19/2026

A large, rigorous randomized controlled trial has just produced the strongest evidence to date that a simple daily multivitamin can meaningfully slow biological aging — particularly in people whose biological age is already running ahead of their chronological age.

The COSMOS ancillary study, published in Nature Medicine (March 2026), examined the effect of two years of daily multivitamin-multimineral supplementation on five DNA methylation-based epigenetic aging clocks in 958 older adults with a mean age of 70 years. This was a prespecified, double-blind, placebo-controlled analysis embedded within one of the largest prevention trials ever conducted. Compliance exceeded 91%.

The epigenetic clocks measured included first-generation clocks predicting chronological age and second-generation clocks — GrimAge and PhenoAge — that predict health span and mortality risk.

The results were clock-specific and clinically informative. The first-generation lifespan clocks showed no significant effect from multivitamin use. The second-generation health-span clocks — those most strongly associated with morbidity, frailty, and all-cause mortality — showed significant slowing: a yearly reduction in PCGrimAge of 0.113 years and PCPhenoAge of 0.214 years compared to placebo.

DunedinPACE, the pace-of-aging clock, significantly increased in the placebo group but remained stable in the multivitamin group, though the between-group difference did not reach statistical significance.

The most striking finding is the subgroup interaction. Among participants with accelerated biological aging at baseline — those whose epigenetic age already exceeded their chronological age — the multivitamin reduced the yearly rate of GrimAge increase by 0.236 years, versus a non-significant 0.013 years among those without baseline age acceleration.

This interaction was statistically significant. In an exploratory analysis of nutritional biomarkers, lower baseline nutrient levels were associated with greater biological age acceleration, and the multivitamin significantly raised folate and lutein specifically among those with accelerated aging — pointing toward micronutrient insufficiency as a plausible driver of the excess aging signal, and its correction as the mechanism of benefit.

Component analysis of GrimAge further revealed that the multivitamin significantly improved DNA methylation-based measures of telomere length, beta-2 microglobulin, cystatin C, and GDF-15 — biomarkers spanning cellular senescence, immune aging, renal function, and mitochondrial stress. Cocoa extract, despite showing cardiovascular mortality benefits elsewhere in COSMOS, had no significant effect on any of the five epigenetic clocks tested.

The authors are measured about what remains to be established — whether epigenetic clock improvements translate to hard clinical endpoints requires longer follow-up and larger studies. But the precision nutrition implication is clear: the patients most likely to benefit from micronutrient supplementation are those already showing biological age acceleration, not those who are nutritionally replete.

Read the full paper here: https://doi.org/10.1038/s41591-026-04239-3

05/18/2026

What if a single AI model — trained only to predict the next measurement in a patient's physiological record — could forecast disease risk, track individual responses to nutrition interventions, and simulate the expected effects of statins, antihypertensives, and GLP-1 agonists, all without being explicitly programmed to do any of these things?

That is precisely what Lutsker et al. demonstrate with HealthFormer, a decoder-only transformer trained on data from the Human Phenotype Project, a longitudinal cohort of over 15,000 deeply phenotyped individuals.

The model tokenizes each participant's health trajectory across 667 measurements spanning blood biomarkers, body composition, sleep physiology, continuous glucose monitoring, gut microbiome, wearable-derived physiology, and behavioral and medication data. From a single self-supervised training objective — predict the next token — the full range of clinical downstream tasks emerges as queries on the same model.

The performance benchmarks across four independent cohorts are significant. Without fine-tuning, HealthFormer improved prediction for 27 of 30 incident disease and mortality endpoints over a standard age-sex-BMI baseline, and outperformed established clinical risk scores — including Framingham CVD and PREVENT-ASCVD — in every direct comparison. Concordance indices reached 0.834 for chronic kidney disease and 0.826 for cardiovascular mortality.

The intervention simulation results are what elevate this paper beyond standard clinical AI benchmarking. In the fully held-out PNP3 personalised nutrition trial, the model predicted individual six-month biomarker responses with correlations of r=0.78 for diastolic blood pressure and r=0.89 for triglycerides — zero-shot. Across 41 randomised trial comparisons covering lipid-lowering agents, antihypertensives, glucose-lowering drugs, exercise modalities, and dietary interventions, predicted direction of effect matched published trial estimates in every case.

The predicted mean fell within the published 95% confidence interval in 30 of 41 comparisons, without any task-specific training.
The authors are precise about what this does and does not represent. HealthFormer generates prognostic stratification under observed intervention exposures — it is not a causal counterfactual engine.

Its under-predictions at the high-potency end of the drug-effect distribution are coherent and mechanistically interpretable. But the directional fidelity across 41 independent trial comparisons, spanning diverse drug classes and physiological systems, marks a meaningful threshold in what generative physiological modelling can achieve.

For those working at the frontier of precision and functional medicine, this paper deserves careful reading. It describes not a finished clinical tool, but the generative substrate from which clinical digital twins may ultimately be built.

Read the full preprint here: https://arxiv.org/abs/2604.27899

05/12/2026

A new study published in Cell Genomics (Liu et al., June 2026) does something most aging research cannot: it moves beyond correlation to establish causality between chronic inflammation and biological age acceleration.

The investigators analyzed four independent human cohorts ranging from young, healthy adults (mean age 25) to elderly individuals with established multimorbidity (mean age 69–72), measuring circulating inflammatory proteins alongside DNA methylation-based epigenetic aging scores. Four epigenetic clocks were assessed — Horvath and Hannum, which predict lifespan, and GrimAge and PhenoAge, which predict health span. The health-span clocks showed markedly stronger and more consistent associations with inflammatory proteins, frailty, and the cumulative burden of age-related diseases including COPD and malignancies. Chronological age, by contrast, was associated only with hypertension. This finding reinforces that biological age, particularly as captured by GrimAge and PhenoAge, is the more clinically meaningful variable.

The central contribution of this paper is its Mendelian randomization analysis. Drawing on protein quantitative trait loci from the UK Biobank (over 33,000 participants) and epigenetic age acceleration GWAS data from a non-overlapping cohort of nearly 35,000, the authors established that four circulating proteins — CXCL9, CXCL10, CCL11, and IL-18 — are causal drivers of epigenetic age acceleration. CXCL9 produced the strongest causal signal, surviving multiple-testing correction and replicating across five independent MR methods. Reverse MR revealed minimal evidence that epigenetic aging causes the inflammatory changes, supporting a directionality in which inflammation precedes and accelerates biological aging.

All four of these proteins are mechanistically connected to the interferon signaling pathway, a finding the authors corroborate through transcription factor enrichment analysis showing that IRF1, STAT1, and STAT3 are the dominant upstream regulators of the EAA-associated protein network. The study also documents that aging is associated with reduced IFN-γ production by lymphocytes in response to microbial stimulation — a finding consistent with immunosenescence, and one that points to a dual pathology: chronic low-grade interferon activation coexisting with impaired adaptive immune responsiveness.

The clinical implication is direct. The interferon pathway is not a passive bystander in biological aging — it is a causal mechanism and a plausible therapeutic target. For practitioners working at the intersection of longevity, functional medicine, and immunology, this paper offers both a mechanistic framework and a rationale for future intervention studies using interferon-modulating agents.

Read the full paper here:

https://doi.org/10.1016/j.xgen.2026.101218

Anil Bajnath, MD

06/08/2026

06/04/2026

06/02/2026

06/01/2026

05/26/2026

05/21/2026

05/20/2026

05/19/2026

05/18/2026

05/12/2026

Address

Opening Hours

Website

Alerts

Contact The Practice

Shortcuts

Share

Category

Monday	9am - 5pm
Wednesday	9am - 5pm
Friday	9am - 5pm
Saturday	9am - 5pm