Author: ajpendo

The regression evidence, extreme LDL targets, safety concerns, and the data-driven case for putting clinical mythology to rest

In my prior essay — “Is Lower Longer Better?” — I described a patient who came to a follow-up confused because someone had told him his LDL of 14 mg/dL was dangerous. He had metabolic syndrome, a coronary calcium score approaching 3,000, and the vascular biology of a man two decades older. His therapy was working. And someone had told him to stop.

That essay was about culture — about what happens when specialists don’t communicate, when clinical mythology overrides data, and when patients get caught in the crossfire.

This essay is about the science.

Because the fears are real — clinicians are genuinely worried about very low LDL, and some of those concerns deserve a direct and systematic response. Others are frankly misinformation that has taken on a life independent of any evidence base. And a subset represent legitimate questions where the data offers strong but not absolute reassurance, and intellectual honesty requires saying so.

I use a target of less than 40 mg/dL in my extreme-risk patients. I am going to explain exactly why that number is defensible, what the evidence says about every major safety concern, and why the clinical intuition that very low LDL is dangerous is not supported by the science we now have.

* * *

## Part I: The Regression Story — Why the Math Has No Floor

Let’s start where the evidence starts: the Cholesterol Treatment Trialists’ (CTT) meta-analysis, the foundational dataset in modern lipidology. Pooling individual participant data from 26 randomized trials and over 170,000 participants, the CTT established a relationship that has now been replicated across every major lipid-lowering trial since: each 1 mmol/L reduction in LDL produces approximately a 22% proportional reduction in major vascular events.¹

This is a log-linear relationship. Not a linear one. Not a threshold model. Log-linear.

What that means mathematically is that the proportional benefit is consistent across the range of LDL values studied — there is no plateau, no inflection point, no floor below which the curve flattens. The risk reduction observed going from 160 to 130 is proportionally similar to the risk reduction observed going from 70 to 50. Each unit of reduction carries its own freight.

This has been confirmed in analyses stratified by achieved LDL level within major trials. In FOURIER, investigators prespecified an analysis of outcomes across achieved LDL categories: those with average LDL below 20 mg/dL had better outcomes than those in the 20–40 range, who had better outcomes than those in the 40–70 range.² Progressively lower MACE incidence down to LDL levels below 10 mg/dL, without a safety offset. When the FOURIER data were overlaid directly on the CTT regression line comparing the effect of PCSK9 inhibition to years of statin therapy — accounting for the short follow-up duration of FOURIER relative to longer statin trials — the data points align on the same curve.³ The mechanism is the same. The math is the same.

The CTT also demonstrated that duration matters independently. A 10–12% reduction in events per mmol/L reduction in the first year of statin therapy rises to a 22–24% reduction per mmol/L in each subsequent year.³ This is the “longer” in “lower longer” — not a slogan, but a quantified phenomenon in the regression data. Earlier initiation, earlier achievement of target, and sustained time at target all compound. An LDL of 30 for fifteen years is not the same exposure as an LDL of 30 for two years. The curves separate with time.

The most powerful demonstration of lifetime exposure comes not from clinical trials but from nature. Individuals with PCSK9 loss-of-function mutations associated with a 28% reduction in mean LDL cholesterol showed an 88% reduction in the risk of CHD over a 15-year interval in the Atherosclerosis Risk in Communities (ARIC) study.⁴ This reduction was dramatically larger than predicted by short-duration statin trials — a discrepancy that quantifies exactly what lifelong low LDL produces that a 2–5 year trial cannot capture.

The first known individual with no immunodetectable circulating PCSK9 — a compound heterozygote for two inactivating mutations — had a strikingly low plasma LDL of 14 mg/dL and was, in detailed clinical evaluations over a decade, an apparently healthy, fertile, normotensive woman with normal liver and renal function.⁵ She was not a medical curiosity. She was a proof of concept.

The regression evidence, the genetic data, and the long-term trial data all point in the same direction and converge on the same conclusion: the math has no floor within the range of values we can clinically achieve.

* * *

## Part II: Who Gets a Target Below 40 — The Extreme-Risk Framework

Guidelines have progressively lower targets as risk increases, but it is worth being explicit about what extreme risk means clinically and why a target below 55 — and in some patients, below 40 — is not aggressive for its own sake. It is calibrated to the biology.

The AACE/ACE 2020 criteria for extreme cardiovascular risk include: progressive ASCVD despite maximally tolerated LDL-lowering therapy; established ASCVD in patients with diabetes, stage 3–4 CKD, or heterozygous familial hypercholesterolemia; premature ASCVD (age under 55 in men, under 65 in women); and prior ASCVD events (recurrent acute coronary syndrome within two years, multivessel disease with recent ACS, PCSK9 inhibitor-eligible patients with recent ACS).⁶

The 2026 ACC/AHA Dyslipidemia Guidelines, published March 13, 2026, codify a Class I recommendation for LDL below 55 mg/dL in very high-risk patients and explicitly acknowledge that pushing further in extreme-risk patients is supported by the evidence.⁷

These are not arbitrary numbers. They are anchored in the regression data. For a patient with a CAC score of 3,000 and metabolic syndrome, or a patient with T1DM and CABG in their forties, or a patient with recurrent ACS on maximal triple therapy — the residual cardiovascular risk at an LDL of 70, or even 55, remains clinically substantial. The question is not whether to push further. The question is whether the safety concerns about doing so hold up under scrutiny.

They do not.

* * *

## Part III: The Safety Myths — Data-Driven Myth-Busting

### Fear #1: “Very low LDL causes cognitive decline and dementia”

This is the fear I hear often, and it is the one most thoroughly contradicted by the evidence.

It has an origin story. Early pharmacovigilance data from PCSK9 inhibitor trials raised signal flags on self-reported neurocognitive events. These signals were not replicated in placebo-controlled analysis and were almost certainly attributable to surveillance bias — patients on a new medication reporting symptoms they attributed to that medication.

The EBBINGHAUS trial was designed specifically to answer this question in FOURIER. Using a validated neuropsychological test battery sensitive to both positive and negative cognitive effects, EBBINGHAUS found no difference between evolocumab and placebo across executive function, memory, or psychomotor speed at a median LDL of 30 mg/dL over 19 months of follow-up.⁸

EBBINGHAUS-OLE then followed these patients long-term and concluded that exposure to very low levels of LDL cholesterol, achieved via PCSK9 inhibition and statin therapy, was not associated with cognitive impairment through long-term follow-up.⁹

The Mendelian randomization data adds a crucial dimension. Genetic variants that mimic lifelong PCSK9 inhibition — individuals born with loss-of-function mutations who have had LDL levels below 30 for their entire lives — show no increase in dementia, cognitive impairment, or neurodevelopmental disorders.¹⁰ This is decades-long exposure, far exceeding any clinical trial follow-up. If very low LDL caused cognitive harm, these populations would show it. They do not.

One legitimate caveat deserves acknowledgment: most trial participants were in their early sixties on average, not the age group at highest risk for dementia. Whether intensive LDL lowering affects cognition differently in those over 75 with pre-existing vulnerability remains an area for further study. Intellectual honesty requires saying that. But the currently available evidence — across short-term trials, long-term extensions, and genetic epidemiology — offers strong reassurance that the “low LDL causes dementia” narrative is not supported by data.

### Fear #2: “Very low LDL will suppress testosterone and sex hormones”

This fear has a plausible biological mechanism, which is partly why it persists. Cholesterol is the precursor for all steroid hormones — cortisol, aldosterone, estrogen, testosterone, DHEA. It is a reasonable hypothesis that radically lowering the substrate would impair the product.

The hypothesis does not survive the data.

The adrenal and gonadal steroidogenic pathways have multiple redundant sources of cholesterol substrate: circulating LDL, intracellular de novo synthesis via HMG-CoA reductase (the same pathway statins target), and HDL-mediated delivery. These tissues do not rely on a single supply line. A phase 3 double-blind randomized controlled trial of evolocumab over 52 weeks investigated steroid hormone and vitamin E levels; findings suggest that the synthesis of steroid hormones is not significantly dependent on circulating LDL levels, with cortisol, ACTH, testosterone, and estradiol all remaining in normal ranges despite sustained very low LDL.¹¹

A 2025 study presented at the European Society of Endocrinology Congress found a correlation between steroid hormones and LDL cholesterol levels but no adrenocortical insufficiency was observed, concluding that the adrenal glands are capable of full adaptation to a profound cholesterol deficiency using it as a substrate for steroid hormone synthesis, and that long-term hypolipidemic therapy does not induce adrenocortical insufficiency.¹²

The PCSK9 loss-of-function data is again instructive. There is no evidence that adrenal, ovarian, or testicular hormone production is impaired even in patients with LDL levels below 15 mg/dL. Individuals born with no functional PCSK9 — lifelong LDL in the low teens — reproduce normally, have normal adrenal function, and show no hormonal deficiency syndrome.

It is worth addressing a nuance here that reflects my own clinical context as an endocrinologist and testosterone specialist. There are observational studies showing an association between lower LDL and lower free testosterone at a population level.¹³ This has been interpreted by some as evidence of causal harm. It is almost certainly a confounding relationship — the same metabolic syndrome that produces elevated LDL also produces elevated SHBG dysfunction and altered gonadal axis tone. Treating LDL aggressively does not cause hypogonadism. What is true is that men with extreme cardiovascular risk who are also hypogonadal deserve evaluation of both conditions on their own terms — and may benefit from testosterone therapy managed appropriately alongside their lipid regimen. These are parallel clinical problems, not a trade-off.

The testosterone concern about very low LDL is, in the current evidence base, clinical mythology. It is not supported by interventional data, physiological mechanistic data, or genetic human models of lifelong very low LDL.

### Fear #3: “Very low LDL causes hemorrhagic stroke”

This is the most legitimate of the safety concerns, and it deserves the most careful treatment.

The SPARCL trial — which tested atorvastatin 80 mg in patients with recent stroke or TIA — found a statistically significant increase in hemorrhagic stroke in the treatment arm (55 vs. 33 events; HR 1.66).¹⁴ This signal was real and was not dismissed. Risk factors in that analysis included male sex, history of hemorrhagic stroke as the qualifying event, age, and poorly controlled hypertension (systolic ≥160 mmHg).

Several contextualizing points are essential. First, the absolute numbers were small and the benefit from reduction in ischemic stroke and cardiac events substantially outweighed the hemorrhagic stroke risk overall. Second, the signal appeared to be concentrated in a specific phenotype: patients with prior hemorrhagic stroke, hypertensive, and male. Third, subsequent PCSK9 inhibitor trials — including FOURIER and ODYSSEY OUTCOMES — did not find significant increases in hemorrhagic stroke, even at median LDL levels of 30–40 mg/dL. There were no differences in serious adverse events, muscle-related events, new-onset diabetes, cataracts, hemorrhagic stroke, or neurocognitive events with evolocumab compared with placebo over up to 8.4 years of follow-up in FOURIER-OLE.¹⁵

The practical clinical implication is real but narrow: in a patient with a history of hemorrhagic stroke, poorly controlled hypertension, and high cardiovascular risk, the decision about LDL target and agent selection warrants individualized discussion. This is a specific subgroup, not a general population concern. For the vast majority of extreme-risk patients without this phenotype, the hemorrhagic stroke signal from SPARCL does not apply and should not be extrapolated.

### Fear #4: “Very low LDL causes cancer”

Statin therapy has no effect on the incidence of, or death from, any type of cancer. This has been established in CTT meta-analyses covering over 175,000 participants.¹ The PCSK9 inhibitor trials are consistent with this — the incidence of cancer was comparable between the very low LDL-C and control groups in the 2025 meta-analysis of six RCTs covering nearly 53,000 patients.¹⁶

The cancer concern about very low LDL arose from early observational epidemiology — some studies noted that patients with cancer had low cholesterol. The causation runs the other way: cancer consumes lipid substrate and lowers cholesterol as a consequence of disease. Reverse causation, not a treatment effect.

### Fear #5: “Very low LDL causes new-onset diabetes”

This is a statin-specific concern that has been inappropriately extended to LDL lowering as a whole.

High-intensity statins do modestly increase fasting plasma glucose and incident diabetes risk, particularly in patients with prediabetes and elevated BMI — the patients already at the highest metabolic risk. The mechanism is likely HMG-CoA reductase inhibition in pancreatic beta cells and skeletal muscle insulin signaling, not the LDL lowering per se.

Studies have not demonstrated an increase in fasting plasma glucose or the incidence of new-onset diabetes associated with PCSK9 inhibitor use. This is a critical distinction. The diabetes signal belongs to statins through a specific mechanism — it does not belong to “low LDL” as a category. Adding a PCSK9 inhibitor to achieve extreme LDL targets does not carry this risk. In fact, for patients who are already on high-intensity statins and are concerned about diabetes risk, optimizing the regimen with a PCSK9 inhibitor may allow consideration of lower-intensity statin dose in some circumstances — a clinical nuance worth discussing.

* * *

## Part IV: The “Experiments of Nature” — What Lifelong Low LDL Actually Looks Like

The most powerful argument for the safety of very low LDL is not a clinical trial. It is the biological record of humans who have lived their entire lives with LDL in the range we are trying to pharmacologically achieve.

Those with PCSK9 loss-of-function mutations, whose LDL-C can be as low as 14 mg/dL, generally show no major coexisting conditions like neurocognitive deficits, diabetes, cataracts, or stroke, suggesting that such low levels are physiologically well-tolerated over a lifetime.¹⁷

This matters because it answers the argument that our trial data is too short. FOURIER-OLE ran nearly nine years. EBBINGHAUS-OLE followed cognitive function long-term. But nature’s experiment is a lifetime. The woman with no functional PCSK9 — LDL of 14, healthy, fertile, with normal hormones and normal cognition — is not an outlier. She is the model.

When clinicians express concern about pushing a patient’s LDL to 20 or 30 with pharmacotherapy, they are in effect saying that a level humans tolerate lifelong without pathological consequence is dangerous to pharmacologically achieve. The burden of proof runs the other direction.

* * *

## Part V: What’s Coming — The Oral PCSK9 Story and Lp(a)

No discussion of extreme LDL targets in 2026 is complete without acknowledging what is about to change therapeutically.

Enlicitide (MK-0616), Merck’s investigational oral PCSK9 inhibitor, completed its Phase 3 CORALreef trials with significant LDL reduction and a safety profile comparable to placebo. Late-breaking Phase 3 data from the CORALreef AddOn trial was presented at ACC.26 in March 2026. If approved, this would be the first oral PCSK9 inhibitor — a daily pill rather than a biweekly or monthly injection — with substantial implications for adherence and the breadth of the population willing to pursue extreme targets.

The adherence problem is real. Many patients who need PCSK9 inhibitors are not on them, and many who start them don’t stay on them. An oral option does not eliminate prior authorization barriers or cost concerns, but it removes the injection barrier that meaningfully limits uptake in primary care and some subspecialty settings.

The Lp(a) story is the other frontier. Extreme LDL reduction is necessary but not sufficient for some patients. Those with elevated Lp(a) carry residual risk that is not attenuated by LDL lowering — Lp(a) is an independent, causal, and largely genetically determined risk factor. RNA-based therapies in late-stage development (olpasiran, pelacarsen) have shown reductions above 90% in Lp(a) levels. The clinical outcomes data is pending. For extreme-risk patients in 2026, Lp(a) measurement is not optional — it is part of the risk characterization that informs how aggressively to pursue LDL targets alongside other strategies.

* * *

## Part VI: Putting It Together — A Clinical Framework

For the clinician reading this: here is how I apply this evidence in practice.

I identify extreme-risk patients using a structured approach: AACE/ACE 2020 criteria as the backbone, augmented by CAC scoring, Lp(a) measurement, and clinical judgment about trajectory. For a 52-year-old with metabolic syndrome, CAC of 3,000, and LDL of 70 on high-intensity statin and ezetimibe — that patient gets a PCSK9 inhibitor and a target below 40 mg/dL. The evidence supports it. The safety data supports it. The 2026 guidelines support it.

I document the rationale explicitly. I communicate with co-managing clinicians before changing my plan, and I expect the same in return. I explain the evidence to patients clearly: the myth of too-low LDL, what the genetic and clinical trial data show, and why the number I’m targeting is not arbitrary.

And when a colleague expresses concern that the LDL is too low — I welcome that conversation. I offer the evidence. I am happy to discuss the nuances, including the legitimate hemorrhagic stroke caveat in the right patient phenotype. What I am not willing to do is de-intensify therapy in an extreme-risk patient based on a safety concern the data does not support, without discussion, in a way that leaves the patient confused about who to believe.

That is not caution. That is harm.

* * *

## References

1. Cholesterol Treatment Trialists’ (CTT) Collaboration. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. *Lancet.* 2010;376(9753):1670–1681.

2. Giugliano RP, Pedersen TR, Park JG, et al. Clinical efficacy and safety of achieving very low LDL-cholesterol concentrations with the PCSK9 inhibitor evolocumab: a prespecified secondary analysis of the FOURIER trial. *Lancet.* 2017;390(10106):1962–1971.

3. Sabatine MS, Wiviott SD, Im K, Murphy SA, Giugliano RP. Efficacy and safety of further lowering of LDL cholesterol in patients starting with very low levels: a meta-analysis. *JAMA Cardiol.* 2018;3(9):823–828.

4. Cohen JC, Boerwinkle E, Mosley TH Jr, Hobbs HH. Sequence variations in PCSK9, low LDL, and protection against coronary heart disease. *N Engl J Med.* 2006;354(12):1264–1272.

5. Zhao Z, Tuakli-Wosornu Y, Lagace TA, et al. Molecular characterization of loss-of-function mutations in PCSK9 and identification of a compound heterozygote. *Am J Hum Genet.* 2006;79(3):514–523.

6. Jellinger PS, Handelsman Y, Rosenblit PD, et al. American Association of Clinical Endocrinologists and American College of Endocrinology Guidelines for Management of Dyslipidemia and Prevention of Cardiovascular Disease. *Endocr Pract.* 2020;26(Suppl 1):1–57.

7. Blumenthal RS, Morris PB, et al. 2026 ACC/AHA Guideline on the Management of Dyslipidemia. *J Am Coll Cardiol.* Published online March 13, 2026. doi:10.1016/j.jacc.2025.11.016

8. Giugliano RP, Mach F, Zavitz K, et al. Cognitive function in a randomized trial of evolocumab (EBBINGHAUS). *N Engl J Med.* 2017;377(7):633–643.

9. Zimerman A, O’Donoghue ML, Ran X, et al. Long-term cognitive safety of achieving very low LDL cholesterol with evolocumab (EBBINGHAUS-OLE). *NEJM Evidence.* 2025;4(1):EVIDoa2400112.

10. Mefford MT, Rosenson RS, Cushman M, et al. PCSK9 variants, LDL-cholesterol, and neurocognitive impairment: The REGARDS Study. *Circulation.* 2018;137(12):1260–1269.

11. Blom DJ, Djedjos CS, Monsalvo ML, et al. Effects of evolocumab on vitamin E and steroid hormone levels: results from the 52-week, Phase 3, double-blind, randomized, placebo-controlled DESCARTES study. *Circ Res.* 2015;117(8):731–741.

12. Klimenko LL, et al. Effect of very low LDL cholesterol levels on steroid metabolome. *Endocrine Abstracts.* 2025;110:P115. [ESPE/ESE 2025 Joint Congress].

13. Rastrelli G, Lotti F, Reisman Y, Maggi M. Relationship between low LDL cholesterol and testosterone. *J Sex Med.* 2018;15(8):1127–1134.

14. Amarenco P, Bogousslavsky J, Callahan A III, et al. High-dose atorvastatin after stroke or transient ischemic attack (SPARCL). *N Engl J Med.* 2006;355(6):549–559.

15. O’Donoghue ML, Giugliano RP, Wiviott SD, et al. Long-term evolocumab in patients with established ASCVD (FOURIER-OLE). *Circulation.* 2023;147(16):1192–1203.

16. Rasheed A, Sultan M, Tahir A, et al. Safety and efficacy of achieving very low LDL cholesterol concentrations with PCSK9 inhibitors: a meta-analysis. *J Clin Med.* 2025;14(13):4562.

17. Nissen SE, Bhatt DL. How low can you go? New evidence supports no lower bound to LDL-C level in secondary prevention [editorial]. *Circulation.* 2023;147(16):1204–1207.

* * *

*Anthony Pick, MD, CDCES, CCD is a board-certified endocrinologist, lipidologist, diabetes care specialist and adisopathy medicine physician practicing at True Health in Deerfield, Illinois. He writes about metabolic medicine, evidence-based practice, and the future of healthcare on Substack. This essay is Part 2 of a series on lipid management. Part 1, “Is Lower Longer Better?”, is available in the archive.*

On labels, complexity, and the cognitive traps that cost patients the right diagnosis Anthony Pick MD, CDCES, CCD | Multisystem Medicine

Every map is a lie. The useful question is whether it is a productive lie (simplification) — one that gets you somewhere — or a dangerous one that leaves you confidently lost.

Medicine runs on maps. We build taxonomies, diagnostic categories, and disease labels because the human mind cannot function without them. Complexity must be compressed. A clinician seeing thirty patients a day cannot hold every nuance of every pathophysiological pathway in working memory simultaneously. The label — Type 2 diabetes, say — is a cognitive scaffold. It tells you roughly where you are, what to expect, which roads are available.

But the scaffold is not the building. And when we mistake the map for the territory, patients pay.

Few areas of medicine illustrate this more clearly — or more consequentially — than diabetes. What we call “diabetes” is not one disease. It is not two diseases. It is a family of metabolic disorders united by a shared endpoint, elevated blood glucose, but driven by radically different mechanisms, following different trajectories, requiring different treatments, and carrying different implications. Collapsing them into a single label is a bit like diagnosing every chest pain as “heart trouble” — sometimes close enough, often not.

This essay practices three disciplines together, and I want to name them before I begin: deep clinical knowledge of the disease itself, careful reasoning about how clinicians arrive at — or fail to arrive at — the right diagnosis, and honest appraisal of the evidence behind the labels we use. Diabetes is one lens. The disciplines apply everywhere.

“The label is a scaffold. It gets you to the building site. It is not the building.”

* * *

## Type 1 Is Not a Moment — It Is a Trajectory

We teach Type 1 diabetes as an event: the immune system attacks the pancreatic beta cells, insulin production collapses, the patient presents in crisis. In reality, Type 1 is a slow-motion process that begins years — sometimes decades — before a single symptom appears.

TrialNet established a three-stage model, now adopted by the ADA and JDRF, that reframes Type 1 as a process rather than an event. **Stage 1:** two or more islet autoantibodies are present, glucose is normal, the patient is asymptomatic — but the immune war has already begun. **Stage 2:** autoantibodies persist and dysglycemia emerges. **Stage 3:** clinical diabetes, the point at which most patients are first diagnosed and at which most clinicians first look.

Staging changes management. A Stage 1 or Stage 2 patient is now eligible for teplizumab, the first FDA-approved therapy to delay progression to Stage 3 — by a median of two to three years in the landmark TrialNet trial. That window is not trivial. It is time for patient education, psychological preparation, and metabolic optimization before the crisis arrives.

The cognitive trap is anchoring: if we only look for Type 1 when the patient collapses into DKA, we have already missed the chance to intervene upstream. The multisystem lens asks not only where the patient is but where they are headed.

Ref: Insel RA et al. Diabetes Care. 2015;38(10):1964–1974. | Herold KC et al. NEJM. 2019;381:603–613.

↗ Coming in this series: Type 1 as Trajectory: Staging, Teplizumab, and the Window Before the Crisis.

* * *

## Type 2 Is Also a Trajectory — and Prediabetes Is Where It Begins

This is the claim that tends to provoke the most pushback, so I want to make it carefully.

Type 2 diabetes affects over 400 million people worldwide. It is real in the same way that *anemia* is real — as a final common pathway, a phenotypic endpoint reached by multiple distinct mechanisms over a long arc of time. Just as we no longer treat all anemia with iron, we should not treat all Type 2 diabetes as a single biological entity, and we should not treat its earlier stages as a separate disease.

The term *prediabetes* implies proximity to a threshold. The biology suggests something different. Microvascular damage begins in the prediabetes range. Cardiovascular risk is elevated. Small fiber neuropathy can precede a formal diabetes diagnosis by years. Prediabetes is not the anteroom to disease. It is the disease, at an earlier stage, when intervention is most effective. Reading prediabetes as Stage 1 of Type 2 — and the diabetes diagnosis itself as a much later inflection point on a continuous trajectory — changes both who gets treated and when.

The trajectory is also not uniform. In 2018, Ahlqvist and colleagues published a landmark cluster analysis using six variables — GAD antibodies, age at diagnosis, BMI, HbA1c, beta cell function, and insulin resistance — and identified five distinct subgroups within what had been labeled Type 2 diabetes. Severe insulin-resistant diabetes carried the highest risk for fatty liver disease and chronic kidney disease. Severe insulin-deficient diabetes (non-autoimmune) carried the highest risk for retinopathy. Mild age-related diabetes followed a more benign course. These are not minor nuances. They predict which organs will fail first, and they should shape which therapies are reached for first.

Biologically, the distinction between **insulin resistance-predominant** and **beta cell failure-predominant** diabetes is fundamental. Insulin resistance — driven by visceral adiposity, ectopic fat, and chronic inflammation — dominates in many patients early in the trajectory. Progressive beta cell exhaustion, driven by glucotoxicity, lipotoxicity, and polygenic predisposition, dominates in others, including many leaner patients of East and South Asian ancestry where Type 2 at normal BMI is common.

Treating these as the same disease with the same first-line algorithm is not precision medicine. It is pattern-matching with a blunt instrument.

”Type 2 diabetes” is a final common pathway, not a single disease. Treating it as one is like treating all anemia with iron.

Ref: Ahlqvist E et al. Lancet Diabetes Endocrinol. 2018;6(5):361–369.*

↗ Coming in this series: Type 2 as Trajectory: Why Prediabetes Is the Disease, Just Earlier. | The Five Faces of Type 2: Phenotyping for Precision in Cardiometabolic Medicine.

* * *

## The Diagnoses That Don’t Fit Either Trajectory

If Type 1 and Type 2 are two trajectories, a meaningful set of patients are on neither — and these are the patients most likely to be misclassified, mistreated, and sent down the wrong therapeutic road. Four categories deserve naming, even briefly.

LADA — Latent Autoimmune Diabetes in Adults.** Shares the autoimmune mechanism of Type 1 but follows a slower trajectory, typically presenting in adults and progressing to insulin dependence over months to years. Accounts for roughly 5 to 10 percent of adults diagnosed with Type 2. The diagnostic test — GAD65 antibody — is rarely ordered at the time of initial diagnosis. The clinical cost is real: these patients are commonly placed on sulfonylureas, which accelerate beta cell exhaustion in the setting of ongoing autoimmune destruction. They fail predictably. ↗ Coming: LADA: When Type 1 Hides Inside Type 2.

Type 3c — Pancreatic Diabetes. This arises from structural or functional disease of the exocrine pancreas: chronic pancreatitis, cystic fibrosis, hemochromatosis, surgical resection, and — critically — pancreatic cancer. Roughly 5 to 10 percent of all diabetes in Western populations, and almost universally misclassified as Type 2. Two clinical points cannot be overstated. First, combined insulin and glucagon deficiency makes these patients exquisitely sensitive to hypoglycemia and poorly suited to aggressive sulfonylurea regimens. Second, **new-onset diabetes in a patient over 50 with weight loss and no metabolic risk factors should raise immediate suspicion for pancreatic malignancy. Pancreatic cancer can present as diabetes before it presents as anything else. ↗ Coming: Type 3c: The Diabetes That Gets Blamed on Something Else.

Type 5 — Malnutrition-Related Diabetes. Recognized by the International Diabetes Federation in 2025 [verify date] after decades of marginalization in Western-dominated literature. Affects primarily young, lean individuals with histories of chronic malnutrition, particularly in early life. Insulin-deficient, autoantibody-negative, ketosis-resistant. Reminds us that the metabolic phenotype is shaped by the entire life history of the organism — genetics, epigenetics, early environment, nutrition. As patient populations diversify globally, the category matters clinically. ↗ Coming: Type 5: The Diabetes the Textbooks Forgot.

Monogenic diabetes (MODY). Roughly 1 to 5 percent of all diabetes, fourteen-plus subtypes described, and dramatically misclassified. MODY 2 (GCK mutations) causes mild, stable fasting hyperglycemia requiring no pharmacological treatment — patients diagnosed as Type 2 and started on metformin or insulin are being overtreated. MODY 3 (HNF1A mutations) is exquisitely sensitive to low-dose sulfonylureas — patients misdiagnosed as Type 1 and placed on insulin can often be transitioned to an oral agent with dramatically better outcomes. Flags that should prompt genetic testing: diagnosis before age 35, strong multigenerational family history, absence of obesity, absence of islet autoantibodies, stable mild fasting hyperglycemia without progression. ↗ Coming: When the Gene Is the Answer: A Clinical Guide to Monogenic Diabetes, MODY and monogenic obesity.

* * *

## When Cortisol Is the Hidden Driver: MACS

There is a category of diabetes-adjacent metabolic dysfunction that sits outside the standard taxonomy entirely — and is almost never considered at the time of diagnosis.

Mild autonomous cortisol secretion (MACS) — previously called subclinical Cushing’s syndrome — refers to low-grade, unsuppressed cortisol excess from adrenal incidentalomas, found in 20 to 50 percent of patients who undergo adrenal imaging for any reason. These patients lack the classic stigmata of overt Cushing’s: no moon face, no purple striae, no proximal myopathy. What they do have, with striking regularity, is a cardiometabolic phenotype that resists conventional management — hypertension, visceral adiposity, dyslipidemia, insulin resistance, impaired glucose tolerance, and accelerated bone loss.

Two recent trials sharpened the picture considerably. The CATALYST trial demonstrated that even modest cortisol excess — defined by failure to suppress below 1.8 mcg/dL on overnight dexamethasone suppression —was present in 23.8% of patients with difficult to control “type 2 diabetes”. The MOMENTUM trial presented at ACC 2026 showed a surprisingly high incidence of cortisol autonomy in resistant hypertension (27.3%).

The clinical implication: a patient labeled as “difficult-to-control Type 2 diabetes” with resistant hypertension and central obesity may have an adrenal incidentaloma driving the phenotype. Without looking upstream at the hormonal architecture, the label sticks, the medications accumulate, and the underlying driver goes untreated.

↗ Coming in this series: The Cortisol You’re Missing: Adrenal Incidentalomas, MACS, and the Cardiometabolic Consequences Nobody Is Treating. | Reading CATALYST and Momentum: A Worked Example in Critical Appraisal.

* * *

## Why These Errors Are Predictable

A pattern runs beneath every misclassification above, and it is not a pattern of insufficient knowledge. It is a pattern of how clinicians think under load.

Anchoring. The first diagnosis sticks. Once the chart reads “Type 2 diabetes”, subsequent clinicians read through that frame. Disconfirming evidence — brittle control, unexpected weight loss, absence of metabolic risk factors — is processed as noise rather than signal. I have tragically seen this anchoring lead to a fatal and potentially avoidable outcome (see upcoming essay on type 3c diabetes).

Availability. Common diagnoses are mentally available; rarer ones require deliberate effort to access. LADA, Type 3c, MODY, and MACS are all common enough to matter clinically but rare enough to fall outside automatic pattern-matching.

Premature closure. The diagnosis is made and the search stops. The patient has diabetes. Which diabetes — and driven by which mechanism — is never asked.

The antidote is not more protocols. Protocols can encode the same cognitive errors at scale. The antidote is a disciplined practice of differential diagnosis applied not just to symptoms but to mechanism — asking not only what the patient has, but why they have it, and whether the label they carry is actually earning its place in their chart.

↗ Coming as a parallel series: **How Doctors Think (and Where They Go Wrong): A Field Guide to Cognitive Bias, Diagnostic Error, Translation of Guidelines and Clinical Decision Support.

* * *

## What This Series Is For

I want to be direct about why I am writing this — and why the diabetes taxonomy is not an academic exercise.

The way medicine is currently organized — by organ, by specialty, by billing code — produces predictable blind spots. The endocrinologist sees the glucose. The cardiologist sees the coronary artery. The gastroenterologist sees the pancreas. No one is structurally required to see the whole system. And so the patient with LADA misclassified as Type 2, failing sulfonylurea after sulfonylurea, cycles through the system without anyone asking the right question. Further diabetes is much more than is reflected in a “glucocentric” view of the disease syndromes. This deeper understanding of the multiple pathophysiology pathways of diabetes is another thread of essays to look out for.

Multisystem Medicine is my name for the discipline that closes that gap. It is not a new specialty in the bureaucratic sense. It is a way of thinking — one that takes seriously the biological reality that human systems are integrated, that metabolism is a network, and that the presenting complaint is rarely the whole story. This essay practices three of the disciplines that constitute it: clinical depth, careful reasoning, and honest appraisal of evidence. There are others. A future essay will name them all and show how they are woven together.

The success criterion is not knowledge, and not guidelines. Guidelines are inert until implemented. A diagnosis is a hypothesis until it changes a treatment. The practice I am building treats translation — the conversion of knowledge into outcomes for the patient in front of you — as the only criterion that matters. Everything else is preamble.

The diabetes taxonomy is the first illustration. The same logic applies across every domain where a reductionist label is doing incomplete work: bone and muscle, hormonal architecture, women’s health, longevity, cardiometabolic medicine. Weaving complexity into clarity — one system at a time, one essay at a time.

“The endocrine system is the operating system of human metabolism. I am trained to read it whole.”

* * *

If this essay resonated, stay tuned for Multisystem Medicine for essays at the intersection of endocrinology, cardiometabolic medicine, and systems thinking.

* * *

Key References

1. Insel RA et al. Staging presymptomatic type 1 diabetes. Diabetes Care. 2015;38(10):1964–1974.

2. Herold KC et al. Teplizumab in relatives at risk for type 1 diabetes. NEJM. 2019;381:603–613.

3. Ahlqvist E et al. Novel subgroups of adult-onset diabetes. *Lancet Diabetes Endocrinol.* 2018;6(5):361–369.

4. Laugesen E et al. Latent autoimmune diabetes of the adult. Diabetic Medicine 2015; 32 (7), 843-852

5. Hart PA et al. Type 3c diabetes mellitus. Lancet Gastroenterol Hepatol. 2016;1(3):226–237.

6. Misra S et al. Malnutrition-related diabetes. Lancet Diabetes Endocrinol. 2022;10(9):675–685.

7. Hattersley AT, Patel KA. Precision diabetes: learning from monogenic diabetes. Diabetologia. 2017;60(5):769–777.

8. CATALYST trial — A phase 4 study of hypercortisolism in patients with difficult-to-control diabetes. American Heart Journal. Vol 267, January 2024, 134-135

9. MOMENMTUM trial — presented at ACC 2026 showing 27.3% prevalence of hypercortisolism in resistant hypertension.

On very low LDL, the myth of “too low,” and what happens to patients when specialists can’t agree on the evidence *Optimized Medicine | Anthony Pick MD, CDCES, CCD

There are moments in clinical medicine that stay with you not because something went wrong, but because something nearly did — and you caught it at the last minute.

A patient of mine, a man in his early fifties with metabolic syndrome, a coronary calcium score approaching 3,000, and the kind of vascular age that belongs in someone two decades older, came to a follow-up appointment confused. He had seen his cardiologist the week before. He was doing well. His LDL was 14 mg/dL — the result of years of carefully titrated therapy, a PCSK9 inhibitor added to his statin and ezetimibe because the evidence demanded it and the risk was undeniable. He had no symptoms. No side effects. The plaques were not growing.

But someone had told him his LDL was too low.

He wasn’t sure whether to keep taking his medications. He didn’t know who to believe. He came in holding his lab slip like a question he didn’t know how to ask.

I have thought about that encounter many times. Not just because of what it meant for his cardiovascular risk — though an LDL of 14 versus 70 mg/dL in a man with his risk burden is not a rounding error — but because of what it revealed about how medicine sometimes works against itself.

—

## What the Evidence Actually Says

Let me be direct about the science, because the clinical folklore about “too low” LDL has a stubborn life that is not supported by data.

The Cholesterol Treatment Trialists’ meta-analysis, which pooled data from 26 randomized trials and over 170,000 participants, established the foundational principle clearly: each 1 mmol/L reduction in LDL reduces major cardiovascular events by approximately 22%, proportionally, regardless of baseline level. There is no plateau. There is no floor below which the benefit disappears.¹²

IMPROVE-IT extended this to ezetimibe — adding it to a statin in post-ACS patients reduced LDL further and reduced events further, with no safety signal.² FOURIER did the same for evolocumab, a PCSK9 inhibitor, achieving median LDL of 30 mg/dL with significant reductions in cardiovascular death, MI, and stroke.³ ODYSSEY OUTCOMES confirmed the findings with alirocumab in a high-risk post-ACS population, with some patients achieving LDL levels below 15 mg/dL.⁴

Then came the long-term data. FOURIER-OLE followed patients for nearly nine years. Those who achieved the lowest LDL levels — including sustained levels below 20 mg/dL — had the best outcomes. Not worse. Not neutral. Better. The curves kept separating. This is the “lower longer” principle operationalized in a clinical trial: duration matters, magnitude matters, and the combination of both matters most.⁵

The new 2026 ACC/AHA Dyslipidemia Guidelines — the first comprehensive update in eight years, published in March 2026 — codified this unambiguously. For very high-risk patients, an LDL goal of less than 55 mg/dL is now a Class I recommendation. For select patients at extreme risk, the guidelines endorse pushing further. The language is deliberate: lower for longer provides greater protection.⁶

—

## The Myth of “Too Low”

The concern about very low LDL causing harm — particularly cognitive harm — has been circulating in clinical hallways for years. It deserves a direct response, because the data is not ambiguous.

The EBBINGHAUS trial was designed specifically to evaluate cognition in patients on evolocumab versus placebo. There was no difference in any cognitive outcome. None.⁷ The open-label extension, EBBINGHAUS-OLE, followed patients with sustained very low LDL and reached the same conclusion.⁸

Mendelian randomization studies — which use naturally occurring genetic variants to simulate lifelong exposure to lower LDL — have not found an association between low LDL and dementia or cognitive impairment.⁹ A 2025 meta-analysis of PCSK9 inhibitor trials confirmed no excess risk for neurocognitive events, diabetes, or cancer in patients achieving very low LDL levels.¹⁰

The fear of “too low” LDL is not grounded in evidence. It is a clinical reflex — understandable given how counterintuitive it can feel to push aggressively against a number that is already very low — but it is a reflex, not a data-driven conclusion. When that reflex results in de-intensifying therapy in a patient with a CAC score of 3,000 and an LDL of 14, it is not caution. It is harm.

—

## The Professional Communication Problem

I want to be careful here, because I am not writing to impugn any individual clinician. Cardiology has extraordinary clinicians. The interventionalists who perform revascularization, the electrophysiologists who manage arrhythmias, the heart failure specialists who manage the most complex patients in medicine — these are colleagues I respect deeply.

But something happens in the space between specialists that is rarely discussed: when a clinician unilaterally modifies a care plan built by another specialist, without communicating, without documentation visible to the patient’s full care team, and often based on concerns that the evidence does not support — the patient bears the consequence.

In the case I described, the patient did not know who was right. He was not equipped to evaluate competing claims about LDL safety from clinicians he trusted. He was confused, and that confusion was entirely a product of a system failure, not a knowledge deficit on his part. He was doing everything right. The system let him down.

I am a board-certified lipidologist. Lipid management is a board-certified subspecialty. When a specialist who is not trained in lipidology changes lipid therapy I have carefully structured — without communicating, without raising the question in a shared note, without even a message to the patient’s care team — and when that change is based on a concern the evidence does not support, it crosses a line. Not a personal line. A professional one. It is a failure of collaborative care that creates clinical risk.

That is a culture problem.

—

## What Collaborative Care Actually Requires

Collaboration is not a soft skill. It is a clinical competency, and in complex patients it is a patient safety issue.

Genuine collaborative care requires several things that are largely absent from how medicine currently operates. It requires communication before changing another specialist’s management, not after or never. It requires shared documentation that surfaces disagreements to the care team rather than obscuring them from patients and colleagues. It requires a shared commitment to the evidence base — not consensus for its own sake, but a willingness to update practice when the data changes.

It also requires intellectual humility. The clinician who is not a lipidologist expressing concern about LDL levels below 20 should first ask: what does the evidence show? Not what does it feel like intuitively, not what did a senior colleague say during training fifteen years ago, but what does the peer-reviewed evidence published in this decade show?

When I raised this question in a large academic health system — when I tried to build the infrastructure for coordinated, evidence-based lipid and cardiometabolic management at scale — I encountered what I now recognize as a familiar pattern: institutional inertia, hierarchical resistance, and the deeply human tendency to defend existing practice against incoming evidence. It is not unique to any person or institution. It is a feature of how large systems preserve their own stability, often at the cost of the patients they serve.

—

## A Note to Patients

If you are managing cardiovascular risk and you have received conflicting guidance from different clinicians about your LDL — whether to push it lower, whether a very low level is dangerous, whether to stop or reduce a medication because your LDL is “too low” — please know the following:

The evidence does not support the concept of “too low” LDL in high-risk patients. Clinical trials with nearly a decade of follow-up have found that lower LDL levels, sustained over time, produce better cardiovascular outcomes without meaningful increases in cognitive impairment, cancer, or other harms. The 2026 ACC/AHA Guidelines have codified this.

Ask your clinician where the concern comes from. Ask them to point to the evidence. You are entitled to that conversation, and a good clinician will welcome it.

If you are receiving conflicting guidance from multiple clinicians and no one is reconciling it, that is not a personal failing. It is a gap in how your care is coordinated. You deserve a care team that talks to each other.

—

## A Closing Thought for Clinicians

The most important line in the new ACC/AHA guidelines may not be about a specific LDL target. It may be the implicit premise behind the entire document: that we now have the tools, the evidence, and the pharmacology to dramatically reduce cardiovascular events in the highest-risk patients — if we use them consistently, if we communicate across specialties, and if we keep the evidence rather than clinical mythology at the center of the conversation.

The question is not whether lower is better. The evidence on that is settled. The question is whether we are organized — professionally and culturally — to act on what we know.

—

## References

1. Cholesterol Treatment Trialists’ Collaboration. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. *Lancet.* 2010;376(9753):1670–1681.

2. Cannon CP, Blazing MA, Giugliano RP, et al. Ezetimibe added to statin therapy after acute coronary syndromes (IMPROVE-IT). *N Engl J Med.* 2015;372(25):2387–2397.

3. Sabatine MS, Giugliano RP, Keech AC, et al. Evolocumab and clinical outcomes in patients with cardiovascular disease (FOURIER). *N Engl J Med.* 2017;376(18):1713–1722.

4. Schwartz GG, Steg PG, Szarek M, et al. Alirocumab and cardiovascular outcomes after acute coronary syndrome (ODYSSEY OUTCOMES). *N Engl J Med.* 2018;379(22):2097–2107.

5. O’Donoghue ML, Giugliano RP, Wiviott SD, et al. Long-term evolocumab in patients with established atherosclerotic cardiovascular disease (FOURIER-OLE). *Circulation.* 2023;147(16):1192–1203.

6. Blumenthal RS, Morris PB, et al. 2026 ACC/AHA Guideline on the Management of Dyslipidemia. *J Am Coll Cardiol.* Published online March 13, 2026. doi:10.1016/j.jacc.2025.11.016

7. Giugliano RP, Mach F, Zavitz K, et al. Cognitive function in a randomized trial of evolocumab (EBBINGHAUS). *N Engl J Med.* 2017;377(7):633–643.

8. Giugliano RP, Keech AC, Murphy SA, et al. Long-term cognitive safety of achieving very low LDL cholesterol with evolocumab (EBBINGHAUS-OLE). *NEJM Evidence.* 2025. doi:10.1056/EVIDoa2400112

9. Mefford MT, Rosenson RS, Cushman M, et al. PCSK9 variants, LDL-cholesterol, and neurocognitive impairment: The REGARDS Study. *Circulation.* 2018;137(12):1260–1269.

10. Rasheed A, Sultan M, Tahir A, et al. Safety and efficacy of achieving very low LDL cholesterol concentrations with PCSK9 inhibitors: a meta-analysis. *J Clin Med.* 2025;14(13):4562.

11. Johnson AE, Swabe GM, Bress AP, et al. Real-world prescribing in accordance with ACC/AHA guidelines for lipid-lowering therapy in high-risk primary and secondary prevention of ASCVD. *Am J Prev Cardiol.* 2025.

12. Nissen SE, Bhatt DL. How low can you go? New evidence supports no lower bound to LDL-C level in secondary prevention [editorial]. *Circulation.* 2023;147(16):1204–1207.

—

*Anthony Pick, MD, CDCES, CCD is a board-certified endocrinologist, obesity medicine specialist, lipidologist, and cardiometabolic lifestyle medicine specialist practicing at True Health (www.truehealth.co) in Deerfield, Illinois, where he leads cardiometabolic and endocrine services. He writes about metabolic medicine, evidence-based practice, and healthcare delivery at Optimized Medicine on Substack.

Optimized Medicine is the pursuit of your best possible biology — through rigorous evidence, systems thinking, and a clinical team designed to make improvement sustainable.

The Problem True Health Exists to Solve

A patient arrived recently with a folder with multiple test results from multiple clinics and no other available medical records, such as physician notes. She has been experiencing chronic fatigue and brain fog that started after a Covid-19 infection. She was concerned about a hormonal imbalance and had been told she has “adrenal fatigue” but NOT confirmed hypoadrenalism. She had purchased a costly cocktail of “adrenal support” supplements.

Her question was simple.

“Why do I still feel this fatigue and brain fog?”

She had not been ignored by the medical system. She had been processed by it.

Everything had been measured. Almost nothing had been understood.

That folder is a map of modern medicine’s integrity deficit — not because clinicians lack intelligence or commitment, but because the structures surrounding patient care increasingly prevent careful reasoning about complex biology.

The Structural Problem

Patients navigating chronic cardiometabolic disease today typically encounter one of three medical ecosystems: the health system, the wellness and functional medicine industry, and the longevity influencer ecosystem.

Each emerged partly in response to the shortcomings of the others. Each contains clinicians acting with genuine integrity. But each also contains structural incentives that often undermine careful clinical reasoning.

The goal here is not to dismiss these models wholesale. It is to examine where they fail — and why a different clinical operating system may be necessary for complex chronic disease.

🏥 Model One: The Health System

This critique is not directed at individual physicians. Many clinicians within large institutions deliver exceptional care under difficult constraints.

The problem is structural.

Large healthcare organizations operate within reimbursement models that reward procedural volume, brief encounters, and departmental revenue streams. These incentives make careful longitudinal reasoning about complex chronic disease difficult to sustain.

When clinicians are responsible for large patient panels within short visits, even excellent physicians find themselves treating individual problems rather than integrating the biological system as a whole.

A patient with metabolic syndrome, insulin resistance, early cardiovascular disease, and sleep-disordered breathing will often be referred to endocrinology, cardiology, pulmonology, and perhaps a dietitian — none of whom communicate meaningfully with each other.

Each specialist treats their organ. Nobody treats the patient.

Access compounds the problem. Meaningful appointments frequently require weeks to months, culminating in encounters that cannot accommodate the complexity of chronic disease.

The result is not negligence. It is fragmentation — systemic, structural, and largely invisible to the patients who experience it.

🌿 Model Two: Wellness and Functional Medicine

The functional medicine movement emerged partly as a legitimate response to the limitations of conventional care. It asks questions conventional medicine sometimes neglects: What is driving this condition? What changed in this patient’s biology? What does the full biological picture look like?

These are important questions. Many practitioners who ask them are genuinely motivated by patient wellbeing, and some practice with real rigor and intellectual integrity.

But the core issue is epistemological.

A practitioner can be entirely well-intentioned and still cause harm when interventions are not grounded in validated clinical evidence. Good intentions do not validate a test or justify an intervention. The same evidence standard that applies to pharmaceuticals must apply to functional protocols, specialty panels, and botanical therapies.

Patterns that warrant scrutiny include:

• Laboratory panels without established clinical utility, ordered because they generate intervention indications rather than because they change management

• Supplement protocols lacking controlled clinical trial evidence, particularly when sold by the same practitioner who prescribes them

• Simplified “root cause” narratives applied to complex multisystem disease — single villains for conditions that require systems thinking

• Diagnostic labels absent from peer-reviewed literature applied to genuine symptoms that deserve rigorous investigation

• Business models dependent on ongoing patient dependency rather than resolution

Some clinicians within functional medicine operate with genuine rigor: validated testing, acknowledged uncertainty, transparent financial relationships, a willingness to refer to conventional specialists. That practitioner is practicing with integrity regardless of the label on their door.

The criticism here is of the commercial ecosystem — not every practitioner within it.

📈 Model Three: The Longevity Influencer Ecosystem

A third model has emerged in the past decade: physician-influencers promoting longevity optimization to large audiences.

Some contributions have been genuinely valuable — particularly the emphasis on metabolic health, earlier cardiovascular risk assessment, and exercise as medicine. These ideas have real clinical merit and have moved mainstream medicine in useful directions.

However, the ecosystem carries its own structural weaknesses.

Several prominent figures have promoted therapies with minimal human evidence — for example, rapamycin use in healthy adults, where most data derive from animal models or disease states, and the risk-benefit profile in healthy humans remains uncertain in peer-reviewed clinical trials. Others have recommended products tied to undisclosed financial relationships or presented speculative extrapolation as established science.

The intellectual framework has merit. The credibility of some of its most visible advocates does not always match it. Patients who built health protocols around these voices deserve better than discovering that the physician-influencer model has its own integrity deficit.

⚖️ The Landscape in Full

✅ Medicine with Integrity

The real distinction is not between conventional and functional medicine. It is between medicine practiced with intellectual honesty and medicine practiced for other ends.

🧠 The Clinical Framework: Systems Biology

Three principles guide the clinical approach.

1. The body is a complex biological system — not a collection of organs, and not a machine awaiting the correct protocol.

2. Clinical decisions follow probability and evidence, not narrative convenience.

3. Sustained health improvement requires patient agency supported by coordinated team infrastructure.

Systems Thinking Over Organ Thinking

Chronic cardiometabolic disease rarely arises from a single cause. Insulin resistance may reflect interacting drivers: visceral adiposity, sleep disruption, chronic stress physiology, sedentary behavior, altered gut biome, and genetic susceptibility. In most patients, several coexist.

The clinical task is not to identify a single villain. It is to determine which drivers matter most for this patient at this moment — and which are modifiable given their biology, circumstances, and resources. That is harder than promising a root cause. It is more honest. And it is far more likely to produce durable results.

Root Cause Analysis as a Clinical Discipline

Understanding accumulates through time, continuity, and repeated observation. Clinical reasoning becomes an iterative process: construct a hypothesis from history and selective testing → implement a targeted intervention → observe the biological response → update the model.

Upstream Mapping — History focuses on physiological turning points: when did function change, and what shifted around it? This reconstructs a biological timeline rather than a diagnostic list.

Hypothesis-Driven Testing — Tests answer specific questions: Does this result alter management? Does it improve risk prediction? If not, it adds noise. More data does not automatically produce more insight.

Circuit-Level Assessment — Metabolic, endocrine, cardiovascular, and inflammatory systems are mapped as an interacting whole, not evaluated in isolation by specialists who never communicate.

Iterative Modeling — The model is continuously updated as interventions produce — or fail to produce — expected responses.

Individual Biology, Not Population Averages

Guidelines describe population averages. Patients are not averages.

Biological variation in drug metabolism, cardiovascular risk, insulin sensitivity, inflammatory response, and body composition frequently changes clinical decisions: pharmacogenomic analysis before initiating medications with known metabolizer variation; DXA body composition rather than BMI as a proxy for metabolic risk; continuous glucose monitoring to characterize metabolic phenotype before prescribing; advanced lipoprotein analysis when standard panels are insufficient for risk stratification.

🏃 Lifestyle as Primary Biology

Lifestyle medicine is not adjunctive. It is primary biology — the foundation upon which everything else operates. Diagnostics and medications operate on top of that foundation, not instead of it.

Cardiorespiratory fitness is among the strongest independent predictors of all-cause mortality across multiple longitudinal cohorts.¹

The Implementation Constraint

Even when a diagnosis is correct and treatment recommendations are sound, outcomes frequently fail to improve.

The limiting factor is often not knowledge. It is execution.

Large trials — including the Diabetes Prevention Program and Look AHEAD — demonstrate that intensive lifestyle intervention significantly improves cardiometabolic outcomes.²⁻³ Yet sustaining behavioral change outside structured support remains the central challenge of chronic disease care.

Chronic disease management depends on behaviors repeated daily over years: sleep timing, physical activity, nutrition, medication adherence, and stress regulation. These behaviors are shaped by more than motivation — they are influenced by work schedules, family obligations, economic constraints, psychological health, and environmental conditions.

Clinical reasoning alone does not change physiology. Human behavior does.

This is why clinical infrastructure matters. The physician identifies the biological drivers. The team translates recommendations into daily action.

Even with strong support, not every patient succeeds. Biology interacts with circumstance. Honest medicine acknowledges this limitation while building the strongest possible environment for success.

👥 The Team: Clinical Infrastructure, Not Support Staff

Systems-based care cannot rely on a single clinician. Multidisciplinary team-based care has repeatedly demonstrated improved glycemic control, cardiovascular risk factor reduction, and adherence compared with physician-only models.⁴⁻⁵

True Health operates as an integrated clinical team in which each member functions at the top of their training within a shared philosophy of care.

This structure is not an amenity. It is a clinical infrastructure.

Without it, many recommendations remain theoretical.

The physician synthesizes. The team implements. The therapeutic relationship sustains it. A clinical recommendation is only as effective as the patient’s capacity to act on it — and the team’s capacity to support that action over time.

Specialty Partnerships: Plug-and-Play Integration

Some clinical domains require dedicated specialty expertise that cannot be fully housed within a single practice. The model here is deliberate plug-and-play integration — pre-built working relationships with concierge-level specialty practices that provide superior access and genuine clinical coordination, not referral into the health system queue.

This is not a referral network. It is a curated ecosystem of aligned specialty partners — each selected for clinical quality, access, and philosophical compatibility with an integrated, patient-centered model of care.

🛌 Sleep Medicine — Sleep-disordered breathing, circadian dysfunction, and chronic insomnia are among the most modifiable upstream drivers of insulin resistance, cardiovascular risk, and metabolic dysfunction — yet they are systematically underdiagnosed and undertreated in standard cardiometabolic care. This practice maintains a working relationship with a dedicated sleep medicine specialty practice, providing patients with access that bypasses the weeks-to-months wait typical of health system referrals.

For chronic insomnia specifically, Cognitive Behavioral Therapy for Insomnia (CBTi) is the evidence-based first-line treatment with stronger long-term outcomes than pharmacotherapy and no dependency risk.⁶ This practice uses DrLullaby (drlullaby.com) as its digital CBTi delivery layer: a clinician-delivered platform backed by the University of Chicago, combining live video visits with secure messaging and wearable integration (including Oura) for data-informed insomnia treatment. This is not a self-help app — it is structured clinical care.

🧠 Brain Health — Cognitive function, neurological resilience, and the intersection of metabolic disease with brain health are increasingly recognized as core components of healthspan — not separate concerns. A working relationship with a concierge-level brain health specialty practice extends the cardiometabolic framework into cognitive longevity: early identification of risk, metabolic contributors to cognitive decline, and integrated management across the metabolic-neurological axis.

🧬 Clinical Genetics — Genetic architecture shapes cardiovascular risk, drug metabolism, nutrient utilization, and disease susceptibility in ways that population-average guidelines cannot capture. A relationship with a dedicated clinical genetics practice allows genomic data to be integrated meaningfully into clinical decision-making — not as a consumer wellness product, but as a tool for genuine risk stratification and personalized treatment planning.

🤝 The Therapeutic Relationship

Technology, testing, tracking metrics and outcomes, and team structure are necessary. They are not sufficient.

The therapeutic relationship — the quality of trust, communication, and genuine investment between clinician and patient — is itself a determinant of clinical outcomes. It shapes whether patients share what matters, whether they act on recommendations, and whether they return when things aren’t working. In this practice, that means the physician knows the patient’s biology, history, and life context — not just their problem list. Recommendations come with reasoning. Uncertainty is named rather than papered over. The patient is treated as an intelligent adult capable of understanding their own biology.

This is not a soft consideration adjacent to real medicine. It is the connective tissue that holds the operating system together.

📊 Measuring What Matters

Clinical philosophy is meaningless without outcomes.

Relevant measures in this practice include: glycemic control and insulin sensitivity improvement; visceral adiposity reduction — measured directly by DXA, not estimated; cardiometabolic risk marker trajectories across lipoproteins, inflammatory markers, and blood pressure; functional capacity as longevity predictors; appropriate medication reduction where biological improvement warrants it; and sustained behavioral adherence over months and years, not weeks.

Longitudinal tracking of these metrics allows clinical reasoning to be tested rather than merely asserted. The model is accountable to the data it produces.

🔧 The Operational Difference: Three Layers of Optimization

The knowledge required to treat cardiometabolic disease already exists in preventive cardiology, endocrinology, and lifestyle medicine. This practice is not built on a novel theory of disease.

The distinguishing feature is the operational environment — and it operates on three distinct, mutually reinforcing layers.

Layer One — Clinical Optimization: Systems biology, root cause analysis, lifestyle as primary biology, individualized evidence-based decision making. The what of care: reasoning carefully about complex biology and treating the whole patient rather than isolated diagnoses.

Layer Two — Operational Optimization: Access design, team architecture, specialty partnerships, care delivery mechanics, and outcome tracking. The how of care: the infrastructure that determines whether clinical knowledge actually reaches the patient consistently, repeatedly, and over the years required for biological change.

Layer Three — Meta-Optimization: Continuous innovation in how the practice itself learns and improves. Systematic outcome tracking creates a feedback loop — clinical decisions inform measurable results, results inform future decisions, and the model evolves. The practice is not a static protocol. It is a learning system designed to get better over time.

• Remote Physiological Monitoring (RPM) and Chronic Care Management (CCM): Structured programs that extend clinical touch between visits, capture real-world physiological data continuously, and generate reimbursable care pathways aligned with CMS evidence standards. RPM transforms wearable and device data from passive tracking into actionable clinical intelligence — flagging trends, triggering outreach, and enabling intervention before problems escalate.

• Integrated technology stack: Purpose-built to connect the care team, patient-facing tools, and physiological data streams into a single coherent clinical workflow — reducing friction, improving continuity, and making the team model operationally scalable.

• Outcome data infrastructure: Longitudinal tracking of cardiometabolic biomarkers, functional capacity, behavioral adherence, and patient-reported outcomes — not as administrative documentation but as a clinical learning system that informs future decisions.

The digital layer is not a feature. It is the nervous system of the operating system.

🏛️ The Honest Case for Concierge Medicine

This practice operates within a membership model. The criticism of concierge medicine — that it reinforces a two-tiered healthcare system — is legitimate and deserves a direct answer.

Many physicians have concluded that careful, longitudinal, reasoning-intensive medicine is structurally impossible inside conventional health system incentives. Membership practices attempt to solve that problem by preserving two resources modern healthcare chronically rations: time and continuity.

The ethical question is not whether concierge medicine exists. It is how it is practiced.

Beyond individual practice, membership models may function as prototype environments — places where integrated, outcomes-focused care can be tested, refined, and eventually translated into approaches larger systems could adopt. The ethical obligation of a practice like this is not only to serve its members well. It is to demonstrate that the model works.

🤝 Who Thrives in This Model

⭐ The North Star

The human body is not a machine awaiting the correct supplement, peptide, infusion, or food restriction protocol. It is a complex biological system shaped by genetics, behavior, environment, and time — one that responds to thoughtful, coordinated, evidence-grounded care in ways no supplement line or influencer protocol can replicate.

Modern medicine already possesses much of the knowledge required to address chronic disease. What it often lacks are environments that allow careful reasoning, appropriate pharmaceutical prescribing based on best evidence, coordinated care, and sustained behavioral implementation over the years required for biological change.

Good medicine is not the performance of manufactured certainty. It is the disciplined, intellectually honest pursuit of understanding — delivered by a team that shares that commitment, sustained across the years required for it to matter.

Optimized Medicine is not a brand. It is not a doctrine.

It is a clinical operating system for practicing medicine thoughtfully, transparently, and in genuine partnership with patients over the years required for health to change.

Anthony Pick— MD, CDCES, CCD · Endocrinology & Cardiometabolic Medicine · True Health · Deerfield, Illinois

References

1. Blair SN et al. Physical fitness and all-cause mortality. JAMA. 1989;262(17):2395–2401.

2. Knowler WC et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. NEJM. 2002;346(6):393–403.

3. Look AHEAD Research Group. Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes. NEJM. 2013;369(2):145–154.

4. Wagner EH et al. Improving chronic illness care: translating evidence into action. Health Affairs. 2001;20(6):64–78.

5. Tricco AC et al. Effectiveness of quality improvement strategies on the management of diabetes. BMJ. 2012;344:e1802.

6. Qaseem A et al. Management of chronic insomnia disorder in adults: a clinical practice guideline from the American College of Physicians. Ann Intern Med. 2016;165(2):125–133.

Why Weight Loss Alone Is the Wrong Target — and Why Body Composition Matters More

A patient loses 40 pounds on a GLP-1 medication. Everyone celebrates. Her blood pressure improves. Her A1c falls. The scale confirms progress.

Then a body composition test adds context. A meaningful portion of that loss came from lean mass. Her grip strength has declined. Her waist circumference remains elevated.

She is lighter. She is not necessarily metabolically stronger.

This is not a failure of treatment.

It is a failure of definition.

Weight Is a Number. Health Is Physiology.

Body weight is a crude aggregate. It does not distinguish between skeletal muscle and visceral fat, subcutaneous fat and ectopic fat, bone and glycogen, or bone and water.

Adipose tissue performs essential physiologic roles: energy buffering, endocrine signalling (leptin, adiponectin), thermal insulation, and mechanical protection. Healthy subcutaneous expansion can be protective. Subcutaneous adipose tissue acts as a metabolic buffer, delaying lipotoxicity when energy intake exceeds expenditure.

When storage capacity is exceeded, ectopic and visceral fat deposition disrupts insulin signalling, promotes inflammation, and drives cardiometabolic disease.

Muscle Determines Metabolic Capacity

Skeletal muscle is the primary site of insulin-mediated glucose disposal and a determinant of resting metabolic rate, strength, and functional independence.

Lean mass losses during caloric restriction are variable and context-dependent but may account for approximately one-quarter to nearly 40% of total weight lost, depending on dietary composition, resistance stimulus, and baseline adiposity.

Strength, particularly grip strength, predicts mortality more reliably than BMI.

In older adults, unmitigated lean mass loss accelerates sarcopenia (loss of muscle mass and function), frailty, and fall risk.

Body Composition in Major Incretin Trials

In STEP-1 (semaglutide 2.4 mg), mean weight loss approached 15%. DXA substudies demonstrated that ~39–40% of the total weight loss was fat-free mass. Participants lost roughly 6–7 kg of fat mass and 4–5 kg of fat-free mass.

In SURMOUNT-1 (tirzepatide), weight reductions ranged from 15–22% depending on dose. Lean mass accounted for approximately 25–30% of total weight loss, with the majority attributable to fat mass reduction.

In SELECT, semaglutide reduced major adverse cardiovascular events by 20% in adults with obesity without diabetes.

Importantly, functional performance outcomes were not systematically impaired in trial populations, suggesting that absolute fat loss outweighed lean mass reductions at a population level. The clinical concern arises at the individual level, particularly in older adults, sarcopenic phenotypes, or patients not engaging in resistance training.

Measuring What Predicts Outcomes

Before advanced imaging, simple tools remain powerful:
• Waist circumference
• Grip strength
• Chair rise testing
• Gait speed

These reflect functional and metabolic resilience more directly than body weight alone.

DXA: Clinical and Technical Nuance

Dual-energy X-ray absorptiometry (DXA) is the most accessible clinical reference method for regional body composition assessment.

It allows quantification of total fat mass, appendicular lean mass, regional fat distribution, estimated visceral adipose tissue, and bone mineral density.

Appendicular lean mass index is central to the diagnosis of sarcopenia. DXA-derived estimates of visceral adipose tissue correlate reasonably with CT-based measurements, though CT and MRI remain the gold standards for precise compartmental quantification in research settings.

DXA uniquely provides bone mineral density assessment, critical when weight reduction may accelerate bone loss, particularly in postmenopausal women and older adults.

The goal of DXA is not imaging for its own sake, but alignment of therapy with measurable tissue-level targets.

Clinical Strategy

The objective is not weight reduction alone.

It is reduction of dysfunctional adiposity (adiposopathy) while preserving skeletal muscle and metabolic reserve.

This requires:
• Resistance training
• Protein intake of ~1.2–1.6 g/kg/day
• Functional monitoring alongside pharmacotherapy

Weight loss is visible.
Muscle preservation is intentional.
Metabolic resilience is the real outcome.

As obesity pharmacotherapy advances, the standard of care must keep pace. The next phase of metabolic medicine is not choosing between weight loss and strength: it is integrating both. When we redefine success beyond the scale, we stop shrinking patients and start strengthening them.

Key References (PubMed Linked)

1. Després JP. Circulation. 2012. PMID: 22949540

2. NCEP ATP III. Circulation. 2002. PMID: 12485966

3. Heymsfield SB. Obes Rev. 2014. PMID: 24447715

4. Weinheimer EM. Nutr Rev. 2010. PMID: 20591106

5. Celis-Morales CA. BMJ. 2018. PMID: 29739772

6. Cruz-Jentoft AJ. Age Ageing. 2019. PMID: 30312372

7. Wilding JPH (STEP-1). NEJM. 2021. PMID: 33567185

8. Jastreboff AM (SURMOUNT-1). NEJM. 2022. PMID: 35658024

9. Lincoff AM (SELECT). NEJM. 2023. PMID: 37952131

After years of concern about endocrine care access there is an uncomfortable but unavoidable conclusion: the system is working exactly as it is designed to work.

Note: This publication is no longer active.
All new essays are published here → [threadmedicine.substack.com]

Subscribe there to receive future work.

Thanks!!

Prolonged wait times, oversized panels, and chronic backlog in endocrinology are not isolated operational failures. They are the predictable result of fixed workforce capacity colliding with steadily rising demand for chronic disease care.

This is not a temporary disruption. It is a structural condition.

📊 Workforce Capacity Has Lagged Demand for Decades

Workforce analyses have warned for years that the supply of endocrinologists would not keep pace with population needs. The Endocrine Society’s commissioned analyses suggest there are approximately 5,000–8,000 practicing clinical endocrinologists in the United States for a population of more than 330 million people with rising or sustained high prevalence of diabetes, obesity, thyroid disease, osteoporosis, PCOS, metabolic liver disease, and hypogonadism (Endocrine Society, 2014; Vigersky et al., 2008).

The Endocrine Society has repeatedly highlighted a persistent national shortage and projected that demand for endocrine services would continue to grow faster than training output due to:

• Slow fellowship expansion
• An aging specialty workforce
• Increasing chronic disease burden
• Lower compensation relative to procedural specialties

Earlier analyses in the Journal of Clinical Endocrinology & Metabolism described widespread access constraints—including multi-month wait times and practices closed to new patients—more than a decade ago (Vigersky et al., 2014). The pattern has not meaningfully improved since.

Despite modest growth in board-certified endocrinologists, access remains constrained because demand has expanded faster than capacity.

💰 The Cardiology Contrast Clarifies the Structural Issue

The difference becomes clearer when contrasted with cardiology.

There are roughly 30,000–35,000 practicing cardiologists in the United States—five to six times the size of the endocrine workforce—serving the same population. Cardiology access problems certainly exist, but median wait times are typically shorter and capacity expansion occurs more rapidly.

Procedural specialties generate downstream revenue through imaging, catheterization, and hospital services. Cognitive specialties like endocrinology generate most of their value through longitudinal decision-making, medication management, and risk reduction—activities that historically receive lower reimbursement (MedPAC, 2023).

When capacity expansion is financially attractive, systems build it.
When it is not, capacity lags demand.

⚖️ Evidence-Based Cognitive Care Does Not Scale Under Current Economics

Endocrinology is defined by longitudinal management of complex chronic disease:

• Diabetes technology and medication titration
• Thyroid and osteoporosis management
• Pituitary and adrenal disorders
• Cardiometabolic risk reduction

These visits require interpretation, counseling, and follow-up over years or decades.

Time-motion studies show physicians already spend as much or more time on documentation and care coordination as in direct patient care (Tai-Seale et al., 2017; Sinsky et al., 2016). There is limited safe room to increase throughput without degrading quality or accelerating burnout.

Under these constraints, shorter visits and larger panels do not solve access. They shift the cost elsewhere. Where that cost lands—primary care, emergency departments, parallel markets, or the quality of decision-making—is the downstream story of this structural mismatch.

🌊 The Demand Reservoir

Endocrinology demand behaves less like a queue and more like a reservoir.

Chronic disease generates ongoing follow-up requirements. As prevalence rises, the number of required visits accumulates faster than incremental capacity can drain (Bodenheimer & Pham, 2010).

This explains why:

• Adding clinic sessions rarely fixes backlog
• Urgent slots do not change long-term access
• Panels refill quickly after temporary relief

The system returns to equilibrium.

* * *

🏥 This Pattern Extends Beyond Endocrinology

Endocrinology is simply an early and visible example.

Primary care workforce projections estimate shortages of tens of thousands of physicians over the coming decade (AAMC, 2023). Behavioral health faces similar structural deficits, with many U.S. counties lacking adequate psychiatric coverage.

Across cognitive specialties, the pattern is consistent:

• Rising chronic disease burden
• Finite clinician supply
• Increasing longitudinal complexity
• Expanding unmet need

These are not isolated operational failures. They reflect a system operating beyond workforce limits. The lines of emergency department patients in chairs and make- shift beds in hallways is a striking reflection of sustained systematic dysfunction.

🔄 System-Level Consequences

When specialty access is constrained, demand does not disappear. It redistributes.

Primary care absorbs higher complexity without additional time or staffing. Emergency departments become default access points (Merritt Hawkins, 2022). Fragmented telehealth and direct-to-consumer services expand into the gap.

These are not purely innovations. They are predictable adaptations to a capacity bottleneck.

* * *

⚖️ A Stable Equilibrium—Not a Temporary Crisis

When:

• Training pipelines expand slowly
• Chronic disease prevalence rises
• Cognitive work is reimbursed less than procedures
• Workforce retirement accelerates

then prolonged wait times are not an anomaly. They are a stable equilibrium.

In systems terms, endocrine access reflects a low-elasticity demand reservoir constrained by finite clinical labor. The same dynamics now shape primary care, mental health, and other cognitive specialties.

📐 SIDEBAR: What One Full-Time Endocrinologist Can Realistically Support

Why access problems persist even when clinicians “work harder.”

Typical health system outpatient template:

• New patient: 45–60 minutes, • Return visit: 20–30 minutes • Documentation, inbox, care coordination: 1–2 additional hours/day

ANNUAL CLINICAL CAPACITY (REALISTIC):

Clinical days per week: 4 days, Working weeks per year: 46 weeks, Patient slots per day: 18–22 slots

Annual visit capacity: 3,500–4,000 visits (Time-motion studies suggest this already pushes sustainable limits.)

Chronic disease follow-up demand:

Typical endocrine follow-up frequency:

• Diabetes on insulin or CGM: 3–4 visits/year

• Osteoporosis: 1–2 visits/year

• Thyroid disease: 1–2 visits/year

• Complex pituitary/adrenal: 2–4 visits/year

Average across a mixed panel: ≈ 2–3 visits per patient per year.

RESULTING SUSTAINABLE PANEL SIZE:

Annual visit capacity: 3,800 visits/year

Visits per patient: 2–3 visits/year

───────────────────────────────────

SUSTAINABLE PANEL: 1,200–1,800 patients

What happens above that threshold:

Once panels exceed ~1,500–1,800:

• Return intervals lengthen

• New patient access collapses

• Inbox volume rises non-linearly

• Clinician cognitive load increases

• Burnout risk accelerates

• Downstream utilization increases

❓ The Real Question

The question is no longer whether large health systems can provide rapid access to cognitive, longitudinal specialty care at scale under current incentives. The data suggest they cannot.

The more relevant question is who absorbs the cost of that structural mismatch:

• Patients
• Primary care
• Cognitive specialists
• Emergency departments
• Parallel care models and telehealth markets
• Or the quality of longitudinal medical decision-making itself

The answer is increasingly ‘all of the above’—and that redistribution is reshaping how endocrine care is delivered outside traditional health systems.

Until the underlying workforce equation changes, the outcome is unlikely to change with it.

——————————————————————————————————————

References

Association of American Medical Colleges (AAMC). (2023). The Complexities of Physician Supply and Demand: Projections from 2021 to 2036.

Bodenheimer T, Pham HH. Primary care: current problems and proposed solutions. Health Aff. 2010;29(5):799-805.

Endocrine Society. (2014). The Endocrinology Workforce: Supply and Demand Projections. Lewin Group white paper. Available at: https://www.endocrine.org/advocacy/workforce

Medicare Payment Advisory Commission (MedPAC). (2023). Report to Congress: Medicare Payment Policy.

Merritt Hawkins. (2022). Survey of Physician Appointment Wait Times and Medicare and Medicaid Acceptance Rates.

Sinsky C, et al. Allocation of physician time in ambulatory practice: a time and motion study in 4 specialties. Ann Intern Med. 2016;165(11):753-760.

Tai-Seale M, et al. Electronic health record logs indicate that physicians split time evenly between seeing patients and desktop medicine. Health Aff. 2017;36(4):655-662.

Vigersky RA, et al. Evolving diabetes clinical care: The endocrinologist and the general internist. J Clin Endocrinol Metab. 2008;93(4):1164-1171.

Vigersky RA, et al. The endocrinology workforce in the United States: a supply-demand analysis. J Clin Endocrinol Metab. 2014;99(9):3112-3121.