RiverRock Methodology
Table of Contents
Why You're Confused
If you're interested in longevity and health optimization, you've probably noticed something frustrating. The advice is contradictory, the influencers disagree with each other, and the science seems to support every position and its opposite.
Standard medicine tells you there's not much you can do for prevention beyond vaccines and managing disease when it shows up. Functional Medicine promises root cause solutions through expansive testing and supplement protocols. Medicine 3.0 (popularized by Peter Attia) promises precision longevity through biomarker optimization and mechanistic interventions.
They can't all be right. But how do you know which approach to trust?
The problem isn't that you lack information. The problem is that most approaches to health optimization are built on flawed epistemology. Flawed ways of knowing what's true. They confuse association with causation, mechanism with outcome, and plausibility with proof. They sell you certainty when the evidence supports only modest confidence. They optimize biomarkers that don't matter and ignore the few things that do.
RiverRock exists because Dr. Z got tired of watching patients get misled by beautiful stories backed by terrible science. This page explains the epistemological framework that distinguishes RiverRock from everything else. Bayesian reasoning, hard clinical endpoints, and intellectual honesty about what we can and cannot know.
Why Popular Approaches Fail
Let's be specific about the failure modes of the three dominant paradigms.
Standard Medicine is excellent at treating acute disease and managing chronic conditions. If you have pneumonia, standard medicine will save your life with antibiotics. If you have diabetes, standard medicine will manage your blood sugar and reduce complications. The problem is that standard medicine is reactive, not proactive. It waits for disease to develop, then treats it. For someone who wants to maximize longevity and healthspan before disease appears, standard medicine has very little to offer beyond "don't smoke, exercise sometimes, and we'll see you when something goes wrong."
Functional Medicine promises to address root causes rather than symptoms. It promotes expansive lab testing, identifies "imbalances," and prescribes supplement protocols to correct them. The rhetoric is appealing. Instead of waiting for disease, we'll optimize your biochemistry now. The problem is that Functional Medicine operates almost entirely on observational studies, mechanistic reasoning, and biomarker associations. None of which establish causation. When these interventions are actually tested in placebo-controlled trials, they almost always fail.
Medicine 3.0 (Peter Attia's framework) is more sophisticated than Functional Medicine. It acknowledges the limitations of standard medicine and emphasizes prevention, longevity, and optimization. It uses advanced testing (coronary calcium scores, DEXA scans, continuous glucose monitors) and mechanistic reasoning to guide interventions. The problem is that Medicine 3.0 still confuses mechanistic plausibility with proof. Just because an intervention makes biological sense doesn't mean it works when tested. The history of medicine is littered with mechanistically plausible interventions that failed or caused harm when rigorously studied.
Let's look at specific examples of how these approaches fail.
Homocysteine and the Biomarker Bamboozle
Homocysteine is an amino acid in the blood. Elevated levels are strongly associated with heart disease, stroke, and dementia. The association is robust. Dozens of observational studies, thousands of patients, statistically significant correlations. The biochemistry is compelling. Homocysteine causes inflammatory signaling, increases collagen production in blood vessel walls, and promotes clotting. Genetic conditions that cause extremely high homocysteine levels (like homocystinuria) result in premature heart disease in children.
The mechanistic story is beautiful. Homocysteine damages blood vessels, leading to heart disease. B vitamins (B6, B12, folate) lower homocysteine levels by 18-27%. Therefore, taking B vitamins should reduce heart disease risk.
This is exactly the kind of reasoning that Functional Medicine and Medicine 3.0 use. It's logical, mechanistically sound, and backed by observational data.
"The baseline homocysteine level predicted risk, but lowering it did nothing."And it's wrong.
Three major randomized controlled trials tested whether lowering homocysteine with B vitamins reduces heart attacks, strokes, or death. VISP, NORVIT, and HOPE-2. All three trials successfully lowered homocysteine levels. None of them reduced cardiovascular events. In fact, some showed trends toward harm.
The baseline homocysteine level predicted risk, but lowering it did nothing.
This is what Dr. Z calls the Biomarker Bamboozle. The fallacy of thinking that changing a correlated biomarker will change the outcome. It's like noticing that Ferrari hood ornaments are strongly correlated with fast cars, then adding a Ferrari hood ornament to your Honda and expecting it to go faster.
The most likely explanation is that some third factor causes both elevated homocysteine and heart disease. Lowering homocysteine doesn't address the underlying cause, so it doesn't reduce risk.
This is not a unique failure. The same pattern repeats across dozens of biomarkers that wellness influencers and longevity doctors promote. The observational data looks convincing, the mechanism is plausible, but when you test the intervention in a placebo-controlled trial, it doesn't work.
Testosterone and Path Dependence
Testosterone supplementation is another example. The logic seems airtight. Muscle mass is associated with longevity. Testosterone increases muscle mass. Therefore, testosterone should increase longevity.
But when you look at the actual trials, testosterone supplementation has never been shown to reduce cardiovascular disease or improve longevity. In fact, non-industry-funded trials show increased cardiovascular risk. Even the industry-funded TRAVERSE trial, which found no increase in heart attacks or death, showed increased risk of pulmonary embolism, atrial fibrillation, and kidney failure.
A Striking Fact
While influencers promote the longevity expectations of TRT, science shows that eunuchs (men castrated before puberty, who have the same testosterone levels as women) live 14-19 years longer than intact men of the same socioeconomic status. This pattern is true across all mammalian species.
The lesson here is path dependence. It matters how you get to an outcome, not just that you get there. You can increase muscle mass with testosterone, but that doesn't give you the longevity benefits associated with muscle mass gained through exercise. You can lower your A1C by donating blood or taking high-dose vitamin C (which screws up the lab equipment), but that doesn't improve your diabetes.
Mechanism is not outcome. Surrogate endpoints (like muscle mass or lab values) are not the same as hard clinical endpoints (like heart attacks, strokes, and death).
Full-Body MRI and Bayesian Math
Full-body MRI screening is marketed as a way to catch cancer early. The technology is actually not that accurate, but the tech bros and influencers always wave that away with "it'll get better". So let's go ahead and show that the problem remains even if that were true. Let's say that the technology improves until the full body MRI is 99% sensitive and 99% specific. That's 1% false negatives and 1% false positives. Very few medical tests are that accurate. Most people assume that a positive result on such an accurate test must be reliable.
But accuracy isn't the only thing that matters. What matters is the pretest probability, which is the likelihood of disease before the test is performed.
Cancer is rare in healthy 35-year-olds. Let's say the true prevalence is 1 in 1,000 (it's much less than this in reality). If we scan 1,000 people, the MRI will catch the 1 real cancer (99% of 1 is approximately 1). The MRI will also generate false positives. 1% of 999 healthy people is approximately 10 false positives.
Now we have 11 positive scans. 1 real cancer and 10 false positives. The probability that any given positive result is actually cancer is 1 in 11, or about 9%. Even with a 99% accurate test, the false positive rate is 91% when screening for rare conditions.
This is not a flaw in the technology. This is a mathematical reality of how diagnostic tests perform in low-prevalence populations.
Math Proves Full-Body MRI Cannot Be Useful (No Matter How Accurate It Gets)
The same principle applies to coronary calcium scores in asymptomatic patients, expansive micronutrient panels in the absence of symptoms, and "optimal" lab ranges that pathologize normal variation.
Medicine 3.0 and Functional Medicine operate as if advanced technology and mechanistic reasoning can overcome these constraints. They cannot.
The Framework
Bayesian reasoning is a framework for updating beliefs in light of new evidence. It formalizes how we should think about probability, uncertainty, and the limits of what we can know.
The core insight is this. The value of a test depends not only on its accuracy but also on the pretest probability. The likelihood of disease before the test is performed. A highly accurate test applied to a low-probability scenario generates more false positives than true positives. A moderately accurate test applied to a high-probability scenario can be highly informative.
A test doesn't "prove" anything. It updates probabilities.
This changes clinical decisions every day.
Example: Strep Throat
A 45-year-old patient has a sore throat but no fever, no exudate, no lymphadenopathy, and no cough. A nurse runs a strep swab according to protocol. The swab comes back positive.
Even with a positive test, the post-test probability of strep throat is only 22%. Why? Because the pretest probability (based on the patient's age and lack of symptoms) is so low that the test cannot provide useful information. The test probably shouldn't have been done in the first place, and the positive result is not actionable.
Now consider a 5-year-old with the same symptoms. Because the pretest probability is higher in children, the post-test probability of a positive test is nearly 70%. Actionable information. Treatment can be instituted with confidence.
But here's where it gets more interesting. If the 5-year-old has exudate, fever, and lymphadenopathy, the pretest probability is so high that even a negative strep swab leaves a residual 15% probability of disease. A throat culture or empiric treatment may be indicated despite the negative test.
This is Bayesian reasoning in practice. The same test, applied to different clinical contexts, yields radically different conclusions.
Why This Matters
Bayesian reasoning explains why so many wellness interventions fail when tested. The pretest probability that a healthy, asymptomatic person has a clinically meaningful deficiency or imbalance is very low. When you run expansive lab panels on healthy people, you're guaranteed to find abnormalities. Not because there's a real problem, but because of statistical noise.
A normal range is defined as the middle 95% of the population. That means for any single test, there's a 5% chance it will flag as abnormal just by chance. When you run 20 tests, the probability of at least one false alarm is 64%. At 50 tests, it's over 92%.
These panels are designed to find something wrong. And once they do, the Functional Medicine or Medicine 3.0 practitioner will tell you a compelling story about why that biomarker matters and what supplement or intervention will fix it.
But if the pretest probability was low, and the test was never validated for use in asymptomatic populations, then the abnormal result is noise, not signal. Trying to "fix" it is chasing a solution to a problem that doesn't exist.
What Actually Works (And How It Works)
The list of interventions with robust evidence of benefit for longevity and healthspan is short. Not 20 things. Not 40 things. Maybe five or six if you're generous.
But here's the critical nuance that most approaches miss. It's not enough to hit the metric. What matters is the method you use to achieve the change. The method is what confers the benefit, not the metric itself.
VO2 Max (But Only Through Exercise)
VO2 max (cardiorespiratory fitness) is one of the strongest predictors of longevity. Unlike blood biomarkers, we have decades of evidence showing that changes in VO2 max achieved via exercise correspond to changes in cardiovascular risk. Moving from the 5th percentile to the 90th percentile has enormous benefits. Moving from the 97th to the 99th percentile has trivial benefits. Diminishing returns apply.
But you cannot shortcut this by taking a supplement that artificially boosts your VO2 max score. Baking soda (sodium bicarbonate) can temporarily improve VO2 max performance by buffering lactic acid. But there is no evidence that taking baking soda improves longevity. The benefit comes from the cardiovascular adaptations produced by exercise, not from hitting a number on a test.
If you could take a pill that increased your VO2 max without exercise, it would not give you the longevity benefit.
The method matters. Everything has diminishing returns.
Strength (But Only Through Resistance Training)
Muscle mass and strength are associated with longevity. But increasing muscle mass via testosterone supplementation does not improve longevity. In fact, it may increase cardiovascular risk. The muscle you build through resistance training delivers the longevity benefit. The muscle you build through exogenous hormones does not.
Why? Because the adaptations that occur during resistance training (improved insulin sensitivity, mitochondrial biogenesis, neuromuscular coordination, metabolic resilience) are what drive the benefit. Testosterone bypasses those adaptations. You get the muscle without the metabolic benefit.
The method matters. Everything has diminishing returns.
Social Connectedness (We Don't Really Know Why)
The situation with social connection and loneliness is interesting. The evidence points to a causal connection, similar to smoking back in the 1950s, meeting ALL of the Bradford-Hill criteria. But, we don't know the exact factors.
It seems to affect introverts and extraverts equally. It doesn't seem to matter whether you like who you're socializing with or dislike them. And the types of socialization seem to matter more than just the amount.
The types of socialization that have been studied are: marriage or living with a partner, immediate family, extended family, friends, and communities or groups. Having some exposure to each type appears to be the most protective, but the data is early.
It also doesn't take very much to get the benefits. It may only take as little one contact with each type per month. Beyond that, the benefits are smaller and smaller.
The method matters. Everything has diminishing returns.
Don't Smoke (But Only Through Quitting)
There is plenty of data that smoking is harmful. It's one of the most well-established results in medical science. It started with the Bradford-Hill criteria, but as it became well known, smoking rates dropped, and the associated diseases dropped as well, constituting a huge experimental proof.
But if you stop smoking by taking up heroin, or methamphetamines, you will not see the benefits. You might laugh at my example, but the recommendations of influencers are using this exact same logic to recommend their supplements and protocols. The foolishness seems obvious in my example here, but if I tell you that some supplement increases VO2Max, Peter Attia and Andrew Huberman will immediately recommend supplementing, because they have not thought deeply about the science and the difference between a metric the underlying biological reality.
The method matters.
Treat Severe Chronic Diseases (But Only With Proven Medications)
Lowering very high A1C improves outcomes in diabetes. But not all methods of lowering A1C are equal. And the benefits of lowering on mortality diminish severely below 7.5 (possibly even causing harm - see ACCORD trial, as well as ADVANCE and VADT).
Lowering A1C via sulfonylureas or insulin does NOT improve longevity, while using GLP-1 agonists (ozempic, mounjaro), some SGLT-2 inhibitors (empaglifozin, but not dapaglifozin), or metformin does improve longevity.
Treating severe hypertension is similar. Most classes of medication work equally. But sepsis and infection, which do lower blood pressure, don't reduce mortality. Beta-blockers, which lower blood pressure AND resting heart rate, seem to be less effective in reducing mortality, and some studies indicate that they could be harmful (ASCOT-BPLA, among others).
Treating hypertension with medications like ACE-I's, ARBs, thiazides, or CCBs, is effective in reducing mortality. It is MOST effective for severe hypertension, and has severely diminishing returns for milder disease.
The method matters. Everything has diminishing returns.
The Principle
This is why biomarker optimization is a trap. You can manipulate almost any biomarker through some intervention. But unless the intervention method itself has been tested in a placebo-controlled trial with hard clinical endpoints, you have no idea whether you're helping or harming.
The interventions that work are not shortcuts. They are the hard things. Exercise. Strength training. Social connection. Smoking cessation. Managing blood pressure and diabetes when present.
The reason the list is short is that most interventions fail when tested. Mechanistic plausibility is not enough. Association is not causation. Surrogate endpoints are not outcomes. And hitting a metric through an untested method is not the same as achieving the outcome through a proven method.
The Method
RiverRock's clinical method follows a three-step process.
Step 1: Reject Certainty Theater
The first step is intellectual honesty. We do not pretend to eliminate uncertainty. We quantify it, communicate it, and make decisions that account for it. We do not recommend interventions that lack evidence of benefit. We do not pathologize normal variation. We do not chase mechanistic theories that have not survived rigorous testing.
This requires discipline. It requires saying no to interventions that sound advanced but lack proof. It requires resisting the cultural pressure to "do something" when doing nothing is the right answer.
Step 2: Optimize What Can Be Optimized
For the short list of interventions that actually work, we help you identify where you are on the curve and optimize to the point where further effort yields minimal return. If you want to pursue the 95th percentile, or the 99th, we can help. But if you think moving from the 99th percentile to the 99.9th percentile will improve your longevity, you're engaging in medical theater, not science.
If you think moving from the 99th percentile to the 99.9th percentile will improve your longevity, you're engaging in medical theater, not science.This is where VO2 max testing, strength assessment, bloodwork, and blood pressure checks come in. We can use these metrics as measurements for when you are into the area of diminishing returns from proven interventions.
You should do cardio, but at some point, more will be harmful. Same for lifting weights. You should socialize, but at some point, there's no more longevity benefit. The metrics aren't there to be manipulated, they are they to guide the proven interventions - when to start, and when returns are diminishing.
Step 3: Maximize MeaningSpan
We have reached the same point of diminishing returns in longevity that exists in other domains. Even if you want to aim for the 99.9th percentile "just to be sure", eventually you still hit the point where you're not even extending your life by months or days. Perhaps you're even reducing it. And if pursuing that "optimization" requires the sacrifice of your joy and meaning, you've really made a poor trade.
Consider Rory Sutherland's example of the Eurostar train between Paris and London. Six million pounds were spent to reduce the journey time by 40 minutes. For 0.01% of that money, they could have put WiFi on the trains, which wouldn't have reduced the duration of the journey but would have improved its enjoyment and usefulness far more. For 10% of the money, they could have paid supermodels to walk up and down the train handing out free wine. People would have asked for the trains to be slowed down.
The psychological and subjective are far more important than the objective. No matter how much money you spend, you will not make that train ride ten times faster. But you could easily make it ten times more useful and enjoyable. Similarly, you're not going to live ten times longer, no matter how loudly Bryan Johnson promotes his delusions.
But, your life could be ten times more useful and enjoyable.
"The goal should not be to maximize the length of life alone, but to maximize the area of life. The product of both the length and the depth."Influencers fixate on trivial improvements (if that) in the objective duration of life, but give no attention to reclaiming and maximizing the subjective depth and meaning of our experience of life.
The goal should not be to maximize the length of life alone, but to maximize the area of life. The product of both the length and the depth. In many circumstances, it may be worth it to trade off a few years of life if you could significantly increase the meaningfulness and depth of your experience.

The RiverRock Overview
This is MeaningSpan. The optimization not just of duration, but the combination of duration and depth. Both the length of life and the meaningfulness of life.
Health is not the absence of disease. Health is the capacity to execute a meaningful life. It is the mental and physical ability to pursue what matters to you.
RiverRock's Methodology begins with your dreams, your values, and your chosen aims. We then work backward to identify the constraints (physical, psychological, relational, financial) that limit your capacity to pursue those aims. Some constraints can be removed. Some must be accepted. Some require trade-offs.
This is medicine as applied philosophy.
RX-Bayes: Bayesian Reasoning in Practice
To implement Bayesian reasoning in clinical practice, Dr. Z built RX-Bayes, a diagnostic tool that calculates real-time probability adjustments based on test characteristics and pretest probability.
The tool allows clinicians to see how changes in pretest probability or test accuracy affect post-test probabilities. It includes pre-populated data for common conditions (strep throat, pneumonia, pulmonary embolus, UTI, breast cancer) and allows users to adjust for uncertainty in sensitivity, specificity, and prevalence.
RX-Bayes makes Bayesian reasoning explicit, transparent, and reproducible. It is not a marketing tool. It is a clinical decision support tool, built for function rather than aesthetics.
See the technical walkthrough →
Go Deeper
This page provides an overview of the RiverRock Method. For those who want to explore further:
- Download the full whitepaper – A detailed technical explanation of Bayesian epistemology, the mathematical critique of Medicine 3.0, and the Root Cause Analysis methodology. Download PDF 2.4 MB • PDF
- Watch the lecture series – Dr. Z has recorded extensive discussions on longevity skepticism, Bayesian medicine, and MeaningSpan philosophy. Topics include:
- Why homocysteine optimization is a biomarker bamboozle
- Why testosterone supplementation doesn't improve longevity
- Why full-body MRI screening generates more harm than benefit
- Why massive lab panels on healthy people are "quantified quackery"
- Read related essays – For more on the philosophy behind RiverRock:
RiverRock exists for people who want rigor instead of theater, and who want their health to power a meaningful life.