The Peter Attia Drive - #392 - Genetic testing： when it's valuable, how to choose the right test, and what to do with the results

00:00:00.000 Hey, everyone. Welcome to The Drive Podcast. I'm your host, Peter Atiyah. This podcast,

00:00:16.540 my website, and my weekly newsletter all focus on the goal of translating the science of longevity

00:00:21.520 into something accessible for everyone. Our goal is to provide the best content in health and

00:00:26.720 wellness, and we've established a great team of analysts to make this happen. It is extremely

00:00:31.660 important to me to provide all of this content without relying on paid ads. To do this, our work

00:00:36.960 is made entirely possible by our members, and in return, we offer exclusive member-only content and

00:00:42.980 benefits above and beyond what is available for free. If you want to take your knowledge of this

00:00:48.100 space to the next level, it's our goal to ensure members get back much more than the price of the

00:00:53.420 subscription. If you want to learn more about the benefits of our premium membership,

00:00:57.480 head over to peteratiamd.com forward slash subscribe.

00:01:03.920 Welcome to today's episode of The Drive. Few areas in medicine generate as much fascination

00:01:10.520 and as much confusion as genetic testing. The basic intuition here is sound. If DNA contributes

00:01:18.240 so strongly to our biological machinery than sequencing, it should, in principle, help us

00:01:24.900 understand health and disease more clearly. It's true that some genetic tests can be genuinely

00:01:30.100 life-changing. But it's equally true that others are barely more useful than a horoscope.

00:01:36.600 Most fall somewhere in between, in a gray zone that is far more nuanced than either the hype

00:01:41.740 or the skepticism would suggest. So today I want to build a practical framework for thinking

00:01:47.240 clearly about genetic testing. Through this, I'll discuss what genetic tests can and can't do,

00:01:54.420 why most genetic tests are probabilistic rather than deterministic, and why directly measuring

00:01:59.780 the phenotype is often more useful than inferring risk from DNA alone. I'll also walk through where

00:02:06.080 genetics can be most informative across the major disease categories, the four horsemen,

00:02:11.680 and where it is much more limited than people assume. Finally, I'll talk about how to choose

00:02:16.980 the right test or type of test, how to interpret the results, and how to avoid the very common

00:02:22.300 mistake of getting more information without getting more clarity. So without further delay,

00:02:28.180 I hope you enjoy this episode of The Drive. Recently, a patient asked me a question that I

00:02:39.000 hear all the time. Should I be doing genetic testing? It sounds like a simple question,

00:02:44.540 but in reality, it doesn't tell me very much because the question can mean a lot of different

00:02:49.920 things. You know, it could mean I'm worried about my risk of Alzheimer's disease should we test for

00:02:54.700 APOE genotype. It could mean my mother died of breast cancer. Should I find out whether or not

00:02:59.900 I carry a BRCA mutation? It could mean, will genetic testing help determine the best medication

00:03:06.180 to address my lipids? Or more commonly, it means something much broader. What diseases am I most

00:03:12.720 likely to get. And that last question is really what most people have in mind. And I, of course,

00:03:18.360 understand why. The notion that your DNA might serve as the kind of blueprint for your future

00:03:24.040 health, that if you could read it closely enough, it would tell you what problems are coming and

00:03:29.880 what to do about them. That is an extraordinarily compelling idea. If we truly had a test that could

00:03:36.620 reliably tell you which diseases you were most likely to develop and exactly how to prevent them,

00:03:41.580 that would be a genuine game-changer for medicine.

00:03:44.520 And that promise, or at least some version of that promise,

00:03:48.100 is what many companies have tried to sell.

00:03:50.720 Genetic testing has been marketed as a way to tell you everything

00:03:54.360 from what diet you should eat, to how you should exercise,

00:03:57.840 to which supplements you need,

00:03:59.480 to how well your body handles things like detoxification or methylation.

00:04:03.980 But this is where we have to slow down and be much more precise,

00:04:08.000 because that's not quite the reality we live in.

00:04:11.580 Genetic testing can absolutely be useful. In some cases, it can be very useful, even life-altering.

00:04:17.580 There are situations where it can meaningfully change screening, influence treatment decisions,

00:04:22.560 clarify risk, or provide critical information for family members. But there are also many

00:04:27.720 situations where it adds very little, where it's oversold, or where the better answers come not

00:04:34.200 from genetics at all, but from directly measuring the phenotype through blood work, imaging,

00:04:40.140 family history, or other clinical evaluation. So the real question is not simply, should I do

00:04:47.080 genetic testing? The real questions are, what exactly am I trying to learn, and is genetic

00:04:53.540 testing the best tool to answer that? If so, what kind of test is actually appropriate?

00:05:00.660 And if I get an answer, will it change anything meaningful about what I do next? That's how I'd

00:05:06.520 like to approach this discussion today. Before getting into where genetic testing is useful,

00:05:12.220 and where it's not, I think it's worth stepping back for a moment and talking a bit about the

00:05:17.220 history, because many of the assumptions people still carry about what genetic testing can deliver

00:05:23.200 are leftovers from an earlier era. We now live in a world where sequencing a human genome feels

00:05:29.900 almost routine. It's far less expensive than it used to be, increasingly accessible, and technically

00:05:35.580 much easier to obtain. But that normalization obscures how recent this capability actually is.

00:05:43.800 The first draft of the human genome was published in 2001, and the Human Genome Project was declared

00:05:50.420 essentially complete in 2003. So we are only a little more than two decades removed from the

00:05:57.060 first time we had a comprehensive map of human DNA. At the time, expectations were extraordinarily

00:06:03.320 high, and to be fair, understandably so. If you suddenly gained the ability to read the full

00:06:09.280 genetic code of a human being, it was not unreasonable to think that many of medicine's

00:06:14.680 biggest problems would become much easier to understand. Cancer, cardiovascular disease,

00:06:19.940 neurodegeneration, psychiatric illnesses, these all seemed like conditions that might become

00:06:25.600 far more tractable once their genetic basis was decoded. This is not exactly what happened.

00:06:32.140 Now, that does not mean the Human Genome Project failed.

00:06:35.000 It was an extraordinary scientific achievement, and it created real value,

00:06:40.020 especially in rare monogenic diseases, in some aspects of oncology and in pharmacogenetics.

00:06:47.300 But it did not deliver on the broadest version of the promises attached to it.

00:06:53.080 The assumption was that once we knew this sequence, we would quickly understand function,

00:06:58.020 that reading the code would tell us more or less directly how disease worked and how to prevent

00:07:04.540 or treat it. Part of the reason that expectation ran into trouble is the sheer scale and complexity

00:07:11.520 of the human genome. It contains roughly 20,000 genes and about 6 billion total base pairs.

00:07:19.660 Each of us differs from one another at millions of places across the genome,

00:07:25.120 including roughly 5 million single nucleotide variants,

00:07:30.820 plus many other insertions, deletions, and structural changes.

00:07:35.540 Some of these variants affect physical traits,

00:07:38.460 some affect our risk for disease,

00:07:40.520 and many appear to do absolutely nothing.

00:07:44.440 And even this description understates the complexity

00:07:47.660 because when people hear genes,

00:07:50.420 they often think only about the part of DNA that codes directly for proteins.

00:07:56.160 But those protein-coding regions of DNA make up only 1.5% of the genome.

00:08:03.920 The vast majority is non-coding, what used to be dismissed as so-called junk DNA.

00:08:10.940 We now know that much of it plays a critical regulatory role,

00:08:14.860 controlling when genes are turned on or off, where, and how strongly proteins are expressed.

00:08:22.340 So the challenge isn't just identifying variations in the coding regions. Variation in the non-coding

00:08:29.780 genome may matter enormously, even when we don't yet fully understand what it means.

00:08:35.960 In hindsight, the $2.7 billion task of sequencing the genome may have actually been the easier part.

00:08:44.860 Accurately interpreting that sequence and what we should do with it is the far bigger challenge.

00:08:50.980 To understand why, it helps to have a working mental model of how genetic variants actually

00:08:56.520 produce disease. DNA is essentially a set of instructions. Those instructions get transcribed

00:09:02.680 into RNA, which is then translated into proteins, the molecular machines that carry out virtually

00:09:09.520 every biological function in the body. A change in those instructions is called a gene variant,

00:09:16.160 or more colloquially, a mutation, though these things mean the same thing.

00:09:20.960 Often the change is inconsequential, but sometimes it results in a protein that doesn't fold

00:09:26.020 correctly, doesn't function as intended, or isn't produced at all. And it is that dysfunctional

00:09:32.860 protein that ultimately shapes the phenotype, the observable, measurable output of the body,

00:09:38.640 from a lab value to a symptom to a disease. This is the central dogma of molecular biology,

00:09:46.180 meaning information flows in one direction, from DNA to RNA, RNA to protein, and protein

00:09:53.840 to phenotype. For a small number of diseases, there is a relatively direct line from gene

00:10:00.900 to dysfunctional protein to disease. But for the conditions that matter most, the story is much

00:10:07.520 more complex. Even diseases where genetics matter a great deal tend to arise from countless

00:10:13.960 interactions among many genes and environmental triggers. Before we dig into specific diseases

00:10:21.080 or tests, it's important that we calibrate our expectations. Genetic testing can be useful,

00:10:27.220 sometimes very useful, but it is not a perfect blueprint for health and it does not replace

00:10:32.760 phenotypic data. Put another way, for genetic testing to be useful, we need to have some

00:10:39.000 confidence that learning the genetic variant will actually inform something clinically.

00:10:44.500 So with this in mind, let's discuss the major limitations we need to consider.

00:10:49.820 The first limitation, and really the most important one, is that most things we talk

00:10:55.080 about in genetics are probabilities, not guarantees. Part of the problem is that most

00:11:00.560 of us were introduced to genetics through a very simplified model in high school biology class

00:11:07.440 genetics is taught through mendelian inheritance we learned about dominant traits recessive traits

00:11:13.760 sometimes called semi-dominant traits we drew punnett squares we crossed a white flower with

00:11:19.680 a red flower and got a white red or pink flower the logic is clean the outcomes are discrete and

00:11:25.920 and the relationship between genotype and phenotype appears relatively straightforward.

00:11:30.720 And for a small number of traits and diseases, that framework is actually useful.

00:11:34.520 There are cases where a single mutation has a large and fairly predictable effect,

00:11:39.780 what we call high-penetrance mutations.

00:11:41.700 And those cases are part of what makes genetics seem so promising in the first place.

00:11:48.340 Huntington's disease is probably the most extreme example.

00:11:51.480 The HTT gene normally contains a stretch of repeated CAG sequences.

00:11:58.820 Those are just the base pairs and shorthand.

00:12:02.220 In Huntington's disease, this CAG repeat is abnormally expanded.

00:12:08.520 When the expanded gene is translated into protein, it produces a mutant Huntington protein that is toxic to neurons, particularly in the striatum and the cortex.

00:12:19.620 If you carry this expansion above the pathologic threshold, you will develop Huntington's

00:12:26.900 disease, full stop.

00:12:28.560 There is essentially no version of that story where the mutation is present and the disease

00:12:33.380 doesn't follow.

00:12:34.580 But that's the exception, not the rule.

00:12:37.220 For the vast majority of conditions that people care about, heart disease, cancer, diabetes,

00:12:43.420 most psychiatric illnesses, that is simply not how genetics work.

00:12:48.100 These are not one-gene, one-disease problems.

00:12:51.800 They are shaped by the combined effects of many genes, each often contributing only a

00:12:57.360 small amount of risk, layered on top of environment, behavior, aging, and chance.

00:13:03.520 Which means you can carry a variant associated with increased risk and never develop the

00:13:08.920 disease, and you can lack any known high-risk variant and still develop the disease anyway.

00:13:14.780 The genetics shift the probability distribution. It doesn't write the ending. People without a

00:13:21.620 BRCA mutation get breast cancer, and people with two copies of APOE4 never develop Alzheimer's

00:13:29.100 disease. The genes matter, but they are almost always operating alongside other genes and other

00:13:35.180 inputs that our tests will likely never fully capture. The second limitation is that our ability

00:13:41.640 to generate genetic data has moved faster than our ability to interpret it. This is especially

00:13:47.900 true as tests get broader. The more of the genome you look at, the more likely you are to find

00:13:54.860 something. But finding something is not the same as finding something useful. And even when a lab

00:14:00.780 does a good job limiting what it reports, broader testing still increases the amount of information

00:14:07.140 that has to be interpreted. We can now sequence a genome for a few hundred dollars, but there are

00:14:13.020 simply too many variants, too many interactions, and too many things we still do not understand.

00:14:18.900 So one of the real paradoxes of genetic testing is that broader testing doesn't always produce

00:14:24.440 more clarity. Sometimes it just produces more ambiguity. You get more data, but not necessarily

00:14:30.620 more understanding. The third limitation is that genetic information is often less informative

00:14:36.760 than directly measuring the phenotype. If the thing you care about can be measured directly,

00:14:42.520 that is usually a better place to start than a genetic proxy for it. If I want to know someone's

00:14:48.620 risk for a heart attack, I can measure their cholesterol and image their coronary arteries.

00:14:54.120 I can take their blood pressure and ask whether or not they're smoking. The phenotype tells me

00:14:59.300 something that is happening right now, something integrated across all the contributing factors,

00:15:04.360 genetic, environment, behavior, age, everything. It's a real-time readout that can be acted upon.

00:15:10.560 Genetic testing is something different. It's reading the source code without necessarily

00:15:15.080 knowing how the program will run. The source code matters, but the program is shaped by a thousand

00:15:22.180 other inputs that the sequence alone doesn't capture. And finally, one of the most under

00:15:28.420 appreciated aspects of genetic testing, the potential psychological weight of the information.

00:15:34.420 I have had patients who found out they did not inherit the same mutation carried by their parents

00:15:39.420 suffered from a debilitating disease, and who were so overwhelmed with relief that they literally

00:15:44.760 started to cry on a call. The results of a genetic test gave them clarity around something they had

00:15:50.700 been dreading for years. That's a real value, even if it doesn't show up in an outcome study.

00:15:56.180 but I've also had patients who received a result indicating an elevated risk for a disease and

00:16:02.400 were consumed by anxiety about it. It didn't make them more proactive, it just made them

00:16:07.740 more frightened in a way that affected their quality of life for years. This matters because

00:16:13.140 information is not automatically useful just because it's true. A result that is likely to

00:16:19.020 produce fear or confusion without changing screening, treatment, or planning in a constructive

00:16:23.920 way has real costs, and those costs need to be a part of the calculation before you ever order a

00:16:30.220 test. Okay, so the obvious question then is, if genetic testing has all of these limitations,

00:16:37.880 how should we think about whether it's worth doing at all? I think it boils down to a few

00:16:43.560 important questions. One, what exactly are you trying to learn? The more specific the question,

00:16:50.520 the better. Two, is genetic testing the best tool for this question, or is it easier and more

00:16:57.540 informative to measure phenotype, the actual biological output, directly? Three, if you get

00:17:04.300 an answer, what will you do differently? In other words, how will this test change your behavior?

00:17:10.820 And four, are you mentally prepared for the answer, whether positive or negative? Genetics

00:17:16.240 should be empowering and informative,

00:17:18.420 so if the results are likely to just scare you,

00:17:21.180 maybe it's not the right choice, at least for you.

00:17:24.220 There aren't necessarily right or wrong answers here,

00:17:27.220 but these are the questions I think people

00:17:29.000 should be thinking about when they consider genetic testing.

00:17:32.140 Layered on top, I think there's one other consideration

00:17:34.360 to help you determine how useful

00:17:36.120 a genetic test is likely to be.

00:17:38.260 And that is, how large is the effect

00:17:41.600 of the genetic signal you're looking for?

00:17:43.720 is this something that nudges the disease risk by 5% or is it something that changes the risk

00:17:50.180 by 10 to 20 fold? And so with this framework in mind, we can now turn to the practical question

00:17:58.640 across the major disease categories, where is genetic testing actually useful and for what

00:18:04.340 kinds of questions? If we start with the biggest threats to lifespan, the first place to go is

00:18:10.760 atherosclerotic cardiovascular disease and metabolic disease. And here, in general, I think

00:18:16.160 the case for routine genetic testing is relatively weak. This is not because genetics don't matter in

00:18:22.460 these conditions. Lipids, blood pressure, insulin resistance, and obesity are very clearly influenced

00:18:28.180 by genetics. But the clinical question is not whether genes play a role, but whether knowing

00:18:34.460 the genotype gives you something more useful than simply measuring the phenotype directly.

00:18:39.340 And most of the time, it does not. Take LP little a. It is almost entirely genetically

00:18:46.040 determined, driven largely by the LPA gene. In fact, it is the most common hereditary driver

00:18:53.060 of cardiovascular disease. But knowing someone carries a variant that influences LP little a

00:18:59.140 doesn't actually tell me what I need to know. I'm still going to measure it directly because

00:19:04.920 the measurement gives me more precise, actionable information than the genotype does. The same

00:19:10.560 logic applies across the board. If I want to know whether someone has hypertension, I measure their

00:19:15.960 blood pressure. If I want to assess insulin resistance, I have direct tools to do that.

00:19:21.360 And of course, if I want to know their LDL cholesterol or ApoB concentration, both things

00:19:27.460 that are highly influenced by genetics, I can simply measure those things and measure their

00:19:33.200 response to therapy. For most of the major drivers of cardiovascular and metabolic disease,

00:19:39.320 we already have access to the things that matter most, and those things are far more actionable

00:19:45.240 than a genetic estimate of predisposition. Now, that does not mean there are no exceptions.

00:19:50.980 Familial hypercholesterolemia is an obvious one. If someone presents with a markedly elevated LDL

00:19:57.460 and a family history suggesting a monogenic lipid disorder, genetic confirmation can be

00:20:03.600 genuinely useful, not necessarily because it changes the initial treatment, but because it

00:20:08.980 can solidify the diagnosis and potentially trigger cascading screenings in relatives who may be

00:20:14.840 affected without knowing it. There's also the rarer situations where specific variants change

00:20:20.740 how we interpret the phenotype all together. SCARB1 mutations, for example, can cause HDL

00:20:28.300 cholesterol to appear elevated in a way that looks falsely reassuring, when in reality,

00:20:34.280 the patient's cardiovascular risk is substantially higher than their lipid panel implies. I actually

00:20:40.520 had a friend that I was able to catch this diagnosis in who had spent years believing

00:20:45.900 that his HDL cholesterol of 100 milligrams per deciliter and his LDL cholesterol of 80 milligrams

00:20:52.600 per deciliter meant he was free and clear of risk, when in reality, a calcium score revealed that he

00:21:00.480 was riddled with disease. These mutations are rare, but they do illustrate where the genotype

00:21:07.060 is actually telling you something that is not immediately obvious from a lipid panel and prompt

00:21:12.400 additional testing. And then there is another category that I think is worth acknowledging.

00:21:17.880 Cases where genetic information shifts behavior rather than truly changing clinical care. I've

00:21:24.480 seen patients agree to start lipid-lowering medication that they'd been resistant to

00:21:29.280 after seeing genetic data that confirmed their risk. I've seen patients who struggled with

00:21:34.320 obesity for years find it meaningful, even relieving, to learn that they carry multiple

00:21:40.940 genetic risk variance for weight gain. Not because it changed the treatment plan, but because it

00:21:46.300 reframed the problem in a way that felt less like a personal failure. That psychological shift is

00:21:52.260 real and I don't dismiss it, but it is different from saying the test revealed something the

00:21:58.000 phenotype could not. In most of these cases, the clinical information was already there. The

00:22:03.200 genetics just changed how the patient related to it. So the conclusion for this category is fairly

00:22:09.520 straightforward. Genetics matter a lot, but for most people, it is not the right first tool.

00:22:15.660 Measure the phenotype. If something in the clinical picture raises a specific question,

00:22:20.840 a family history that doesn't add up, an unusually extreme lab value, a presentation that suggests a

00:22:27.320 monogenic disorder, then consider whether genetic testing adds something. But as a routine starting

00:22:33.640 point, the phenotype almost always beats the genotype here. Now, once you move outside of

00:22:40.160 ASCVD into inherited cardiac conditions, the question changes a bit. This is where genetic

00:22:46.180 testing can become more compelling because there are real inherited syndromes involving arrhythmia,

00:22:53.300 cardiomyopathy, and structural heart disease where genotype may reveal risk that is not obvious from

00:22:59.120 routine labs or standard cardiovascular risk markers. Of course, these are also conditions

00:23:04.740 we can test for, but may not do so nearly as regularly. For individuals with no obvious risk

00:23:10.800 factors or symptoms, we may not perform an EKG, or we may run these tests once and then never

00:23:16.960 run them again if everything appears normal. Conditions like atrial fibrillation, however,

00:23:22.000 can arise later in life, so a normal EKG earlier in life doesn't necessarily preclude a future

00:23:29.020 problem. Knowing about genetic risk may also prompt more regular testing that can catch a

00:23:34.860 potential life-threatening condition earlier on. So the options here are to run these somewhat

00:23:40.500 more complex tests on everyone or find a way to determine who should get regular testing.

00:23:46.220 Family history can be helpful. Patients with a family history of sudden cardiac death

00:23:50.240 or known cardiac issues are probably good candidates for routine screening.

00:23:54.800 But family history alone may not give us the whole picture. Some mutations are incompletely

00:23:59.760 penetrant, meaning some but not all of the carriers are affected, so it's entirely possible

00:24:05.660 to have a genetic risk factor without a family history. Further, a vague family history of heart

00:24:11.840 disease could mean any number of different conditions, and not all patients are able to

00:24:16.480 provide a complete, detailed history that would inform more advanced tests. So in this domain,

00:24:22.180 genetic testing can sometimes uncover a more specific inherited risk that may change what

00:24:28.060 tests you order and when. This does not mean everybody should go out looking for every

00:24:33.900 conceivable arrhythmia or cardiomyopathy gene, but compared with ASCBD, this is clearly a category

00:24:40.260 where genetics can provide more incremental value. So I would kind of put this in the middle category,

00:24:45.740 not an area for broad routine testing in everyone, but clearly more defensible when the personal or

00:24:52.020 family history points in that direction. Cancer is where the conversation becomes much more nuanced,

00:24:59.120 because cancer is, after all, a genetic disease. But most cancer is not due to inherited DNA.

00:25:07.140 The vast majority of cancers arise from what are called somatic mutations, as opposed to germline

00:25:13.320 mutations. These are acquired mutations, not the inherited ones, and they won't appear on standard

00:25:20.180 genetic tests. In fact, by most estimates, only about 5% of cancers are attributable to inherited

00:25:26.720 germline mutations. Put another way, most people who have cancer would not find anything unusual

00:25:33.760 in their inherited genetics. So the absence of cancer-predisposing mutations certainly does not

00:25:40.480 rule out the possibility of developing cancer later in life. But 5% matters because the cancers

00:25:48.100 that fall into this category tend to involve highly penetrant mutations that carry substantial

00:25:53.680 elevated lifetime risk, and because knowing about them changes management in meaningful ways.

00:25:59.880 The clearest examples, which I've already mentioned, are BRCA1 and BRCA2. Women who carry 1.00

00:26:05.400 BRCA mutations face drastically elevated lifetime risks of breast and ovarian cancer, high enough

00:26:12.540 that enhanced screening, chemoprevention, and even prophylactic surgery can be appropriate.

00:26:18.100 but these are not the only breast and ovarian cancer genes. They are also associated with an

00:26:25.020 increased risk of cancers such as pancreatic and prostate cancer, which broadens both the

00:26:30.540 clinical implications and the family history clues that matter. Similarly, Lynch syndrome,

00:26:36.580 caused by mutations in mismatch repair genes, dramatically increase the risk for colorectal

00:26:42.980 cancer and several other cancers, and knowing your status changes screening intensity and

00:26:48.640 frequency in ways that saves lives. These types of conditions also have implications for family

00:26:55.280 members. A father who carries a mutation in BRIP1 that predisposes to ovarian cancer is not going to

00:27:03.160 be concerned about his risk for cancer in an organ that he does not have, but knowing that he carries

00:27:08.800 this variant is a very good reason for his daughters to get tested. Just as importantly,

00:27:14.500 an apparent lack of family history does not entirely rule out the risk for having one of

00:27:19.920 these mutations. While not the norm, there are certainly cases of patients carrying potent

00:27:25.780 cancer-predisposing mutations with an unremarkable family history. So does this mean everyone needs

00:27:32.720 genetic testing for cancer? Not necessarily. For someone with no meaningful family or personal

00:27:39.460 history of cancer, the pretest probability of finding a high penetrance cancer mutation is

00:27:45.940 low enough that it's hard to justify routine testing on a population basis. That said, I do

00:27:52.580 think cancer is one area where an individual may opt for testing without notable family history

00:27:59.100 because unlike cardiovascular disease, there is no biomarker we can assess, and the relative

00:28:05.480 impact of these mutations tends to be quite high. Now, to be clear, most people will have a negative

00:28:11.900 result here, and while that negative result can be reassuring, it's important that we don't lose

00:28:16.680 sight of the fact that 95% of cancers that arise are still somatic mutations. There is one important

00:28:25.160 technical point I want to make here, because it's one of the most common sources of false

00:28:30.060 reassurance I encounter. Consumer genetic tests oftentimes only test a few different cancer

00:28:37.540 predisposing mutations. The original 23andMe test, for example, assessed only three pathogenic

00:28:44.500 mutations in BRCA1 and BRCA2, but there are thousands of known pathogenic variants in these

00:28:51.220 genes. So if a patient tells me that their 23andMe results came back negative for BRCA, that does not

00:28:58.280 really tell me that they don't have an important cancer mutation. It just tells me that they don't

00:29:03.220 have one of the three more well-studied ones. For meaningful cancer genetic risk assessment,

00:29:09.640 you need clinical-grade panel testing, not a consumer genotype product. So cancer is one of

00:29:16.440 the clearest areas where germline genetic testing can be very useful, but only when it is used to

00:29:23.940 answer the right question, with a very clear understanding of what the test is and is not

00:29:29.240 covering. Neurodegenerative disease is a very different category, because here the balance

00:29:35.560 between understanding risk and actionability is much shakier, and for obvious reasons,

00:29:40.720 these diseases tend to be the most emotionally complex for patients to deal with. The most

00:29:46.980 familiar example is, of course, ApoE, something we've talked about a lot on this podcast over

00:29:54.460 the years. ApoE4 is the strongest common genetic risk factor for Alzheimer's disease,

00:30:00.860 and it can shift risk in a meaningful way. Individuals with two copies of ApoE4 may have

00:30:08.240 risk of Alzheimer's disease up to 15 times higher than someone who does not have the mutation.

00:30:14.980 But it is still not destiny. Not everyone who is homozygous for ApoE4, meaning has two copies,

00:30:22.260 develops Alzheimer's disease, and roughly 50% of people with Alzheimer's disease do not carry even

00:30:28.260 a single ApoE gene at all, let alone two copies. That said, I think that knowing ApoE status can

00:30:35.980 be useful for a few reasons. There are emerging therapeutic strategies, such as opacetrapib,

00:30:41.660 which we've also covered on this podcast, being studied that specifically show promise in APOE4

00:30:48.660 carriers based on the hypothesis that APOE4 affects lipid metabolism in both the brain and

00:30:54.880 periphery. The data are still early, but it's a plausible example of where genetic information

00:31:00.040 could begin to inform therapeutic decisions in a more personalized way,

00:31:04.860 even for diseases as ravaging as Alzheimer's.

00:31:08.280 It may also make us more aggressive about managing other modifiable Alzheimer's disease risk factors,

00:31:14.940 such as lipids and metabolic health, if we know the patient is starting from a higher baseline risk.

00:31:21.240 For some patients, particularly those with a family history of dementia,

00:31:25.340 knowing their APOE status is less about medicine and more about planning,

00:31:30.040 it is far easier to discuss finances and long-term care preferences many years before a crisis than

00:31:37.100 during one. Beyond APOE, there are rare but highly penetrant mutations that drive early-onset or

00:31:44.700 familial forms of severe neurodegenerative disease, including Alzheimer's disease, Parkinson's

00:31:50.260 disease, Huntington's disease, as I mentioned earlier, and ALS. Diseases such as Parkinson's

00:31:56.400 disease and ALS do not have one clear common genetic risk factor. Instead, only about 10%

00:32:03.260 of cases are due to known genetic mutations. And so for someone with a parent or sibling affected,

00:32:09.680 especially at an unusually young age, genetic testing may be an appropriate option.

00:32:16.440 But broad population screening rarely makes sense in this category. The prevalence of these

00:32:21.900 mutations are low, and unlike hereditary cancer mutations, the results don't yet really map onto

00:32:28.420 established interventions. So testing here is almost always a personal decision driven by a

00:32:34.220 specific family circumstance. More than any other type of disease, the value of testing here depends

00:32:40.200 on the question being asked. Are you trying to estimate Alzheimer's disease risk for curiosity's

00:32:45.820 sake? Are you trying to determine whether or not you carry the mutation for a devastating familial

00:32:51.740 syndrome? Are you trying to make treatment decisions or prevention decisions? Or are you

00:32:56.820 really asking a planning question, gaining information that may help inform long-term

00:33:01.840 financial, career, or even care decisions? All of these are valid questions, but ones that should

00:33:09.260 be very carefully considered before testing. This is also the category where the psychological

00:33:15.100 dimension of testing matters most. Here the question is not just whether the information is

00:33:20.700 true, but whether it is likely to be useful for that particular person. There are no universal

00:33:26.560 right answers. It is deeply patient-specific, and it requires an honest conversation before the test

00:33:32.900 is ever ordered. Once you move into diseases outside of the four horsemen, areas like mental

00:33:39.620 health and complex chronic conditions, the same themes we saw in ASCVD return. Genetics clearly

00:33:47.520 matter, but that does not mean that the current genetic testing changes care in a meaningful way.

00:33:53.980 The genetic influences on mood disorders, stress reactivity, and substance use are real,

00:33:59.820 but the variants identified so far explain only a modest fraction of overall risk,

00:34:05.700 and they do not yet guide treatment in a way that outperforms standard clinical care.

00:34:11.660 Of course, that may change in the future, but we are just not there yet. And that makes this space

00:34:16.680 particularly vulnerable to the hype that outruns the science, especially the functional medicine

00:34:23.300 style panels that test common variants and then build a story around detoxification,

00:34:29.720 methylation, inflammation, or neurotransmitter balance. The leap from this variant plays some

00:34:37.120 role in this pathway to therefore this is the supplement protocol you need is almost always

00:34:44.260 much larger than the evidence justifies. MTHFR is a prime example of this. Variants in this gene

00:34:51.460 are real, and they do alter folate metabolism to some degree. But the thing that often gets

00:34:56.860 left out is just how common these variants are. Up to 40% of the population carries one or two

00:35:04.560 copies of some of these variants. Think about what that means for a moment. If MTHFR variants

00:35:11.820 were driving meaningful disease, we would expect to see it clearly in population-level data.

00:35:19.380 But of course, we don't. The fact that these variants are so prevalent is itself strong

00:35:24.740 evidence that their average effect is small, because natural selection tends to weed out

00:35:30.560 variants that cause serious harm. And yet, MTHFR has become one of the most over-ordered and

00:35:37.940 over-interpreted findings in functional medicine. Patients are routinely told that their fatigue

00:35:43.800 or brain fog or anxiety or a dozen other non-specific symptoms are caused by MTHFR,

00:35:52.060 and they are placed on aggressive methylation protocols based on a variant that, for the vast

00:35:59.680 majority of people, is clinically irrelevant. The mere presence of an MTHFR variant is not

00:36:06.700 a diagnosis or an explanation for mysterious symptoms or reason to spend a fortune on specialized

00:36:13.180 just-for-you supplement stacks. A mutation can be biologically interesting without being clinically

00:36:20.200 actionable. MTHFR is perhaps the clearest illustration of that distinction in all of

00:36:26.900 clinical genetics, but it is far from the only one. In fact, once you know what to look for,

00:36:32.860 a pattern emerges. Find a common variant with at least some science, inflate its importance,

00:36:40.220 and then use it to justify a supplement protocol. Oftentimes, these variants are explained in a way

00:36:47.060 that is akin to personality tests or astrological signs. The description is just broad enough that

00:36:53.940 practically everyone can read it and say, wow, that describes me perfectly, without realizing

00:36:59.700 that the information applies to everyone. COMT and other neurotransmitter-related genes are

00:37:05.740 another favorite for this playbook. A common COMT variant is often used to tell people that

00:37:12.560 they are a fast or slow metabolizer of dopamine, that it explains their personality, their stress

00:37:19.220 response, and, yet again, which supplements they should take. There is real biology here. Many of

00:37:25.700 these gene variants do affect dopamine metabolism, but the jump from a common context-dependent

00:37:32.800 variant to a bespoke personality profile or supplement protocol is wildly overconfident.

00:37:40.140 Then there are the so-called detox panels, tests that assess common variants in cytochrome

00:37:47.300 P450 enzymes and related pathways and present them as a window into your liver's ability to

00:37:54.660 quote, handle toxins. The premise sounds scientific, and the results often come back

00:38:01.140 with color-coded charts implying that your detox pathways are somehow compromised and in need of,

00:38:07.820 wait for it, supplemental support. What these reports usually leave out is that the liver's

00:38:13.880 detoxification machinery is extraordinarily redundant. When one pathway is less active,

00:38:21.060 others often compensate. There is no recognized clinical syndrome of poor detox in otherwise

00:38:29.380 healthy people based on common variants in these genes. The body has been solving this problem for

00:38:35.800 a very long time without personalized supplement stacks. Neutrogenomic tests that are marketed as

00:38:43.140 a way to identify your perfect diet based on your DNA function similarly. There are some studies

00:38:49.660 linking mutations in genes like FTO to better outcomes from certain diets, but these DNA-specific

00:38:56.620 diets often only perform as well as, if not worse, than conventional wisdom. You don't need a fancy

00:39:04.840 genetic test to tell you to eat fruits and vegetables and get regular exercise. For these

00:39:10.580 functional tests, we should be very skeptical, not of genetics itself, but of the claim that

00:39:16.420 current testing can turn common variants into recommendations that are more specific, more

00:39:22.060 reliable, or more effective than good clinical care. In many of these settings, it simply cannot.

00:39:29.820 One final area worth mentioning is pharmacogenetics. Unlike the kinds of functional medicine tests that

00:39:37.600 I just described, pharmacogenetics is a very real and potentially very impactful way to utilize

00:39:43.240 genetic information. Rather than asking about risk for disease, these tests address something

00:39:49.620 that is far more practical. How might I respond to a medication, and could that help guide which

00:39:56.200 one I choose? This is a very different kind of question, and genetics tend to perform much better

00:40:02.080 when the question is that specific. This is especially relevant in areas where treatment

00:40:08.020 is often trial and error, where side effects may be severe, or where metabolism varies

00:40:14.200 meaningfully from one person to another. If someone has already struggled with medication

00:40:19.480 tolerability, or if there are several reasonable treatment options and no obvious reason to choose

00:40:25.660 one over the other, pharmacogenetic information may help refine the decision. For example,

00:40:31.920 Plavix is one of the most commonly prescribed antiplatelet medications. In order for this

00:40:37.420 drug to work, it needs to be activated by a specific enzyme called CYP2C19. About 10% of

00:40:45.100 the population have gene variants that make this enzyme completely non-functional, meaning their

00:40:51.840 body can't convert Plavix into a usable compound. For someone who needs reliable platelet inhibition

00:40:59.080 after a stent or another vascular event and carries one of these loss-of-function mutations,

00:41:04.680 they can instead be prescribed a different drug that does not require CYP2C19.

00:41:11.720 A very different example is HLA-B58 and the drug very commonly used to treat uric acid

00:41:19.660 called allopurinol. Here the issue is not whether the drug will work, but whether it's safe to give

00:41:25.900 the drug at all. Patients who carry HLA-B58 are at a substantially increased risk of developing

00:41:33.220 a potentially life-threatening hypersensitivity reaction to the drug.

00:41:38.520 In fact, the effect of this is so clear that testing for this gene prior to giving allopurinol

00:41:45.400 has become standard of care for us in our practice, and hopefully for any other physician

00:41:50.100 out there listening.

00:41:51.600 Other genes can help guide decisions about dosages or determine whether an entire class

00:41:57.940 of drugs, as opposed to one particular medication, may be contraindicated.

00:42:03.220 Pharmacogenetics does not necessarily dictate the answer, but it can help inform it.

00:42:07.820 This is a more modest claim than what is often promised in the broader genetic marketplace,

00:42:13.440 but it is also a much more defensible one.

00:42:16.100 So if there's one place where inherited genetic testing may be most clinically useful for

00:42:21.080 otherwise common conditions, it may be here, not in trying to predict disease in a broad

00:42:26.640 abstract way, but in helping optimize a specific treatment decision.

00:42:31.500 Okay, now that we've gone through the major disease categories, I think it's helpful to

00:42:36.540 zoom out and see how they could compare side by side. We've put a summary table in the show notes,

00:42:42.560 and I'd encourage you to take a look, but let me walk you through the important takeaways.

00:42:47.300 There are really two axes that matter here. How large is the effect of the genetic variant,

00:42:54.160 meaning does it have a dramatic change on risk or only nudge it slightly, and how much does

00:42:59.440 knowing about it actually change what you do clinically. When you lay things out this way,

00:43:05.680 hereditary cancer panels covering things like BRCA and Lynch syndrome sit squarely in the upper

00:43:12.040 right quadrant, high effect size and high action ability. On the opposite end, consumer variant or

00:43:19.440 SNP tests for things like MTHFR or COMT sit in the lower left of this 2x2. Low effect and

00:43:28.920 virtually no response clinically to the intervention. And most other categories fall

00:43:35.800 somewhere in between. Pharmacogenetics, for example, has a moderate effect size,

00:43:40.540 but a relatively high actionability. It may not tell you whether or not you get a disease,

00:43:45.160 but it can meaningfully change how a disease is treated. APOE is interesting because the effect

00:43:51.540 size is real, but actionability remains somewhat limited, or at least it does for now. The main

00:43:57.280 point is that these two dimensions don't always move together. A variant can have a large biological

00:44:02.540 effect but still not change what you do, and a variant with a more modest effect can still be

00:44:08.320 highly useful if it shifts specific clinical decisions. So let's say you've worked through

00:44:14.400 this framework and you have a specific question and you've determined that genetics is actually

00:44:19.560 the right tool to answer it. You've thought through what you want to do with the results

00:44:24.360 and you're mentally prepared for whatever comes back. The next question is, which test?

00:44:30.120 And this is where a lot of people and a lot of clinicians get tripped up, because one of the

00:44:35.760 biggest mistakes people make is assuming that all genetic tests are more or less interchangeable,

00:44:41.500 as though getting genetic testing is a single thing. But it isn't. There are many different

00:44:47.560 kinds of genetic tests, and they differ enormously in what they measure, how much of the genome they

00:44:54.260 cover, how reliable they are for a given question, and how clinically useful the results are likely

00:45:01.080 to be. The key principle, then, is that the type of test you choose should be determined by the

00:45:08.220 question you are trying to answer. More specifically, it should be determined by how much of the genome

00:45:14.500 you actually need to look at to answer the question reliably. And in general, you want the test that

00:45:21.540 captures what you need without unnecessary additional data that is likely to generate

00:45:26.400 confusion rather than clarity. There's a natural temptation here to assume that more is better.

00:45:32.220 And for some people, in some cases, a comprehensive sequencing-based test may, in fact, be a good

00:45:38.480 option. But for many of us, that much data is overwhelming. The human genome, as we've discussed,

00:45:44.160 is enormous. The data files used to meaningfully interpret that sequence can exceed 100 gigabytes.

00:45:52.120 Collecting more of it does not automatically mean we will get more insight.

00:45:56.140 Sometimes it just generates more noise.

00:45:58.980 With that in mind, let's walk through the main categories, from narrowest to broadest.

00:46:04.100 Single gene or single mutation tests sit at one end of the spectrum.

00:46:08.940 They look for a specific variant, usually because clinical presentation or family history

00:46:14.060 already points towards it.

00:46:15.680 If your mother carries a known BRCA1 mutation and you want to know whether you inherited

00:46:21.000 it, that is a very specific question, and a targeted test is the right choice. It's precise,

00:46:27.340 relatively inexpensive, and gives you a clean answer. This is genetics at its best. Narrow

00:46:32.580 question, specific test, interpretable result. The limitation, of course, is that it only answers

00:46:38.300 the question you asked. If the question is too narrow, you may miss something important that

00:46:42.880 lies just outside of the test's scope. Genotyping arrays are the technology underlying most

00:46:50.500 direct-to-consumer products. They scan for hundreds of thousands of common single nucleotide

00:46:58.020 polymorphisms, or SNPs, known positions in the genome where people commonly differ from one

00:47:05.520 another. These tests can be useful for ancestry and physical traits, but because they only look

00:47:11.860 at common variants, they will miss rarer but far more clinically significant mutations.

00:47:18.060 A negative result on a consumer SNP test can create false reassurances because these tests

00:47:24.520 capture only a narrow slice of the variants that may matter clinically.

00:47:29.420 A related but distinct concept worth addressing here are polygenic risk scores.

00:47:35.700 Rather than reporting individual SNPs, a polygenic risk score aggregates the effects of thousands

00:47:42.600 of common variants across the genome into a single composite score meant to reflect overall

00:47:49.020 genetic predisposition to a given disease relative to the population. The appeal here is obvious. It

00:47:55.540 sounds like it should be more information than any single variant, and at the population level,

00:48:00.580 these scores can capture real signal. But at the individual level, the evidence is quite

00:48:06.100 underwhelming. This is an active and genuinely interesting area, but it is still very early

00:48:12.400 stage. When paired with other tests or analyses, such as in the myriad MyRisk test, they may help

00:48:19.740 to further stratify risk, but for now, I don't find these tests particularly useful on their own.

00:48:26.320 Gene panels are up next. Rather than scanning the whole genome for common variants,

00:48:31.420 A panel sequence is a defined set of genes known to be relevant to a specific condition or disease category.

00:48:40.180 A hereditary cancer panel, for example, might include BRCA1, BRCA2, PALB2, CHECK2, Lynch syndrome genes, and dozens of others,

00:48:49.980 all sequenced with enough depth to detect rare high-impact variants, not just the common ones.

00:48:56.160 Pharmacogenetic panels work similarly, covering the key metabolic genes relevant to drug response.

00:49:03.080 Panels tend to be the right tool when you have a specific clinical question and a defined set of

00:49:08.840 genes that are well-established as relevant to that question. They are more expensive than SNP

00:49:15.060 tests, but often covered by insurance when there is a clinical indication and the results are far

00:49:20.700 more meaningful for health decisions. Whole exome sequencing and whole genome sequencing

00:49:26.240 sit at the broadest end of the spectrum. Whole exome sequencing covers all protein coding regions

00:49:33.000 of the genome, roughly 1.5% of total DNA, but the region where the majority of known disease-causing

00:49:39.460 mutations occur. Whole genome sequencing covers everything, including the non-coding regions,

00:49:46.360 though our ability to interpret variants in those regions remains quite limited.

00:49:52.280 Both generate enormous amounts of data, and interpretation is highly dependent on the

00:49:57.820 quality of the sequencing analysis. These tests are most appropriate for unexplained or complex

00:50:03.580 presentations, a patient with a rare disease that hasn't been characterized, or a situation

00:50:08.420 where a panel has come back negative, but clinical suspicion remains high.

00:50:13.300 For most routine health questions, sequencing can likely answer the question, but it may provide more information than is needed.

00:50:21.940 It can generate incidental findings and create more questions than answers.

00:50:26.400 We've also put a comparison table in the show notes that lays out each of these test types, what they measure, and where they're most appropriate.

00:50:35.800 Let me highlight where I think the most important distinctions lie.

00:50:39.420 The biggest mistake I see people making is treating a consumer SNP test as though it were a clinical-grade gene panel.

00:50:48.240 These are fundamentally different tools.

00:50:51.200 A SNP test is scanning for common variants.

00:50:54.240 It's good for ancestry, it's fun, but it's not designed to answer clinical questions about disease risk.

00:51:00.300 A gene panel, by contrast, is sequencing specific genes in depth looking for rare high-impact

00:51:07.380 mutations that the SNP test will miss entirely.

00:51:10.660 That's why a negative BRCA result on a consumer test is not the same as a negative result

00:51:15.560 from a hereditary cancer panel.

00:51:18.400 On the other end, whole exome and whole genome sequencing gives you the most amount of data,

00:51:23.200 but that doesn't automatically make them the best choice.

00:51:26.020 more data means more incidental findings, more variants of uncertain significance,

00:51:30.940 and more interpretive complexity. For most people with a defined clinical question,

00:51:36.100 a well-chosen panel is going to give you a cleaner, more interpretable answer than sequencing

00:51:41.600 everything and hoping the signal emerges from the noise. One practical point that can be confusing

00:51:48.580 is that these categories are not completely mutually exclusive in what they can find.

00:51:54.700 The same pathogenic variant might be detectable on a targeted test, a disease-specific panel,

00:52:00.360 or a whole genome or exome sequence. So the choice is not simply about whether a variant

00:52:06.400 is theoretically on the test. It's about whether the test is designed and validated to answer your

00:52:12.260 specific question reliably, and whether you only want to answer that question or that answer plus

00:52:19.980 a larger amount of additional information. In other words, broad tests can often include

00:52:25.640 the narrow answer, but they also bring more data beyond the scope of your question.

00:52:31.220 For any genetic test that will inform a meaningful medical decision, I would strongly encourage

00:52:36.540 using a CLIA-certified laboratory with demonstrated expertise in the relevant area.

00:52:42.540 The Clinical Laboratory Improvements Amendments, or CLIA, regulations set a minimum standard for

00:52:49.220 laboratory quality, which at least tells you the lab completing the tests has been inspected and

00:52:54.880 approved for human samples. Beyond that, you also want a lab that has deep experience in the

00:53:00.500 specific domain. A lab that specializes in hereditary cancer genetics is going to give you

00:53:05.280 a more reliable and better contextualized result than a general purpose sequencing facility.

00:53:12.040 And before you order anything, look carefully at the data privacy policies. Genetic data is

00:53:18.300 uniquely sensitive. It is permanent. It is shared with biological relatives who may or may not

00:53:24.180 consent to data sharing. And the downstream implications of how it is stored and used

00:53:29.520 are worth understanding before you hand it over. And finally, make sure you understand exactly

00:53:34.740 what a test does and does not cover. Not at a general level, but specifically for that test.

00:53:40.820 What mutations does it detect? What does it miss? What will a negative test actually mean?

00:53:47.600 These are questions worth asking explicitly and ideally working through with a clinician

00:53:53.160 or genetic counselor before the test is ordered, not after the results come back.

00:53:58.360 Now, of course, ordering the test is actually the easy part.

00:54:02.160 Once the results come back, the real question is what to actually do with them.

00:54:06.760 A genetic result is not like most other lab values.

00:54:10.020 You cannot simply glance at it in the portal and move on.

00:54:12.880 The more consequential the test, the more deliberate the follow-up needs to be.

00:54:17.460 And ideally, you have already thought through the major possibilities before the test was

00:54:22.600 ever ordered.

00:54:23.820 One thing worth noting up front, a negative result is not always a clean bill of health.

00:54:28.500 It means no pathologic variant was found on the specific test ordered, which is useful

00:54:33.740 and sometimes very useful, but it does not override a strong phenotype or family history.

00:54:39.000 and it does not mean something wasn't missed simply because it wasn't tested for. A negative

00:54:44.140 result deserves the same careful interpretation as a positive one. With that in mind, I find it

00:54:49.300 useful to sort genetic findings into a few broad categories. The first is a result that confirms

00:54:54.860 something already suspected. If a lab test or family history suggests genetic condition,

00:55:01.300 such as familial hypercholesterolemia, genetic testing can confirm its presence or absence.

00:55:06.640 This may not change clinical management, but it can increase confidence in the diagnosis, solidify the plan, and inform testing for other family members in addition to provide coverage for medication.

00:55:18.240 The second and most valuable is a result that identifies a novel but actionable risk, something that wasn't necessarily on the radar before the test, but that points to a clear next step that may not have otherwise been considered.

00:55:30.580 These can be from a test that was performed specifically to answer this question, such as someone who doesn't have a strong family history of cancer, but opts to complete a hereditary cancer panel or an incidental finding from a broader test.

00:55:44.980 For these results, knowledge of risk can inform more advanced cancer screening, and a clear action plan can be made.

00:55:51.620 The third category is a result that adds context but not necessarily new action, a variant associated with a structural cardiac condition in a patient who already had a normal echocardiogram, for example, or a metabolic risk factor that is already being tracked through phenotype.

00:56:12.180 Not every finding demands a new intervention.

00:56:15.780 The fourth and most difficult is a result that points to a risk with no RCT-level action plan

00:56:23.280 available. I think dementia risk in patients is probably one of the most common examples we see

00:56:29.820 here. We don't really have validated screening tests or even well-established preventive

00:56:36.040 strategies. Of course, there are many things that we think there are compelling and suggestive data

00:56:42.580 for, but it's not quite at the same level of cancer screening for a woman with a BRCA mutation.

00:56:49.020 The value here is less about established medical action and potentially more about being on the

00:56:56.280 front edge of what prevention looks like and considering more planning or even perspective.

00:57:01.680 That does not make the result less helpful, but it does shift what useful might look like.

00:57:07.300 Every result in any of these categories should ultimately come back to a question.

00:57:11.460 What now? Do we confirm a diagnosis? Do we increase our screening? Do we change treatments? Do we inform family members? Or do we simply document the finding without changing the management? If it's the last of these, I would call into question the purpose altogether. The test is just the information-gathering step. The clinical value comes entirely from what happens next.

00:57:34.600 So, if I had to compress all of this into a single answer to the question, should I be doing

00:57:41.020 genetic testing, my obvious response now would be, it depends. Some patients want all available

00:57:47.940 information about their health, full stop, and for them, comprehensive testing may be worth it,

00:57:54.220 even knowing its limitations. Others have a specific clinical question where more

00:57:59.640 constraint testing is the right tool, and some are perfectly content to leave it alone. All of those

00:58:06.100 positions are reasonable. But regardless of where you land, the framework is still the same. Start

00:58:11.880 with the question, determine whether genetics is the right tool to answer it, choose the test that

00:58:16.840 matches the question, and think through what you'll do with the results before they arrive.

00:58:21.900 To make this more concrete, I think there are a few buckets we can use to think through genetics.

00:58:28.660 The best use case, by far, is for something like BRCA.

00:58:33.540 These mutations are highly penetrant with clear actionability.

00:58:37.320 Most people do not have these mutations, but for those who do, learning about them can be life-saving.

00:58:44.220 Genetics, at its worst, are the direct-to-consumer-style tests that are marketed for health purposes,

00:58:49.260 the tests that look at common, low-effect variants like COMT and MTHFR

00:58:54.960 and treat them as gospel for justifying supplement protocols that evidence simply doesn't support.

00:59:01.860 Most aspects of health are going to sit somewhere in the middle, where genetic testing can be

00:59:06.200 informative but may not be quite as clearly actionable or with as much supporting evidence.

00:59:11.620 For patients with questions about risk that can't be answered with lab testing,

00:59:16.740 such as predicted medication response, genetic testing can sometimes offer insight.

00:59:21.640 I think ApoE deserves its own place within this category. It isn't highly actionable in the

00:59:28.980 traditional sense, but that certainly doesn't make it useless. For patients with ApoE4,

00:59:33.860 we may be more aggressive in reducing other risk factors for Alzheimer's disease,

00:59:38.300 such as aggressive managing of lipids, promoting greater insulin sensitivity,

00:59:42.880 and early adoption of treatments like GLP-1 agonists. Or it may serve as the behavioral

00:59:48.380 lever to help keep a person motivated and stick with a lifestyle intervention. We may not use

00:59:54.600 this information the same way we would use pharmacogenetics, but it can matter for stratifying

00:59:59.780 risk and long-term planning. Beyond these categories, the clinical utility for genetics

01:00:04.740 is less clear. Seeking out genetic information purely out of curiosity is not an illegitimate

01:00:11.260 reason to test, provided you recognize that you may not get more clarity from the tests.

01:00:17.440 Genetic testing is a tool, and like every tool we have in medicine, it has real strengths and

01:00:22.920 real limitations. And its value depends almost entirely on how thoughtfully it is used. It is

01:00:29.700 not a blueprint. It does not tell you everything. And it will sometimes raise more questions than

01:00:36.040 it answers. But when the question is clear, the test matches the question, and the answer changes

01:00:42.400 something meaningful, that is when genetic testing earns its place. That is when it stops being just

01:00:49.440 an interesting data point and starts being genuinely useful. The principle I'd leave you

01:00:55.180 with is simple. Test with intention. Know what you're looking for, know what you'll do when you

01:01:01.360 find it out, and know what you will do if you don't. Everything else follows from that.

01:01:06.800 Thank you for listening to this week's episode of The Drive. Head over to peteratiamd.com

01:01:13.980 forward slash show notes if you want to dig deeper into this episode. You can also find me

01:01:20.000 on YouTube, Instagram, and Twitter, all with the handle peteratiamd. You can also leave us

01:01:25.580 review on Apple Podcasts or whatever podcast player you use. This podcast is for general

01:01:31.620 informational purposes only and does not constitute the practice of medicine, nursing,

01:01:35.500 or other professional healthcare services, including the giving of medical advice.

01:01:40.320 No doctor-patient relationship is formed. The use of this information and the materials linked to

01:01:45.980 this podcast is at the user's own risk. The content on this podcast is not intended to

01:01:51.320 be a substitute for professional medical advice, diagnosis, or treatment. Users should not disregard

01:01:56.640 or delay in obtaining medical advice from any medical condition they have, and they should

01:02:00.920 seek the assistance of their healthcare professionals for any such conditions. Finally,

01:02:05.840 I take all conflicts of interest very seriously. For all of my disclosures and the companies I

01:02:11.100 invest in or advise, please visit peteratiamd.com forward slash about where I keep an up-to-date

01:02:18.400 and active list of all disclosures.

The Peter Attia Drive - May 18, 2026

#392 - Genetic testing： when it's valuable, how to choose the right test, and what to do with the results

Episode Stats

Summary

Transcript