Three Numbers Are Already on Your Last Blood Test. Nobody Did the Math.

You know the appointment. Fasting since the night before, blood drawn, then a phone call or a portal message a week later. Cholesterol is up a bit. Let's watch it. Maybe we start you on something. See you in a year.

You went home with one number and a vague sense that you had been weighed and found wanting.

Here is what I want you to know. On that same page, in the same panel your doctor already ordered, there is a ratio nobody calculated. It costs nothing extra. It takes ten seconds with the calculator on your phone. And in the research it tracks heart risk at least as tightly as the number you were actually told about.

That is not a conspiracy. It is a habit. The lipid panel was designed to answer one question, and it has been answering that same question since the 1970s, so that is the question that gets asked.

The conversation happening away from the clinic

There is an argument running on X right now that is worth your attention, and worth your scepticism too.

Mark Kaplan has been posting his own lab work over six years and making a case that three numbers matter more than the one you get told: HOMA-IR, the triglyceride to HDL ratio, and hs-CRP. His framing is blunt. "Your doctor checks your LDL and writes a prescription. I checked two hundred markers and changed what I ate."

Sama Hoole posted a timeline of how the standard advice got written in the first place, running from the launch of Crisco in 1911 to today.

Ken D Berry MD pointed at a randomized trial that put a very low carbohydrate diet head to head against the DASH diet in adults with high blood pressure and prediabetes or type 2 diabetes.

I went and read the underlying research for all of it. Some of it holds up well. Some of it is real but oversold. One claim does not survive a check at all, and I am going to show you that one too, because you should know what my standard is before you decide whether to trust anything else I tell you.

Number one: the ratio you already own

Take your last lipid panel out of the drawer. Divide your triglycerides by your HDL.

That ratio is a rough read on whether your body is handling carbohydrate well. When insulin is working properly, triglycerides tend to sit low and HDL tends to sit higher. When insulin is not working properly, triglycerides climb and HDL falls, and the ratio between them opens up like a pair of scissors.

The study people cite here is Gaziano 1997 in Circulation. People in the highest quarter of the triglyceride to HDL ratio had a relative risk of heart attack of 16.0 compared with the lowest quarter.

Now let me tell you what that number actually is, because the honest version is more useful to you than the impressive version. It was a case-control study, 340 heart attack patients matched to 340 people without one. The confidence interval ran from 7.7 all the way to 33.1, which is a very wide range and tells you the true figure is uncertain. Blood was drawn after the heart attacks, which muddies cause and effect. And it compares the extremes, top quarter against bottom quarter, which is the comparison that always produces the biggest-looking number.

So: a strong signal from a small study, not a law of physics. What it is good for is telling you which direction you are pointing.

If you are in Canada, read this part twice. Almost every threshold you will find online for this ratio was calculated using American units, milligrams per decilitre. Your Ontario lab reports in millimoles per litre. The two are not interchangeable, and people compare their Canadian ratio to an American cut point every day and conclude they are fine when they are not.

The conversion is roughly 2.3. Take your Canadian ratio and multiply it by 2.3 to compare it with anything you read from a US source.

Worked example. Triglycerides 1.5, HDL 1.2, both in mmol/L. Your ratio is 1.25. Multiplied by 2.3, that is about 2.9 in American terms, which most sources would call a middling to poor result, not a good one.

In Canadian units, very roughly, under 0.9 is a good place to be and over 1.3 is worth a conversation. I want to be straight that those cut points are convention rather than consensus. No one has run the trial that validates a single universal threshold. Use it as a direction of travel, not a diagnosis.

Number two: the one your doctor probably did not order

HOMA-IR combines fasting glucose with fasting insulin to estimate how hard your pancreas is working to keep your blood sugar normal. That last part is the whole point. Your glucose can look perfectly fine for years while your insulin quietly doubles to keep it there. Glucose alone tells you the ship is level. Insulin tells you how many people are bailing.

Kaplan cites 65 studies and 516,325 people. That is a real paper and the figures are exact: Gast 2012 in PLOS ONE, free to read. Per one standard deviation higher, HOMA-IR carried a relative risk of cardiovascular events of 1.46, compared with 1.21 for fasting glucose and 1.04 for fasting insulin on its own.

The caveats, since you are owed them. The 65 studies cover the whole review; the HOMA-IR result itself rests on about 17 of them. The authors found signs of publication bias for exactly that measure. And these are observational studies, so this is association, not proof that lowering the number lowers your risk.

There is also a practical wrinkle. Fasting insulin is not usually part of a routine panel in Ontario and is generally not covered, so getting HOMA-IR means asking for it and likely paying for it. Ask your doctor whether it makes sense in your case. And note that insulin assays vary between labs, so a result from one lab is not cleanly comparable to another.

Number three: inflammation, and where the story gets honest

hs-CRP measures background inflammation. Kaplan's claim is that it beat LDL over 30 years in 27,939 women.

Both halves are true, but they come from two different papers. The famous 2002 study followed those women for eight years. The 30-year follow-up was published in the New England Journal of Medicine in 2024, same cohort, 3,662 cardiovascular events. Comparing highest to lowest fifth: hs-CRP carried a hazard ratio of 1.70, LDL 1.36, lipoprotein(a) 1.33.

So hs-CRP did come out ahead of LDL. But 1.70 against 1.36 is a modest margin, not a rout, and the researchers' actual conclusion was that all three together beat any one of them alone. Also worth knowing, because it is the kind of thing that gets left out: when researchers used genetics to test whether CRP itself causes heart disease, the answer came back no. CRP appears to be the smoke, not the fire. That does not make it useless. Smoke is worth knowing about. It just means chasing the number itself is not the goal.

Notice what all three numbers have in common. Not one of them is cholesterol. All three are reading the same underlying thing from different angles: how well your body is handling fuel.

Why this is not a story about effort

There is a reason I want you looking at a measurement rather than at your own record of trying.

Kaplan also posted the loop he says drives weight gain, and this one is worth understanding even if you never order a single extra test. Insulin resistance pushes the body to store fuel rather than burn it. More stored fat means more leptin, the hormone that is supposed to tell your brain you have plenty. When leptin stays high for long enough, the brain stops registering the message. You stay hungry while carrying more than enough energy on your body.

The core of that is well established. In the study that opened this field, leptin levels in people with obesity averaged more than four times those of lean people, and the authors concluded that most people with obesity are insensitive to the leptin they are already producing. Not deficient in it. Deaf to it.

Kaplan's version overreaches when he says it is not calories. Energy still has to balance somewhere. But the useful half of his point stands, and it is the half you have never been offered: the reason your appetite kept overruling your intentions was not a shortage of character. It was a signal that stopped arriving. That is a mechanism, and a mechanism can be changed. A character flaw just gets carried around.

This is also why the order of operations matters so much. Fix the fuel and the hunger signal starts working again, and everything downstream of it gets easier.

How the advice got locked in, and one claim that does not hold

Hoole's timeline is a good piece of writing and most of it checks out. Crisco really was launched in 1911 from cottonseed oil that had been an industrial nuisance. Linoleic acid in the body fat of American adults really did climb from about 9% in 1959 to about 21% in 2008 (Guyenet and Carlson 2015, free to read).

And the two trials he names are real, and they are uncomfortable ones.

In the Sydney Diet Heart Study, 458 men who had recently had a coronary event were randomized, and the group told to replace animal fat with safflower oil and safflower margarine died more often, not less. The causes of death were left out of the original 1978 publication and only surfaced when the data was recovered and reanalysed in 2013. In the Minnesota Coronary Experiment, 9,423 people were randomized to corn oil and corn oil margarine. Cholesterol fell nicely. Deaths did not, and at autopsy the intervention group had more heart attacks in the tissue, not fewer.

Both deserve their caveats. Sydney's all-cause mortality result sat right at the edge of significance. Minnesota lost roughly three quarters of its participants within the first year, which is a serious problem for any study trying to measure something slow.

Now the part I promised. The timeline says Procter and Gamble handed the proceeds of a 1948 radio contest to the American Heart Association, implying the maker of Crisco bought the body that would go on to write the fat guidelines. That one does not hold up. The AHA's own account is that Procter and Gamble sponsored the airtime for the show, and the money, somewhere between 1.5 and 1.75 million dollars, came from more than two million listeners who mailed in contributions. The money did transform the AHA from a small professional society into a national organization, which is a real and interesting fact. But the sponsor did not write the cheque.

I am flagging it because that is the claim a critic reaches for first, and when it falls it drags the true parts down with it. The accurate version of this history is strong enough. It does not need the extra push.

What actually happened when someone ran the test

Berry pointed at the trial worth knowing about: Saslow 2023 in Annals of Family Medicine, free to read in full. Adults with high blood pressure plus prediabetes or type 2 diabetes were randomized to a very low carbohydrate diet or to DASH, the diet usually recommended for blood pressure.

Over four months the low carbohydrate group came out ahead on all three measures. Systolic blood pressure fell about 4.6 mm Hg further, HbA1c 0.21 percentage points further, and weight about 8.8 pounds further.

Read the fine print with me. This was 94 people over four months, and the blood pressure result only just cleared the significance line. Both groups got a real support package, weekly sessions, cookbooks, coaching emails, self-monitoring, so this shows a low carbohydrate approach beating DASH inside a supported program, not a diet sheet working on its own. Some of the funding came from a foundation sympathetic to this way of eating, and the lead author declared a conflict involving software used in the trial. The authors themselves asked for larger and longer studies.

One more thing about that trial, and I want to be careful here. More people in the low carbohydrate arm reduced or came off blood pressure and glucose medication. That happened under medical supervision inside a clinical trial. If you eat this way and your numbers move, your medication may need to change, and that is a conversation with your doctor, arranged in advance, not something you do on your own.

What about the seed oil argument

Kaplan's six-step model has insulin resistance damaging the artery wall first and linoleic acid from seed oils making LDL easier to oxidize. The chemistry underneath it is not made up. Polyunsaturated fats have double bonds and can oxidize; saturated fats do not and cannot. Oxidized LDL really is found in plaque.

But a mechanism is a hypothesis until outcomes test it, and the outcome evidence here is genuinely split. The Cochrane review of randomized trials found that cutting saturated fat reduced cardiovascular events by about 17%, with no effect on death from any cause, and found no clear advantage to replacing it with polyunsaturated fat rather than carbohydrate. A 2017 reanalysis that separated the better-controlled trials found no benefit at all from swapping in n-6 polyunsaturated oils. A 2026 review in Annals of Internal Medicine found the benefits concentrated almost entirely in people already at high risk.

So the honest position is this. That seed oils are the single cause of heart disease is not established. That insulin resistance sits upstream of a great deal of it is on much firmer ground. If you want a practical rule that survives either outcome, cook with fats that were food before they were processed and spend your attention on the carbohydrate load and the whole food, which is where the stronger evidence points anyway.

One thing to do this week

Find your last lipid panel. It is in a drawer, or in your online portal, or your clinic can send it.

Divide the triglycerides by the HDL. If you are in Canada, multiply that answer by 2.3 before you compare it to anything you read online.

That is it. No appointment, no cost, no new regime. You will have one number that tells you something the phone call did not, and if it is not where you want it, you now know which conversation to have and which question to ask.

If you want the fuller picture of how insulin resistance builds quietly for years before anything shows up on a standard test, that is laid out here: The Insulin Loop. And if you would rather start at the plate than the lab, the Protein, Fat and Fibre Plate Guide is free.

You have probably been told for years that your problem was effort. Look at your own numbers first. Effort applied to the wrong measurement is how a decade goes by.

If you want a guide who has climbed this and will be there every step of the way, book my free discovery call. Fifteen minutes, no cost, and you leave knowing where you stand on the trail.

With thanks to Mark Kaplan, Sama Hoole and Ken D Berry MD, whose posts sent me to the papers behind this piece. Where I have disagreed with a detail, it is because I went and read the source, which is the point of publishing in the open.

Graham Kingma is a health and wellness coach, not a physician. Nothing here is medical advice, and nothing here is a reason to change or stop a prescription. Interpreting your bloodwork, ordering additional tests, and adjusting medication are all conversations with your doctor.

Source

| Gaziano 1997, TG:HDL RR 16.0 | Verified. Case-control, 340 cases, CI 7.7 to 33.1, extreme-quartile comparison. All disclosed in text. |

| Ridker 30-year, hs-CRP 1.70 vs LDL 1.36 | Verified, NEJM 2024, n=27,939. The post corrects Kaplan's conflation of the 2002 (8-year) and 2024 (30-year) papers. |

| CRP not causal | Verified, Mendelian randomization, BMJ 2011, 194,418 participants. |

| Gast 2012, 65 studies, 516,325 people | Verified exactly. HOMA-IR RR 1.46 vs glucose 1.21 vs insulin 1.04. Disclosed: HOMA-IR arm rests on ~17 studies, Egger p<0.01 for publication bias. |

| Saslow 2023 keto vs DASH | Verified. n=94, 4 months, SBP -4.59 (p=.046), HbA1c -0.21 (p=.034), weight -8.81 lb (p=.0003). Disclosed: support package in both arms, funding, COI, authors' call for larger trials. |

| Sydney Diet Heart | Verified. Corrected from Hoole's "458 on safflower oil": 458 randomized, 221 in the intervention arm, 86% post-MI, oil plus margarine. All-cause p=0.051 disclosed. |

| Minnesota Coronary Experiment | Verified. n=9,423, institutionalized population, autopsy IRR 1.90. ~75% first-year attrition disclosed. "43 years in a basement" framing dropped as exaggerated (partial results were published in 1981 and 1989). |

| Linoleic acid 9% to 21% | Verified. Guyenet and Carlson 2015, 9.1% (1959) to 21.5% (2008). |

| P&G handed proceeds to AHA | FALSE and openly corrected in the post. AHA's own account: P&G sponsored the airtime, listeners contributed the money. Amount inconsistent across AHA's own documents, so given as a range. |

| Cochrane, Hamley, Steen on SFA and PUFA | Verified. Presented evenhandedly, per the banned-claims rule against presenting AHA saturated-fat positions as settled. |

| Canadian unit conversion, factor 2.3 | Checked arithmetically. TG mmol/L x 88.57 and HDL mmol/L x 38.67 gives a ratio conversion of 2.29. Worked example verified. |