In April I published my own LDL history: a peak of 127 mg/dL before treatment and 41 on a statin plus ezetimibe. FLOWPROMOTE's combination arm started at 127.4 and finished at 42.5. That's where the resemblance ends. I take those drugs for prevention, to keep a lifetime of LDL exposure low, and every patient in this trial was already past that point, with a coronary narrowing that limited blood flow.

This newsletter has spent the year working toward the question those patients pose. LDL causes atherosclerosis, and that argument is settled. Plaque shows up early, often in people whose risk scores call them low risk, and left untreated it keeps growing, as it did in the NATURE-CT cohort I covered in June. AMUNDSEN showed how long the benefit of a very low LDL can take to show up as fewer events. What none of those pieces could show was the artery itself responding to treatment in a randomized test of how low to go.

FLOWPROMOTE, published October 30 in JAMA Cardiology, is that test.¹ In 104 adults with newly diagnosed coronary disease who had never taken a statin long term, treatment that brought LDL to about 43 mg/dL took more plaque off the coronary arteries over 18 months than treatment that brought it to about 70. Plaque volume fell in both groups, and it fell further where LDL went lower.

In both groups, one part of the plaque grew anyway: the calcium. That's the finding most likely to be misread, by patients holding a repeat calcium score and by the clinicians who order one, and it makes sense once you see where the rest of the plaque went.

FLOWPROMOTE ran at three Danish hospitals that use CT angiography as the first test for suspected angina. To qualify, a patient needed a coronary narrowing greater than 50% on CT and a CT-derived fractional flow reserve of 0.80 or less, a computed estimate that the narrowing limits blood flow, and they had to meet predefined criteria for starting with medication rather than a procedure. Of 294 patients who met the flow criterion, 60 went straight to invasive angiography and 104 were randomized. They averaged 64 years old; 79% were men and 98% were White. None had been on long-term statin therapy, although about a quarter had started one in the weeks before the first scan (median, 25 days).

Half were assigned to atorvastatin 40 mg and half to rosuvastatin 40 mg plus ezetimibe 10 mg, open label, for 18 months. The authors call these "intensive" and "more intensive," and the first label is accurate: atorvastatin 40 mg counts as high-intensity statin therapy under US guidelines, so this was good against better, not a weak statin against a strong one.

Each patient had CT angiography at baseline, 9 months, and 18 months, with the first scan's settings repeated at every follow-up. Plaque was measured along the proximal segments of all three major coronary arteries and their larger branches with semi-automated software (Autoplaque, from Cedars-Sinai) by a reader who didn't know the patient's treatment, clinical details, or where a scan fell in the sequence. Follow-up imaging was nearly complete: 102 of 104 patients had both follow-up scans.

About 1 in 5 patients had their regimen changed along the way (9 on atorvastatin, 11 on the combination), mostly for muscle pain or raised liver enzymes, and only 2 stopped statin therapy altogether. Because the trial was open label, it can't separate drug effect from expectation for symptoms like muscle pain, the problem at the center of September's piece on statin nocebo.

Two structural points before the numbers. The trial's primary endpoint was the change in CT-derived fractional flow reserve, and on that endpoint it was neutral: the combination didn't improve the flow estimate significantly more than atorvastatin did. It did take more plaque off the narrowings themselves, and where more came off, flow tended to improve more, a correlation rather than a randomized result. Those findings appeared in JACC in February.² The plaque results in this paper are prespecified secondary endpoints.^1,3 And the Novo Nordisk Foundation funded the trial and had no role in its design, analysis, or write-up.

The combination did what it was built to do. Mean LDL fell from 127.4 to 42.5 mg/dL (65%) on rosuvastatin plus ezetimibe and from 119.7 to 69.5 mg/dL (38%) on atorvastatin.

Plaque followed. Total plaque volume fell by an average of 69.8 mm³ (21.1%) with the combination and 39.0 mm³ (13.6%) with atorvastatin, so the combination took off 30.9 mm³ more (95% CI, 4.5 to 57.2; P = .02). All of that decline came from noncalcified plaque, the softer fibrous and fatty tissue, including any lipid-rich core, that made up nearly 9 of every 10 cubic millimeters of plaque at baseline. It fell 28.2% with the combination and 20.0% with atorvastatin (difference, 28.9 mm³; P = .02), and some reduction showed up in 51 of 52 patients on the combination versus 39 of 51 on atorvastatin.

Three things about those numbers deserve a second look.

First, what's randomized and what isn't. There was no untreated arm, and there shouldn't have been: nobody would leave flow-limiting coronary disease untreated for 18 months to build a control group. The 21% and 14% are before-and-after changes within each arm, and the randomized result is the gap between them. For a sense of the untreated trajectory, go back to NATURE-CT, where noncalcified plaque roughly doubled over about 5 years in a lower-risk group taking no lipid-lowering drugs, measured with different software.⁴ FLOWPROMOTE looks like that picture run in reverse.

Second, the size of the LDL gap. The combination lowered LDL 38.6 mg/dL more than atorvastatin did, which is almost exactly 1 mmol/L, the unit the Cholesterol Treatment Trialists use. Across 26 randomized trials and about 170,000 people, each 1 mmol/L of LDL lowering reduced the annual rate of major vascular events by just over a fifth, with no threshold in the range studied.⁵ FLOWPROMOTE didn't count events, but it's the imaging counterpart to Ez-PAVE, which I covered in April: targeting an LDL below 55 mg/dL instead of below 70 cut a composite of cardiovascular events, including revascularization, at 3 years (6.6% vs 9.7%; hazard ratio, 0.67).⁶

Third, the slope. Across the whole trial, lower achieved LDL went with more regression of total and noncalcified plaque, but it wasn't associated with the change in calcified plaque. Because the arms differed by two drugs, that dose-response across all patients is what points to LDL as the common pathway, although unlike the arm-to-arm comparison it isn't protected by randomization. Most patients finished low: the 90th percentile of achieved LDL was about 85 mg/dL, and only 1 patient ended above 116. The data describe a continuous slope in that range, not a threshold.

The response was also fast. By my arithmetic from the averages in the paper's Table 3, total plaque in the combination arm was already about 53 mm³ lower at 9 months, more than the atorvastatin arm lost over the full 18. When I wrote about AMUNDSEN, the point was that plaque remodeling in the first year doesn't tell us when it starts to show up as fewer events. FLOWPROMOTE adds that the remodeling itself can begin within months, even if the event curves take years to follow.

Here's the result that will generate confused patient messages. Calcified plaque volume rose in both arms, by an average of 14.6 mm³ on atorvastatin and 12.6 mm³ on the combination, with no significant difference between them (P = .65), and achieved LDL didn't predict how much it rose.

This isn't a quirk of one trial. In a patient-level analysis of 8 intravascular ultrasound trials, coronary calcium rose in every treatment group and more with statins, including the high-intensity group whose atheroma regressed; the authors concluded that statins promote calcification independent of their effect on plaque size.⁷ In PARADIGM, plaque in statin users grew more slowly overall but calcified faster, and developed new high-risk features less often.⁸ A later analysis of the same PARADIGM registry, looking at plaque density, linked statins to shrinking low-attenuation and fibro-fatty plaque and a shift toward denser calcium.⁹

The leading interpretation is that as the lipid-rich core shrinks, what remains heals and calcifies, and dense calcium tends to travel with more stable plaque. In June I called calcium the scar tissue of atherosclerosis, and in a box marked as speculation I wrote that in treated cohorts the noncalcified components tend to stall or regress while the remaining calcium shifts toward a denser, more stable form. FLOWPROMOTE fits the first half of that, with the randomized comparison showing more regression at the lower LDL, and is consistent with the second. The stability part still rests on observational data, since no trial randomizes people to more calcium.

The clinical translation doesn't depend on settling that. FLOWPROMOTE measured calcified plaque volume, not repeat calcium scores, but a calcium score is built from the same calcium and gives denser deposits more weight. On a statin it can rise while the plaque that responds to treatment shrinks, so a higher score on treatment is not, by itself, evidence that treatment is failing. Set that beside NATURE-CT, where noncalcified plaque kept growing in people with calcium scores of 100 or less, about half of them zero, and the score's job comes into focus. It's useful for deciding whether someone needs treatment and poor at grading treatment once it has started. LDL or apoB does that job.

When I wrote about KETO-CTA in October, the problem was how easily plaque numbers bend: change the method or the reader, and the same 200 scans gave different answers. This result agrees with what I already believe, which is exactly when the same standards matter most.

FLOWPROMOTE clears most of the bars that study didn't. It randomized patients to a comparison, and its plaque endpoints were prespecified. Its reader was blinded not only to treatment but to where each scan fell in the sequence, which keeps expectations about change out of the measurement. One reader and one platform measured both arms, so whatever bias the platform carries lands on both sides of the comparison, and the trial lost almost no one to follow-up.

It still has real limits. The trial was neutral on its primary endpoint, CT-derived blood flow, so the plaque differences above are secondary findings that support the result rather than settle it.² It's an imaging study of 104 people, with too few clinical events to count: 3 nonemergent PCIs for worsening symptoms, no heart attacks, no deaths. It was open label and changed two things at once, a more potent statin and ezetimibe. The patients had flow-limiting disease, hadn't been on long-term statins, and were mostly White men, so the result doesn't automatically extend to someone who has taken a statin for a decade or to small plaque found on a screening scan, where, as I wrote about REACT, the right target is still unsettled.

Repeat measurements are also noisier than most people assume. When this group scanned 101 patients twice, an hour apart, they rated the reproducibility of plaque volumes as modest (the FLOWPROMOTE paper calls it good), though agreement was best at the whole-patient level this trial used.¹⁰ "Regression" in this paper means any decrease at all, so a count like 51 of 52 includes changes small enough to sit inside that noise. On percent atheroma volume, the metric of older ultrasound trials such as SATURN and GLAGOV, the arms moved 2 and 4 percentage points and the difference between them missed significance (P = .08). Don't set FLOWPROMOTE's 21% beside the roughly 1-point changes in those trials;^11,12 CT volumes across the major arteries and their branches and ultrasound percentages in a single segment aren't the same measurement. Finally, the plaque software comes from Cedars-Sinai, where one coauthor reports royalties and holds a patent; the paper discloses both.

None of that makes the result weak. It makes FLOWPROMOTE a well-run surrogate study pointing the same way as decades of outcome trials, which is a much stronger position than a surrogate standing alone.

FLOWPROMOTE doesn't prove that an LDL of 43 prevents more heart attacks than an LDL of 70; outcome trials such as IMPROVE-IT and Ez-PAVE carry that argument, and this one was never built to count events. What it adds is a view of the mechanism moving the way those trials predicted. In people with established coronary disease, plaque came back down on treatment, the lower LDL took more of it off, and the change was already visible at 9 months. The calcium rose in both arms, which is what treated plaque tends to look like rather than a sign that treatment failed. For a patient with coronary disease, the question isn't whether they're on a statin. It's whether their LDL is low enough to keep taking plaque off the wall.