Time for a proper deep dive into the SELECT trial results. Lincoff et al. published the full results in the NEJM last year, and I think the implications are still under-appreciated in both the cardiology and obesity medicine communities.[1]

Trial design overview: SELECT was a randomized, double-blind, placebo-controlled, event-driven trial. N = 17,604 adults aged 45+ with BMI >= 27, established CVD (prior MI, stroke, or PAD), but without diabetes. Randomized 1:1 to semaglutide 2.4 mg SC weekly vs placebo. Median follow-up 39.8 months.

Primary endpoint: 3-point MACE (composite of CV death, non-fatal MI, non-fatal stroke).

Key results:

• Primary endpoint: HR 0.80 (95% CI: 0.72-0.90), p < 0.001
• CV death: HR 0.85 (95% CI: 0.71-1.01) — note this did not reach significance individually
• Non-fatal MI: HR 0.72 (95% CI: 0.61-0.85)
• Non-fatal stroke: HR 0.93 (95% CI: 0.74-1.15)
• All-cause mortality: HR 0.81 (95% CI: 0.71-0.93)

The NNT over ~3.3 years is approximately 67 for MACE and 53 for all-cause mortality. For context, that's in the same ballpark as statin therapy in secondary prevention.

What strikes me most: the event curves separate early — within the first 12 months — well before maximum weight loss is achieved. This strongly suggests the cardiovascular benefit is not purely mediated through weight reduction. There may be direct anti-inflammatory or anti-atherosclerotic effects at play.

Thoughts from the cardiology-minded folks here? Is this practice-changing for you?

Excellent summary. I want to push back slightly on one point — the NNT calculation. The NNT of 67 for MACE is calculated over the full ~40-month follow-up, but the trial was event-driven and stopped when sufficient events accrued. If you extrapolate the ongoing divergence of the KM curves, the NNT over 5 years would likely be significantly lower (better).

Also worth highlighting the subgroup analyses. The treatment effect was consistent across virtually all pre-specified subgroups — age, sex, race, baseline BMI, history of MI vs stroke vs PAD, statin use, baseline hsCRP. The interaction p-values were all non-significant. That kind of consistency is exactly what you want to see in a well-powered outcome trial.

One subgroup that raised eyebrows: patients with baseline BMI < 30 showed a trend toward less benefit (HR ~0.88) compared to BMI >= 35 (HR ~0.72), though the interaction was not significant. This doesn't mean we should restrict use, but it's worth watching in future analyses.

the event curves separate early — within the first 12 months

This is the critical observation. By month 12, mean weight loss was only about 8-9%, yet the CV benefit was already apparent. Compare this to bariatric surgery data where CV risk reduction correlates more tightly with weight trajectory over years. Something else is going on mechanistically.

I ran some back-of-the-envelope mediation analyses based on the published data. If you assume weight loss is the sole mediator and use the LOOK AHEAD trial's weight-CV relationship as a reference[2], a ~15% weight loss would predict roughly a 10-12% relative risk reduction in MACE, not 20%.

The "excess" benefit (~8-10% additional RRR) likely reflects direct vascular and anti-inflammatory effects. Supporting this:

• hsCRP decreased by 38% vs placebo in SELECT — this is a massive anti-inflammatory signal
• GLP-1 receptors are expressed on monocytes, macrophages, and vascular endothelial cells[3]
• Preclinical data shows semaglutide reduces plaque inflammation and macrophage infiltration independent of weight change

The mechanistic story is still being worked out, but I think we're looking at a drug that operates through at least three parallel pathways: metabolic (weight loss, insulin sensitization), anti-inflammatory (hsCRP, IL-6), and direct vascular (endothelial function, plaque stabilization).

As a cardiologist in practice — yes, this is practice-changing. But the implementation is messier than the trial data suggests.

My biggest practical concern: in SELECT, semaglutide was added on top of guideline-directed medical therapy. 90% were on statins, 85% on antiplatelets, 80% on ACE-I/ARBs, 60% on beta-blockers. These patients were well-treated. So the 20% MACE reduction is on top of optimal medical therapy.

The real-world question: how many of my secondary prevention patients are actually on optimal therapy? Many are not. Before adding a $1,500/month GLP-1 RA, shouldn't I make sure they're on a high-intensity statin and an SGLT2 inhibitor first?

I'm not arguing against the data. I'm arguing that the clinical priority algorithm matters enormously for cost-effectiveness.

Valid point, and it's exactly what the ACC/AHA guidelines committee is wrestling with. But I'd reframe it: SELECT gives us a new tool in the toolbox. The sequencing question is important but shouldn't delay adoption for patients who are already on optimal therapy and still have residual risk.

Consider a patient who's had an MI, is on rosuvastatin 20mg, aspirin, lisinopril, metoprolol, and has a BMI of 33. Their LDL is 55 mg/dL. They're doing everything right. Before SELECT, our only option for further risk reduction was PCSK9 inhibitors (which target LDL, not residual inflammatory risk). Now we have semaglutide targeting a completely different axis of residual risk.

The all-cause mortality benefit (HR 0.81) is particularly compelling. We don't see all-cause mortality benefits from many interventions in secondary prevention beyond statins and RAAS inhibitors. This puts semaglutide in elite company.