Four Years of Declining Power Data. Same Training Load. Normal Bloodwork. Then I Found the Metric That Explained Everything — And It Wasn't in Any Panel My GP Could Run.

1:06. Then 1:04. Then 1:02.

Same benchmark loop. Same bike. Same training load. Three consecutive years of getting slower on a course I have ridden five hundred times.

I am 52 years old. I have been racing amateur — cycling, some triathlon — since my late thirties. I track everything. Oura ring for HRV and sleep quality scoring. Power meter on both bikes. Garmin for GPS and interval data. Everything pulled into TrainingPeaks. Full bloodwork panel twice yearly. I am not a person who guesses.

And for three years, the data told me something I did not want to accept.

Peak power: declining. Not dramatically in the first year — roughly 8%, enough to notice on climbs but manageable. I adjusted. Added intensity work. Extended the Zone 2 base. The data did not respond. By year two, peak power was down 15% from the prior year's peak. Recovery was extending: where I had reliably been ready for a hard effort in 24 hours, the number had moved to 36, then 48. HRV trending lower across rolling 30-day averages. Sleep quality scores declining even though sleep duration was identical. I was not overtraining. Training load was unchanged. Inputs were the same. Outputs were moving the wrong way, season after season.

My GP reviewed a full bloodwork panel. Everything was within normal range. I did not expect a different answer, but I had to rule it out.

The bloodwork was normal. The performance data was not. Something upstream was broken.

I spent four months assuming detraining despite the training log showing otherwise. I spent two months adjusting recovery protocols — cold exposure, extended tapers, more structured sleep hygiene. The data moved briefly and returned to its trend. I had been competing long enough to know the difference between a training problem and something upstream of training. This was something upstream.

THE MECHANISM

The explanation was not in my bloodwork. It was in a paper published in Cell Metabolism in 2016.

The subject was an enzyme called CD38 — part of the immune system's cellular signalling apparatus, present in virtually every tissue. CD38's primary function is not NAD+ metabolism. But as a secondary function, it degrades NAD+ — nicotinamide adenine dinucleotide, the molecule that powers every mitochondrion in every muscle and every processing function in your brain. The 2016 research established that CD38 activity approximately doubles between the ages of 20 and 50.

Follow-up research published in Nature Metabolism in 2020 reinforced this in human tissue. The conclusion: the majority of age-related NAD+ decline is not caused by the body producing less NAD+. It is caused by the body destroying it faster. The drain is the mechanism. The production side is a secondary issue.

Understanding this changed how I read four years of performance data.

NAD+ is not a general wellness molecule. It is the functional substrate of mitochondrial energy production. Your mitochondria run on it. When NAD+ levels decline — whether through the CD38 drain or production shortfall — mitochondrial output decreases. Power declines. Recovery extends. The adaptation response to training weakens. And none of this appears in a standard bloodwork panel, because standard panels do not measure tissue NAD+. It shows up in the numbers that actually matter: the watts, the times, the recovery windows. It shows up as a trend line pointing the wrong way despite everything being done correctly.

This is why the data was moving in the wrong direction for three years. The engine's fuel supply was being degraded faster than the training log could reflect. The only readout was in the performance metrics themselves.

It also explained why eight months of NMN supplementation in 2023 produced no measurable effect. I had done it correctly — pharmaceutical-grade product, documented purity, appropriate dose, consistent timing. Nothing moved.

THE DELIVERY PROBLEM

Oral NMN does not deliver NAD+. This statement requires precision.

NMN is a precursor. When you swallow an NMN capsule, the molecule must survive partial gut degradation, enter systemic circulation, and then be converted through a multi-step enzymatic process before becoming functional NAD+. Each step involves losses. Each step depends on gut enzyme activity and conversion efficiency, both of which decline with age. Bioavailability research suggests that what arrives at the tissue level is substantially less than the label dose — and the gap widens every year past forty.

In a 52-year-old with elevated CD38 activity, the oral NMN problem compounds: the delivery format is inefficient, and the degradation enzyme is running at approximately double its baseline rate on the other end. You are filling a tank through a partially blocked funnel while the drain is running faster than it ever has.

Researchers in the longevity forum communities who had been working through this mechanistically arrived at the same conclusion independently over several years: the question is not which precursor to take. The question is whether precursors are the right format at all for someone dealing with both problems simultaneously.

The alternative is subcutaneous injection of NAD+ directly. The molecule does not need to survive gut absorption. It does not require enzymatic conversion. It enters systemic circulation as NAD+ because it was always NAD+. The bioavailability comparison between oral precursors and injectable NAD+ is not marginal. It is structural. You are bypassing the entire problem set.

This is the format that was producing the loop times I was watching my training partner post at 51.

THE PROTOCOL AND THE DATA

I have been using the NADPure protocol for four months.

NADPure is a subcutaneous injection pen for at-home use. Each pen ships as high-purity NAD+ powder in a pre-measured 500mg dose — batch-verified, lab certificate available on request. Powder form is not a compromise. It is the pharmaceutical standard for sensitive injectables: the compound does not enter solution until you activate it, approximately 30 seconds before use. This matters for molecular integrity in a way that pre-filled liquid pens cannot match — you are injecting freshly activated NAD+, not a solution that degraded somewhere between a production facility and your door. The injection is fine-gauge subcutaneous, administered to abdominal tissue. The sensation is mild and brief. Total time: under 90 seconds.

The data at four months:

Benchmark loop time: from 1:02 to 55 minutes. A 12% improvement over the prior year. I am posting times at 52 I was not posting at 49 or 50 or 51.

Peak power output: recovering from its three-year trough. Not yet back to the personal best from four years ago, but trending in the correct direction for the first time since I started tracking the decline.

HRV: recovery scores trending upward across rolling 30-day averages. The 48-hour recovery window between hard efforts has returned to 24 hours. This was the first metric to change — within three weeks of starting the protocol. Recovery was the leading indicator.

Sleep quality scores: stable to improved. More consistent deep sleep readings. Less night-to-night variability.

I am now putting in times on my benchmark segments that are creating friction in the club with riders seven to ten years younger. I do not mention this for competition. I mention it because when I was watching the trend line deteriorate for three years, I needed to know that the direction was reversible. That it was not the inevitable trajectory that everyone — coaches, GPs, the general narrative about ageing and performance — had been implying it was.

It is reversible. The data is the data. The power meter does not lie. The Strava segments do not lie.