NAD+ Before And After: What Was Measured, and In What

An NAD+ before and after is almost always one of three things: a pair of blood assay results, a pair of subjective energy ratings, or a pair of outputs from a biological age model. They look like comparable evidence and they fail in three completely different ways.

The blood assay is the one that deserves the most attention, because it looks the most objective and is the most fragile. It is also where the site’s usual question, what instrument produced this number, does the most work.

NAD+ before and after: the second measurement is not one thing

Start by refusing the summary. A statement that concentrations rose after an intervention describes a laboratory result. A statement that someone felt more energetic describes a self report. A statement that a biological age score fell describes the output of an algorithm. Presenting the three as interchangeable evidence of the same underlying event is the first error, and it happens before any of the individual problems below.

Each also answers a different question. The assay asks whether a marker moved. The rating asks how someone experienced a period of time. The score asks what a model predicts from a set of inputs. Only the first is about the molecule at all.

The analyte degrades between the needle and the instrument

NAD+ is chemically unstable in a collected sample. Once blood leaves the body, enzymes present in the sample continue consuming it, and the concentration begins falling immediately. How far it falls depends on how quickly the sample was processed, the temperature it was held at, whether an inhibitor was added, and how long it sat before extraction.

This is a well recognised problem in the analytical literature rather than a hypothetical one, and it has a specific consequence for paired samples. If the before sample and the after sample were handled differently, the difference between them partly records the handling. A sample processed promptly and one that waited will differ whether or not anything happened in the person.

Nothing about a printed result reveals which occurred. A number arrives with two decimal places and no account of its own provenance, and the decimal places suggest a precision that the sample handling may have destroyed before the instrument saw it.

Assay method is part of the result

There is a second layer beneath the sample. Different laboratories measure this coenzyme by different methods, using different sample preparation, different standards, and in some cases measuring different compartments: whole blood, isolated blood cells, or plasma. Those compartments do not contain the same concentrations and are not interchangeable.

A before and after pair is therefore only interpretable if both measurements used the same method, the same compartment and the same laboratory, ideally in the same analytical run. Comparing a result from one provider against a result from another compares two methods as much as two timepoints, and the difference between methods can exceed the difference an intervention would produce.

Which intervention the paired result came from

Even a technically sound paired assay carries the question that governs everything on this compound. Human before and after data showing a rise in blood NAD+ concentrations exists, is replicated, and comes overwhelmingly from trials of oral precursors: nicotinamide riboside or nicotinamide mononucleotide taken by mouth by human participants.

That is a real result about a real intervention. It is not a result about an infusion, and a page reproducing it beside a photograph of a drip has substituted one intervention for another without saying so. When a paired assay is offered as evidence, the first question is not whether the numbers moved. It is which molecule was administered, and by which route.

Self rated energy after a paid infusion

The subjective record has a different failure mode and a simpler one to describe.

A person who has paid a substantial sum, travelled to a clinic, sat for an hour or more receiving attention from staff, and been told what to expect is not a neutral instrument. Everything in that sequence pushes the after rating upward independently of the contents of the bag, and none of it can be subtracted afterwards.

Two further effects operate on top. People tend to seek an intervention when they feel worse than usual, and periods of feeling unusually bad are typically followed by improvement regardless of what was done, which biases every uncontrolled before and after in the same direction. And the people who report are the people who returned, while those who noticed nothing are absent from the record entirely.

Biological age outputs are model predictions

The third readout is the most persuasive and the furthest from a measurement. A biological age score is produced by feeding inputs into a model that was trained to predict chronological age or mortality risk in some reference population.

The output is a prediction, and it inherits everything about the model: which inputs were used, which population it was trained on, and how much of its output varies between repeat samples from the same unchanged person. That last property, test to test variability, is what determines whether a change between two scores means anything, and it is rarely reported alongside the score.

The presentation compounds the problem. A score expressed in years invites reading as though it were an age, which is a quantity everyone knows how to interpret. It is not an age. It is a number on an arbitrary scale that the model’s authors chose to express in familiar units, and a movement of a few years in that output does not correspond to anything a person has gained or lost. Two different providers running two different models on the same person will not agree, and neither result is checkable against a physical reference, because none exists.

What would make a paired record worth reading

The requirements are unglamorous and they are the same ones that make any comparison interpretable. Both samples drawn, handled, stored and analysed identically, ideally in one run. A comparison group receiving something else, so that the natural course of feeling better or worse can be separated from the intervention. Outcomes named before the first measurement, so the moving one is not selected afterwards. Assessors who do not know which group they are assessing. And every enrolled person accounted for at the end, including those who stopped.

Almost no before and after presented on this topic satisfies more than one of those. That does not make the people reporting them dishonest. It makes their records unable to answer the question the reader brought.

There is a modest version worth mentioning, because the alternative to a rigorous study is not always nothing. A person who fixes in advance what they will measure, uses one laboratory and one method throughout, records the intervention including which molecule and which route, notes everything else that changed over the same period, and reports the result whether it moved or not has produced something more honest than the usual account. It is still a single uncontrolled case and it still cannot separate the intervention from the passage of time. What it can do is avoid the retrospective selection that makes most of these records worthless.