Mapping Evidence Gap Between NMN and NR for Metabolic Outcomes: A Systematic Review, Transitivity Assessment, and Indirect Comparison Meta-Analysis

This systematic review and indirect comparison meta-analysis concludes that current evidence is structurally insufficient to reliably compare the metabolic effects of NMN and NR due to systematic asymmetries in dosing, population demographics, and assay methods, necessitating future head-to-head trials with harmonized protocols to generate interpretable comparative data.

The Gist

Imagine you are standing in a massive supermarket aisle filled with two very popular energy drinks: Drink A (NMN) and Drink B (NR). Both are marketed as "super-boosters" that will fix your metabolism, help you lose weight, and make you feel younger. They are made by different companies, sold in different countries, and cost a fortune.

Everyone wants to know: "Which one is actually better?"

You might expect scientists to have run a big, fair race between the two drinks to see the winner. But here is the twist: No one has ever run that race.

This paper is a team of researchers (from Johns Hopkins and the University of Richmond) who decided to investigate why we don't have that answer yet. They didn't just look at the drinks; they looked at the entire history of every test ever done on them.

Here is the story of their findings, explained simply.

1. The "Apples and Oranges" Problem

The researchers tried to compare the results of studies on Drink A with studies on Drink B. But they quickly realized they were trying to compare apples to oranges (or maybe, apples to a completely different fruit grown in a different climate).

Here is why the comparison was impossible:

• The Dose was Wildly Different: Imagine if every test on Drink A gave people a tiny sip (250 mg), while every test on Drink B gave people a giant bucket (1,000 to 2,000 mg). You can't say Drink B is "stronger" just because people drank more of it. It's like saying a firehose puts out more water than a garden hose, but you never tested them at the same pressure.
• The People Were Different: The tests for Drink A were mostly done on people in Japan and India. The tests for Drink B were mostly done on people in Europe and the USA. These groups have different diets, different genes, and different lifestyles. It's like testing a winter coat in Florida and a summer dress in Alaska, then trying to decide which is better for "all weather."
• The Rulers Were Broken: When they tried to measure the "energy boost" (NAD+ levels), the scientists used different measuring cups. One group used "milliliters," the other used "liters." They couldn't even add the numbers together because the units didn't match.

2. The "Ghost Race" (Indirect Comparison)

Since they couldn't run a direct race, the researchers tried a clever trick called an "Indirect Comparison."

Think of it like this:

• You have a runner, Alice, who ran against a Tortoise (Placebo).
• You have another runner, Bob, who also ran against a Tortoise.
• You want to know if Alice is faster than Bob.

Since you don't have a race between Alice and Bob, you try to guess by looking at how much faster they were than the Tortoise.

The Result: The researchers did this math for 14 different health markers (like blood sugar, cholesterol, and blood pressure).

• Did Alice beat Bob? No.
• Did Bob beat Alice? No.
• Did they even move? Not really.

The math showed no clear winner. But more importantly, the researchers said, "We can't trust this math at all."

3. Why the Math Failed (The "Very Low" Certainty)

The paper gives these results a rating of "Very Low Certainty."

Imagine you are trying to guess the winner of a chess tournament, but:

1. You only have scores from 5 games.
2. Alice played in a noisy, hot room.
3. Bob played in a quiet, cold room.
4. They used different chess sets.
5. The referee for Alice was her brother; the referee for Bob was his neighbor.

In this scenario, even if you do the math perfectly, the answer is meaningless. The researchers found that the "evidence gap" is so wide that we simply cannot tell which supplement is better right now.

4. The "Recipe" for a Real Answer

The paper doesn't just say "we don't know." It acts like a mechanic giving you a checklist to fix the car. They say that to finally answer the question, future scientists need to run a Head-to-Head Race with these specific rules:

1. Equal Doses: Give both drinks the exact same amount of "fuel" (molar equivalence). If Drink B needs 1,000 mg, Drink A needs about 1,150 mg to be fair.
2. Same People: Test them on the same type of people (e.g., people with high blood sugar or obesity), not just healthy volunteers.
3. Same Rulers: Use the exact same lab tests to measure the results.
4. Longer Time: Run the race for at least 6 months, not just 2 or 3 months.
5. No Bias: The company selling the drinks shouldn't pay for the test.

The Bottom Line

Don't buy one because you think it's "better" than the other.

According to this paper, the current science is like a puzzle with half the pieces missing and the picture on the box is blurry.

• Is NMN better? We don't know.
• Is NR better? We don't know.
• Are they both just expensive water? We don't know.

The only thing the researchers know for sure is that we haven't built the right experiment yet to find the answer. Until a proper, fair, head-to-head race is run, the "winner" remains a mystery.

Technical Summary

1. Problem Statement

Nicotinamide mononucleotide (NMN) and nicotinamide riboside (NR) are the two leading oral NAD+ precursor supplements marketed for metabolic health and anti-aging. Despite billions in annual sales and a rapidly expanding clinical trial literature, no head-to-head randomized controlled trial (RCT) has directly compared NMN and NR. Consequently, it is impossible to determine which precursor is more effective for metabolic outcomes.

Previous systematic reviews have analyzed NMN and NR separately against placebo, but none have addressed the structural incompatibility between the two evidence bases. The core problem is whether an indirect comparison (via a common placebo comparator) is methodologically feasible given the known heterogeneity in dosing, population demographics, and biomarker reporting between NMN and NR trials.

2. Methodology

The authors conducted a systematic review and indirect comparison meta-analysis following PRISMA 2020 guidelines and registered the protocol with PROSPERO.

• Data Sources: A comprehensive search of five databases (PubMed, Embase, Scopus, Web of Science, Cochrane CENTRAL) from January 2018 to May 2025.
• Eligibility Criteria:
  • Population: Adults (≥18 years).
  • Intervention: Oral NMN or NR.
  • Comparator: Placebo.
  • Outcomes: Quantitative metabolic markers (glycemic, lipid, hepatic enzymes, blood pressure, anthropometrics, NAD+ levels).
  • Study Design: Randomized Controlled Trials (parallel or crossover).
• Included Studies: 15 RCTs (5 NMN, 10 NR) involving 740 participants. Only 8 studies (4 NMN, 4 NR) contributed data to the quantitative network meta-analysis (NMA).
• Statistical Analysis:
  • Pairwise Meta-Analysis: Random-effects models (DerSimonian-Laird) for NMN vs. Placebo and NR vs. Placebo.
  • Indirect Comparison: The Bucher method was used to estimate NMN vs. NR effects via the common placebo node (Star-shaped network geometry).
  • Transitivity Assessment: A formal evaluation of effect modifiers (age, sex, BMI, dose, duration, ethnicity, disease status) to determine if the assumption of transitivity (similarity of trial populations) held.
  • Certainty of Evidence: Assessed using the GRADE/CINeMA framework.
  • Sensitivity: Leave-one-out (LOO) analyses and exclusion of high-risk-of-bias studies.
• Software: Primary analysis in Python (scipy, numpy); validation in R (metafor).

3. Key Contributions

This paper makes three distinct methodological and clinical contributions:

1. First Formal Evidence Gap Characterization: It is the first study to systematically diagnose why a reliable comparison between NMN and NR is currently impossible, moving beyond simple narrative reviews to a quantitative structural analysis.
2. Quantification of Structural Heterogeneity: The authors quantified the "transitivity threats" that invalidate indirect comparisons, specifically highlighting a massive molar dose asymmetry and geographic/ethnic segregation of trial populations.
3. Methodological Benchmarking: It demonstrates that when an evidence network is structurally incompatible, a well-conducted indirect comparison will correctly yield "Very Low" certainty results, serving as a warning against over-interpreting such data.

4. Key Results

A. Structural Heterogeneity and Transitivity Failure

The analysis revealed that NMN and NR trials are systematically asymmetric across every major effect modifier, violating the transitivity assumption required for indirect comparison:

• Dose Disparity: NR trials used doses 1.9 to 9.2 times higher than NMN trials on a molar basis.
  • NMN: 250–300 mg/day (~0.75–0.90 mmol/day).
  • NR: 500–2,000 mg/day (~1.72–6.88 mmol/day).
  • Implication: The "NR" arm in an indirect comparison is effectively a high-dose intervention compared to the low-dose "NMN" arm.
• Population Segregation: NMN trials were predominantly conducted in East Asian populations (Japan, India), while NR trials were predominantly Western (Europe, North America). This introduces confounding due to genetic variations in NAD+ metabolism enzymes and dietary differences.
• Biomarker Incompatibility: NAD+ pharmacodynamic measures were not harmonized. One NMN study reported cellular NAD+/NADH in pmol/mL, while a key NR study reported blood NAD+ in µM, rendering direct pharmacodynamic comparison impossible.

B. Pairwise Meta-Analysis (vs. Placebo)

• Glycemic Outcomes: Neither NMN nor NR showed significant effects on fasting glucose, HbA1c, HOMA-IR, or fasting insulin.
• Lipid Profile: No significant effects on total cholesterol, LDL, HDL, or triglycerides. (Note: NMN showed a non-significant trend toward triglyceride reduction, p=0.063).
• Blood Pressure & Anthropometrics: No significant changes in systolic/diastolic BP, BMI, or body weight.
• Hepatic Enzymes: No significant changes in ALT or AST.

C. Indirect Comparison (NMN vs. NR)

• Statistical Significance: Across 14 metabolically comparable outcomes, no indirect comparison reached statistical significance (all p > 0.10).
• Confidence Intervals: All confidence intervals were wide and crossed the null, reflecting extreme imprecision.
• Certainty of Evidence: Using GRADE/CINeMA, all 14 indirect comparisons were rated "Very Low" certainty.
  • Reasons for downgrading: Serious indirectness (population/dose heterogeneity), imprecision (small sample sizes), and risk of bias (industry funding, secondary endpoints).
• Sensitivity Analysis: Excluding individual studies (Leave-One-Out) did not alter the non-significant nature of the results, except for a single nominal significance in triglycerides upon excluding the Conze 2019 study (which did not survive multiple testing correction).

5. Significance and Implications

Clinical Practice

• No Evidence for Superiority: Current data does not support choosing NMN over NR (or vice versa) for any metabolic health indication. Marketing claims of metabolic superiority for either precursor are unsupported.
• Interpretation of "Null" Results: The lack of significant difference does not mean the drugs are equivalent; it means the evidence base is too structurally flawed to distinguish them.

Future Research Recommendations

The authors propose a definitive roadmap for future head-to-head trials to close this evidence gap. A valid trial must address six specific barriers:

1. Equimolar Dosing: Future trials must use molar-equivalent doses. Based on molecular weights (NMN: 334.2 g/mol; NR Chloride: 290.7 g/mol), 1,000 mg of NR is molar-equivalent to ~1,150 mg of NMN. Current NMN trials (250–300 mg) are severely under-dosed for comparison.
2. Harmonized Biomarkers: Standardized whole-blood NAD+ assays reported in µmol/L must be used as a primary pharmacodynamic endpoint.
3. Target Population: Trials should enroll metabolically at-risk populations (e.g., HOMA-IR ≥ 2.5, prediabetes) to avoid "floor effects" seen in healthy volunteers.
4. Duration: Minimum 24 weeks to allow for detectable changes in HbA1c and metabolic reprogramming.
5. Sample Size: At least 100 participants per arm to achieve adequate statistical power (80% power to detect a 5 mg/dL difference in fasting glucose).
6. Independent Funding: Trials must be funded independently of supplement manufacturers to mitigate reporting bias.

Conclusion

This study concludes that the current evidence base for NMN and NR is structurally insufficient for reliable indirect comparison. The barriers are quantifiable (dose, geography, assay incompatibility) and modifiable. Until head-to-head trials meeting the proposed rigorous criteria are conducted, evidence-based guidance on the comparative efficacy of NMN versus NR for metabolic health remains impossible.