Where the Kavalactones Go: A Mass Balance of the American Kava Beverage

How to read this paper — The partition estimates are exploratory. They are computed as a residual, they inherit every error upstream of them, and in one cultivar they came out slightly negative — a mathematical impossibility and an honest signal of where the model's precision runs out. The direction of the finding is well supported. The exact percentages are not.

A root assay is not a serving claim

Kava has been prepared as an aqueous beverage across Oceania for at least three thousand years. The pharmacologically relevant constituents are the six major kavalactones, whose relative proportions define a cultivar's chemotype, written as a six-digit rank string. Kavain is the constituent most closely tied to the characteristic acute effect and has a documented interaction with GABAₐ receptors.

Consumption in the United States has grown rapidly and is no longer confined to Pacific Islander communities. Almost all of that kava is imported. Domestic cultivation is new, and there has been no published account of whether American-grown material behaves like Pacific material when it is actually prepared as a drink.

The deeper problem is that the trade's standard measurement does not describe the product. Kavalactone content is routinely determined on dried, ground root by organic solvent extraction. Those figures characterize the raw material well. They do not tell a preparer how much kavalactone ends up in a cup, because traditional preparation is a mechanical process in water, not a solvent extraction. Studies that measured finished beverages directly report widely varying per-serving content, and the classic Fijian work of Duve and Prasad established decades ago that a large fraction of the available material never leaves the root.

The question nobody had closed — Measuring the beverage tells you what came out. It does not tell you what stayed behind, and it does not tell you what physical form the delivered material is in. Kavalactone aqueous solubility is low and differs by more than an order of magnitude between compounds — so if so little can dissolve, how is a strong drink possible at all? Answering that requires weighing and assaying every fraction, including the spent root, on the same material.

The method: short warm-water extraction

SWWE is not an invention of this paper — it is a written-down version of what American kava bars already do with fresh-frozen root. Fixing it in writing is what makes it testable.

1. 454 g of frozen homogenate thawed at 23 °C for approximately ten hours.
2. Combined with 1,250 mL of filtered tap water at 38 °C, adjusted to pH 8.0.
3. Macerated in a Vitamix 5200 (1,800 W) on high for 30 seconds.
4. Transferred to a 150 µm nylon straining bag and kneaded by hand for roughly 40 seconds, until flow slowed to about one drop per second.
5. Repeated for three cycles with fresh water each time — 3,750 mL added in total. Measured final beverage volume was 4,000 mL; the excess is moisture released from the thawed plant material itself. A serving is 125 mL throughout.

The spent root left in the bag — the marc — was dried at 38 °C for 48 hours to 3.0% moisture or less, verified on a moisture analyzer, then weighed and submitted for assay alongside the beverage and the raw material.

One condition, not a comparison — Warm slightly alkaline water was used because it is common practice among American preparers, not because this study tested it against alternatives. No comparison of temperature or pH was performed and none is claimed.

The mass-balance framework

Three quantities were measured directly for each run: the kavalactone mass in the starting material, in the finished beverage, and in the dried marc. Extraction efficiency is beverage over starting mass; retention is marc over starting mass; closure is the extent to which the two sum to unity.

The partition model then splits the delivered kavalactone. Dry-matter pass-through — the proportion of starting dry mass not recovered in the marc — was measured gravimetrically per run. Multiplying it by the starting kavalactone mass gives the particle-associated kavalactone, KLₚ. Whatever remains after subtracting KLₚ from the measured beverage total is the non-particulate fraction, NPF.

What NPF is, and what it is not — NPF is a residual, not a measurement. It absorbs everything the subtraction does not capture: genuinely dissolved kavalactone, colloidal and emulsified material fine enough to pass 150 µm, and the accumulated error in KLₚ. Calling it “water-soluble” would be wrong, and this paper does not. The model also assumes solids crossing the mesh carry kavalactone at bulk concentration — and since kavalactones concentrate in the root cortex and in resinous material that may fragment preferentially, that assumption is unlikely to hold exactly.

Analysis

All analyses were performed by Flora Research Laboratories under job identifier J25-0930-H, with HPTLC and HPLC completed on 9 October 2025. Kavalactones and flavokavains were determined by HPLC-UV on an Agilent 1200 RRLC with a Poroshell 120 SB-C18 column at 55 °C, detecting at 240 nm for methysticin, dihydromethysticin, kavain and dihydrokavain and 355 nm for yangonin and desmethoxyyangonin; eight-point calibration from 5 to 250 µg/mL, R² ≥ 0.999. Reported values are the mean of duplicate injections.

The material: identified, beverage grade, and not anomalous

Identity of each lot was confirmed by high-performance thin-layer chromatography against an authenticated Piper methysticum reference and a kavain standard, following the identity test of the British Herbal Pharmacopoeia. All three lots matched the reference profile. Two are noble-classified by chemotype: Kali-Hiwa at 463251 and Kali-Rogu at 423561, the latter the profile conventionally described as Borogu. Kali-ISA returned 254631, an Isa-type profile led by dihydrokavain and dihydromethysticin rather than kavain, and is described as such throughout rather than as noble.

Table 1. Starting material: total kavalactone and flavokavain content of the dried root, with chemotype. Kali-Hiwa laterals were a separate single preparation.

• Kali-ISA — Total KL (%): 14.29 · Total FK (%): 0.354 · Chemotype: 254631
• Kali-Hiwa — Total KL (%): 7.09 · Total FK (%): 0.133 · Chemotype: 463251
• Kali-Hiwa (laterals) — Total KL (%): 10.64 · Total FK (%): 0.234 · Chemotype: 463251
• Kali-Rogu — Total KL (%): 9.98 · Total FK (%): 0.162 · Chemotype: 423561

Total kavalactone content ranged from 7.09% for Kali-Hiwa to 14.29% for Kali-ISA — spanning the range typical of Pacific commercial material. Flavokavain content was low in all three, from 0.133% to 0.354%, including in the Isa-type lot, consistent with beverage-grade rather than tudei-type material.

Flavokavain load relative to kavalactone content, computed from Table 1 and checked against the harmonized multinational beverage limits.

• Kali-ISA — Total FK (% w/w): 0.354 · FK as % of total KL: 2.48% · Beverage-grade limits: Passes both
• Kali-Hiwa — Total FK (% w/w): 0.133 · FK as % of total KL: 1.88% · Beverage-grade limits: Passes both
• Kali-Hiwa (laterals) — Total FK (% w/w): 0.234 · FK as % of total KL: 2.20% · Beverage-grade limits: Passes both
• Kali-Rogu — Total FK (% w/w): 0.162 · FK as % of total KL: 1.62% · Beverage-grade limits: Passes both

Derived here from the Table 1 figures rather than reported in the paper. The harmonized multinational kava standards set the beverage limit at not more than 0.4% w/w combined flavokavains and not more than 5% of total kavalactone content; all four lots clear both, the tighter of the two by roughly half.

How much comes out, and how strong the drink is

Warm-water extraction recovered most of the available kavalactone. Overall transfer across the study was 85.1%. Kali-ISA was the most complete and the most consistent at 90.4 ± 1.2%, Kali-Rogu close behind at 84.1 ± 0.5%, and Kali-Hiwa both lower and far more variable at 80.7 ± 7.3%. The single lateral-root run gave 78.7%. Between 9.5% and 19.3% of the starting kavalactone remained in the marc — a substantial reservoir and a real loss to the preparer.

Extraction efficiency by lot

Figure A.

• Kali-ISA — 90.4 %
• Kali-Rogu — 84.1 %
• Kali-Hiwa — 80.7 %
• Kali-Hiwa laterals — 78.7 %
• Overall — 85.1 %

Percentage of the kavalactone mass in the starting root that reached the beverage. Triplicate lots shown at their mean; laterals n = 1.

Table 2. Extraction efficiency, dry-matter pass-through and per-serving potency by lot. Triplicate values are mean ± SD of three process replicates from one lot.

• Kali-ISA (n = 3) — Extraction efficiency (%): 90.4 ± 1.2 · Dry-matter pass-through (%): 71.1 ± 1.4 · Total KL per 125 mL (mg): 447 ± 5.8
• Kali-Hiwa (n = 3) — Extraction efficiency (%): 80.7 ± 7.3 · Dry-matter pass-through (%): 78.3 ± 5.1 · Total KL per 125 mL (mg): 205 ± 19
• Kali-Hiwa laterals (n = 1) — Extraction efficiency (%): 78.7 · Dry-matter pass-through (%): 74.4 · Total KL per 125 mL (mg): 334
• Kali-Rogu (n = 3) — Extraction efficiency (%): 84.1 ± 0.5 · Dry-matter pass-through (%): 71.7 ± 1.3 · Total KL per 125 mL (mg): 228 ± 1.2
• Overall — Extraction efficiency (%): 85.1 · Dry-matter pass-through (%): — · Total KL per 125 mL (mg): 188–452 (range)

Efficiency was not uniform across the six kavalactones. In every cultivar dihydrokavain transferred most completely and yangonin least — 78.0% against 88.4% in Kali-Rogu, 76.4% against 86.5% in Kali-Hiwa, and a narrower 88.5% to 92.3% in Kali-ISA. That ordering tracks published aqueous solubility, which runs from 8.1 mg/100 mL for dihydrokavain down to 0.3 for yangonin at 21 °C, but the effect is modest against a large particle-transported background.

Against the published beverage literature

Placed alongside the small body of work that measured finished beverages rather than root, the servings produced here are strong. Normalized to a common 125 mL, traditional Hawaiian preparations give approximately 89 mg, traditional-method infusions approximately 115–173 mg, and Micronesian sakau approximately 174–250 mg. The present range of 188–452 mg sits at or above the top of that span.

Total kavalactone per 125 mL serving, against prior beverage studies

Figure B.

• This paper (American) — Low end of reported range: 188 mg · High end of reported range: 452 mg
• Micronesian sakau [19] — Low end of reported range: 174 mg · High end of reported range: 250 mg
• Aqueous infusion [18] — Low end of reported range: 115 mg · High end of reported range: 173 mg
• Traditional Hawaiian [17] — Low end of reported range: 89 mg · High end of reported range: 89 mg

Normalized to 125 mL where the original serving volume differed. Indicative only: preparation methods, root-to-water ratios and analytical methods differ across all four studies. The Hawaiian figure is a single value, so its low and high ends coincide.

Table 8. Per-serving total kavalactone content compared with prior beverage-focused studies, normalized to 125 mL where the original serving volume differed.

• This paper — Material: Kali-ISA, Kali-Hiwa, Kali-Rogu · Method: Warm-water, 150 µm · Serving as reported: 125 mL · Per 125 mL: 188–452 mg
• Brown et al. 2007 — Material: Traditional Hawaiian kava · Method: Cold-water, traditional · Serving as reported: 250 mL, 178 mg · Per 125 mL: ≈ 89 mg
• Jhoo et al. 2006 — Material: Water infusion, traditional method · Method: Aqueous infusion · Serving as reported: 240 mL, 220–332 mg · Per 125 mL: ≈ 115–173 mg
• Balick & Lee 2002 — Material: Micronesian sakau · Method: Traditional · Serving as reported: 125–180 mL, ≈ 250 mg · Per 125 mL: ≈ 174–250 mg

Where the kavalactone physically is

This is the central result. Of the kavalactone delivered to the cup, the particle-associated share was 78.7% for Kali-ISA, 85.3% for Kali-Rogu and 97.2% for Kali-Hiwa. The corresponding non-particulate residuals were 21.3%, 14.7% and 2.8%; the lateral-root preparation split 94.6% to 5.4%. A one-way ANOVA across lots gave F(2,6) = 85.3, p < 0.0001, η² = 0.97 — by a wide margin the strongest effect in the dataset.

Particle-associated versus non-particulate kavalactone in the finished beverage

Figure C.

• Kali-ISA — Particle-associated (KLₚ): 78.7 % · Non-particulate residual (NPF): 21.3 %
• Kali-Rogu — Particle-associated (KLₚ): 85.3 % · Non-particulate residual (NPF): 14.7 %
• Kali-Hiwa — Particle-associated (KLₚ): 97.2 % · Non-particulate residual (NPF): 2.8 %
• Kali-Hiwa laterals — Particle-associated (KLₚ): 94.6 % · Non-particulate residual (NPF): 5.4 %

Share of the kavalactone delivered to the cup. The two series sum to 100% within each lot. The Kali-Hiwa split should be read as “very high, upper bound uncertain” — see the caution below.

The drink is therefore a suspension in which the great majority of the active material is carried on solids fine enough to pass a 150 µm mesh, not dissolved in the water. This resolves the apparent contradiction between low kavalactone solubility and high beverage potency: potency does not depend on dissolution.

Two cautions, and they are not small — First, in Kali-Hiwa two compounds returned slightly negative non-particulate values — −2.9% for yangonin and −0.9% for methysticin. Negative residuals are impossible, and their appearance signals that for this lot the KLₚ estimate slightly exceeded what the beverage actually contained. Kali-Hiwa also had the poorest gravimetric closure (12.4 ± 5.5%) and the highest process variability (CV 20.2% on marc dry weight against 4.9% for Kali-ISA), so the two observations are consistent: incomplete closure inflates KLₚ and deflates NPF. Second, the 97.2% figure should be read as “very high, upper bound uncertain” rather than as a precise value.

Table 3. Partitioning of delivered kavalactone between the particle-associated fraction (KLₚ) and the non-particulate residual (NPF), as a percentage of the beverage total. n = 3 per lot.

• Methysticin — ISA KLₚ: 80.4 · ISA NPF: 19.6 · Hiwa KLₚ: 101 · Hiwa NPF: −1.47 · Rogu KLₚ: 88.4 · Rogu NPF: 11.6
• Dihydromethysticin — ISA KLₚ: 79.0 · ISA NPF: 21.0 · Hiwa KLₚ: 97.8 · Hiwa NPF: 2.22 · Rogu KLₚ: 85.4 · Rogu NPF: 14.6
• Kavain — ISA KLₚ: 78.2 · ISA NPF: 21.8 · Hiwa KLₚ: 94.1 · Hiwa NPF: 5.94 · Rogu KLₚ: 83.1 · Rogu NPF: 16.9
• Dihydrokavain — ISA KLₚ: 77.0 · ISA NPF: 23.0 · Hiwa KLₚ: 90.5 · Hiwa NPF: 9.48 · Rogu KLₚ: 81.2 · Rogu NPF: 18.8
• Yangonin — ISA KLₚ: 79.9 · ISA NPF: 20.1 · Hiwa KLₚ: 103 · Hiwa NPF: −2.94 · Rogu KLₚ: 92.0 · Rogu NPF: 8.00
• Desmethoxyyangonin — ISA KLₚ: 79.4 · ISA NPF: 20.6 · Hiwa KLₚ: 99.9 · Hiwa NPF: 0.10 · Rogu KLₚ: 87.6 · Rogu NPF: 12.4
• Total kavalactones — ISA KLₚ: 78.7 · ISA NPF: 21.3 · Hiwa KLₚ: 97.2 · Hiwa NPF: 2.81 · Rogu KLₚ: 85.3 · Rogu NPF: 14.7
• Total flavokavains — ISA KLₚ: 75.7 · ISA NPF: 24.3 · Hiwa KLₚ: 91.9 · Hiwa NPF: 8.14 · Rogu KLₚ: 84.0 · Rogu NPF: 16.0

Values above 100 and below zero are reported as computed rather than truncated, because truncating them would hide the precision limit they reveal.

Chemotype by phase

Chemotype was largely conserved from root to cup. Kali-ISA gave 254631 in the raw material, in the beverage and in the particle-associated fraction alike; Kali-Rogu gave 423561 across the same three; Kali-Hiwa shifted by a single adjacent pair, from 463251 in the root to 462351 in the bulk beverage. A drinker receives, in rank terms, essentially the profile the grower selected for.

The interesting divergence is in the fractions that are not the bulk beverage. The residual marc was consistently enriched in the less water-mobile compounds — Kali-ISA shifted 254631 to 256431, Kali-Rogu 423561 to 342651. The non-particulate fraction moved in the opposite direction, toward dihydrokavain and kavain: 245631 for Kali-ISA, 245631 for Kali-Rogu and 245163 for Kali-Hiwa. All three place dihydrokavain and kavain in the top two positions — the expected ordering if aqueous solubility is what distinguishes this fraction.

Table 4. Chemotype rank strings by lot and by fraction. Digits denote 1 desmethoxyyangonin, 2 dihydrokavain, 3 yangonin, 4 kavain, 5 dihydromethysticin, 6 methysticin, in descending order of abundance.

• Kali-ISA — Raw root: 254631 · Beverage: 254631 · Particle fraction: 254631 · Non-particulate fraction: 245631
• Kali-Hiwa — Raw root: 463251 · Beverage: 462351 · Particle fraction: 463251 · Non-particulate fraction: 245163
• Kali-Rogu — Raw root: 423561 · Beverage: 423561 · Particle fraction: 423561 · Non-particulate fraction: 245631

That pattern is suggestive rather than settled. In Kali-Rogu the dihydrokavain-over-kavain ordering that drives the string is statistically supported (t(2) = 14.0, p = 0.005), but in Kali-Hiwa the same comparison is not (t(2) = 1.14, p = 0.375), and one Kali-Hiwa replicate returned 425631 instead of 245163. Kali-Hiwa is also the lot whose NPF is smallest and least reliable, so its phase-resolved chemotype rests on the thinnest evidence in the paper.

Does the accounting actually work?

A mass-balance reconstruction is only useful if it predicts what direct measurement finds. Across ten individually validated beverage runs, the mean absolute relative difference between the mass-balance estimate and the HPLC measurement of total kavalactone per serving was 2.8% (SD 2.2%; 95% CI 1.2–4.4%) — 0.4% for Kali-ISA, 3.2 ± 2.0% for Kali-Hiwa across four runs, and 4.5% for Kali-Rogu.

• 2.8% — Mean absolute difference, mass balance vs. direct HPLC (n = 10)
• 3.5 mg — Paired bias, 95% CI −1.8 to 8.7 mg, p = 0.173
• R² 0.996 — Regression fit, slope 1.00 — range-driven, not a precision claim

A paired t-test gave a mean difference of 3.5 mg (t(9) = 1.48, p = 0.173). This is an absence of detected bias, not a demonstration of equivalence; with ten runs the test could not have detected a small systematic offset. Bland–Altman gave limits of agreement from −11.0 to +17.9 mg. Flavokavain agreement was considerably looser at 7.9 ± 5.0%, as expected given the much smaller absolute quantities involved.

Table 5. Agreement between mass-balance-calculated and HPLC-measured content per 125 mL, as mean absolute relative difference (%).

• Methysticin — Kali-Hiwa (n = 4): 2.17 · Kali-ISA (n = 3): 2.60 · Kali-Rogu (n = 3): 3.38 · Pooled (n = 10): 2.7 ± 2.3
• Dihydromethysticin — Kali-Hiwa (n = 4): 2.19 · Kali-ISA (n = 3): 0.46 · Kali-Rogu (n = 3): 5.37 · Pooled (n = 10): 2.6 ± 2.6
• Kavain — Kali-Hiwa (n = 4): 5.83 · Kali-ISA (n = 3): 1.23 · Kali-Rogu (n = 3): 3.29 · Pooled (n = 10): 3.7 ± 3.5
• Dihydrokavain — Kali-Hiwa (n = 4): 5.12 · Kali-ISA (n = 3): 1.20 · Kali-Rogu (n = 3): 5.79 · Pooled (n = 10): 4.1 ± 3.2
• Yangonin — Kali-Hiwa (n = 4): 5.01 · Kali-ISA (n = 3): 1.66 · Kali-Rogu (n = 3): 4.07 · Pooled (n = 10): 3.7 ± 3.8
• Desmethoxyyangonin — Kali-Hiwa (n = 4): 2.97 · Kali-ISA (n = 3): 0.55 · Kali-Rogu (n = 3): 4.62 · Pooled (n = 10): 2.7 ± 2.4
• Total kavalactones — Kali-Hiwa (n = 4): 3.21 · Kali-ISA (n = 3): 0.44 · Kali-Rogu (n = 3): 4.47 · Pooled (n = 10): 2.8 ± 2.2
• Total flavokavains — Kali-Hiwa (n = 4): 12.5 · Kali-ISA (n = 3): 4.93 · Kali-Rogu (n = 3): 4.66 · Pooled (n = 10): 7.9 ± 5.0

The excluded run

One preparation is reported but excluded from the pooled statistics, and the reason is stated here rather than buried. During the Kali-Hiwa series, Beverage 3 was subject to a documented processing deviation — the material was mishandled during preparation, and it was not clear at the time whether that would affect the result. Beverage 4 was run as a pre-planned contingency precisely because of that uncertainty. Beverage 3 subsequently returned a 12% discrepancy against 1.3–5.8% for every other run in the series.

It is excluded on the documented handling deviation, which was recorded before its assay result was known. It is not excluded as a statistical outlier: a Grubbs' test gives G = 1.406 against a critical value of 1.481, so it does not meet the criterion for statistical exclusion. It remains in the table under its own number, and Beverage 4 remains Beverage 4 — renumbering would misrepresent the experimental record.

Table 6. Per-run validation detail for the Kali-Hiwa series, showing all preparations including the excluded Beverage 3.

• Laterals (Bev. 1, n = 1) — DW₀ (g): 127.6 · Marc (g): 32.68 · KL measured (mg): 338 · KL calculated (mg): 334 · Diff. (%): 1.3
• Bev. 1 — DW₀ (g): 114.6 · Marc (g): 30.28 · KL measured (mg): 178 · KL calculated (mg): 188 · Diff. (%): 5.8
• Bev. 2 — DW₀ (g): 114.6 · Marc (g): 18.60 · KL measured (mg): 233 · KL calculated (mg): 225 · Diff. (%): 3.6
• Bev. 3 (excluded — processing deviation) — DW₀ (g): 114.6 · Marc (g): 22.80 · KL measured (mg): 237 · KL calculated (mg): 209 · Diff. (%): 12
• Bev. 4 — DW₀ (g): 114.6 · Marc (g): 25.70 · KL measured (mg): 197 · KL calculated (mg): 201 · Diff. (%): 2.1
• Pooled (n = 4) — DW₀ (g): — · Marc (g): — · KL measured (mg): — · KL calculated (mg): — · Diff. (%): 3.2 ± 2.0

Closure and process variability

Closure — how completely the measured fractions account for the starting mass — was excellent for Kali-ISA at −0.4 ± 1.5%, moderate for Kali-Rogu at 10.1 ± 0.5%, and poor and erratic for Kali-Hiwa at 12.4 ± 5.5%. Closure quality is the single best predictor of how far a given lot's partition estimates should be trusted. Kali-ISA's are the most reliable in the study; Kali-Hiwa's the least.

Table 7. Gravimetric closure and process variability by lot.

• Kali-ISA — Closure gap (%): −0.4 ± 1.5 · CV, marc dry weight (%): 4.9 · CV, beverage solids (%): 0.2
• Kali-Rogu — Closure gap (%): 10.1 ± 0.5 · CV, marc dry weight (%): 4.6 · CV, beverage solids (%): 2.7
• Kali-Hiwa — Closure gap (%): 12.4 ± 5.5 · CV, marc dry weight (%): 20.2 · CV, beverage solids (%): 14.3
• Kali-Hiwa laterals — Closure gap (%): −2.9 · CV, marc dry weight (%): — · CV, beverage solids (%): —

Table 9. Inferential tests supporting the validation, the cross-lot comparisons and the near-tie chemotype rank positions.

• Mass balance vs. HPLC, total KL (n = 10) — Test: Paired t · Statistic: t(9) = 1.48 · p: 0.173 · Interpretation: No bias detected; limited power
• Mass balance vs. HPLC agreement — Test: Bland–Altman · Statistic: Bias 3.5 mg · p: — · Interpretation: LoA −11.0 to +17.9 mg
• Mass balance vs. HPLC correlation — Test: Regression · Statistic: r = 0.998 · p: — · Interpretation: R² = 0.996; range-driven
• Extraction efficiency by lot — Test: One-way ANOVA · Statistic: F(2,6) = 3.95 · p: 0.080 · Interpretation: Large effect (η² = 0.57); not confirmed
• Particle partitioning by lot — Test: One-way ANOVA · Statistic: F(2,6) = 85.3 · p: <0.0001 · Interpretation: Highly significant (η² = 0.97)
• DK vs. K in NPF, Kali-Rogu — Test: Paired t · Statistic: t(2) = 14.0 · p: 0.005 · Interpretation: Rank order supported
• DK vs. K in NPF, Kali-Hiwa — Test: Paired t · Statistic: t(2) = 1.14 · p: 0.375 · Interpretation: Not resolvable; provisional
• M vs. K in marc, Kali-ISA — Test: Paired t · Statistic: t(2) = 6.84 · p: 0.021 · Interpretation: Rank order supported
• DK vs. DHM in marc, Kali-ISA — Test: Paired t · Statistic: t(2) = 0.12 · p: 0.912 · Interpretation: Near-tie; explains run variation
• Y vs. K in marc, Kali-Rogu — Test: Paired t · Statistic: t(2) = 1.28 · p: 0.329 · Interpretation: Near-tie; not established

What it means

The contribution is the closed accounting itself. Previous studies have measured what a kava beverage contains, and previous studies have measured what root contains, but the two have rarely been measured on the same material with the spent root weighed and assayed as well. Doing so converts extraction from an unknown into a quantity, and it makes possible a question a beverage assay alone cannot answer: not just how much kavalactone reached the cup, but in what physical form it travelled.

The answer is that it travelled mostly as particulate. Published aqueous solubilities for the six kavalactones at 21 °C span 8.1 mg/100 mL for dihydrokavain down to 0.3 for yangonin — figures which, taken alone, could not produce a 450 mg serving in 125 mL of water by any amount of dissolution. Something else must be doing the work, and the mass balance identifies it. Mesh aperture, grind fineness and the mechanical energy of kneading are the primary levers on beverage strength, and water temperature and pH — the variables preparers most often argue about — act on a comparatively small share of the total. The companion white paper reached the same conclusion for a mechanized 70 µm straining system on partly overlapping material, which is a useful convergence given how different the two apparatuses are.

There is a direct analytical consequence. Any method that assays only the filtered or centrifuged liquid phase of a traditional kava beverage will systematically report a small fraction of what a consumer actually receives — on these numbers, somewhere between about 3% and 21% of it. That is a large enough error to matter for potency verification, for compliance testing, and for any exposure estimate built on beverage assays.

The phase-resolved chemotype result is more tentative but points somewhere interesting. Across all three lots the non-particulate fraction placed dihydrokavain and kavain in its top two positions, regardless of the raw-root chemotype. If that holds up, a drinker is receiving two chemically distinct deliveries at once: a bulk particulate dose carrying the cultivar's own profile, and a smaller dissolved-and-colloidal dose enriched in the two most water-mobile compounds. Kavain being one of them is not incidental, since it is the constituent most closely tied to the acute effect. Whether the two fractions differ in rate or completeness of absorption is entirely unaddressed here and would need pharmacokinetic work to answer.

Cultivar behaviour differed markedly, and the temptation to attribute that to genetics should be resisted. Kali-ISA extracted more completely, closed better and varied less than Kali-Hiwa on every measure. But each cultivar is represented by one lot, and lot-level factors — harvest handling, root-to-rhizome ratio, homogenate texture, moisture at freezing — are fully confounded with cultivar identity in this design. The ANOVA on partitioning is highly significant, but what it establishes is that these three lots behaved differently, not that these three cultivars do.

Finally, the extraction efficiencies reported here are high against the historical literature. Duve and Prasad's Fijian work reported recoveries around 22–29% under some conditions and 81–83% under others, and lower figures still circulate as rules of thumb in the trade. The 80.7–90.4% found here reflects a specific combination of frozen homogenate, high-speed maceration, warm water and three sequential cycles, and should not be read as a general property of kava.

What this changes in practice

For preparers and kava bars

Grind and mesh are the strength controls. Because the great majority of delivered kavalactone crosses the strainer as fine solid, anything that changes how much solid gets through changes potency roughly in proportion. Standardizing the straining bag and the grind standardizes most of the variability in the product. Standardizing water temperature alone does not.

A settled beverage is not a uniform beverage. If most of the active material is particulate, it separates on standing, and a serving poured from the top of an unstirred vessel will not carry the same dose as one poured from the bottom. Agitation before pouring is a dose-control step, not presentation.

There is real material left in the bag — between 9.5% and 19.3% of the starting kavalactone, even after three cycles. Whether recovering it is worth the dilution it costs is a business decision, but it should be a deliberate one.

For growers and producers

A root assay is not a serving claim. Stating that material is 14% kavalactone says nothing reproducible about milligrams per cup unless mesh aperture, grind, water volume, temperature, cycle count and agitation are all fixed and disclosed. The gap between the two numbers is precisely what this paper measures.

Selecting for chemotype works. Rank order was substantially conserved from root to beverage across all three lots, with changes confined to adjacent positions. The profile a grower selects for is broadly the profile a drinker receives — which was worth confirming rather than assuming.

American-grown material is not anomalous. All three cultivars confirmed as Piper methysticum, produced beverage-grade material with low flavokavain content — two noble by chemotype and one, Kali-ISA, an Isa-type profile — and extracted into servings at or above the strength reported for traditional Pacific preparations. Nothing in this dataset suggests that greenhouse cultivation in California yields chemically unusual kava.

For regulators and standards bodies

A specification that regulates only raw material is incomplete. The same root prepared two ways delivers materially different doses. A workable specification either fixes the preparation parameters or states the dose as a range conditional on them.

Liquid-phase-only assays substantially understate delivered dose — capturing roughly 3% to 21% of what the consumer actually receives. Any compliance testing, potency verification or exposure assessment built on such a method will be systematically wrong in the same direction. Existing quality-standardization proposals address raw-material composition; the delivered-dose question is separate and largely unaddressed.

Mass balance is auditable in a way beverage assay alone is not. Because the framework requires the spent root to be weighed and assayed, it produces an internal consistency check: the fractions either close or they do not, and when they do not, that failure is visible rather than hidden. Closure quality is reported here for exactly that reason.

Limitations

• Replication is at the level of process, not biology. Three runs per cultivar came from one characterized lot each. Nothing here can separate cultivar genetics from lot properties. Confirming the differences are genetic requires material from independent plants and independent harvests.
• The partition model is indirect. KLₚ is inferred from gravimetric dry-matter pass-through, not measured, and rests on the assumption that transferred solids carry kavalactone at bulk concentration. NPF, being the residual, absorbs every error in that inference. The negative values in Kali-Hiwa are the model announcing its own precision limit. Direct fractionation — centrifuging the beverage and assaying pellet and supernatant separately — is the single most valuable next experiment.
• The validation set is small. Ten runs support a mean absolute difference of 2.8%, which is reassuring, but the paired t-test's failure to detect bias is a statement about power as much as about the method.
• Matrix-specific recovery was not established. The HPLC method is adapted from a validated procedure, but spike-and-recovery data specific to this beverage matrix, and to the dried marc matrix, were not generated. Both differ substantially from dried root powder.
• Sampling detail is incompletely documented. The interval between homogenization and aliquot withdrawal, and whether the beverage was re-agitated immediately before sampling, were not recorded run by run. In a suspension where most of the analyte is particulate, both matter.
• One condition was tested. Temperature, pH, mesh aperture, cycle count and maceration time were fixed at values reflecting common American practice. Nothing here identifies an optimum or supports a recommendation that one setting is better than another.

What this paper does not establish — It does not establish that these cultivars differ from one another — it establishes that these three lots did. It does not establish that the partition percentages are accurate to the decimal place; it establishes that the particle-associated share is large, and gives best current estimates of how large. It does not establish that the non-particulate fraction is dissolved material — that fraction is an operational residual that also contains colloidal and emulsified matter and accumulated model error. It does not establish that warm alkaline water is better than any alternative, because no alternative was tested. And it does not address absorption, bioavailability, subjective effect or safety, none of which were measured.

Disclosures and conflicts of interest

All three authors are commercially engaged in the kava trade and are owners and operators of kava nurseries and kava businesses in California and Florida, which produce the American cultivars evaluated here. T.B. holds an ownership interest in Root of Happiness, which was the named client on the analytical reports commissioned for this study. J.B. is affiliated with Nakamal at Home and M.M. with Kavafied, both kava businesses. The plant material was grown by T.B. in the American Kava Association research greenhouse. This paper is published by the American Kava Association and the authors are members of that organization.

The authors therefore stand to benefit commercially from favourable findings about American-grown kava, and this paper does report favourable findings about it. Against that: the analytical data were produced by an independent accredited laboratory from samples submitted without any of the interpretations offered here, none of the interested parties participated in the analysis, and the limitations section states plainly where the evidence is weakest — including the negative partition values, the confounding of cultivar with lot, and the absence of matrix-specific recovery data. Readers should judge the findings on that basis.

No comparison against any competing preparation method or apparatus was performed, and no claim of superiority over manual preparation or over any other device is made or supported here. This work received no external grant funding; all materials, analytical services and laboratory costs were borne by the authors.

Use of generative artificial intelligence — During preparation of the paper the authors used Claude Opus 4.6 (Anthropic) to compile analytical data into tabular form, construct tables and check mass-balance arithmetic, and ChatGPT 5.2 (OpenAI) for grammar and language editing. The authors reviewed and edited all output and take full responsibility for the content.