# Sulforaphane

Rating: Early — Mostly animal or cell studies; few or uncontrolled human data
Type: Compound
Updated: 2026-10-15
Canonical: https://heretic.bio/articles/sulforaphane-nrf2-activation-and-clinical-evidence.html

> Sulforaphane is an isothiocyanate generated from glucoraphanin in cruciferous vegetables, richest in broccoli sprouts, and the most-studied natural activator of the Nrf2 defense pathway. Its human RCT ledger is real but mixed: small trials show a replicated benefit on schizophrenia negative symptoms and clean pollutant-detoxication data, while prediabetes and lung-chemoprevention trials missed their primary endpoints. The practical catch is conversion: glucoraphanin supplements only work if your gut microbes or an added enzyme supply myrosinase.

> Sulforaphane (SFN) is an isothiocyanate from cruciferous vegetables — generated from the glucosinolate **glucoraphanin** by the enzyme **myrosinase** when plant tissue is damaged. Broccoli sprouts are the richest source (Fahey/Talalay, PNAS 1997, PMID [9294217](https://pubmed.ncbi.nlm.nih.gov/9294217/)). It is the most-studied natural activator of the **Nrf2 (NF-E2-related factor 2) cytoprotective pathway** , with a human RCT literature now spanning autism, schizophrenia, ADHD, diabetes, airborne-pollutant detoxication, and cancer-chemoprevention biomarkers. This file maps that literature with this library's evidence hierarchy — including the nulls and missed endpoints.

Sulforaphane was absent from this library until now despite being one of the few supplements where investigators have run repeated **double-blind RCTs in humans with psychiatric and metabolic endpoints** — arguably closer to a low-dose pharmacological Nrf2 agonist than a "vegetable extract."

* * *

## Mechanism

### 1\. Nrf2–Keap1–ARE activation (the primary axis)

  * SFN is an electrophilic isothiocyanate that covalently modifies cysteine sensors on **Keap1** , releasing **Nrf2** to translocate to the nucleus and bind Antioxidant Response Elements (Dinkova-Kostova et al., PMID [12193649](https://pubmed.ncbi.nlm.nih.gov/12193649/)).
  * Transcriptional output: **phase-II detoxication and antioxidant enzymes** — glutathione S-transferases (GSTs), NQO1, heme oxygenase-1 (HO-1), γ-glutamylcysteine synthetase (GSH synthesis), UDP-glucuronosyltransferases.
  * This is **hormetic** in the strict sense: a mild electrophilic provocation that upregulates endogenous defense capacity, mechanistically parallel to the stressor-adaptation logic in [Overview - Biohacking and Optimization](biohacking-and-optimization-overview.html) and [Cold and Heat Therapy](cold-and-heat-therapy.html).
  * Demonstrated in human airway: oral SFN dose-dependently increased phase-II enzymes in nasal epithelium (Riedl et al., PMID [19028145](https://pubmed.ncbi.nlm.nih.gov/19028145/)).



### 2\. Beyond Nrf2

  * **HDAC inhibition** : SFN metabolites (sulforaphane-cysteine) inhibit histone deacetylases — epigenetic chemoprevention mechanism (Myzak lineage, PMID [16652377](https://pubmed.ncbi.nlm.nih.gov/16652377/)). The schizophrenia trialists explicitly cite HDAC inhibition as a rationale (PMID [39832347](https://pubmed.ncbi.nlm.nih.gov/39832347/)).
  * **NF-κB suppression** and downstream inflammatory cytokines (reviewed PMID [40686501](https://pubmed.ncbi.nlm.nih.gov/40686501/)).
  * **Anti-AGE/RAGE activity** — isothiocyanates interfere with advanced-glycation signaling (PMID [39770075](https://pubmed.ncbi.nlm.nih.gov/39770075/)), mechanistically adjacent to [Advanced Glycation and Crosslink Breakers](advanced-glycation-and-crosslink-breakers.html).
  * **Hepatic gluconeogenesis suppression** — Nrf2-dependent downregulation of glucose-production genes (preclinical + T2D context; tested clinically below).



### 3\. Geroprotection framing

  * Reviews now position cruciferous isothiocyanates as candidate **geroprotectors** acting on oxidative stress, inflammation, and proteostasis (Adv Pharmacol 2025, PMID [40716930](https://pubmed.ncbi.nlm.nih.gov/40716930/)); neuroprotective mechanisms reviewed in PMID [40686501](https://pubmed.ncbi.nlm.nih.gov/40686501/) and 31334360.



* * *

## Bioavailability — the part most users get wrong

Form | Content | Conversion | Notes  
---|---|---|---  
Fresh broccoli sprouts | Glucoraphanin + endogenous myrosinase | Efficient; chewing releases both | ~30–100× glucoraphanin of mature heads (PMID [9294217](https://pubmed.ncbi.nlm.nih.gov/9294217/)); dose varies by cultivar/seed  
Glucoraphanin supplement (most common) | Precursor only | Requires **gut-microbiota myrosinase** | Conversion highly variable between individuals (see microbiome section)  
SFN-rich extract | Pre-formed sulforaphane | None needed | Chemically labile; heat-sensitive  
Glucoraphanin + mustard-seed myrosinase | Precursor + exogenous enzyme | Restored to near-sprout levels | Mustard myrosinase significantly increased SFN bioavailability (Sci Rep 2026, PMID [41692762](https://pubmed.ncbi.nlm.nih.gov/41692762/))  
  
  * Plasma quantitation requires thiol-blocking LC-MS/MS methods (PMID [37584212](https://pubmed.ncbi.nlm.nih.gov/37584212/)) — an analytic reason older supplement studies report wildly variable kinetics.
  * **Microbiome dependency** : baseline gut microbiota predicts systemic SFN exposure. Responders in the Gothenburg prediabetes trial carried a _Bacteroides_ -encoded transcriptional regulator (gene operon) required for glucoraphanin→SFN conversion; operon abundance correlated with serum SFN (Nat Microbiol 2025, PMID [39929977](https://pubmed.ncbi.nlm.nih.gov/39929977/)). Ties directly to [Gut Microbiome and Postbiotics](gut-microbiome-and-postbiotics.html) — your sprout supplement may be a no-op without the right converter organisms.
  * Cooking destroys myrosinase: boiled broccoli behaves like a glucoraphanin-only supplement.



* * *

## Human Evidence (RCTs and trials, by condition)

### Neuropsychiatry — the most active 2024–2026 frontier

Trial | Population | Design | Result | PMID  
---|---|---|---|---  
Singh 2014 (PNAS) | 44 young men, moderate-severe ASD | DB-RCT, 50–150 µmol SFN/day, 18 wk + 4 wk  | ABC and SRS improved vs placebo; benefit **reverted after discontinuation** | 25313065  
Singh 2021 (Mol Autism) | Children with ASD | DB-RCT + metabolomics | Identified SFN-responsive metabolite fingerprints | 34034808  
Children ASD RCT 2023 (Nutrients) | Children with autism | Prospective RDB | Positive behavioral signal | 36771424  
Huang 2025 (J Clin Psychiatry) | 77 schizophrenia, add-on to antipsychotics | DB-RCT, high-dose SFN (1,700 mg glucoraphanin-tablets ≈ Avmacol Extra Strength) × 24 wk | PANSS negative symptoms ↓ (p=.01); **d = 0.8** at wk 24; not mediated by depression/cognition | 39832347  
Chen 2025 (J Psychiatr Res) | 42 chronic schizophrenia (same program, [NCT04521868](https://clinicaltrials.gov/study/NCT04521868)) | DB-RCT × 24 wk | PANSS 5-factor negative score ↓ (p=.007); **cognition (MCCB) null** | 40133020  
Kassar 2025 (BMC Psychiatry) | 5 RCTs | **Meta-analysis** | Negative symptoms improved; cognition not | 41184790  
Eur Arch Psychiatry 2026 | Nrf2 activators (SFN trials included) | **Meta-analysis** | Signals maintained across Nrf2-activator class | 42154264  
Ghannadi 2025 (Clin Neuropharmacol) | 63 children (6–11 y) ADHD, add-on to methylphenidate | DB-RCT, 30 mg/d × 8 wk | ADHD-RS total/inattention/hyperactivity ↓, teacher- and parent-rated, **Cohen's d ≈ 1.0–1.4** ; side effects = placebo | 41604557  
  
**Interpretation:** the schizophrenia negative-symptom signal is replicated across two independent analyses of the Xiangya program plus meta-analytic aggregation — one of the few nutritional compounds with a consistent psychiatric effect size (d = 0.8). Cautions: Chinese single-center programs dominate, sample sizes are small (42–77), and cognition is consistently null. ASD data rest on a small famous trial plus follow-ons (systematic review: PMID [33013262](https://pubmed.ncbi.nlm.nih.gov/33013262/)).

### Cognition / dementia prevention

  * 42-month DB-RCT pilot (n=25 completers aged 63–90, 30 mg/day glucoraphanin): significant group×time interaction favoring GLR on Memory Performance Index (p=.012); immediate and delayed recall better (p<.001, p=.012); MCI subgroup drove the effect (p=.029). **Funded by KAGOME Co. with funder involvement** — pilot-grade only (Front Nutr 2026, PMID [41669080](https://pubmed.ncbi.nlm.nih.gov/41669080/)).



### Metabolic

Trial | Population | Result | PMID  
---|---|---|---  
Bahadoran 2012 | 81 T2D patients, broccoli sprout powder 5/10 g vs placebo × 4 wk | ↓ fasting insulin, ↓ HOMA-IR (dose-dependent) | 22537070  
Dwibedi 2025 (Nat Microbiol) | 74 drug-naive **prediabetes** , BSE vs placebo × 12 wk | **Missed primary endpoint** (target −0.3 mmol/L; achieved −0.2 mmol/L, p=.04); responder subgroup (mild obesity, low IR) −0.4 mmol/L; responder status predicted by _Bacteroides_ myrosinase-gene operon | 39929977  
NAFLD pilot 2026 | Women with NAFLD, broccoli sprouts ± Pilates | Liver-enzyme improvement | 42228225  
Muscle recovery 2026 | Eccentric-damage model, glucoraphanin DB-RCT | Short-term recovery endpoints | 41754227  
  
The prediabetes null-with-subgroup is the honest headline: sulforaphane is **not** a general glucose drug; hepatic-gluconeogenesis effects appear phenotype- and microbiome-dependent (diabetes-complications nutraceutical review: PMID [37703962](https://pubmed.ncbi.nlm.nih.gov/37703962/)).

### Detoxication of airborne pollutants (the strongest pharmacodynamic data)

  * Qidong, China crossover (n=50): glucoraphanin-rich and SFN-rich broccoli-sprout beverages increased urinary excretion of pollutant-derived mercapturic acids (benzene, acrolein metabolites) — sustained, dose-tractable (Kensler 2012, PMID [22045030](https://pubmed.ncbi.nlm.nih.gov/22045030/)).
  * Dose-dependent benzene detoxication confirmed in a dedicated RCT (Am J Clin Nutr 2019, PMID [31268126](https://pubmed.ncbi.nlm.nih.gov/31268126/)). This is the cleanest human demonstration of induced phase-II metabolism by any dietary compound.



### Cancer chemoprevention (biomarker trials — mixed)

  * **Prostate** : SFN and metabolites measurable in human prostate tissue after oral dosing (PMID [36014767](https://pubmed.ncbi.nlm.nih.gov/36014767/)); chemopreventive biomarker activity in men scheduled for biopsy (PMID [31155953](https://pubmed.ncbi.nlm.nih.gov/31155953/)). Earlier PSA-doubling-time trials (Cramer 2001; Alumkal 2010) were mixed and are not PMIDs-cited here — treat prostate indication as biomarker-grade only.
  * **Breast** : 54 women pre-biopsy, glucoraphanin supplement modified biomarkers in blood and breast tissue (Atwell 2015, PMID [26511489](https://pubmed.ncbi.nlm.nih.gov/26511489/)).
  * **Lung (2025, the key null)** : 43 former smokers, 95 µmol/day × 12 months, bronchial biopsy endpoints. **Histopathology: no significant effect.** Ki-67 proliferation index reduced; apoptosis markers unaffected (Cancer Prev Res 2025, PMID [40041932](https://pubmed.ncbi.nlm.nih.gov/40041932/)). Twelve months of high-dose SFN did not reverse bronchial dysplasia — the most important negative in the file.
  * Systematic review of SFN in breast cancer progression/radiation outcomes exists (PMID [40013196](https://pubmed.ncbi.nlm.nih.gov/40013196/)) but is low-certainty.



### Airway / asthma

  * Improved bronchoprotective response in asthmatics via Nrf2-mediated gene expression (PMID [26369337](https://pubmed.ncbi.nlm.nih.gov/26369337/)); broccoli-sprout beverage RCT increased airway antioxidant gene expression in polluted environments (PMID [27130714](https://pubmed.ncbi.nlm.nih.gov/27130714/)); neutrophilic airway inflammation proof-of-concept (PMID [27450419](https://pubmed.ncbi.nlm.nih.gov/27450419/)).



### Stroke / neuroprotection (drug-grade delivery)

  * **SFX-01** (stabilized SFN, Evgen Pharma) 300 mg BD × 28 days after subarachnoid hemorrhage: multicenter DB-RCT (SAS study) — safe, SFN measurable in plasma **and CSF** ; vasospasm/delayed-ischemia efficacy not established from the published primary results (Transl Stroke Res 2025, PMID [39028412](https://pubmed.ncbi.nlm.nih.gov/39028412/)). A pharma-grade SFN molecule reaching CSF is itself a proof-of-concept for CNS delivery.



* * *

## Animal and In-Vitro Evidence (separate tier)

  * STZ-diabetic rats: SFN partially normalized glycemia, insulin sensitivity, TAG, urea/creatinine — but **worsened total cholesterol and transaminases** at the same doses (J Med Food 2012, PMID [22925073](https://pubmed.ncbi.nlm.nih.gov/22925073/)). A rarely-mentioned counter-signal worth keeping in the ledger.
  * Nrf2/KEAP1/ARE in diabetic kidney disease — extensive mechanistic base (PMID [36139066](https://pubmed.ncbi.nlm.nih.gov/36139066/)).
  * Alzheimer's models: preclinical neuroprotection plausible via Nrf2 + anti-amyloid mechanisms (PMID [33805772](https://pubmed.ncbi.nlm.nih.gov/33805772/)); no human AD trial.
  * Immune enhancement review (PMID [33535560](https://pubmed.ncbi.nlm.nih.gov/33535560/)) — mostly in-vitro.



* * *

## Evidence Ledger (honest accounting)

Claim | Verdict  
---|---  
Induces human phase-II enzymes | **Strong** (airway biopsies, pollutant-metabolite excretion)  
Schizophrenia negative symptoms (add-on) | **Promising, replicated small** (d=0.8; meta-analyzed; single-program risk)  
ASD behavior | **Promising, small** (one classic + follow-ons; reverses on discontinuation)  
ADHD (add-on) | Single RCT, large effects, needs replication  
Prediabetes glucose lowering | **Null on primary endpoint** ; subgroup responders only  
Lung-chemoprevention histopathology | **Null** (12-mo biopsy trial)  
Prostate/breast chemoprevention | Biomarker-grade only  
General cognition | Pilot-grade (funder-run, n=25)  
Cognition in schizophrenia | **Consistently null** (MCCB)  
  
## Safety and Conflicts

  * Tolerability consistently placebo-level; main adverse events are mild GI (39929977). No severe signals in the RCTs above.
  * **Thyroid** : comprehensive systematic review found no significant effect of Brassica vegetables on thyroid function in humans (PMID [38612798](https://pubmed.ncbi.nlm.nih.gov/38612798/)) — the goitrogen fear is mostly in-vitro/iodine-deficient-animal territory. Prudence in overt hypothyroidism remains reasonable.
  * **Conflicts are structural** : trial materials often supplied by supplement makers (Avmacol/Nutramax in the schizophrenia program; Lantmännen in Gothenburg; KAGOME funded the cognition pilot with funder involvement; Fahey/Talalay held Brassica-related IP disclosed in 39929977). Academic enthusiasm and commerce are entangled here as usual — rate accordingly.



* * *

## Practical Context

  * **Dose context from trials** : 50–150 µmol SFN/day (Singh ASD); ~30 mg glucoraphanin/day (cognition pilot, ADHD trial); up to 1,700 mg/day of glucoraphanin-containing extra-strength tablets (schizophrenia); 95 µmol/day (lung trial); 300 mg BD stabilized SFN (SFX-01). No single validated "optimal" dose exists; trials cluster around the equivalent of a few dozen mg glucoraphanin with myrosinase, or tens of µmol preformed SFN.
  * **Form matters more than brand** : glucoraphanin-only products depend on your microbiome (39929977). Sprout powder + mustard myrosinase (41692762), fresh sprouts, or SFN-rich extracts bypass that dependency.
  * Fresh 3–4-day-old broccoli sprouts: a practical whole-food route; glucoraphanide content varies by seed lot — dose consistency is inherently poor.
  * Heating: light steaming preserves glucoraphanin while destroying myrosinase (boiling leaches glucosinolates). Adding raw mustard powder to cooked broccoli is the kitchen-level implementation of 41692762.
  * Effects on Nrf2-dependent enzymes are pharmacodynamic (measurable in days); psychiatric trials ran 8–24 weeks before full separation — this is not an acute nootropic.
  * Any evaluation should separate **systemic exposure** (microbiome-dependent) from **pharmacodynamic response** — without myrosinase conversion, the compound never reaches target.



## Notes — Open Questions

  * Does the schizophrenia negative-symptom effect (d=0.8) replicate outside the Xiangya/NKI program? Independent replication is the single highest-value missing study.
  * Can responder status (_Bacteroides_ myrosinase-operon abundance) be turned into a practical pre-test for metabolic indications?
  * The STZ-rat cholesterol/transaminase worsening (22925073) has no human follow-up — worth watching lipid/liver panels in long-duration users.
  * SFX-01 pipeline (Evgen) is the bridge from supplement to drug — track it as pharma-grade SFN evidence.
  * Interaction with [Autophagy](autophagy.html) and NAD+ pathways ([NAD+ and Sirtuin Activators](nad-and-sirtuin-activators.html)) is mechanistically plausible but clinically untested; combine claims cautiously.



* * *

## Sources

1. PubMed 9294217 (PubMed 9294217)
2. PubMed 12193649 (PubMed 12193649)
3. PubMed 19028145 (PubMed 19028145)
4. PubMed 16652377 (PubMed 16652377)
5. PubMed 39832347 (PubMed 39832347)
6. PubMed 40686501 (PubMed 40686501)
7. PubMed 39770075 (PubMed 39770075)
8. PubMed 40716930 (PubMed 40716930)
9. PubMed 41692762 (PubMed 41692762)
10. PubMed 37584212 (PubMed 37584212)
11. PubMed 39929977 (PubMed 39929977)
12. NCT04521868 (NCT04521868)
13. PubMed 33013262 (PubMed 33013262)
14. PubMed 41669080 (PubMed 41669080)
15. PubMed 37703962 (PubMed 37703962)
16. PubMed 22045030 (PubMed 22045030)
17. PubMed 31268126 (PubMed 31268126)
18. PubMed 36014767 (PubMed 36014767)
19. PubMed 31155953 (PubMed 31155953)
20. PubMed 26511489 (PubMed 26511489)
21. PubMed 40041932 (PubMed 40041932)
22. PubMed 40013196 (PubMed 40013196)
23. PubMed 26369337 (PubMed 26369337)
24. PubMed 27130714 (PubMed 27130714)
25. PubMed 27450419 (PubMed 27450419)
26. PubMed 39028412 (PubMed 39028412)
27. PubMed 22925073 (PubMed 22925073)
28. PubMed 36139066 (PubMed 36139066)
29. PubMed 33805772 (PubMed 33805772)
30. PubMed 33535560 (PubMed 33535560)
31. PubMed 38612798 (PubMed 38612798)