§B3 · Grade-D exclusions
A service that will not recommend something is more credible than one that always has an answer
18 indications the library must never map to a product, each with the corpus source for why. This is a feature, not an appendix.
The rule
D — Not established — evidence absent or negative. No entry graded D may be recommended. Grade D appears in the exclusion section only.
Enforced, not merely stated: a grade-D protocol can never reach approved status, and the
recommender returns 404 NO_APPROVED_PROTOCOL for its condition code.
Two different refusals happen at this counter, and they fail differently. Conflating them is a mistake: one is a clinical-evidence boundary; the other is a legal one.
Why the grade is not a discouragement
Walt's neighbour Renée is certified for chronic pain but leads with insomnia. Primary insomnia is Grade D — not established, may never be recommended — and grade D is not a discouragement, it is a structural impossibility: POST /api/protocols/[id]/approve returns 422 for evidence_grade='D', so a grade-D protocol can never reach status='approved', and GET /api/shelf-observations/recommend returns 404 NO_APPROVED_PROTOCOL for its condition code. Priya cannot get a product list for insomnia even if she wants one.
The grade is not arbitrary. CFPC 2021 Rec 4 states cannabis "is not an appropriate therapy" for insomnia pending further research; the sleep improvement in pain RCTs is a co-benefit only (Busse 2018, BMJ — "small-to-very-small improvement in sleep quality", moderate certainty). The same D applies to anxiety disorder as the treated condition, PTSD, CINV from dispensary products, any strain-name/indica-sativa sleep claim, and terpene-based matching (build/PROTOCOL-LIBRARY-V1-sleep-anxiety-ptsd-nausea.md, Grade-D exclusion table).
Do not use — Grade D exclusion table
Chronic-pain library · 10 rows
The following pain-adjacent indications must not be mapped to products by this library. They are listed so the library cannot silently drift into claiming them.
| Indication | Grade | Why excluded (corpus-sourced) |
|---|---|---|
| Primary insomnia (sleep as the treated condition) | D | CFPC 2021 Rec 4: cannabis "is not an appropriate therapy" for insomnia pending further research. Sleep improvement in pain RCTs is a co-benefit only (Busse 2018, moderate certainty). |
| Anxiety disorder (as treated condition) | D | Same CFPC negative statement; no RCT of dispensary products for anxiety in corpus; THC anxiogenic at high doses. |
| PTSD | D | CFPC negative statement; no PTSD RCT in corpus (nabilone-PTSD data wider literature — UNVERIFIED here, deliberately not cited). |
| Glaucoma | D | Evidence AGAINST: 1 RCT of 6 patients, no benefit (Allan 2018). |
| IBD (Crohn's/UC) as disease treatment | D | Antón 2022 citing Kafil et al. Cochrane (Inflamm Bowel Dis 2020): "evidence is not robust." Palliative symptom support only, never induction of remission. |
| Opioid-dose reduction as an effect claim | D | Noori et al. SR/MA (via Busse 2018): opioid-sparing "remain[s] uncertain due to very low certainty evidence." Supported only as prescriber-led taper strategy, not as an outcome promise. |
| Any strain-name / indica-sativa-targeted claim | D | Not chemistry: Jikomes/Smith 2022 PLoS ONE (six states, labels don't track chemistry); Cyca 2023 (names invented/renamed for sales). |
| Terpene-based matching for efficacy | D | CONTESTED in-vitro conflict (Santiago 2019, Finlay 2020 vs Raz 2023), no clinical confirmation. Terpenes may be recorded as COA data; never used for matching. |
| Suppository dosing | D | No data in corpus; rectal THC-hemisuccinate literature UNVERIFIED — not staged. |
| CINV from dispensary products | D | CINV belongs on prescription synthetics (Marinol label); no modern herbal-cannabis CINV RCT in corpus. Out of scope for v1. |
Do not use — Grade D exclusion table
Sleep · Anxiety · PTSD · Nausea/CINV library · 8 rows
The following must not be mapped to products by this library. Listed so the library cannot silently drift into claiming them.
| Indication | Grade | Why excluded (corpus-sourced) |
|---|---|---|
| Primary insomnia (sleep as the treated condition) | D | CFPC 2021 Rec 4: cannabis "is not an appropriate therapy" for insomnia pending further research (cfpc-guidance2021.txt). Sleep improvement in pain RCTs is a co-benefit only (Busse 2018, moderate certainty). |
| Anxiety disorder (as treated condition) | D | Same CFPC negative statement; no RCT of dispensary products for anxiety in corpus; THC anxiogenic at high doses (memo 03 §1.6). |
| PTSD | D | CFPC negative statement; no PTSD RCT of dispensary products in corpus. Note: a preliminary nabilone-for-nightmares RCT (Jetly 2015, Psychoneuroendocrinology 51:585-8) is cited within the Health Canada review (hc-html.txt) — the underlying trial text is NOT staged; the citation is on-disk but the trial's numbers are UNVERIFIED and not used. |
| CINV from dispensary products | D | CINV evidence is labeled prescription-synthetic (Marinol 5 mg/m² per label; nabilone 0.2–6 mg/day per Busse 2018). No modern herbal-cannabis CINV RCT in corpus (memo 03 §1.2). Out of scope for v1. |
| Any strain-name / indica-sativa-targeted sleep claim ("indica for sleep") | D | Not chemistry: Jikomes/Smith 2022 PLoS ONE (six states, labels don't track chemistry); Cyca 2023 (names invented/renamed for sales). |
| Terpene-based sedation/matching (linalool-for-sleep, myrcene-for-calm) | D | CONTESTED in-vitro conflict (Santiago 2019, Finlay 2020 vs Raz 2023); André 2024 PRISMA review: exploratory signals only, no clinical confirmation. Terpenes may be recorded as COA data; never used for matching. |
| PTSD nightmare suppression as a product claim | D | Only nabilone (prescription synthetic) has preliminary RCT text even cited (Jetly 2015, unstaged); no dispensary claim is derivable. |
| Sleep-apnea or hypnotic-taper replacement | D | No corpus source. |
How the refusal is actually delivered
The wording is the deliverable. Four scripts, verbatim from the library, for when a cardholder asks for product matching on a Grade-D item.
Sleep (primary insomnia)
"For insomnia as the main problem, I've got to be straight with you: the clinical guideline the library grades against — Canadian Family Physician's cannabis guidance — says cannabis is not an appropriate therapy for insomnia until further research comes in. The sleep improvement studies show is a side-benefit in people using it for pain. What I can do: if your sleep trouble rides on your pain (or another condition you're certified for), there is a graded protocol for that, and sleep often follows. Or, if you're already using cannabis at night, I'll do a harm-reduction review — get you to a measured, lowest-effective pattern instead of escalating. What I won't do is sell you a product claiming it treats insomnia."
Anxiety disorder
"The evidence for cannabis treating an anxiety disorder is the thinnest we stock — contested, dose-dependent, and THC can worsen anxiety at higher doses. The guideline says it's not an appropriate therapy for that. If your anxiety is tied to a treated condition, we can talk about a cautious CBD-only, low-dose, stop-rule-heavy trial — and I'll tell you up front it's a trial with explicit exit points, not a promise. If it's the primary thing you're here for, the right move is your certifying practitioner or a mental-health clinician; I'm happy to coordinate the medication-safety side from here."
PTSD
"For PTSD the honest answer is: the evidence does not support me matching a product to it. The guideline includes PTSD in its 'not an appropriate therapy' list. There is a small trial of a prescription synthetic (nabilone) for nightmares — that's something your prescriber, not the dispensary, would weigh. What I can do is a full medication-interaction screen on anything you're already taking or considering, and I can be the pharmacy point-person if your clinician initiates a cannabinoid trial. But I won't recommend a dispensary product for PTSD."
Nausea/CINV
"For chemotherapy nausea, the real evidence is on the prescription side — dronabinol and nabilone are labeled drugs, used after standard antiemetics fail, and they come from your oncology team. No dispensary product has the same evidence behind it for CINV, and I'd be doing you a disservice to imply otherwise. What I bring to the table: if your oncologist does start dronabinol or nabilone, I'll run the full interaction screen against your chemo regimen and everything else you take. That's where I can actually help."
(All four scripts end with the same pivot: offer the interaction screen and the graded adjacent protocol; never end on a bare refusal.)
The referral path
Every one of the four scripts (sleep, anxiety, PTSD, CINV) ends on the same pivot, and the pivot is the product: offer the interaction screen and the graded adjacent protocol; never end on a bare refusal. For Renée that means: run the pain protocol if her certification supports it (sleep may follow as a co-benefit, stated as a co-benefit); or run a harm-reduction review of what she is already taking at night; or route her to her certifying practitioner or a mental-health clinician while Priya keeps the medication-safety side. For PTSD the referral is explicit — "that's something your prescriber, not the dispensary, would weigh" — and for CINV it is the oncology team, with Priya offering the interaction screen against the chemo regimen.
GAP-18 · blocks a pilot
A refusal leaves no record.
There is no refusal, decline, or out-of-scope entity anywhere in the schema. The most defensible act the pharmacist performs in a week — a documented, scripted, evidence-cited refusal — leaves no row. Compliance cannot audit that refusals happen; the pharmacist has no record proving she declined; per-consult billing cannot count it; and the refusal rate, which is the best single quality signal the programme has, is unmeasurable.