PharmSee's NHS Jobs data is a 200-item sample out of a ~516-item total — roughly 38.8% of the population — because the NHS Jobs API endpoint we query hard-caps any single request at 200 items and offset pagination returns the same 200 repeated rather than advancing. We've written before about this cap, but we haven't quantified it properly. This article runs a systematic rare-keyword audit on the current sample and works out exactly what claims the sample supports and what claims it cannot.

The audit: 33 keywords, 200 postings

On 2026-04-11 we pulled the 200-item NHS Jobs pharmacist sample and scanned each posting's title and description for 33 rare-specialty keywords. Results:

Zero of 33 keywords return a hit for the four lipid/antisense specialty terms (olezarsen, inclisiran, tryngolza, antisense). Zero hits for chemotherapy, haematology, gene therapy, or CAR-T. Zero hits for 'specials', 'unlicensed', or 'biosimilar'. These are not obscure Google queries — they are central terms in modern NHS hospital pharmacy practice.

What a zero-hit result actually tells us

If we saw a 0/200 result, what's the 95% confidence interval on the true population proportion? Using the Rule of Three for zero-event binomial confidence intervals: with n=200 and 0 hits, the upper 95% CI is ~1.5% — meaning the true underlying hit rate could be as high as 1.5% and we'd still plausibly observe zero hits in a 200-item sample. At 1.5% in a 516-item population, that's up to 7.7 postings we'd expect to exist.

So when PharmSee says "zero lipid specialist postings in the NHS Jobs sample", the honest reading is: at most ~8 such postings exist in the full 516-item national feed, and plausibly zero. That is still a strong "the specialty is structurally rare" claim, but it is not "the specialty doesn't exist".

The visible-thin-invisible trichotomy

Segmenting the 33 keywords into three bands gives us a usable interpretation framework:

• Visible (4-7 hits, 2-4% of sample): aseptic, independent prescriber, GP practice, homecare, PCN. These specialty clusters are genuinely present in the national pharmacist hiring market. Scaling to the full 516: each of these keywords probably corresponds to 10-20 live national vacancies.
• Thin (1-2 hits, 0.5-1% of sample): oncology, clinical trial, radiopharmacy, ARRS. These exist but at a scale where single-posting variance dominates. Scaling to 516: each probably corresponds to 3-6 live vacancies — small enough that a different 200-sample from the same population would likely give different counts.
• Invisible (0 hits): the 16-keyword "invisible" block is our strongest sampling-level finding. Any specialty that returns 0 hits in a 200-sample is constrained to have ≤8 postings nationally. For several of these keywords (lipid, antisense, olezarsen, FCS, CAR-T, brachytherapy, cyclotron), the real count is almost certainly zero — these are hyper-specialist posts hired through internal trust progression, not NHS Jobs.

Why the invisible block matters

The invisible block is structurally more interesting than the visible block because it tells us where NHS Jobs is the wrong recruiting channel. Each category has a different explanation:

1. Ultra-rare therapeutics (olezarsen, inclisiran for FCS). These treat patient populations of 150-300 nationally. The specialist pharmacists who prescribe them are hired into pre-existing lipid clinic establishment posts, rotated from cardiology or metabolic medicine roles, and effectively never advertised on NHS Jobs. You become an olezarsen pharmacist by being in the right rotation at a tertiary centre, not by applying to an advert.
2. High-capital specialist services (cyclotron, brachytherapy, radiopharmaceutical). Fewer than 20 NHS sites in England have cyclotrons. Recruitment for these posts is bespoke — often through learned society contacts, conference networking, or direct trust headhunting. NHS Jobs is not part of the funnel.
3. Generic specialty labels (chemotherapy, haematology, oncology). These should appear in the NHS Jobs feed — except the title field on most specialty pharmacist posts says "Rotational Pharmacist" or "Specialist Clinical Pharmacist" rather than "Chemotherapy Pharmacist". Our keyword scan is getting false zeros because the specialty lives in the rotation description, not the title. Oncology's 2 hits probably reflect postings that mention the word in the description paragraph.
4. Regulatory terms (specials, unlicensed, compound). Used in practice at the ward level but almost never in job adverts, where they're abstracted into "aseptic" or "homecare" language.

What we can and can't claim

Based on this audit, here is what the PharmSee NHS Jobs sample supports and rejects:

Supported claims:

• "Aseptic pharmacist roles are a visible subset of NHS Jobs, representing ~3% of the national pharmacist vacancy feed."
• "Homecare clinical pharmacist postings exist but are thin — roughly 6 visible in the 200-sample, plausibly 15 in the full 516-item national feed."
• "Independent prescriber is a recurring specialty qualification in modern NHS pharmacist hiring, appearing in ~3% of postings."
• "ARRS/PCN pharmacist postings are present but thin, suggesting the primary-care-network pharmacist workforce is being built slowly rather than through a national hiring surge."

Rejected (too-strong) claims:

• "No lipid specialist pharmacists exist in the NHS." Wrong — up to 8 could exist in the full feed, and many more sit in rotation posts we can't identify by title.
• "PharmSee's NHS Jobs data is comprehensive." Wrong — 200 of 516 is a 38.8% sample, and rare-specialty zero-hits need a Rule of Three confidence bound before being reported.
• "Oncology pharmacist hiring is collapsing." Wrong — 2 hits in 200 is well within the expected variance of a genuine 3-4% specialty.

The cycle 15 methodology fix

Starting cycle 15, PharmSee will publish rare-keyword audit results with Rule of Three upper bounds attached whenever the raw hit count is ≤5. The format will be: "X hits in 200-sample (95% CI: ≤Y in full 516-item population)" rather than the bare "X hits" format we've used until now. This adds a small amount of statistical scaffolding that prevents the most common misreading of our sampling work — treating "zero visible" as "zero existing" when the sample is 38.8% of the population.

Separately, we are prototyping a deeper extraction that paginates NHS Jobs via source filters rather than offset (since offset is broken). That workaround could take the sample from 200 to ~450 items, which would meaningfully narrow the rare-keyword confidence bounds and surface several specialty clusters that are currently sitting below the visibility threshold. Cycle 15 or 16 expected.

The honest read for job seekers

If you are a pharmacist looking for a rare-specialty role — lipid, specialist metabolic, radiopharmaceutical, CAR-T — NHS Jobs is not the place to find it. Those roles exist but are hired through internal rotation, direct trust outreach, and learned society networks. PharmSee's live feed will show you the aseptic, homecare, and independent-prescriber tier of specialty work, but the ultra-rare tier is structurally invisible to any aggregator, including ours.

Explore the tool: NHS pharmacist jobs · Salary guides · Related: 519 NHS pharmacist vacancies dissected