The costs — what addiction does to the public purse, and what the public purse does to the argument

The NHS numbers — alcohol, obesity, opioids, pornography

Alcohol

The most up-to-date authoritative UK figure is the Institute of Alcohol Studies 2024 study, the first nationwide reassessment in over twenty years. Headline numbers for England, annual:

• NHS and healthcare: £4.91bn (hospital admissions, A&E, ambulance, primary care)
• Lost productivity: £5.06bn — 44 per cent of which is presenteeism rather than absenteeism
• Criminal justice and wider crime: £14.58bn
• Total societal cost: £27.44bn — a >40 per cent real-terms uplift on the 2003 Cabinet Office estimate

The older "£3.5bn NHS / £7bn productivity" figures still in circulation pre-date this and trace back to PHE 2016 and the 2003 Cabinet Office Strategy Unit work. Both are outdated. The Drinks Business pushed back on the £27.44bn methodology, mostly on the value-of-statistical-life loading inside the crime component. Methodological dispute, not bad faith.

Obesity

The "£6bn NHS / £27bn wider economy" couplet is the policy stock phrase, but it has drifted from its primary source. The £27bn wider-economy figure originates in the Government Office for Science 2007 Foresight report "Tackling Obesities: Future Choices". More recent OHID / DHSC figures put NHS spend on obesity at around £6.5bn per year. Two recent re-estimates have pushed the total UK figure much higher:

• Frontier Economics 2022 (commissioned by Novo Nordisk): total UK annual cost of overweight + obesity £98bn when productivity, quality of life and informal care are valued. Full report.
• Tony Blair Institute 2024: Unhealthy Numbers puts the comprehensive UK figure at £126bn when all indirect costs are bundled.
• IFS Report R266, 2023: The costs of obesity — a sceptical, careful breakdown.

Honest read: the £27bn figure is now nearly two decades old and methodologically loose. The Frontier and TBI updates are larger but were published by parties with an interest in GLP-1 uptake. Flag the funding when you cite them. The lab tech on the previous page is right to do so.

Tobacco (the comparator class)

ASH 2024: cost of smoking to society in England £21.8bn per year, up from the earlier £17.04bn estimate. Tobacco duty revenue is around £10bn — i.e. on ASH's framing, societal cost comfortably exceeds revenue. This is the figure Snowdon contests directly below.

Opioids

The UK has no headline equivalent. US numbers dominate the literature. Florence et al., 2021: US economic burden of opioid use disorder plus fatal opioid overdose in 2017 was $1,021bn ($471bn OUD + $550bn fatal overdose). 85 per cent of that figure is statistical-life valuation, not cash outflow. The Joint Economic Committee 2022 extrapolation to 2020 puts it at ~$1.5 trillion, 37 per cent above 2017. These are real welfare losses with no excise-revenue offset, which is why the opioid case partially breaks the Snowdon model below.

Pornography

There is no credible NHS or UK economic-cost figure for pornography use. The only "porn in the NHS" coverage is a 2010 think-tank survey of fertility-clinic spending on stimulus material at ~£700 per trust per year. Note this gap honestly. Any microsite claim of a porn-as-addiction economic burden is currently unsupportable in the peer-reviewed cost-of-illness literature.

The Snowdon counter-thesis

Chris Snowdon, Head of Lifestyle Economics at the Institute of Economic Affairs, makes a structurally consistent argument across three publications. Killjoys: A Critique of Paternalism (IEA, 2017) is the rhetorical statement. Alcohol and the Public Purse (IEA DP 63, 2015) and Smoking and the Public Purse (IEA DP 84, 2017, with Mark Tovey) are the empirical core.

The argument, stated as strongly as Snowdon would state it:

1. The "cost to the NHS" headline is gross, not net. It is not netted against the excise duty paid by the same consumers, nor against the lifetime state expenditure those consumers do not incur because they die before drawing decades of pension, social care and dementia treatment.
2. On the alcohol side: Snowdon's 2015 paper computes gross state cost of £3.9bn against alcohol-tax revenue of £10.4bn — net £6.5bn benefit to the Treasury. He frames it as "drinkers subsidise non-drinkers by £6.5bn a year".
3. On the tobacco side: the 2017 paper finds gross cost ~£4.6bn against tobacco duty ~£9.5bn, plus avoided pension and social care from early mortality, giving a net saving of ~£14.7bn per year, rising to £19.8bn in the headline number.
4. The thought-experiment generalisation: if everyone stopped drinking, smoking and overeating, the Treasury would be worse off, because excise revenue and avoided end-of-life spending exceed lifetime healthcare costs.
5. The deeper move: public-health economics systematically counts the costs of vice and ignores the savings. It is not arithmetic, it is moral accounting dressed as economics.

van Baal 2008 — the inconvenient empirical finding

The Snowdon argument is not invented. It sits on top of a genuinely peer-reviewed empirical finding that the public-health lobby has spent fifteen years trying not to discuss.

van Baal PHM, Polder JJ, de Wit GA, Hoogenveen RT, Feenstra TL, et al. (2008). Lifetime medical costs of obesity: Prevention no cure for increasing health expenditure. PLoS Medicine 5(2): e29.

Method: simulation model, Dutch cohort, age 20 baseline, three groups — healthy-living (non-smoker, BMI 18.5–25), obese non-smoker, smoker. Tracked lifetime healthcare expenditure to death.

This is the load-bearing empirical paper. Anyone arguing the lifetime-cost thesis without citing van Baal is bluffing; anyone dismissing the thesis without engaging van Baal is dodging. The accompanying PLOS Medicine perspective piece accepted the maths while arguing prevention is justified on welfare grounds, not cost grounds. That distinction — welfare versus cost — is the one the rest of this page turns on.

The opposing case — steelmanned

The standard public-health response (Sheron and Gilmore at the Lancet Commission on Liver Disease, ASH, the BMA, the RCP) does not really dispute the van Baal arithmetic. It rejects the frame.

1. Lifetime accounting prices lives at zero. Saying smokers are cheaper because they die before drawing a pension is correct only if the years of life lost are valued at nil. The CDC's $1.02 trillion opioid figure is 85 per cent statistical-life valuation — once you put any real number on a QALY, the IEA arithmetic flips.
2. Externalities are systematically under-counted in the Snowdon model. Drunk driving, domestic violence, foetal alcohol syndrome, second-hand smoke, the carer burden on family members. Productivity losses (the £5bn IAS figure for alcohol; £13bn ASH figure for smoking) fall on third parties.
3. Distributional truth: harm concentrates in deprived deciles. Treasury net-benefit hides a transfer from the dying poor to the surviving rich.
4. Compression of morbidity is not guaranteed. Smokers and obese people often die slowly and expensively, not quickly and cheaply.
5. Tax is not consent. The libertarian "they paid their duty, leave them alone" frame assumes informed adult choice — addiction, by definition, compromises that assumption.

Global cost-per-class

Two observations. First, on absolute scale, tobacco and obesity are the two large classes — each measured in trillions globally, each an order of magnitude above the opioid crisis even at its US peak. Second, the porn line is doing more work than it looks. If pornography were a public-health emergency on the scale that the most vocal critics imply, you would expect at least one credible cost-of-illness estimate. There isn't one. Twenty years in.

Tirzepatide — the actuarial maths

Efficacy

SURMOUNT-1 (NEJM 2022, n=2,539, non-diabetic, 72 weeks): mean weight loss 15.0% / 19.5% / 20.9% at 5 / 10 / 15 mg vs 3.1% placebo. Roughly 9 in 10 lost ≥5%.

SURMOUNT-5 (NEJM 2025, head-to-head): tirzepatide −20.2% vs semaglutide −13.7%. 47% greater relative loss; 19.7% of tirzepatide arm vs 6.9% of semaglutide arm lost ≥30% body weight.

SURMOUNT-4 (JAMA 2024, withdrawal): after 36-week run-in, discontinuation produced ~14 percentage-point regain within a year, cardiometabolic improvements reversed in proportion. This is chronic suppressive therapy, not a course.

Mortality and cardiovascular

SELECT (Lincoff et al., NEJM 2023, n=17,604, obese non-diabetic with established CVD, ~40-month follow-up): semaglutide reduced 3-point MACE by 20%, HR 0.80 (95% CI 0.72–0.90). All-cause mortality HR 0.81 (0.71–0.93) — a 19% relative reduction in death from any cause.

SURPASS-CVOT (NEJM Dec 2025, n=13,165, T2D + ASCVD): tirzepatide vs dulaglutide. Non-inferior on MACE-3 (12.2% vs 13.1%). ~16% relative all-cause mortality reduction vs an already cardioprotective comparator.

SURMOUNT-MMO (n≈15,000, obese non-diabetic) reads out 2027–28 and will convert the extrapolation below into a head-to-head finding.

Life-years on the table

Lost to obesity + hypertension (50-year-old, BMI 32, non-smoker, hypertensive). Drawing on Prospective Studies Collaboration, Lancet 2009 (900,000 adults, 57 cohorts: BMI 30–35 reduces median survival by 2–4 years), Global BMI Mortality Collaboration, Lancet 2016 (HR 1.44 for BMI ≥30 vs normal), and Peeters et al., Annals of Internal Medicine 2003 (Framingham life-table): central estimate 3 to 5 life-years lost to the BMI 32 + hypertensive composite.

Gained on tirzepatide (extrapolation, not direct evidence). Applying SELECT's 19% all-cause mortality HR to the ~30-year residual life expectancy of a 50-year-old, the order-of-magnitude gain is 2 to 3 life-years, contingent on continued use.

Net. Compliant long-term use: ~1.5 to 3 net life-years recovered, with the bulk of the gain coming from reduced CV mortality rather than weight loss itself.

Mounjaro — the worked case for the lab tech

The lab tech on the Steelman page says Mounjaro is "cheating". The actuarial answer is in the row above. The consistency answer is below.

1. We do not call insulin cheating diabetes. We do not call statins cheating familial hypercholesterolaemia. We do not call SSRIs cheating depression — at least not in clinical settings. We do not call methadone cheating opioid dependence. We do not call warfarin cheating atrial fibrillation. The "cheating" frame is reserved, almost exclusively, for weight — and within weight, almost exclusively for women. The asymmetry is the giveaway. The objection is not medical, it is moral.
2. The biology refutes the willpower frame directly. The set-point physiology documented at Rockefeller for forty years (see Pathway § Set-point) shows that the reduced-obese body actively defends prior fat mass through a measurable metabolic suppression of 300–400 kcal/day below predicted, persisting indefinitely. The drug shifts the defended set-point downward. That is its mechanism. Asking the patient to do without it is asking them to white-knuckle a homeostatic system every day for the rest of their life. A few patients can; most cannot; this is not a moral ranking.
3. Side-effect profile, honest read.

   Concern	Evidence	Verdict
   GI (nausea, diarrhoea, vomiting)	39–49% of SURMOUNT-1, mostly mild-moderate, titration-related	Real, common, tolerable
   Pancreatitis	Meta-analysis: no significant increase; incidence <0.2%	Overhyped
   Gallbladder / cholelithiasis	RR 1.52 for biliary disease (12-trial meta-analysis)	Real, driven by rapid weight loss
   Thyroid C-cell carcinoma	Rodent finding; humans have ~1/40th the GLP-1R density; no clinical signal	Largely overhyped
   Lean mass loss	~26% of total loss; better fat:lean ratio than diet-only	Real but proportionate; mitigate with resistance training
   Retinopathy	SUSTAIN-6 signal in long-standing diabetics during rapid glycaemic correction	Real for diabetics, not for non-diabetic obesity use

4. The genuine worries — not "cheating", but cost, supply, lifetime safety, and off-label cosmetic creep. UK retail Mounjaro runs ~£150–£250 per month depending on dose and pharmacy. NICE TA1026 found cost-effectiveness only at BMI ≥35 with comorbidity; the BMI 30–34.9 cohort sits above the willingness-to-pay threshold. Persistent supply constraints 2023–25. Longest RCT exposure is ~3 years; cardiovascular and oncological signal at 15–20 years is genuinely uncertain. The legitimate frontier worry is off-label cosmetic use in BMI <27 populations — the side-effect burden with the smallest mortality dividend.

Honest summation

Where Snowdon is right. The "cost to the NHS" number is gross, not net, and is routinely cited as if it were the bottom line. That is sloppy. Excise duty on tobacco (£10bn) and alcohol (£10.4bn) is real revenue that public-health advocates structurally exclude from their cost-of-illness tables. The van Baal 2008 finding — healthy-living cohorts have the highest lifetime healthcare spend — is robust, peer-reviewed and consistently replicated. Public-health cost estimates have a documented tendency to grow whenever they are recalculated (alcohol: £21bn → £27.4bn; obesity: £27bn → £98bn → £126bn). Treat each escalation with the same scepticism Snowdon brings.

Where Snowdon deflects. "Early death is a fiscal win" is a Treasury argument, not a welfare one. It only works if QALYs are priced at zero. The model systematically under-weights externalities falling on non-users. The "drinkers subsidise non-drinkers" framing rests on average drinkers — harm and cost concentrate in the heaviest-drinking decile, and average-versus-marginal sleight-of-hand obscures it. Opioids break the model entirely; there is no excise-revenue offset for prescription or illicit opioid use, and the US $1.5 trillion figure is real welfare loss. The libertarian "adult informed choice" frame collapses under addiction's neurobiology.

The defensible middle ground. The headline "addiction is bankrupting the NHS" claims are inflated, gross-of-revenue and methodologically loose. The Snowdon arithmetic is correct as Treasury accounting but morally evasive as welfare economics. Lifestyle-related disease is an expensive welfare loss, not a fiscal crisis. Policy built on the fiscal-crisis framing tends to be paternalist theatre; policy built on welfare-loss accounting has to defend itself on welfare grounds, not on cost-to-the-taxpayer scaremongering.

That answers the cost question. The next page asks the better one: what works.