J. Konstapel, Constable Research, Leiden
© 2026 J. Konstapel, Constable Research, Leiden. All rights reserved. Quotation with attribution is permitted; reproduction of the whole requires written permission.
1. One question, asked of everything
Where Illness Comes From, Completed is a free reference work of 1,253 pages. It covers every chapter of the International Classification of Diseases, eleventh revision (ICD-11), without exception. One entry treats one condition.
Every entry has the same eight sections, in the same order: what it is, measured, where it starts, cure status, prevention, treatment, simple means, gaps.
The uniformity is the instrument. The same eight questions are put to tuberculosis, cataract, diverticulitis and depression. The answers can then be laid side by side.
Three of the eight do the real work. Can it be cured? What prevents it? And what can be done where there is no industry, no cold chain and no money?
The book says of itself that the third question “turned out to be the one that reorganises the book”. This essay follows that thread. It concentrates on what is unusual in the book: the findings, the cases and the method that a reader will not find assembled anywhere else.
2. Four words, used narrowly
The spine of the book is a cure status with four values.
- SELF-LIMITING. The body completes the discharge itself. Intervention adds nothing.
- CURABLE. A treatment exists that removes the condition.
- CONTROLLABLE. The state can be held. The deviation remains.
- RELIEF ONLY. Nothing removes it.
The words are kept narrow on purpose. A treatment that must continue for life does not cure. Phenylketonuria on a lifelong diet is therefore CONTROLLABLE. So is haemochromatosis held by repeated phlebotomy. Gastric MALT lymphoma, cleared by one course of antibiotics, is CURABLE. The book names it as the type case.
Where the honest answer differs by stage, cause or strain, the label is split. Colorectal cancer in the bowel wall has a five-year relative survival of 91.3 per cent. With liver and lung deposits it has 16.9 per cent. One word for both would be false at one end.
The narrowness has consequences that the book enforces against itself. A penile prosthesis was once called “a mechanical cure” in one entry. The native tissue is destroyed and its function substituted, so the deviation remains. The label was corrected. Trachoma appeared twice. One entry split it three ways and reserved RELIEF ONLY for established corneal blindness. The other simply said CURABLE. For a book whose subject is what can and cannot be cured, the book notes, “that omission is the whole of the matter”.
Prevention carries a second scale: STRONG, MODERATE, WEAK, CONTESTED. A compound grade takes the lower value. CONTESTED overrides the others, because disagreement among sources is a statement about the state of the evidence, not its strength. The checking resolved 239 compound grades this way.
3. The simple means, audited
A simple means, in the book’s definition, prevents, cures or controls a condition without a long industrial supply chain behind it. A sachet of salt and glucose. A bar of soap. A bed net. A lamp at the right hour. Thirty minutes of walking. A mother’s chest.
Part 3 audits this class. Its dossier holds forty-six interventions, graded as fifty-nine rows. About half of what is commonly claimed for the class does not survive the grading. Both halves are printed, and the order is fixed: what holds first, what does not hold second.
What holds. Reduced-osmolarity oral rehydration solution, at a UNICEF indicative price of US$0.05 a sachet. Against intravenous rehydration, across seventeen trials and 1,811 children, the treatment-failure risk difference is 4 per cent. Roughly one child in twenty-five on oral therapy still needs a drip. Insecticide-treated nets: all-cause child mortality rate ratio 0.83, high certainty. Immediate kangaroo mother care: 28-day mortality 12.0 against 15.7 per cent in 3,211 infants. That is a number needed to treat of 27, and no equipment beyond the mother’s body. Magnesium sulfate for eclampsia, an inorganic salt solution at about US$0.33 an ampoule. Toothbrushing in intensive care: hospital-acquired pneumonia risk ratio 0.67 across fifteen trials. Exercise for knee osteoarthritis: twelve points less pain on a scale of 100, high certainty.
What does not hold, or holds less. The kitchen version of rehydration solution has no mortality estimate at all. Its documented risk is hypernatraemia from mismeasured salt. “Measurement is exactly the step that packaging removes.” Vitamin A supplementation is CONTESTED: a mortality risk ratio of 0.76 without the two-million-child DEVTA trial, 0.88 with it, and 0.96 for DEVTA alone. The 72 per cent protection of folic acid is recurrence prevention, in women with a previous affected pregnancy. For first occurrence there is one trial, not significant. Handwashing reduces diarrhoea episodes by about 30 per cent; no trial establishes fewer deaths. Building latrines without changing their use has not reduced diarrhoea in cluster-randomised trials. Social connection carries one of the largest associations in epidemiology, an odds ratio of 1.50 for survival. The widely repeated comparison with “fifteen cigarettes a day” has no numeric basis in the papers.
Light therapy shows both sides in one entry. As treatment for seasonal affective disorder its effect size is 0.84. In the single head-to-head trial against fluoxetine, light reached 0.80 and the drug 0.24. The same review records hypomania at a number needed to harm of 8. That figure, the book notes, “is rarely mentioned when light is described as side-effect-free”.
Two counter-cases. Vaccination is STRONG for effect by a wide margin: 154 million lives saved since 1974. Yet it sits outside simple means by construction, because every vaccine but one needs an unbroken 2 to 8 °C chain. Amoxicillin is the second case. It is off patent and costs about five US cents a tablet. Its active ingredient runs through more than ten industrial steps and up to about 70 kg of waste per kilogram. There were eight active US drug master files, held by seven companies. “Off-patent describes a legal status. It says nothing about how many steps, how much waste, or how few producers.”
From this the book draws its central reading: “a simple means is not a cheap means. It is a means whose chain closes at the point of need.”
The short list. No validated catalogue of simply producible essential medicines exists. What can be asserted, item by item, is a list of eleven: oral rehydration salts, zinc sulfate, iodised salt, folic acid, ferrous salts, vitamin A capsules, chlorhexidine 7.1 per cent gel, magnesium sulfate injection, sodium chloride and glucose solutions, paracetamol, oral vitamin D.
One WHO production manual exists for the whole class, for oral rehydration salts (2006). For three to four million sachets a year it specifies a mixing room of about 30 m², a packing room of about 45 m² and a quality-control laboratory of about 15 m². Part 3 closes on one sentence: “Simple does not mean improvised. The shortest chain in medicine still has a laboratory in it.”
4. Where the cheapest cures sit
Across the chapters the third question produces a map. Some of its points are well known to specialists and almost unknown outside them.
Cancer cured by antibiotics. Early gastric MALT lymphoma is cleared in roughly 60 to 80 per cent of cases by one to two weeks of generic antibiotics costing a few dollars. Helicobacter pylori eradication more broadly clears the organism in 70 to 94 per cent of courses. Ulcer recurrence falls from about 60 to 70 per cent a year to under 5 per cent. The organism is a Group 1 carcinogen behind 75 to 90 per cent of non-cardia gastric cancer.
Metastatic cancer cured with off-patent medicines. Four cases stand out. Widely metastatic testicular cancer: roughly 90 to 95 per cent of good-risk patients cured by bleomycin, etoposide and cisplatin. Stage IV Hodgkin lymphoma: 82.8 per cent five-year relative survival on four listed medicines. Acute promyelocytic leukaemia: over 90 per cent cured by all-trans retinoic acid and arsenic trioxide, without chemotherapy in non-high-risk disease. Childhood acute lymphoblastic leukaemia: roughly 90 to 94 per cent cured in high-income settings on a protocol built entirely from off-patent listed drugs.
Treatment that is the removal of blood. In haemochromatosis each 500 mL unit removed takes 200 to 250 mg of iron with it. A homozygote de-ironed before cirrhosis has normal life expectancy. The consumables are a needle, a bag and a nurse. Where the patient donates, the cost to the system is negative. In polycythaemia vera the CYTO-PV trial compared haematocrit targets: cardiovascular death or major thrombosis fell from 9.8 to 2.7 per cent.
Food as cure, and food that is not food. Severe wasting is cured by ready-to-use therapeutic food, with recovery around 80 to 90 per cent in outpatient programmes. The instrument that finds the cases is a paper tape around the upper arm, “and a mother can use it on her own child”. Phenylketonuria shows the limit. The heel-prick test costs a few dollars. The treatment is diet. But the phenylalanine-free amino-acid mixture is a manufactured medical food costing roughly US$5,000 to 25,000 a patient-year. “There is no generic pathway for medical foods.”
The eye, the ear and the skin. Cataract surgery at the Aravind Eye Care System cost US$29.02 per operation in 2013. Its own manufacturing arm brought the intraocular lens to US$3. The complication rate was 1.5 per cent, reported at about half that of the British National Health Service. In the PROSPER trial, 750 tea pickers in Assam received reading glasses. Productivity rose 21.7 per cent overall and 31.6 per cent among those over fifty. Cerumen impaction, a WHO-listed cause of hearing loss, is removed with water and a rubber bulb. The Epley manoeuvre cures roughly 80 per cent of positional vertigo in one session of about two minutes.
The heart and the lungs. Five low-priced interventions carry measured mortality benefits. Benzathine benzylpenicillin cuts rheumatic fever recurrence by 80 to 90 per cent, at about US$0.59 a vial. It suffers repeated global shortages “because the margin is too small to hold manufacturers in the market”. One chewed aspirin at the onset of infarction cuts 35-day vascular mortality by 23 per cent. Oral amoxicillin from a community health worker cuts child pneumonia mortality by 35 to 42 per cent. Dexamethasone and spironolactone complete the five. Beside them stand measures that are not drugs at all. Bystander resuscitation doubles to triples survival from cardiac arrest. Prone positioning in severe respiratory distress needs four or five people, pillows and a protocol, for a number needed to treat of 6.
Oxygen, the constraint that returns. In the infectious diseases, where nothing simple works, oxygen is the recurring limit. It is the rate-limiting treatment in childhood pneumonia and the intervention that turns severe RSV bronchiolitis from often fatal to usually survivable. “It cannot be improvised.” The minimum chain is a concentrator with uninterrupted electricity, or cylinder logistics.
Three measures that are not clinical. In the chapter on reproduction, the perinatal period, congenital anomalies and injury, three of the largest reported effects come from outside the clinic.
The first is Sri Lanka’s pesticide regulation. The suicide rate peaked at 47 per 100,000 in 1995 and then fell about 70 per cent. The reported result is about 93,000 lives over twenty years, at under US$50 per life saved. The bans removed specific agents; total pesticide use still rose.
The second is folic acid in flour. A neural tube closes by the twenty-eighth day after conception, before most pregnancies are recognised. Fortification removes the timing problem. Neural tube defects fell 25 to 50 per cent wherever it was done, at one to ten cents per person per year. “The intervention is a decision taken at a mill.”
The third is mass dog vaccination. Seventy per cent coverage interrupts rabies transmission, at about a dollar per dog.
None of the three falls inside a health-service budget. The book names this as a recurring reason why they are underfunded relative to their effect.
5. The two ends of one list
The WHO Model List of Essential Medicines, 24th edition, now holds both ends of the price range. Oral morphine costs cents a dose. Pembrolizumab costs about US$53,000 a year at the United States cash price. Blinatumomab costs about US$180,000 a course at United States list.
The morphine end is the more telling. About 56.8 million people need palliative care each year; roughly 14 per cent receive it. In 2018, 79 per cent of the world’s population consumed 13 per cent of the morphine used for pain. That morphine was 1 per cent of the 388 tonnes manufactured worldwide: about 3.9 tonnes. Oral morphine is the only intervention that appears in the simple-means section of every cancer entry. It is also the one most consistently unavailable. The barrier is legislation, licensing and fear of diversion, not price, patent or technology.
The same pattern appears in insulin. Human insulin is priced at 1.2 to 1.8 times its estimated production cost. Glargine is priced at 5.6 to 7.8 times its own. “The economic problem is in the analogues, not in insulin.”
6. The chain
Part 4 asks the question that no single entry can answer. A medicine exists, it is cheap to make, and it is not available. How short can the production chain be, and who may legally make what? The answer is given as law and institution, never as a recipe.
Three Dutch cases. Each has a published price before and after registration.
- Chenodeoxycholic acid for cerebrotendinous xanthomatosis: from €308 to about €153,300 per patient-year after orphan registration in 2017. That is a factor of 498.
- Cysteamine eye drops for cystinosis: the pharmacy preparation at €1,839.24 per patient-year against €53,300 for the registered product. A factor of 29.
- Mexiletine for myotonia: €6.95 against €23.39 a capsule. A factor of 3.4.
The multiples are the headline. The book states that they are not the most important finding.
What actually binds. The Amsterdam preparation of chenodeoxycholic acid did not stop on law. The minister wanted it to continue. The competition authority later fined the rights holder €19.6 million for abuse of a dominant position. The preparation still stopped for sixteen months, because the national institute found unknown impurities in the starting material above the pharmacopoeial limit. “The binding constraint on the short chain is qualification of the active substance, not the legal exemption.”
Registration closes the short chain. The cysteamine preparation existed first and worked. Registration did not add a treatment. It moved an existing one inside the authorisation regime. After that the preparing pharmacies could no longer supply it. The Dutch counter-move, from June 2021, is for academic hospitals and a compounding pharmacy to register their own preparations. The book calls it “the short chain accepting the long chain’s paperwork in order to keep the short chain’s cost base”.
A precedent. In January 2021 the Dutch national health care institute advised against reimbursing registered mexiletine. It advised bringing the pharmacy preparations into the basic package instead. The regulation followed in July 2021. The compounded route became the reimbursed standard of care at roughly a third of the registered price.
Two legal architectures. The European exemption is relational. Directive 2001/83/EC exempts the magistral and the officinal formula, tied to the pharmacy’s own patients, with no volume ceiling. The American Drug Quality and Security Act of 2013 does the opposite. Its section 503B creates a licensed industrial tier with no patient relationship at all, regulated as manufacture. On 30 September 2026 there were 98 registered outsourcing facilities.
Off-patent is not secure supply. Across 554 generic active ingredients, 56.3 per cent have five or fewer certified manufacturers worldwide. In 2025 the Netherlands had 1,134 medicines in shortage, affecting 3.5 million people. The medicines named are levothyroxine, rifampicin, salbutamol, prednisolone, quetiapine, lithium and oxazepam: old, cheap, off-patent and preparable. Pharmacy preparation appears once in the entire national shortage report. The two datasets describe the same medicines from opposite ends and have never been joined. Appendix B calls this the only question in Part 4 with no published answer at all.
Where the chain cannot be shortened. Monoclonal antibodies: the process defines the molecule. Cell therapy: the product is the patient’s own cells. India’s point-of-care CAR-T is expected at about US$50,000 against about US$400,000 in the United States. That is a factor of 8, not the factor of hundreds that compounding achieves for a small molecule. Radiotherapy: the constraint is the workforce, not the machine count. Dialysis: a permanent consumables line. Transplantation shortens that chain. By the second year after transplant the annual cost is €15,018 against €92,616 for in-centre haemodialysis, about a sixth.
The asymmetry of mRNA. An mRNA vaccine cannot be made in a pharmacy. It does not need a global industry either. One production line can make 100 million doses a year with seventy-five to eighty staff. The chain is “short in headcount and footprint and long in technical specificity”. The binding constraint is technology transfer, not scale.
7. What no longer holds
Most reference works record what is currently believed. Several chapters of this one also record what was believed and has been retired. Readers will meet these claims in older textbooks, in clinical habit and in marketing.
Some of the retired certainties are striking. In idiopathic pulmonary fibrosis, the standard immunosuppression of decades raised mortality from 1 to 8 per cent in PANTHER-IPF. Adaptive servo-ventilation in heart failure increased cardiovascular mortality in SERVE-HF. Arthroscopic partial meniscectomy did no better than sham surgery in FIDELITY. Paracetamol was identical to placebo for acute low back pain in PACE, with seventeen days to recovery in both arms. In head injury, the CRASH trial found increased mortality with corticosteroids. In neonatal encephalopathy, cooling helps in high-income settings and raised mortality in the HELIX trial in South Asia. It is a treatment “that works where it was first tested and does not travel”.
Diverticular disease, the entry that prompted the book, contains two reversals. Nuts, seeds and popcorn were forbidden for decades. In a cohort of 47,228 men, popcorn carried a diverticulitis hazard ratio of 0.72: significantly protective. And antibiotics for uncomplicated diverticulitis, standard for seventy years, showed no demonstrated benefit in the DIABOLO and AVOD trials. Hospital stay was shorter without them. The book notes that the failure also puts the infective model of the disease in doubt, and that this question is open.
One observation runs through the chapter on the gut, the frame and the waterworks. Structural findings on imaging are largely age-normal. Disc degeneration is present in about 37 per cent of people without symptoms at twenty and about 96 per cent at eighty. Full-thickness rotator cuff tears are present in over half of people above eighty, many of them painless. “A finding that is present in nearly everyone of a given age is a property of the age, not of the complaint.”
8. The reservoir
The cancer chapter collects the autopsy evidence in one place. It comes from people who died of something else and whose organs were then examined systematically.
Incidental prostate cancer is present in 5 per cent of men under thirty and 59 per cent of men over seventy-nine. Occult differentiated thyroid cancer is found in 4.1 per cent on partial examination of the gland and 11.2 per cent on whole-gland examination. The difference is the point: “how much cancer is found depends on how hard the pathologist looks.” Occult invasive breast cancer is found in 0.85 per cent; in-situ cancer in 8.9 per cent.
The size of the reservoir tracks the size of the overdiagnosis problem. Prostate and thyroid have large reservoirs and large overdiagnosis. Breast has a small invasive reservoir and a moderate, contested figure concentrated in in-situ disease. South Korea added thyroid ultrasound to health checks; incidence rose roughly fifteenfold between 1993 and 2011 while mortality stayed flat. Nasopharyngeal carcinoma is the control case. No occult reservoir is found at autopsy, and screening by plasma Epstein-Barr virus DNA gives a large stage shift without a measurable overdiagnosis penalty.
The book adds a second list that is rarely made: where overdiagnosis has never been quantified. Gastric, liver, kidney and bladder cancer. Cervical intraepithelial neoplasia, colonic adenoma, uterine fibroids, incidental meningioma, monoclonal gammopathy and monoclonal B-cell lymphocytosis. Several of these have reservoirs as large as the measured ones, or larger.
Childhood cancer carries its own quantity. Paediatric tumours have about 0.13 mutations per megabase against 1.8 in adults, roughly fourteen times fewer. There is almost nothing for checkpoint inhibitors to work on, and checkpoint blockade has broadly failed in paediatric oncology. The immunotherapies that do work in children target a lineage antigen. Survival diverges threefold between rich and poor countries. The book states that the divergence is not biological. A leading cause of death is abandonment of treatment, driven by travel cost, lost income and accommodation.
9. Gaps that are content
The book treats the absence of a number as a finding in its own right.
A whole ICD-11 chapter sits outside the global burden accounts. No sleep-wake disorder is a level-three cause in the Global Burden of Disease study. That includes obstructive sleep apnoea, which on the only global estimate affects 936 million adults aged 30 to 69. The funding statement of that study is the single word ResMed.
Chapter 17 of ICD-11, conditions related to sexual health, has existed in released form for seven years. Not one of its categories has a prevalence estimate measured against its own definition. Erectile dysfunction carries a published global prevalence range of 3 to 76.5 per cent. “A 25-fold spread is not a prevalence estimate. It is a measurement failure.”
Seventeen circulatory and respiratory conditions carry no global burden figure in the tables retrieved. Heart failure alone affects about 64 million people; allergic rhinitis about 400 million.
Four WHO fact sheets give their headline figure without a data year. Two expected fact-sheet addresses return pages with no Key Facts and no body. “Any work citing a ‘WHO fact sheet’ for either is citing a document that does not exist.”
10. The frame underneath
Part 0 sets out the rules the book applies. They come from the Vacuum.Net theory, governed by its Foundational Paper, eighth edition, version 8.1. The theory was built at the bottom, with one strand and one test, and no constants given in advance. The book applies it upward and does not repeat the derivation.
The fishnet. The standing image is a fishnet under tension. The net is one strand. Where it crosses itself there is a crossing; between crossings lie openings, the meshes. A winding is a stretch of strand that returns on itself in step. A knot is windings that hold each other in three directions. One rule decides what stays: a winding persists only if it returns on itself in step. What does not close passes its tension to the meshes beside it.
The register. At a crossing the strand passes over, passes under, or does not cross. These are written +1, −1 and 0: a trit. Counting in balanced ternary needs no separate minus sign; negation is mirroring. Three is also the cheapest base. The cost of writing numbers in base r goes as r / ln r. That gives 2.73 for base three against 2.89 for base two.
Windows. Three crossings form a window. It is closed when its three trits add to a multiple of three. Of the 27 possible windows, 9 close. Three are uniform, with three equal trits: the form of a crystal, holding and not turning. Six are mixed, with three different trits: a form that keeps turning while it stays closed. Every content changes; the closure holds. The book calls this the form of living matter. Where a uniform stretch meets a mixed stretch, the window on the seam stands open and carries tension. “That seam is where this book finds most of its deviations.”
The test. When one form reads another there are three outcomes. +1: it fits, and stays. −1: it does not fit, and is taken apart. 0: it is not read, and simply remains. The third outcome carries most of the book. “Health is not the absence of misfits; misfits arise every day, in everyone, in thousands. Health is a test that reads fast enough.”
Illness read through the test. The body carries a reading layer that runs the test at cell scale. There are places where it must not discharge what it reads: the gut wall at every meal, the eye, the child in the womb, which is by signature half foreign. So the body carries official zero-tokens that turn the local reading to zero. Clinicians call them checkpoints. Without them the body destroys itself; with them it has a blind spot. A test that reads too weakly lets misfits through. A test that reads too sharply discharges healthy tissue, which is autoimmunity. “One layer, one dial, two directions of failure.”
Cancer then sits in one sentence: a tumour is the remainder of a selection. For years every variant without the zero-token was discharged and every variant with it was not. The autopsy reservoir of Section 8 is the measured form of that remainder: altered tissue that the test around it held. Ageing is read as lost phase-lock — heart-rate variability, deep sleep, hormone tides drifting out of step — so lifespan is read as coherence-span. Tuberculosis in a quarter of the world, carried and mostly never expressed, is the zero outcome at the scale of a population.
A check at the bottom. Part 0 also prints the theory’s measured anchor. The threshold for gravitational anomalies, a₀, comes out as c·H₀/2π. With H₀ = 70 km s⁻¹ Mpc⁻¹ this gives 1.084 × 10⁻¹⁰ m s⁻², against about 1.2 × 10⁻¹⁰ measured from galaxy rotation curves. With H₀ = 67 it gives 1.050 × 10⁻¹⁰. Both are printed, because the two measurements of the Hubble constant disagree. For the book the point is one only: the frame was not built for biology and bent to fit.
Readings, not mechanisms. The book uses the frame to order, never to supply numbers. Every number comes from the standing literature. Where a measured fact is set beside the rules, the sentence is marked “Read in the net” and closes with “This is a reading, not a mechanism.” There are twelve such readings in the book. In place of cause the book uses four relations: selection by fit, identity over scale, constitutive relation, and correspondence. A dictionary maps every word in the entries — thread, winding, the test, tension, depth, the legitimate zero — onto exactly one element of Part 0.
Two readings show how the frame earns its place in a medical book. On diverticula: “the wall is a mesh under tension, and it yields where it is already open.” The vascular penetration points of the colon wall are the pre-existing openings. Nothing new has to be made for a pouch to form. And on simple means, quoted above: a means whose chain closes at the point of need. Price and patent are read at the wrong depth. What decides is whether the thing can be completed where it is used.
11. A book that shows its own accounting
The last distinctive feature is the apparatus. It is unusual for any reference work, and it makes this one usable at a known level of confidence.
The dagger. Every numeric claim was checked against a named source during compilation. A figure read off its source, or resting on a named source not opened again, is printed plainly. A figure whose source could not be opened at all carries a dagger (†). The book counts 2,361 daggers against 6,352 figures resting on a named source. A figure with no source at all was struck, and the sentence says in words what is known without it. “No figure in this book is an estimate.”
The Register of Contradictions. Appendix A records thirty-one places where a single WHO fact sheet contradicts itself. Hepatitis C appears as about 47 million chronic infections in the key facts and 50 million in the sum of the body’s own regional figures. Dengue risk appears as about four billion people and as 5.6 billion. A further table lists pairs of sources of comparable standing that disagree, printed both ways with no winner picked. Four of these differ in method or vintage rather than in the world, and the book says which.
The book against itself. Three conditions turned out to have a full entry in two chapters, written by different hands from different dossiers. Merging the two fibroid entries found one visible disagreement and eight that nobody had seen. “The lesson is not about fibroids.” Two competent writers produced nine disagreements, and only one concerned the headline answer.
An invented number. A number needed to treat of about 50 for aspirin in secondary prevention circulated through the preparation, plausible and correctly cited. The cited paper publishes event rates of 6.7 and 8.2 per cent a year, and no number needed to treat. Recomputed, the figure is 67 a year and 33 over two years. “The figure had no origin.”
The drift. Appendix C records 3,388 corrections over four rounds of checking by readers who had not produced what they read. The dominant finding is that the book “systematically presented itself as better sourced than it was” — everywhere, by every hand, in the same direction, and never by a large step. About 1,480 provenance tags were lowered. “Nothing was discovered to be false by this. What was discovered is that the book did not know what it knew.” The checks themselves were checked. A comparison of entry titles had silently skipped all forty-two infectious-disease entries because they were headed one level higher. An assembly tool had dropped 541 daggers at line breaks. Both faults were found and corrected, and both are recorded.
12. What the book makes visible
Read across its 1,253 pages, the book produces two lists that no other single source assembles.
The first is a list of interventions that are cheap, old and effective, with their measured effects and their honest limits. Many sit outside the clinic: at a mill, in a ministry of agriculture, in a veterinary campaign. Others are in the clinic but not in reach, because a chain does not close where they are needed.
The second is a list of conditions where nothing simple works, named as plainly as the successes. Glaucomatous field loss has never been recovered by any treatment. Cochlear hair cells do not regenerate. Multimorbidity remains CONTROLLABLE after thirty-eight trials and 18,073 participants showed no effect of structured deprescribing on admissions or quality of life.
Between the two lists runs the fishnet of Part 0. It does not explain the numbers. It orders them by one question, asked at every depth: does it close where it is used?
Annotated references
The references below are those that carry the most weight in the book and in this essay. Each has two parts: why it is worth reading, and where useful, how to read it. The book itself cites several hundred sources; its appendices name the sources it could not open.
A. The book and its frame
1. Konstapel, J. (2026). Where Illness Comes From, Completed: A Reference Work on Every Chapter of Human Illness, Its Prevention, Its Cure, and What Can Be Done With Simple Means. First edition, version 1.0, October 2026. Constable Research, Leiden. Why read? The subject of this essay. 1,253 pages, every ICD-11 chapter, one eight-section entry per condition. Reading advice: Start with “How to use this book” (pp. 12–13). Then read Part 3 (pp. 1221–1230) and Part 4 (pp. 1231–1240) before any entry. The appendices (pp. 1241–1253) explain how much weight each figure carries. Use the entries as a reference, one condition at a time.
2. Konstapel, J. (2026). Where Illness Comes From. Ninth edition, September 2026. constable.blog. Why read? The predecessor. It walks one frame from first chemistry to the human being and takes its cases where they illustrate a layer. It is not a catalogue. Diverticulitis does not appear in it; that absence prompted the completed book. Reading advice: Read for the frame and the layer-by-layer argument. Use the completed book for any particular condition.
3. Konstapel, J. (2026). The Vacuum.Net Theory: Foundational Paper. Eighth edition, version 8.1, 17 October 2026. constable.blog. Why read? The governing statement of the rules in Part 0: one strand, closure as the only selection rule, memory, context, scale, and the ternary register. Reading advice: Read the five starting rules and the window count first. The a₀ derivation shows how the frame was made testable at the bottom before it was walked upward.
B. Burden and classification
4. GBD 2023 Collaborators (2025). Global Burden of Disease Study 2023. The Lancet, published 12 October 2025. Why read? The book’s burden basis: 292 causes of death, 375 diseases and injuries, 204 countries, 1990–2023. Reading advice: Note what is not in it. No sleep-wake disorder is a level-three cause, and heart failure is not a cause line. The per-condition Results Tool is the single source that would retire the most daggers in the book.
5. World Health Organization (2019–). International Classification of Diseases, 11th Revision (ICD-11 MMS). Why read? The scaffold of the book. Its chapters 07 (sleep-wake) and 17 (sexual health) are new in ICD-11, and both lack measured epidemiology. Reading advice: Read the chapter introductions and the extension codes (Chapter X). The HA40 aetiology axis shows how ICD-11 moves causation off the diagnosis and onto a separate coordinate.
6. Ferlay, J. et al. / International Agency for Research on Cancer. GLOBOCAN 2022 and GLOBOCAN 2024. Why read? Cancer incidence and mortality by site. The book prints both editions side by side. Reading advice: Cervical cancer moves from 662,301 to 604,196 cases between editions. The book reads this as a methodological revision, not a trend. Do not compare editions as if they were a time series.
C. Simple means and their audit
7. World Health Organization (2006). Oral Rehydration Salts: Production of the New ORS. Geneva: WHO. Why read? The only operational WHO production manual for a simple means. Formulation, raw-material specifications, quality-control tests, premises and shelf life. Reading advice: Read the premises specification. It is the concrete form of the book’s sentence that the shortest chain still has a laboratory in it.
8. Imdad, A. et al. (2022). Vitamin A supplementation for preventing morbidity and mortality in children from six months to five years of age. Cochrane Database of Systematic Reviews. Why read? Nineteen trials, 1,202,382 children, all-cause mortality risk ratio 0.88. The central source for one of the book’s CONTESTED grades. Reading advice: Read the sensitivity analysis with and without the DEVTA trial. The familiar 24 per cent rests mainly on trials before 1994.
9. Magpie Trial Collaborative Group (2002). Do women with pre-eclampsia, and their babies, benefit from magnesium sulphate? The Lancet 359:1877–1890. Why read? 10,141 women in 33 countries. Eclampsia 1.9 against 0.8 per cent, a number needed to treat of 91, from an inorganic salt solution. Reading advice: Read it beside the UNICEF price of about US$0.33 an ampoule. Few interventions combine this effect with this simplicity of manufacture.
10. WHO Immediate KMC Study Group (2021). Immediate “kangaroo mother care” and survival of infants with low birth weight. New England Journal of Medicine 384:2028–2038. Why read? 3,211 infants in five countries. 28-day mortality 12.0 against 15.7 per cent, a number needed to treat of 27, with no equipment beyond the mother’s body. Reading advice: The trial stopped early for benefit, which overstates effect size. The book prints that qualifier with the figure.
11. Clasen, T. F. et al., for the HAPIN Investigators (2022). Liquefied petroleum gas or biomass fuel for cooking and effects on birth weight. New England Journal of Medicine. Why read? 3,200 households. Personal fine-particle exposure fell 66 per cent, verified. Birth weight did not change. Reading advice: Read it with RESPIRE (Guatemala) and CAPS (Malawi). Large attributable fractions and null trials stand side by side, and the book leaves the question open.
12. Holt-Lunstad, J., Smith, T. B. & Layton, J. B. (2010). Social relationships and mortality risk: a meta-analytic review. PLOS Medicine 7(7):e1000316. Why read? 148 studies, 308,849 participants, survival odds ratio 1.50. Larger than many treatment effects in the book. Reading advice: Look at the sub-measures. Complex social integration is strongest and most confounded. Living alone, the hardest measure, is weakest and not significant. The “fifteen cigarettes” comparison is not in the paper as a number.
13. Sherrington, C. et al. (2019). Exercise for preventing falls in older people living in the community. Cochrane Database of Systematic Reviews. Why read? 108 trials, 23,407 participants. Rate of falls 0.77 at high certainty: the only high-certainty result in the book’s chapter on sexual health, the non-disease chapters and the cross-cutting states. Reading advice: Balance and functional training carry the effect. Tai Chi needs nothing at all.
14. Leong, D. P. et al. (2015). Prognostic value of grip strength: findings from the Prospective Urban Rural Epidemiology (PURE) study. The Lancet 386:266–273. Why read? 139,691 people in 17 countries. Each 5 kg less grip strength raises all-cause mortality by a hazard ratio of 1.16. Reading advice: Read it as a measurement that takes seconds and reports on the whole person.
D. Cancer: reservoir, mutation burden, prevention
15. Bell, K. J. L. et al. (2015). Prevalence of incidental prostate cancer: a systematic review of autopsy studies. International Journal of Cancer 137:1749–1757. Why read? 29 autopsy studies. 5 per cent under thirty, 59 per cent over seventy-nine. Reading advice: Set the 59 per cent beside a lifetime risk of dying of prostate cancer of roughly 2 to 3 per cent.
16. Furuya-Kanamori, L. et al. (2016). Prevalence of differentiated thyroid cancer in autopsy studies over six decades: a meta-analysis. Journal of Clinical Oncology 34:3672–3679. Why read? 12,834 autopsies. 4.1 per cent on partial and 11.2 per cent on whole-gland examination. Reading advice: The difference between the two figures is the finding: the amount of cancer found depends on how hard one looks.
17. Thomas, E. T. et al. (2017). Prevalence of incidental breast cancer and precursor lesions in autopsy studies. BMC Cancer 17:808. Why read? 2,363 autopsies. Invasive cancer 0.85 per cent, in-situ 8.9 per cent. Reading advice: Explains why breast overdiagnosis is moderate and concentrated in in-situ disease.
18. Vaccarella, S. et al. (2016). Worldwide thyroid-cancer epidemic? The increasing impact of overdiagnosis. New England Journal of Medicine 375:614–617. Why read? About 470,000 women and 90,000 men overdiagnosed in twelve countries over two decades. Reading advice: Read with the South Korean incidence curve: a fifteenfold rise with flat mortality.
19. Gröbner, S. N. et al. (2018). The landscape of genomic alterations across childhood cancers. Nature 555:321–327. Why read? 961 tumours. 0.13 mutations per megabase against 1.8 in adults. Reading advice: Read it with the clinical consequence the book draws: checkpoint blockade has broadly failed in children, while lineage-antigen therapies work.
20. de Martel, C. et al. (2020). Global burden of cancer attributable to infections in 2018. The Lancet Global Health 8:e180–e190. Why read? 2.2 million infection-attributable cancers, about 13 per cent. H. pylori alone about 810,000. Reading advice: Only four agents are counted. Epstein-Barr virus, HHV-8, HTLV-1 and Schistosoma haematobium are not, which affects several of the book’s cancer entries.
21. World Health Organization and International Agency for Research on Cancer (2026). Analysis of 2022 cancer cases attributable to modifiable risk factors. Nature Medicine. Why read? 37.8 per cent of 2022 cancer cases, about seven million, with a published derivation. Reading advice: The book prefers this figure to the widely quoted but unsourced 30 to 50 per cent. Attributable is not the same as preventable now.
E. Trials that retired standing practice
22. Daniels, L. et al. (2017). Randomized clinical trial of observational versus antibiotic treatment for a first episode of CT-proven uncomplicated acute diverticulitis (DIABOLO). British Journal of Surgery 104:52–61. Why read? 528 patients in 22 Dutch centres. No difference in recovery; hospital stay shorter without antibiotics. Reading advice: Read with AVOD (Chabok, British Journal of Surgery 2012). The trials were not powered for rare outcomes and excluded the immunosuppressed. The supported conclusion is “no demonstrated benefit”.
23. Strate, L. L. et al. (2008). Nut, corn, and popcorn consumption and the incidence of diverticular disease. JAMA 300:907–914. Why read? 47,228 men over eighteen years. Popcorn hazard ratio 0.72 for diverticulitis. Decades of dietary restriction pointed the wrong way.
24. Crowe, F. L. et al. (2011). Diet and risk of diverticular disease in Oxford cohort of EPIC. BMJ 343:d4131. Why read? 47,033 participants. Highest against lowest fibre fifth: hazard ratio 0.59 in the abstract and 0.58 in the results section of the same paper. Reading advice: Read with Peery et al., Gastroenterology 2012, where higher fibre went with more diverticulosis. The outcomes differ: clinical disease against pouches seen at colonoscopy. No trial has randomised fibre.
25. Idiopathic Pulmonary Fibrosis Clinical Research Network (2012). Prednisone, azathioprine, and N-acetylcysteine for pulmonary fibrosis (PANTHER-IPF). New England Journal of Medicine 366:1968–1977. Why read? Stopped early. Mortality 8 against 1 per cent. The heaviest reversal in the respiratory chapter.
26. Sihvonen, R. et al. (2013). Arthroscopic partial meniscectomy versus sham surgery for a degenerative meniscal tear (FIDELITY). New England Journal of Medicine 369:2515–2524. Why read? No benefit over sham surgery. The operation continues at scale. Reading advice: Read with the book’s section on age-normal imaging findings; the two explain each other.
27. Thayyil, S. et al. (2021). Hypothermia for moderate or severe neonatal encephalopathy in low-income and middle-income countries (HELIX). The Lancet Global Health 9:e1273–e1285. Why read? No reduction in death or disability; increased mortality. Reading advice: The clearest case in the book of a treatment that works where it was tested and does not travel.
F. Price, supply and the chain
28. World Health Organization (2025). WHO Model List of Essential Medicines, 24th List. Technical Report Series 1064. Why read? The governing edition for the book. It holds oral morphine at cents a dose and pembrolizumab at tens of thousands of dollars a year. Reading advice: Distinguish a selected medicine from a therapeutic equivalent. The book treats the distinction as load-bearing.
29. UNICEF Supply Division. UNICEF Supply Catalogue, read 19 October 2026. Why read? The working substitute for the retired MSH price guide. Indicative US-dollar prices with material numbers that can be rechecked. Reading advice: The catalogue returns different prices for the same material number on different pages. Treat its figures as indicative and dated, never as a price.
30. Management Sciences for Health. International Medical Products Price Guide. Last edition 2017; website retired after 30 June 2024. Why read? For its history. Four of the book’s research lines named it as the source that would settle the weakest data. It no longer exists as a live product. Reading advice: Any price traceable to it is a 2017 price.
31. Gotham, D., Barber, M. J. & Hill, A. (2018). Production costs and potential prices for biosimilars of human insulin and insulin analogues. BMJ Global Health 3:e000850. Why read? Production cost estimates per phial for human insulin and three analogues. Human insulin is priced near cost; glargine at five to eight times its own. Reading advice: The estimates are for phials only, while current prices cover pens and cartridges too. A second, lower estimate of about US$1 a vial exists; the book prints both.
32. European Parliament and Council (2001). Directive 2001/83/EC on the Community code relating to medicinal products for human use, Articles 3(1), 3(2) and 5(1). Why read? The two sentences on which the whole European short chain rests: the magistral and the officinal exemption. Reading advice: Note what is absent. There is no volume ceiling; the limit is the pharmacy’s own patient relationship.
33. United States Congress (2013). Drug Quality and Security Act, sections 503A and 503B of the Federal Food, Drug, and Cosmetic Act. Why read? The opposite architecture: a licensed industrial compounding tier with no patient relationship, regulated as manufacture.
34. Autoriteit Consument & Markt (2021). Decision on Leadiant Biosciences: abuse of dominant position in the pricing of CDCA-Leadiant. The Hague, 19 July 2021. Why read? The €19.6 million fine in the chenodeoxycholic acid case. The authority held magistral preparation to be an exception to orphan exclusivity and a room for lower prices. Reading advice: Read it with the book’s finding that the preparation stopped on impurities in the starting material, not on law.
35. Zorginstituut Nederland (2021). Advice on reimbursement of registered mexiletine, 14 January 2021. Why read? A national authority prefers the pharmacy preparation to the registered orphan product and brings the preparation into the basic package.
36. Inspectie Gezondheidszorg en Jeugd (2023). Circular on supply of pharmacy preparations beyond the pharmacy’s own patients. Why read? Six conditions that define where the short chain may serve a population in the Netherlands. Reading advice: The circular says openly that this route rests on enforcement discretion, not yet on statute.
G. Palliative care and multimorbidity
37. World Health Organization (2020). Palliative Care. Fact sheet, 5 August 2020, citing the International Narcotics Control Board. Why read? 56.8 million people in need, about 14 per cent reached. In 2018, 79 per cent of the world’s population used 13 per cent of the morphine for pain. Reading advice: Compute the tonnage: 1 per cent of 388 tonnes is about 3.9 tonnes for most of humanity.
38. Temel, J. S. et al. (2010). Early palliative care for patients with metastatic non-small-cell lung cancer. New England Journal of Medicine 363:733–742. Why read? Better quality of life, less depression, less aggressive end-of-life care, and a median survival 2.7 months longer. Reading advice: Read with the 2017 Cochrane review. The quality-of-life benefit survives pooling; the survival benefit does not reach significance.
39. Barnett, K. et al. (2012). Epidemiology of multimorbidity and implications for health care, research, and medical education. The Lancet 380:37–43. Why read? 1.75 million patients in Scotland. More multimorbid people are under sixty-five than over it. Deprivation brings onset forward by ten to fifteen years.
H. Cardiovascular risk
40. Yusuf, S. et al. (2004). INTERHEART. The Lancet 364:937–952; and O’Donnell, M. J. et al. (2016). INTERSTROKE. The Lancet 388:761–775. Why read? Nine and ten modifiable factors with combined attributable risks of 90.4 and 90.7 per cent. Reading advice: Both are case-control studies. An attributable risk from that design is not a demonstrated preventable fraction. The phrase “90 per cent of heart attacks are preventable” is not what either study measured.
41. ISIS-2 Collaborative Group (1988). Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17,187 cases of suspected acute myocardial infarction. The Lancet 2:349–360. Why read? One chewed aspirin at the onset of infarction: 23 per cent less 35-day vascular mortality, for under one cent.