The HPV vaccine: what is established and what is disputed

Key findings

Global scope · Country detail for Australia, the United Kingdom, Sweden, Denmark, Japan, the United States, Germany, Poland, Colombia and India · Evidence cutoff: 11 September 2026

The disagreement is narrower than it appears. The vaccine’s defenders and its most cited critics agree on most of the facts below. They disagree mainly about whether trials designed the way these were can measure rare harms reliably. On prevention of the cervical lesions that precede cancer, the two sides are close. S05 S08

The disease is large and unevenly distributed. WHO’s cancer agency estimates 604 196 new cervical cancer cases and 279 583 deaths worldwide in 2024, over half in Asia and a fifth in Africa. WHO separately estimates HPV caused about 620 000 cancers in women and 70 000 in men in 2019, counting throat, anal, penile, vulval and vaginal cancers too. S01 S03

The trials were not a textbook ideal, and were never designed to be. No randomised trial has shown that the vaccine prevents cancer, because pre-cancerous cell changes cannot ethically be left untreated for the 15 to 20 years cancer takes to develop; trials measured those changes instead. Almost all comparison groups received an aluminium-containing injection or another vaccine rather than salt water. Protection in 9- to 14-year-olds, the group most countries vaccinate, was inferred from antibody levels rather than measured. Nearly every trial was manufacturer-funded. S32 S05 S08

What the trials could not measure has since been measured in populations. A Cochrane review of November 2025 pooled 225 studies covering over 132 million people and found roughly 80% lower cervical cancer risk among those vaccinated at or before 16, at moderate certainty. An English mortality analysis published in June 2026 found no cervical cancer deaths among women aged 20 to 24 in 2020–2024. Both are observational. S06 S22

Documented side effects are mostly short-lived; the disputed ones remain disputed. Arm pain affects about 85% of recipients, headache about 13%. Fainting is listed as uncommon, severe allergic reaction at about 1.7 per million doses. A French study of 2.2 million girls found a probable excess of Guillain-Barré syndrome of one to two cases per 100 000, unreplicated elsewhere; the agency judged the benefit still far greater. Claims of chronic fatigue, complex regional pain syndrome, ovarian failure and infertility have been examined repeatedly and not confirmed. S15 S13 S06

The commercial stake is large and one-sided. Merck’s Gardasil products earned USD 63.9 billion between 2007 and 2025, peaking at USD 8.9 billion in 2023 — about a seventh of the company’s sales that year. Research that measures the effect of industry funding finds it acts on which questions are asked and how results are framed, not on how carefully trials are run. S61 S62

Almost nowhere is it compulsory. The vaccine is recommended and publicly funded in 162 countries. Five United States jurisdictions require it for school entry, with exemptions. Poland has a draft regulation making it obligatory from 1 January 2027 for ages 9 to 15. Global coverage of girls is 33%, against a WHO target of 90% by 2030. S04 S48 S43

Method and status. Sources are regulatory documents, systematic reviews by supportive and by critical authors, national registries and official statistics. Research, calculations and writing were AI-assisted. No independent human or specialist medical review has taken place. This is not medical advice and contains no recommendation for or against vaccination.

1. What the virus is, and what it does

Human papillomavirus is a family of more than 170 related viruses. Most cause nothing worse than warts. About a dozen are classified as high-risk because, over many years, a persistent infection can turn normal cells cancerous. S10

The virus passes through skin-to-skin and skin-to-mucosa contact, usually sexual. It is the most common sexually transmitted infection in the world. WHO states that almost all sexually active people are infected at some point, generally without symptoms, and that in about 90% of people the body clears the infection on its own. S03

The minority of infections that persist are the problem. Persistent infection of the cervix causes about 95% of cervical cancers. WHO states that it usually takes 15 to 20 years for cervical cancer to develop after infection. Two virus types, HPV 16 and HPV 18, account for around 76% of cervical cancers on the WHO figure; the 2018 Cochrane review puts the same figure at approximately 70%. The difference reflects different type-attribution studies rather than a dispute. S02 S05

This long delay is the single most important fact for understanding the argument. It explains why the vaccine is given to children long before any risk arises, why no trial could have measured cancer, and why the strongest evidence on cancer only became available roughly fifteen years after vaccination began.

The cervix is not the only site. HPV is also found in cancers of the oropharynx — the middle part of the throat, including the tonsils and base of the tongue — and of the anus, penis, vulva and vagina. WHO’s estimate is that HPV caused about 620 000 cancers in women and 70 000 in men in 2019, and that cervical cancers make up over 90% of HPV-related cancers in women. Throat cancer linked to HPV is rising in men in high-income countries and, unlike cervical cancer, has no screening programme. S03 S49

A distinction matters here. Some published totals count every cancer occurring at a site where HPV is sometimes found, which produces much larger numbers. The figures used in this article count only cancers estimated to be caused by HPV.

2. The size of the problem

The International Agency for Research on Cancer, part of WHO, publishes modelled global estimates under the name GLOBOCAN. Its 2024 estimates, issued on 8 July 2026, put cervical cancer ninth among all cancers for new cases and tenth for deaths: 604 196 new cases and 279 583 deaths. S01

The regional split is uneven. Asia accounts for 336 441 cases and 141 639 deaths; Africa for 124 627 cases and 73 397 deaths; Latin America and the Caribbean for 66 200 and 33 068; Europe for 58 901 and 24 612; Northern America for 15 417 and 5 588; Oceania for 2 610 and 1 279. Africa has 20.6% of cases but 26.3% of deaths, which reflects later detection and less access to treatment rather than a more aggressive disease. S01

Rates adjusted for age and population size show a different picture from raw counts. Malawi has the highest recorded incidence rate at 93.4 per 100 000 women and the highest mortality rate at 55.2. Western Europe, Northern America and Australia–New Zealand sit near the bottom of the international range, largely because organised screening has been in place for decades. S01

In Germany, the national cancer registry centre records about 4 300 women diagnosed with cervical cancer in 2023 and about 1 413 deaths, with a relative five-year survival of 69%. The Robert Koch Institute estimates that about 10 000 people a year in Germany develop an HPV-caused cancer, roughly 7 000 women and 3 000 men. S40 S39

These are estimates, not counts. GLOBOCAN models national figures from registry data of varying completeness; the poorer the registry, the wider the real uncertainty. Two consecutive GLOBOCAN rounds can differ for methodological reasons as well as real change: the 2022 round produced 662 000 cases and 348 000 deaths, the 2024 round 604 196 and 279 583. That difference should not be read as a fall of 68 000 deaths in two years. S01 S02

WHO has set a target of eliminating cervical cancer as a public health problem, defined as fewer than four new cases per 100 000 women per year. The route to it is summarised as 90–70–90 by 2030: 90% of girls fully vaccinated by age 15, 70% of women screened twice with a high-performance test by ages 35 and 45, and 90% of women with disease treated. S02

3. The vaccines, and how they were licensed

The vaccines contain no virus and no viral DNA. They contain virus-like particles: empty protein shells with the outer shape of the virus and nothing inside. The immune system responds to the shape. Because there is no genetic material, the vaccine cannot cause an infection. S15

Three products dominate. Gardasil, covering types 6, 11, 16 and 18, was approved in the United States in 2006. Cervarix, covering 16 and 18, followed in 2007 in Europe and 2009 in the United States. Gardasil 9, adding types 31, 33, 45, 52 and 58, was approved in 2014 and is now the only HPV vaccine used in the United States. A Chinese-made vaccine, Cecolin, covering types 16 and 18, received WHO prequalification on 4 October 2021 — the quality assessment that allows United Nations agencies and the vaccine alliance Gavi to buy a product — and is now widely used in lower-income programmes. S10 S60

Each contains an adjuvant — a substance added to provoke a stronger immune response. Gardasil and Gardasil 9 use an aluminium compound called amorphous aluminium hydroxyphosphate sulfate, referred to below as the aluminium adjuvant. Cervarix uses a different aluminium-based formulation. The adjuvant is central to the safety dispute described in section 5. S15

The schedule has shortened twice. Originally three doses, then two for younger adolescents, and since 2022 WHO has accepted a single dose for girls aged 9 to 20 as an alternative giving comparable protection. By the end of 2025, 93 countries used a one-dose schedule and more than 85% of girls aged 9 to 14 vaccinated that year received a single dose. S36 S04

Price varies by more than fivefold according to a country’s income. For countries supported by Gavi, the vaccine alliance, UNICEF lists the Chinese-made bivalent vaccine at USD 2.90 per dose for 2025 and 2026 and GlaxoSmithKline’s bivalent vaccine at USD 4.60. For middle-income countries above a defined income threshold the same GSK product is listed at USD 14.14. S50

4. What the trials tested, and what they did not

The user-facing question is whether the vaccine was investigated completely and ideally. The honest answer has two parts: the programme met the standards regulators apply, and it departed from a textbook ideal in five identifiable ways. Both parts are documented in the regulators’ and reviewers’ own records.

The endpoint was not cancer. No randomised trial measured whether the vaccine prevents cervical cancer. An expert meeting convened by IARC and the United States National Cancer Institute recorded the reason plainly: “Protection against invasive cancer was not used as an end-point because the standard of care requires treatment of premalignant disease, which is recognized as being on the causal pathway to invasive cancer.” In other words, once a trial detects pre-cancerous change, it must be treated; the trial therefore cannot observe whether it would have become cancer. WHO and regulators endorsed grade 2 and 3 cervical intraepithelial neoplasia — abnormal cell changes, abbreviated CIN2 and CIN3 — and persistent infection as substitute measures, with reduction in disease “being verified by post-licensure monitoring.” S32

The comparison group was rarely salt water. This is set out in section 5.

The target age group was not tested for efficacy. Most programmes vaccinate at ages 9 to 14. The trials that measured protection against lesions enrolled women aged 15 to 26. Protection in younger children was inferred by showing that their antibody levels after vaccination were at least as high as in the age group where protection had been demonstrated — a method called immunobridging. It is accepted for licensing. It is not direct evidence of benefit. The Cochrane network meta-analysis published in November 2025 states the position without qualification: “no data on pre-cancer outcomes were available for vaccination under age 15 years.” S07 S33

Follow-up in the randomised phase was short. The pivotal trials followed participants for around four years. The trial supporting Gardasil 9 collected efficacy data over 42 months. Longer evidence comes from extension studies and national registries: a Nordic follow-up reported durable protection against HPV 16/18-related high-grade lesions for at least 12 years with a trend continuing to 14, and no sign of waning antibody response. Those extensions are no longer randomised comparisons. S33 S46

Almost all the trials were funded by the manufacturers. The 2018 Cochrane review states: “All but one of the trials was funded by the vaccine manufacturers.” Manufacturer funding is normal in vaccine development and does not by itself invalidate a result, but it is a declared interest that applies to nearly the whole randomised evidence base. Section 6 sets out what is known about the effect of that arrangement, and section 7 sets out the size of the commercial interest behind it. S05

A sixth point concerns access rather than design. The critics who obtained the manufacturers’ full clinical study reports from the European Medicines Agency reported that after more than three years of requests they were still missing trials covering over 21% of all participants. Independent verification of the complete dataset has therefore not been possible. S09

Against these limits, the randomised evidence base is large by vaccine standards. The 2025 Cochrane network meta-analysis covers 60 randomised trials and 157 414 participants, with follow-up from seven months to 11 years, and obtained clinical study reports for 33 of them. S07

5. The comparison-group dispute

In a drug trial, the comparison group ideally receives something with no effect, so that any difference in side effects can be attributed to the product. In the HPV vaccine trials this mostly did not happen.

The numbers are not in dispute. Counting from the manufacturers’ own clinical study reports obtained from the European Medicines Agency and GlaxoSmithKline, 48 289 of 48 595 control participants — 99% — received an active comparator: either an aluminium-containing injection or a hepatitis vaccine. Only 306 received a saline injection. S08

Two arguments follow from the same fact.

The design rationale is that an injection containing the aluminium adjuvant without the virus-like particles produces similar arm pain and swelling, which keeps participants and investigators from working out who received what. Where a hepatitis A vaccine was used instead, as in the Costa Rica trial of Cervarix, the control group also received a genuine health benefit rather than nothing. Blinding protects the measurement of benefit as well as harm.

The objection is that if the comparison injection can itself cause reactions, the trial cannot show how often the vaccine causes them relative to no injection at all. Critics also point to what participants were told. A 2024 analysis of Danish trial documents by Lucija Tomljenovic and Leemon McHenry reports that recruitment material and consent forms described the comparator as saline or an inactive substance when the actual comparator contained the proprietary aluminium adjuvant. Those authors write from a position critical of the vaccine, and their paper is an argument about research ethics rather than a finding about the vaccine’s effects; the documentary point it makes about consent wording has not been rebutted in the sources examined here. S51

Three qualifications belong with this dispute. First, it affects the measurement of harm far more than the measurement of benefit, because an aluminium injection would not be expected to prevent HPV infection. Second, it is not unique to HPV vaccines; adjuvant comparators are common in vaccine trials. Third, the question of whether aluminium adjuvants themselves cause chronic disease has been studied separately: a Danish cohort of 1 224 176 children published in the Annals of Internal Medicine in July 2025 found results incompatible with moderate or large increases in autoimmune, allergic or neurodevelopmental disorders, while noting that small increases could not be excluded for some rarer outcomes. S59

6. Who paid for the evidence

The 2018 Cochrane review records the funding position in one sentence: “All but one of the trials was funded by the vaccine manufacturers.” That is the fact behind the conflict-of-interest objection, and no one disputes it. S05

Why it happens. Medicines law puts the burden of proof on the company that wants the licence. A full European marketing-authorisation application must contain the applicant’s own pharmaceutical, pre-clinical and clinical test results. United States regulation makes the trial sponsor accountable for monitoring, for the accuracy of the data and for compliance with the protocol. The party with the commercial interest is therefore the party the law obliges to produce and submit the evidence. Manufacturer funding of licensing trials is how the system is built, not a deviation from it. S64

Whether it is standard. It is. Roughly seven tenths of drug-trial money in the United States comes from industry. In a sample of 1 977 trials registered in 2020 and 2021, vaccine trials specifically were 42.8% industry-funded and 35.4% publicly funded — a more balanced picture than for medicines generally, but still one in which the manufacturer pays the largest single share. HPV vaccines are not unusual in this respect. They are also not exempt from what the arrangement does to a body of research. S63

What funding demonstrably does. This has been measured. A Cochrane methodology review by Lundh and colleagues, covering 75 studies of the question itself, found that industry-sponsored research more often reported favourable efficacy results — risk ratio 1.27 (1.17 to 1.37), moderate-quality evidence — and more often reached favourable conclusions, risk ratio 1.34 (1.19 to 1.51), low-quality evidence. On harms the estimate was 1.37, but the confidence interval ran from 0.64 to 2.93 and the evidence was rated very low quality, so nothing follows from it. S62

The detail that matters most is where the difference did not appear. The review found no difference between industry and non-industry studies in how participants were randomised, how allocation was concealed, how completely people were followed up, or whether outcomes were selectively reported. Industry-funded trials were in fact more often at low risk of bias from blinding, risk ratio 1.25 (1.05 to 1.50). What did differ was the fit between findings and conclusions: industry studies showed less agreement between their own results and the conclusions drawn from them, risk ratio 0.83 (0.70 to 0.98). The authors’ summary was that company sponsorship “leads to more favorable efficacy results and conclusions” and reflects “an industry bias that cannot be explained by standard ‘Risk of bias’ assessments.” S62

That finding is precise, and it cuts in both directions. Commercially funded trials are not run more carelessly; on the measurable quality criteria they are often run better. The influence operates earlier and later than the conduct of the trial — in which question is asked, which comparator is chosen, which outcome is made the primary one, and how the result is summarised. Those are exactly the features the critics identify in sections 4 and 5: a substitute endpoint rather than cancer, an aluminium-containing comparator rather than salt water, and a design built to detect benefit rather than harm. The general research finding and the specific criticism describe the same mechanism.

What it does not establish. A measured tendency across a literature is not a finding about any particular result. The efficacy conclusions have since been reproduced by parties with no commercial interest: national cancer registries, two Cochrane reviews, and a 2026 evidence review whose authors declared no pharmaceutical funding and had their declarations checked by an outside party. Sponsorship bias is a reason to have a result checked independently. In this case the independent check has largely been carried out, and it pointed the same way. S06 S10

Whether independent institutions could do it instead. In part they already have, and it changed policy. The Costa Rica Vaccine Trial was sponsored and funded by the United States National Cancer Institute under a government contract, with the institute and Costa Rican investigators responsible for the design, the conduct, the data, the analysis and the published paper; the manufacturer supplied vaccine. That trial produced the first evidence that a single dose protects. The ESCUDDO trial, which enrolled more than 20 000 girls aged 12 to 16, was run by the National Institutes of Health with the National Cancer Institute, the Gates Foundation and WHO’s cancer agency. DoRIS in Tanzania and KEN SHE in Kenya were publicly and philanthropically funded. The single-dose schedule now used by 93 countries rests largely on work that no manufacturer paid for. S65 S34 S35 S04

The obstacle is cost, and who carries it. Mechanisms exist: Italy’s medicines agency runs an independent research fund financed by a levy equal to 5% of the promotional spending that pharmaceutical companies incur in the country, which paid for 282 studies totalling about EUR 43.5 million between 2005 and 2018. That is a working model. It is also roughly the budget of a single mid-sized trial, set against a registration programme that enrols tens of thousands of participants over several years. S66

The two positions, stated plainly. One holds that the party standing to profit should not also be the only one measuring the harm, and that public money should at minimum pay for the safety arm and for independent confirmatory trials. The other holds that the manufacturer has both the resources and the legal duty, and that the place to intervene is the safeguards — advance registration of the protocol, independent data monitoring, inspection, and publication of the full study reports. The HPV case supplies evidence for both. The study reports were in fact released by the European Medicines Agency, which is the only reason the critical review described in section 9 exists at all. The researchers who read them also reported that after more than three years of requests they were still missing trials covering over 21% of participants. S09

7. What the vaccines have earned

Reported sales are a matter of public record, because listed companies must disclose them. Profit is not. Neither Merck nor GSK reports profit for an individual product — segment level is as far as the filings go — and development costs are not broken out either. Every figure in this section is revenue, and no profit figure for any HPV vaccine can be derived from public filings.

Merck. Gardasil and Gardasil 9 dominate the market. Merck’s annual filings with the United States Securities and Exchange Commission record worldwide sales rising from USD 1.48 billion in 2007 to a peak of USD 8.89 billion in 2023, then USD 8.58 billion in 2024 and USD 5.23 billion in 2025. The total for 2007 to 2025 is USD 63.9 billion. The launch year 2006 is not included, because no figure for it was found in the filings. S61

One accounting fact belongs with that series. Until 31 December 2016, Merck sold vaccines in most major European markets through a joint venture with Sanofi. In the company’s own words, the reported figures “do not reflect sales of vaccines sold in most major European markets through the Company’s joint venture, Sanofi Pasteur MSD, the results of which are reflected in Equity income from affiliates.” From 2017 Merck recorded those European sales directly. The cumulative total is therefore a lower bound, and part of the rise between 2016 and 2018 is a change in what was counted rather than growth in what was sold. S61

Set against the company as a whole, Gardasil was 14.8% of Merck’s total sales in 2023, 13.4% in 2024 and 8.0% in 2025. The 39% fall in 2025 was concentrated in one market. Merck paused shipments to China in February 2025, later extended the pause to the end of the year citing weak demand and high stock levels, and withdrew a previously stated target of USD 11 billion in annual Gardasil sales by 2030. Sales outside the United States fell from USD 6.80 billion in 2023 to USD 2.59 billion in 2025, while United States sales rose from USD 2.08 billion to USD 2.64 billion. S61 S68

GSK. Cervarix was always the smaller product and is now a marginal one. GSK reported worldwide turnover of GBP 187 million in 2009 and GBP 242 million in 2010, then GBP 506 million in 2011 — its peak, which the company attributed to Japan’s national programme starting at the end of 2010. After Japan suspended its recommendation, sales fell 37% to GBP 172 million in 2013 and a further 26% to GBP 118 million in 2014. Worldwide turnover was about GBP 88 million in 2015, of which GBP 3 million in the United States. GSK stopped supplying the American market in 2016, citing very low demand against broader-spectrum competitors, and has not reported Cervarix as a separate line since. S67

Chinese manufacturers. Cecolin and Walrinvax now account for a large share of doses given, particularly in programmes supported by Gavi, where the vaccine costs USD 2.90 to USD 4.60 per dose against the USD 14.14 charged to middle-income countries above a defined income threshold. Their makers do not disclose HPV revenue in a form comparable to Merck’s product line, and their company results are dominated by the post-pandemic fall in Chinese vaccine demand. No like-for-like figure can be given, and none is offered here. S50 S68

What the figure establishes, and what it does not. It establishes the size of the commercial interest: one product family, one company, roughly USD 64 billion of revenue over nineteen years, at its height about one seventh of that company’s turnover. That is the stake behind every declaration of interest in this article, and it is a legitimate thing for a reader to weigh when judging who produced which piece of evidence. It does not, on its own, say anything about whether the vaccine works or whether it is safe. A large revenue figure is equally consistent with a widely used product that does what it claims. Those questions are settled on different evidence, and sections 8 to 11 set out what that evidence is.

8. What the trials found about benefit

The 2018 Cochrane review of 26 trials and 73 428 participants reported the clearest results in women aged 15 to 26 who had no high-risk HPV infection when they enrolled. In that group, high-certainty evidence showed CIN2 or worse associated with HPV 16/18 falling from 164 to 2 cases per 10 000, and CIN3 or worse from 70 to 0 per 10 000. Counting lesions caused by any HPV type rather than only 16 and 18, CIN2+ fell from 287 to 106 per 10 000. S05

The effect was smaller in two circumstances. Among women counted regardless of whether they were already infected, CIN2+ associated with HPV 16/18 fell from 341 to 157 per 10 000. In women vaccinated between ages 24 and 45 and not selected by infection status, the review found risks “similar between vaccinated and unvaccinated women.” The vaccine prevents new infection; it does not clear an existing one. S05

The 2025 Cochrane network meta-analysis reported the same pattern in updated form: in females aged 15 to 25, CIN2+ risk ratio 0.70 (95% confidence interval 0.56 to 0.88) counting any HPV type, and 0.40 (0.30 to 0.54) for types the vaccine covers, both moderate certainty. Anogenital warts fell by 25 cases per 1 000 participants, high certainty. Treatments for HPV-related disease fell by 12 per 1 000, moderate certainty. S07

The critical review of the clinical study reports by Lars Jørgensen, Peter Gøtzsche and Tom Jefferson reached benefit conclusions in the same direction but weaker. Over about four years it found HPV-related carcinoma in situ in 367 vaccinated participants versus 490 controls, a risk ratio of 0.73 with a confidence interval reaching 1.00, and HPV-related treatment procedures in 1 018 versus 1 416, a risk ratio of 0.71. Their headline on benefit was that the vaccines “decreased HPV-related cancer precursors and treatment procedures.” S08

For men, randomised evidence exists for infection and visible lesions rather than cancer. A trial of the quadrivalent vaccine in men aged 16 to 26 reported prevention of 86% of persistent infections with the four covered types and 90% of external genital lesions in the per-protocol group; in men who have sex with men, 95% of anal infections and 75% of high-grade anal lesions. No randomised trial has measured prevention of throat cancer in men. S45

Single-dose evidence has accumulated since 2022. A randomised trial in Kenya reported efficacy above 97% against persistent HPV 16 or 18 infection at 36 months after one dose. A Tanzanian trial in girls aged 9 to 14 found antibody responses after one dose still durable at five years. These trials measure infection and antibodies, not lesions or cancer. S34 S35

9. What the trials found about harm

The approved product information for Gardasil 9 rests on seven clinical studies covering 15 776 people, with side effects recorded on a card for 14 days after each injection. It lists injection-site reactions in 84.8% of recipients within five days of any visit and headache in 13.2% within 15 days, both “usually mild or moderate in intensity.” It records that 0.1% of participants stopped because of an adverse experience. S15

The frequency table lists dizziness, nausea, fever, tiredness, injection-site itching and bruising as common, meaning between 1 in 100 and 1 in 10. Swollen lymph nodes, fainting sometimes accompanied by jerking movements, vomiting, hives, joint pain, muscle pain, weakness, chills and malaise are listed as uncommon, between 1 in 1 000 and 1 in 100. Hypersensitivity is rare. Anaphylaxis — severe allergic reaction — is listed with frequency “not known,” because it was identified after marketing. S15

A separate table lists events reported after marketing of the quadrivalent vaccine, all at unknown frequency: injection-site cellulitis, idiopathic thrombocytopenic purpura, anaphylactoid reactions, bronchospasm, acute disseminated encephalomyelitis and Guillain-Barré syndrome. The document is explicit about what such reports establish: “Because these events were reported voluntarily from a population of uncertain size, it is not possible to reliably estimate their frequency or to establish, for all events, a causal relationship to vaccine exposure.” That caveat cuts both ways — it is the reason spontaneous reports cannot prove harm, and also the reason they are not dismissed outright. S15

WHO’s Global Advisory Committee on Vaccine Safety puts anaphylaxis at approximately 1.7 cases per million doses and describes fainting as a common anxiety or stress-related reaction to injection rather than an effect of the contents. Its 2017 statement, by which point more than 270 million doses had been distributed, said the committee “considers HPV vaccines to be extremely safe.” S16

On serious harm, the two systematic reviews reach different conclusions from overlapping data.

The 2018 Cochrane review found serious adverse events in 669 per 10 000 controls versus 656 per 10 000 vaccinated, risk ratio 0.98 (0.92 to 1.05), high certainty. It also recorded a mortality imbalance: 11 deaths per 10 000 in control groups against 14 per 10 000 with vaccine, risk ratio 1.29 (0.85 to 1.98), rated low certainty. The review noted that numbers were low overall, that deaths were more frequent in older women, that the reported causes were judged unrelated to vaccination, and that “no pattern in the cause or timing of death has been established.” On pregnancy, it found no increased risk of miscarriage or termination, and rated the evidence on congenital abnormalities and stillbirth as uncertain. The 2025 network meta-analysis found serious adverse events at risk ratio 0.99 (0.94 to 1.04), high certainty, and no deaths reported. S05 S07

The Jørgensen review of clinical study reports found overall serious harms essentially level — 1 404 versus 1 357, risk ratio 1.01 (0.94 to 1.08) — and deaths 45 versus 38, risk ratio 1.19 (0.65 to 2.19), a difference compatible with chance. Its two positive findings were general harms, 13 248 versus 12 394, risk ratio 1.07 (1.03 to 1.11), and serious nervous system disorders, 72 versus 46, risk ratio 1.49 (1.02 to 2.16). The authors labelled the second an exploratory analysis and concluded that “the extent to which the HPV vaccines’ benefits outweigh their harms is unclear.” They also stated their own principal limitation: serious harms were incompletely reported for 72% of participants, and the trials “were primarily designed to assess benefits and were not adequately designed to assess harms.” S08

That nervous-system finding is the single most contested number in the literature, and a commentary published alongside it in the same journal set out why. Hilda Bastian argued the authors were “on very shaky ground,” noting that the analyses “show no statistically significant increase in any individual serious or fatal adverse event”; that the original protocol had targeted postural orthostatic tachycardia syndrome and complex regional pain syndrome and was amended twice when that proved unfeasible, making the reported result post-hoc rather than pre-specified; and that risk ratios “unambiguously require knowing how many individuals were affected,” data the authors did not have, so that counting symptom events rather than affected people inflates apparent frequency. Bastian’s overall assessment was that the review confirms the benefit findings but does not provide a rigorous assessment of the alleged harms. S09

10. What happened after roll-out: disease

The evidence that the randomised trials could not produce has since come from national registries, which follow whole populations for far longer.

Sweden linked records for 1 672 983 girls and women aged 10 to 30 between 2006 and 2017. Invasive cervical cancer was diagnosed in 19 vaccinated and 538 unvaccinated women, a cumulative incidence of 47 versus 94 per 100 000. After adjustment, the incidence rate ratio was 0.12 (0.00 to 0.34) for those vaccinated before age 17 and 0.47 (0.27 to 0.75) for those vaccinated between 17 and 30. S19

Public Health Scotland reported in January 2024 that no cervical cancer cases had been detected in women fully vaccinated at age 12 to 13 since the Scottish programme began in 2008. The agency paired the finding with a caution from its own clinical lead: the vaccine “can’t protect against them all and it’s still important that you go for regular cervical screening.” S21

In England, an analysis of every cervical cancer death among women aged 20 to 34 from 2001 to 2024, published in The Lancet on 17 June 2026, found no cervical cancer deaths at all among women aged 20 to 24 in the years 2020 to 2024. About nine in ten of that cohort had been vaccinated at age 12 or 13. The authors estimated that roughly 23 deaths would have occurred without vaccination, and about 200 deaths avoided in England to the end of 2024. Against a 2000–2014 baseline, mortality fell about 80% among women aged 20 to 24 and about 69% among women aged 25 to 29. The study is observational and was funded by Cancer Research UK. S22

The most comprehensive synthesis is the Cochrane review published on 24 November 2025, which pooled 225 studies covering more than 132 million people. Across five cohort studies with 4 390 243 women, the adjusted risk ratio for cervical cancer after vaccination was 0.37 (0.25 to 0.56). For those vaccinated at or before age 16, covering 4.54 million person-years, the risk ratio was 0.20 (0.09 to 0.44) — an 80% reduction. CIN3+ fell 74% in that group, risk ratio 0.26 (0.12 to 0.56). Anogenital warts fell 53% in the long term. All these were rated moderate certainty. The authors note that risk of bias across included studies ranged from moderate to critical. S06

Not every recent study produces large numbers. A multinational study published in The Lancet Primary Care in March 2026, using primary-care and registry data from the United Kingdom, Catalonia and Norway and a method designed to imitate a randomised trial, estimated vaccine effectiveness at 15 years of 42% (6 to 64) against CIN2+ and 58% against surgical removal of cervical tissue. It could not estimate effectiveness against invasive cervical cancer at all, because there were too few cases — the participants were still young and follow-up reached at most 16 years. S23

Australia offers the longest programme record. It vaccinated girls from 2007 and boys from 2013 and moved early to HPV-based screening. Genital wart diagnoses in young heterosexual men, who were not themselves vaccinated in the first phase, fell 49% during the female-only period and 89% once boys were included. Australia is projected to eliminate cervical cancer by 2035. Coverage there has slipped to 83.0% across both sexes by age 15 in 2023, below the national target, and screening participation has also declined. S24 S25

One caveat applies to all of this. Registry comparisons are not randomised. People who accept vaccination may differ systematically from those who do not, in screening attendance, smoking or sexual behaviour, and statistical adjustment can only partly correct for that. The scale and consistency of the findings across countries with different health systems is the main argument against attributing them to selection effects; it is not a proof.

11. What happened after roll-out: safety monitoring

Independent of the trials, national health systems have examined whether vaccinated populations develop specific conditions more often.

In 2013, a Danish and Swedish cohort followed 997 585 girls aged 10 to 17, of whom 296 826 received 696 420 doses, and looked for 53 autoimmune, neurological and clotting outcomes up to 180 days after each dose. It found no evidence of association. S17

The largest contrary finding is French. In September 2015, the national medicines agency ANSM and the health insurance fund published a cohort of 2.2 million girls aged 13 to 16, about 840 000 of them vaccinated, examining 14 autoimmune conditions. For twelve of the fourteen and for the group as a whole, it found no association. For inflammatory bowel disease and for Guillain-Barré syndrome — a rare condition in which the immune system attacks peripheral nerves, usually with recovery — it found statistically significant associations, and concluded that an increased risk of Guillain-Barré after HPV vaccination “appears probable,” of the order of one to two additional cases per 100 000 girls vaccinated. The agency’s conclusion was that “the expected public health benefits of this vaccination remain far greater than the risks.” Its current guidance repeats the figure and adds that the finding requires confirmation, since no other study worldwide has reproduced it. S13 S14

The 2025 Cochrane population-level review examined the conditions most often raised online and found, at moderate certainty, no increased risk of postural orthostatic tachycardia syndrome, chronic fatigue syndrome or myalgic encephalomyelitis, paralysis, complex regional pain syndrome, premature ovarian failure, infertility, or changed sexual activity. On Guillain-Barré it found evidence suggesting no increased risk, rated low certainty — a weaker rating that leaves the French signal unresolved rather than refuted. It also recorded a gap: no studies reported on community rates of serious adverse events generally. S06

In 2015 the European Commission, at Denmark’s request, asked the European Medicines Agency to examine complex regional pain syndrome and postural orthostatic tachycardia syndrome. The agency’s committee reviewed published research, trial data, reports from patients and clinicians, member state data, expert advice and material from patient advocacy groups. It concluded on 5 November 2015 that the evidence does not support a causal link, that rates in vaccinated girls did not differ from expected rates, and that no change to the product information was warranted. It cited background rates of about 150 cases per million girls and young women aged 10 to 19 per year for each condition — a figure that matters, because both conditions occur in unvaccinated adolescents at rates high enough to generate coincidental cases in any large vaccination programme. S11

The Nordic Cochrane Centre formally complained in May 2016 that the agency had mishandled the review. The European Ombudsman examined the complaint in case 1475/2016/JAS and concluded in October 2017 that there had been no maladministration and that the agency’s conflict-of-interest policy had been complied with. The critics have not accepted that outcome. S12

The most recent independent synthesis is an evidence review by the Vaccine Integrity Project at the University of Minnesota’s Center for Infectious Disease Research and Policy, funded by the Rockefeller Foundation and a foundation grant with no pharmaceutical funding, and published in 2026. It screened more than 5 000 abstracts, evaluated studies published between September 2024 and January 2026, and combined them with the two Cochrane reviews, synthesising 274 studies in total. Its overall finding was that the evidence “continues to show no association between HPV vaccination and serious adverse events.” S10

That review also documented the one place where recent evidence diverges. Of three new studies of postural orthostatic tachycardia syndrome, two reported a positive association. A self-controlled study of commercially insured United States girls and women found an increased risk of autonomic dysfunction after adjusting for age, incidence rate ratio 1.23 (1.08 to 1.41), driven by 18- to 26-year-olds at 1.40 (1.12 to 1.75), with no significant increase among 9- to 17-year-olds at 1.14 (0.97 to 1.35). A case series using the United States spontaneous reporting database found a reporting odds ratio of 10.39 (7.75 to 13.93) among serious reports. The reviewers rated that second study at critical risk of bias, on the ground that it compared HPV vaccination in adolescents with vaccines mostly given in infancy, when the condition is essentially unknown, so the comparison largely measures age rather than vaccine. Their conclusion was that “the preponderance of evidence, including high-quality data from randomized trials, continues to indicate no association between HPV vaccination and POTS.” S10

Litigation has run in parallel. Product liability claims against Merck, mostly alleging postural orthostatic tachycardia syndrome and premature ovarian insufficiency, were consolidated into a federal multidistrict litigation in North Carolina. The presiding judge dismissed the bulk of them, finding that the plaintiffs’ expert evidence could not support a causal conclusion. In June 2026 Merck agreed to pay more than USD 50 million to resolve more than 200 cases without admitting liability, stating that research “continues to support the safety and efficacy of our HPV vaccines” and that the settlement was less than its anticipated defence costs. A settlement without admission establishes no medical fact in either direction. S55

12. Four episodes that shaped public opinion

Public confidence in this vaccine has collapsed in several countries after specific events, and the pattern is consistent enough to be part of the evidence.

India, 2010. A demonstration project run by the organisation PATH vaccinated around 24 000 girls aged 10 to 14 in two states. It was suspended in April 2010 after seven deaths. Government investigations concluded that the deaths had unrelated causes including snake bite, malaria, epilepsy and suicide. A parliamentary committee nonetheless criticised the delay in making the deaths public and the absence of an independent investigation, and campaigners raised questions about consent procedures in the communities involved. The episode is cited by critics as evidence of institutional carelessness and by defenders as evidence that deaths after vaccination are not deaths from vaccination. S30 S31

Japan, 2013. After media reports of girls with chronic pain and walking difficulties, the health ministry suspended its proactive recommendation in June 2013 while keeping the vaccine formally available. Coverage fell from over 70% to below 1%. A survey of 71 177 women in Nagoya, published in 2018, found no significant increase in any of 24 reported post-vaccination symptoms among vaccinated women, though it did find higher odds of hospital visits for heavy or irregular menstrual bleeding and severe headache. The recommendation was restored in April 2022, nine years later, with a catch-up programme for those who missed it. S26 S27

Denmark, 2015. In March 2015 the public broadcaster TV 2 aired a documentary, The Vaccinated Girls — Sick and Abandoned, presenting accounts of young women who believed the vaccine had made them ill. Compliance fell from over 90% to under 30%. A public information campaign followed and uptake returned to its earlier level, helped by catch-up doses, but an estimated 26 000 fewer girls started the course than would otherwise have done. In February 2018 TV 2 acknowledged that its programme had contributed to the failure of the vaccination programme, while saying that had not been its intention. S28

Colombia, 2014. Colombia’s school-based programme reached over 90% of the target group in its first year. In May 2014, 15 girls at one school in Carmen de Bolívar were admitted to hospital after vaccination; video of girls fainting spread through national and social media, and more than 600 cases were reported across the country. Health authorities investigated and concluded there was no organic link to the vaccine, classifying the episode as mass psychogenic illness. Uptake nonetheless fell from 98% for the first dose and 88% for the full course in 2012 to 14% and 5% by 2016. S29

The common feature is that in each case official investigation found no causal link, and in each case the investigation did not restore confidence by itself. Denmark is the one clear recovery, and it took a deliberate communication campaign over roughly three years.

13. The critics, named

Three groups of expert criticism can be distinguished, and they are not equivalent.

Methodologists working from the trial documents. Peter Gøtzsche, a co-founder of the Cochrane Collaboration and former head of its Nordic centre, Tom Jefferson, and Lars Jørgensen have published the most substantial critical work. In July 2018 they argued in BMJ Evidence-Based Medicine that the 2018 Cochrane review was incomplete, had missed roughly half the eligible trials, and had not adequately handled reporting bias or author conflicts of interest. Cochrane’s editor-in-chief replied on 3 September 2018 that the criticisms were overstated and would not change the review’s conclusions, while accepting that some points would inform the next version, and separately corrected the authorship to comply with its conflict-of-interest policy. Gøtzsche was expelled from the Cochrane governing board later that month, in a dispute that was partly about this review and partly about wider governance. Their 2020 review of the clinical study reports, discussed in section 9, is the substantive output of that line of work. S52 S53 S08

Assessing this work requires separating two questions. Their procedural criticisms — that adjuvant comparators limit harm detection, that trials were designed around benefit, that full data access was not granted — are recognised as legitimate issues by people who disagree with their conclusions, including Bastian. Their substantive harm finding is the contested part, and the objections to it are specific and technical rather than dismissive.

Researchers making stronger causal claims. A second group has argued that the vaccine causes specific syndromes. The claims that have been tested — complex regional pain syndrome, postural orthostatic tachycardia syndrome, premature ovarian insufficiency, chronic fatigue — have been examined by the European Medicines Agency, WHO’s safety committee, national cohorts and two Cochrane reviews, and have not been confirmed. Some publications in this group have been retracted or heavily criticised for methodology. Their arguments are documented in this article where an official body has responded to them, which is the fairest available test.

Scientists whose positions are misreported. Diane Harper, a lead investigator on the original clinical trials, is widely quoted in anti-vaccination material as having turned against the product. She has denied the quotations attributed to her, stating that she did not say the vaccine could be more dangerous than cervical cancer. Her own publications support vaccination: in 2021 she co-authored a paper in The Lancet Infectious Diseases titled “Elimination of cervical cancer depends on HPV vaccination and primary HPV screening.” Readers encountering a striking quotation from a named scientist have a practical check available: look for what that scientist has actually published. S54

There is also criticism that is neither about safety nor about efficacy. Some public health researchers argue that in countries with strong screening programmes the additional benefit per pound spent is smaller than in countries without them, and that screening must continue regardless because the vaccine does not cover every cancer-causing type. That argument is compatible with the vaccine working exactly as described.

14. Where the disagreement actually lies

Setting the positions side by side narrows the dispute considerably.

Question	Broad agreement	Genuine dispute
Does HPV cause cervical cancer?	Yes, established	None identified
Do the vaccines reduce pre-cancerous lesions?	Yes; critics’ own review found reductions	Size of effect in unselected populations
Were trials designed to detect rare harms?	No; both sides say so	Whether this leaves harms undetected or merely unmeasured
Was the comparison group inert?	No, in 99% of cases	Whether this invalidates the safety conclusions
Is there a proven link to POTS, CRPS or ovarian failure?	Not established by regulators or Cochrane	Whether absence of proof reflects absence of effect or inadequate study
Guillain-Barré syndrome	Not found in most national studies	One French study found a probable excess of 1–2 per 100 000
Does the vaccine prevent cancer?	Registry evidence points strongly that way	Whether observational data can settle it

The deepest divide is not about any single number. It is about what to conclude when a well-conducted study finds nothing. One position holds that after hundreds of millions of doses, multiple national cohorts and two independent Cochrane reviews, the repeated failure to find a signal is itself strong evidence that no substantial signal exists. The other holds that if the instruments were built to measure benefit, their silence about harm carries less weight — and that an exploratory finding of 72 versus 46 serious nervous system events should have prompted a properly designed trial rather than a rebuttal.

Neither position can be resolved by citing more of the same kind of study. What would resolve it is a large randomised trial with a genuinely inert comparator and pre-specified neurological outcomes. No such trial has been conducted, and after two decades of licensed use it is unlikely that one will be, because randomising children to no protection against a preventable cancer would not now pass an ethics committee. That is a real and permanent limit on what can be known, and it is worth stating plainly rather than arguing around.

15. Recommended, funded, or required

The premise that HPV vaccination is becoming compulsory for teenagers is, as of the cutoff date, largely not yet the case.

The vaccine was in national programmes in 162 countries in 2025, including 15 new introductions that year. In almost all of them it is recommended and publicly funded, not required. Global coverage of girls with a first dose stands at 33%, up from 17% in 2019 but far from the 90% target. S04

In the United States, five jurisdictions — Hawaii, Puerto Rico, Rhode Island, Virginia and the District of Columbia — require HPV vaccination before a particular school grade, with exemptions available. This is a condition of school entry rather than a legal duty to vaccinate. National coverage of at least one dose among 13- to 17-year-olds was 78% in 2024, below the 91% for tetanus-diphtheria-pertussis booster and 90% for meningococcal vaccine. S48 S38

United States federal recommendations have been under unusual scrutiny. Since February 2025 the Department of Health and Human Services has changed several long-standing vaccine recommendations, and the reconstituted advisory committee has taken particular interest in aluminium adjuvants. A Congressional Research Service summary updated on 29 July 2026, listing every recommendation change made in 2025 and 2026, contains no change to HPV vaccine recommendations. HPV remains on the committee’s working agenda. S44

In Germany, the standing vaccination committee recommends HPV vaccination for all girls and boys from age 9 to 14, with two doses at least five months apart and catch-up to age 17. Coverage measured at age 15 was 55% for girls and 34% for boys in December 2024. Vaccination is not compulsory. S39

England’s school programme, once close to 90%, has not recovered from the pandemic. In the 2024/25 academic year, coverage in school year 9 was 75.3% for girls and 70.5% for boys, ranging from 62.6% and 57.7% in London to 81.1% and 76.2% in the East of England. S37

Poland is the clearest case of a country moving from voluntary to compulsory. A universal free programme began in June 2023 for 12- and 13-year-olds and was extended on 1 September 2024 to all children aged 9 to 14. The national immunisation schedule for 2026 still lists HPV as recommended rather than obligatory. Uptake has been low: a Ministry of Health answer to a parliamentary question, published in December 2025, reported 585 580 children vaccinated across the 2010–2016 birth cohorts, or 20.15%, against a national target of 60% by 2028. A peer-reviewed cross-sectional analysis of the programme’s first phase put participation at 8.67%. A draft amendment to the regulation on compulsory vaccinations, published on the government legislation portal on 20 October 2025, would add HPV to the compulsory list from 1 January 2027 for children from completed age 9 to completed age 15, beginning with those born in 2018. At the cutoff date this is a draft in the legislative process, and readers should verify its current status before relying on it. S41 S42 S57 S43

Whether a recommended vaccine should become a legal obligation is a question about the proper limits of state authority over children’s medical care, not a question that evidence about the vaccine can answer. Countries with very similar evidence have reached opposite answers.

16. What is still unknown

Several questions remain open, and they are open for both sides.

How long protection lasts. Antibody responses and protection against high-grade lesions have held for at least 12 to 14 years in follow-up studies, with no evidence of waning. Nobody has thirty-year data, because the vaccine is not thirty years old. Whether a booster will eventually be needed is unresolved. S46

Whether other virus types will fill the gap. If vaccination removes the types it covers, other high-risk types could in principle become more common. Surveillance so far has not found systematic increases in non-vaccine types, and has found evidence of cross-protection against types genetically related to HPV 16. Specialists nonetheless describe it as too early to rule out, and one recent study reported a 59% increase in one type, HPV 66. Continued monitoring is the agreed position. S47

Benefit in men beyond infection. Randomised evidence in men covers infection and visible lesions. Prevention of HPV-related throat cancer, the fastest-growing HPV cancer in high-income countries, has not been demonstrated in a randomised trial and would take decades to demonstrate. The 2026 evidence review identified one study reporting a protective effect against head and neck, oesophageal, anal and penile cancers in adolescent and young adult males, and noted the absence of dose-comparison efficacy data in males entirely. S10

Rare harms below the detection threshold. A harm occurring in, say, one in 200 000 people cannot be detected by a trial of 15 000, and is difficult to distinguish from background in observational data. The French Guillain-Barré signal is exactly in this range and remains unreplicated. The honest statement is that no substantial rare harm has been identified despite extensive search, and that a very rare one could not be excluded by the methods used.

Whether single doses hold up over decades. Single-dose protection is supported by randomised trials measuring infection and by immunogenicity data, and has been adopted by 93 countries. Its long-term performance against lesions and cancer is still being established. S04 S34

Screening remains necessary. No available vaccine covers every cancer-causing type. Every authority cited here, including those reporting the largest benefits, states that vaccinated women should continue cervical screening. S21

Method and evidence limits

This article draws on regulatory product information and assessment documents from the European Medicines Agency, the French ANSM and the United States Food and Drug Administration; systematic reviews from Cochrane and from authors critical of the vaccine; an independently funded 2026 evidence review; national cancer registries and immunisation statistics; company annual filings lodged with the United States Securities and Exchange Commission; and the published record of the disputes themselves, including the responses of the bodies criticised. Where an author has a declared position for or against the vaccine, this is stated in the text.

Figures are reported with their source, year and definition. Where two authoritative sources give different figures for the same quantity — the share of cervical cancers caused by HPV 16 and 18, or the global burden in successive GLOBOCAN rounds — both are shown rather than one selected. Coverage figures use each country’s own definition of age group, sex and dose count and are not exactly comparable.

Sales figures come from the companies’ own filings and were read from the original documents, with two limits stated in the text: no figure for the 2006 launch year was found, and Merck’s own footnotes record that figures up to 2016 exclude most European sales. The later Cervarix figures and the reason for Merck’s 2025 fall in China rest on trade reporting rather than a filing. The statements about how commonly trials are industry-funded, about the legal obligation on the applicant, and about Italy’s independent research fund were established from summaries of the published sources rather than from the full documents.

Several load-bearing documents could not be retrieved directly. The full texts of The Lancet’s June 2026 mortality analysis and March 2026 effectiveness study, and of the Cochrane network meta-analysis, returned access restrictions; their figures here come from the publishing institution’s own release, from the accessible abstract, or from the open-access version, and are labelled accordingly. The 2018 Cochrane review and the November 2025 Cochrane population-level review were read in full from open-access copies.

The status of Poland’s draft regulation was established from the legislative record as reported in December 2025 and February 2026. Whether it has since been signed and published in the official journal was not confirmed and should be checked before the date is relied on.

Research, calculations, charts and writing for both language editions were AI-assisted. No independent human review, specialist medical review, peer review or fact-check by a third party has taken place. Nothing here is medical advice, and the article expresses no recommendation for or against vaccination in any individual case. A person deciding about vaccination for themselves or a child should discuss it with a qualified clinician who knows their history.

Sources and exact reference points

• S01 · gco.iarc.who.int — GLOBOCAN 2024 cervix uteri fact sheet, dated 8 July 2026: rank, case and death totals, age-standardised rates, continental breakdown, national extremes

• S02 · who.int — Cervical cancer fact sheet, 3 July 2026: 2024 totals, 76% HPV 16/18 share, 95% HPV attribution, elimination threshold, 90–70–90 targets

• S03 · who.int — HPV and cancer fact sheet, 5 March 2024: 620 000 female and 70 000 male cancers in 2019, 90% clearance, 15–20 year interval, schedule

• S04 · who.int — Immunization coverage fact sheet, 15 July 2026: 33% first-dose coverage in girls, 162 countries, 93 one-dose schedules, 85% single-dose share

• S05 · cochranelibrary.com — Arbyn et al, Cochrane 2018, abstract: 26 trials / 73 428 participants, CIN2+ and CIN3+ risks per 10 000 by baseline status and age, serious adverse events RR 0.98, mortality RR 1.29, pregnancy outcomes, manufacturer funding

• S06 · cochranelibrary.com — Henschke et al, Cochrane, 24 November 2025, abstract: 225 studies / 132 million people, cervical cancer RR 0.37 and RR 0.20 for age ≤16, CIN3+ RR 0.26, anogenital warts, specific adverse-event findings and certainty ratings

• S07 · cochranelibrary.com — Bergman et al, Cochrane network meta-analysis, 24 November 2025: 60 RCTs / 157 414 participants, CIN2+ RR 0.70 and 0.40, serious adverse events RR 0.99, absence of cancer and under-15 pre-cancer data

• S08 · pmc.ncbi.nlm.nih.gov — Jørgensen, Gøtzsche and Jefferson, Systematic Reviews 9:43 (2020): comparator composition 48 289 / 48 595, benefit and harm risk ratios, exploratory nervous-system analysis, authors’ stated limitations

• S09 · pmc.ncbi.nlm.nih.gov — Bastian, Systematic Reviews 9:41 (2020), commentary: protocol amendments, event-versus-person counting, missing trials for over 21% of participants

• S10 · vaxintegrity.cidrap.umn.edu — Vaccine Integrity Project, CIDRAP, 2026 systematic review: funding and interests statement, 274 studies, POTS study-by-study appraisal, effectiveness findings, male and dose-comparison gaps

• S11 · ema.europa.eu — EMA news, 5 November 2015: Article 20 referral outcome, background rates of CRPS and POTS, data examined, conclusion

• S12 · ema.europa.eu — EMA on European Ombudsman case 1475/2016/JAS, October 2017: no maladministration finding

• S13 · ansm.sante.fr — ANSM and Cnamts, September 2015: 2.2 million girls, 14 conditions, Guillain-Barré “probable”, 1–2 extra cases per 100 000, benefit-risk statement

• S14 · ansm.sante.fr — ANSM thematic page, updated 16 June 2023: listed adverse effects, unreplicated status of the Guillain-Barré finding, absence of autoimmune signal

• S15 · ema.europa.eu — Gardasil 9 Summary of Product Characteristics, section 4.8 and 5.1: safety database of 15 776, 84.8% and 13.2% figures, full frequency table, post-marketing table, adjuvant identification

• S16 · who.int — WHO Global Advisory Committee on Vaccine Safety: anaphylaxis 1.7 per million doses, syncope characterisation, doses distributed, 2013–2017 statements

• S17 · pubmed.ncbi.nlm.nih.gov — Arnheim-Dahlström et al, BMJ 2013: 997 585 girls in Denmark and Sweden, 53 outcomes, no association found

• S19 · nejm.org — Lei et al, New England Journal of Medicine, 2020: 1 672 983 Swedish women, 19 versus 538 cancers, incidence rate ratios by age at vaccination

• S21 · publichealthscotland.scot — Public Health Scotland, January 2024: no cases in women fully vaccinated at 12–13 since 2008, screening caution

• S22 · qmul.ac.uk — Queen Mary University of London release for Sasieni and Falcaro, The Lancet, 17 June 2026: zero deaths at ages 20–24 in 2020–2024, counterfactual estimate, deaths avoided, percentage falls, funder

• S23 · thelancet.com — Alcalde-Herraiz et al, The Lancet Primary Care, March 2026: target trial emulation across UK, Catalonia and Norway, 42% and 58% effectiveness, inability to estimate invasive cancer

• S24 · thelancet.com — Australian sentinel surveillance 2004–18: 49% and 89% reductions in genital warts in young heterosexual men

• S25 · mja.com.au — Yuill et al, Medical Journal of Australia 2026: elimination projected by 2035, 83.0% coverage in 2023, declining screening participation

• S26 · tandfonline.com — Resumption of active HPV recommendation in Japan, 2022: suspension June 2013 to April 2022, coverage collapse, catch-up programme

• S27 · pubmed.ncbi.nlm.nih.gov — Suzuki and Hosono, Nagoya study, 2018: 71 177 women surveyed, no significant increase in 24 reported symptoms, menstrual and headache findings

• S28 · pubmed.ncbi.nlm.nih.gov — Hansen et al, Vaccine 2020: Danish decline after the March 2015 documentary and subsequent recovery

• S29 · pubmed.ncbi.nlm.nih.gov — Simas et al, 2018: Carmen de Bolívar episode, over 600 reported cases, official mass psychogenic illness finding, uptake collapse to 14% and 5%

• S30 · icmr.gov.in — Indian government committee final report on the PATH HPV demonstration project

• S31 · nature.com — Nature India, 2013: parliamentary scrutiny of the suspended Indian trial, consent and reporting criticisms

• S32 · ncbi.nlm.nih.gov — IARC and US National Cancer Institute expert meeting, chapter 1: why invasive cancer was not used as an endpoint, endorsed surrogates, reliance on post-licensure monitoring

• S33 · fda.gov — FDA clinical review, Gardasil 9 (STN 125508/0): trial V503-001 design, 14 215 participants, CIN2+ primary endpoint, 42-month follow-up, immunobridging basis

• S34 · nature.com — KEN SHE trial three-year results, Nature Medicine 2023: single-dose efficacy against persistent HPV 16/18 infection

• S35 · thelancet.com — DoRIS long-term extension, Lancet Global Health 2024: five-year immunogenicity after one dose in girls aged 9–14

• S36 · who.int — WHO, 20 December 2022: single-dose schedule recommendation

• S37 · gov.uk — UK Health Security Agency, published 29 January 2026: year 8 and year 9 coverage by sex, regional range, comparison with pre-pandemic levels

• S38 · cdc.gov — CDC, National Immunization Survey-Teen 2024: 78% one-dose HPV coverage against 91% Tdap and 90% MenACWY

• S39 · rki.de — Robert Koch Institute, STIKO recommendation as at 6 June 2025: ages 9–14, two doses, catch-up to 17, coverage at 15, German HPV cancer burden

• S40 · krebsdaten.de — Centre for Cancer Registry Data, Robert Koch Institute: German cervical cancer cases and deaths, five-year survival

• S41 · szczepienia.pzh.gov.pl — Polish National Institute of Public Health, 2026 immunisation schedule: HPV listed as recommended, ages 9–14, free provision and reimbursement rules

• S42 · termedia.pl — Polish Ministry of Health reply to parliamentary question no. 13432, published 2 December 2025: 585 580 vaccinated in the 2010–2016 cohorts, 20.15%

• S43 · infor.pl — Draft amendment to the Polish regulation on compulsory vaccinations, published on the Government Legislation Centre site 20 October 2025: HPV obligation from 1 January 2027, ages 9–15, 2018 birth cohort first

• S44 · everycrsreport.com — Congressional Research Service IN12684, updated 29 July 2026: complete list of CDC vaccine recommendation changes in 2025 and 2026; no HPV change listed

• S45 · nejm.org — Giuliano et al, New England Journal of Medicine 2011: quadrivalent vaccine efficacy in men aged 16–26 against persistent infection and external genital lesions

• S46 · thelancet.com — Fourteen-year Nordic follow-up, eClinicalMedicine 2020: durability of protection and antibody response, no evidence of waning

• S47 · academic.oup.com — Journal of Infectious Diseases: assessment that genotype replacement remains unlikely but cannot yet be excluded

• S48 · hpvroundtable.org — National HPV Vaccination Roundtable: the five US jurisdictions with school-entry requirements and their exemption provisions

• S49 · onlinelibrary.wiley.com — de Martel et al, International Journal of Cancer 2017: worldwide burden of cancer attributable to HPV by site, country and type

• S50 · unicef.org — UNICEF Supply Division HPV vaccine price data, January 2026: per-dose prices by product and country income tier

• S51 · journals.sagepub.com — Tomljenovic and McHenry, International Journal of Risk and Safety in Medicine 35 (2024): Danish trial documents on comparator description and consent wording; authors write from a position critical of the vaccine

• S52 · pubmed.ncbi.nlm.nih.gov — Jørgensen, Gøtzsche and Jefferson, BMJ Evidence-Based Medicine, July 2018: criticism of the Cochrane review as incomplete and biased

• S53 · cochrane.org — Cochrane editor-in-chief response, 3 September 2018: criticisms judged overstated, conflict-of-interest correction, points carried into the next version

• S54 · pubmed.ncbi.nlm.nih.gov — Harper et al, Lancet Infectious Diseases 2021: “Elimination of cervical cancer depends on HPV vaccination and primary HPV screening”

• S55 · insurancejournal.com — Insurance Journal, 8 June 2026: Merck settlement of over USD 50 million covering more than 200 cases, no admission of liability, prior dismissal on expert evidence

• S57 · doi.org — Healthcare (2025), cross-sectional analysis of the Polish national HPV programme: 8.67% participation in the initial phase, regional disparities

• S60 · path.org — PATH, 4 October 2021: WHO prequalification of Cecolin (Xiamen Innovax), the first Chinese HPV vaccine to be prequalified, and what prequalification permits

• S59 · acpjournals.org — Andersson and Hviid et al, Annals of Internal Medicine, 15 July 2025: Danish cohort of 1 224 176 children, 50 chronic outcomes, aluminium-adsorbed vaccines

• S61 · sec.gov — Merck & Co. Forms 10-K for financial years 2009, 2010, 2013, 2016, 2019, 2022 and 2025: Gardasil/Gardasil 9 worldwide sales by year, the US and non-US split, total company sales, and the Sanofi Pasteur MSD joint-venture footnote

• S62 · cochranelibrary.com — Lundh, Lexchin, Mintzes, Schroll and Bero, Cochrane methodology review MR000033 (2017), abstract: 75 papers, risk ratios for favourable results and conclusions, the blinding finding, and the authors’ conclusion on industry bias

• S63 · pmc.ncbi.nlm.nih.gov — Funding sources of therapeutic and vaccine clinical trials, 2020–2021: 1 977 registered trials, industry and public shares overall and for vaccine trials specifically

• S64 · ema.europa.eu — European Medicines Agency pre-authorisation guidance and Directive 2001/83/EC: the applicant must submit its own pharmaceutical, pre-clinical and clinical test results; US sponsor accountability under FDA regulation

• S65 · dceg.cancer.gov — US National Cancer Institute: the ESCUDDO trial and the Costa Rica Vaccine Trial, their public funding, and where responsibility for design, conduct, data and publication sat

• S66 · aifa.gov.it — Italian Medicines Agency independent clinical research programme: the 5% levy on promotional expenditure, and 282 studies totalling about EUR 43.5 million between 2005 and 2018

• S67 · sec.gov — GSK Annual Report 2011 (SEC exhibit) and the FY2013 and FY2014 results releases: Cervarix turnover for 2009, 2010 and 2011, the Japan-driven peak, and the subsequent declines

• S68 · fiercepharma.com — Trade reporting of Merck’s February 2025 pause of Gardasil shipments to China, its extension to the end of 2025, and the withdrawal of the USD 11 billion 2030 target