Book summary “Bad pharma”: summary

Introduction

You trust your doctor to prescribe the best medicine based on solid science. When you swallow a pill, you assume rigorous tests proved it works better than older options. Yet pharmaceutical companies routinely hide negative trial results, leaving you vulnerable to biased evidence.

Ben Goldacre reveals how the $65 billion global drug industry systematically distorts research. You will see how companies bury unfavorable data, design rigged trials, and corrupt regulators to protect profits.

The book unmasks these hidden mechanisms across several key areas. You will discover how missing data distorts science, how new drugs are developed, and how regulators fail you. You will also learn how trials are manipulated and how simpler methods could fix the system.

How do drug companies hide unfavorable trial results from doctors and patients? Why do regulators approve medications on low evidentiary bars? What stops independent researchers from exposing flawed studies?

1. Missing Clinical Trial Data

Missing clinical trial data is far more dangerous than biased marketing because it completely blinds doctors and researchers to the true effects of treatments.

While marketing distortions and spin from drug reps can be bypassed by consulting independent systematic reviews of good-quality trials, withheld trials mean that nobody can know the true efficacy or harms of drugs. This leads to real-world patient suffering and death based on bad evidence.

Why missing data matters

Unpublished negative results create a heavily distorted medical literature that leads doctors to prescribe ineffective or harmful treatments. Drug companies often bury unfavorable trial results, leaving doctors to make prescription decisions based solely on positive published reports. When all trials are finally pooled together, treatments that appeared safe and effective in print often perform no better than sugar pills and carry worse side effects.

For the antidepressant reboxetine, published trials showed it was safe and effective. Unpublished trials involving significantly more patients revealed it was no better than a dummy sugar pill and caused worse side effects. This pattern involved 254 patients in the single published positive placebo trial for reboxetine against 1,657 patients worth of data left unpublished showing patients on reboxetine did worse.

How much data is missing?

Comparison of FDA regulatory files with published literature reveals that negative trials are systematically erased from public view. Investigators checking all trials submitted to the FDA for antidepressant marketing approvals found that positive studies were almost always published. Negative studies were either buried or misleadingly rewritten to make the drugs look successful.

A 2008 study of FDA files for twelve antidepressants found thirty-eight positive trials and thirty-seven negative trials. In the published academic literature this turned into forty-eight positive papers and only three negative ones. These numbers covered 74 studies in total representing 12,500 patients' worth of data during the 1987 to 2004 market entry period for the studied antidepressants.

Industry sponsors routinely use confidential contracts and gagging clauses to control data and forbid researchers from publishing unflattering results. Academic institutions and ethics committees frequently permit contracts that grant pharmaceutical companies the right to monitor trials, stop them early, or completely block the publication of data that threatens commercial interests.

A study of clinical trials approved in Copenhagen found that 98 per cent were industry-sponsored. Nearly all featured restrictions on publication rights, and half gave the sponsor outright ownership of the data or final approval over publication. This review examined 44 trials in total, alongside cases such as the £145,000 offered in a gagging settlement to whistleblower Aubrey Blumsohn.

Drug regulators withhold comprehensive trial data from the public and maintain chaotic, unsearchable archives that protect corporate interests. European and national regulators possess detailed Clinical Study Reports submitted during drug approvals but routinely refuse to share them with independent scientists, citing commercial confidentiality or patient privacy.

Nordic Cochrane Centre researchers spent three and a half years fighting the European Medicines Agency in a battle overseen by the European Ombudsman to obtain study reports for weight-loss drugs like orlistat and rimonabant, enduring a conflict lasting from June 2007 to November 2010. Similar review documents can reach extreme lengths, such as a 400 pages long document from the FDA on pregabalin.

What can be done?

Solving the crisis of missing trial data requires strict legislation, mandatory public registries, universal clinical study report disclosures, and severe penalties. To protect patients from avoidable harm, regulatory authorities must enforce a one-year publication deadline for all human trials, outlaw gagging clauses, make all clinical study reports publicly available, and mandate patient-level data sharing for independent researchers.

• In legislation, gagging clauses should be made illegal, with no possibility of quibbles.
• All professional bodies should take a clear stand on unpublished trial data, declare it clearly as research misconduct, and state that it will be handled like any other form of serious misconduct that misleads doctors and harms patients.
• All research staff involved in any trial on humans should be regarded as jointly and severally liable for ensuring that it is published in full within one year of completion.
• No person should be allowed to conduct trials in humans if a research project they are responsible for is currently withholding trial data from publication more than one year after completion.
• No trial should be approved without a firm guarantee of publication within one year of completion.

2. Where New Drugs Come From

New drugs are primarily developed by copying existing molecular mechanisms or through high-throughput screening of large molecule libraries in laboratories. Researchers look for molecules that target specific biological mechanisms, such as enzymes, to alter bodily functions. Screening involves synthesizing hundreds or thousands of molecules and testing them in lab dishes to measure desired changes, but much of the resulting data is locked in company vaults or discarded. There are 20,000 genes in the human body, meaning building blocks are reused and drugs often have unintended toxic effects at higher doses.

Candidate molecules undergo screening and destructive animal testing before moving to human trials that increasingly rely on globalized outsourcing and vulnerable populations.

Before human trials, candidate molecules undergo rigorous and destructive animal testing to determine basic toxicity, maximum tolerable doses, and breakdown products. Animals, typically rodents and dogs, are dosed to evaluate blood concentrations, breakdown products, and serious toxicities like cancer or organ damage. Animals are frequently given increasing doses until death or severe toxicity occurs to find lethal thresholds, establishing baseline safety before human administration. Practolol, a beta-blocker molecule structurally similar to propranolol, caused severe multi-system oculomucocutaneous syndrome unexpectedly in humans despite animal tests.

Early trials

Initial human safety trials are often filled by unemployed people, students, and marginalized individuals who participate for easy money rather than altruism. Due to back-loaded payments and few economic alternatives, participants are disincentivized from reporting side effects or withdrawing from unpleasant and painful studies. Only 20 per cent of drugs entering phase 1 trials go on to be approved and marketed. Typical payments range from $200 to $400 a day for phase 1 trial participants.

Eli Lilly recruited homeless alcoholics from a local shelter for trials, and the publication Guinea Pig Zero documented participant experiences in phase 1 units.

CROs and trials around the world

Pharmaceutical companies increasingly use Clinical Research Organisations, known as CROs, to conduct trials globally to benefit from dramatically lower costs, poorer regulatory oversight, and vulnerable populations. This commercial shift prioritizes marketing and favorable results over neutral scientific inquiry, risking data integrity and population relevance. In 2010, CROs reached $20 billion in global revenue, representing about a third of pharma R&D spending. By 2008, these organisations ran 9,000 trials with over two million participants across 115 countries.

GlaxoSmithKline aimed to move half its trials to low-cost countries, noting that a trial costing $30,000 per patient in the US can be done in Romania for $3,000.

Conducting trials in developing nations introduces issues such as language barriers, lack of independent ethics reviews, high financial inducements relative to local wages, and questionable applicability of results to Western populations. Pfizer conducted an experimental Trovan antibiotic study during a meningitis epidemic in Kano, Nigeria, in 1996 without proper informed consent, while effective treatment was available nearby, later leading to a lawsuit and leaked diplomatic cables revealing investigator pressure. 11 children died during the Trovan trial in Nigeria, resulting in a $75 million out-of-court settlement paid by Pfizer in 2009.

3. Bad Regulators

Drug regulators suffer from regulatory capture, approving medications on low evidentiary bars and keeping safety monitoring opaque.

You might assume government agencies thoroughly test every medicine before it reaches the market. In reality, drug regulators socialize with industry lobbyists and share a common technical background. A revolving door allows former public regulators to take lucrative jobs advising drug companies. Thomas Lonngren stepped down as executive director of the EMA and gave four days' notice before starting work as a private consultant to the pharmaceutical industry. Corporate Europe Observatory documented 15 similar cases of senior EU officials passing through this revolving door between government and industry.

Approving a drug

Regulators routinely approve new drugs based on the low bar of showing they are better than a placebo rather than testing against existing effective treatments. A 2011 paper reviewed 197 new drugs approved by the FDA from 2000 to 2010 and found a full third had no evidence comparing them with the best currently available treatment. Only 70 percent had data showing they were better than other treatments. Meanwhile, 50 percent of drugs approved by the EMA between 1999 and 2005 were studied in comparison with other treatments.

Surrogate outcomes

Drugs are frequently approved based on surrogate outcomes, such as blood test results, which may not translate into real-world benefits like reduced mortality. Commercial pressures and short-term incentives discourage long-term follow-up studies, and history shows that improving surrogate outcomes can sometimes paradoxically increase real-world harm. The Cardiac Arrhythmia Suppression Trial tested anti-arrhythmic drugs that successfully prevented abnormal heart rhythms, but a proper trial measuring death had to be stopped early because the drugs drastically increased the risk of death.

Monitoring side effects

Post-market safety monitoring is severely compromised by underreporting of side effects, industry obstruction, and a culture of regulatory secrecy. Pre-approval trials are too small to catch rare adverse events, relying instead on spontaneous reporting systems and database studies. A study size follows the rule of three, where 500 patients studied can only spot side effects occurring more frequently than 1 in 166, and 3,000 patients spot side effects affecting more than 1 in 1,000.

Regulators keep safety planning documents secret and are structurally reluctant to withdraw approved drugs because it would admit past failures. Pfizer took nearly two years to update statin labels with warnings for depression, sleep disturbances, and memory loss because a marketing authorization holder disagreed with the wording. This delay affected 4 million people prescribed statins in the UK, while the estimated rate of adverse events actually fed back to the MHRA in Britain was only 1 in 20.

Fixing this broken system requires concrete operational changes:

• Require drug companies to provide data showing how their new drug compares against the best currently available treatment before market entry.
• Use regulatory and healthcare funder influence to force companies to produce more informative trials measuring real-world outcomes.
• Put all information about safety, efficacy, and adverse events into the public domain to allow independent academic analysis.
• Create a better market for communicating clear, quantitative risks and benefits through tools like drug facts boxes.

4. Bad Trials

Trials are routinely distorted through ideal patient selections, inappropriate comparators, truncated stopping rules, and switched primary outcomes.

You might assume published clinical trials give you a clear picture of how treatments work. In reality, researchers and companies manipulate trial design and analysis in systematic ways.

Outright fraud involves complete data fabrication, such as inventing patients and studies. Anaesthetist Dr Scott Reuben published at least twenty pain trials that were entirely fabricated without conducting studies on patients. Similarly, British obstetric surgeon Malcolm Pearce fabricated a trial of two hundred women and invented a fictitious pharmaceutical company. A 2009 systematic review combining survey data from twenty-one studies on scientific malpractice showed 2 per cent of researchers admitted to fabricating or falsifying data. In the same review, 14 per cent of researchers suspected colleagues of fabricating data.

Design flaws

Trials often test drugs on freakishly ideal patients who bear little resemblance to real-world populations. A 2011 Finnish study on hip fracture drugs found that out of 7,411 patients, exclusions for being male, of the wrong age, or lacking osteoporosis meant trial data only strictly applied to about one in seven patients. For asthma, only 6 per cent of general population patients were eligible for the trials forming international consensus guidelines. This lack of external validity exaggerated cost-effectiveness estimates, turning a $20,000 cost per avoided gastrointestinal bleed estimated by NICE into $100,000 when calculated using GP records for real-world patients.

Seeding trials are covert viral marketing campaigns disguised as research projects to recruit physicians and alter prescribing habits. Internal documents for Merck's ADVANTAGE trial on Vioxx and Pfizer's STEPS trial on Neurontin proved they were designed as marketing devices executed by sales representatives. The ADVANTAGE trial targeted 5,000 patients using six hundred doctors. The STEPS trial on Neurontin involved 2,759 patients and recorded 73 serious adverse events and 11 deaths.

Statistical manipulation

Unflattering data is routinely massaged through selective reporting, composite endpoints, and strategic risk presentation.

Researchers frequently switch their pre-specified primary outcomes after seeing trial results, turning chance findings into discoveries. In trial 329 of paroxetine in children, the original protocol specified two primary and six secondary outcomes with no difference between drug and placebo. Published papers highlighted four positive results out of twenty-seven measured outcomes as if they were the main results.

Trials also bundle multiple endpoints into composite outcomes or use per-protocol analyses that exclude dropouts. The UKPDS diabetes trial reported a 12 per cent reduction in a massive composite outcome, but the improvement was mostly driven by a reduction in laser treatments for eyes rather than changes in death or blindness rates. Stopping trials early based on interim results exploits random statistical variation and overstates treatment benefits by about a quarter. Media coverage then exaggerates these findings using relative risk reductions, such as reports on the JUPITER trial highlighting a 54 per cent reduction in heart attacks while ignoring that the placebo rate of 0.37 events per one hundred person years only fell to 0.17 on rosuvastatin.

You can avoid these pitfalls by demanding rigorous methodological safeguards:

• Include all patients prescribed a treatment in the analysis regardless of whether they stopped taking it.
• Perform an intention-to-treat analysis rather than a per-protocol analysis.
• Set up stopping rules specified before the trial begins.
• Ensure stopping rules are extreme enough to be unlikely to be triggered by expected chance variation over time.

5. Bigger, Simpler Trials

Medicine suffers from a staggering scale of ignorance regarding which widely used treatments are best. This failure leads to unnecessary suffering and death. Many important diseases lack clear comparative data on standard treatments because research is driven by commercial expense.

Embedding large, simple pragmatic trials directly into everyday medical practice can eliminate our staggering scale of ignorance at minimal cost.

A past BMJ paper noted that strokes are so common that if every stroke patient in the world were entered into a trial comparing best treatments, enough patients would be recruited in twenty-four hours to answer the question.

Designing simple trials

Large, simple, pragmatic trials can be seamlessly embedded into everyday GP surgeries using existing electronic health records to compare standard treatments at almost zero cost. Instead of separate expensive administrative studies, routine software can prompt a doctor to randomise a patient between two standard options, automatically gathering long-term outcome data from computerised notes without requiring extra appointments or paperwork.

A pilot study funded at £500,000 was set up in the UK to compare two widely used statins, atorvastatin and simvastatin, by adding a single prompt button into GP prescription software. This design detected a 2 per cent difference in effectiveness that could save vast numbers of lives globally. You can participate in this research framework through simple clinical interactions:

• GP opens prescription screen for a patient needing a statin.
• An extra screen appears stating genuine uncertainty between the two drugs.
• Doctor presses a big red button to randomly assign the patient to one of the two treatments.
• Patient outcomes and medical notes are tracked automatically through existing electronic health records without extra forms.

Regulatory barriers

A bizarre regulatory paradox allows doctors to choose treatments arbitrarily on a whim with zero oversight, yet imposes mountainous red tape on doctors trying to randomly assign patients between those same treatments. While everyday arbitrary prescription requires no special justification, attempting to generate valid knowledge via randomisation subjects researchers to disproportionate regulatory scrutiny and ethics committees that stall research and harm patients.

In the CRASH trial for head injuries, ethics committees delayed randomising unconscious patients between two widely used treatments. During that time, a steroid treatment later proven to kill patients continued to be used unchecked, resulting in two and a half extra deaths for every hundred people treated with steroids. Local trial centres introduced an average delay of 1.2 hours due to family consent requirements.

Lengthy twenty-minute consent processes for comparing standard treatments do not improve patient understanding and actively threaten the validity of pragmatic trials. Evidence shows people remember more information from short forms than long ones. Imposing twenty-minute hurdles makes trials expensive and skews participation toward unrepresentative doctors and patients willing to stop routine care.

You can improve participation by adopting streamlined methods:

• Offer participation in trials of standard treatments based on a brief consent form of no more than a hundred words.
• Provide an option to access more detailed explanatory material for anyone who wants it.

Conclusion

Modern medical evidence is systematically distorted across the entire lifecycle of drug development and regulation.

You face a medical system where commercial interests routinely corrupt the information base of clinical practice. The central insight is that bad pharma is not a collection of isolated bad apples, but a systemic failure driven by hidden trial data, biased research designs, captured regulators, and corrupted academic publishing. When you look across the evidence, you see that half of all clinical trials vanish, regulators act as industry service agencies, and doctors rely on marketing masquerading as science.

To fix this broken ecosystem, you can demand transparency and integrity through specific practical actions:

1. Insist that all past and present clinical trials publish their full protocols and raw data openly.
2. Support independent researchers who conduct pragmatic trials comparing existing treatments without commercial funding.
3. Prescribe and consume generic drugs whenever possible to bypass the financial incentives of new brand-name medications.
4. Read medical journals with intense skepticism, recognizing that published papers often exaggerate benefits and conceal harms.
5. Push policymakers to completely separate drug regulation and funding from pharmaceutical industry fees.

10 Key Ideas

1. Missing clinical trial data blinds doctors and researchers to the true efficacy and harms of medical treatments, leading to real-world patient suffering and death based on bad evidence.
2. Pharmaceutical companies routinely bury unfavorable trial results, leaving doctors to make prescription decisions based solely on positive published reports.
3. Candidate molecules undergo screening and destructive animal testing before moving to human trials that increasingly rely on globalized outsourcing and vulnerable populations.
4. Drug regulators suffer from regulatory capture, approving medications on low evidentiary bars and keeping safety monitoring opaque.
5. Trials are routinely distorted through ideal patient selections, inappropriate comparators, truncated stopping rules, and switched primary outcomes.
6. Unflattering data is routinely massaged through selective reporting, composite endpoints, and strategic risk presentation.
7. Embedding large, simple pragmatic trials directly into everyday medical practice can eliminate our staggering scale of ignorance at minimal cost.
8. Modern medical evidence is systematically distorted across the entire lifecycle of drug development and regulation.
9. You face a medical system where commercial interests routinely corrupt the information base of clinical practice.
10. To fix this broken ecosystem, you can demand transparency and integrity through specific practical actions.