Bad Pharma

Why do so many medical decisions rest on distorted or incomplete information? In his deep investigation of the modern pharmaceutical evidence system, Ben Goldacre argues that medicine has been systematically misled — not by a few rogue actors, but by ordinary institutional incentives that favour positive results, obscure negative findings, and let secrecy thrive. The book reveals that clinical evidence, instead of being a transparent record of discovery, is often a curated story where industries, regulators, journals, and even universities participate in selective storytelling. Understanding this requires seeing how multiple failures interact: selective publication, bad trial design, institutional complicity, weak regulatory enforcement, and the pervasive influence of marketing.

The Evidence Distortion Cycle

The first distortion begins with industry-funded research. Systematic reviews consistently show that company-sponsored trials are more likely to report favourable results — not because drugs magically work better under corporate sponsorship, but because companies control what is studied, how it’s analysed, and what gets published. Selective publication, biased comparators, and post-hoc analytic flexibility create an illusion of consistent success. Individual doctors reading the literature rarely see the dozens of invisible negative trials that never reached print.

The next distortion arrives through institutional silence. Universities, journals, and ethics committees — institutions designed to uphold integrity — often sign contracts giving sponsors power to block or delay publication. When researchers fight for access to data, they can face retaliation. Patients who consented under the promise of helping science are misled when sponsors retain control of results. Each layer protects commercial confidentiality over collective safety.

What Missing Data Costs Real People

The harm of hidden trials is not abstract. Goldacre recounts the story of reboxetine, an antidepressant once prescribed in good faith by doctors. Published data made it seem effective; the full dataset later showed that the majority of trials were negative and unpublished. Patients had been exposed to a drug that offered no benefit and more harm. Similar tragedies — from the TGN1412 immune disaster in healthy volunteers to the lorcainide anti-arrhythmic study that presaged 100,000 excess deaths from successor drugs — show that every missing dataset is a potential catastrophe delayed.

At the population level, the cumulative cost is staggering: wasted billions on marginal or ineffective therapies, distorted guidelines, and a betrayal of the social contract between patients and medicine. These are not peripheral problems of academic integrity — they are central to how health systems allocate trust, money, and care.

Regulatory and Cultural Complicity

Regulators, whose mission is to guard the public, often become inadvertent gatekeepers of secrecy. Agencies such as the EMA and MHRA hold vast troves of Clinical Study Reports (CSRs) — the comprehensive trial dossiers that show everything from protocols to adverse event logs — yet routinely block access under the banner of commercial confidentiality. In cases like paroxetine in children or Tamiflu for influenza, regulators and manufacturers long withheld critical data while governments spent hundreds of millions based on incomplete evidence.

Attempts to fix transparency — trial registries, publication rules, and registration mandates — often became what Goldacre calls “fake fixes.” Without vigilant enforcement and retrospective coverage of old trials, registries like clinicaltrials.gov or EudraCT create an appearance of openness but leave most missing data unrecovered. The right laws without the will to apply them become theatre.

The Deep Structure of Bias

Even published trials can mislead. Design choices — surrogate endpoints, idealised patients, weak comparators — and analytic tricks like outcome switching or subgroup dredging can turn a neutral result into “proof” of efficacy. Marketing-driven “seeding trials” blur the line between research and salesmanship, while ghostwriting, reprint sales, and conflicted continuing medical education (CME) shape doctors’ perceptions long after the data leave the lab.

Underlying all of it is a conflict of interest structure that normalises bias while denying its influence. Financial ties, professional incentives, and institutional dependence on industry funding make distortion predictable. Transparency is therefore not an optional virtue; it’s the minimum condition for scientific trust.

Rebuilding Trust

Goldacre’s solution is neither cynicism nor simple regulation. It’s a systematic cleanup: full public disclosure of all past and present trials, open access to CSRs, independent auditing of trial registries, named payments from companies to doctors, and embedded pragmatic trials that use routine health data to test real-world questions. The message is blunt but hopeful — evidence-based medicine can only fulfil its promise when the evidence itself is open to scrutiny.

Central Idea

Medicine fails when data are distorted or missing. The cure is radical transparency — not just for future trials, but for the millions of patients whose care still depends on hidden evidence from the past.

This is a call to civic engagement as much as to scientific integrity: if you care about safer drugs, fairer decisions, and honest science, you must demand that every trial — past, present, and future — be open to view.

When companies or regulators conceal trial results, real patients suffer. The most tragic cases show how suppression moves harm from paper to bedside. Reboxetine’s negative trials, buried for years, led thousands to use an ineffective antidepressant. The lorcainide study—kept secret for decades—let doctors continue prescribing similar anti-arrhythmics that likely killed over 100,000 Americans. And the TGN1412 catastrophe illustrated how a single unpublished human immunology signal could have prevented volunteers from life-threatening reactions. Missing data aren’t statistical quirks; they are lives miscounted and lessons unlearned.

The Systemic Effects of Secrecy

Withholding results distorts every level of healthcare. Doctors make prescriptions based on an incomplete literature. Regulators evaluate safety without visibility of older trials. Health services waste funds on inflated claims of benefit. Patients, who consent believing they're contributing to science, see their participation turned into invisible commercial advantage. Science relies on the sharing of all outcomes, especially the negative, because failures narrow uncertainty and prevent repetition of error.

This secrecy persists because professional bodies rarely treat non-publication as misconduct. Institutions that should enforce transparency — ethics committees, journals, Royal Colleges — often remain silent. Fixing this means declaring data withholding not just poor practice, but unethical behaviour against patients’ contributed trust.

Key Lesson

Incomplete evidence harms real people. Every unpublished dataset is an ethical breach against the patients who participated and those who later depend on their results.

Not all published trials are trustworthy — many are subtly engineered to flatter a sponsor’s product. Goldacre dissects these design tricks, showing how bias can be built before data collection even starts. Selecting unrepresentative patients makes results look cleaner than reality; testing a new drug against a weak comparator makes it shine; relying on surrogate laboratory markers (like cholesterol or tumour size) bypasses the harder proof of real patient benefit. These methods rarely amount to overt fraud but operate through statistically respectable choices driven by commercial incentives.

Outcome Switching and Data Dredging

Across medicine, authors change their primary outcomes after seeing results. In the gabapentin and paroxetine trials, half of the pre-specified endpoints vanished, replaced by new “positive” ones introduced later. Because statistical tests assume one primary endpoint, measuring many and reporting only the most favourable almost guarantees false discovery. Post-hoc subgroup analysis compounds the illusion — finding spurious “benefits” for arbitrarily sliced populations, such as specific ages or regions. Unless a subgroup was pre-declared or biologically plausible, these effects are mirages.

Marketing Disguised as Science

Some trials are not even designed to inform medicine at all. “Seeding trials” like Merck’s ADVANTAGE (for Vioxx) and the Neurontin STEPS project were marketing campaigns masquerading as research, involving hundreds of clinicians each enrolling only a handful of patients. The aim: to familiarise prescribers with a drug and encourage ongoing prescribing. Data quality, consent, and patient safety became secondary. Such pseudo-research corrupts the scientific literature and deceives both doctors and patients about legitimacy.

Practical Message

If a result looks too good to be true, check its comparator, endpoints, and registration history. The architecture of bias is built into trial design long before the paper is published.

Regulatory agencies and trial registries should safeguard openness, yet they are often complicit in concealment. Clinicaltrials.gov, created in 2000, and the 2007 FDA Amendments Act promised full registration and result reporting — but audits showed fewer than 20% of eligible studies complied. In Europe, EudraCT remained partly secret until 2012 despite claiming to be public. These are “fake fixes”: institutional gestures that generate the illusion of reform without meaningful enforcement.

Regulatory Obstruction

Regulators hold enormous archives of clinical study reports (CSRs), yet they often deny requests for public release. The paroxetine-in-children case revealed that UK law required submission only for licensed indications, letting GlaxoSmithKline withhold negative trials legally but unethically. When the EMA refused to share CSRs on orlistat and rimonabant, it took a three-year battle and an Ombudsman ruling to force disclosure. Some regulatory bodies have even destroyed old CSRs under “retention policies,” erasing decades of evaluative evidence.

Why Full CSRs Matter

The Tamiflu controversy exemplifies why partial transparency isn’t enough. Governments Stockpiled the drug during flu scares based on a meta-analysis that relied mostly on unpublished data from Roche. When Cochrane reviewers demanded the full CSRs, the company offered only redacted summaries with confidentiality conditions. Their persistence revealed hidden adverse events and inconsistencies, undermining global health spending decisions. Without CSRs, you cannot audit methods, check adverse events, or replicate analyses — the foundation of science.

Core Principle

Transparency demands enforcement, retrospective coverage, and independent auditing. Registries and regulators that hide data serve bureaucracy, not patients.

Pharmaceutical marketing operates under the guise of professional education. Companies spend more on promotion than on research and development, reaching doctors through multiple coordinated channels: drug representatives, sponsored continuing education, key-opinion-leader events, journal reprints, and ghostwritten literature. Each pathway influences prescribing behaviour, often imperceptibly.

Marketing Through Information

Drug reps are the friendly face of influence. Evidence shows doctors who meet reps prescribe the featured product more and at higher cost, while believing they remain objective. Sponsored CME and journal reprints buttress the same effect: teaching sessions that cite the sponsor’s drug more often and journals dependent on advertising revenue hesitant to criticise major clients. Ghostwritten research — like the paroxetine and olanzapine cases exposed in court — turns marketing scripts into apparent academic authority.

The Numerical Tricks of Spin

Even honest numbers can deceive through presentation. Reporting “relative risk reduction” (a 50% improvement) rather than “absolute risk reduction” (2 percentage points) transforms modest results into headline miracles. Trials with negative outcomes can be spun through selective framing or reclassifying success. Systematic reviews show that abstracts frequently omit main negative findings, overemphasising secondary positives. Media headlines amplify the distortion, shaping public expectation and policy.

Conflict of Interest as a Structural Disease

Financial links predict opinion. Studies on drugs like rosiglitazone show that authors with company ties are up to six times more likely to publish positive conclusions. Because guideline panels and patient groups also receive industry money, bias extends from research to policy. Goldacre supports complete transparency — “Sunshine”-style databases of named payments, itemised and public — alongside disclosure of ghostwriting, contractual gag clauses, and full authorship attributions. Only then can readers weigh evidence without blind trust.

Key Takeaway

In medicine, influence is most effective when invisible. Transparency about payments, authorship, and methodology is the only antidote to manufactured consensus.

Goldacre ends on constructive ground: cleaning up bad evidence requires new trial architecture and institutional reform. One practical model is the “pragmatic embedded trial” — using existing electronic health record systems to randomise patients into comparisons of standard treatments (for example, prescribing one statin versus another). Outcomes like heart attacks or deaths are tracked automatically through health records, drastically lowering cost and bias while maintaining real-world relevance.

Advantages of Embedded Trials

These trials study genuine patients, not filtered volunteers, provide large sample sizes, and follow outcomes for years using routine data. They restore medicine’s link to everyday practice rather than isolated laboratory precision. The pilot comparing simvastatin and atorvastatin demonstrated feasibility for about half a million pounds — less than the cost of a single large commercial trial.

Yet regulations designed for high-risk experimental drugs often suffocate such low-risk research with excessive paperwork and narrow consent requirements. Goldacre argues for proportional oversight: brief, clear consent for low-risk randomisation and a public ethic where participating in learning health systems is part of good citizenship.

Beyond Trials: A Transparency Charter

Repairing the evidence base means retrospective accountability: full release of all withheld trials; permanent public access to CSRs; institutional sanctions for non-reporting; and publication of enforcement audits so the public can see whether laws are followed. The remedy is systemic — cultural, legal, and technical. Only when transparency is assumed instead of exceptional can evidence-based medicine regain its credibility.

Final Reflection

We already have the tools to fix medicine’s evidence failures: open data, digital health records, independent analysis, and honest incentives. What remains is moral will — the decision to prioritise truth over comfort.