For 12 years, Roy A. Teel Jr. was involved in human research trials at the University of California, Los Angeles, but he didn’t don a lab coat.

As a research subject with progressive multiple sclerosis, he was involved in five medical trials.

When his condition didn’t improve, he retired on disability in 2011. The whole ordeal left a sour taste in his mouth.

“As long as there is a profit to be made, it is very, very unlikely that you are going to see any significant changes in how trials work,” Teel, now an author, told Healthline. “For those patients who are injured or lose their lives in clinical trials they are nothing more than collateral damage or the cost of doing business. There is no direct harm to the researcher, institution, or the drug company.”

The old analogy of breaking a few eggs to make an omelet seems fitting.

In searching for new and potentially life-saving drugs, mistakes and adverse events —including deaths — are seemingly inevitable.

The U.S. Food and Drug Administration (FDA) is the gatekeeper in ensuring the science is sound and the drugs brought to market are relatively safe and used appropriately.

Getting there is an often long, expensive, and complicated journey for drug makers.

According to the Tufts Center for the Study of Drug Development, the cost of bringing a new prescription drug to market comes in at an estimated $2.6 billion.

Almost half of that estimate accounts for revenues a drug company doesn’t receive while a new drug is in development. Still, almost $1.4 billion is listed as “out of pocket costs,” including $312 million for research and development.

“Companies learn a lot by failure,” Peter Pitts, president of the Center for Medicine in the Public Interest and former FDA associate commissioner for external relations, told Healthline. “That’s how science works.”

Read More: Why Some Drugs Cost So Much and Others Don’t »

How One Detail Can Skew Results

The latest study under fire involves rivaroxaban (Xarelto), the blood-thinning medication manufactured by Bayer, marketed by Janssen, a Johnson & Johnson company, and endorsed by celebrities like Arnold Palmer and Kevin Nealon.

In an investigation published in the British Medical Journal, researchers pointed out faults in the single trial used to gain approval in the United States and Europe.

The trial — ROCKET-AF — pitted Xarelto against warfarin, the most widely prescribed oral blood thinner drug in North America. The trial was overseen by the Duke Clinical Research Institute (DCRI).

Before ROCKET-AF began in 2002, a device that measured international normalized ratio (INR), important numbers when testing blood thinners, were found to be delivering lower results.

The FDA later recalled the device. A warning letter issued in 2005 detailed erroneous high and low results, which can lead to death or serious injury such as major bleeding. Too much of the drug can cause uncontrolled bleeding while too little can lead to a stroke.

The device’s use wasn’t published in phase III trials, so the potentially faulty numbers went unnoticed for years. The FDA did, however, warn Xarelto manufacturers about false or misleading ads for downplaying the risks of the drug. It also rejected wider use of the drug in early 2014 when the medicine was already bringing in $1.5 billion a year.

In September, Bayer and Johnson & Johnson announced findings from a 45,000-patient trial — including data from the ROCKET-AF study — that rates of major bleeding in patients with atrial fibrillation remained low.

“Bayer is committed to supporting physicians and patients in the safe and responsible use of Xarelto,” Dr. Michael Devoy, chief medical officer of Bayer HealthCare, said in a press release.

Read More: Emergency Rooms Facing Shortages of Important Drugs »

Study Sponsors Influence Outcomes

While doctors and scientists are calling for an independent investigation into the ROCKET-AF trial, a review of the data released in October found Xarelto was similar in safety and efficacy to warfarin.

Of the 16 researchers listed on the study, all but three received consulting fees from, were on advisory boards for, or were employees of Bayer, Janssen, or Johnson & Johnson, according to the disclosure statements.

Overall, nine companies have spent $25 million on 88,548 doctors related to Xarelto, according to ProPublica’s “Dollars for Docs” project.

While these potential conflicts of interest may seem alarming to some people, for those familiar with pharmaceutical research, it’s just another day at the office.

According to a recent review of meta-analyses of antidepressant drugs, when a drug company employee authors research on one of their drugs, it’s 22 times less likely to contain negative statements. Nearly a third of 185 analyzed studies had authors that were drug company employees and 79 percent had some conflict of interest, the review stated.

Read More: Drugs Used in Jimmy Carter’s Treatment Among a New Generation of Immune Therapies »

How Drug Trials Work

Drug companies front the money to develop a drug and bring it to market; it’s their research and they hold the keys to the data.

But, in order to receive approval by the FDA, they have to follow certain steps in the clinical process.

First, a potential drug is tested on animals — typically rodents, dogs, and primates — to determine toxicity.

Ira S. Pastor, CEO of Bioquark Inc., said it’s widely acknowledged animal models remain poorly predictive for humans, yet remain a mandatory cornerstone behind years and millions of dollars of early drug development activities.

“Penicillin kills guinea pigs and produces birth defects in rats. Aspirin is poisonous to cats. Cancer has been ‘cured in mice’ thousands of times and dozens of drugs found safe in animals are later withdrawn from market due to adverse drug events in humans,” Pastor told Healthline.

From there, a drug moves to a Phase I study — or first-in-human trials — which often has a small sample size of healthy adults.

Here, again, the focus is to determine toxicology. Should it be deemed safe to a certain extent then a Phase II trial, with a few dozen to a few hundred of potential patients to determine the drugs’ efficacy at treating a certain disease or condition, is launched.

On Feb. 29, AstraZeneca announced that a Phase IIb trial for tremelimumab to treat mesothelioma, a rare and deadly cancer of the lining of the lungs or abdomen, didn’t “meet its primary endpoint of overall survival,” meaning it will not proceed to a Phase II trial or be considered by the FDA.

The “gold standard” of testing is the randomized, double-blind trials, where neither patient nor researcher knows if a drug or a placebo is being administered. These are typically done in Phase III trials involving several hundred to 3,000 people.

Should a drug be shown to be safe and effective, it’s sent to the FDA for approval. The FDA determines whom the drug can be given to and what the drug can be used to treat.

Some drugs can receive “accelerated approval” where they hit the market after favorable Phase II results with the caveat of doing a follow-up study while the drug is being used in real-life patients. Drugs that make their way through this process are most often for life-saving medications such as cancer and public health emergencies like HIV/AIDS.

Since 2007, drug companies have registered their trials with ClinicalTrials.gov, operated by the U.S. National Institutes of Health. As of March 1, there were 209,563 studies registered, 105,573 of which were for drug or biologic therapies.

Read More: Would the Trans-Pacific Pact Boost Drug Prices? »

When Trials Go Wrong

In 1993, 15 people participated in a trial for the experimental hepatitis B treatment fialuridine. Five of them died and two others needed life-saving liver transplants.

In 2006, six volunteers were given an antibody — TGN1412 — 500 times lower than what had been deemed safe in animal studies. After their first dose, all were hospitalized with multiple organ failure.

In January, a French trial of 128 healthy volunteers had to be stopped. Of the 90 who received increased doses of the drug, six participants fell ill and one died. Since it’s common for pharmaceutical companies to not release the makeup of the molecules they’re testing, outsiders have been left in the dark about the substance used in the trial.

These alarming stories are indeed rare and regulatory agencies like the FDA are working to revamp the drug trial process in the wake of these adverse events, including the need for better mechanisms to safely test first-in-human studies.

In his book, Bad Pharma, Dr. Ben Goldacre examined the problems that arise in modern trials, including poor designs, faulty analysis of the data, exaggerated benefits, and downplayed harmful side effects.

While outright fraud is rare, Goldacre wrote, more trials are impacted by recruiting too few patients, stopping the trial early or late, testing drugs against something that doesn’t work, testing against relatively meaningless outcomes, ignoring patients who drop out, switching focus of the trial midway through, and spinning the results in a favorable light.

The results of these studies when bolstered, such as the case of a trial about intensive blood sugar control for diabetics, can permeate medical knowledge.

“There is a terrifying reality revealed by this study: rumours, oversimplifications, and wishful thinking can spread through the academic literature, just as easily as they do through any Internet discussion forum,” Goldacre wrote.

Another common occurrence is that when a trial doesn’t find its intended results, it often doesn’t see the light of day.

According to data from ClinicalTrials.gov, of the more than 1.2 million trials registered from 2009 to today, 90,381 — or less than 8 percent — have posted their results.

Brad Thompson, PhD, chief executive officer of Oncolytics Biotech, a startup focusing on oncology with five ongoing phase II studies, said when pharmaceutical companies pay for research the positive results are what become published in academic journals.

Journals themselves aren’t interested in studies with negative results, so getting all pertinent data available is a problem.

“No one wants to promote a negative study,” he told Healthline, “but there are checks and balances in place.”

Pitts says the FDA has an “extremely robust process” and if a company was found falsifying data it “would be out of business.”

“Could a company hide data? They could, but they’d be caught,” he told Healthline.

Read More: ‘Right to Try’ Movement Wants Terminally Ill to Get Experimental Drugs »

Fixing Faulty Data Takes Time

If a drug hits the market on faulty data, it’s difficult to get it revoked as regulators can’t act unless there are safety concerns.

When this does occur, the FDA issues “black box warnings” that alert physicians to an increased risk of adverse effects.

On Monday, the FDA announced it will require a “black box” warning for Bayer’s permanent birth control device Essure as well as directing Bayer to study heightened risks for woman, including unplanned pregnancies, pain, and other complications. One review linked 303 fetal deaths to Essure, Reuters reported.

The initial studies found 97 percent of women could rely on the device, but only 25 percent of the 926 women enrolled in the program were studied for effectiveness two years after the device was implanted. Of those 926 women, 181 didn’t even undergo the procedure, according to an article in the New England Journal of Medicine.

Bayer was supposed to do two five-year follow-up studies, but neither was registered, one remains unpublished, and the other wasn’t widely circulated.

“The problems of inadequately rigorous premarketing and postmarketing studies, unregistered clinical trials, and incomplete and delayed dissemination of results are not unique to Essure,” the NEJM authors wrote. “The 13-year history of Essure emphasizes the necessity for thorough examination and timely reporting of patient outcomes in well-conducted premarketing clinical trials and dedicated follow-up in postmarketing studies. Only then will we better understand the risks and benefits of various devices.”

One case of hidden unfavorable data regarded the antidepressant Paxil when maker GlaxoSmithKline deliberately hid two studies showing the drug had modest results compared with a placebo and could increase the risk of suicide in children. In 2012, GSK pleaded guilty and agreed to pay $3 billion for fraud, including failing to report safety data regarding Paxil, Wellbutrin, and Avandia.

But when pharmaceutical companies pay out fines, it’s most often for marketing their drugs for “off label” uses or those that the drugs have not been approved to treat.

Pitts, who is also the global food, drug and policy expert at YourEncore, says these kinds of cases highlight the importance of proper labeling language so physicians are aware of a drug’s risk as part of an “ongoing, imperfect system.”

“It’s complicated,” he said. “Nothing is for free and there’s no product without risk.”

Read More: Prescription Drugs Are Leading to Heroin Addictions »

Trials Don’t Reflect Real-World Patients

Patients looking to enroll in a trial can check ClinicalTrials.gov, but Tom Krohn, chief development officer of Antidote, says that’s an overly scientific and confusing process for the average patient.

For example, there are more than 1,000 different ways to say “not pregnant” in trial recruitment data.

Antidote — a patient recruitment service for clinical trials — is one company helping to close the gap between researchers and trial participants. Those conducting the trials, typically pharmaceutical companies, dictate what specific patients they are looking to recruit.

“It’s a big challenge,” Krohn told Healthline. “From the patient’s perspective, they are looking for help for their disease and a clinical trial is one way to do it.”

Scientific discovery is beholden to imperfect human behavior, which means controlling for variables is even more difficult. People leave trials. They stop taking medication.

“You can’t force people to follow-up that don’t want to. You can’t force a patient to keep taking a drug,” Thompson said. “Those are the realities of dealing with humans.”

More than half of drugs destined for U.S. patients are tested overseas. Medically, this can create problems as they may metabolize drugs differently than Americans.

In poorer countries — India, Malawi, Thailand, etc. — where more drug trials are being conducted, patients will often lie to stay on drugs in trials because it’s the only way they can get needed medication, Thompson said.

Pastor says these trials also often exclude parts of the population who will end up taking the drugs and they affect each person differently, showing flaws in the drug-approval process.

“The more we learn, the more we realize that every patient’s disease is a rare disease,” he said. “One need only look at the sheer number of withdrawals, as well as adverse drug events (ADE) and deaths, associated with approved, marketed drugs — 2 million ADE and 100,000 fatalities annually in U.S. alone — which have gone through decades of human testing and use, to realize that something is very wrong with the current model.”

Andrea LaFountain, PhD, a cognitive psychologist with eight years of healthcare experience, says data from clinical trials vastly over-promise health benefits because on average people take half of their prescribed medications.

In trials, if a person takes less than 90 percent, “they are terminated from the trial and their data is scrubbed from the record.”

“Clinical trials do not factor in this lower consumption rate in the real world,” she told Healthline. “And even when pharmaceutical companies support patients with coupons and reminder programs, adherence rates do not lift to the levels seen in clinical trials.”

While the clinical trial process isn’t perfect, it’s currently the complicated way drugs go from the lab to the real world.

“Of course there needs to be improvements,” Thompson said. “There always needs to be improvements, but (we) can only get them with this kind of commentary.”