Two days before the execution, Charles Ryan, the director of the Arizona Department of Corrections, drove to a Phoenix parking lot early in the morning. He was meeting a man he had never seen, though they had spoken by phone a couple of times. The man, a doctor, had once assisted Ryan’s predecessor. Now Ryan needed his help: The state was putting Jeffrey Landrigan to death on October 26, 2010, its first execution in three years. Death-penalty states often struggled to find medical professionals willing to participate, but the doctor had agreed after a five-minute phone call. He got into Ryan’s SUV for the hourlong ride to the Arizona State Prison Complex in Florence. Driving through the desert, the men talked about Iraq. The doctor was a veteran, and Ryan had helped set up the country’s prison system after the U.S. invasion. 

In Florence, Ryan introduced the doctor to the medic and prison guards who would assist in the execution. One prison guard lay on a gurney, and the doctor rehearsed his method. While most states executed prisoners using peripheral IVs in the arms, the doctor preferred to set a central line in the femoral vein in the prisoner’s groin. It was a more invasive procedure, but, in his opinion, a more reliable and comfortable one. The team walked through the execution several times. Later, Ryan drove the doctor back to Phoenix.

The following evening, a U.S. District Court judge issued a stay of Landrigan’s execution—a temporary postponement until Arizona revealed the source of its lethal drugs. The state had kept the source a secret because it had gone to extraordinary lengths to buy the drugs it would use in a three-drug cocktail, and it feared public disclosure would jeopardize future access. (The state defended its position by citing the same law that has kept the doctor’s name confidential.) Without knowing the source, Landrigan’s lawyers said, it was impossible to ensure the drugs would work properly and satisfy the Eighth Amendment protection against cruel and unusual punishment. Despite the holdup, Ryan and the doctor returned to Florence on October 26. If a higher court lifted the order, they would be ready to proceed.

The U.S. Court of Appeals for the Ninth Circuit maintained the stay. “The State’s gamesmanship is unseemly at best, and inhumane at worst,” the judges wrote. Arizona appealed to the Supreme Court, whose five conservative justices quickly voted that same day to let the execution proceed. “Speculation cannot substitute for evidence that the use of the drug is ‘sure or very likely to cause serious illness and needless suffering,’” they wrote in a one-page order, quoting a 2008 opinion, Baze v. Rees, in which they held that three-drug lethal injection was constitutional. Landrigan’s execution was back on track.

The Arizona State Prison in Florence early on Tuesday, October 26, 2010—hours before Jeffrey Landrigan’s scheduled execution. Associated Press / Matt York

At 9:31 p.m. that night, the prison guards strapped Landrigan to a gurney. The doctor placed a sterile drape over Landrigan and let him know that he would soon feel a prick in his groin from the local anesthetic. A medic used an ultrasound machine to locate Landrigan’s femoral vein, which the doctor penetrated with a needle. He struggled, however, to slide the wire—which would guide a catheter that would deliver the drugs into the vein—along the needle: The gurney strap and Landrigan’s underwear got in the way. The doctor had a prison guard rotate Landrigan’s foot to better expose his groin and succeeded on his second try. He completed the central line and stitched the catheter to Landrigan’s thigh.

As he lay on the gurney, Landrigan thanked his family for attending and spoke his last words: “Boomer Sooner,” a lyric from the University of Oklahoma’s fight song. An execution team member began pushing the first drugs at 10:15 p.m. Eleven minutes later, the doctor declared Landrigan dead. For some reason, Ryan asked the execution team to administer the backup drugs to Landrigan’s corpse, but without any blood pressure, it was difficult to depress the syringe. The doctor warned that Landrigan’s vein might rupture, and Ryan called them off. The doctor’s work was done, and Ryan paid him $18,000 cash. He claimed he didn’t learn until later, from news accounts, that the drug that had killed Jeffrey Landrigan was illegal. 

Lethal injection was first adopted in Oklahoma in 1977 as a less violent alternative to the gas chamber and the electric chair. Over the next 25 years, almost every death-penalty state copied Oklahoma’s three-drug formula: first the barbiturate sodium thiopental to knock the prisoner out, then the paralytic pancuronium bromide to immobilize him, and finally potassium chloride to stop his heart. The second and third drugs would cause intense suffering on their own, but the Supreme Court ruled that the method was constitutional in Baze: As long as the thiopental rendered the prisoner unconscious, he would be insensate to the agonizing effects of the next two drugs.1 Just one year after the Baze decision, though, in late 2009, the pharmaceutical company that sold thiopental to every death-penalty state, Hospira, reported a shortage.

As a consequence, the death penalty has undergone in the past five years its biggest transformation since states began switching to lethal injection decades ago. As thiopental disappeared, states began executing prisoners with experimental one-, two-, or three-drug cocktails. States have essentially been improvising what is supposed to be one of their gravest and most deliberate duties, venturing deep into the shadows to carry out executions. They have turned to mail-order pharmaceutical suppliers and used untested drugs. They have sidestepped federal drug laws, minimized public disclosure, and, on multiple occasions, announced changes to execution protocols just hours before prisoners were set to die. The machinery of death in the United States has become a kluge. 

In April, the Supreme Court acknowledged this when it heard oral arguments in Glossip v. Gross. A group of prisoners from Oklahoma—including Richard Glossip, a convicted murderer challenged the state’s use of a drug called midazolam because they feared it would not anesthetize them. The court had hoped Baze would obviate future lethal injection lawsuits, but the thiopental shortage had stripped the decision of any practical relevance almost as soon as it was issued. Now, just seven years later, the justices were considering whether they should invalidate a specific method of execution for the first time in U.S. history. The court’s decision won’t overturn the death penalty, but it will define the way we practice it for years to come. 

Though the challenge comes from Oklahoma, it is Arizona that provides the best case study of the rapid, slipshod evolution of lethal injection since Baze. The desert state hasn’t executed the most prisoners since the thiopental shortage began—that distinction belongs, as always, to Texas—but it has used more methods than any other state, killing prisoners with four different drug combinations. No other state has been quite so dogged in its determination to carry out executions. And no other state has left so detailed a paper trail. Judges, lawyers, and journalists (most notably Michael Kiefer at The Arizona Republic) have brought much of the abuses to light over the years, but the story has been told in disparate pieces: a deposition here, an uncovered email there. The complete narrative is more troubling than any one of its components.

In Landrigan’s case, the search for thiopental was particularly frantic. On September 21, 2010 the state Supreme Court scheduled Landrigan’s execution and gave Ryan only ten days to announce whether or not he had the drugs to carry out the death sentence. At that point, Ryan had no drugs and no leads. He and his employees began calling other state prison officials. They got lucky in Arkansas: Prison officials there had learned about a British wholesaler called Dream Pharma, run out of the back room of a West London driving school. Ryan’s deputy, Charles Flanagan, cold emailed Dream Pharma, laying out everything that Arizona needed. “I cannot stress enough how critical and time sensitive this is,” he wrote.

Flanagan had an unlikely background for the execution business: In the 1970s and 1980s, he worked as a professional dancer with the Dutch National Ballet in Amsterdam before he moved to Arizona in 1985 and became a prison guard. After he wrote to Dream Pharma, he checked with the state’s traditional pharmaceutical broker, the Minnesota Multistate Contracting Alliance for Pharmacy (mmcap), to see if it could help purchase the drugs from Dream Pharma. “It’s pretty likely that the product they have is not approved by the FDA,” a company representative from mmcap replied.2 Unapproved thiopental “might cause trouble you don’t want in this particular situation,” she wrote, and then continued: 

In my opinion, the Dream Pharma website leaves something to be desired. It is nothing like the pharmaceutical wholesale distribution websites we use here in the United States. It makes me wonder whether Dream Pharma is reputable and where exactly the medication would be coming from. There is a “gray market” in the pharmaceutical industry, and in this particular instance you need to be sure that the product is actually thiopental and that it is going to work.

The representative said she could help Arizona buy a substitute barbiturate. But Flanagan decided to stick with Dream Pharma, which had replied quickly and enthusiastically to his email. The mmcap rep was right, though, that Flanagan would need to be careful around the Food and Drug Administration (FDA). Local FDA officials had put a hold on Arkansas’s thiopental when it landed at the FedEx hub in Memphis. Ryan and Flanagan could not afford any such delay if they wanted to execute Landrigan on time. 

Flanagan hired a customs broker—a kind of fixer for the complicated importation process—to help arrange the order. Then he reached out to U.S. Customs and Border Protection, which referred him to the Phoenix office of the FDA. There, Flanagan met an FDA official named Dave Thomas, who usually spent his days checking boxes and seemed eager for the opportunity to participate in more important matters. He gave Flanagan his personal cell phone number and told him not to hesitate to call on weekends. (Attempts to contact Thomas were unsuccessful.)

Although the order of thiopental would land first in Memphis, Flanagan arranged it so that the drugs would not be inspected until the package arrived in Phoenix. On September 24, Thomas wrote an email to customs letting them know, “There will be no FDA detention or inspection of these products.” (When Thomas was shown this email later in a deposition, he said he had no recollection of writing it.) He told Flanagan not to worry: “If there are any problems whether in Memphis or Phoenix, please let me know. I’ll ensure an expedient release.”

Arizona’s thiopental touched down in Phoenix and was then sent to Los Angeles by mistake, but it made it back to Phoenix by September 29, two days before the court deadline. Thomas personally inspected the shipment, signed off on it, and released it to the custody of the Arizona Department of Corrections. As this was all being worked out, the California Department of Corrections had reached out to Ryan to ask for help in its own search for thiopental. One day after Arizona obtained its fresh supply, prison employees from California arrived in Florence to pick up twelve grams of the drug, which Ryan had agreed to give them. Scott Kernan, a California corrections official, thanked him in an email: “You guys in AZ are life savers.”

The federal government never wanted to be involved in state executions. So long as states purchased thiopental from Hospira, the FDA had been able to avoid getting involved. When the states turned to unapproved drugs from overseas, however, matters became more complicated. Three days before Landrigan was set to be executed, one of Landrigan’s attorneys, a federal defender in Phoenix named Dale Baich, wrote a letter to the chief counsel at the FDA. “I am requesting that you expeditiously investigate whether the State of Arizona and its agencies are in possession of a non-FDA-approved drug, and whether the State intends to use it on a human being,” Baich wrote. The FDA did not respond to Baich, but the letter alarmed high-level officials.

Close-up of sodium thiopental in a tray. Thiopental is used as a barbiturate general anesthetic. Universal Images Group via Getty Images

In November, after Landrigan’s execution, the FDA heard from the British government, which was unhappy about the importations and wanted export controls to prevent its drug companies from profiting off U.S. executions. Soon, the FDA was “treating this as an emergency matter,” according to internal emails. “The department is getting a lot of pressure from the White House,” the director of the FDA’s Division of Import Operations wrote to his colleagues. That fall, the FDA put holds on shipments of Dream Pharma thiopental that Arizona, South Carolina, and California had been able to order before the British government cut them off. 

The agency was buying time as it figured out what to do. If the FDA rejected the thiopental imports, it would risk a political backlash in conservative states. (Flanagan left angry messages about the detained thiopental and threatened to go to the press.) If it allowed the imports, it might break the law. The latter route, the FDA apparently decided, was less risky. On January 4, the agency issued a statement. “Reviewing substances imported or used for the purpose of state-authorized lethal injection clearly falls outside of FDA’s explicit public health role,” it read. “Accordingly, FDA chooses to continue to defer to law enforcement on all matters involving lethal injection.” Within a few days, the FDA released the shipments of thiopental to the state departments of corrections.3

Not long after the FDA released the thiopental, the Drug Enforcement Administration swooped in. Thiopental was a controlled substance, and most of the states that had imported the drug hadn’t registered for a DEA import license. Consequently, they had run afoul of the Controlled Substances Act, which regulated the movement of drugs with the potential for recreational abuse. In 2011, the DEA confiscated thiopental from Georgia, Arkansas, South Carolina, Kentucky, Alabama, and Tennessee. For some reason, the agency left Arizona alone, even though it, too, lacked the import license. Arizona used the Dream Pharma thiopental again in an execution on March 29, 2011. 

The next federal intervention came from the Justice Department in the spring of 2011, as Arizona prepared to execute a man named Donald Beaty on May 25, 2011. By that point, federal defenders had traced the thiopental trade to Dream Pharma. The Justice Department asked Arizona not to use the drug, since it had been imported in violation of the Controlled Substances Act, and said the DEA would be happy to help Arizona ensure that future orders complied with the law.

In response, Ryan announced the day before the execution that the state would replace thiopental with a similar drug called pentobarbital. Beaty’s lawyers had no idea where Arizona had gotten the pentobarbital and had only 24 hours to investigate its quality. They were not successful, and Beaty was executed as planned. In 2011, Arizona carried out two more executions using pentobarbital, pancuronium bromide, and potassium chloride. Pentobarbital had the same vulnerability as thiopental, though: The only FDA-approved source was the Danish company Lundbeck. On July 1, 2011, Lundbeck announced that it would not sell to state departments of corrections and would require all buyers to sign a contract forbidding them from reselling the drug for use in executions.4 The death penalty states recognized that they had a problem: Drug companies didn’t want to be publicly associated with lethal injections. The next stop for some states, like Texas and Georgia, was compounding pharmacies, state-regulated facilities that make custom mixtures of drugs for the specific needs of individual clients.5

At the same time, states began acting more secretively than ever. Public scrutiny was creating both logistical and legal challenges, and the states’ response was to basically obstruct it. The Georgia Legislature, for instance, passed a law called the “Lethal Injection Secrecy Act” saying the state would never reveal the source of its execution drugs, not even under court order. Critics warned that these surreptitious methods courted disaster. Landrigan’s execution had already demonstrated how prison officials were willing to bend rules and cut corners in order to carry out death sentences. Landrigan had not appeared to suffer as he died, but it seemed only a matter of time before a prisoner did. 

If there’s one state to rival Arizona in its death-chamber persistence over the past several years, it’s Ohio. The Midwestern state began experimenting with lethal injection after a particularly gruesome episode in 2009 in which a condemned prisoner, Romell Broom, exited the death chamber at the Southern Ohio Correctional Facility alive because his executioners failed to get an IV line into his veins. In a good-faith effort to make lethal injection less complicated, Ohio then began using a single-drug protocol: A botched execution with only thiopental or pentobarbital would make the prisoner extremely groggy, whereas a botched execution with pancuronium bromide and potassium chloride could suffocate him through chemical paralysis or burn him.6

Over the past few years, eight states have executed prisoners using just one drug—including the more active death-penalty states, like Texas, Georgia, and Missouri—while several more have said they will do so in their next executions. In February 2012, Arizona switched to a single dose of pentobarbital after Ryan discovered just days before an execution that one of the drugs it intended to use had expired. The appeals court denied a stay but rebuked Arizona for “leaving the courts with a rolling protocol that forces us to engage with serious constitutional questions and complicated factual issues in the waning hours before executions.” Arizona carried out the following seven executions with a single dose of pentobarbital.

The Knock-Out Drugs

A large part of the scramble for lethal injection drugs stems from the need for a powerful knock-out drug. In recent years, manufacturers and supply chains have been shut down, forcing states to get creative about substitutes.

Sodium Thiopental

The first drug in three-drug lethal injection, thiopental was used by every state with lethal injection until 2010. States began turning to other drugs after thiopental’s sole FDA-approved manufacturer, Hospira, reported a shortage in late 2009. It was last used in an execution in 2011.


A similar barbiturate to thiopental, pentobarbital was first used in Oklahoma in 2010. Its sole manufacturer, the Danish company Lundbeck, banned buyers from using its drugs in executions in 2011. Some states have been able to continue buying it from loosely regulated compounding pharmacies. It has also been used in one-drug executions.


First adopted in Ohio as part of a two-drug mix, midazolam was later used to replace thiopental or pentobarbital in three-drug lethal injection. It is not a barbiturate and is not usually used on its own in medicine for anesthesia. The Supreme Court is currently deciding whether it is an adequate replacement for thiopental and pentobarbital.

The switch to a single-drug lethal injection did not actually address the problem of the Broom execution. A one-drug execution with pentobarbital still required intravenous access. So Ohio also added a “Plan B” that could be injected into a prisoner’s muscle: a mixture of two drugs, midazolam and hydromorphone. Neither had been used before in an execution. And neither drug is generally used on its own in medicine to induce and maintain anesthesia: Midazolam is most often administered as a sedative to calm patients before surgery, while hydromorphone is a powerful painkiller.

To determine the drugs’ viability, lawyers for Ohio corresponded with an anesthesiologist from the University of Massachusetts Medical School named Mark Dershwitz, who assured them that a large dose of the two drugs in combination would knock out a prisoner and stop his breathing. In emails obtained last year by the new republic, Ohio’s lawyer worried that the combination could leave a prisoner “gasping for air in a hyperventilating fashion, with eyes still open,” an image that might incite litigation. But Dershwitz wrote that this would be unlikely: The two drugs were used commonly in combination for conscious sedation. “I have never seen the scenario described (in my 27-year career) although I will admit nothing is impossible,” he wrote. 

Once midazolam was enshrined in the Ohio protocol, other states—first Florida, then Oklahoma—began using it to replace pentobarbital in the three-drug cocktail, even though it was a completely different class of drug. When Ohio ran out of pentobarbital in late 2013, it then tweaked “Plan B” into “Plan A,” announcing it would kill prisoners with an intravenous injection of midazolam and hydromorphone. 

It was far from a perfect solution. Ohio used midazolam and hydromorphone in an execution for the first time on January 14, 2014 on a prisoner named Dennis McGuire. The execution took 26 minutes, the longest lethal injection in Ohio history, and witnesses said McGuire clenched his fists and appeared to choke for ten minutes. 

Oklahoma’s first experience with midazolam, which it used in sequence with vercuronium bromide and potassium chloride, was even more alarming. On April 29, 2014, a convicted murderer named Clayton Lockett woke up partway through his execution. “The drugs aren’t working,” he said. “This shit is fucking with my mind.” The warden called off the execution, but by that point it was too late for Lockett, who died 43 minutes after the lethal injection began. An autopsy revealed just how badly Oklahoma had butchered the execution. A paramedic had tried three times to insert the IV needle into Lockett’s arm before a doctor took over and attempted to stick it in Lockett’s jugular. That failed too, so the doctor then tried to insert the IV in the subclavian vein beneath Lockett’s collarbone, while the paramedic tried his right arm and his right foot. These efforts also failed, so the doctor then went for the vein in Lockett’s groin. But the doctor missed, so the drugs pooled in Lockett’s soft tissue. Lockett appeared to be conscious and fighting the onset of chemical paralysis.7

Despite these unpropitious precedents, Ryan announced that Arizona would copy Ohio’s midazolam and hydromorphone combo for the execution of Joseph Wood that July. Arizona said the drugs were FDA-approved and domestically manufactured but refused to disclose any other information. The appeals court stayed the execution. “We, and the public, cannot meaningfully evaluate execution protocol cloaked in secrecy,” the judges wrote. The Supreme Court, however, lifted the stay with no explanation. Wood’s execution began at 1:52 p.m.

It was the longest execution in modern U.S. history. Wood gasped and snorted for one hour and 57 minutes as Ryan ordered the executioners to inject the prisoner with 15 different doses of midazolam and hydromorphone. It was exactly the scenario that the Ohio lawyer originally feared. Wood’s lawyers called a judge midway through the procedure to try to get it stopped, but they did not succeed. Afterward, Ryan insisted that Wood’s execution had not been botched, but Arizona has nevertheless put executions on hold for now. 

Arizona will likely resume executions after the Supreme Court issues an opinion for Glossip v. Gross. Although the case came out of Oklahoma—and the central matter is the use of a single drug, midazolam—it involves many of the same issues lawyers have been trying to draw attention to since Landrigan’s execution back in 2010: the evasive tactics of the state, the widespread use of untested drugs. When the Court heard arguments in April, it was a federal defender from Phoenix, an attorney named Robin Konrad, who presented the prisoners’ case.

Midazolam, Konrad argued, could not maintain a comalike state of unconsciousness; painful stimuli, like suffocation or burning, could cause a prisoner to be “jolted” into consciousness, where he would be powerless to express his agony because of the paralytic drug. Any execution with midazolam, she argued, posed an unconstitutional risk of suffering. The state’s lawyer, meanwhile, argued that midazolam worked just fine. 

The decision will probably hinge, as it often does, on the vote of Justice Anthony Kennedy and maybe Chief Justice John Roberts. Neither gave much indication of his leaning during oral arguments. The other justices, however, turned the occasion into what Slate’s Dahlia Lithwick called “a proxy war about ideology and politics and the ugly rift between the justices on how we feel about killing people in America.” Justice Sonia Sotomayor accused Oklahoma’s attorney of dishonesty, while Justice Samuel Alito said death-penalty opponents were waging a “guerrilla war” against capital punishment. The justices entertained a long thought experiment about whether it would be constitutional to burn someone at the stake if he were properly anesthetized. Overall, though, the debate was, essentially, more about medicine and pharmacology than it was about the law. Unless the Court remands the case to a lower court, the decision will likely rest on nine judges’ interpretations of technical data from a few medical studies. 

And the outcome may only accelerate the cycle of semi-illicit experimentation. If the Court allows the drug, Ryan has already said that Arizona will use a three-drug combination with midazolam; but if the Court accepts the arguments of the plaintiffs, and rejects midazolam, states may just come up with new methods of execution altogether. Two weeks before oral arguments in Glossip, Oklahoma—the birthplace of lethal injection—added a backup method, hypoxia by nitrogen gas, to its books. The Oklahoma representative who wrote the bill sold it with a guarantee: “It’s foolproof.”

  1. Without thiopental, Chief Justice John Roberts noted in the plurality opinion, an execution would pose a “substantial, constitutionally unacceptable risk of suffocation from the administration of pancuronium bromide and of pain from the injection of potassium chloride.”
  2. In fact, Hospira was the only FDA-approved manufacturer of thiopental, so all other sources were unapproved.
  3. In 2012, a district court held that the FDA clearly broke the law. “The FDA appears to be simply wrapping itself in the flag of law enforcement discretion to justify its authority and masquerade an otherwise seemingly callous indifference to the health consequences of those imminently facing the executioner’s needle,” Judge Richard Leon wrote. A circuit court upheld the decision in 2013. By that time, however, the victory was largely academic. Arizona had not used thiopental in an execution in over two years.
  4. By the end of 2011, the European Union had followed Great Britain’s lead and placed strict export controls on barbiturates that could be used in U.S. executions.
  5. South Dakota was the first state to use compounded pentobarbital in an execution. It was later found to be contaminated with fungus.
  6. In Baze, the plaintiffs’ attorneys actually proposed a one-drug execution as a safe alternative to the three-drug protocol they were challenging. This was, after all, how veterinarians all over the country put animals to sleep.
  7. In the Glossip oral arguments, Chief Justice Roberts said the problem with Lockett’s execution was IV administration, not midazolam. But Ohio specifically chose midazolam because it believed it would work even if injected into a prisoner’s muscle instead of his bloodstream.