Bottles of the sedative midazolam are shown at a hospital pharmacy in Oklahoma City. (Associated Press)

David Waisel is an associate professor of anesthesia at Harvard Medical School.

In theory, executions by lethal injection using three drugs can be efficient, effective and humane. But lately, in practice, they have been anything but.

Take the execution of Kenneth Williams last week in Arkansas: Media and attorney witnesses reported movements indicating that Williams did not lose consciousness after receiving the first drug. He struggled to breathe while the sedative — known as midazolam — took effect. His “head began rocking forward and back,” and his chest “began convulsing up from the table,” one witness said. A minute later his “convulsions became more violent, and there was an audible groan of pain.” Four minutes after the start of the execution, Williams was “gasping for air” and appeared to be “repeatedly clenching and unclenching” his jaw.

Williams’s responses were not surprising. Mounting evidence suggests that midazolam does not anesthetize inmates during executions, as shown by movement and difficulty breathing (each a sign that someone isn’t anesthetized) long after injection in at least four other executions since 2014. Midazolam is clearly unfit for use in lethal injection.

If lethal injection goes according to plan, the first drug, an anesthetic, makes the inmate unconscious; the second, a paralytic, stops all movement; and the third, potassium chloride, stops the heart. The first drug is key to ensuring a humane execution, because the paralytic and potassium will cause pain and suffering if the inmate is not anesthetized. For this reason, the first drug must have certain characteristics: rapid onset, so that the inmate does not go through an extended period of difficulty breathing; analgesia, to block the burning pain of potassium chloride; and consistency of effect for a given dose.

In the four executions performed in Arkansas last month, the first drug administered to the inmates was midazolam. It is a sedative that does not provide rapid onset, analgesia or a consistent effect (even with repeated doses). When midazolam is used, executions predictably go awry. It is virtually impossible to say how awry, though, because Arkansas conceals much about how its executions are performed.

Assuming Williams’s execution followed the state’s protocol, he was initially administered 500 milligrams of midazolam, after which he continued to move and show signs of consciousness for five or more minutes. He eventually stopped moving, so he may have received a second 500-milligram dose, but this is not known. At some point after receiving the midazolam, he was administered 100 milligrams of the paralytic drug and then an equivalent dosage of potassium chloride.

The descriptions of Williams’s reactions to midazolam — particularly the bucking of his head and the rising and falling of his chest — demonstrate that the drug did not render him unconscious and that he was experiencing airway obstruction. In other words, the little we do know suggests that Williams was struggling to breathe and was likely experiencing suffocation.

Midazolam’s obvious inadequacies are exacerbated by the use of the paralytic drug and potassium chloride. The sole purpose of the paralytic drug is to make executions look calm and peaceful. It forestalls further movement and provides a “chemical curtain” to preserve the tableau, hiding whether inmates are aware while paralyzed (a gruesome experience) and preventing them from recoiling violently from the burning sensation of potassium chloride. The paralyzing drug allows for the appearance of serenity that enables the state to describe the execution as “flawless.”

The harm of the paralytic drug goes beyond an individual execution. Using a chemical curtain to hide the response to the most painful part of the execution — the injection of potassium chloride — gives the false impression that midazolam is an appropriate drug for use in executions and allows states to claim success for executions that “look okay.”

The state’s self-serving statements that Williams’s execution was flawless and proceeded according to plan do not make it so, especially when numerous eyewitnesses contradict the version of events the state is promoting. We need more information if we want any meaningful evaluation of Williams’s execution and the three others conducted in Arkansas last month. That should include the precise times the paralytic drug was administered and the results of complete autopsies, including toxicology testing.

Regardless, violent and painful executions will continue as long as we attempt to use midazolam as an anesthetic. A drug cannot do what it does not do. The cynical and expeditious use of paralytic drugs will whitewash some of the more spectacular suffering, providing cover for those who want to play down “convulsing up from the table” and “gasping for air.” To expect otherwise violates the laws of pharmacology and physiology.