Lethal Injection for Execution: Chemical Asphyxiation?

Teresa A. Zimmers; Jonathan Sheldon; David A. Lubarsky; Francisco López-Muñoz; Linda Waterman; Richard Weisman; Leonidas G. Koniaris


PLoS Med. 2007;4(4) 

In This Article


Most US executions are beset by procedural problems that could lead to insufficient anesthesia in executions. This hypothesis has been supported by findings of low postmortem blood thiopental levels and eyewitness accounts of problematic executions. Herein we report evidence that the design of the drug scheme itself is flawed. Thiopental does not predictably induce respiratory arrest, nor does potassium chloride always induce cardiac arrest. Furthermore, on the basis of execution data and clinical, veterinary, and laboratory animal studies, we posit that the specified quantity of thiopental may not provide surgical anesthesia for the duration of the execution. Thus some inmates may experience the sensations of pancuronium-induced paralysis and respiratory arrest.

In the United States and Europe, techniques of animal euthanasia for clinical, laboratory, and agricultural applications are rigorously evaluated and governed by professional, institutional, and regulatory oversight. In university and laboratory settings, local oversight bodies known as Animal Care and Use Committees typically follow the American Veterinary Medical Association's guidelines on euthanasia, which consider all aspects of euthanasia methods, including drugs, tools, and expertise of personnel in order to minimize pain and distress to the animal. Under those guidelines, lethal injections of companion or laboratory animals are limited to injection by qualified personnel of certain clinically tested, Food and Drug Administration-approved anesthetics or euthanasics, while monitoring for awareness.

In stark contrast to animal euthanasia, lethal injection for judicial execution was designed and implemented with no clinical or basic research whatsoever. To our knowledge, no ethical or oversight groups have ever evaluated the protocols and outcomes in lethal injection. Furthermore, there are no published clinical or experimental data regarding the safety and efficacy of the three-drug lethal injection protocol. Until the unprecedented and controversial use of bispectral index monitoring in the last two North Carolina lethal injections,[19] no monitoring for anesthesia was performed. Given this paucity of knowledge and documentation, we sought to evaluate available data in order to determine the efficacy of the three drug protocol.

The designers of lethal injection intended that each of the drugs be fatal independently and that the combination provide redundancy.[2] Moreover, in legal challenges to the death penalty, the leading expert witness testifying on behalf of the states routinely asserts that 3 g of thiopental alone is a lethal dose in almost all cases.[14] The data presented here, however, suggest that thiopental alone might not be lethal. First, extrapolating from clinical use, the lowest dosages used in some jurisdictions would not be expected to kill. Calculated dosages in North Carolina executions using 3 g of thiopental ranged from 10 to 45 mg/kg. Assuming inmates are roughly the same size across jurisdictions, the dose range would be 17-75 mg/kg in California, where 5 g of thiopental is used, and 6.6-30 mg/kg in Virginia and other jurisdictions, which use 2 g. Thus, at the lowest doses, thiopental would be given near the upper range of that recommended for clinical induction of anesthesia (3-6.6 mg/kg) -- clearly not a dose designed to be fatal.[20] Second, the calculated doses used across lethal injections are only 0.1-2 times the LD50 (dose required to kill 50% of the tested population) of thiopental in dogs (37 mg/kg), rabbits (35 mg/kg), rats (57.8 mg/kg), and mice (91.4 mg/kg).[21,22] Third, intravenous delivery of thiopental alone is not recommended by The Netherlands Euthanasics Task Force, which concluded "it is not possible to administer so much of it that a lethal effect is guaranteed",[23] even in their population of profoundly ill patients.

The most compelling evidence that even 5 g of thiopental alone may not be lethal, however, is that some California inmates continued to breathe for up to 9 min after thiopental was injected. This observation directly contradicts testimony of that state's expert witness, who asserted that "this dose of thiopental sodium will cause virtually all persons to stop breathing within a minute of drug administration" and that "virtually every person given 5 grams of thiopental sodium will have stopped breathing prior to the administration of the pancuronium bromide".[24] The witness has made identical statements regarding 3 g of thiopental.[14] Indeed, the clinical literature is replete with examples of patients experiencing respiratory failure after even low doses of thiopental.[25] Others, however, experience merely transient, nonfatal apnea. Of course, for inmates who did not stop breathing with thiopental alone, it is impossible to know whether the thiopental solution was correctly mixed, whether the entire dose was administered intravenously, or whether the apparent resistance was due to bolus dosing or individual variation. It remains possible, however, that bolus dosing of 5 g of thiopental alone might not be fatal in all persons. Indeed, nonhuman primates given as much as 60 mg/kg (the mass equivalent of 6 g for a 100 kg man) experienced prolonged sleep, but ultimately recovered.[26]

If thiopental does not reliably kill the inmates, then perhaps death is effected by potassium chloride. Rapid intravenous or intracardiac administration of 1-2 mmol/kg potassium chloride under general anesthesia is considered acceptable for euthanasia of large animal species; thus the 1.11-2.35 mmol/kg doses given in North Carolina's lethal injections ought to be fatal. If potassium chloride contributes to death through cardiotoxicity, however, cardiac activity ought to cease more quickly when potassium is used than when it is not. Indeed, such is the principle behind the animal euthanasia agent, Beuthanasia-D Special, in which the cardiotoxic effects of phenytoin synergize with the central nervous system-depressive effects of pentobarbital, accelerating death over pentobarbital alone.[27] In contrast, our analysis shows that use of potassium chloride in North Carolina's Protocol C did not hasten death (defined as flatlining of the ECG) over Protocol A, which used thiopental and pancuronium alone. Moreover, in California executions, ECG flatlining was noted from 2 to 9 min after potassium chloride administration. This observation contrasts sharply with reports of accidental bolus IV administration of concentrated potassium chloride solution, in which patients experienced complete cardiopulmonary arrest almost immediately upon injection.[28] The North Carolina and California data together suggest that potassium chloride might not be the lethal agent in lethal injection.

Given that neither thiopental nor potassium chloride can be construed reliably to be the agent of death in lethal injection, death in at least some inmates might have been due to respiratory cessation from the use of pancuronium bromide. The typical use of 0.06-0.1 mg/kg pancuronium bromide under balanced anesthesia produces 100% neuromuscular blockade within 4 min, with approximately 100 min required for 25% recovery.[29] The doses used in North Carolina were some 3-11 times greater than the typical intubation dose, and thus would be expected to produce more rapid paralysis of many hours duration and complete respiratory arrest.[30] Indeed, pancuronium might have been the agent of death even in inmates who ceased breathing coincident with or shortly after injection of pancuronium, rendering permanent the thiopental-induced apnea. In addition, because pancuronium bromide is effective even when delivered subcutaneously or intramuscularly, pancuronium is likely the sole agent of death when IV catheter misplacement or blowout impairs systemic delivery of the other two drugs. In such cases death by suffocation would occur in a paralyzed inmate fully aware of the progressive suffocation and potassium-induced sensation of burning. This was likely the experience of Florida inmate Angel Diaz, whose eyes were open and mouth was moving 24 min into his execution and who was pronounced dead after 34 min. Findings of two 30-cm burns over both antecubital fossae prompted the medical examiner to conclude that the IV lines were misplaced and the drugs were delivered subcutaneously.[31]

Executions such as Diaz's, in which additional drugs were required, constitute further evidence that the lethal injection protocols are not adequate to ensure a predictable, painless death. Court documents and news reports indicate that at least Virginia,[32] California,[10] and Florida[31] have administered additional potassium chloride in multiple executions when the inmate failed to die as expected. If a Virginia execution takes too long and if the inmate fails to die, the protocol indicates that additional pancuronium and potassium chloride should be injected, although there is no provision for additional thiopental.[32] In cases such as Diaz's, additional drugs may have been required due to technical problems with delivery, but it remains possible that in others, the standard drug protocol failed to kill.

Given the uncertainty surrounding the mechanism of death and low postmortem blood thiopental levels in some executed inmates,[7] one must ask whether adequate anesthesia is maintained to prevent awareness and suffering. Medical experts on both sides of the lethal injection debate have asserted that 3 g of thiopental properly delivered should reliably result in either death or a long, deep surgical plane of anesthesia.[13,14] In support of this contention, continuous or intermittent thiopental administration was formerly used for surgical procedures lasting many hours. In one study, 3.3-3.9 g given to patients over 25-50 min resulted in sleep for 4-5.5 h.[33] Depth and duration of thiopental anesthesia depends greatly upon dose and rate of administration, however, and bolus dosing results in significantly different pharmacokinetics and duration of efficacy than administration of the same quantity of drug at a lower rate.[22]

In the modern practice of anesthesia, thiopental is used solely to induce a few moments of anesthesia prior to administering additional agents. Anesthesiologists are taught to administer a small test dose while assessing patient response and the need for additional doses.[20] Such stepwise administration and evaluation has been the practice from the first reports of thiopental usage in 1934, due to the known potential for barbiturate-induced respiratory arrest.[34] It was early recognized that age, body composition, health status, anxiety, premedication, and history of substance abuse clearly influence response to thiopental, with some individuals showing marked resistance to standard doses[35] and others fatal sensitivity.[25] Thus the historical and modern clinical use of thiopental results from its cautious application to prevent respiratory arrest both in the typical patient and the abnormally susceptible. In consequence, there is almost no information about duration of anesthesia following large bolus doses of thiopental in unpremedicated patients, and there are few living anesthesiologists with clinical experience relevant to lethal injection protocols.

Unlike in clinical medicine, however, bolus injection of thiopental is regularly practiced in laboratory animals and veterinary medicine. Standard texts specify from 6 to 50 mg/kg thiopental, depending on the species, for 5-10 min of anesthesia,[36] including 18-22 mg/kg for 10-15 min of anesthesia in dogs, pigs, sheep, and swine.[37] Such dosages are conservative guidelines based on average responses of animals in experimental trials ( Table 1 ), with the assumption that respiration and depth of anesthesia will be assessed in individual animals prior to onset of the procedure. (In addition, thiopental is not recommended for painful procedures in animals.) Withholding or administering additional dosages would compensate for individual variation in response.

Although species differences complicate pharmacological comparisons from animals to humans, animal studies are the basis for virtually all human drug trials. According to FDA guidelines, toxicity endpoints for drugs administered systemically to animals are typically assumed to scale well across species when doses are normalized to body surface area (i.e., mg/m2).[38] Calculating the human equivalent dose (HED) as recommended by the FDA[39] gives a more conservative estimate of thiopental equivalencies across species than does using simple mg/kg comparisons ( Table 1 ). Swine in particular are regarded as an excellent model of human cardiopulmonary and cerebrovascular physiology, with comparable size, body composition, and brain perfusion rates.[40] Comparing the HED for thiopental anesthesia in swine to lethal injection dosages, we conclude that at least some inmates at the lower end of the thiopental dose range might have experienced fleeting or no surgical anesthesia, while others at the higher end of the range might have received doses predicted to induce more prolonged anesthesia ( Table 1 ). Such a prediction is impossible to evaluate, however, because any evidence of suffering would be masked by the effects of pancuronium.

Our study is necessarily limited in scope and interpretations. Given the secrecy surrounding lethal injections, we were able to analyze only a small fraction of the 891 lethal injections in the US to date. Indeed, the majority of executions actually take place in states such as Texas and Virginia, where the protocols and procedural problems are likely similar to the ones described, but where the states are unwilling to provide information.[7] Not only are available data limited, however, medical literature addressing the effects of these drugs at high doses and in combination is nonexistent, emphasizing the failure of lethal injection practitioners to design and evaluate rigorously a process that ensures reliable, painless death, even in animals. In consequence, the adequacy of anesthesia and mechanism of death in the current lethal injection protocol remains conjecture.

Despite such limitations, our analysis of data from more forthcoming states along with reports of problematic executions and judicial findings[41] together indicate that the protocol of lethal injection for execution is deeply flawed. Technical difficulties are clearly responsible for some mishandled executions, such as Diaz's. Better training of execution personnel and altering delivery conditions may not "fix" the problem,[41,42] however, because the drug regimen itself is potentially inadequate. Our analysis indicates that as used, thiopental might be insufficient both to maintain a surgical plane of anesthesia and to predictably induce death. Consequently, elimination of pancuronium or both pancuronium and potassium, as has been suggested in California,[41] could result in situations in which inmates ultimately awaken.

With the growing recognition of flaws in the lethal injection protocol, 11 states have now suspended the death penalty, with nine of those seeking resolution of issues surrounding the process.[42] In California and Florida, commissions of experts have been charged with evaluating and refining lethal injection protocols. As deliberations begin, we suggest that the secrecy surrounding protocol design and implementation should be broken. The available data or lack of data should be made public and deliberations should be open and transparent.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: