What killed George Floyd: Part 11 — Prone position

One of the key points of dispute in the trial was the effect of prone position.  The issue of prone position as a cause of death has a rather long and variable history. People also sometimes conflate the effects of prone position in special cases with that of prone position during police restraint.  So let’s start with that.

If you review the literature on prone position, you will find two situations where the evidence of a deliterious effect are fairly strong, that of sudden infant death and sudden death in epilepsy.  Both of these, however, are special cases.  In the case of sudden infant death, the proposed mechanism is that the developing infant is at risk of having his or her head down in some soft material, such as bedding or against a caretaker.  As the child breathes, the local concentration of carbon dioxide increases around the face.  For one reason or another —  perhaps decreased arousability due to the degree of neurologic development, inability to move the head due to development of neck muscles, peculiarities of the structure of the upper airway, or something else, the infant cannot move his or her head to access fresh air.  The more the infant breathes, the higher the carbon dioxide concentration, and eventually the child dies of carbon dioxide intoxication.  This has been experimentally produced in dog models.  Thus, the likelihood of sudden death rises with the likelihood of this kind of situation.  Placing a child in a bed with fluffy pillows and quilts increases risk.  Sleeping in the same bed with an adult has an increased risk, at least if there are other issues involved (e.g. excessive fatigue, inebriation, drug use, obesity, etc.).   Sleeping with an infant on a couch has a 66 times increased risk of sudden infant death.

The second area is that of death in epilepsy.  The situation is somewhat similar, except that the patient cannot adjust his or her access to air because of seizure activity, sometimes complicated by aspiration of gastric contents.

Neither of these situations apply in this case, and literature associated with the above are not applicable to this case.

In the modern era, the first emphasis on prone position as a cause of death in police restraint came from observations made by Don Reay, who was a Medical Examiner in Washington State.   Dr. Reay first became interested in deaths in custody in the 1980s where he did seminal work on neck holds.  He then became interested in prone position and started by doing physiologic testing that showed prolonged recovery from exercise if volunteers were placed in prone position(1). He then noted in 1992  a series of three people who were found dead in the back of patrol cars after restraint(2).  All were prone, and all were “hog tied,” with their arms and legs tied and connected (e.g. arms behind the back and knees flexed).

Breathing, one might remember, is due to the combination of movement of the chest wall *and* movement of the diaphragm.  Dr Reay posited that prone position was more likely to be dangerous because it decreased the excursion (expansion) of the abdomen, which made the use of the diaphragm in breathing more difficult resulting in failure to compensate for metabolic acidosis (an idea that will rise again).  Accordingly, the victim would have to work his or her chest wall more, and would become exhausted.  He wasn’t sure how the hog-tie issue played in.  He also suggested that the increased exertion associated with struggling with police would increase the likelihood of exhaustion.  This article lead to the generalized acceptance that prone position in police restraint was dangerous (though there was some pushback, for instance by Elizabeth Laposada (3) ).

Ron O’Halloran, Larry Lewman, and Janice Frank added more cases to this, with series of 11, and then another of 21 cases of similar deaths (4,5). Bell, et al. also published a series of 30 cases around this time, of positional asphyxia in other settings, noting neck hyperflexion, prone position involving occlusion of the airways, and the importance of comorbidities (e.g. dementia). Alcohol intoxication and aspiration played an important role in these deaths(6).  In 1995, Stratton, et al. published 2 cases of people who similarly died in prone position and hog-tied, but his cases differed in that the deaths occurred during transport to the hospital rather than at the scene of arrest (7).  Ross  published another series of similar deaths in 1996(8).

The early literature strongly supported a role of prone position in death in custody.  After the articles by Reay and O’Halloran, there was an explosion of observational studies and case studies that confirmed the correlation between drug use, “excited delirium”, hog-tying, prone position and death in custody. Parkes replicated Reay’s findings regarding recovery from exercise(9).

In Canada, Pollanen et al. described 21 cases that combined excited delirium, cocaine use, police restraint, and sudden death.  Of these 21 cases, 18 had been placed in prone position.  Pollanen was a little less certain in his conclusions, stating: Restraint may contribute to the death of people in states of excited delirium, and further studies to test this hypothesis are recommended.(10)

This resulted in a number of editorials claiming that prone position was deadly.  However, as always happens in scientific journeys, people started asking some hard questions.  First, it was noted that almost all of these cases involved sympathomimetic intoxication. If prone position were so deadly, why was it deadly only in people taking amphetamines or hallucinogens?  Others noted that most people who resist arrest end up prone while being handcuffed.  That’s thousands and thousands of people.  If prone position was inherently dangerous, then lots of people — and not only people who are delirious — should be dying.  It’s a little like saying “We have a series of 50 people who died in custody and they were all wearing pants.  Pants must be deadly!”  And what’s this about hog-tying?  Sometimes yes, sometimes no.  This problem of conflating drug effect, prone position, hog-tying, etc. without controls raised a lot of questions, but really answered none of them.

So, skeptics started showing up.  One review by Chan et al. criticized the exercise studies noting that measurement methodologies were potentially flawed.   They noted a study (the paper of which I have not been able to read) that did not replicate the exercise stress findings of Reay and Parkes, but instead found no significant effect(11). A more rigorous study by Chan et al. found some changes associated with exercise and prone position, but less than would be important in any practical sense — oxygenation went from normal to low normal, essentially (though this will come up later)(12)

Chan et al. continue(13):

Based on these findings, factors other than body positioning appear to be more important determinants for sudden, unexpected deaths in individuals in the hogtie custody restraint position. Illicit drug use (including sympathomimetic, hallucinogenic, and psychomotor stimulant drugs), physiologic stress, hyperactivity, hyperthermia, catechol hyperstimulation, and trauma from struggle may be more important factors in the deaths of these individuals. Although restraints in general increase the psychological and physiologic stress on the individual, no evidence suggests that body position alone causes hypoventilation, respiratory compromise, or positional asphyxia in the hogtie custody restraint position.

Just to make things more complex, Roeggla et al. looked at the effects of hog tying and prone position.  They put six volunteers in hobble restraints in upright and prone position.  They found(14):

No change was observed in the investigated cardiopulmonary parameters after hobble restraint in the upright position. After hobble restraint in the prone position, mean forced vital capacity decreased by 39.6%, mean forced exspiratory volume by 41%, mean end-tidal carbon dioxide increased by 14.7%, mean heart rate decreased by 21.3%, mean systolic blood pressure decreased by 32.3%, mean diastolic blood pressure decreased by 26.1% and mean cardiac output decreased by 37.4% (P for all reported changes < 0.01).

Roeggla’s work was criticized because of the technical aspects of measuring carbon dioxide using a noninvasive finger probe, which has high variability depending on where the sensor is placed (15)

In the end, Dr. Reay retracted his position. In 1998 he wrote (16):

We concluded that such a prone hog-tied position promoted a delay in respiratory recovery.  This was viewed as potentially harmful because it interfered with respiratory recovery.  More recent studies using arterial blood gas determinations refute our earlier work regarding the hog-tied prone position.

But the battle lines had been drawn.  By 2003, the issue was up in the air.  As one review article noted(17):

The extant literature on restraints does not represent a body of knowledge upon which clinicians can base continued uncritical use of this quasi-therapeutic and security measure. Research is needed to provide clinicians with data on both the risk factors and adverse effects associated with restraint use, as well as data on procedures that will reduce these coercive measures. Research is also needed to determine what individual and combined risk factors contribute to injury and death. Finally, research is needed to determine efficacious and effective alternative measures to restraint.

While this was going on,  people looking to blame prone position raised again the issue of metabolic acidosis — the increase in the acidity of the blood during exercise.  Since struggling against police constitutes exercise, it’s not surprising that people who struggle with police are acidotic(16).  Basically, normal metabolism works by creating acid in the body, mostly carbonic acid, which is dissolved carbon dioxide or other acids such as lactic acid.  Under normal circumstances, we get rid of acid as we make it, either by breathing out the carbon dioxide, or by putting acid in the urine.   If our blood is more acid than it should be, it’s called “acidosis.”  In contrast, if the blood is too alkaline, it’s called “alkalosis.”

If most of the acid is due to carbonic acid, usually because we people can’t breathe enough to get rid of excess carbon dioxide, it’s called “respiratory acidosis.”  When its due to other acids created by ingesting acids (like aspirin), or  breakdown of tissues or energy substrates (producing lactic acid), it’s called “metabolic acidosis.”  Since these other acids are removed through the kidneys, metabolic acidosis can occur in renal failure — and there’s a long list of other things.  During heavy exercise, some metabolic acidosis is normal.  To compensate, we breathe harder, creating a “respiratory alkalosis” that balances it out until we get rid of the metabolic acids.  Generally speaking, because acidosis is a common consequence of activity, we are pretty tolerant of it.  We are less tolerant of alkalosis.  Nonetheless, severe acidosis is deadly.  Normal blood pH is between 7.35 and 7.45.  Prolonged periods of pH below 7.20 carries a mortality of up to 57%.

So, maybe prone position makes acidosis worse not because it *causes* acidosis, but because it makes it harder to do respiratory alkalosis to compensate for the metabolic acidosis produced by struggling.  At least that’s the idea, which Reay raised in a more general sense in the early 1990s and then abandoned.  Hick, et al. note(18):

These cases suggest that a profound metabolic acidosis is associated with cardiovascular collapse following exertion in a restrained position. It is important that public safety personnel recognize the significant life threat when a patient remains combative despite restraints. Avoiding the hobble restraint position and emphasizing side rather than prone positioning may eliminate some of the problems that contribute to the pathophysiology of such deaths.

Since the death of George Floyd, there have been a flurry of articles supporting  metabolic acidosis as the culprit.   Two notable articles are by Weedn and Steinberg.

Weedn et al seem to exhume Dr. Reay’s position and wrap it in new clothes (19):

The prone position restricts ventilation and diminishes pulmonary perfusion. In the setting of a police encounter, metabolic demand will be high from anxiety, stress, excitement, physical struggle, and/or stimulant drugs, leading to metabolic acidosis and requiring significant hyperventilation. Although oxygen levels may be maintained, prolonged restraint in the prone position may result in an inability to adequately blow off CO2, causing blood pCO2 levels to rise rapidly. The uncompensated metabolic acidosis (low pH) will eventually result in loss of myocyte contractility.

What is old is new again.  Sternberg states (20):

Overall, most findings revealed that individuals subjected to physical prone restraint experienced a decrease in ventilation and/or cardiac output (CO) in prone restraint. Metabolic acidosis is noted with increased physical activity, in restraint-associated cardiac arrest and simulated encounters. A decrease in ventilation and CO can significantly worsen acidosis and hemodynamics. Given these findings, deaths associated with prone physical restraint are not the direct result of asphyxia but are due to cardiac arrest secondary to metabolic acidosis compounded by inadequate ventilation and reduced CO.

The Resurrection Men bringing back Reay’s ideas sound pretty good, except for one big problem. These are postulated explanations for increased risk were it to exist, but it is not demonstration that the risk does in fact exist.  There is a great danger in “reverse engineering” a risk like this.  Sometimes it’s very seductive to have a great theory for something and then use that theory to “prove” that it’s the case.  First you have to show that there is, in fact a greater risk of injury with prone position.  *Then* you can attempt to show that your theory is correct in explaining why it is more risky.

And that’s where the metabolic acidosis theory starts to fail.  While there are retrospective studies that demonstrate that subjects have died in prone position, there have been no studies that demonstrate that they are *more likely* to die in prone position.

Let’s say, for instance, that I see a bunch of people who die while taking cold showers.  I suspect that cold showers are a problem, so I collect some of those cases and publish a series “Five cases of death in cold showers, a single institutional retrospective study.”  I propose a theory that taking cold showers will lead to a massive increase in sudden cardiac death.  I postulate that this will happen because of the autonomic effect of cold water and the vascular changes associated with cold skin.  It sounds good, and is consistent with other autonomic trigger literature.  Some other offices read my article and start collecting cases and suddenly there are five small series of people dying taking cold showers.   People start making the diagnosis of death by cold shower.  This increase in the diagnosis of cold shower deaths in turn results in articles that “demonstrate” the epidemic of cold shower death (this is so-called “diagnostic drift”). The government decides to regulate how cold showers can be.

But what if that’s not the case?  What if you took a hundred thousand people, put 50,000 in warm showers and 50,000 in cold showers and found out that 100 people died in the warm shower and 50 people died in the cold shower? You have a great explanation for why something *might* be a risk, but the prospective study shows that it isn’t really such a risk at all.

An example of this, by the way, is Finnish sauna.  In the 1970s, people noted that Finns liked saunas *and* Finns had an elevated rate of sudden cardiac death.  Thus, it was reasoned, Finnish sauna bathing probably increased the risk of sudden death.  One study from that time noted (21):

In addition, this study indicates that sudden exposure to intense heat, as in sauna baths, Turkish baths, and presumably the “coffee baths” becoming fashionable in Japan, may contain an element of risk in those prone to coronary insufficiency. This group includes a number of middle-aged men with tiredness, lethargy, and “fibrositis” in the chest and shoulder who will seek out the most gruelling of these artificial environments and stay in longest to “sweat it out of their system.” Apart from descriptions such as “my heart was thumping” and “I thought my heart was going to stop” appearing in recent articles on the subject (Observer, 1971), sudden syncopal attacks just on standing to leave the sauna are not uncommon.

Now, 50 years later, the dominant belief is the opposite — that Finnish saunas have a protective effect.  As an aside, Turkish saunas are a different matter.

In order to see whether or not prone position is more dangerous than other positions, one needs to do a prospective study where the positions are recorded in all cases.  If prone position is more dangerous, then more people will die in prone position.

Well, there have been multiple such prospective studies.  None of them show increased deaths due to prone position.  As Vilke notes in his review (22):

Four studies evaluated real world police use of force events that resulted in prone positioning.  Hall et al. reviewed restraint and police use of force events and in 1255 subjects and reported that 42.8% were left in a prone position and none of the subjects died. Hall et al. subsequently reviewed restraint and police use of force events and in 4828 use of force events, with over 2000 of the subjects were restrained in a prone position. There was only one death in this study population, and that subject was not in a prone position. Ross and Hazlett reported that in 110,173 arrests, 1085 incidents resulted in prone positioning of the subject. In this study population, they reported no deaths. Lasoff et al. reported that in 2431 use of force incidents, 63.1% ended up being placed in a prone restraint position and no fatalities were noted.

The studies referred to are in the reference list (23-26).

This is not to say that there do not exist people who are sufficiently vulnerable so that they would die if placed in prone position.  There is a particular danger in obese people with severe cardiopulmonary disease, and in people with chronic obstructive pulmonary disease(27).   In this particular subset of vulnerable people, the response to prone position depends on their particular vulnerability.  As one author notes regarding prone restraint in chronic obstructive pulmonary disease(28):

However, three subjects could not tolerate the prone position due to a clinical deterioration in symptoms. The response to the prone position with or without wrist restraint appears highly individual, with some individuals tolerating the prone position with no measurable clinical effects and others suffering a clinical deterioration in symptoms. The reasons for this individual variation remain unclear.

But there is essentially no evidence that prone position, per se, is dangerous in your average person. In the trial of Officer Chauvin the prosecution expert witnesses, as I remember, were in the odd position of claiming both that the studies that show little danger don’t count because they were performed on healthy people *and* that prone position was so dangerous that even healthy people are at risk.  The literature simply doesn’t support this.

Are there metabolic changes associated with prone position?  Certainly.  Are they life-threatening?  Not in people who are not particularly vulnerable.  Like the other things I’ve written about so far, it certainly represents yet another stressor. It is, I believe, dangerous, however, to load multifactorial deaths onto one causative agent — because eliminating just that agent will not save lives.  I’ll write about that at the end of all these posts.


  1. Reay, DT, Howard, JD, Fligner, CL, Ward, RJ. Effects of positional restraint on oxygen saturation and heart rate following exercise. Am J Forensic Med Pathol 1988;9:16–8.
  2. Reay DT, Fligner CL, Stilwell AD, Arnold J. Positional asphyxia during law enforcement transport. Am J Forensic Med Pathol 1992;13(2):90-97
  3. Laposata, Elizabeth A. M.D. Positional Asphyxia during Law Enforcement Transport, The American Journal of Forensic Medicine and Pathology: March 1993 – Volume 14 – Issue 1 – p 86
  4. O’Halloran R, Lewman L. Restraint asphyxiation in excited delirium.  Am J Forens Med Pathol.  1993 14(4):289-295
  5. O’Halloran R, Frank J. Asphyxial death during prone restraint revisited.  Am J Forens Med Pathol 2000 21(1):39-52.
  6. Bell MD, Rao VJ, Wetli CV, Rodriguez RN. Positional asphyxiation in adults: a series of 30 cases from the Dade and Broward county Florida medical examiner offices from 1982-1990. Am J Forensic Med Pathol 1992;13:101
  7. Stratton S J, Rogers C, Green K: Sudden death in individuals in hobble restraints during paramedic transport. Ann Emerg Med May 1995;25:710-712.
  8. Ross DL. An analysis of in-custody deaths and positional asphyxiation. Police Marksman 1996;March/April:16
  9. Parkes J. Sudden Death during Restraint: A Study to Measure the Effect of Restraint Positions on the Rate of Recovery from Exercise. Medicine, Science and the Law. 2000;40(1):39-44.
  10. M. S. PollanenD. A. ChiassonT. J. CairnsJ. G. Young Unexpected death related to restraint for excited delirium: a retrospective study of deaths in police custody and in the community
  11. Schmidt MA, Snowden T, Clin J. The effects on oxygen saturation and heart rate of persons due to positional restraint [unpublished report]. San Diego: San Diego Regional Public Safety Training Institute, 1996.
  12. Chan TC, Vilke GM, Neuman T, Clausen JL. Restraint position and positional asphyxia. Ann Emerg Med 1997;30:578.
  13. Chan, Theodore C. M.D.; Vilke, Gary M. M.D.; Neuman, Tom M.D. Reexamination of Custody Restraint Position and Positional Asphyxia, The American Journal of Forensic Medicine and Pathology: September 1998 – Volume 19 – Issue 3 – p 201-205
  14. Roeggla, M., A. Wagner, M. Muellner, A. Bur, H. Roeggla, M.M. Hirschi and G. Roeggla (1997). Cardiorespiratory consequences to hobble restraint. Weiner Klinische Wochenschrift, 109/10: 359-361.
  15. Savaser, Davut J., et al. “The effect of the prone maximal restraint position with and without weight force on cardiac output and other hemodynamic measures.” Journal of Forensic and Legal Medicine 20.8 (2013): 991-995.
  16. Reay, Donald T. “Death in custody.” Clinics in laboratory medicine 18.1 (1998): 1-22.
  17. Mohr WK, Petti TA, Mohr BD. Adverse Effects Associated with Physical Restraint. The Canadian Journal of Psychiatry. 2003;48(5):330-337. doi:10.1177/070674370304800509
  18. Hick, J.L., Smith, S.W. and Lynch, M.T. (1999), Metabolic Acidosis in Restraint-associated Cardiac Arrest: A Case Series. Academic Emergency Medicine, 6: 239-243.
  19. Weedn, Victor, Alon Steinberg, and Pete Speth. “Prone restraint cardiac arrest in in-custody and arrest-related deaths.” Journal of forensic sciences 67.5 (2022): 1899-1914.
  20. Steinberg A. Prone restraint cardiac arrest: A comprehensive review of the scientific literature and an explanation of the physiology. Med Sci Law. 2021 Jul;61(3):215-226. doi: 10.1177/0025802420988370. Epub 2021 Feb 25. PMID: 33629624.
  21. Taggart, Peter, Peter Parkinson, and Malcolm Carruthers. “Cardiac responses to thermal, physical, and emotional stress.” Br Med J 3.5818 (1972): 71-76.
  22. Vilke, Gary M. “Restraint physiology: A review of the literature.” Journal of forensic and legal medicine 75 (2020): 102056.
  23. C.A. Hall, A.M. McHale, A.S. Kader, L.C. Stewart, C.S. MacCarthy, G.H. Fick Incidence and outcome of prone positioning following police use of force in a prospective, consecutive cohort of subjects J Forensic Leg Med, 19 (2) (2012), pp. 83-89
  24. C. Hall, K. Votova, C. Heyd, et al. Restraint in police use of force events: examining sudden in custody death for prone and not-prone positions J Forensic Leg Med, 31 (2015), pp. 29-35
  25. D.L. Ross, M.H. Hazlett A prospective analysis of the outcomes of violent prone restraint incidents in policing
    Forens Research Crim Int J, 2 (1) (2016), pp. 1-10
  26. D. Lasoff, C.A. Hall, W.P. Bozeman, T.C. Chan, E.M. Castillo, G.M. Vilke Proning: outcomes of use of force followed with prone restraint J Forensic Med, 2 (2) (2017), pp. 1-3
  27.  Palmon SC, Kirsch JR, Depper JA, Toung TJ: The effect of the prone position on pulmonary mechanics is frame-dependent.Anesth Analg 1998;87:1175– 80.
  28. Meredith, C., Taslaq, S., Kon, O. M., & Henry, J. (2005). The cardiopulmonary effects of physical restraint in subjects with chronic obstructive pulmonary disease. Journal of Clinical Forensic Medicine, 12(3), 133–136. doi:10.1016/j.jcfm.2004.10.013





Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.