Friday, April 10, 2015

Epidemiology and Toxicology of Aircraft Assisted Pilot Suicides

I thought I would add a few facts to the speculation about what is really known about the epidemiology and toxicology involved in aircraft assisted suicides. It turns out that there are substantial studies that have been written.  If you are a bottom-line kind of person and want to avoid further reading, I can tell you that the events are rare especially events involving commercial aircraft where the incident is ruled a suicide by aviation authorities.  The events are so rare that prediction is doubtful.  In many cases the descriptions of suicidal statements and behavior occur on the day of the events and there are further extenuating circumstances like the use of alcohol and other intoxicants.  If you are really interested in these events, there are numerous places where you can see the analysis of what happened and what the ruling was by the National Transportation Safety Board (NTSB). 

The media reaction is similar to what is seen following mass shootings in the United States.  After the initial shock, there is typically a period of speculation about the causes of the disaster of the form: “What motivates a person to do something like this?”  There is the invariable dissection of their life in the media.  Were they bullied?  What was their personality like? What was on their computer?  Were there any clues that were missed that suggested that one day they would start shooting people?  Were psychiatrists involved?  How did they get the firearms?  When all of those familiar touchstones are exhausted (and it does not take long), the analysis starts to take on the characteristics of groups with agendas.  Gun advocates will suggest that this person was not a typical gun owner and therefore tighter gun laws are not needed.  Gun control advocates will provide the counter arguments that usually involve how easy it was for this person to get a gun.  There is a political impasse largely due to the power of the gun lobby and some politicians start to talk about “being in the wrong place at the wrong time.”  Mental health advocates, especially anyone who wants to talk about the real problems of mental illness and violence are as disenfranchised as the gun control advocates.  Nothing ever happens.  The screening advocates step up and suggest that many of these incidents could be prevented if we just “screened” enough people.  Anyone familiar with Bayesian statistics knows why that won’t work and may cause more harm than good. 

After that impasse, a second wave of speculation starts driven largely by people who ascribe to the theory that psychiatric medications and psychiatric treatment can cause homicidal behavior.  There are a couple of schools of thought on that one.  The first has to do with medications and the idea that specific medications like SSRIs can lead to homicidal behavior.  The other has to do with the fact that seeing a psychiatrist is associated with homicidal behavior and therefore psychiatric treatment must at some level cause homicidal behavior or at the very least the psychiatrist is responsible for not stopping it.  As I explored in a previous posts – there is not a shred of evidence that any of that is true.  There is however more evidence about pilot safety, pilot use of antidepressants, and incidents ruled pilot suicide than I have seen discussed in the media.  Here are a few bits of solid data to ponder during the expected swell of speculation about causes, who is to blame, and possible solutions.

1.  The denominator is huge:  

When the FAA or NTSB looks at all certified pilots in the US that includes a total of roughly 620,000 people per year including classifications for student, recreational, sport, private and commercial.  Roughly 1/3 of the FAA certified pilots are classified as commercial.  The US government also collects detailed statistics on the total number of passengers flown per year (815.3 million), the total number of flights per year (9.821 million) and a host of associated statistics on the Bureau of Transportation Statistics web site. 

2.  The numerator is very small:  

A quick glance at the table below on either antidepressant use by pilots or the total incidents rules as suicide shows that a small proportion of the total deaths are associated with either suicide or antidepressant use.  The proportions of the total pilots in the data base is much smaller and the rates of both suicide and antidepressant use are much lower than expected on a population wide basis.  Data from the Aviation Safety Network suggests that there were 8 to 10 incidents involving commercial aircraft and pilots since 1976 or about 9 in the last 40 years.

3.  The data on pilot use of antidepressants in fatal crashes: 

 Until about 2006, the FAA prohibited the use of antidepressants by commercial pilots.  They have since modified their stance to allow for specific antidepressants.  The European Aviation Safety Administration has publicly posted information of the safety of pilots and necessary screening for psychiatric disorders as well as prohibitions on certain diagnoses.  There have been studies that look at positive toxicology for antidepressants in the cases of fatally injured pilots.  These studies have looked for the presence of tricyclic antidepressants (TCAs) and selective serotonin reuptake inhibitors (SSRIs) in in fatal crashes.  Tricyclic antidepressants were the predominant antidepressants prescribed before the approval and release of fluoxetine in 1987.  One study by Dulkadir, et al looked at fatal crashes between 1990 and 2012.  In this study the researchers received biological samples from 7,037 fatally injured pilots out of a total of 8,429 fatal accidents.  2,664 were positive for drugs on toxicological analysis.  Of those positive samples TCAs were found in 31 samples, TCAs alone in 9 and TCAs with other drugs in 22.  None of the pilots involved reported TCA use during their aviation medical exam.  The authors point out that at the time covered in this study that TCAs were not approved for pilot use and that selective serotonin reuptake inhibitor antidepressants or SSRIs were approved on a case by case basis.  That is a prevalence of TCA use in this database is less than 0.5% ( 31/7,037 aviators).  That number is much lower than estimates of population wide use of antidepressants.

Where the blood levels were determined they clearly indicate that some overdoses had occurred (see Table II and III).  Blood concentrations greater that 1,000 ng/ml are usually very consistent with overdoses and that is the case with nortriptyline and imipramine/desipramine in these tables.  The authors were able to determine that the TCAs were prescribed for depression in three cases, pain in two cases, and chronic insomnia in one case.  Other antidepressants were listed along with opioids, anticonvulsants, cold medications, antihypertensives, benzodiazepines, muscle relaxants, diabetes medications and ethanol were detected but the epidemiology was not reported.  In both the studies by Akin and Dulkadir “drugs and alcohol and/or a medical condition” was given as “a probable cause or contributing factor in about 1/3 of the accidents where antidepressants were detected.

There was an earlier study of the epidemiology of SSRIs in pilot fatalities from 1990-2001 (Akin, et al) that showed they were involved in 61/4,128 pilot fatalities or a total of 1.48%.

The available data suggests that pilot suicide by aircraft is very rare and much lower than the pilot suicide rate by all methods.  There is also a suggestion that the suicide rate in pilots has actually decreased.  Searching the NTSB database yielded 74 fatal accidents using the search term "suicide" dating back to 1966. 

Explanations given in the article for the fewer pilots taking TCAs was that they are more toxic and less preferred agents.  Certainly in the 1990s SSRIs were heavily promoted along with the medical treatment of depression.

4.  Intoxicants are found in toxicology specimens –

The study by Canfield, et al identified a greater percentage of specimens that were positive for cannabinoids (relative to antidepressants) and additional performance impairing drugs in 38% of the individuals who tested positive to cannabinoids.  They also looked at the mean THC concentration in the blood and concluded that during 1997-2001 it was 2.7 ng/ml and for 2002-2006 it was 7.2 ng/ml.  The rate of increase in THC levels over those years exceeded the increase in cannabis potency as reported by the National Institute of Drug Use (NIDA) over the same years (2.7 fold as opposed to 1.5 fold).  Some authors have concluded that THC levels between 2 and 5 ng/ml represent the lower and upper ranges of significant impairment from cannabis use on performance tests measuring driving skill (see Ramaekers, et al) in recreational cannabis users.

The study by Bills, et al looked at the toxicology in a cohort of 36 pilots who committed suicide by aircraft during a 21 year period from 1983 and 2003.  Each suicide case was matched against 2 randomly selected control accidents.  In this study, the pilot characteristics included positive toxicology for alcohol, prescription drugs, and illegal drugs in 24.3%, 21.6%, and 13.5% of cases respectively.  An exhaustive list of drugs found was not available in the paper.  The authors were also not able to compare the toxicology of the cases to controls because 84% of the controls survived and their toxicology was unknown.  

5.  The baseline rate of pilot suicide is low or is it? -

Bialik looked at the issue of workplace suicide, the data quality estimates for pilots in the US.  One of the key references was a paper by Tiesman, et al that looked at the issue of workers who suicide in the workplace.  It used databases from the CDC (National Occupational Mortality Surveillance (NOMS)) and  Bureau of Labor Statistics (Census of Fatal Occupational Injury (CFOI)).  The NOMS database has no granularity and does given intentional self harm as a search parameter.  Unfortunately only "transportation occupations"  can be searched grouped by age, race, and sex.  I did not find the number of deaths or the PMR (Proportionate Mortality Ratio) to be useful.  The NOMS did have granularity with specific occupations and there was a homicide definition but none for suicide or intentional self harm.  Bialik concludes that pilots in general may have a slightly higher rate of suicide than the population in general but there are problems with that estimate and he was able to consult with an epidemiologist from the CDC.

Another approach to looking at this issue to to find a study with a very well characterized database that looks at the occupational issue.   Roberts, et al meets that criterion in a 2013 study of high-risk occupations for suicide.  The researchers looked at the numbers of suicides and numbers in all occupations in England and Wales for specific time intervals.  They determined the 30 occupations with the highest suicide rates (generally greater than 20/100,000).  In comparing the time intervals (1979–80, 1982–83) to  (2001–2005) they determined shift in the ranking and discussed possible causes of those changes.  Pilots were not listed in the top 30 occupations by suicide rate.  The only transportation workers listed were "rail transport operating staff".  They noted that suicide rates for professional occupations decreased over the time interval studied while there were sharp increases in the suicide rates for manual occupations.  As a comparison the 2013 suicide rate in the US was 12.6 per 100,000.

6.   The accident rate due to suicide attempts in commercial aviation is lower than that found in general aviation - 

These incidents are tracked  by the Aviation Safety Network and their web site currently lists intentional incidents and accidents caused by pilots dating back to 1976 in commercial flights.   There is a separate list of aircraft accidents caused by pilot suicide and that lists 9 suicides in the same time period but proportionally more associated fatalities. 

7.  Pilots can already self report substance use problems - 

There have been some suggestions that screening would be enhanced if pilots could self report problems without the fear of recrimination - the same way that licensed health care professionals are allowed to do in many states.  The focus would be on treatment rather than punishment.  The health care professional experience demonstrates that this leads to significantly more self reports and that is consistent with the goal of public safety.  Since pilot certification occurs at the federal level and health care professional licensing occurs at the state level - there is an opportunity to develop a more standardized approach to the potentially compromised pilot that depends more on self-report than screening.  There is currently an "occupational substance abuse treatment program" called HIMS that states at least part of their goal is to preserve careers.  A broader focus to include voluntary self- report of psychiatric conditions and suicidal thinking would result in more referrals for treatment and potentially impact the suicide rate.

8.  Aviation regulators and the aviation industry collect data that the healthcare industry can only marvel at - 

Reading through the sheer amount of data and how it is acquired it is evident that anyone involved in aviation has a single-minded focus on safety.  The methods of data acquisition through flight recorders and the checks and balances on the ground are far superior to any safety standards in the health care industry in the United States.  As a basic thought experiment, can you imagine recording similar outcome data from patients rapidly discharged from hospitals in the US?  I am talking about real data and not the survey that the nurse hands a patient after they have coached them on what to check off. 

I don't have to imagine what that data would look like.  I know what that data looks like and it is quite ugly.  It is more than a little ironic that health care experts, especially in this case psychiatrists and other behavioral experts are going to rush in and correct what is wrong with the aviation industry.  By comparison, health care measurement and incident analysis is all smoke and mirrors.  They don't know how to collect relevant data and many of the outcome measures are strictly political and meaningless.  If anything we should be bringing in aviation safety experts to run hospitals instead of MBAs.

With what I have read, I doubt that there is any possible improvement beyond voluntary reporting and making sure that there is always a second crew member in the cabin on commercial airliners.  In some of the commercial aircraft crashes the planes were stolen by staff who were not pilots and crashed.  But in the case of air disasters that resulted in multiple passenger deaths a second person in the cabin is a clear safeguard.  I am not an expert on how many people are in air crews, but I know that there is also a flight engineer in the cabin in some cases.  Given that these incidents are rare by any combination of numerators and denominators that are chosen and the fact that screening for rare events is generally not successful, screening for these rare events is not likely to work.  Flight crews currently undergo random urine toxicology to prevent the use of intoxicants that can impair the ability of a pilot.  Anecdotal evidence would suggest that is useful, but in the case of addictions there are often attempts to circumvent this intervention or use a drug that is not detectable.  The experience of health care professional screening programs would suggest that voluntary reporting can both improve public safety and preserve careers.  That seems like a useful approach for pilots.

Most importantly, the aviation industry is a model for safety assurance and the investigation of incidents where there were lapses.  It holds many lessons for the health care industry.        

George Dawson, MD, DFAPA

Akin A, Chaturvedi AK. Selective serotonin reuptake inhibitors in pilot fatalities of civil aviation accidents, 1990-2001. Aviat Space Environ Med 2003; 74(11):1169–76

Canfield DV, Dubowski KM, Whinnery JE, Lewis RJ, Ritter RM, Rogers PB.  Increased cannabinoids concentrations found in specimens from fatal aviation accidents between 1997 and 2006. Forensic Sci Int. 2010 Apr 15;197(1-3):85-8. doi: 10.1016/j.forsciint.2009.12.060. Epub 2010 Jan 13. PubMed PMID: 20074884.

Zeki Dulkadir,  G├╝lhane, Arvind K. Chaturvedi, Kristi J. Craft, Jeffery S. Hickerson, Kacey D. Cliburn. Antidepressants Found in Pilots Fatally Injured in Civil Aviation Accidents.  Federal Aviation Administration, Office of Aerospace Medicine, Nov 2014.

Lewis RJ, Johnson RD, Whinnery JE, Forster EM. Aircraft-assisted pilot suicides in the United States, 1993-2002. Arch Suicide Res. 2007;11(2):149-61. PubMed PMID: 17453693.

Russell J. Lewis, Estrella M. Forster, James E. Whinnery, Nicholas L.  Webster.  Aircraft-Assisted Pilot Suicides
in the United States, 2003-2012  Civil Aerospace Medical InstituteFederal Aviation Administration. Oklahoma City, OK 73125
February 2014

Ungs TJ. Suicide by use of aircraft in the United States, 1979-1989. Aviat Space Environ Med. 1994 Oct;65(10 Pt 1):953-6. PubMed PMID: 7832739.

Bills CB, Grabowski JG, Li G.  Suicide by aircraft: a comparative analysis.  Aviat Space Environ Med. 2005 Aug;76(8):715-9. PubMed PMID: 16110685.

Ramaekers JG, Moeller MR, van Ruitenbeek P, Theunissen EL, Schneider E, Kauert G. Cognition and motor control as a function of Delta9-THC concentration in serum and oral fluid: limits of impairment.  Drug Alcohol Depend. 2006 Nov 8;85(2):114-22. Epub 2006 May 24. PubMed PMID: 16723194.

Roberts SE, Jaremin B, Lloyd K. High-risk occupations for suicide. Psychol Med. 2013 Jun;43(6):1231-40. doi: 10.1017/S0033291712002024. Epub 2012 Oct 26. PubMed PMID: 23098158; PubMed Central PMCID: PMC3642721.

Total FAA Certified Pilots:

Aviation x Antidepressant Medline Search April 2015:

Carl Bialik. We Don't Know How Often Pilots Commit Suicide.  FiveThirtyEight (a very sophisticated blog)


  1. Agree on two people in the cabin.

    Also, I wasn't able to find a great deal of information on it, but what do you think of behavioral observation programs like the nuclear power industry?

    Knowing his girlfriend has stated she observed some issues with him, I wonder why she didn't report it. Or did I miss that she had?

  2. I think a better approach is a team approach like the one I employed on inpatient psychiatric units. Regular meetings, esprit de corps, and the understanding that authoritarian approaches have major limitations. Knowing and agreeing why random tox screens need to be done is different than feeling like they are arbitrary and an invasion of privacy. Knowing that your supervisor has an eye toward your safety and career as well as the regulations is also helpful.

    I think behavioral observations can be useful but they need to be balanced with a knowledge of boundaries is a supervisor who has a lot of self awareness.

  3. Eventually the copilot or at least a 2nd copilot will not be on the plane but in an office building monitoring the passenger craft like a drone. So when that plane descends to 20k over the Alps, they take over including the cockpit door.

    In Japan drones have all but eliminated crop dusters. It's coming to commercial craft although for many reasons I think there will always be two pilots.

    1. It will probably not be too long before planes are crash proof in the typical sense. For example when the low altitude warning comes on with instruction to pull up - the plane does it automatically instead of waiting for the pilot and if the pilot or pilots are incapacitated they will be landed remotely.

      But I think the stats show a very high degree of safety given what could happen and the fallibility of humans.

  4. Let us all pat ourselves on the back for the above comments....this article came out today: