March 15, 2011

Mortality on Mount Everest, 1921-2006: A Descriptive Study

Tuesday, January 20, 2009
(Click here for Research Paper)
Re: 1. Firth PG, Zheng, H, Windsor JS, Sutherland AS, Imray CH, Moore GWK, Semple JL, Roach RC, Salisbury RA. Mortality on Mount Everest, 1921-2006: descriptive study. BMJ 2008; 337: a2654
2. Supplemental Data.

I was asked by a member of the IMF to submit a brief discussion of a recent medical study performed about deaths on Mount Everest. Some years ago I was in Delhi and was pleased to be able to visit the headquarters of the IMF. I was impressed with the library and the IMF commitment to helping climbers from around the world. I also appreciated the extensive information that was available, data that can help to improve the safety of climbers. In my early years I was a member of the Mountain Club of South Africa; now that I live in Boston I am now a member of the American Alpine Club. I think climbing associations are vital to further the aims of mountaineering, so I am pleased to provide some information to further the IMF’s goals and help improve the safety and enjoyment of climbers.

The original article may be downloaded without cost from the British Medical Journal (BMJ) website, or accessed via the IMF website which has a link to the article. Additional information, which includes more a detailed plot of deaths, deaths at and below base camp, the details and importance of summit times, and a further informal discussion, is available for free online at
What did this study examine, and how was it done?
The study set out to examine the circumstances of all deaths on Everest expeditions between 1921 and 2006. We identified all deaths on Everest expeditions by searching the Himalayan Database and expedition books. We then located accounts of these deaths by examining reports in the Database, expedition books, climbing journals, or on the Internet, or by directly contact with witnesses. Four doctors, all of whom had climbed on Everest, classified the deaths using a descriptive classification system. Three had summited Everest; one had turned back at 8300m. All four reviewers had a particular interest in high altitude physiology, and had practical experience in managing high altitude illness. The classifications were then pooled and differences were resolved by consensus. The degree of disagreement between the independent assessments was measured to give an indication of the reliability or certainty of the final classification system. The final results can be found on the website of the Himalayadatabase and in the BMJ article.

We then looked at specific subsets of events. As most mountaineers are interested in what happens higher on the mountain we focused on events above base camps, defined as the last encampment before technical (roped) climbing began. As the majority of summits were via the ‘standard’ Southeast and North ridge routes, during the spring season, and in the last 25 years, we also examined this group in detail. This minimized variables such as season, the differing difficulty of the technical routes, and the changes in expedition styles that have occurred over the decades. We looked at the specific circumstances of the deaths of those that died after climbing above 8000m.
To examine the effect of weather, we used measures of barometric pressure at 9000m obtained from a dataset at the National Center for Environmental Prediction. This has been used previously in studies of weather conditions on Everest.

What were the most significant findings?
For the entire study period, the mortality rate of mountaineers above base camp was 1.3%. Among climbers it was 1.6%, among sherpas it was 1.1%.
Deaths could be classified as involving trauma (objective hazards or falls), as non-traumatic (high altitude illness, hypothermia or sudden death) or as a disappearance (unwitnessed death, body not found). During the spring seasons on the standard routes, most climbers died above 8000m during a summit bid. The death rate on the north was 3.4%; on the south side it was 2.5%. Most summits occurred during good weather windows, and most deaths also occurred during fair weather. When storms played a role in deaths, many climbers tended to die on a single day; however the majority of deaths occurred on days when weather had not markedly deteriorated.
Most mountaineers who died above 8000m died during descent from the summit. Climbers died at a much higher rate during the descent than sherpas. Late summit times were associated with subsequent death. Many developed symptoms suggestive of high altitude cerebral edema (HACE): confusion, loss of consciousness, and a staggering gait. Symptoms of high altitude pulmonary edema (HAPE) were rare in those that died. An common early sign was marked fatigue, as reflected in reports that the mountaineer looked exhausted, the tendency to fall behind other climbers on the party, and a late summit time.
So: Most climbers die during summit attempts, when they are climbing above 8000m. On well-established routes, with relatively few technical difficulties, with most climbers climbing during good weather windows, many of the climbers who died developed confusion and inco-ordination. These symptoms often became debilitating during the descent. Slower climbing speeds and late summit times were early signs associated with subsequent deaths.

While obviously technical difficulties, bad weather, lack of protective ropes, difficulty of rescue and breakdown in teamwork all play roles in deaths at extreme altitude, a significant fatal effect of very high altitude therefore appears to impairment of thinking and co-ordination. Excessive fatigue appears to be an early warning sign.

How did people die above 8000m?
We classified deaths by a descriptive technique. Of 94 deaths of mountaineers who reached 8000m, we classified deaths as those that involved a fall (34%), disappeared (29%), high altitude illness (11%), suddenly death (5%), hypothermia (2%) or we couldn’t decide exactly a single classification (15%) – usually between HACE or hypothermia. We used this technique since it allowed us to classify a very varied series of accounts, of a very varied set of circumstances, in an objective and reproducible way that minimized our own interpretation or bias. From this foundation, we could then examine the deaths in more detail and try to look for factors we speculated might have caused the fatalities. This was therefore a descriptive classification, not necessarily one that demonstrates the underlying ‘cause’ of death
Of these deaths, a large number involved climbers that developed neurological problems before falling or disappearing – confusion or staggering gait. This would suggest that neurological problems were an underlying ‘cause’ of many of these deaths.

How reliable is this classification system?
Since an observer examining an account might read different things into what happened, we used four reviewers, all of who examined the accounts independently of one another. For all 212 deaths during Everest expeditions, there was an initial unanimous agreement for 165 (78%) of classifications. This number does not give a full measure of agreement – if, for example, only one reviewer disagreed on the remainder, this would be less disagreement than if all four reviewers came up with four completely different classifications. So we used a statistical test called the Fleiss Kappa test that measure the degree of inter-rater agreement. The kappa value was 0.63, which implies ‘substantial agreement.’ It is impossible to accurately classify all deaths due to the variability in accounts, the differing circumstances, and the possibility of unknown causes. If one has a classification system that gives 100% agreement, this would probably be too insensitive a reflection of complex events. We felt the measure of agreement reflects the best balance between agreement and uncertainty in a classification system of an uncertain subject matter.

What causes the confusion and loss of co-ordination?
We felt that these symptoms were consistent with cerebral edema (HACE), caused by the low oxygen atmospheric content at extreme altitude. HACE is due to inadequate acclimatization to the low oxygen levels, which results in the blood vessels in the brain leaking fluid into the surrounding brain tissue (edema). Confusion and loss of co-ordination follow. Direct deterioration of the brain’s functioning may also occur, although relatively little is known about this at present.
Hypothermia or extreme cold can also cause these symptoms. In some cases we felt the cold was the primary cause of the symptoms, in others we concluded it was due to HACE, and in some cases we couldn’t determine which was the primary cause. A climber who is incapacitated by HACE will easily develop hypothermia in the extreme cold above 8000m. However we felt it was unlikely that hypothermia was the sole or primary cause of neurological deterioration in most cases.
We also looked for signs of visual disturbances (snow blindness or retinal hemorrhages) that might masquerade as confusion or uncoordinated gait. Although we know that these are significant problems at extreme altitude, we didn’t find evidence to suggest that visual problems were a common primary explanation for the symptoms. We didn’t think the symptoms could be explained by visual disruption alone.
Other physical features associated with extreme altitude, such as high ionic radiation, are not know to cause these effects in the relevant time frame.
We therefore often interpreted the described symptoms as evidence of fluid leaking from blood vessels in the brain, due to inadequate acclimatization to the low oxygen availability at extreme altitude.

Couldn’t the deaths and neurological symptoms on descent simply be due to climbers running out of oxygen?
Running out of supplemental oxygen probably doesn’t help the functioning of the brain. However, we saw similar patterns of mortality in climbers who died while climbing without supplemental oxygen – i.e. most of these climbers died during the descent. This suggests the problems cannot simply be explained by loss of oxygen. In addition, it is not uncommon for climbers to have failures of their delivery sets while climbing up. However we didn’t detect any deaths due to this. We interpreted this as implying that mountaineers developed some problem that heightened their susceptibility to loss of supplemental oxygen. This might include sub-clinical HAPE and/or HACE that is not obvious when oxygen-enriched air is used, but which becomes obvious when the supplemental oxygen runs out.

So is HAPE rare at extreme altitude?
This study suggests that it is unusual for non-survivors to develop florid HAPE. This may be because mountaineers who get HAPE turn back early and survive, because those with severe HAPE are still able to get down and live, because those prone to HAPE get it lower down the mountain, or because the use of supplemental oxygen (which lowers the blood pressure in the lungs) alters the response to altitude. Another caveat is that this study is the first to examine the problems in the ‘death zone’ in detail – perhaps other future studies may have other results. However we detected very few deaths that clearly involved HAPE. So although HAPE may occur at extreme altitude, it was rarely a clear or obvious factor in fatalities.

So how do you tell if you are getting HACE?
Unfortunately once you get HACE you are in severe trouble – your insight is impaired, since you are confused and may not realize what is happening. As you are confused and uncoordinated, it is difficult for other mountaineers to rescue you since they have to physically haul you down the mountain. In the absence of a large team, this is hard to do (but not impossible – organized teams can get even severely impaired climbers down to lower altitude where they can make good recoveries).
The earliest sign seems to be marked fatigue – sense of extreme exhaustion, falling behind the group, and making the summit later in the day.

How late is late? Everyone is exhausted at 8000m – how fast do the individual need to climb?
Unfortunately we don’t have records of what time mountaineers left summit camps at, or what camp they left from on the North Side. There is also no record of the exact oxygen flow rates that climbers used, which will impact on their climbing speed. Almost all survivors make the summit by around noon to very early afternoon. However assuming most leave around 10pm - midnight, this means a maximum of about 12 –14 hours. The risk of death associated with slower climbing speeds and later summit times goes up dramatically after this. Graphs of summits times of survivors and non-survivors are available at

In recent years large numbers of climbers make summit bids on the same days. Measuring ascent speed against other mountaineers is another yardstick. Mountaineers who died tended to fall behind the others in their team.

What about other symptoms?
Other early signs of HACE include nausea, vomiting and headache. These symptoms seemed rare in those that died. The precise mechanisms that lead to these symptoms are unknown – one theory is that they may be related to stretching of the sensitive membranes covering the brain as it swells. We speculate that at extreme altitude, the brain does not swell as much before debilitating confusion and loss of coordination set in, so headache and nausea are not reliable early warning signs. However perhaps witnesses simply didn’t report these events, or those mountaineers with these symptoms turned back earlier and survived. Further study will be needed to confirm this.
NOTE: Although this is speculation, it seems that fatal cases of HACE are not reliably preceded by headache or vomiting. Although one might assume that climbers who die are simply victims of their own stubbornness in not turning back earlier, this may not be the case. Rather they mistake the early warning sign (severe exhaustion) for the expected fatigue experienced while climbing at extreme altitude. Warning signs such as headache or nausea may not occur. As HACE sets in, they become confused and lack insight into their condition. Although this study has not proved this (it is very difficult to demonstrate the definite absence of something), this speculation should be considered by climbers at these altitudes.
Couldn’t the late summit time-death association simply be explained by more time at risk of being caught in bad weather?
We were somewhat surprised to see how rarely bad weather played a role in deaths at very high altitudes during the spring. We were able to use weather data collected over years to look at an association of deaths with changes in atmospheric conditions. The most useful marker was barometric pressure, although other weather markers have also been collected. Most summit bids occur in good weather – and most days when deaths occur have reasonable weather too. When the weather does turn bad it can kill a lot of people if it catches entire teams out on the mountain. However the 1996 “Into Thin Air” storms are quiet unusual events. This may be because weather forecasting has improved, and so mountaineers tend to pick the right time to try for the summit.

Is it possible to acclimatize to 8000m?
This study cannot directly address this issue. Certainly the low descent mortality rates among the sherpas suggest that mortality rates amongst climbers could be lower. Because they are employed to place ropes and transport equipment, sherpas tend to spend more time at very high altitudes. But as they have been through a competitive process to win jobs as sherpas, there is a selection bias in this group. Many are born and live at high altitude (Sherpas or ethnic Tibetans) and may have adaptive advantages to high altitude. So it is difficult to tell if they are sending more time at very high altitude because they acclimate better, or if they are better acclimated because they spend more time at very high altitude, or both. Other factors, such as greater climbing experience or the tendency to climb in groups may also play a role. So the ability of lowlanders to acclimate to very high altitudes remains ill defined and needs further work.
Any other notes of caution?
This is a single study – later studies, differing interpretations and advances in the understanding of high altitude physiology at extreme altitude may alter the conclusions mountaineers should draw. A catastrophic event such as a mountaineering death is often the result of the convergence of multiple factors. While we have examined the effect of high altitude illness, other factors play a role in fatalities. The mountaineer should simply use this study as one guide in assessing the many complex decisions that need to be made to ensure safety in high places.
I hope this is of help to Himalayan mountaineers. The Himalayas are the highest range in the world, and the challenges of extreme altitude are unique to this area. An understanding of the events in the ‘death zone’ is essential to safe mountaineering.
Thank you for allowing me the opportunity to post on your website. Happy and safe climbing!

Paul Firth, MBChB
Department of Anesthesia and Critical Care, Massachusetts General Hospital
Boston, Massachusetts, USA